University of Groningen Aging and cognitive control Dekker ... · of performance measures on a battery of executive tasks, rather than a single (unitary) factor. Nevertheless, there

University of Groningen

Aging and cognitive controlDekker, Mark Roel

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2009

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Dekker, M. R. (2009). Aging and cognitive control. s.n.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 14-06-2020

https://www.rug.nl/research/portal/en/publications/aging-and-cognitive-control(effe9a17-0a24-4ba5-a592-4b6ae75ac3f1).html

https://www.rug.nl/research/portal/en/publications/aging-and-cognitive-control(effe9a17-0a24-4ba5-a592-4b6ae75ac3f1).html

1

Introductie

RIJKSUNIVERSITEIT GRONINGEN

AGING ANd COGNITIVE CONTROl

Proefschrift

ter verkrijging van het doctoraat in de

Gedrags- en Maatschappijwetenschappen

aan de Rijksuniversiteit Groningen

op gezag van de

Rector Magnificus, dr. F. Zwarts,

in het openbaar te verdedigen op

donderdag 18 juni 2009

om 14.45 uur

dooR

Mark Roel dekker

geboren op 4 juni 1972

te Wolvega

2

Promotores: Prof. dr. R. de Jong Prof. dr. W.H. Brouwer

Beoordelingscommissie: Prof. dr. J.M. Bouma Prof. dr. ir. N.M. Maurits Prof. dr. K.R. Ridderinkhof

3

Introductie

3

chapteR 1

4

chapteR 1

chapteR 1

5

Introductie

ChAPTER 1 Introduction 7

ChAPTER 2 age-related changes in event-related prospective memory performance: a comparison

between four prospective memory tasks 23

ChAPTER 3 cognitive aging and task switching 55

ChAPTER 4 patterns of aging and cognitive control 75

ChAPTER 5 Summary and general discussion 127

BIBlIOGRAPhy

135

SAmENVATTING151

dANKwOORd157

6

chapteR 1

7

Chapter 1

Introduction

Introductie

8

It is well established that performance of people on cognitive tests declines with

age (e.g. Schaie, 1994; Verhaeghen & Salthouse, 1997). Several theories have

been proposed to explain cognitive decline with age. Most of these theories can

be categorized either as general factor theories or specific loss theories.

The currently dominant general factor theory is the global speed hypothesis

(Salthouse, 1991, 1996). The global speed hypothesis states that age-related

cognitive decline can be attributed to a decrease in the speed with which elementary

cognitive operations are carried out. The decrease in information processing speed

places limits on performance levels that can be reached on most cognitive tasks

(Birren, 1956; Cerella, 1985; Eearles, Connor, Smith & Park, 1997; Salthouse,

1996). Evidence for the global speed hypothesis is mostly derived from mediational

analyses, and studies show that the relation between age and cognition is reduced

or diminished when statistically controlling for speed measures which relate to

both the cognitive measures and age.

Among specific loss theories, age-related differences are often connected to age-

related functional and structural changes in the brain. Specific loss theories are

inspired by evidence that some cognitive abilities are not at all susceptible to age

effects. This group of theories focuses on age-related changes in specific cognitive

processes. Typically, in these theories a distinction is made between behavior that

is dependent on executive control processes and supposedly affected by advancing

age, such as switching between tasks, divided attention and selective attention

(McDowd & Shaw, 2000; Braver, Barch, Keys, Carter, Cohen, Kaye, Janowsky,

Taylor, Yesavage, Mumenthaler, Jagust & Reed, 2001; de Jong, 2001; Kramer,

Larish & Strayer, 1995; Mayr & Kliegl, 1993), and behavior which is less or not

dependent on control processes and is free from age-related effects.

A prominent theory of the relation between aging and changes in executive

control function is the frontal lobe hypothesis of aging. This hypothesis holds that a

differential decline in old age of neural tissue in the frontal lobe occurs (Raz, 2000;

Van der Molen & Ridderinkhof, 1998; West, 1996). Correspondingly, cognitive

functions that are related to the frontal lobe are hypothesized to be affected by

age more strongly than cognitive functions related to other regions of the brain.

Performance on executive control functions have been found to rely strongly on

the integrity of the frontal cortex as evidenced by studies with patients with frontal

lobe lesions, and with neuroimaging studies (West, 1996; Fuster, 1997)

chapteR 1

9

Introductie

In this introductory chapter first executive control functions and the relation

to other psychological constructs (intelligence, intention and goal activation)

and neurobiological constructs (frontal lobe function, neuromodulation) will be

discussed. Then these aspects of cognitive control functions will be discussed in

relation to cognitive aging. Finally, an outline of the rest of the thesis will be given

and the other chapters will be introduced.

1.1 Cognitive control The concepts of cognitive control and executive functions are often used

interchangeably. The term ‘cognitive control’ is used more in the context of the

cognitive processes and in the field of cognitive and experimental psychology.

‘Executive function’ is used in the field of neuropsychology and in the context of

functions/outcomes of the cognitive system that accommodate tasks in every day

life situations. Executive functioning is conceptualized as super-ordinate control

and as a top-down system sub-serving several basic cognitive domains (Baddeley,

1998; Shallice, 1982). In this thesis ‘cognitive control’ and ‘executive functions’ will

also be used interchangeably.

Cognitive control refers to organizing and monitoring cognitive processes. It is

crucial for making behavior adaptive and controlling behavior according to

intentions and internal goals. The concept of cognitive control implies endogenous

control of processes and, in that sense, can be contrasted with situations in which

behavior is cued, triggered or prompted explicitly from or by the environment.

Situations which place a high demand on cognitive control processes are situations

in which tasks to be performed are novel or characterized by weak environmental

support. Examples of those are situations in which an automatically triggered

response needs to be suppressed, and situations in which two or more tasks need

to be performed at the same time or in succession (see e.g. Baddeley, 1986; Meyer

and Kieras, 1997; Monsell, 1996; Norman and Shallice, 1986).

Cognitive control has been conceptualized as a unitary process that is the same in

different situations. It has also been conceptualized as a composite of fractionated

control processes. In the latter case, cognitive control process operations in

one situation can be different and distinct from those operating in another

situation. Several models of cognitive control have been developed from both

conceptualizations.

10

chapteR 1 Introductie

1.1.1 Models of cognitive control

Two of the first and most influential models of cognitive control are the model

by Baddeley and Hitch (1974) and the model developed by Norman & Shallice

(1986).

The model by Baddeley and Hitch (1974) proposed that short-term memory

involves not only passive storage but also active manipulation of information that

is attended to in order to transfer the information to long-term memory. The

model fractionates working memory into two ‘slave systems’, and an attentional

control system. The slave systems are a visuospatial sketchpad capable of holding

and manipulating visuospatial information and a phonological loop performing

a similar function for verbal information. The control system, termed central

executive, coordinates the two slave systems and links to long-term memory. The

central executive in the model is not fractionated, which makes the model prone to

the objection that the central executive is just a convenient homunculus (Baddeley,

1996).

The model of Norman and Shallice (1986) is based on a production-system

architecture. In this model specialized routines (thought and action schemata) are

triggered by conditions which are linked to specific schemata by if-then statements

(productions). If a condition triggers multiple schemata, a mechanism termed

‘contention-scheduling’ is used to prevent response conflict and errors. Control by

contention-scheduling works by means of lateral inhibition between the triggered

schemata, resulting in a dynamic of more and less activated schemata. Eventually

the most strongly activated schema will direct action or thought. Contention

scheduling is efficient in producing routine behavior in well-known situations.

Norman & Shallice (1986) outline five situations in which this routine, automatic

control will not be sufficient for optimal performance (see also Burgess, 1997).

These are situations that require planning or decision making, situations that

involve error correction or troubleshooting, novel situations, dangerous situations

and situations that require the overcoming of a strong habitual response. To enable

the model to incorporate optimal behavior in these situations, Norman & Shallice

(1986) introduced a second, higher order, level of control termed Supervisory

Attentional System (SAS). The SAS exerts control in order to accommodate

behavior to the overall goals of a person. In line with these goals, it can intervene

in the contention scheduling. This schema modulation intervention by the SAS is

realized by biasing the operation of contention scheduling by additional activation

or inhibition of lower level schemata. In terms of this model executive control (by

chapteR 1

11

Introductie

SAS) is necessary because the automatic processes of the contention scheduling

mechanism are inadequate with respect to goal-directed behavior. The SAS implies

constant monitoring of the results of actions and comparing it to the goals one has

set (Brouwer and Fasotti, 1997).

Several researchers have proposed, and worked on, deconstructing SAS or the

central executive (Myiake, Friedman, Emerson, Witzki and Howerter, 2000; Stuss,

Shallice, Alexander and Picton, 1995; Tranel, Anderson and Benton, 1994). These

studies are often inspired by findings of dissociations in performance among

executive tasks, for instance some patients fail on one executive test but not on

another (e.g. Godefroy, Cabaret, Petit-Chenal, Pruvo & Rouseaux, 1999).Also,

computational models of cognitive control have been developed (e.g. Kimberg

& Farah, 1993; Cohen, Dunbar & Servan-Schreiber, 1990) that provide a more

detailed account of control processes than the unitary SAS or Central Executive,

but are mostly constructed to model the performance characteristics of a specific

task. Also other studies, focusing on individual differences (e.g. Lehto, 1996,

Lowe and Rabbitt, 1997; Myiake et al., 2000; Robbins et al. 1998; Burgess, 1997;

Burgess, Alderman, Evans, Emslie and Wilson, 1998; Duncan et al., 1997), mostly

using batteries of executive tasks, provide evidence for a non-unitary view of

executive functions. A consistent pattern in the results of these studies is that the

intercorrelations among different executive tasks are low and often statistically not

significant. Also, exploratory factor analysis tends to yield multiple separable factors

of performance measures on a battery of executive tasks, rather than a single

(unitary) factor. Nevertheless, there are weaknesses and limitations connected

to these correlational approaches and consequently to the conclusions that can

be drawn from them (Baddeley, Della Sala, Gray, Papagno and Spinnler, 1997;

Rabbitt, 1997). The lack of strong correlations between different executive tasks

can also be a result of differences in the non-executive performance requirements

of the different tasks.

Miyake et al. (2000) conducted an individual differences study on executive

functions to provide an empirical basis for developing a theory that specifies how

executive functions are organized and what role they play in complex cognition.

They focused on separability of the three most postulated executive functions in

the literature: shifting of mental set, monitoring and updating of working memory

representations, and inhibition of prepotent responses. Furthermore they aimed

at specifying the relative contribution of these executive functions to complex

tests that are commonly used to assess executive functioning, by means of latent

12


variable analysis. Three tasks for each of these executive functions were selected

to be performed by the participants. To address the issue of the separate nature

of the three executive functions Miyake et al. (2000) performed confirmatory

factor analysis (CFA). They found that a three-factor model (in which each factor

represents one of the three postulated executive functions) provided a better fit to

the data then a one-factor model (which postulated unseparated/unitary executive

function). They also tested two-factor models (in which two of the three executive

functions were postulated as one factor) and even though these models provided

reasonable fits, none provided a better fit than the three-factor model. The model

that provided the worst fit, though, was an alternative 3-factor model in which

the three factors were independent. Based on these findings Miyake et al. (2000)

concluded that though the three executive functions are distinguishable, they are

not completely independent and there is some underlying commonality. They

also found that shifting, updating and inhibition abilities differentially contribute

to performance on commonly used executive tasks, such as the Wisconsin

Card Sorting Test (WSCT), the Tower of Hanoi (TOH) and the Random Number

Generation task. Together with other theoretical proposals (Duncan et al., 1997)

they argue that a simple dichotomy between a ‘unitary’ and a ‘non-unitary’ view

on cognitive control will not suffice to help reconcile the controversy, and both

views should be taken into account.

1.1.2 Cognitive control and the frontal lobes

In the field of neuropsychology, performance deficits in executive functions are

often found as a consequence of lesions to the frontal lobes (West, 1996; Fuster,

1997). Consequently much research has been conducted on the relation between

executive functions and the integrity of the frontal lobes. Frontal lobe damage

usually leads to behavioral deficits like a tendency to perseverate and distractibility.

These impairments are not in a specific cognitive or behavioral domain, but rather

in the organization, control and monitoring of cognitive abilities (Duncan, 1986,

Shallice, 1988, Monsell, 1996). Most researchers agree that the frontal lobes play

an important role in the executive control of behavior and many studies have

linked the two (Shallice & Burgess, 1991; Stuss, 1992; Stuss & Benson, 1984).

As a result the term ‘frontal function’ is sometimes used metaphorically to denote

executive functions, even though they are conceptually and empirically not identical

(Baddeley, Della Salla, Gray, Papagno & Spinler, 1997; Burgess, 1997; Myiake et al.

2000). Many of the tests that are referred to as “frontal tests” appear sensitive to

damage in other areas in the brain (Della Sala, Gray, Spinnler & Trivelli, 1998). In

chapteR 1

13

Introductie

other words, patients with lesions outside the frontal lobes can show severe deficits

in tests of executive functioning (Anderson, Damasio, Jones & Tranel, 1991; Reitan

& Wolfson, 1994). Moreover, not all patients with frontal lesions show problems

with tasks of executive functions (Shallice & Burgess, 1991).

1.1.3 Cognitive control, general intelligence and goal activation

In every day life, executive functions make it possible to guide behavior intentionally

according to internal goals. This means that executive functioning makes it possible

to deal with novel conditions, to solve complex problems, to adapt to unexpected

circumstances and to perform multiple tasks at the same time (see e.g. Cahn-

Weiner, Boyle & Malloy, 2002; Grigsby, Kaye, Baxter, Shetterly & Hamman, 1998).

Most of these abilities have also been linked to intelligence and it is therefore no

coincidence that several researchers in psychology have studied executive function

in relation to general intelligence.

Intelligence tests reliably predict performance across a wide range of different

tasks assessing information-processing speed (Eysenk, 1986; Jensen, 1985, 1987),

working memory capacity (Carpenter, Just & Shell, 1990; Just & Carpenter, 1992;

Kyllonen & Krystal, 1990) and the efficiency of allocating attention to subtasks

within highly complex tasks such as driving (Duncan, 1990).

In the psychometric literature on intelligence, a distinction parallel to the distinction

between control functions and automatic functions exists (Phillips, 1997), namely

between “fluid intelligence” (gf) required for performance on tasks that are novel

and complex on one hand, and “crystallized intelligence” (Horn, 1982; Horn &

Cattell, 1966) on the other. Studies by Duncan and co-workers (1995, 1997)

suggest that intelligence tests designed to tap fluid intelligence (i.e. Cattell &

Cattell, 1960), are very good predictors of the level of executive functioning.

A general finding in studies in which a battery of cognitive tasks is administered,

is that correlations between the performance measures are always positive.

To explain this phenomenon, Spearman (1927) introduced the concept g as a

factor that plays some role in performance on all tests. Duncan et al. (1996,1997)

proposed that this g factor reflects the efficiency of a general goal activation

process. More specifically they suggest that g reflects “…a frontal process of mental

programming, or constructing an effective task plan by activation of appropriate

goals or action requirements” (Duncan et al. 1997). When this goal activation

14


fails, a phenomenon Duncan et al. (1996) term goal-neglect occurs. Goal neglect

is defined as a disregard of a task requirement even when it has been understood

and remembered (Duncan 1995). Although the severity of goal neglect has been

reported to be increased in patients with frontal lobe lesions (Duncan et al. 1996),

it has been found that under conditions of novelty, multiple task requirements and

weak error feedback it also occurs in the normal population.

Duncan et al. (1997) addressed the relation between general intelligence or

Spearman’s g and goal activation function. They administered a battery of

conventional executive control tasks and a battery of test that are not usually

considered as measuring executive function. They found positive but low

correlations between the executive tests. Nevertheless, the overall aggregated

indices of the two batteries correlated strongly. Duncan et al. (1997) argued that the

common element in these tests reflect the contribution of goal activation function

to all these tests. In a second experiment they tested this idea by administering a

battery with executive tests as well as a test developed to measure goal neglect

and a test for measuring g. In this experiment they found again low correlations

between performance indices of different tasks within the executive task battery,

but a close relationship between the overall index of performance on the executive

task battery and goal neglect and g. Based on these findings Duncan et al. (1997)

posed that each test is weakly influenced by goal activation. This goal activation

function can be indicated by averaging performance over a number of tests, or by

a single purer measure of goal neglect. From this viewpoint a general intelligence

test consists of a composite or a battery of a diverse set of cognitive tests all sharing

the need for goal activation in order to perform. In the line of reasoning of Duncan

et al. (1997) this idea can also be reversed in the sense that any battery of diverse

and heterogeneous cognitive tasks will produce an aggregated performance index

that corresponds to g and to goal neglect or goal activation function.

A task domain in which goal neglect has systematically been studied is prospective

memory. Typically in these tasks, subjects are required to place a task on hold,

mostly while performing another (background) task, until a trigger (time or an

event) occurs, after which the prospective task is required to be resumed. Duncan

et al. (1997), for instance, used a task in which the background task was to

monitor two streams of random letters and digits and speak out loud every letter

that appeared on one side. Occasionally a central, symbolic cue (‘+’ or ‘-‘) was

presented, indicating subjects on which side to continue reading. The goal or

intention that subjects were to keep activated while performing the background

chapteR 1

15

Introductie

task was to respond to the cue. Other studies on prospective memory have found

performance to be dependent on the delay between instruction of the prospective

memory component and the moment of carrying out the intention (Brandimonte,

Einstein & McDaniel, 1996) and the saliency of cue signaling the prospective task

(Einstein & McDaniel, 1996; Maylor, 1996). Most studies have found that goal

neglect in prospective memory tasks can be attributed to failures to act upon the

instruction at the required moment, and not to simply forgetting the instructions

(e.g. Brandimonte et al., 1996).

The ‘goal activation’ conceptualization implies a probabilistic view on limitations

of cognitive control. If cognitive control and its limitations are conceptualized as

probabilistic instead of absolute, then intra-individual variability and within task

variability of performance become much more relevant, instead of solely inter-

individual and between task differences (Nieuwenhuis, 2001). Several studies have

been performed to explore this view on executive control and its limitations. This

view sprouts from the work previously mentioned on goal activation by Duncan

and colleagues (Duncan, 1995; Duncan, Emslie, Williams, Johnson & Freer, 1996)

and on intention activation by De Jong and colleagues (De Jong, 2000; De Jong,

Berendsen & Cools, 1999). These studies directly or indirectly address the question

of whether performance limitations on performance in cognitive control tasks

reflects intrinsic, fundamental or static limitations or whether it reflects a failure to

fully or consistently utilize the capabilities of executive control or a combinations

of these factors.

De Jong et al. (1999) concentrated on this issue with respect to Stroop-type

interference and residual switch costs. Stroop interference is commonly attributed

to an involuntary consequence of processing the meaning of a color word (e.g.

blue) when the task is to respond to the color the word is printed in. This would

then result in the usually slower and less accurate responses when the meaning of

the color word and the color it is printed in are different (incongruent trials) than

when it is the same (congruent trials). De Jong et al. (1999) argue that even if

subjects would be able to completely prevent word meaning from influencing task

performance, they might not always fully exploit this ability. If the task conditions

do not induce subjects to fully use the ability to prevent irrelevant information

having influence on task performance, this would result in the same Stroop effect.

To test this idea De Jong et al. (1999) used a spatial version of the Stroop task

and varied the pace in which the stimuli were presented, reasoning that a fast

pace would induce or help subjects to stay focused on the instructed task and

16


a slow pace would give rise to fluctuations at attentional state across trials and

result in failures to fully bring to bear one’s ability to inhibit the processing of the

irrelevant information. The results supported evidence that Stroop interference is,

at least partly, attributable to a failure to fully and consistently use capabilities of

control. This account is in line with the demonstration by Logan and Zbrodoff

(1979) of a decrease of the magnitude of the Stroop effect when increasing the

relative frequency of demanding, non-corresponding trials (see also Kane & Engle,

2003; West 1999; Heathcote, Popiel and Mewhort, 1991; Mewhort, Braun and

Heathcote, 1992)

1.1.4 Psychometric issues of cognitive control

Apart from the complexity of the relation between cognitive processes and

physiological processes, measuring cognitive control poses several difficulties such

as construct validity and task-purity.

Task-purity and construct validity

According to Miyake et al. (2000), “…Because executive functions necessarily

manifest themselves by operating on other cognitive processes, any executive

task strongly implicates other cognitive processes that are not directly relevant to

the target executive function”. This states the problem of task-purity of tasks that

are used to measure executive function. Because performance on tests that are

designed to tap executive control is dependent on executive control processes

as well as on component processes, several attempts have been made to isolate

control processes like inhibition. Rabbitt (1997), though, points out that the

poor construct validity of terms such as “planning”, “inhibition”, “impulsivity”

and “memory for context” is, at least in part, due to the fact that these terms

come from the language of every day subjective experience, while the constructs

are used by researchers for hypothetically distinct “components” of executive

behavior.

Furthermore, several studies have examined the relation between indices of

different tasks that are designed to measure the same executive process. Construct

validity of executive processes would imply significant intercorrelations between

different indices that tap the same postulated executive process. There is not

much evidence for this type of construct validity for indices of executive control

processes. An indicative example is a study by Shilling, Rabbitt and Chetwynd

(2002) investigating inhibitory function in older adults. They administered four

analogues of the Stroop interference paradigm to a group of older participants,

chapteR 1

17

Introductie

and found no significant correlations between these different versions of the

Stroop task. Also in other studies, intercorrelations between indices of the same

postulated executive process are predominantly moderate to low and not higher

than the intercorrelations with indices for other types of processes (Duncan et al.

1997; Rabbitt, 1997, Kramer et al.,1994, Nieuwenhuis et al. 2004).

1.2 Cognitive Aging

1.2.1 Global versus specific loss hypotheses of cognitive aging

The global speed hypothesis states that age-related cognitive decline can be

attributed to a decrease in the speed with which elementary cognitive operations

are carried out. The decrease in information processing speed places limits on

the performance that can be reached on most cognitive tasks (Eearles, Connor,

Smith & Park, 1997; Salthouse, 1996). This hypothesis is contrasted to hypotheses

that contribute age-related effects on performance on executive control tasks to

specific control deficits. With regard to several cognitive control functions the

evidence on this issue is mixed (McDowd and Shaw, 1999). For example age-

related differences in the Stroop-effect have been found to be accountable by

the global speed hypothesis (e.g. Salthouse and Meinz, 1995; Verhaeghen &

De Meersman, 1998) as well as by a specific-deficit account (e.g. Hartley, 1993).

Another general factor that has been postulated to be able to account for a large

part of age-related variance performance on cognitive tasks is general intelligence.

For instance, as mentioned above in relation to psychometric issues of measuring

executive control, Shilling et al. (2002) administered several different versions of

the Stroop task. They found that age differences in inhibitory function as measured

with the Stroop task, were not independent of fluid intelligence. They suggest

that if subjects of different ages are matched in terms of intelligence scores they

will not differ in terms of their ability to cope with executive control tasks (see also

Rabbitt et.al 2001). Moreover, intelligence test scores have been found to act as

a powerful mediator of age effects on simple tests of memory and information

processing speed (Rabbitt, 1993). Apparently, empirical overlap exists between the

consequences of different one-factor theories. In the paragraph of cognitive control

it was discussed that most performance measures of executive and non-executive

tests correlate positively but weakly (Duncan, 1997). Many tests, as argued by

Duncan (1997), share the influence of a general (g) factor, and correspondingly

an aggregate indicator of performance on these tests correlates strongly with

measures of intelligence as well as with measures of goal neglect. Taken together,

18


these findings suggest that a large part of the age-related variance in performance

on most cognitive tasks may be explained in terms of a single factor that is strongly

related to performance on intelligence tests. Still, some performance measures of

cognitive tasks differentiate better between age-groups than other measures. A

theory aimed at explaining these differentiated patterns from a neuropsychological

perspective is the frontal lobe hypothesis of aging.

1.2.2 Frontal lobe hypothesis of cognitive aging

From the viewpoint of a specific loss hypothesis, researchers search for more

specific explanations for differential declining of cognitive functions as people age.

The most dominant process-specific theories of cognitive aging concern ‘executive

functions’ or ‘cognitive control’. As people age, executive functions are found to

be compromised (Bryan & Luszcz, 2001; Mayr, Spieler & Kliegl, 2001; Rabbitt

et al. 1997; Wecker, Kramer, Wisniewski, Delis & Kaplan, 2000). Furthermore

the executive function theory of aging finds support from neurobiological and

neuropsychological studies on aging. It has been demonstrated that executive

functioning is highly dependent on the integrity of the frontal lobes. West (1996)

presents an overview of the relation between advancing age and functional and

structural changes in the frontal lobes that occur earlier than in other regions in

the brain. These changes involve a relatively greater loss of volume of frontal areas

than of other cortical areas, a local decrease in the number of synapses, atrophy

of dendritic processes, and reduced efficiency in cellular mechanisms that support

the synthesis and transmission of neurotransmitters (West, 1996). Together these

findings result in the frontal lobe hypothesis of aging. This hypothesis states that

cognitive functions supported by the prefrontal cortex show signs of age-related

decline at an earlier age and to a greater degree than cognitive functions supported

by other brain structures (West, 1996; Perfect, 1997).

Some studies lead to a more differentiated view on this relation (for review see e.g.

Band et al. 2002; Greenwood 2000). First, the age-related changes in the structure

and function of the brain have been found to be differentiated, also within the

frontal cortex. Neuronal loss has been found to differ even between cortex layers

within the same subregions of the dorsolateral prefrontal cortex (Uylings & De

Brabander, 2002). Second, neurobiological changes may not correspond in a direct

manner to cognitive changes. Third, related to the former, several methodological

problems exist with measuring cognitive control. As discussed above in the section

cognitive control, construct validity of many tasks that are designed to indicate

chapteR 1

19

Introductie

executive control is unclear (e.g. Rabbitt, 1997). Different indices that are designed

to measure the same construct are often not strongly correlated (e.g. Burgess and

Shallice, 1997, Rabbitt, 1997, Shilling et al., 2002). These complications call for a

more differentiated view on the relation between age-related changes in cognition

and neurobiological changes.

1.2.3 Neuromodulation models of cognitive aging

According to Li, Lindenberger and Sikström (2001), several theories of cognitive

aging try to explain age-related cognitive changes in terms of a decline of

cognitive resources. Li et al. (2001) argue that those resource-reduction theories

are confronted with major difficulties. They posit that the different resources such

as working memory, attention regulation or processing speed are interdependent.

Working memory as a resource is for example not independent of attentional

control mechanisms (Braver et al. 2001) or processing speed. Furthermore, Li and

Lindenberger (2001) criticize the circular nature of these accounts. To overcome

the difficulties of the resource-reduction theories they argue for the need of an

integrative account that cuts across neural, information-processing and behavior

levels.

Welford (1977) argued that any increase with age of random activity in the brain

(“neural noise”) would affect a wide range of both sensori-motor en intellectual

performances. Evidence from neurobiological and other studies is accumulating

about changes with age in the dopaminergic neurotransmitter systems playing

an important role in age-related changes in cognition and behavior (Backmann

et al., 2000; Goldman-Rackic and Brown, 1981; Li and Lindenberger 1999;

Braver et al. 2001; Nieuwenhuis et al. 2001). Two computational theories have

tried to explicate the relationships between age-related decline in dopaminergic

modulation, neural information processing fidelity and cognitive aging. Both

theories manipulate the signal-to-noise ratio of neural information processing and

in that way regulating neurons’ sensitivity to afferent signals (Li and Lindenberger,

2001). Li et al. (2000) focus in their computational theory on relating aging

deficits and cortical representational distinctiveness to dopaminergic modulation

and neural information processing fidelity. Braver et al. (2001) focus on relating

age-related impairments in regulating context representations and maintenance

with dysfunction of the dopaminergic system in the dorsal lateral prefrontal

cortex. In general the theoretical frameworks have in common linking “…

attenuated neuromodulation to increased neural noise and less distinctive cortical

20


representations in the aging brain, and finally on to cognitive aging deficits” (Li

and Lindenberger, 2001). Given the role of dopamine in the frontal cortex, these

models of (dopaminergic) neuromodulation provide important insights in the

nature of aging at both the biological and cognitive level and could lead to a

refinement of the frontal lobe hypothesis.

1.2.4 Aging and goal activation

Age effects in executive control tasks might reflect intrinsic limitations in executive

control capabilities or failures to fully or consistently utilizing such capabilities or a

combination of these factors (De Jong et al., 1999). Intention activation in relation

to cognitive control has been discussed above. This perspective is strongly related

to the concept of goal-neglect, put forward by Duncan et al. (1996, 1997). Goal

neglect is defined as a disregard of a task requirement even when it has been

understood and remembered (De Jong, et al. 1999, Duncan et al. 1996).

Goal selection and goal maintenance under conditions of novelty or weak

environmental support is seen as a primary function of the frontal lobe (De Jong,

2001). Research into the relation between intention or goal activation and cognitive

aging have focused primarily on three experimental paradigms, prospective

memory tasks (i.e. Maylor, 1996; Einstein & McDaniel, 1990; Duncan et al. 1996;

West & Craik,1999), switch-task (i.e. De Jong, 2001) and cue-tasks (Nieuwenhuis

et al. 2000, 2004). These paradigms require subjects to be proactive and anticipate

for performance in future trials. More specifically these tasks have in common that

for performance to be optimal subjects need to consistently endogenously initiate

a preparatory action or process.

1.3 Outline of thesis

In this thesis three studies are reported that explicitly study cognitive control

function in relation to aging.

From the perspective of the frontal lobe hypothesis of cognitive aging, prospective

memory tasks are particularly interesting and relevant, because prospective

memory requires planning and keeping a prospective intention activated during

performance on another task. Both functions are generally believed to involve the

frontal lobes. Thus, according to the frontal lobe hypothesis of cognitive aging,

robust age-related performance differences on prospective memory tasks should be

chapteR 1

21

Introductie

expected. However, the evidence on age effects on prospective memory abilities is

not entirely consistent. Older adults have been found to perform as well as younger

adults in every day tasks such as making telephone calls or taking medicine at

particular times of the day (see Maylor, 1996, for a review). It is possible that

this age-invariance is due to older adults adopting efficient strategies (for example

strategic use of retrieval cues). In prospective memory tasks developed for use

in laboratory settings, the use of external memory aids and other strategies can

be better controlled. Several studies using laboratory prospective memory tasks

have found age effects on performance (Einstein, McDaniel, Richardson, Guynn

& Cunfer, 1995; Park, Hertzog, Kidder, Morrell & Mayhorn, 1997; Cherry &

LeCompte, 1998; Duncan et al. 1996; Maylor, 1998; West & Craik, 1999, but see

Einstein & McDaniel, 1990). Moreover, prospective memory tasks are especially

sensitive for demonstrating goal neglect (Duncan et al., 1996; see also Nieuwenhuis

et al., 2004). In chapter 2, a study is presented testing the sensitivity and reliability

of four different prospective memory tasks for assessing effects of normal aging.

Based on previous evidence, the tasks were differentiated on various dimensions

such as perceptual saliency of prospective target events, frequency of occurrence

of prospective target events, complexity of prospective-memory instructions,

and provision of feedback after prospective-memory errors. The role of goal

maintenance (or maintaining prospective intentions) and basic mental speed as

mediators for age effects on prospective memory performance are discussed.

Chapter 3 reports a study on age-related differences in task-switching. Switching

between tasks requires the application of cognitive control function (Allport &

Wyllie, 2000)). In switch tasks, usually three types of trials can be distinguished.

On fixed-task trials, that constitute control blocks, the task remains the same

throughout a block of trials. On non-switch trials the task to be performed is the

same as on the previous trial. On a switch trial the task to be performed is different

than the task on the previous trial. The differences in performance between switch

trials and non-switch trials are referred to as local switch costs. The differences

in performance between blocks with fixed-task trials and blocks in which non-

switch trials and switch trials are intermixed, are referred to as global switch costs.

Local switch costs have been found to decrease, but mostly not diminished when

subjects are given ample time to prepare in advance for the task switch (preparation

interval) resulting in, what is referred to as, residual switch-costs. Several studies on

the relation between task-switching and cognitive aging have been conducted (i.e.

Eenshuistra, Wagenmakers & De Jong, 2000; Kray (2005); Kray and Lindenberger

(2000); Kramer, Hahn & Gopher, 1999; Meiran et al., 2001). Age-related differences

22


in global switch-costs are found in most studies. This finding seems to be restricted,

though, to paradigms in which response sets of the different tasks overlap, and

it appears to be mediated by strategical aspects. Age-related differences in local

switch costs are found in most but not in all studies. It appears that when taking

into account baseline performance and practice, these age-related effects become

modest or diminished. With respect to residual switch costs in old adults, it remains

unclear whether these can be attributed solely to intermittent failures to engage in

advance preparation when ample time is provided, as has been found with young

adults by De Jong (2000) and by Nieuwenhuis and Monsell (2002), or to intrinsic

limitations to prepare in advance by endogenous means only. In chapter 3, a study

on global, local and residual switch costs in relation to age, is reported. Specifically,

the study was aimed at clarifying possible differential contributions to residual

switch costs between young and old adults. To accomplish this, a RT distribution

modeling approach developed by De Jong (2000) was used.

As reviewed above some theories posit that age-related differences can be

accounted for solely by one general aspect of cognition (e.g. Birren, 1956;

Cerella, 1985; Eearles et al., 1997; Salthouse, 1996), while other theories hold

that that effects of age on cognition are more specifically limited to executive

control functions (e.g. Bryan & Luszcz, 2001; Mayr, Spieler & Kliegl, 2001; Wecker,

Kramer, Wisniewski, Delis & Kaplan, 2000). In chapter 4, a study is reported that

focuses on performance on several executive control functions in relation to aging.

The study is part of a larger research project in which the relationships between

age on executive control and on memory functions are investigated, using an

extensive battery of retrospective and prospective memory tasks and executive

control tasks. In chapter 4 patterns of performance of older and younger adults

on measures of executive control, fluid intelligence and performance on standard

neuropsychological indices are reported. Age-related effects on executive control

functions are discussed by focusing on the results regarding age-related within-task

variability of the different tasks. The relation between different executive control

measures and the effect of aging on this relation are discussed by focusing on

the results regarding age-related between-task variability and these results are

modeled using structural equation modeling.

chapteR 1

23

Introductie

Chapter 2

Age-related changes in event-related prospective memory performance:

A comparison between four prospective memory tasks

Vogels, W. W. A., Dekker, M. R.,

Brouwer, W. H., & de Jong, R.

Brain and Cognition (2002); 49, 341–362.

chapteR 2

2.1 IntroductionThe frontal lobes are commonly associated with executive functioning (Fuster,

1989; Moscovitch & Winocur, 1992). For instance, the frontal lobes are assumed

to play a critical role in mental activities such as formulating plans, initiating

actions, monitoring on-going behavior, and evaluating outcomes (Glisky,

1996). Also, maintaining and manipulating information in working memory are

frequently listed as key functions subserved by the frontal lobes (Baddeley, 1986).

Neurophysiological and neuroanatomical evidence suggests that changes in old

age may occur earlier and progress more rapidly in frontal areas than in most

other parts of the brain (West, 1996). According to the frontal lobe hypothesis

of cognitive aging, these distinct patterns of brain aging should be reflected in

correspondingly distinct time courses of changes in mental abilities, with mental

functions that involve the frontal lobes being particularly susceptible to effects of

normal aging (Craik, 1986; West, 1996).

With respect to memory, the frontal-lobe hypothesis would predict that age-

related deficits should depend strongly on the degree to and manner in which

the frontal lobes are involved in the performance of specific memory tasks. In this

theoretical context, the distinction between retrospective and prospective memory

tasks has recently attracted a great deal of attention in research on memory

aging. Retrospective memory tasks require remembering things in the past, with

recall and recognition tasks being prime examples. Prospective memory tasks, in

contrast, require remembering to perform an intended action at some time in

the future. Though prospective memory tasks come in many forms, they typically

require the subject to perform a background or cover task with the additional

instruction to perform some type of action (e.g. clapping hands or pressing a

designated key) when, at some point during performance of the background

task, a cue (e.g. sounding of a bell or appearance of a target word) is presented.

From the frontal-lobe hypothesis of cognitive aging, prospective memory tasks

are particularly interesting and relevant because such tasks require planning and

impose the requirement to keep the prospective intention activated during ongoing

performance of a background task, both functions that are generally believed to

involve the frontal lobes.

It is important to point out that retrospective and prospective memory tasks do

not represent relatively pure instances of, respectively, nonfrontal and frontal tasks.

For one, retrospective memory tasks may greatly differ in the degree to which

they depend on frontal-lobe functioning. Thus, implicit memory tasks are usually

24

chapteR 2Age-related changes in event-related prospective memory performance:


regarded as requiring minimal involvement of the frontal lobes, whereas there is

ample evidence for critical frontal-lobe involvement in various types of recognition

and recall tasks (for review, see Perfect 1997). Second, a substantial degree of

overlap probably exists between component processes underlying retrospective

and prospective memory tasks. Prospective tasks require forming cue-action

associations and maintaining these at a sufficiently high level of activation for

the association to be retrieved and the action to be triggered when the cue is

presented (Manntyla, 1996). Forgetting of the relevant associations, or an inability

to remember the action associated with a prospective cue, represents a major

potential cause of prospective memory failures. For example, in a recent study

(Cherry and LeCompte, 1999) 50% of a group of old, low-ability (as assessed

by educational and socio-economic status) participants failed to spontaneously

recall the prospective cue or associated response during task debriefing. However,

virtually all participants that failed this initial recall test, succesfully recalled the

cue and associated response after some prompting (see Duncan et al., 1996,

for similar results). Also, several studies have managed to exclude retrospective

memory problems as a factor responsible for prospective memory failures (e.g.,

Brandimonte et al., 1996; but see Burgess and Shallice, 1997, for some possible

exceptions).

The theoretical perspective that we adopt here holds that intention retrieval and

triggering of the requisite response in prospective memory tasks is a stochastic

process with a probability of success that jointly depends on the momentary

state of activation at which the associative cue-action encoding is maintained and

the trigger strength or potency of the cue. Thus, low probabilities of success will

be associated with low maintained levels of activation and weak cues, and high

probabilities with high levels of activation and potent cues. Following suggestions

by Craik and Kerr (1996), transient lapses of intention, causing intentions to

temporarily fall below retrieval threshold, are thought to occur over the course

of background task performance, causing temporal fluctuations of prospective

memory success and failure. Results consistent with this notion were reported

by Maylor (1996) and by West and Craik (1999). These investigators found that

overall prospective memory performance could be represented in terms of two

ratios: a forgetting ratio (ratio of prospective memory hits that were followed by

a prospective memory miss) and a recovery ratio (ratio of prospective memory

misses followed by a prospective memory hit). Typically, both these ratios have

been found to assume non-zero values. Given these findings, and the evidence

presented earlier for succesfull recall of task instructions during debriefing, it seems

25

chapteR 2

26



reasonable to suggest that momentary lapses of intention over the course of task

performance, and not a complete forgetting of the demands of the prospective

memory task, are primarily responsible for prospective memory failures.

Age-related effects in prospective memory tasks

As noted previously, prospective memory tasks require planning and impose

the requirement to keep the cue-action association activated during ongoing

performance of a background task, both functions that presumably involve the

frontal lobes. According to the frontal-lobe hypothesis of cognitive aging, robust

age-related effects on performance in prospective memory tasks should therefore

be expected. The evidence on age effects on prospective memory abilities is not

entirely consistent, however. For instance, older adults generally perform at least as

well as younger adults in everyday tasks such as mailing postcards on particular days

or making telephone calls or taking medicine at particular times (see Maylor, 1996,

for review). However, they appear to do so by adopting efficient retrieval cues or

strategies, which may be hypothesized to compensate for declines in the ability to

initiate and maintain self-initiated mental activity. More pertinent evidence should

therefore come from tasks in which the use of external memory aids and other

potentially relevant activities can be controlled, as in laboratory settings.

Though the evidence on possible age effects in prospective memory tasks from

laboratory studies is also somewhat mixed, a fairly consistent picture has started

to emerge in the literature. Consistent age effects have been found in time-based

prospective memory tasks, that require individuals to carry out the prospective

action after a given period of time has elapsed (e.g., Einstein, McDaniel, Richardson,

Guynn, & Cunfer, 1995; Park, Hertzog, Kidder, Morrell, & Mayhorn, 1997). In

contrast, age invariance has been found in some event-based prospective memory

tasks, in which a cue serves as a prompt for prospective remembering and action.

For instance, Einstein and McDaniel (1990) reported age invariance in an event-

based task when the functional demands of the background short-term memory

task were equated across age groups. As suggested by Einstein & McDaniel (1990)

this difference in results might well be due to time-based prospective memory tasks

being more difficult because they require more self-initiated retrieval processes.

More recent evidence suggests however that age invariance in event-related

prospective memory tasks may be the exception rather than the rule. Cherry and

LeCompte (1998), using essentially the same task as that employed by Einstein

and McDaniel (1990), reported age invariance when comparing high-ability young

and old groups but marked age effects when comparing low-ability groups. Also,

chapteR 2chapteR 2Age-related changes in event-related prospective memory performance:


27

robust age effects in event-related prospective remembering have been reported

in several other variants of prospective memory tasks (e.g., Duncan et al., 1996;

Maylor, 1998; West & Craik, 1999). Interestingly, higher forgetting ratios and

lower recovery ratios have been reported for older as compared to younger adults,

suggesting that age-related declines in event-based prospective memory reflect a

greater incidence of momentary lapses of intention in older adults (Maylor, 1996,

1998; West & Craik, 1999). In summary, the evidence on age effects in prospective

memory performance appears to be reasonably consistent with predictions derived

from the frontal-lobe hypothesis of cognitive aging.

The present research

The present research is part of an ongoing project in which the relationships between

age effects on executive control functions and on memory functions are investigated,

using an extensive battery of retrospective and prospective memory tasks and ‘executive

control’ tasks. Thus, the immediate and foremost aim of the present research was to

compare four different prospective memory tasks with respect to their suitability for

inclusion in the overall task battery. Criteria of special interest were sensitivity to age

and reliability (as assessed by split-half methods). To facilitate the assessment of (split-

half) reliabilities, only event-related prospective memory tasks were included. The four

tasks, and the reasons for their initial conclusion, will be described next.

Letter monitoring task

The letter monitoring task was developed by Duncan and coworkers (1996).

Participants were presented with streams of letters and digits that were presented

on either side of fixation on a computer screen. They were to watch out for letters

at one, initially instructed, side of fixation only and repeat them aloud when

they occurred. In addition, they were instructed to look out for an occasionally

presented plus or minus sign at fixation which served as a cue directing them to

report letters in the right (plus sign) or the left (minus sign) stream. As discussed

in detail by Maylor (1998), this task can reasonably be viewed as a special type of

prospective memory task, special because the delay between the instructions and

the prospective target event is relatively short and the rate at which prospective

targets occur is relatively high. The task was selected here because Duncan and

coworkers presented evidence that it is (a) specifically sensitive to integrity of

frontal-lobe functioning and (b) sensitive to age (Duncan et al., 1996, Experiment 3).

As documented in detail by Duncan et al. (1996), prospective memory success

and failure in this task present a pattern that differs from the fluctuating pattern

chapteR 2

28



observed in other prospective memory tasks (Maylor, 1998; West & Craik, 1999).

After the first correct response to a prospective target was recorded, subsequent

forgetting (success followed by a failure) almost never occurred. Thus, participants

differed almost exclusively in terms of the number of prospective memory failures

before they, often after some degree of prompting, became consistently successful.

Duncan et al. (1996) suggested that this pattern of results could be explained in

terms of goal neglect, which they defined as the “disregard of a task requirement

even though it has been understood and remembered” (p. 257). However, they

also acknowledged that it might be possible to observe goal neglect after initial

success in this task if, for example, the interval between prospective targets were

prolonged. Note that the apparent absence of subsequent forgetting in this task

prohibits assessment of split-half reliabilities.

Word comparison task

The word comparison task was introduced by West & Craik (1999). On each

trial, participants were presented with a word pair and indicated their judgment

whether the words were from the same or a different semantic category. This

choice of background task is interesting because much evidence indicates that

semantic knowledge remains stable into later adulthood (West & Craik, 1999).

Indeed, West & Craik found younger and older adults to perform at similar levels

on this background task. From a methodological perspective, then, this choice of

task serves to equal total time spent on the task as well as relative task difficulty for

younger and older adults.

Prospective cues were defined as word pairs with both words being printed in

a non-standard color or letter case; participants were instructed not to make a

category judgment for such perceptually deviant word pairs but to press a separate

key instead. Occasionally presented word pairs with only one of the words printed

in a non-standard color or letter case served as prospective lures, and required

a regular category judgment response. It should be noted that the prospective-

memory instructions for this task are quite complex, and may thus be expected

to impose considerable demands on participants’ ability to keep the relevant cue-

action associations suitably activated during performance of the background task.

Also note that, in contrast to the letter monitoring task, prospective targets did

allow for the background task to be performed; in that sense these targets can be

considered to be much weaker prospective cues than the second-side instructions

in the letter monitoring task.

With this task West and Craik (1999) obtained several interesting age effects



29

on prospective memory performance. First, whereas younger adults responded

with nearly perfect accuracy for prospective cues as well as prospective lures,

older adults committed about 20% errors for prospective cues as well as lures

on average. Also, clear evidence for temporal fluctuations in prospective-memory

success and failure was found for older adults. In addition, whereas younger and

older adults did not differ with respect to the speed of regular semantic-judgment

responses, they differed strikingly with respect to the speed at which they made

correct prospective responses. Younger adults were much faster in making correct

prospective responses as compared to regular semantic-judgment responses; in

contrast, correct prospective responses were much slower than semantic-judgment

responses in older adults. As discussed by West and Craik (1999), this pattern of

results is nicely consistent with the notion that older adults have a diminished

ability to maintain the cue-action associative encoding in a highly activated and

highly accessible state, resulting in slower and more error-prone responses for

prospective cues.

Pictures task

In the two previous tasks, prospective cues or targets are defined in terms of

perceptual qualities that set them apart from the inputs associated with the

background task, more strongly so in the letter monitoring task. Also, these

defining perceptual qualities were kept constant throughout the experiment. Both

factors should contribute to the saliency or potency of prospective memory cues

to trigger prospective remembering. In the present task, we sought to diminish

the trigger strength of prospective events, thereby enhancing the dependency of

prospective memory performance on the ability to maintain relevant cue-action

associative encodings in a highly activated state.

In the pictures task, participants made speeded semantic judgments of pictures

of objects. As in the word comparison task, this choice of background task was

expected to result in approximately equal total time spent on the task as well

as relative task difficulty for younger and older adults. Prior to a block of trials,

participants were told to look out for a specific object (defined verbally) and to

immediately press a separate button when this object was presented at some time

during the upcoming block. Thus, a new prospective target was defined for every

trial block. Also, the prospective target was defined verbally, thus ensuring that

recognition of the target would depend on semantic processing or categorical

perception. Finally, the prospective target did not differ in any perceptually salient

manner from the other objects in a block; that is, it was effectively embedded in

chapteR 2

30



the regular stream of events. All of these factors may be hypothesized to limit the

perceptual saliency of the prospective target in the present task.

‘Three in a row’ task

The background task was a four-choice reaction time task, with four letters mapped

onto two key press responses. Participants were instructed to watch for any letter

appearing three consecutive times and to press a separate key in reaction to the

third consecutive appearance of a letter. Because letters were randomly intermixed,

the probability of such a three-in-a-row prospective event was 1/16.

As in the pictures task, the prospective event was not defined in terms of a

perceptually salient difference from background-task stimuli; identifying the

prospective event required monitoring of stimulus sequences in this task. Most

important, however, was the inclusion of two conditions in this task. In the no-

feedback condition, following a prospective-memory error, i.e. when the regular

key associated with the letter was pressed, the next letter was presented as usual.

In the feedback condition, a prospective-memory error resulted in the computer

program to wait until the special prospective key had been pressed. Thus, only in

the feedback condition were participants explicitly reminded of the prospective-

memory instruction after every prospective-memory failure and forced to execute

the appropriate prospective action. As compared to the no-feedback condition, the

provision of explicit and immediate feedback upon prospective-memory errors was

expected to greatly reduce the prevalence of such errors in the forced condition,

especially so in older adults. More specifically, in terms of forgetting and recovery

ratios such feedback was expected to markedly enhance recovery ratios and,

though perhaps to a lesser extent, to decrease forgetting ratios.

2.2 Method

Participants

This study was conducted as part of a larger experiment consisting of various

memory- and learning-tasks that were designed to place different demands on

different aspects of memory and learning, some supposedly strategy-sensitive,

thereby trying to disentangle frontal from more specifically mnestic components.

Participants for the older adults group were recruited by either an advertisement

in a newspaper for elderly or they were drawn from a participant pool at the

Groningen University. Younger adults were either drawn from a participant pool



31

or they were introduced by other participants. Exclusion criteria were forgetfulness

due to (mild) dementia (Cognitive Screening Test; De Graaf & Deelman, 1991) or

other neurological disorders. All travelling expenses were paid and every participant

received a participation fee of 90 Dutch guilders.

The participant sample consisted of 32 persons. Table 2.1 shows the demographic

characteristics of the participant sample. Groups differed with regard to educational

level, F(1,30)= 8.72, p < 0.01, with younger adults being more highly educated

than older adults.

Table 1.1: Characteristics of participants.

*rated on a 7-point scale, 1= unfinished primary school, 7= university degree, 4 compares to 9 years of

successful formal education, 5 compares to 10 or 11 years of successful formal education.

Each participant was tested individually on two different days, three hours a day,

with no more than three days in between.

General Procedure

A questionnaire was sent to all participants with general questions about a

participant’s profession, education, medical history, drug use and a rating of

subjective well-being. Participants brought their filled in questionnaire when

visiting the laboratory for the first day of the study. Specific procedures for each

prospective memory task are described further on, when each task is discussed in

more detail. At the beginning of the first day, older adults were given the Cognitive

Screening Test.

Each day contained two prospective memory tasks, separated by at least one hour

during which participants were engaged in other tasks. All prospective memory

tasks were administered by use of a computer. The computer screen was placed at

Younger adults (N=16) Older adults (N=16)

M (SD) Range M (SD) Range

Age 20.38 (2.13) 17-24 72.00 (3.93) 66-79

Education* 5.81 (0.4) 5-6 4.88 (1.2) 3-6

Sex (M/F) 6/10 7/9

chapteR 2

32



a comfortable viewing distance of about 50 cm. and ample armrest was present

for the participant in front of the keyboard. Lighting was adjusted for convenient

screen reading. Verbal responses in the letter monitoring task were recorded on

tape and scored on-line by the experimenter.

Letter monitoring task

Material. White characters were presented on a black background screen of a VGA-

monitor. Uppercase letters and numbers were presented in a large, easily readable

font, each character measuring about 1.0 x 0.5 cm.

Procedure and design. Each trial began with the word READY in the middle of the

screen. Next, the participant pressed the spacebar, by means of which the word

disappeared and, after a blank interval of 500 ms, the (Dutch equivalent of the)

instruction WATCH RIGHT or WATCH LEFT appeared in the middle of the screen.

At the beginning of the experiment, the experimenter pressed the spacebar to

start a trial, in order to prevent participants from going too fast. After the practice

trials, participants were able to continue the task themselves at a steady pace.

The instruction remained on the screen for 1 s, and after a blank interval of again

1 s the stimulus sequence began. A sequence consisted of a series of ten frames,

each frame being displayed for 200 ms, separated from one another by a blank

interval of 200 ms duration. Each frame consisted of a pair of letters or digits

presented side by side in the middle of the screen, one to the left and one to

the right of the center of the screen and separated by a distance of 2.5 cm. Ten

frames appeared in turn when, after the tenth, a + or – symbol of similar size to

the characters was presented at the center of the screen. This symbol was also

presented for 200 ms. and was separated by an blank interval of 200 ms from the

preceding and following frame; in this manner, the rhythm of the presentation

remained undisturbed. Following the symbol came three more frames, of which

the first always contained a randomly selected pair of digits and the last two a

randomly selected pair of letters. Of the first 10 pairs of characters, five were letter

pairs, with letters randomly selected but without replacement from the set of all

letters excluding I and O, while the other five were digit pairs, with digits randomly

selected from the range 1-8. A different random sequence was used for each trial

but, to facilitate on-line scoring by the experimenter, the same sequence was used

for each trial for all participants.

Participants were instructed to watch for letters on the side indicated by the initial

instruction and to read them aloud when they occurred. Letters on the incorrect

side were to be ignored. The + or – symbol indicated which side to watch for the



33

last three frames, a + indicating right, a – indicating left; this symbol is referred

to as the second side instruction or SSI (Duncan et al., 1996). On half the trials

(WATCH RIGHT followed by + or WATCH LEFT followed by -) the participant was to

stay with the initial side for the last part of the trial, while on the other half (WATCH

RIGHT followed by – and WATCH LEFT followed by +) he or she was to switch sides.

Thus, a perfectly correct report consisted of five letters from the appropriate side

for the first part of the trial, followed by two from the appropriate side for the last

part, all letters being different.

The experiment started with a training phase to familiarize participants with the letter

monitoring task. Instructions for that phase were written on paper and described only

two of the three basic task requirements in turn: 1] to report letters and ignore digits,

and 2] to report letters only on the correct side as indicated by instructions presented

on screen at the start of the trial. Participants were run through one complete example

on a piece of paper that contained initial instructions and the first ten frames. Then

followed a series of four or, when necessary, more practice trials. Practice trials contained

only the initial instruction and the ten first frames. Practice trials, in series of four trials,

were repeated until the participant correctly reported at least three of the five possible

letters; no participant needed more than eight practice trials.

Subsequently, participants were instructed to watch out for the SSI and to use

it appropriately. An example of a complete trial, now also including the SSI and

subsequent stimuli, was shown on paper and explained in order to ensure that the

participant had correctly understood instructions. Next there followed a block of

12 experimental trials, arranged such that in each successive sub-block of 4 trials

(sub-blocks 1a: trial 1-4; 1b: 5-8; 1c: 9-12) there was one trial of each possible type

(WATCH RIGHT followed by a +, WATCH RIGHT followed by a -, etc.) in random

order. After each sub-block in which the SSI had not been used properly, the

participant was asked if he or she had seen the + and – symbols; this served as a

mild and unspecific form of prompting.

Precise rules were used to control prompts given by the experimenter and answers

to questions of the participant. If participants reported no letters on a trial, or

reported a digit, this error was pointed out. If letters from the wrong side were

reported, however, no feedback was given. To questions of participants, the

experimenter replied with exactly the information already given in the instructions.

After participants had finished the task, they were asked to write down in their own

words what exactly they had to do in the task.

chapteR 2

34




Materials. For the words used in this task, 400 word pairs were drawn from 50

categories. Dutch words, corresponding to the Battig & Montague (1969) category

norms, were taken from Hudson (1982). Because of the difference in frequency of

Dutch and American words, and because of differences in culture, some categories

used by West & Craik (1999) were excluded from the word list constructed for

this task. Excluded categories were female names, male names, diseases, flowers,

colleges or universities, and members of clergy. Eight pairs were selected from

each category, two pairs from each of exemplars 1-4, 5-8, 9-12 and 13-16. For

instance, from the category fruits, the first (e.g. apple) and third (e.g. banana)

most frequently reported category members would be used to form a related pair.

Four hundred unrelated pairs were formed by randomly combining exemplars

from different categories. This resulted in each word appearing twice in the task,

once in a related pair and once in an unrelated pair.

Procedure and design. In each block of 40 trials, 38 trials (19 related and 19 unrelated

word-pairs) were presented in lowercase, grey letters. On these trials, participants

had to press the v-key with the left index finger if the words belonged to the same

category, or press the n-key with the right index finger if the words belonged to

different categories. Of the 40 trials in each block, one was a prospective cue trial and

one a prospective lure trial. For prospective cue trials both words appeared either in

grey, uppercase or in green, lowercase letters. Participants were instructed to press

the b-key with their preferred index finger when such a trial occurred, regardless

of whether the words belonged to the same category or not. For prospective lure

trials, one of the words was presented either in grey, uppercase letters or in green,

lowercase letters, and the other word in grey, lowercase letters. When such a trial

occurred, participants were instructed to ignore the partial prospective cue and

make a regular category judgement response. Half of the prospective cue trials and

prospective lure trials embodied a color change and half a case change.

Participants received written instructions at the beginning of the task. They then

worked through a short series of trials on paper that contained every possible trial

type, under supervision of the experimenter who corrected mistakes and answered

remaining questions. Participants were instructed that the category judgement

and prospective memory components of the task were equally important.

Words were presented in the middle of the screen, one below the other and

separated by a blank line, until a response was made. After a response the words



35

were replaced by a mask (x-symbols) for 1s. The 800 trials were presented in 20

blocks of 40 trials, each block consisting of 20 related and 20 unrelated pairs. The

presentation order of word-pairs was randomized, as was the occurrence of the

prospective target and lure target within a block.

During the task, the participant was given two opportunities to take a short break.

After completion of the task, participant wrote down in their own words what they

had to do in the task.

Pictures task

Materials. Black and white pictures of foods, objects, and animals from Cycowicz

(1997) and Snodgrass & Vanderwart (1981) were used. The pictures were 400 x

300 pixels (filling about one third of the screen) and presented on a VGA monitor.

In total 810 pictures were presented to the participants. Because of the original

amount of pictures, 260 of Snodgrass and 400 of Cycowicz, some pictures appeared

more than once, or a different exemplar from the same semantic category (e.g.

both an armchair and a kitchen chair) appeared.

Procedure. Participants were instructed to indicate for each object whether they

thought it was: 1] alive and natural, 2] not alive and natural, or 3] not alive and

not natural, and to indicate their choice by pressing the keys, 1, 2 or 3 respectively.

Every picture remained on the screen for 5 s maximally, or disappeared when the

participant pressed a key. After 500 ms ISI the next picture appeared on the screen.

With every picture, the possible key-choices and their particular meaning, was

shown on the screen.

At the beginning of each trial block, the prospective-memory instruction appeared

on the screen that told the participant to look out for a particular object (e.g., chair;

the critical object was defined verbally) to appear in the next block. When that

object appeared, participants were instructed not to make a category judgment

response but to press the 0-key instead. Each prospective target was used only

once and neither this picture nor a picture from a semantically related category

was shown again during the remainder of the task.

The task consisted of 35 blocks, each block containing one prospective target. Block

length varied from 13 to 35 trials. The first 10 trials of a block never contained a

prospective target, and the block ended two trials after the prospective target had

been presented. The mean number of trials per block was 23.

chapteR 2

36



Participants received written instructions and were given no practice trials. During

the task, there were two opportunities to take a short break. After the task had

been completed, participants wrote down in their own words what they had to

do in the task.


Materials. The letters used were A, B, C, and D, all written in capitals and in a

large and easily readable font. The letters where white and displayed on a black

background screen of a VGA monitor.

Procedure and design. The background task was a four-choice reaction time task.

One of the four stimulus letters was presented on each trial. Letters A or C required

pressing the v-key with the left index finger and the letters B or D pressing the

n-key with the right index finger. Letters remained on the screen until a response

was made. Errors were indicated by a beep of 1 s duration, and had to be corrected

before the task continued. The next stimulus appeared 1500 ms after a correct

response was registered. Each block consisted of 64 trials and participants received

feedback on screen regarding number of errors and mean response latency at

the end of a block. Order of stimuli within a block were randomized, with the

restriction that the critical prospective event, i.e. a letter appearing three times

consecutively, occurred exactly four times in each block, and that a letter could

appear consecutively at most three times.

The experiment consisted of 11 trial blocks. The first two blocks and the last block

were background-task only blocks, in which only the background task had to be

performed. The remaining blocks were prospective-memory blocks, in which

participants received an additional prospective-memory instruction. This instruction

told them to look out for any letter appearing three consecutive times and, when this

occurred, not to press the v-key or the n-key but to press the spacebar instead. For the

prospective task there were two conditions. In four prospective-memory blocks, the

program simply continued after a prospective-memory miss (no-feedback condition).

In the other four blocks, the program halted after a prospective-memory miss, to

continue only after the prospective action (i.e., pressing of the space-bar) had been

registered (feedback condition). For even-numbered participants, the series of eight

prospective-memory blocks began with two blocks in the no-feedback condition,

then two blocks in the feedback condition, then two more blocks in the no-feedback

condition, and finally two more blocks in the feedback condition. For odd-numbered

participants, this ordering of the feedback and no-feedback conditions was reversed.



37

Initial written instructions pertained to the background task only. Participants were

instructed to respond as fast as possible but not to make more than three errors

in a trial block. Prior to the first prospective-memory block, participants received

further written instructions regarding the prospective-memory component, and

they worked through a short series of trials on paper that contained a prospective

target event, with the experimenter providing feedback and answering possible

questions. During the instructions, no mentioning was made of the peculiarities

of the prospective-memory feedback condition. Prior to the final, background-task

only block, participants were informed that the prospective-memory requirement

was no longer relevant. After having completed this task, participants wrote down

in their own words what they thought they had to do in this task.

2.3 ResultsUnless indicated otherwise, an alpha level of .05 was used in all statistical tests. For

each of the four tasks, results for background task performance will be presented

first, followed by results for prospective-memory task performance.

Letter-monitoring task.

The first twelve experimental trials were examined with the same procedure as

used by Duncan et al. (1996). The twelve trials were divided in 3 sub-blocks of

4 trials. The use of the second-side-instruction (SSI) was assessed by means of a

dichotomization. A trial was considered as ‘passed’ if more letters were reported

from the correct side than from the incorrect side both before and after the SSI.

A sub-block was scored as ‘passed’ if it contained at least one ‘passed’ switch trial

(when the SSI was different from the initial instruction) and one ‘passed’ stay trials

(when the SSI and initial instruction were the same).

The distribution of the number of failed sub-blocks across participants was

almost identical for younger and older adults. Thirteen younger and twelve

older participants passed all sub-blocks. Five, two younger and three older, of

seven participants, who did not pass all sub-blocks failed one sub-block and two

participants, one of each age-group, failed two sub-blocks. None of the participants

failed all three sub-blocks. One younger participant with one failed sub-block failed

the last sub-block and one older participant with one failed sub-block failed the

second sub-block. The other five participants failed the first sub-block (in case of

the one failed sub-block) or failed the first two sub-blocks (in case of the two failed

sub-blocks). An overview of results from more detailed analyses of performance

aspects is presented in Table 2.2.

chapteR 2

38



Table 2.2: Letter-monitoring task. Means (M) and standard deviations (SD) of performance on

the background task (proportion correct responses before SSI), and of performance

on the prospective memory task (proportion of switches correct) per subblock.

Old Young

M SD M SD

Background task performance

Subblock 1 0.94 0.09 0.93 0.19

2 0.96 0.08 0.96 0.08

3 0.92 0.12 0.94 0.10

Prospective memory performance

Subblock 1 0.81 0.36 0.81 0.31

2 0.88 0.29 0.91 0.20

3 0.97 0.13 0.94 0.17


Background task performance was defined as letter-monitoring task performance

before the SSI. The total number of letters to be reported before the SSI was five.

Younger and older adults correctly reported an average of 4.72 and 4.70 letters,

respectively (F<1). No differences in performance between having to report the

letters on the right side and letters on the left side were found.


Prospective memory performance in the letter-monitoring task was defined as the

proportion of correct use of the SSI on switch trials. For this analysis only trials were

used for which the performance on the letters before the SSI could be considered

as ‘passed’ (again performance was considered as ‘passed’ if more letters were

reported from the correct side than from the incorrect side). According to this

procedure, ten older adults and ten younger adults showed perfect prospective

memory performance. Both age groups showed an increase in mean prospective

memory performance over sub-blocks (F(2,29)=4.8, p<.05), and no overall

difference in prospective memory performance between age groups was obtained

(F<1). The ‘perfect’ prospective memory performance of most participants



39

prohibited assessment of forgetting and recovery ratios and of split-half reliabilities.


Five older participants were excluded from analyses. These participants performed

extremely poorly on the prospective memory task, exhibiting excessive numbers

of false alarms or misses, and they all failed to correctly reproduce the task

instructions at task debriefing. Two of these participants also performed poorly on

the background task, with accuracy more than two standard deviations below the

group mean. All relevant performance results are presented in Table 2.3.

Table 2.3: Word comparison task. Mean (M) and standard deviations (SD) of reaction times (RT)

and proportion of correct responses on (not lure) background task trials, on lure trials and on

prospective memory trials.

Old Young

M SD M SD


RT (ms.) 1605 313 1469 160

Proportion correct 0.91 0.03 0.88 0.03

Performance on lure trials

RT (ms.) 2242 507 2048 351



RT (ms.) 1772 434 1285 307


chapteR 2

40



Background task performance.

Response speed was similar for younger and older adults (F(1,25)=2.2), with older

adults being somewhat more accurate (F(1,25)=5.84, p< .05).

Prospective memory performance.

Younger adults performed more accurately than older adults on the prospective

memory trials (F(1,25)=9.8, p<.01). Also, prospective memory hits were much

faster for younger than for older adults (F(1,25)=11.7, p<.01). Older adults tended

to make more false alarms on lure trials than young adults, but this difference did

not approach significance (F(1,25)=1.85). Also, correct reaction times on lure trials

were not significantly different for the two age groups (F(1,25)= 1.38).

Forgetting and recovery ratios were computed following the procedure outlined

by Maylor (1996). An instance of forgetting is defined as a prospective memory

hit followed by a prospective memory miss, whereas an instance of recovery is

defined as a prospective memory miss followed by a prospective memory hit. The

forgetting ratio was computed by dividing the number of forgetting instances by

the number of opportunities for forgetting, and the recovery ratio by dividing the

number of recovery instances by the number of opportunities for forgetting.

Mean forgetting and recovery ratios for both age groups are presented in Table

2.6. Forgetting ratios were higher for older than for younger adults (F(1,24)=7.36,

p< .05), and recovery ratios were lower for older adults, although this latter

difference failed to reach significance (F(1,15)=2.5). For older adults, forgetting

and recovery ratios were significantly different from zero (t(10)=3.4, p<.01, and

t(8)=6.6,p<.001, respectively). For younger adults, this was the case only for

recovery ratios (t(7)=12.8 p<.001).

Cue accessibility was assessed by comparing response latencies of correct

background task responses with those of prospective memory hits. For younger

adults, prospective memory hits were faster than correct background task

responses, whereas the opposite pattern was found for older adults; this interaction

of trial type and age group was significant (F(1,25)=8.1, p<.01).

Reliability

Split-half reliability was assessed by correlating prospective memory performance

and background task performance in odd-numbered with that in even-numbered

trial blocks. Results are presented in Table 2.7.



41

Pictures task

Two older participants were excluded from statistical analyses, because they never

made a prospective response and failed to correctly reproduce task instructions

during debriefing. Both participants performed within the normal range on the

background task. All relevant results are presented in Table 2.4.

Table 2.4: Pictures task. Means (M) and standard deviations (SD) of reaction times (RT) on

background task trials, and prospective memory trials, and of the proportion correct responses on

prospective memory trials.

Old Young

M SD M SD


RT (ms.) 1626 239 1066 186


RT (ms.) 2040 749 1366 243

Proportion correct .69 .23 .84 .11


Older adults responded more slowly than younger adults in the background task

(F(1,28) = 51.97, p<.001).


Prospective memory accuracy was higher for younger than for older adults (F(1,28)

= 5.10, p<.05). Also, prospective memory hits were faster for younger than for

older adults (F(1,28)=11.61,p<.01).

Forgetting and recovery ratios are presented in Table 2.6. A significant effect of

age group was found for recovery ratio (F(1,28)= 13.80, p<.01), with younger

adults being more likely to recover, but not for forgetting ratio (F(1,28)=1.06). The

forgetting and recovery ratios differed significantly from zero (older: forgetting

ratio t(13)=3.8, recovery ratio t(13)=7.2; younger: forgetting ratio t(15)=5.5,

recovery ratio t(13)=22.5; all tests p<.001).

chapteR 2

42



To assess cue accessibility, latencies of background task responses were compared

with latencies of prospective memory hits. Response latencies of prospective

memory hits were larger than those of background task responses (F(1,28)=14.0),

p<.01), and similarly so for both age groups (F(1,28)<1).

Reliability

Split-half reliabilities, assessed by correlating prospective memory performance

and background task performance in odd-numbered with that in even-numbered

trial blocks, are presented in table 2.7.


Trials with reaction times slower than four seconds were excluded from analysis.

Also, four participants, one younger and three older adults, were excluded from

statistical analyses, as they scored very few prospective-memory hits. Two of

these participants (both older) failed to correctly reproduce prospective memory

instructions during task debriefing (re-analysis of the data with only these two

participants being excluded from analysis yielded essentially the same pattern of

results as that presented below). These four participants did press the key assigned

to the prospective memory response at some regular trials (mean of four times

during the task), but this frequency of false alarms was not different from that of

the other participants. For the three older participants that were excluded, mean

response latency and accuracy in the background task were within two standard

deviations of the means of the other older participants. The excluded younger

participant responded at a normal speed on the background task but put in the

lowest accuracy score of all participants (.91 proportion correct). All relevant results

are presented in Table 2.5.



43

Table 2.5: ‘Three in a row’ task. Mean (M) and standard deviations (SD) of reaction times (RT)

and proportion correct responses on background task trials, and prospective memory trials during

the background task, prospective memory without feedback, and prospective memory with

feedback conditions.


To examine the effects of prospective memory instructions on background task

performance, performance on ‘background task only’ blocks (the first and last block)

was compared with ‘background task plus prospective memory’ blocks (pooled

across conditions feedback and no-feedback). Prospective memory instructions did

not affect accuracy of background task performance (F(1,26) = 3.15). Older adults

were somewhat more accurate than young adults (F(1,26)= 5.44). No interaction

of trial type and age group on accuracy was found (F<1).

Response latencies showed a different pattern. As compared to the ‘background

task only’ condition, reaction times in the prospective memory conditions were

strongly elevated (F(1,26) = 76.8, p<.001). Also, older participants responded more

slowly than younger participants (F(1,26) = 12.78, p<.01), but the interaction of

trial type and age group was not significant (F(1,26)= 1.82).

Within the prospective memory condition, no main effect of feedback or an

interaction of this factor with age group was obtained for either accuracy or speed

of background task performance (F’s < 1).

Old Young

background only No feedback feedback background only no feedback feedback

M SD M SD M SD M SD M SD M SD

Background task

RT (ms.) 622 152 713 228 722 172 462 60 525 70 539 86

Proportion correct 0.97 0.01 0.98 0.02 0.98 0.09 0.97 0.02 0.97 0.01 0.97 0.02

Prospective memory

RT (ms.) 703 336 756 262 400 48 417 75

Proportion correct 0.86 0.11 0.87 0.11 0.84 0.12 0.88 0.11

chapteR 2

44



Prospective memory performance.

Overall prospective memory accuracy was not different for the two age groups

(F<1). Prospective memory hits were substantially slower for older as compared

to younger adults (F(1,26)= 27.21, p< .001). The main effect of feedback and the

interaction of this factor with age group on prospective memory accuracy and

speed of prospective memory hits did not approach significance.

Forgetting and recovery ratios were computed for the two feedback conditions

separately, but since this gave essentially similar results, the conditions were

pooled. The mean ratios are presented in Table 2.6. No significant effects of age

group on the ratios were found. T-tests revealed that forgetting and recovery ratios

were different from zero for both age groups (older: forgetting t(12)=5.4,p<.001,

recovery t(12)=17.5,p<.001; younger: forgetting t(14)=5.3,p<.001, recovery

t(14)=20.8, p<.001).

Cue accessibility was assessed by comparing mean RT for correct background task

responses with that for prospective memory hits. For younger adults, prospective

hits were substantially faster than correct background task responses. In contrast,

for older adults, prospective hits were slightly slower than correct background task

responses. Both the main effect of trial type (F(1,26)= 6.8)) and the interaction of

this factor with age group were significant (F(1,26)= 10.2, p< .01).

Reliability

Split-half reliability was assessed by correlating prospective memory performance and

background task performance for the four odd-numbered prospective memory blocks

with that for the four even numbered block. Results are presented in Table 2.7.

Table 2.6: Forgetting and recovery ratios per task.

Forgetting Recovery

Old Young Old Young

M (SD) M (SD) M (SD) M (SD)

Word comparison .31 (.30) .06 (.15) .69 (.32) .90 (.20)

Pictures .25 (.25) .18 (.13) .58 (.30) .91 (.15)

‘Three in a row’ .16 (.10) .18 (.13) .86 (.18) .90 (.17)



45

Table 2.7: Guttman split-half reliabilities computed for younger and older adults separately

and for the two groups collapsed. Reliabilities are computed for accuracy performance on the

backgroundtask (backgr), and for accuracy of the prospective memory performance (PM).

2.4 DiscussionIn the four different tasks different numbers of participants were excluded from

analyses as prospective memory non-performers. One younger adult was exclud-

ed from the ‘three in a row’ task analysis; all other prospective memory non-per-

formers, three in the ‘three in a row’ task, five in the word comparison task, and

two in the pictures task, were older adults. Two older adults were excluded in all

three tasks, in each case based on failures to reproduce the relevant instructions

(in the letter-monitoring task analysis, no participants were excluded) during task

debriefing. The older prospective memory non-performers in the ‘three in a row’

and pictures tasks performed well within normal range on the associated back-

ground tasks, with respect to accuracy as well as response latency. This suggests

some degree of specificity of their prospective memory problems in these tasks.

Two of the five prospective memory non-performers in the word comparison

task also performed very poorly with respect to accuracy in background task per-

formance, suggesting a profound misunderstanding of task instructions; the other

three non-performers did well on the background task.

In general then, complete failures to follow prospective memory instructions were

limited almost entirely to the group of older adults. In some cases such complete

failures could be traced to failures to understand or remember the relevant

instructions. Also, such failures typically occurred in conjunction with normal levels

of background task performance.

Letter-monitoring task

The present results for the letter monitoring task are particularly clear in revealing

no age-related differences in prospective memory performance, with performance

split-half reliability (n) All Old Young

backgr PM backgr PM backgr PM

Word comparison task .89 (27) .95 (27) .92 (11) .93 (11) .82 (16) .92 (16)

Pictures task .91 (30) .93 (14) .77 (16)

‘Three in a row’ task .85 (28) .53 (28) .78 (13) .31 (13) .83 (15) .72 (15)

chapteR 2

46



being close to perfect for both age groups. This outcome stands in marked contrast

to that reported by Duncan et al. (1996), who found distinctly poorer prospective

memory performance in their sample of older adults as compared to younger

adults. The reasons for these different findings are not entirely clear. A potentially

relevant difference in experimental procedure is that we provided participants with

some limited opportunity to practice the letter monitoring background task before

introducing the prospective-memory instructions, whereas participants in the

Duncan et al. (1996) study were introduced to the full set of task requirements right

from the start. Our reason for incorporating such an initial but limited practice phase

was that we had found in pilot testing that older adults in particular tended to be

“thrown off” by the rapid rate of stimulus presentation in the letter monitoring task,

and needed some practice before being able to perform this task at a reasonable

level of accuracy (i.e., a level of accuracy that permitted accurate assessment of

participants’ ability to follow second-side instructions). Conceivably, however,

this opportunity to initially ‘master’ the background task, may have significantly

enhanced our participants’ ability to subsequently incorporate the additional

prospective-memory requirements into an overall goal structure governing task

performance (Duncan et al., 1996). In addition, our sample of older adults may

have differed from the sample studied by Duncan et al. (1996, Experiment 2) in

consisting primarily of relatively high-functioning elderly. Consistent with this latter

conjecture is the finding by Duncan et al. (1996) that prospective memory failures

in the letter monitoring task is strongly characteristic of people from the lower

part of Spearman’s g distribution in the normal population (see also Cherry &

LeCompte, 1999).

Given the present results and those reported by Duncan et al. (1996) and as robust

age differences in prospective memory performance were obtained for two of the

four tasks in the present study, it would appear that the letter monitoring task,

in the form used by Duncan and coworkers and in the present study, may be a

particularly useful tool for detecting relatively severe cases of frontal dysfunctioning;

on the other hand, it is relatively insensitive, with prospective memory performance

typically near ceiling, to more subtle forms of decline of frontal functioning as may

occur in the course of normal aging.


Results for the word comparison task replicated those reported by West and Craik

(1999) reasonably well. No significant age effects on performance in the semantic

background task were found. Older adults made many more prospective memory



47

errors than young adults but, unlike what has been reported by West and Craik

(1999), they did not commit more false alarms on lure trials; the reasons for this

latter discrepancy are not clear. Non-zero forgetting and recovery ratios provided

clear evidence for temporal fluctuations of prospective memory success and failure.

More specifically, the higher incidence of prospective memory errors in older adults

could be shown to result from a higher forgetting ratio and a lower recovery ratio

in the elderly. Finally, whereas for younger adults prospective memory hits were

faster than regular background task responses, the opposite pattern was found for

older adults. Following West and Craik (1999), this robust pattern of slower and

more error-prone responses to prospective targets found for older adults, strongly

suggests a diminished ability to maintain prospective cue-action associative

encodings in a highly activated and accessible state.

As we have noted previously, the prospective-memory instructions for the word

comparison task are quite complex and are thus likely to impose considerable

demands on participants’ ability to establish the relevant cue-action associative

encodings and to maintain them in a suitably state of activation during background

task performance. The present results, and those reported by West and Craik

(1999), appear to confirm these expectations. Thus, the present word comparison

task would appear to be suitably sensitive to relatively mild forms of frontal

dysfunctioning that may occur in the course of normal aging. From a methodological

perspective, then, the word comparison task and the letter monitoring task may

have complementary sensitivities, the former being most useful to detect and

distinguish between relatively mild forms of frontal dysfunctioning and the latter

most useful in cases of more severe forms of frontal dysfunctioning. Also, split-half

reliabilities for the word comparison task were very high, for both background

task performance and prospective memory performance. This suggests that the

word comparison task may represent not only a sensitive but also a reliable tool for

assessing age effects on prospective memory abilities.

One potential drawback of the word comparison task concerns the relatively large

proportion of prospective memory non-performers among the older participants

in the present study. As no such problems were reported by West and Craik (1999),

this may simply have been a case of bad luck. However, this aspect of our results

may also indicate that several elderly experienced considerable difficulties in arriving

at a suitable representation of the complex task instructions in working memory.

When working memory capacity is defined as the capacity to attain as well as

maintain temporary goal representations (e.g., De Jong, Cools, & Berendsen,

chapteR 2

48



1999), age-related reductions in this capacity might result in a larger proportion of

prospective-memory nonperformers in the elderly as well as in reduced consistency

of prospective-memory success amongst prospective-memory performers in the

elderly. Thus, as long as the proportion of nonperformers remains acceptably low,

the presence of some proportion of nonperformers does not disqualify the task as

a suitable tool for studying age effects on prospective memory abilities.

Pictures task

Large and robust age differences in prospective memory performance were

obtained in the picture categorization task. The higher incidence of prospective

memory errors in older adults could be traced to a lower recovery ratio and,

though to a much lesser extent, a higher forgetting ratio in the elderly. Several

characteristics of the task may have contributed to this outcome. The prospective

target was defined verbally, so that target recognition necessarily involved

semantic processing or categorical perception. Also, prospective targets did not

differ in any perceptually salient way from non-target events, thereby enhancing

the dependency of correct prospective memory performance on the ability to

maintain the relevant target-action encoding in a highly activated state. Finally, a

new prospective target was defined in every block, thus reducing possibilities for

automatization of prospective memory performance that may develop when the

same prospective target is encountered repeatedly. The relative contributions of

these task characteristics to age sensitivity of prospective memory performance in

this task remain to be sorted out.

Based on extensive evidence that semantic knowledge remains stable in later

adulthood (Horn & Donaldson, 1976), we expected younger and older adults to

perform at similar levels on the background picture categorization task. However,

and unlike what we found in the word comparison task, older adults were

substantially slower than younger adults in eliciting their category judgments. Task

debriefing suggested a possible reason for this unexpected result. Several of the

pictures used turned out to be rather ambiguous with respect to their belonging

to one of the relevant categories (i.e., natural vs. not-natural and living vs. non-

living). As extensive evidence suggests that older adults tend to favor accuracy of

performance over speed more strongly than younger adults, interspersion of such

ambiguous stimuli may well have induced a more conservative response strategy in

our elderly participants. It is not clear whether and to what extent such a potential

age-related strategy difference with respect to background task performance may

have been responsible for the age effects on prospective memory performance



49

in the present task. In future applications of the picture categorization task such

ambiguities should be carefully avoided.

Split-half reliability of prospective memory accuracy in the picture categorization

task was quite high. Thus, the present results suggest that this task may provide a

quite sensitive and reliable tool for assessing age effects on prospective memory

abilities.


In line with much previous evidence (e.g., Salthouse, 1996), regular responses

in the three-in-a-row task were substantially slower for older as compared to

younger adults. More importantly, no age effects on prospective memory accuracy

were obtained in this task. One possible explanation for this latter result can be

excluded. Reaction times in the background task were substantially elevated in

the prospective memory condition as compared to the background task-only

condition, but this was the case for both age groups. This would appear to render

unlikely the possibility that older adults, by emphasizing the prospective memory

task component more strongly than younger adults, may have strategically

compensated for their diminished prospective memory abilities (Cherry &

LeCompte, 1999; Maylor, 1998).

Surprisingly, provision of feedback after prospective memory errors, by forcing

participants to execute the appropriate prospective action, failed to have any

effect on prospective memory performance. This result contrasts markedly with the

beneficial effects of feedback and prompting on prospective memory performance

that were reported by Duncan and coworkers (1996). A potential explanation of

this outcome, which may also serve to explain the lack of age effects on prospective

memory accuracy and the relatively low split-half reliability of prospective memory

accuracy in the present task, is that many participants performed at ceiling levels

of accuracy even in the no-feedback condition. It seems likely that the relatively

high frequency of occurrence of the prospective target event (every 16th trial on

average) may have rendered this prospective memory task too easy, thus severely

limiting its sensitivity to detect possible age-related declines in prospective memory

abilities. We intend to more systematically study the effects of relative frequency

of prospective target events on prospective memory performance in future studies

that will employ adapted versions of the present task.

Interestingly, prospective memory hits were faster than regular background task

responses for younger adults, whereas the opposite pattern was found for older

chapteR 2

50



adults, similar to what was found in the word comparison task. This suggests that

younger adults may in fact have been able to maintain the target-action associative

encoding in a more highly activated and accessible state than older adults (West &

Craik, 1999), and that the speed at which prospective actions can be initiated may

in some cases provide a more sensitive measure of prospective memory abilities

than prospective memory accuracy. Of course, these possibilities should at present

be considered speculative and open to further investigation.

Correlational analyses

Given the small sample sizes in the present study, results from correlational

analysis of the present results should be considered preliminary. Nevertheless,

as such analyses can be used to address a number of important theoretical and

practical issues, we will report here the results of correlational analyses that were

performed in order to elucidate the nature of the age effects on prospective-

memory accuracy in the two tasks that produced such effects, the word

comparison task and the pictures task.

A sizable zero-order correlation, computed with younger and older adults pooled

and with all prospective memory non-performers left out, was obtained between

prospective-memory accuracies in the word comparison and the pictures task

(r=0.54, p<.01). With age partialed out, this correlation dropped markedly to

0.38 (p<.05). Also, the quasi-partial correlation, proposed by Salthouse (1994)

as a measure of the proportion of shared age-related variance, was 0.85. Each of

these two tasks was argued earlier to provide a quite sensitive and reliable tool

for assessing age-related changes in prospective memory abilities. The present

finding of a substantial, and largely age-related, correlation between prospective

memory accuracies in these two tasks additionally suggests that the age

sensitivities of the two tasks can be largely traced to the same underlying factors;

in that sense, the word comparison task and the pictures task would seem to

provide almost equivalent tools for assessing age-related changes in prospective

memory abilities. Note that these results might also be taken to indicate a certain

degree of construct validity for the theoretical concept of prospective memory

ability.

Various reports of distinct relationships between ‘basic speed of performance’ and

indices of a large variety of other mental abilities, have given rise to the influential

and powerful notion that individual differences, including age-related differences,

in a vast range of mental abilities can be largely accounted for in terms of direct



51

or indirect effects of individual differences in basic mental speed (e.g., Salthouse,

1996). For present purposes, this theoretical perspective would hold that the

age effects on prospective memory performance obtained in this study might

be explained largely, if not exclusively, in terms of age effects on basic mental

speed. In order to address this issue, we selected mean reaction time in the

three-in-a-row task as the best available measure of ‘basic mental speed’. Zero-

order correlations of mental speed and prospective memory accuracy were 0.80

(p<.01) and 0.31 (p<.10) for the word comparison and pictures task, respectively.

With mental speed partialed out, the correlation of prospective memory

performance and age dropped from 0.53 to 0.26 for the word comparison

task, and from 0.42 to 0.32 for the pictures task. These results are reasonably

consistent with a mental-slowing account of the age effects on prospective

memory accuracy obtained in the present study.

Though we cannot exclude such a mental-slowing account of the present results,

we believe that closer scrutiny points to an alternative interpretation. To develop

the argument, we will focus on the very high zero-order correlation (r=0.80)

between our measure of ‘basic mental speed’ (i.e., mean reaction time in the

three-in-a-row task) and prospective-memory accuracy in the word comparison

task. Several considerations are relevant here. First, age effects on speed of

performance in the background word comparison task were rather small and

insignificant. This replicates previous findings of relative sparing of semantic

processing in old age, and would seem inconsistent with any simple account in

terms of an age effect on basic mental speed. Second, prospective memory hits

in the word comparison task were faster than regular background responses for

younger adults, whereas this difference was reversed for older adults; the same

pattern was also found in the three-in-a-row task but, for reasons that remain

to be clarified, not in the pictures task. Importantly, such a pattern cannot

be accounted for in terms of age effects on basic mental speed, whereas an

explanation in terms of age differences in cue accessibility, i.e., the ability to keep

the cue-action associative encoding in a highly activated and accessible state,

readily presents itself, as we discussed earlier. If this line of reasoning is accepted,

how should the strong correlation between ‘basic mental speed’ and prospective

memory accuracy in the word comparison task be explained?

We suggest that this explanation becomes apparent when it is realized that

speed of performance in the three-in-a-row task may be determined not so

much by basic mental speed but by the ability to keep the four stimulus-

chapteR 2

52



response associations in this task in a highly activated and accessible state.

Indeed, Gottsdanker (1980) used the notions of a limited preparatory capacity

and the need to divide this capacity among the various competing stimulus-

response associations in order to keep the associations activated, to develop a

comprehensive account of effects of number of stimulus-response alternatives

on reaction time (i.e., Hick’s law). In summary, what we suggest is that average

response speed in the three-in-a-row task cannot serve as a pure index of ‘basic

mental speed’ but, rather, reflects the capacity or ability to keep the various

relevant stimulus-response associations activated so that the requisite response

can be quickly and effectively triggered by an imperative stimulus. In this

conception, correlations between prospective memory performance and speed

of reaction-time performance are mediated by individual differences in this latter

capacity, rather than by individual differences in basic mental speed. We believe

that this theoretical approach provides a more plausible account of the results

of the present study, and may also apply as an alternative theoretical account

in many other studies where individual differences in basic mental speed have

been argued to underlie patterns of inter-correlations between various indices of

mental abilities.

2.5 ConclusionsOf the four prospective memory tasks examined in this study, the word comparison

task and the pictures task were found to provide sensitive and reliable tools for

assessing effects of normal cognitive aging on prospective memory abilities.

Because the four tasks varied in a largely non-systematic fashion on various sorts

of dimensions, which dimensions underlie the sensitivity of a task, or lack thereof,

to age effects on prospective memory performance cannot be determined on the

basis of the present results; Nevertheless, the present study has achieved its main

and important practical aim of identifying two tasks that provide suitable tools for

research on aging and prospective memory performance, especially in studies that,

as in the present study, employ healthy and relatively high-functioning elderly.

From the theoretical perspective on control of prospective memory performance

that we adopted in this paper, the word comparison task and the pictures task

have somewhat complementary characteristics. Prospective targets in the word

comparison task could be distinguished from regular stimuli in the background

task on the basis of salient perceptual differences and thus constituted relatively

potent retrieval cues. However, prospective memory instructions in this task were

quite complex and thus presumably imposed considerable demands on people’s

ability to represent the relevant prospective intentions in working memory and



53

keep these intentions sufficiently activated for them to be reliably triggered by

the retrieval cue. In contrast, prospective memory instructions in the pictures

task were relatively simple and straightforward. In this task, detection of the

prospective target was made relatively difficult by means of several manipulations.

As a consequence, the prospective target constituted a relatively weak retrieval

cue, requiring the prospective intention to be maintained in a highly activated

state for it to be reliably triggered by the weak retrieval cue. Thus, both tasks, but

in different ways, focussed on people’s ability to maintain prospective intentions

properly activated and accessible. Consistent with this perspective, prospective

memory accuracies in the two tasks were substantially correlated, the correlation

being mediated largely by effects of age on the accuracies.

54

chapteR 3

55

Cognitive aging and task switching

Chapter 3Cognitive aging and task switching

56

3.1 IntroductionAging is commonly known to be accompanied by a decline in performance in

everyday life and in the laboratory. Presently, theoretical considerations about

effects of normal aging on cognitive functions are often concerned with cognitive

control functions. These functions imply selection, scheduling, monitoring and

coordination of processes that are responsible for perception, memory and action

(Kramer, Hahn and Gopher, 1999). In general, performance in tasks that place

a high demand on cognitive control functions has been found to be negatively

affected by old age (Braver, Barch, Keys, Carter, Cohen, Kaye, Janowsky, Taylor,

Yesavage, Mumenthaler, Jagust and Reed, 2001; de Jong, 2001; Kramer, Larish

& Strayer, 1995; Mayr & Kliegl, 1993). The concept of cognitive control implies

endogenous control of processes and, in that sense, can be contrasted with

situations in which behavior is cued, triggered or prompted explicitly from the

environment. Evidently, situations which place a high demand on cognitive control

processes are situations in which tasks to be performed are novel or characterized

by weak environmental support.

In task-switching situations, a certain task is to be performed in face of a set of

potentially applicable alternative tasks. A consistent finding in the literature on

task-switching is when a task is repeated, performance is better than when the

current task is different from the previous one. Usually this switch cost is found in

accuracy as well as in the speed of performance and has usually been found to be

reduced but not diminished when ample time was provided to prepare for the new

task (‘residual switch-costs’). Switch-costs are widely believed to reflect control

processes that are engaged if one has to switch between two or more competing

tasks (Allport and Wyllie, 2000).

Given that some evidence suggests that the lateral prefrontal cortex is involved in

task-switching (Braver, Reynolds and Dondaldson, 2003; Dreher, Keochlin, Ali and

Grafman, 2002; Dove, Pollman, Schubert, Wiggins and Von Cramon, 2000) and

that cognitive control processes play a dominant role and that switching between

tasks is common aspect of daily life behavior, it is of importance to study whether

and how switching between tasks is affected by advancing age. Several studies

have examined the effects of aging on task-switching and results are in some

respects not very consistent. A short overview of task-switching paradigms used

will be given.

chapteR 3

57


Task-switching paradigms

In task-switching paradigms one of several alternative tasks (usually choice reaction

time (RT) tasks) is to be performed on every trial. If on the current trial the task to

be performed is different from the task on the previous trial, it is called a switch

trial; if the task is the same as the previous task, it is called a non-switch trial. If

tasks are to be performed in an unpredictable order, a cue or an instruction signal

precedes the presentation of the stimulus, signaling the task to be performed (e.g.

Meiran, 1996). If there are only two possible tasks, sometimes a cue is used that

does not explicitly cue the task at hand, but merely signals to switch to the other

task (Salthouse, Fristoe, McGuthry and Hambrick, 1998). In some paradigms,

switch trials occur in a predictable order, simple alternating order (Allport, Styles

and Hsieh, 1994), or alternating in a more complex manner. In some ‘predictable

order’ paradigms tasks or switches are cued implicitly (e.g. Rogers and Monsell,

1995) or explicitly (e.g. Kramer, Hahn and Gopher, 1999) and in other, the subjects

need to keep track of the alternating tasks internally (e.g. Kray and Lindenberger,

2000). Furthermore, the stimuli employed are typically ambiguous as to which task

is to be performed, making it necessary to process the cue or to keep track of the

task-sequence. In some studies though, univalent stimuli are used, in which case

the stimulus itself provides information about what the relevant task is. Usually a

preparation interval (PI) is provided between the cue or the previous response (in

the paradigms without cues) and the imperative stimulus in order to give subjects

time to prepare for the upcoming task and to make it possible to study residual

switch costs. To study the effect of PI, the length of the interval is varied in a

blockwise manner or on a trial-by-trial basis. Finally, paradigms differ in whether or

not the same responses are used in the tasks (response-set overlap).

Switch-costs

Heterogeneous blocks, in which switch trials and non-switch trials are mixed,

can be contrasted with homogeneous blocks, in which the same task has to be

performed on all trials. Although performance on non-switch trials in heterogeneous

blocks is usually better than on switch-trials, often it does not reach the level of

performance in homogeneous blocks. The difference in performance on non-

switch trial in heterogeneous block on one hand and performance on (fixed) trials

in homogeneous block on the other are called ‘global switch-costs”. Theoretical

accounts for global switch-costs can be divided in two general theoretical frames.

One account suggests that the slowing of non-switch RTs compared to fixed RTs,

is caused by involuntary priming (e.g. Allport and Wylie, 2000). An alternative

explanation for global switch-costs is that it reflects the demand to maintain in

58

chapteR 3 Cognitive aging and task switching

working memory two task-set instructions and to coordinate these instructions with

information of the actual appropriate task-set (Kray and Lindenberger, 2000, Rogers

and Monsell, 1995). Another account suggests that incomplete disengagement of

prior task sets or incomplete task set reconfiguration on non-switch trials is due

to control strategies subjects adopt (e.g. De Jong, 2000). According to the latter

account subjects might use the knowledge of having to perform the other task

again shortly, by voluntarily not completely disengaging the previous task-set,

although they may be capable of doing so.

The difference in RT between non-switch trials and switch trials in heterogeneous

blocks is usually sizable and is called ‘local switch-cost”. Compared to short

PI’s, switch-costs are often found to be considerably reduced (but usually not

eliminated) at long PI’s (‘residual switch costs’). The exact cause of residual switch-

costs is a topic of some controversy. Different accounts of residual switch costs

have been proposed. One account, by Rogers and Monsell (1995), proposes a

distinction between an endogenous and an exogenous component of task-set

reconfiguration. The endogenous component can be initiated and carried out

without external support or triggering, but cannot attain a fully reconfigured task-

set. For the reconfiguration to complete a relevant stimulus is needed. According

to this account, residual switch-costs reflect this latter exogenous component. The

idea that residual switch costs reflect an fundamental involuntary limitation, is also

underlying theories that suggest that switch costs are due to involuntary processes

of persisting activation of the previously relevant task set or persisting inhibition of

the previously irrelevant task set (e.g. Allport and Wylie, 2000; see also Goschke,

2000).

According to another account, full preparation in advance by endogenous means

only is possible, but it stresses the optional and probabilistic nature of preparation

(De Jong, 2000; De Jong et al. 1999). According to this account, residual switch

costs reflect intermittent failures to engage in advance preparation. The different

accounts do not rule each other out and could be complementary. Using

distributions of RTs instead of mean RTs, De Jong (2000) developed a modeling

technique to estimate the relative contributions of the two causes of residual

switch-costs. Using this technique he concluded that, at least for normal young

adults, residual switch-costs could be solely attributed to intermittent failures to

take advantage of opportunities for advance preparation, and found no evidence

for a fundamental preparatory limitation.

chapteR 3

59


Global, local and residual switch costs and aging

Kray and Lindenberger (2000) studied adult age differences in task-switching,

giving young and old adults two tasks to switch between, in an alternating runs

design (AABBAA . . .) with bivalent stimuli (every stimulus contained the features

of both tasks) and overlapping responses. Additionally, to assess generality, three

switch tasks were used, one with verbal, one with figural and one with numeric

stimuli. Mean number of years of formal education of the participants was 12.

No explicit cues were used, but subjects were instructed to keep track of the task

sequence and perform ‘task A’ for two trials and then switch and perform ‘task B’

for two trials etc. To be able to test not only whether switch costs were larger for

older adults but also whether switch costs were disproportionately larger, Kray

and Lindenberger (2000) used natural logarithms of RTs to analyze the data. This

method results in testing ratios between RT’s on different trial types (switch, non-

switch, fixed task) instead of the absolute difference. On none of the switch tasks

reliable age differences were found in local switch costs. Two preparation intervals

(RSI, 200 and 1200ms) were used and varied in a blockwise manner. At long RSI

trials switch-costs still existed, thus residual switch-costs were evident, but no

age differential effect of RSI on switch-costs was found. They did however find a

significant effect of age group on global switch-costs.

In an attempt to replicate these age-related effects on global switch-costs found

by Kray and Lindenberger (2000) and to further specify the boundaries of task

switching and aging, Mayr (2001) conducted two task switching experiments with

young and old subjects. On every trial a verbal cue was shown, informing the

subject about which task was to be performed. This cue was used for reducing the

working memory load that could have been critical in the Kray and Lindenberger

(2000) study, because subjects in that study needed to keep track themselves which

task was at hand. In the first experiment a cue-stimulus interval (CSI, which could

be used as a preparation interval) of 600 milliseconds was used and in the second

experiment CSI was 1200 ms. Old adults, in both experiments, exhibited larger

local switch costs than young adults. In the second experiment ambiguity of the

stimuli and response overlap was manipulated. Local switch costs were unaffected

by response-set overlap and only moderately affected by stimulus ambiguity. Mayr

(2001) also pointed out that age differences in local switch costs were modulated

by response repetition effects. Old adults exhibited a larger response repetition

benefit in the case of non-switch trials than young adults, and this effect was

reversed in the case of switch trials.

60


Salthouse et al.(1998) used data from 161 subjects who had a mean of 15 years

of education. They report that 37 subjects did not complete the tasks within a

given time or did not understand one or more of the tasks. They furthermore

noted that the mean age of the excluded subjects was higher than the mean age

of the subjects who were included in the analysis Salthouse et al.(1998) conducted

three switch tasks, all three consisting of two tasks between which subjects had to

switch. All stimuli were ambivalent and in two of the switch-tasks the pair of tasks

had overlapping response sets. Every tenth trial a rectangle appeared simultaneous

with and around the stimulus, indicating a task-switch had to be made. Old adults

exhibited larger switch costs than young adults. Salthouse et al. (1998) tested a

structural model with one common factor, mediating relations between age and all

variables used in the study. They found no significant direct relation between age

and RT on switch trials of the different tasks. In other words, the relations between

age and switch RT’s were mediated by the common factor.

Kramer, Hahn and Gopher (1999), studying local switch costs, used a paradigm

in which the same task was to be performed for 4 to 10 trials in a row, after which

subjects had to switch to another task. A verbal instruction (cue) was presented

for 1000 ms and a 200ms cue-stimulus interval (CSI) was used. The responses for

the two tasks were non-overlapping and the stimuli were bivalent. They found

considerably larger switch costs for old adults than for young adults, though this

age-related effect diminished after three sessions of 1½ hours of practice. In a

second experiment the occurrences of switch trials were predictable (every 5 trials)

and cued by a colored box presented 200, 400, 800 or 1600 ms (preparation

interval, RSI, varied in a blockwise manner) before stimulus onset. Age-related

effects were found on switch costs at all RSIs, but these effects disappeared after

practice except for the 200 ms RSI blocks, at which old adults kept demonstrating

larger switch costs than young adults throughout the sessions. In a final experiment

the task-switch was not cued on trial-level but subjects were to keep track of the

number of trials themselves and were told that they had to switch every fifth trial,

increasing working memory load. Results of this experiment showed somewhat the

same pattern on the first session as the previous experiment, but in contrast to it,

considerable smaller practice and RSI effects were found on switch costs for the old

adults. At all sessions and all PI’s switch-costs were significantly larger for old adults

than for young adults. Furthermore, to test whether or not the difference in switch

costs between age groups could be explained in terms of general slowing, they

calculated the amount of age-related variance in RT’s on switch trials after removing

age-related variance attributed to performance on the non-switch trials by means

chapteR 3

61


of hierarchical regression. The results suggested that the processes that support

non-switch trial performance and those that support switch trial performance are,

at least in part, differentially influence by aging. Kramer et al.(1999) reported the

mean years of formal education in experiment 2 to be 15.4 and 16.0 for young

and old adults respectively.

Meiran, Gotler and Perlman (2001) conducted three experiments. They reported

results of analysis on logarithms of RT’s (as did Kray and Lindenberger (2000))

when it was discrepant with the analysis on raw RT’s,. In the first experiment

they tested 16 young and 20 old adults. Four of the old adults performed below

chance level on trials with incongruent stimuli and were excluded from analysis.

The participants were members of a kibbutz, as an attempt to match the age

groups in background, education (a mean of 13 years) and occupation. Meiran

et al.(2001) used a task-switching design in which explicit, spatial cues preceded

every stimulus indicating which task was at hand. The three experiments differed

in whether response-sets overlapped or not, and different numbers of preparation

intervals (PI) were used. Age-related differences in local switch-costs were found

in all experiments, but, in experiment 3, when only one PI (of 117 ms) was used

these age-related differences were not disproportional (no significant effect when

analyzing log transformed RTs). Additionally, residual switch-costs were absent for

young adults but present for old adults, when using non-overlapping response-sets

(experiment 2).

Span (2002) used a task-switch paradigm with overlapping response-sets and

ambiguous stimuli. Tasks were cued, and three randomly selected CSI’s (66, 166

and 500ms) were used. A group of 42 adult subjects, aged from 19 to 84 years

old, participated. Span (2002) used linear and exponential models to describe the

age-related changes in RTs on the different trial types. They found that age-related

changes in response speed in homogeneous blocks (fixed task trials) were best

described using a simple linear function, while age-related changes in response

speed in the heterogeneous blocks (non-switch trials and switch trials) were

better described by an exponential function. These results can be interpreted as

demonstrating age-related differences in global switch-costs. They found no age-

related differences in local switch-costs.

Some of the described studies were aimed specifically at the issue of age-related

differences in global switch-costs. The data on this issue are fairly consistent,

showing responses on non-switch trials in heterogeneous block to be slower than

62


responses in homogeneous blocks, and an enlargement of this effect in old adults.

The study by Mayr (2001) found this age-related difference in global switch-costs

to persist when the heterogeneous and homogeneous blocks were matched in

number of stimulus-response rules, but only when the tasks converged on the

same responses.

Eenshuistra, Wagenmakers and De Jong (2000) forced old adults to stress speed

over accuracy which resulted in global switch costs to decrease to almost zero

(experiment 3). From these results they concluded that global switch costs (in

old adults) have a strategic basis and that these costs are due a conservative

response criterion and the use of different strategies in different (homogeneous

and heterogeneous) task situations.

In summary, age-related differences in global switch-costs are found in most

studies (see also Kray, 2005), but this finding seems to be restricted to paradigms

in which response sets of the different tasks overlap and some evidence suggests it

to be mediated by strategical aspects. Age-related differences in local switch costs

were found in most but not in all studies. It appears that, especially when taking

into account baseline performance and practice, these age-related effects became

modest or disappeared. A special case of local switch cost is residual switch cost.

Some of the studies reported whether or not residual switch costs were affected

by age (Kray and Lindenberger, 2000; Kramer et al., 1999, experiments 2 and 3;

Meiran et al., 2001, experiments 1 and 2). The results are mixed, ranging from

no age-related difference in residual switch costs and age-related differences

diminishing after practice to clear age-related differences.

It should be noted that in all of these studies, conclusions were derived solely

on the basis of mean RTs. As suggested by De Jong (2000), limiting analysis to

the level of mean RT, will often obscure important behavioral phenomena that

may be visible most clearly at the level of RT distributions. In the present study,

a detailed inspection of switch costs and, specifically, residual switch-costs, was

performed, as proposed by De Jong (2000). Generally in the studies discussed

above, old adults were, like young adults, capable of making effective use of

opportunities for advance preparation during prolonged PIs. Nevertheless, with

the exception of experiment 2 of Meiran et al. (2001), residual switch costs were

found. How these residual switch-costs relate to the initial switch-costs (at short

PIs) and the question about the role of aging therein is unclear. More specifically,

it remains unclear whether residual switch-costs in old adults can be attributed

chapteR 3

63


solely to intermittent failures to engage in advance preparation when ample time

is provided for that at long PIs, as has been found with young adults by De Jong

(2000) and by Nieuwenhuis and Monsell (in press), or to fundamental limitations

to prepare in advance by endogenous means only. For highly educated old adults

Eenshuistra et al. (2000) found evidence suggesting that residual switch costs in

elderly might not be attributable solely to failures to engage.

In the present study, performance of old adults was compared with performance

of young adults, and was aimed at clarifying possible differential contributions

to residual switch costs between young and old adults. To accomplish this, a

RT distribution modeling approach developed by De Jong (2000) was used. An

alternating runs paradigm (Rogers and Monsell, 1995) was developed in which

task-order and thus the occurrence of switch trials was predictable (AABBAA . .

.) and cued only implicitly (by way of clockwise positioning of the stimuli in the

four quadrants of a rectangle and using position as a cue). In this manner task-

switches were not explicitly cued. Furthermore, stimuli were ambiguous as to what

task was to be performed and responses of the two tasks were the same, thus

the dependency on self-initiated endogenous control processes remained high,

without the subjects needing to count the trials (thus minimizing working memory

load unrelated to task-switching). Three response-stimulus intervals were used as

PIs.

Furthermore, in order to match age groups, the younger group was not recruited

from the university population but was recruited in the same way as the older

group. In this way it was attempted to match at least level of education of the age

groups and not to restrict the sample to academics.

3.2 Methods

Subjects.

Twelve older subjects (6 men and 6 women) and 12 younger adults (6 men and

6 women) participated in this study. The older subjects ranged in age from 65 to

80 years (M=73.2, SD=5.0) and the younger subjects’ age ranged from 18 to 24

(M=20.5, SD=2.2). All subjects responded to an advert in the local newspaper

asking for participants in a study of aging and cognition. From the respondents

a quasi random sample was taken with the following constraints: the mean level

of education of the two groups was not to differ reliably, subjects needed to have

normal or corrected-to-normal vision. The mean level of education was 5 (SD=.9)

64


and 5.2 (SD=.8) for older and younger subjects respectively, using a scale of 1

(primary education) to 7 (academic education). A computerized form of the Raven

advanced progressive matrices was administered with odd numbered problems

presented on one day and even numbered problems on the other day, counter

balanced over subjects. Old adults solved on average 10.7 (SD=4.9) and young

adults solved on average 20.3 (SD=6.0) of the 36 problems in 40 minutes. The

difference between groups on Raven scores was significant (t(22)=4.9, p<.001).

Apparatus and tasks

Stimuli and instructions were displayed on a 15’’ VGA monitor controlled by an

IBM compatible computer using MEL2 (Schneider, 1988). Each block started with

the presentation of the stimulus-frame, which was a square (10 by 10 cm) divided

in four quadrants. The stimulus-frame remained on the screen throughout of the

block. On the sequential trials the position of the presented stimuli was varied

predictable in a clock wise manner, presenting the stimulus on the first trial in

the upper-left quadrant, on the second trial in the upper-right quadrant etc. The

stimulus was a ‘smiley’ and was manipulated in two dimensions. First, the face could

be round (diameter of 3 cm) or oval (3 cm vertical and 2.5 cm horizontal), and

second, the mouth of the face could be sad (corners of the mouth down) or happy

(corners of the mouth up). The experiment consisted of three conditions, two fixed

conditions and a mixed condition. In the mixed condition (heterogeneous blocks)

the relevant dimension was the form of the face if the stimulus was presented in

one of the upper quadrants of the stimulus frame, and form of the mouth was

relevant when the stimulus was presented in one of the lower quadrants. In the

fixed condition (homogeneous blocks) ‘face’, the form of the face was always the

relevant condition, and in the fixed condition ‘mouth’, the form of the mouth was

always the relevant dimension.

The subjects task was to respond with a keypress using their left index finger (V

key) or using their right index finger (N key) depending on the stimulus and the

relevant dimension. If the relevant dimension was the form of the face, the left key

had to be pressed if the face was round and the right key when the face was oval. If

the relevant dimension was form of the mouth, the left key had to be pressed when

the form of the mouth was ‘happy’ and the right key had to be pressed when the

form of the mouth was ‘sad’. A response had to be made within five seconds and

the stimulus remained on the screen for 5 seconds or until a response was made.

If an incorrect response or no response was given, a feedback-tone was presented

for 500 milliseconds. After a response had been made, a response stimulus interval

chapteR 3

65


of 100, 600 or 1500 milliseconds was randomly selected as the preparation interval

(PI) for the next trial.

Procedure and design.

Participants were sitting approximately 50 centimeters from the screen. The

experiment was administered with every subject on two days, both sessions

endured approximately two and a halve hours. The two days of testing were the

same, except for the order of the homogeneous blocks which was reversed the

second day and counterbalanced over participants. The trials were presented in

blocks of 40. Both sessions started with a training phase of 3 blocks of one of the

fixed condition, then 3 blocks of the other fixed condition and finally 3 blocks of

the mixed condition. The experimental phase consisted of 3 blocks of the mixed

condition, then 1 block of both fixed conditions, again 3 blocks of the mixed

condition, again 1 block of both fixed conditions, and finally again 3 blocks of the

mixed condition.

During the training phase, before the task started, the condition was explained

to the participants, and by simulating the task on paper by means of a booklet

(on every page a stimulus) it was tested whether or not the instructions had been

understood. If the participants performed accurately on the ‘paper’ test, the task

was started . In the training, as well as in the experimental phase, each block

started with on screen instruction about the condition at hand, and instruction

about speed and preparation. Participants were explicitly instructed to respond as

fast and as accurate as possible, and to use the time (if provided) between response

and stimulus presentation two prepare for the next task.

General linear model (GLM) was used to perform analysis of variance on the

repeated measures, using preparation interval (PI) and trial type (fixed, non-switch

or switch) as within subject factors, and age group as between subjects factor. To

be able to distinguish local and global switch costs, a priori contrasts were chosen

(for fixed, non-switch and switch trials the local and global switch cost contrasts

were respectively 0, 1, 1 and 1, 1, 0). The reported GLM analyses were performed

on the raw RTs, but for proportionality reasons, additionally, as done by Meiran

et al. (2001), results of GLM analysis on log transformed RTs are reported when

discrepancies in statistical significance were present. For detailed analysis of residual

switch costs, the “generalized mixture model” introduced by De Jong (2000) was

used, in order to investigate whether residual switch-costs are attributable to

switch trial specific post stimulus processing or to a failure to consistently engage

66


in advance preparation during PI. De Jong (2000) describes a two state mixture

model for long PI switch trials, and uses two other trial types to provide estimates

for the basis states that contribute to the mixture. The distribution of RT’s on non-

switch trials was used here to provide an estimate of the prepared state (fast basis

distribution), and the distribution of RTs on switch trials with a short PI (100 ms)

was used as an estimate of the unprepared state (slow basis distribution). The

following relation between the cumulative density function (CDF) for RT in the

mixture condition and the CDFs for RT in the two basis states is posited by the

model:

Fmixture(t) = α Fnon-switch trials (t-δ) + (1- α) Fswitch, short PI trials (t),

in which α reflects the probability of advance preparation during PI, which will

approximate the maximum value one if the mixture distribution overlaps the

fast basis distributions. If, on the other extreme, subjects engage in advance

preparation on none of the trials the mixture distribution will resemble the slow

(unprepared state) distribution of short PI switch trials, and will result in the

minimum value of α (zero). If the fast part of the mixture distribution resembles the

fast basis distribution and at the slow part crosses over to resemble the slow basis

distribution, this would result in an α of 0.5. The second parameter (δ) is employed

to account for the possibility that even when provided ample time to prepare in

advance a subject can consistently not reach the same response speed on switch

trials with long PI as on non-switch trials. In the case of a subject who consistently

engages in advance preparation (high α), this will result in the mixture distribution

resembling the fast basis distribution, but shifted in the direction of longer RTs

(consistently throughout the distribution function). This shift is represented by δ

(in milliseconds). For each subject, the three RT distributions were partitioned in

five bins. A multinomial maximum-likelihood method (MMLM) of Yantis, Meyer

and Smith (1991) is used to determine the maximum likelihood estimates of α and

δ and to compute goodness-of-fit statistics (G2) for the model (see De Jong (2000)

for a more detailed description).

3.3 Results

Analyses of variance

General linear model (GLM) was used to perform analysis of variance on the

repeated measures, using PI and trial type (fixed, non-switch or switch) as within

subject factors, and age group as between subjects factor and accuracy and RT’s

chapteR 3

67


as dependent variables. To be able to distinguish local and global switch costs,

a priori contrasts were chosen (for fixed, non-switch and switch trials the local

and global switch cost contrasts were respectively 0, 1, 1 and 1, 1, 0). The

reported GLM analyses were performed on the raw RTs, but for proportionality

reasons, additionally, as done by Meiran et al. (2001), results of GLM analysis on

log transformed RTs are reported when discrepancies in statistical significance

were present. For the analysis of reaction times (RT), correct RTs, larger than 200

milliseconds, after at least two successive correct trials were used.

Accuracy.

With respect to overall accuracy no reliable difference was found between age-

groups (F(1,22)=1.4, p=.24). Significant local switch-costs, but no global switch-

costs were found in accuracy (resp. F(1,22)=7.3, p<.05, F(1,22)=2.7, p=.112). See

table 3.1 for accuracy of older and younger adults on the different trialtypes.

Table 3.1: Means and standard deviations of accuracy of young and old adults

RSI Young Old

Fixed task trials 100 0.97 (.03) 0.98 (.02)

600 0.96 (.04) 0.97 (.03)

1500 0.95 (.04) 0.98 (.03)

Non-switch trials 100 0.93 (.03) 0.94 (.03)

600 0.93 (.03) 0.94 (.03)

1500 0.92 (.05) 0.95 (.03)

Switch trials 100 0.96 (.02) 0.96 (.05)

600 0.96 (.03) 0.96 (.05)

1500 0.95 (.04) 0.96 (.04)

Global switch costs Pooled 0.004 (.02) 0.02 (.04)

Local switch costs Pooled 0.03 (.03) 0.02 (.05)

68


Reaction times.

Mean individual global and local switch-costs were computed for the separate days

of testing to assess test-retest reliability. Cronbach’s α for global switch costs was

0.83 (r = 0.71) and was 0.79 (r = 0.67) for local switch-costs.

Figure 3.1 shows the relevant data for the analysis of mean RTs. All main effects

were found to be significant, the within subject effects trial type (F(1,21)=128.1,

p<.001) and PI (F(1,21)=39.9, p<.001) as well as the between subject effect, age-

group (F(1,22)=39.2, p<.001).Trials with long PIs were responded to faster than

short PI trials and old subjects were slower than young subjects. Global switch-costs

(contrasting trial type non-switch against fixed) and local switch-costs (contrast

between switch and non-switch trials) were found to be significant (respectively:

F(1,22)=57.0, p<.001 and F(1,22)=197.3, p<.001).

1200

1100

1000

900

800

700

600

500

old adults

RT (

ms)

PI (ms) 100 600 1500

young adults

PI (ms) 100 600 1500

fixed

switch

non-switch

Figure 3.1: Mean correct reaction times as a function of preparation interval (PI) and trial type.

Global switch-costs were smaller for young than for old adults (F(1,22)=16.1,

p<.01). Local switch-costs were smaller for young adults than for old adults

(F(1,22)=7.0, p<.05), but this interaction group effect could be interpreted as being

proportional, because in the analysis on the log transformed RT’s this effect seized

to be significant (F(1,22)<1). Local switch-costs were reduced with increasing PI

(F(1,22)=147.8, p<.001). This reduction was larger for old than for young adults,

but again not disproportionately so (raw RT: F(1,22)=6.3, p<.05, logRT: F(1,22)<1).

The PI pattern was confirmed by analysis of separate PIs, which showed that initial

chapteR 3

69


local switch-costs (at the shortest PI) were larger for old adults than for young

adults (F(1,22)=8.3, p<.01, though for logRTs: F(1,22)<1). Mean residual switch-

costs (at long PI) were not significantly affected by age group (F(1,22)=2.7, p=.12).

RT distributions.

In order to find reliable mixture estimates, in the distribution modeling the data

of the two days of testing were pooled. The obtained fit of the mixture model

was good for the old group (G2(24)=18.9, p<.76) and for the young group (G2

(24)=14.5, p<.94). Across groups the mean estimate of α was .61 (see table 3.2

for group means) and no age-related difference was found on α (F(1,23)<1, see

figure 3.2 for the pooled CDFs). In Figure 3.2 the RT distribution of switch-trials at

long PIs can be seen to be close to the fast (non-switch) distribution at the fast RT

tail, while crossing over to the slow distribution (switch at short PIs) at the slow RT

tail. The crossover reflects α. Furthermore apart from the crossover of the mixture

distribution another aspect is shown in figure 3.2.

0.90

0.80

0.70

0.60

0.50

0.40

0.30

0.20

0.10

old adults

Cum

ulat

ive

prob

abili

ty

young adults

500 1000 1500 2000 500 1000 1500 2000

No-switchSwitch, short PISwitch, long PIModel fit

RT (ms) RT (ms)

A B

Figure 3.2: Cumulative distribution functions for the groups of old and young adults as a function

of preparation interval (PI) and trialt type.

The mixture distribution of the young adults at the fast RT tail is almost overlapping

the fast RT basis distribution, while at the slow tail it crosses over to the slow basis

distribution. In the case of the group with old adults, the mixture distribution is

shifted from the fast distribution in the direction of slow RTs. This aspect reflects the

70


models’ δ. The age groups did significantly differ with respect to δ (F(1,23)=8.3,

p<.01). For the older group the mean δ (78 ms) estimated by the model significantly

differed from zero (t(11)=6.4, p<.01), and the mean δ (22 ms) for younger group

did not significantly differ from zero (t(11)=1.6, p<.14).

Table 3.2: Mixture model fit parameters and goodness of fit for old and young adults.

Young and old adults in this study behaved similarly with respect to the use of the

preparation interval at switch trials. In both groups failures to engage in advance

preparation were equally likely contributing to residual switch-costs. For the young

group, residual switch-costs could be explained solely by failures to engage. In

contrast, for the mixture distribution of the older adults to be modeled best, the

model needed a positive δ. These results suggest that when subjects did engage

in advance preparation on a switch trial, younger adults were able to attain a fully

prepared state (as on a non-switch trial) while older adults were not. Although the

CDF’s (see figure 3.2) and the interpretation of the distribution modeling were

straightforward when pooled over groups, it should be stressed that there were

notable individual differences. To illustrate this the CDF’s of 3 older participants are

shown in figure 3.3. Figure 3.3a shows a case of an older subject who engaged in

advance preparation on very few switch trials with long PI. The RT distribution on

long PI switch trials largely overlaps the RT distribution of the slow basis distribution

(switch, short PI) resulting in a very low α (.15). The mixture distribution could be

described almost solely by only one of the basis distributions, thus leaving little

room for an interpretation in terms of a mixture. Related to this, the estimate of δ

(119 ms) should be interpreted cautiously, though the CDFs in figure 3.3a clearly

show that this subject was on switch trials with long PI not able to reach the same

level of performance as on non-switch trials. Figure 3.3b shows the distributions of

another old participant. Clearly, this participant engaged in advance preparation

often, but could not attain a fully prepared task-set when ample time was provided

to do so. In other words the mixture distribution (switch, long PI) could be

described by the fast basis distribution (non-switch) in the fast RT tail, but shifted

α δ

mean (sd) range mean (sd) range

Old adults 0.57 (0.22) 0.14 - 0.85 78 (38) 20 - 160

Young adults 0.64 (0.18) 0.33 - 0.95 22 (49) -46 - 107

chapteR 3

71


in the direction of slow basis distribution (switch, short PI) in the long RT tail,

resulting in a high α (0.75) and a sizable δ (87 milliseconds). The model fitted

the data exceptionally well with these parameters (G2(2)=.02). Finally, figure 3.3c

shows the data of a participant who performed in a similar manner as most young

adults. At the fast part, the mixture distribution could be described best in terms of

the fast basis distribution, while the slow part of the mixture distribution could be

described best by the slow basis distribution, resulting in a α of 0.60 and a δ of -1.

A B

C

Figure 3.3: Cumulative distribution functions as a function of preparation interval (PI) and trial

type of three individual subjects from the old adults group.

0.90

0.80

0.70

0.60

0.50

0.40

0.30

0.20

0.10

Cum

ulat

ive

prob

abili

ty

500 1000 1500 2000

RT (ms)


0.90

0.80

0.70

0.60

0.50

0.40

0.30

0.20

0.10

Cum

ulat

ive

prob

abili

ty

500 1000 1500 2000

RT (ms)


0.90

0.80

0.70

0.60

0.50

0.40

0.30

0.20

0.10

Cum

ulat

ive

prob

abili

ty

500 1000 1500 2000

RT (ms)


72


3.4 DiscussionThe results of the present study were consistent with the general findings of previous

studies on task-switching and aging. Global switch-costs were disproportionally

larger for old adults than for young adults. This result is consistent with several

studies including the study by Mayr (2001), who found age-related differences

in global costs to be restricted to task situations of complete response-set overlap

and ambiguous stimuli, which was the case in the present study. Local switch-

costs were larger for old adults, but not disproportionally so. This is also generally

consistent with the existing literature.

The present study was specifically aimed at whether or not residual switch-costs in

old and young adults were attributable to the same factors. At the level of mean

RTs no significant age-related differences were found in residual switch-costs. In the

present study, residual switch-costs were examined in a more detailed way, using

analysis of RT distributions. The distributional analysis revealed residual switch-costs

in younger adults to be attributable completely to failures in engaging in advance

preparation, replicating results reported by De Jong (2001) and by Nieuwenhuis

and Monsell (2002). For the group with old adults, though, this failure to engage

(FTE) account of residual switch-costs did not suffice. For the distributional model

to accurately fit the data of the old group a mean δ of about 79 milliseconds had

to be included in the model. Evidently engaging in advance preparation did, for

old adults, not suffice to attain a similar level of preparation as on non-switch trials.

Thus, extra time after the onset of the imperative stimulus was needed, suggesting

that old adults in this study were to rely more on exogenous control with respect

to task switching, or that old adults, when switching tasks, did not attain a strong

enough representation of the task-set to hold under the high interference situation

that comes with the (ambiguous) stimulus.

At short PI’s it is usually assumed that at the moment the stimulus is presented,

preparation for the assigned task is not completed, resulting in hampered

performance and evidently time-consuming task-set configuration related

processing such as selection, retrieval, engaging and/or disengaging needs to be

done in order to perform the task fast and accurately. If stimuli are ambiguous

and there is no explicit informative cue, these processes need to be initiated

endogenously (see also Kray, 2005).

Mayr and Liebscher (2001) note that minor or absent age-related effects in

local switch-costs would be consistent with the “LTM retrieval” interpretation of

local switch costs proposed by Mayr and Kliegl (2000). Mayr and Kliegl (2000)

chapteR 3

73


hypothesized that local switch-costs reflect a process of actively retrieving task-set

related information from LTM. They tested this hypothesis by constructing a switch

task with two semantic tasks (low demands on LTM retrieval) and two episodic

tasks (high demands on LTM retrieval). They found increased switch costs when

participants had to switch to one of the episodic tasks and inferred that a major

component of accurate switching of tasks is retrieval of information from LTM

(between task retrieval) and this interferes with retrieval processes required for the

primary task (within task retrieval). Mayr and Liebscher (2001) note that since speed

of memory retrieval has been found not to be strongly affected by age (Wingfield,

Lindfield and Kahana, 1998), minor or no disproportional age-related effects on

local switch costs would not be completely surprising. The work of Mayr & Kliegl

(2000) suggests that accurate task-switching depends on retrieving information

from LTM, and they suggest that residual switch costs could be explained in terms

of retrieval failures resulting from the high interference situation. This explanation

is in line with the results of older adults reported in the present study. Although

speed of LTM retrieval may not be strongly affected by aging, the ability to

endogenously attain a strong enough retrieval to maintain this retrieval in face

of the high interference situation the (ambiguous) stimulus poses, may well be.

Related to this is research studying prospective memory, which typically measures

whether an intention can be maintained in the face of high interference situations.

Several studies have dealt with aging and prospective memory, and old adults are

often found in these studies to show worse performance than young adults under

conditions of relatively low environmental support and relatively high interference

(Einstein, Smith, McDaniel and Shaw, 1997; Maylor, 1996; Vogels, Dekker and De

Jong, 2002). From the results of the distributional analyses it follows that, although

old adults did not have more problems with engaging in advance preparation, the

product of the old adults’ advance preparation during the PI at a task-switch was

not sufficient for reaching the same prepared state as in the non-switch situation.

The time it takes to respond (even when engaged in advance preparation during PI)

for old adults in the present study could reflect a decrement of quality of retrieval

of the task set information from LTM.

The present study adds to the knowledge of aging and switch costs, in the sense

that it specifically addressed the differences in manifestation of residual switch

costs in young and old adults. Further study is needed to examine whether the

limitation of old adults to prepare in advance by endogenous means only is really

fundamental or is affected by strategic or other factors like training or stimulus-

ambiguity.

74

chapteR 3

chapteR 3

75

Patterns of aging and cognitive control

Chapter 4Patterns of aging and cognitive control

76

4.1 IntroductionThis study focuses on how several cognitive control functions are influenced by

aging, and how these functions relate to each other with respect to age related

influences. Most of the current theories in the field of cognitive psychology and

neuropsychology concerning age-related changes in cognitive functions can be

separated into two groups.

One group of theories posits age-related changes in different cognitive functions to

be mediated by a general aspect of cognition (general factor theories); one of these

theories is the global speed hypothesis (e.g. Salthouse, 1991, 1996). The global

speed hypothesis states that age-related cognitive decline can be attributed to a

decrease in the speed with which elementary cognitive operations are carried out.

The decrease in information processing speed places limits on the performance

that can be reached on most cognitive tasks (Birren, 1956; Cerella, 1985; Eearles,

Connor, Smith & Park, 1997; Salthouse, 1996).

The other group of theories focuses on age-related changes in specific cognitive

functions. Typically, in this group of theories a distinction is made between

behavior that is dependent on executive control processes and supposedly affected

by advancing age, such as switching between tasks and divided attention and

selective attention (McDowd & Shaw, 2000; Braver, Barch, Keys, Carter, Cohen,

Kaye, Janowsky, Taylor, Yesavage, Mumenthaler, Jagust & Reed, 2001; de Jong,

2001; Kramer, Larish & Strayer, 1995; Mayr & Kliegl, 1993), and behavior that is

less or not dependent on control processes and is relatively free from age-related

effects. This specific loss hypothesis has found support from aging research in the

field of neuroscience (West, 1996) in the sense that the frontal lobes, which are

particularly vulnerable to effects of aging (e.g. Craik & Grady, 2002), are strongly

involved in executive control functions of cognition (Fuster, 1997; Kramer,

Humphrey, Larish & Logan & Strayer, 1994; Duncan, Burgess & Emslie, 1995).

Theories and their methodologies

Research initiated from the different viewpoints or theories usually employs different

analytical approaches. These differences in methodology can lead to contradictory

results (see e.g. Bashore and Smulders 1994).

An influential methodology used to support the global slowing hypothesis is to

compare older and younger adults’ performance on executive control tasks as well

on tasks that measure basic speed. Using regression analysis, it is tested whether

chapteR 4

77


any age-associated variance in executive control tasks remains after individual

differences in speed on the other tasks have been taken into consideration. For

example, studies by Salthouse (1985) and Salthouse and Babcock (1991) showed

that age-related variance in a number of different cognitive tasks is significantly

reduced or eliminated after individual differences in speed have been partialled

out.

In research that sprouts from specific-process theories, usually analysis of variance

is utilized as the statistical method, and often the measure of the specific process

of interest is experimentally or statistically controlled for general processes that

are not of interest, in such a way that age-related deficits with respect to the

specific cognitive function is evidenced by an interaction of age by condition. For

example, in research on task switching, reaction times (RT) in conditions during

which participants need to switch between performing several tasks are compared

with RT’s in conditions during which participants need to perform the same task

on every trial. The difference in RT’s between conditions are termed global switch

costs or mixing costs. In most studies on aging and task switching, older adults

exhibit larger global switch costs than younger adults (Kray and Lindenberger,

2000; Meiran et al. 2001)

Present study

In the present study a task battery with executive control tasks was administered

to four age-groups. Each of the tasks consists of different conditions which are

postulated to differ in the demand on executive control. Interactions between

age-group and condition on the performance indices of these tasks indicate age-

effects in the manipulated executive control function. Also, a task that has often

been used as an indicator for basic speed in studies supporting the global speed

hypothesis was included in the task battery.

The relationships between the performance indices on the different tasks and the

pattern of age-effects on these relationships was examined by using structural

equation modeling.

Cognitive control tasks

In the present study a task battery is used with tasks from which different aspects of

cognitive control functioning can be derived (suppression of prepotent responses,

task-switching, working memory, divided attention, sustained attention, planning)

and tasks of which the performance indices have been shown to be related to or to

78

chapteR 4 Patterns of aging and cognitive control

mediate age-related differences in performance on cognitive control tasks (symbol

digit substitution test, Raven progressive matrices).

Suppression of prepotent responses

Several studies have found that older adults show more failures to inhibit prepotent

response tendencies than younger adults (e.g. De Jong, 2001; Hasher, Zacks & May,

1999; Kramer et al., 1994; Nieuwenhuis et al. 2004; West, 1996). In general, active

inhibition of responses is thought to be mediated by prefrontal structures (Roberts,

Hager & Heron, 1994). Tasks used to investigate failures of the suppression of

response tendencies include the Stroop task and the Wisconsin Card Sorting Test

(WCST). In the Stroop task subjects need to suppress the naturally dominant task

of word reading. In the WCST subjects are to suppress a sorting rule, which has

become prepotent by means of experimental practice. A common and important

correspondence of these and other inhibition of prepotent response tasks, is that

as a prerequisite of accurate performance participants need to 1.) select and

execute the accurate response, while 2.) endogenously and actively suppressing

the inaccurate response tendency.

With the paradigm used in the present study an attempt is made to distinguish

between these two aspects of task performance. The ‘pro- anti-and central-cue’

(PAC) paradigm, introduced here, is adapted from cue-tasks used by Nieuwenhuis,

Broerse, Nielen & De Jong (2004) and De Jong (2001). These cue-tasks are related

to pro- vs. anti-saccade tasks. Performance on pro- vs. anti-saccade tasks have also

been studied with respect to age-related effects. Studies using pro- vs antisaccade

taks typically revealed comparable performance by younger and older adults in

terms of errors and time needed to initiate correct antisaccades compared to

prosaccades (Munoz, Broughton, Goldring & Armstrong, 1998; Fischer et al.

1997). Using a cue task, though, Nieuwenhuis et al. (2000) did find performance

of high functioning old adults on the anti-cue condition to be worse than young

adults even at SOA’s of 1.5 seconds (see Nieuwenhuis et al. 2004 for a discussion

on the discrepancy of the findings of age effects between the pro- vs. antisaccade

and pro- vs. anticue paradigms)

In the cue-tasks, from which the current PAC-task is adapted, a visuo-spatial cue

is presented peripherally. Subsequently, after an unpredictable duration of the

stimulus onset asynchrony (SOA), the target stimulus is presented. Depending

on the blocked (pro-cue or anti-cue) condition the stimulus is presented on the

same location of the cue or on the location on the opposite side of the cue. The

chapteR 4

79


cues used are known to capture overt and covert attention in an automatic way

(Theeuwes, Atchley & Kramer, 2000; Yantis, 1998), in other words, induce a

prepotent response tendency. The stimulus incorporates a dichotom feature that

needs to be identified and responded to by the participant.

In the pro-cue condition the prepotent response is the appropriate response

and the performer can use the cue to direct attention to the correct location. In

this respect, the anti-cue condition is different in two ways. Firstly, in the anti-

cue condition the performer needs to suppress the prepotent response tendency.

Secondly, in the anti-cue condition attention, needs to be endogenously directed

to the opposite location of the cue. Thus, age differences in anticue performance

could be due to either difficulty with suppressing prepotent response tendency,

difficulty with endogenously initiating the direction attention or a combination of

these two. To be able to distinguish between these two influences on performance

a third, central-cue condition was introduced. During this condition there is no

need for suppressing a prepotent response tendency, but there is the need to

endogenously direct attention to the correct location.

Task-switching

Research on cognitive control processes has focused on behavior in dual-task

situations or interference situations. In the last decade much research has been

conducted to study the dynamics of cognitive control processing under conditions

of switching between cognitive tasks (Allport, Styles & Hsieh, 1994; Gopher, 1996;

Rogers & Monsell, 1995). The main focus of task-switching studies are switch costs,

which are differences in reaction time (RT), but also in accuracy performance,

between situations in which the same task is repeated and situations in which

switches between tasks have to be made.

Switch costs may reflect the time and effort needed for task-set reconfiguration

(Rogers & Monsell, 1995; Rubinstein, Meyer & Evans, 2001, De Jong 2000) or

the time and effort needed to overcome “task-set inertia”, interference due to

positive priming of the now-inappropriate task-set, and negative priming due to

the persistence of inhibition applied to the now-appropriate task-set (Allport et al.

1994; Meuter & Allport, 1999; Mayr & Keele, 2000).

An important distinction needs to be made in research on task switching with

respect to the actual measurement of switch costs. Most of the studies on task

switching (Allport, Styles & Hsieh, 1994; Botwinick, Brinley & Robbin, 1958;

80


Gopher, 1996; Kluwe, 1997) compared performance in task homogeneous blocks

(same task on all trials) with performance in heterogeneous block (alternating tasks

on consecutive trials). Consequently switch costs were determined by performance

differences between task heterogeneous and task homogeneous blocks (global

switch costs). Other researchers (initiated by Rogers & Monsell, 1995) evaluated

switch costs by comparing performance on trials in which the task to be performed

is different from the task on the previous trial with performance on trials in which

the task is a repetition of the previous trial (local switch costs). Paradigms in which

both types of switch costs can be evaluated have been utilized in several studies

now (e.g. Kray & Lindenberger, 2000, Kray, Li and Lindenberger, 2002; De Jong,

2001; for a review see Monsell, 2003).

Switching between different tasks and the cognitive control processes that underlie

task-switching is of interest also in the context of the present study. Most studies

concerned with effects of aging on task-switching found age-related deterioration

in performance in task-switching situations (De Jong, 2001; Hartley, Kieley &

Slabach, 1990; Kramer, Hahn & Gopher, 1999; Kray & Lindenberger, 2000; Mayr,

2001; Salthouse, Fristoe, McGuthry & Hambrick, 1998). Age-related differences in

global switch-costs are found in most studies, but this finding seems to be restricted

to certain conditions (response sets of the different tasks overlap) and it has been

postulated to be mediated by strategical aspects. Eenshuistra, Wagenmakers and

De Jong (2000), for example, forced old adults to stress speed over accuracy which

resulted in global switch costs to decrease to almost zero (experiment 3). From

these results they concluded that global switch costs (in old adults) have a strategic

basis and that these costs are due a conservative response criterion and the use of

different strategies in different (homogeneous and heterogeneous) task situations.

Age-related differences in local switch costs are found in most but not in all studies.

It appears that when taking into account baseline performance and practice, the

age-related effects become modest or diminished. A special case of local switch

cost is residual switch cost, which can only be reliably measured if at least one

very short PI and one long PI are included in the paradigm. Some studies reported

whether or not residual switch costs were affected by age. Mostly these studies find

that both younger and older adults are able to reduce switch costs with increasing

preparation time (Cepeda et al., 2001; Kramer et al., 1999; Meiran et al., 2001;

De Jong, 2001; Kray and Lindenberger, 2000, Kray, 2005). Kramer et.al (1999),

reported that older adults were not able to reduce switch costs under conditions

which place high demands on working memory.

chapteR 4

81


Several task-switching paradigms can be distinguished (for review, see Monsell,

2003). Paradigms can be differentiated on the manner in which participants

are informed about which task to perform on a particular trial (see Kray, 2005).

Sometimes, the task at hand is explicitly cued (Cepeda et al. 2001; Kramer et

al. 1999; Mayer and Liebscher, 2001; Meiran, 2001), while other paradigms the

task at hand is cued implicitly (for example by predicable task order) or memory

based (De Jong, 2001; Goschke, 2000; Kray and Lindenberger, 2000; Myiake et al.,

2004). The task switching paradigm used in the present study is the alternating-

runs paradigm developed by Rogers & Monsell (1995). The paradigm incorporates

homogeneous blocks and heterogeneous blocks, in which participants need to

alternate between task A and task B in predictable sequences (…AABBAABB…),

using the same stimuli for the two tasks. The task that is to be performed is cued

implicitly by the position of the stimulus. Using this paradigm, both general and

specific switch costs can be quantified.

Divided attention

Several studies have shown age-related difficulties with dividing attention across

two or more tasks (Brouwer, Waterink, Van Wolffelaar & Rothengatter, 1991;

Korteling, 1993; McDowd, Vercruyssen & Birren, 1991; Ponds, Brouwer &

Wolffelaar, 1988; Salthouse, Rogan & Prill, 1984). A review and meta-analysis is

presented by Verhaeghen, Steitz, Sliwinsky & Cerella (2003) and showed that costs

of dual-task processing on latency are larger in old adults than in young adults and

larger than predicted from general slowing. In accuracy measures costs of dual-task

processing were also found but no age specific deficit was associated with these

costs.

Dual task costs in the present study will be examined using a paradigm adapted

from Ponds, Brouwer & Van Wolffelaar (1988), whoe studied divided attention

in old adults and closed head injury patients (see e.g. Withaar, 2000; Riese,

Hoedemaeker, Brouwer & Mulder, 1999). The two tasks are a lane-tracking task

and a dot-counting task. Basic individual abilities on the tracking task are controlled

by means of adapting the level of difficulty to the extent that all participants reach

the same level of single-task performance. The performance measure used for the

dot-counting task is a composite of speed and accuracy of responses, in this way

controlling for age-related differences in speed-accuracy trade-offs.

82


The trailmaking test (Reitan, 1958) is included in the task battery, because of its

wide used and presumed sensitivity to executive function (Kuhlman, Little and

Sekuler, 2006). It is used as a measure of attentional functions in many individual

test batteries and in clinical practice. The trailmaking test consists of two versions (A

and B) and is a paper and pencil test. The A version requires participants to draw a

single, continuous line through randomly located items on a sheet of paper, which

are numbered form 1 to 25, in increasing numerical order. In version B, items are

intermixed labeled by number and letters and the line is to be drawn alternating

between the two kind of items (1-A-2-B-…-L-13). Difference in completion times

between version-B and version-A has been linked to executive functions such as

planning of actions and switching between tasks. Advancing age is usually related

to slower performance on the B-version in comparison with the A-version. In

several studies it was found that, when controlling performance on the B-version

for performance on the A-version, performance on the B-version was no longer

related to age (e.g. Salthouse & Fristoe, 1995; Salthouse et al. 2000). Other studies

did find a relation between age and performance measures on the B-version after

controlling for performance on the A-version (Keys and White, 2000; Kuhlman,

Little and Sekuler, 2006; Salthouse et.al 1996).

Sustained attention

Some studies have focused on sustained attention and aging, and the results are

not entirely consistent. Berardi, Parasuraman & Haxby (2001) did not find age-

related differences in the ability to sustain attention over time, using a high event

rate digit-discrimination task. On the contrary, using an auditory match to sample

task, Chao & Knight (1997) did find impaired sustained attention with age, in

addition to increased distractibility. Also, Filley & Cullum (1994) found performance

decrements in a group of old adults (70-90 years) on another sustained attention

test (numerical attention test) as compared with a younger group of old adults (50-

69 years). To measure the ability to sustain attention over time in the present study

the sustained attention to response test (SART) was used, which was developed by

Robertson, Manly, Andrade, Baddeley & Yiend (1997). The sustained attention test

has been shown be predictive of every day attentional failures and action slips (see

Manly, Robertson, Galloway & Hawkins, 1999).

Control in working memory

Many experimental and neuropsychological tasks have been developed and used

to manipulate the involvement of working memory. Essential in these tasks is that

information should be kept on-line and should be available for operations necessary

chapteR 4

83


for accurate performance. Several studies found that the efficiency and capacity of

working memory is affected by age (see e.g. Raz, Gunning-Dixon, Head, Dupuis &

Acker , 1998; Salthouse, 1994).

A theoretical account of cognitive control and cognitive aging that is related to

working memory has been put forward by Braver and colleagues (Braver, Barch,

Keys, Cohen, Taylor, Carter, Kaye, Janowsky, Yesavage, Mumenthaler, Jagust and

Reed, 2001; Braver and Barch, 2002; Barch, Braver, Racine, Satpute, 2001) . They

modeled a link between the decline with aging in the function of dopamine system

(DA) projection to the prefrontal cortex (PFC) and cognitive changes (with respect

to attention, active memory and inhibition) observed in aging. Braver and Barch

(2002) postulate that this neural mechanism is integral to the representation,

maintenance and updating of context information such as goal representations.

Context is in their view a component of working memory, subserving both

mnemonic and control functions simultaneously (Braver et al., 2001). Indeed,

Braver and Barch (2002), found that older adults show deficits in cognitive control

due to an impaired context processing mechanism.

An often used working memory paradigm is the n-back task, in which response

decisions are to be based on the stimulus presented n trials ago (e.g. Cohen,

Perlstein, Braver & Nystrom, 1997; Jonides, Schumacher, Smith, Lauber, Awh,

Misnoshima & Koeppe, 1997). Keys et al. (2002) examined sensitivity to aging of

several working memory tasks. They found that the 2-back condition of the n-back

task was most sensitive to aging. Most working memory tasks are versions of a span

task. In the present study, in addition to an n-back task, two versions of a working

memory span task are used, the forward-backward span test and an adaptation

of modified digit span task (Daily, Lovett & Reder, 2001). The modified digit span

paradigm is constructed in such a way that the probability that the subjects can use

the most efficient strategy to use in such span tasks, sub-vocal rehearsal, is being

kept to a minimum.

Planning

Several executive control functions (such as keeping representations of information

and goals active in working memory, suppressing prepotent responses, task

switching) are important functions underlying planning. In every day live,

constantly, activities are planned by structuring and composing series of steps

to follow. The ‘Zoomap’ test from the was administered to assess executive

control aspects of planning. The Zoo map test is a sub-test from the task battery

84


‘Behavioural Assessment of Dysexecutive Syndrome’ (BADS) (Wilson, Alderman,

Burgess, Emslie and Evans, 1996). The test consists of two versions. In one of the

versions, the planning demands for the participant are minimized by means of

explicitly instructing not only the end goal of the test but also the manner in which

to reach the end goal. In the other version only the end goal is instructed and the

participant needs to plan the way to reach the end goal.

Speed

Studies on cognitive aging that are aimed at examining the mediation of speed on

the relation between cognitive functions and aging, mostly used the performance

measure of the digit-symbol substitution test (DSST) as measure for speed. The

use of DSST as a measure of basic speed is not without controversy though. For

a critical discussion see the discussion section of this chapter and Parkin and Java

(1999,2000). This test was included in the task battery of the current study also.

Fluid intelligence

The short version of the Raven standard progressive matrices was used as an

indicator of fluid intelligence (gf). In the context of exploring the construct validity

of ‘frontal tests’ Rabbitt, Lowe & Shilling (2001) discuss and compare the results

of two studies using ‘frontal tests’ measuring executive function. One study was

performed with ‘normal’ young and old adults (Lowe, Rabbitt & Shilling, 2000).

The other study was done on a group of patients of mixed aetiology and a control

group (Wilson, Alderman, Burgess, Emslie & Evans, 1996). Rabbitt et al. (2001)

noted that measures from different frontal tests correlate only modestly and often

not significantly within the control group of the Wilson et al. study. Moreover when

subsequently variance associated with individual differences in intelligence was

partialled out, most of the significant associations between the tests disappeared. In

the patient group, though, many correlations were markedly higher and remained

significant after variance associated with individual differences in intelligence was

partialled out. Rabbitt et al. (2001) noted that performance on frontal tasks as well

as performance on other (non-frontal) tasks by participants who do not suffer from

focal deficits or damage, may be well-predicted by global task performance indices

(such as intelligence scores), because of the common dependence on gf of these

measure of performance. In contrast, in groups of patients with frontal lesions,

performance on frontal tasks may be more strongly determined by the extent and

site of brain damage rather than by gf.

If this line of reasoning is accurate, together with findings that the proportion of

individuals with frontal decrements in a population increases with age, one would

chapteR 4

85


expect less dependence of performance on frontal tests on gf in older groups than

in younger groups of participants. This, indeed, is what Lowe et al.(2000) found

in their longitudinal study with two groups of older adults. The results from that

study were not replicated, though, within another sample. Also Duncan (1995)

and Duncan et al. (1995) found that within groups of patients with frontal lesions

individual differences in performance on frontal tests can be largely accounted for

by individual differences in scores on fluid intelligence test.

To be able to scrutinize the relation between a measure of gf and executive control

measures and the age-effects on these relation, in this study the Raven Standard

Progressive Matrices (SPM) was administered.

4.2 Method

Participants

Eighty subjects from four different age groups participated. The participants

were recruited by advertisements in local newspapers and from referrals by

other participants, and were paid for completing two sessions of approximately

3 hours; each on different days during which also tasks were administered that

were used for another study in which the participants were also screened for

dementia. A standard questionnaire revealed that none of the subjects had serious

health problems. All participants had normal or corrected to normal vision. Older

participants were living independently in their own homes. A description of the

relevant characteristics of the age groups is given in table 4.1.

Table 4.1: Characteristics of the sample. Educational level was scored with a dutch educational

coding system (Verhage, 1964) from one to seven, in which one and seven are coded for

respectively no or unfinished elementary school and academically educated.

Age group 20-30 45-60 60-70 70+

N (f/m) 20 (12/8) 20 (12/8) 20 (12/8) 20 (12/8)

Age

Mean 24 (3.1) 53 (3.6) 64 (2.6) 76 (3.2)

Range 20-31 45-58 60-69 70-82

Education 5.8 (.49) 5.4 (.75) 5.3 (.79) 5.4 (1.5)

86


Apparatus and general procedure

For the tasks that were administered using a computer (SART, divided attention

task, switch task, cue task, n-back task and MODS-R), stimuli and instructions

were displayed on a 15’’ VGA monitor controlled by an IBM compatible

computer using MEL2 (Schneider, 1988), or dedicated software (dual task

paradigm). During these tasks, participants were seated approximately 50

centimeters from the screen, in a dimly lit room.

The order of the tasks and the placement of tasks in the first or second session

were controlled in a way that each task administered in the first session for half

of the participants (of each age-group) and in the second session for the other

participants. Also, each task was administered for half of the participants of

each age-group during the first part of the session and for the other participants

during the second part of the session.

Detailed description of tasks

Trailmaking (Reitan, 1958)

Both subtasks of the trailmaking test (A and B), require the participant to draw

an uninterrupted line between a series of 25 circles as quickly as possible. The

route is indicated by the content of the circles. In trailmaking A, in the circles are

randomly distributed numbers and the numbered circles have to be connected in

ascending order. In trailmaking B, there are both numbers and letters in the circles,

and the participant has to alternate between both types of stimuli in ascending

order. Consequently, in the B-version attention has to be divided between different

response categories, a criterion that is not applicable to the A-version. The derived

measures in the current study are the time to complete the A-version, the time

the complete the B-version and the B-A difference. To interpret age by version

interaction (cf. Kliegl et al., 1994), age differences in baseline performance were

taken into account by use of log-transformed completion times.

Zoomap (BADS)

The Zoo map test is a sub-test from the task battery ‘Behavioural Assessment of

Dysexecutive Syndrome’ (BADS) (Wilson, Alderman, Burgess, Emslie and Evans,

1996). The test consists of two conditions. In both conditions subjects are asked to

show how they would visit designated places on a map of a zoo without breaking

chapteR 4

87


the rule of using designated paths once only. The first condition of the test requires

the subjects to plan the order in which to visit the locations. The second condition

gives explicit instructions about in which order the locations are to be visited, in

this way excluding the need for the participant to plan the route. The measure

used are the accuracy-scores of both conditions and the total profile score (see

Wilson et al. (1996) for a more detailed description of the scores).

Sustained attention to response test (SART)

The SART was developed by Manly, Robertson, Galloway and Hawkins (1999). In

the present study the test consisted of a block of 225 trials during which single digits

were presented centrally on the computer screen for 250 milliseconds and then

replaced by mask (a cross with surrounding ring) for 900 milliseconds, resulting in

an onset-to-onset interval of 1150 milliseconds. The digits were displayed in one

of five randomly assigned fonts with heights between 12 and 29 millimeters. Both

digit and mask were white against a black background. Subjects were required to

respond to the digits by pushing the spacebar on the keyboard with the exception

of the target, the number 3, which required no response and occurred 25 times in

a pre-fixed quasi-random fashion. The measure used from this task was the mean

accuracy on the target trials. The experimental phase of the test was preceded by

a block of 18 practice trials, two of which contained the target digit.

Divided attention task (continuous tracking and dot-counting task)

A dual-task paradigm was used for indicating divided attention abilities. The two

tasks the the paradigm consisted of were a continuous tracking task and a dot-

counting task. Variants of the paradigm have been used in prior studies (e.g. Withaar,

2000; Riese, Hoedemaker, Brouwer, Mulder, Cremer & Veldman, 1999). For the

continuous tracking task, a road scene was projected on the computer screen. The

subjects were told to ‘drive’ as straight as possible in the middle of the right lane. This

was complicated by course deviations which were produced by an unpredictable

low frequency noise signal. The subjects were told that the deviations were due to

“side-wind”. The deviation signal consisted of a distortion signal composed of three

superimposed sine waves (1/15, 1/7.5 and 1/3.75 Hz.) and resulted in horizontal

displacements of the road scene. To compensate for this distortion, subjects had to

react by means of a steering wheel placed in front of them. During the tracking task,

lateral position on the road was measured continuously. This was transformed into

scores of the standard deviation of the lateral position (SDLP) in consecutive periods

of 15 seconds. The measure used for performance on the tracking task during a block

was the mean SDLP during that block.

88


The dot-counting task consists of a subject-paced visual analysis task. It required a

rapid reaction by means of pushing the button on the left side or the right side of

the wheel with the corresponding thumb, depending on whether the number of

displayed dots was or was not nine (either eight or ten). For the presentation of the

stimuli, on each trial, random positions from a 10 x 5 matrix on the road scene, just

above the horizon, were chosen to present 8 or 9 or 10 dots. The dot patterns were

presented until the subject responded, but maximally for a duration of 10 seconds,

after which a mask, showing all dots on the 50 positions, was presented for 500

milliseconds. Subjects were instructed to carry out the task as fast and as accurately

as possible. The measure used for performance on the dot-counting task was the

number of correct responses in a block, which comprises speed and accuracy, since

the presentation was subject-paced.

In the single-task condition of the tracking task, no dots appeared on the screen.

During the single-task condition of the dot-counting task, in addition to the dot-

patterns, the road scene was displayed, but road position remained correct and no

steering was required.

Procedure

After receiving general instructions for the task, subjects were given one minute to

get used to the steering wheel (tracking task only scenario). Before the experimental

conditions were run, single task difficulty of lane-tracking was established

individually using an adaptive task procedure during the training blocks. The

adaptation consisted of stepwise adjustment (2 blocks of 12 consecutive periods

of 15 seconds) of a multiplication factor which determined the strength of the

”side-wind”. If performance improved in a 15 second period, the “side-wind”

was increased during the next 15 seconds period, if it had deteriorated, the “side-

wind” was decreased. During the first 15 seconds period of the second adaptation

block mean of the “side-wind” factor of the first adaptation block was used. At the

end of the second adaptation block a stable level of steering quality was achieved

for all subjects. The mean “side-wind” factor of the second adaptation block was

used to individually set the “side-wind” during the remainder of the task, resulting

in the assumption that the difficulty of tracking task was equal for each participant.

After the two adaptation blocks, single-task performance on the tracking task was

assessed during a block of three minutes. During the next block, subjects were

given two minutes of training on dot-counting, after which, during a block of three

minutes, single task performance on the dot-counting task was assessed. In the

chapteR 4

89


next block, subjects were given one minute to get used to the dual task scenario,

followed by a block of 3 minutes to assess dual-task performance.

Forward span and backward span

In the forward condition, the subject had to repeat strings of digits ranging from

3 to 8 digits in ascending order. In the backward condition the strings had to be

repeated backward. Every string length was represented by 2 items. The test was

stopped when both items of a string length were repeated incorrectly. The score

consisted of the total number of correctly repeated strings.

Switch task

Each block started with the presentation of the stimulus-frame, which was a square

(10 by 10 cm) divided in four quadrants. The stimulus-frame remained on the

screen throughout the block. On subsequent trials the position of the presented

stimuli was varied predictably in a clock wise manner, presenting the stimulus on

the first trial in the upper-left quadrant, on the second trial in the upper-right

quadrant etc. The stimulus was a ‘smiley’ and was manipulated in two dimensions.

First, the face could be round (diameter of 3 cm) or oval (3 cm vertical and 2.5 cm

horizontal), and second, the mouth of the face could be sad (corners of the mouth

down) or happy (corners of the mouth up). The experiment consisted of three

conditions, two fixed conditions and a mixed condition. In the mixed condition

(heterogeneous blocks) the relevant dimension was the form of the face if the

stimulus was presented in one of the upper quadrants of the stimulus frame, and

form of the mouth was relevant when the stimulus was presented in one of the

lower quadrants. In the fixed condition (homogeneous blocks) ‘face’, the form of

the face was always the relevant condition, and in the fixed condition ‘mouth’, the

form of the mouth was always the relevant dimension.

The subjects task was to respond with a keypress using their left index finger (V

key) or using their right index finger (N key) depending on the stimulus and the

relevant dimension. If the relevant dimension was the form of the face, the left key

had to be pressed if the face was round and the right key when the face was oval. If

the relevant dimension was form of the mouth, the left key had to be pressed when

the form of the mouth was ‘happy’ and the right key had to be pressed when the

form of the mouth was ‘sad’. A response had to be made within five seconds and

the stimulus remained on the screen for 5 seconds or until a response was made.

If an incorrect response or no response was given, a feedback-tone was presented

for 500 milliseconds. After a response had been made, a response stimulus interval

of 100, 600 or 1500 milliseconds was randomly selected as the preparation interval

90


(PI) for the next trial.

Each of the two fixed conditions and the mixed condition were trained in two

blocks of 40 trials. Before training, a condition was exercised unpaced with stimuli

in a booklet (presented by the experimenter) instead of on the computer screen.

Only after the subject showed complete understanding of the condition the

training of the condition began. If mean accuracy in the second training block of

a condition was below 85%, the training block was repeated. After the training

block the experimental fase began with four blocks of 80 trials of the mixed

condition, then two blocks of 40 trials of one of the fixed conditions, then 80 trials

of the mixed condition and finally 2 block of the other fixed condition. The order

of the fixed conditions was counterbalanced across subjects (in training as well

as experimental blocks). At the beginning of each block, the instruction for the

condition (re)appeared.

Symbol-digit substitution test (DSST)

The symbol-digit substitution test involves showing subjects a series of nine novel

symbols paired with numbers 1 to 9. Next, the subject was given a random

sequence of numbers, and they were asked to substitute the first seven numbers

with the corresponding symbol as training. Subsequently, the test required the

subjects to substitute the remaining symbols with the corresponding numbers as

quickly and accurately as possible within a period of 90 seconds. The measure

used for performance on this test was the number of correctly substituted symbols

within the 90 seconds.

Raven standard progressive matrices (shortened version)

Sets B,C and D from the Standard Progressive Matrices (SPM) (Raven & Court,

1988) were used. Each set consists of 12 items. An item contained a figure with

a missing piece. Below the figure were six (set B) or eight (set C and D) pieces

of which only one could correctly complete the figure. The subjects task was to

identify the correct piece. Within a set, items are arranged in increasing order

of difficulty. Within 20 minutes subjects were required to do as many items as

possible. The measure used for performance on the test was the number of

correct items in 20 minutes.

N-back task

The n-back task consisted of three conditions, the 0-back, 1-back and 2-back

task, which were each administered in two blocks of 33 trials. The order in which

chapteR 4

91


the conditions were administered was counterbalanced across subjects. A trial

consisted of a white letter (consonant) presented in upper case or lower case

on a black background on the screen for a duration of 500 milliseconds, after

which the screen was blanked (black) for 2500 milliseconds. The subjects were

required identify the letter, to push the ‘/’ on the keyboard with their right index

finger if the letter was a non-target, and in case of a target the subjects were

required to push the ‘z’ with their left index finger. The letter specified as target

was different for each condition. In the 0-back condition the target letter was the

letter ‘g’. In the 1-back condition a letter was a target if the letter was the same

as on the previous trial. In the 2-back condition a letter was a target if the letter

was the same as the letter on two trials before the current trial. The identification

of a letter as a target or non-target was irrespective of the case of the letter. The

onset-to-onset interval of the trials was 3000 milliseconds and for the subjects

this was also the maximum time to respond. In a block ten target trials were

presented. Before the experimental task was started, explanation and exercises on

paper were given to the subject. The measures that were derived from the task

were proportion of hits and false alarms and d’, an index of accuracy that takes

into account both hits and false alarms.

Modified Digit Working Memory Span Task (MODS-R)

The span task used was based on the modified digit span task (Daily, Lovett &

Reder, 2001). In each trial, 3 or 5 random digits were presented one at a time in

already present circles on the screen (first digit in the most left circle, the second

in the second to left circle, etc.) for a duration of 500 milliseconds (except for

the final digit, which was presented for a duration of 600 milliseconds) and an

inter-stimulus interval of 100 milliseconds. After the final digit of the string was

presented a new string of digits was presented in the same manner. During a

trial this was repeated for a total of three, four, five or six times, representing the

span-length of the trial. The subjects were to read aloud each presented digit and

were required to remember the final digit of each presented string for later recall.

After the final digit of the final string was presented the subjects were prompted

to type the memory digits in the order in which they were presented as fast and as

accurately as possible. In the two blocks a total of 32 trials were presented (8 trials

of each span-length). Between the blocks subjects were allowed to pause shortly.

PAC-task

In this cue-task, a variant of tasks used by De Jong (2001) and Nieuwenhuis,

Ridderinkhof, De Jong, Kok & Van Der Molen (2000), stimuli were presented in

92


one of four boxes (see figure 4.1) against a black background on the screen. The

stimuli consisted of schematic face in the shape of a circle. The mouth differentiated

between a happy and a sad face. In all conditions the subjects were required to

identify the shortly presented stimulus, as happy or sad. The box in which the

stimulus appeared was cued differently in the three experimental conditions. In

the procue condition the cue consisted of a brief flickering of the target box. In the

anticue condition the box opposite to the target box flickered. In the central cue

condition an arrow was displayed at the position of the fixation cross, pointing to

the target box.

Figure 1 shows an example of a sequence of events in a trial of one of the

conditions. First a fixation cross was presented for 1000 milliseconds, then after 200

milliseconds the cue was presented, which consisted of (in the pro- and anticue

conditions) one of the boxes disappearing or (in the central cue condition) an

arrow being presented at the position of the fixation cross, for the duration of 83

milliseconds. After a stimulus onset asynchrony (SOA) of 200 or 1400 milliseconds,

starting at cue offset, the face was displayed for an individually set duration, after

which the mask was presented. The mask display remained on the screen until

500 milliseconds after a response had been given and consisted of a white filled

rectangle which was presented over the mouth.

Figure 4.1: Example of a sequence of events for a pro-cue trial (see text for the specific duration

of the displays).

Before the experimental conditions the subjects were able to exercise the task

on paper in an unpaced interactive manner with the experimenter. After the

exercises on paper, the subjects received a practice set consisting of a block of

120 trials and a block of 60 trials, during which the cue consisted of all four

boxes disappearing for 83 milliseconds (neutral cue). During the practice set the

time was determined that the mouth should be displayed, before being masked,

in order to yield 67% correct responses with the neutral cue for that particular

Fixation displayF ixation disp layT arget display Mask displayCue display

chapteR 4

93


subject. This was done by means of a staircase tracking algorithm. The resulting

individual target duration was used as onset-to-onset interval of the stimulus and

the mask in the experimental conditions. The three experimental conditions were

each administered in one block of 90 trials. During the experimental condition

one quarter of the trials exhibited a neutral cue (all four boxed flickering in the

anti- and procue conditions, and presentation of all four arrows simultaneously

in the central cue condition). The order in which the conditions were performed

was counterbalanced across subjects.

4.3 Results and discussion

Trailmaking

The means and standard deviations of the age-groups on the trailmaking A,

trailmaking B and the B-A difference are shown in table 4.2. The time needed for

completing the test versions was affected by age group (F(3,76)=11.5, MSE=1473,

p<.001,) and more time was needed for the B version than for the A version

(F(1,76)=130.2, MSE=665, p<.001). The interaction of test version and age group

was significant when using the absolute time measures (F(3,76)=7.7, MSE=665.3,

p<.001). Interestingly there was no significant age-group by test-version effect

when using the log transformed measures (F(3,76)=2.1, MSE=.05, p>.1), which

suggests that the additional time time needed by the older groups with the

B-version is proportional to the additional time the younger adults needed. To

gain more detailed insight in the differences between the four age groups, in

table 4.13 the results are shown of the pairwise comparisons of the B-A difference

measure using Tukey HSD. This analysis revealed that the difference between time

to complete the A and B version was significantly larger for the oldest group in

comparison with the youngest two age-groups. Again these effects seized to be

significant when using the log-transformed data.

Table 4.2: Means and standard deviations of completion time (seconds) of the trailmaking test in

seconds for the four age groups.

Age group 20-30 50-60 60-70 70+

Trailmaking A 35 (14) 42 (15) 51 (21) 56 (15)

Trailmaking B 62 (19) 72 (17) 104 (58) 132 (59)

B-A 27 (14) 30 (18) 53 (45) 76 (53)

94


Difference in completion times between version-B and version-A has been linked to

executive functions such as planning of actions and switching between tasks. Advancing

age is usually related to slower performance on the B-version in comparison with the

A-version. In several studies it was found that, when controlling for performance on

the A-version, the age effect of age on performance on the B-version was no longer

significant (e.g. Salthouse & Fristoe, 1995; Salthouse et al. 2000). Other studies did find

a relation between age and performance measures on the B-version after controlling

for performance on the A-version (Keys and White, 2000; Kuhlman, Little and Sekuler,

2006; Salthouse et.al 1996). The current results show performance differences between

the task-versions not to be dispropionately larger for older than for younger adults,

consistent with the studies by Salthouse et al. (1995, 2000).

In addition to differences in mean completion times with respect to both A and B versions

of the trailmaking test, there were sizable age group differences in standard deviations,

especially with respect to completion times of the B version of the test. To illustrate this

aspect of the data figure 4.2 shows the individual completion times of both versions

in a scatterplot. It shows that there was one extreme performer in the 60’s group with

respect to both the A and B versions. The scatterplot also shows that on the B version

some subjects in the oldest two groups were not performing worse than the subjects in

the other age groups, and that there was a greater spread in the oldest groups.

Figure 4.2: Scatterplot with completion times of trailmaking A and trailmaking B (circles: 20-35yrs., triangles: 50-60yrs; squares: 60-70yrs; crosshairs: 70+yrs). The presented regrssion-lineis derived from the performance of the youngest group.

25 50 75 100

A (in s.)

100

200

300

B (

in s

.)

chapteR 4

95


Digit-symbol substitution test

Table 2 shows the results of the symbol-digit substitution test. The number

of symbols correctly substituted in 90 seconds was affected by age group

(F(3,76)=15.8, MSE=103.6, p<.001). Pairwise comparisons using Tukey HSD (table

12) revealed the youngest group to perform significantly better than the three

other age groups, while differences between the latter three age groups did not

reach significance.

Table 4.3: Means and standard deviations of the number of correct symbol-digit substitutions

performed in 90 seconds.

These results are consistent with previous findings in the literature (e.g. Salthouse,

1993, Parkin & Java, 2000). The interpretation of the DSST measure is more

controversial than the result in itself, though. In studies presenting evidence

supporting a general speed factor, often the performance on DSST is used as

a measure of perceptual speed (e.g. Salthouse, 1993; Bryan and Luszcz, 1996;

Dunlosky & Salthouse, 1996). In these studies, when the DSST is used as a covariate,

usually a very large share of the age-related cognition measure variance is removed.

Parkin and Java (2000) discuss the interpretation of DSST as ‘pure’ measure for

perceptual speed. First, they argue that performance on DSST is dependent on

memory, and that DSST speed would be enhanced if the associations between

digits and symbols were learned because less reference to the code table would be

required. Furthermore, Parkin and Java (2000) argue, the DSST measure is not only

related to perceptual speed and memory, but also to intelligence. The IQ sensitivity

of the DSST-measure (a sub-test of the WAIS-R; Wechsler, 1987) is evidenced by its’

correlation of 0.65 with IQ on the WAIS-R.

Raven standard progressive matrices (shortened version)

Table 3a shows the results of the short version of the standard Raven test. The test

showed decreased performance with aging (F(3,76)=14.5, MSE=29.5, p<.001).

Age group 20-30 50-60 60-70 70+

N correct substitutions 65 (10) 52 (10) 49 (11) 43 (9)

96


Table 4.13 shows the results of the post-hoc pairwise comparisons and reveals

significant differences between the oldest group and the two youngest groups

and between the 60-70 group and the youngest group. Table 4.4a also shows that

standard deviations vary between age-groups. This is mainly due to all participants

of the youngest group performing with a score higher than 26, while some of the

participants of the two oldest groups perform just as well and some perform much

worse (see table 4.4b).

The Raven SPM test has widely been used as an indicator of fluid intelligence. The

test was included in the task-battery because fluid intelligence measures have been

found to be sensitive to age differences and to frontal functioning.

The finding of age-related decline of unadjusted scores on this fluid intelligence

test, corresponds with previous research (Holland & Rabbitt, 1990; Horn &

Masunaga, 2000, Parkin & Java, 2000). According to Duncan et al. (1995, 1997)

fluid intelligence corresponds to executive function and to the integrity of the

frontal lobes. From this viewpoint, the age-related effect on performance on the

Raven SPM provides support for the frontal lobe hypothesis of aging. There is

evidence that crystallized intelligence and fluid intelligence dissociate with age.

Salthouse et al. (1996), for instance, found vocabulary to remain invariant or even

show improvement with age.

Table 4.4a: Means and standard deviations of the number of correctly solved problems of the

Raven standard progressive matrices series B, C and D (each comprising of 12 problems).

Table 4.4b: Number of participants of each age-group performing within different score-clusters

on the Raven SPM.

RavenSPM score

Age-group 0-6 7-12 13-18 19-24 25-30 31-36

20-30 0 0 0 0 4 16

50-60 0 0 0 1 10 9

60-70 0 2 0 2 9 7

70+ 0 2 3 7 5 3

Age group 20-30 50-60 60-70 70+

Score Raven SPM short 33.2 (2.8) 30.3 (3.3) 26.8 (6.8) 22.5 (7.3)

chapteR 4

97


Zoomap (BADS)

The results of the zoomap test are presented in table 4.5. ANOVA on the profile score

of the zoomap-test, showed a main effect of age group (F(3,76)=4.1, MSE=1.69,

p<.01). Tukey’s HSD (table 12) showed that the profile score for youngest group

was significantly higher than for the oldest group. Additionally, repeated measures

analysis was performed on the scores of the two conditions. Performance during

the first condition, which required planning from the subject, was worse than

performance on the second condition during which no planning was required

from the subject (F(1,76)=124.5, MSE=11.954, p<.001). Performance was also

affected by age group (F(3,76)=3.6, MSE=48.467, p<.05). In the condition during

which the participant needed to plan the route, performance of older adults was

clearly worse than for younger adults, while when no planning was needed, the

participants from the different age-groups performed just as well (age-group by

condition: F(3,76)=3.4, MSE=11.954, p<.05). Thus, the age-group effect on the

profile score can be attributed to age-group differences in performance in the

planning condition.

Table 4.5: Means and standard deviations of the accuracy scores of the two conditions and of the

total profile score on the Zoomap test.

Age group 20-30 50-60 60-70 70+

Planning 4.4 (5.1) 2.2 (4.8) 1.1 (5.2) -.6 (4.9)

No planning 8.0 (.2) 8.0 (0) 7.8 (.4) 7.8 (.72)

Profilescore Zoomap 3.0 (1.3) 2.4 (1.2) 2.0 (1.4) 1.6 (1.3)

Forward span and backward span

Table 7 shows the data from the forward and backward span. The pattern of results

in both the forward and backward span test showed slightly better performance

for the youngest age group as compared to the older age groups. Analysis of

variance revealed no significant effect of age group, though, on either forward

or backward span performance (respectively: F(3,76)=1.2, MSE=4.870, p>.3;

F(3,76)=1.6, MSE=4.382, p>.15).

It is commonly assumed that, due to the need for executive control in addition

to short term memory, the backward span is more strongly affected by aging

than forward span. Indeed, the manual of the WAIS-III, of which the forward and

98


backward span are sub-tests, reports a larger discrepancy between forward and

backward span for older adults than for younger adults. Also, Babcock & Salthouse

(1990) reported a meta-analysis that indicated a larger effect of age on backward

than on forward digit span. On the other hand, Verhaeghen et al. (1993) found,

using a meta-analysis on several studies, that age effects were not different between

the forward and backward span. Although Bopp and Verhaeghen (2005) did find

different age-related effects for the forward and backward span, when examining

the averaged data, age differences in backward and forward digit span were not

significantly different. Park, Lautenschlager, Hedden, Davidson, Smith & Smith

(2002) also show evidence for an equivalent age-related effect on forward and

backward span. Moreover Meyerson, White and Hale (2003) and Hester, Kinsella

and Ong (2004), analyzing the standardization sample of WAIS-III, found that the

age-related performance decline was equivalent for both tests. Gregoire and Van

der Linden (1997) came to a similar conclusion when analyzing the data from the

standardization sample for the French adaptation of the WAIS-R.

In conclusion, although performance on the backward span is commonly believed

to be affected more strongly by aging than performance on the forward span,

most studies reveal age invariance with respect to the difference of performance

between the forward and backward span. The current results are consistent with

these studies. The inconsistency with the study of Babcock and Salthouse (1990)

remains unresolved.

Table 4.6: Means and standard deviations of the number of correct trials in the forward and

backward span tests (prior to two successive incorrect trials)

Sustained attention to response test

Table 5 shows the accuracy performance of the age groups on targets in the SART.

The main effect of age group was significant (F(3,76)=2.8, MSE=.033, p<.05).

The post-hoc analysis revealed non of the pairwise age group comparisons to be

significant, but table 4.7 reveals the greatest difference in performance between

the 20-30 group and the three older groups. No other studies were found utilizing

the SART to study sustained attention in normal aging. Manly, Robertson, Galloway

and Hawkins (1999) found that the performance on the SART was predictive of self-

Age group 20-30 50-60 60-70 70+

Forward span (WAIS-) score 7.8 (2.2) 6.7 (2.0) 6.9 (2.4) 6.7 (2.3)

Backward span (WAIS-) score 8.1 (1.9) 6.8 (2.0) 7.1 (1.8) 6.9 (2.6)

chapteR 4

99


reports of everyday attentional failures and action-slips (CFQ, Broadbent, Cooper,

Fitzgerald and Parkes, 1982) in both brain injured and normal participants. Several

studies have shown that old adults do not report more cognitive failures in everyday

activities than do young adults (Rabbitt, 1990; Kane et al. 1994; Kramer et al.,

1994). Taking these findings together, one might expect an absence of age-related

effects on performance on the SART. Thus, the current finding of a difference

between young and old adults is inconsistent with this indirect argument. It goes

beyond the scope and possibility of the present paper to resolve this inconsistency.

The finding of an age-related effect on sustained attention is in line with Chao and

Knight (1997), who made use of another sustained attention task.

Table 4.7: Means and standard deviations of the proportion of correctly omitted responses.

Divided attention test

Table 6 shows performance of the age groups on the adaptation of tracking difficulty,

and on the tracking task and dot-counting task in single-task and dual-task conditions.

The performance measure used for dot-counting was the number of correctly counted

dots, resulting in an aggregate score of accuracy and speed of the responses. An

important assumption for using this measure as a performance index is that subjects

respond on the basis of counting and not ‘speed-guessing’. This assumption was tested

individually, and one subject of the youngest group was evidently randomly guessing

and not counting during the dual task condition (very fast responses and accuracy

of .5). Data of that subject were therefore excluded from further analysis on this test.

Single task performance.

As indicated by the adaptation factor (table 6), tracking performance was affected by

age-group (F(3,75)=8.7, MSE=6.521, p<.001). As shown in Table 4.13 the adaptation

factor of the youngest group was significantly different from the three older age

groups, while the differences between the older groups did not reach significance.

Using the adaptation factor, tracking task difficulty was individually adapted to

reach equal tracking performance between the groups in a single task situation.

This adaptation was successful as revealed by the absence of a significant effect

Age group 20-30 50-60 60-70 70+

Prop.correct no resp. .82 (.18) .69 (.18) .70 (.17) .68 (.19)

100


of age-group on tracking performance on the single task condition (F(3,75)=1.1,

MSE=46.767, p>.35). Also, performance on the dot-counting task in the single-task

condition was not affected by age group (F(3,75)=1.4, MSE=92.079, p>.25). Thus,

the age-groups showed similar performance on the tasks in single task condition,

making these suitable control conditions for the dual task conditions.

Dual-task performance.

As shown in table 4.8, subjects performed worse on the tracking task in the

dual task condition as compared to the single-task condition (F(1,75)=67.9,

MSE=140.816, p<.001). Furthermore this dual-task effect interacted with age

group (F(3,75)=4.5, MSE=140.810, p<.01). Pairwise comparisons (see table 4.13)

showed that the youngest group exhibited significantly smaller divided attention

costs in tracking performance than the two oldest age groups.

With respect to the dotcounting task, performance was also better in the single task

condition than in the dual task condition (F(1,75)=75.0, MSE=14.652, p<.001).

The interaction effect of age group and condition did not reach significance

(F(3,75)=2.5, MSE=14.650, p>.06). In summary, performance on both tasks

decreased when subjects were confronted with the dual task condition.

As is the case for most of the indices of executive function used in the present paper,

evidence on the relation between aging and dual-task costs is mixed. Some studies

failed to demonstrate age-related decline in dual-task performance (Salthouse and

Somberg, 1982; Tun and Wingfield, 1994; Wickens, Braune and Stokes, 1987).

Many other studies have found that aging is associated with increased dual-task

costs (e.g. Brouwer et al., 1991; Holtzer, Stern, Rakitin, 2005; Ponds et al., 1988;

Verhaeghen et al., 2003). Dual-task methodology has been used for measuring

executive functions (Baddeley and Hitch, 1974; Baddeley, 2001; Meyer and Kieras,

1997). Hartley (1992) suggested that age-related costs of dual-tasking may be

due to compromised performance by older adults on the single task conditions.

In the paradigm used here, the level of performance on both the single tasks

were successfully adapted to be age-invariant. Moreover, although the perceptual

domain of the tasks is the same, the stimuli of the tasks are not ambiguous and the

tasks do not require the same responses. Accordingly, the age-related differences

in dual-task costs in the present study are most likely not due to interference at the

motoric or perceptual level, but to dealing with the need for dividing attention

between the two tasks.

chapteR 4

101


From a more qualitative viewpoint, the youngest group showed large dual-task

costs in dot-counting performance while keeping tracking performance relatively

stable. Instead, the three older groups showed relatively more dual-task costs in

tracking performance as compared to dot-counting performance. This pattern of

results suggest that older adults exert different strategies than younger adults with

respect to coping with demands of dividing attention between two tasks.

Table 4.8: Means and standard deviations of the divided attention task measures. The used

adaptation factor is the mean adaptation score of the second adaptation block. Tracking

performance is measured by the standard deviation of the mean lateral position during a

block (lower number means better performance). Dual-task costs in tracking are calculated by

subtracting single task performance from dual task performance. Dot-counting performance is

measured by the number of correct categorized dot patterns (as consisting of 9 dots or not (8 or

10)). Dual-task costs in dot-counting are calculated by subtracting dual task performance from

single task performance.

N-back task

Table 8 shows the mean performance of the age groups in the three conditions of

the n-back task. Analysis was performed on d’ data. The main effect of condition

was found to be significant (F(2,150)=103.8, MSE=.204, p<.001). There also was

a significant main effect of age group (F(3,75)=3.8, MSE=.510, P<.05), and post-

hoc analysis showed the youngest age group to perform better than the oldest

group, while the other pairwise comparisons between the age groups were not

significant (see table 4.13). Keys et al. (2002) found the 2-back condition to be

most sensitive to aging in comparison with 0-back, 1-back and 3-back conditions.

In table 4.9 it can be seen that in this study the most pronounced performance

Age group 20-30 50-60 60-70 70+

Adaptation factor 8.2 (3.0) 4.8 (2.1) 5.1 (2.4) 4.7 (2.4)

Tracking performance singletask 39.7 (5.8) 43.2 (6.2) 42.8 (7.3) 42.8 (7.8)

Tracking performance dual task 44.0 (10.8) 58.8 (17.4) 62.7 (14.5) 65.3 (20.1)

Dual-task costs in tracking performance 4.2 (13.8) 15.6 (16.8) 19.9 (16.5) 22.5 (19.4)

N of correct dot-countings in single task 54.4 (10.7) 48.9 (8.3) 50.0 (7.0) 49.4 (11.7)

N of correct dot-countings in dual task 47.4 (10.8) 46.4 (6.1) 44.4 (4.7) 43.4 (10.6)

Dual-task costs in dot-counting performance 7.1 (5.3) 2.6 (5.4) 5.6 (4.8) 6.0 (6.0)

102


difference between the age-groups were also found in the 2-back condition. The

interaction between condition and age-group, though, did not reach significance

(F(6,150)=1.7, MSE=0.204, p>.1).

Table 4.9: Means and standard deviations of the proportion of correctly and incorrectly identified

targets (resp. hitrate and f.a.rate) and of the derived d’.

MODS-R working memory task

The results of the MODS-R working memory task are presented in table 4.10.

Performance on the MODS-R working memory task was significantly affected by

the to be remembered span-length (F(3,228)=167.1, MSE=112.622, p<.001).

Although the mean performance of the older groups was, consistently over span-

lengths, worse than performance of the youngest group the main effect of age

group was not significant (F(3,76)=2.6, MSE=816.685, p>.05). An interaction of

age group and span-length was not found (F(9,228)<1, MSE=112.618, p>.7).

In accordance with the results on the forward and backward span tests, discussed

earlier in this section, no significant effect of age-group was found in the MODS-R

task, which is also a digit span task. The MODS-R (Daily, Lovett and Reder, 2001)

Condition 20-30 50-60 60-70 70+

Hitrate 0-back .96 (.06) .97 (.04) .92 (.20) .93 (.09)

1-back .92 (.10) .88 (.10) .87 (.16) .82 (.18)

2-back .83 (.20) .80 (.14) .73 (.17) .67 (.18)

f.a.rate 0-back .01 (.01) .01 (.02) .01 (.01) .01 (.01)

1-back .01 (.01) .01 (.01) .02 (.04) .04 (.10)

2-back .02 (.02) .03 (.03) .03 (.04) .08 (.13)

D’ 0-back 3.75 (.43) 3.74 (.27) 3.73 (.40) 3.61 (.42)

1-back 3.55 (.49) 3.34 (.49) 3.36 (.53) 3.01 (.89)

2-back 3.06 (.73) 2.77 (.63) 2.59 (.72) 2.19 (.83)

chapteR 4

103


differs from the typical span tasks in that it prevents participants to utilize subvocal

rehearsal as a performance strategy. The use of this rehearsal strategy was restricted

by means of presenting irrelevant digits, that were to be read aloud by the subject,

in the time interval between the presentation of the to be remembered digits,

and consequently resulted in virtually continuous articulation (Daily et al., 2000).

Furthermore, the filler items increased the delay before recall, adding to the

working memory demands of the task.

Using ACT-R as cognitive architecture Daily et al. (2000) presented a computational

model of individual differences in working memory. They proposed that, holding

prior knowledge and strategic approaches relatively constant, individual differences

in performance on working memory tasks can be largely attributed to differences

in their amount of source activation (W). Source activation is a kind of activation

used in ACT-R to maintain goal-relevant information in an available state relative

to information that is less relevant to the goal. They were able to accurately model

individual behavior on the MODS-R task by varying only the source activation

parameter (W). In another study Lovett et al. (2002) found that the individual

source activation parameter could be estimated on basis of the MODS-R task

and used to produce fits to individual participant data from another qualitatively

different working memory task.

Together with the modeling work of Lovett and Daily and coworkers, the pattern

of results suggest compromised source activation as an explanation of the patterns

of aging on performance on cognitive tasks, but provide no convincing evidence

supporting this explanation.

104


Table 4.10: Means and standard deviations of the percentage correctly recalled spans.

PAC-task

Figure 4.3: Graphs of the mean proportion of correct responses of the four age groups in the pro-

cue (filled line), anti-cue (dotted) and central cue (striped) conditions on trials with informative

cues and long and short SOA’s.

Cohort 20-30 50-60 60-70 70+

Span length 3 88 (11) 82 (17) 78 (17) 79 (19)

4 72 (15) 70 (18) 63 (21) 60 (22)

5 63 (12) 52 (18) 52 (18) 50 (21)

6 50 (14) 46 (11) 40 (15) 39 (18)

0.96

0.90

0.84

0.78

0.72

0.66

20-30 yrs

prop

ortio

n co

rrec

t

50-60 yrs

0.96

0.90

0.84

0.78

0.72

0.66

60-70 yrs

prop

ortio

n co

rrec

t

70+ yrs

SOA SOA 200 1400 200 1400

chapteR 4

105


Accuracy performance on the cue task is shown in figure 4.3 and, in more detail,

in table 4.11. Analysis of the trials with an informative cue revealed a main effect

of condition (F(2,150)=114.1, MSE=.007, p<.001) and a main effect of SOA

(F(1,75)=301.9, MSE=.008, p<.001). Best performance was reached in the pro-cue

condition and with long SOA’s. There was a main effect of age group (F(3,75)=4.7,

MSE=.036, p<.01), reflecting better performance of the youngest group as

compared to the two oldest groups (see table 4.13 for pairwise comparisons).

The interaction between condition and SOA was also significant (F(2,150)=13.9,

MSE=.005, p<.001). Figure 4.3 shows the performance in the central cue condition

to benefit more strongly from a long SOA than performance in the pro-cue and

anti-cue condition. Age group interacted with condition (F(6,150)=2.3, MSE=.008,

p<.05), reflecting larger differences between conditions for older adults than for

younger adults. One-way ANOVA’s with age-group as between subject factor

on accuracy performance in the three conditions showed that performance in

the pro-cue condition was not significantly affected by age-group (F(3,78)=1.7,

MSE=.007, p>.15), while there were significant age-related difference in the anti-

cue and central-cue condition (resp. F(3,78)=5.7, MSE=.009, p<.01; F(3,78)=4.56,

MSE=.008, p<.01). The three-way interaction of age groups, condition and SOA

was borderline significant (F(6,150)=2.1, MSE=.005, p<.06).

The finding that, for older adults, anticue performance was worse than procue

performance, even at long SOAs, is consistent with findings in previous studies

using a cue-task (De Jong, 1999; Nieuwenhuis et al. 2004). The current results

also show this asymptotic difference between the two conditions in younger

adults, while the previous studies found no difference in performance at long SOAs

between the two conditions. It is not clear to what this difference between the

present and previous studies, in performance of younger adults, can be attributed.

One difference in method that possibly could have played a role is the difference in

characteristics of participants. The previous studies used undergraduate students,

while in the present study the use of academic participants was avoided as much

as possible in order to match the educational level of the older groups.

106


C

on

dit

ion

C

ue

SOA

2

0-3

0

50

-60

6

0-7

0

70

+

Pr

o

Info

Sh

ort

0.

87 (

.11)

0.

81 (

.12)

0.

83 (

.13)

0.

84 (

.11)

Pr

o

Info

Lo

ng

0.98

(.0

3)

0.96

(.0

6)

0.91

(.0

9)

0.92

(.0

8)

C

entr

al

Info

Sh

ort

0.

81 (

.11)

0.

72 (

.15)

0.

68 (

.13)

0.

73 (

.14)

C

entr

al

Info

Lo

ng

0.98

(.0

3)

0.90

(.1

4)

0.90

(.1

0)

0.91

(.0

6)

A

nti

In

fo

Shor

t 0.

71 (

.12)

0.

67 (

.09)

0.

65 (

.09)

0.

64 (

.08)

A

nti

In

fo

Long

0.

89 (

.14)

0.

87 (

.08)

0.

75 (

.14)

0.

75 (

.14)

Pr

o

Neu

t Sh

ort

0.

68 (

.16)

0.

67 (

.15)

0.

63 (

.15)

0.

71 (

.16)

Pr

o

Neu

t Lo

ng

0.71

(.1

8)

0.69

(.1

4)

0.68

(.1

4)

0.67

(.1

4)

C

entr

al

Neu

t Sh

ort

0.

75 (

.17)

0.

60 (

.17)

0.

61 (

.17)

0.

61 (

.17)

C

entr

al

Neu

t Lo

ng

0.75

(.1

1)

0.64

(.1

5)

0.57

(.1

5)

0.63

(.1

6)

A

nti

N

eut

Shor

t 0.

68 (

.17)

0.

66 (

.13)

0.

61 (

.16)

0.

61 (

.16)

A

nti

N

eut

Long

0.

79 (

.14)

0.

71 (

.15)

0.

67 (

.19)

0.

64 (

.19)

Table

4.1

1: M

eans

and s

tandard

dev

iati

ons

of a

ccura

cy per

form

ance

(pro

por

tion

cor

rect

) on

the

pro

-cue,

anti

-cue

and c

entr

al c

ue

condit

ions

div

ided

into

info

rmati

ve c

ues

(in

fo)

and n

eutr

al c

ues

(neu

t) a

nd lo

ng a

nd s

hor

t st

imulu

s on

set

asy

nch

rony

(SO

A)

by

age

gro

up.

chapteR 4

107


Table

4.1

1: M

eans

and s

tandard

dev

iati

ons

of a

ccura

cy per

form

ance

(pro

por

tion

cor

rect

) on

the

pro

-cue,

anti

-cue

and c

entr

al c

ue

condit

ions

div

ided

into

info

rmati

ve c

ues

(in

fo)

and n

eutr

al c

ues

(neu

t) a

nd lo

ng a

nd s

hor

t st

imulu

s on

set

asy

nch

rony

(SO

A)

by

age

gro

up.

Recently, studies on aging using similar tasks have been conducted (De Jong et

al. 1999; Nieuwenhuis et al. 2000, 2004). Nieuwenhuis et al. (2000) used an

integrated pro vs. antisaccade and pro- vs. anticue task. In addition to accuracy

on the identification as performance measure they recorded eye-movements. The

results of the Nieuwenhuis et al. (2000) study showed older adults to be impaired

at actively suppressing prepotent eye movements towards the location of the cue.

This was indicated by an increased percentage of inappropriate, reflexive saccades,

and by a task specific slowing of antisaccade generation. Interestingly, Nieuwenhuis

et al. (2000, experiment 1) also found, only in the anti-cue condition for older and

not for younger adults, that there was a strong relationship between SOA and

the onset latency of anti-saccades. For older adults, longer SOA’s were associated

with slower (initiation of) anti-saccades. Based on this relationship Nieuwenhuis et

al. (2000) hypothesized that the older subjects adopted a strategy in which they

attempted to take advantage of the onset of the target stimulus to trigger the

antisaccade. Only when the target did not appear promptly after the cue, the older

subjects engaged in endogenous initiation of the antisaccade. Nieuwenhuis et al.

(2000) interpreted this as older adults exploiting external support to help achieve

the task goal (target identification) with reduced effort (see also Nieuwenhuis et al.,

2004). To test this hypothesis, Nieuwenhuis et al. (2000) conducted an additional

experiment in which they neutralized the unique exogenous qualities of the target

stimulus by presenting the distractors (without the identifiable feature) in the non-

target locations simultaneous with the target stimulus. This manipulation of the

cue-task necessitated fully endogenous initiation of the required antisaccade. The

results showed, in contrast to the first experiment, that age differences in accuracy

performance in the anticue condition were small and not larger than in the pro-

cue condition. Moreover, eye-movement data indicated that this reduction in age

difference was caused by decreased age differences in antisaccade speed, especially

at the longer SOAs. Nieuwenhuis et al. (2000) concluded that when the possibility

of the use of exogenous control is minimized and the use of endogenous, voluntary

control is necessary to attain and maintain adequate task performance, older adults

can and do engage in endogenous control.

In the present experiment, we used an alternative approach to minimize the use of

exogenous control. Instead of manipulating the target, we used a third condition

with an alternate, symbolic cue. In this central cue condition there was no need

to suppress a prepotent response as was necessary in the anticue condition.

Furthermore, in the central cue condition, the cue in itself did not serve as a reflexive

trigger of attention to the appropriate location, as it was in pro cue condition. In

108


other words, the possibility of inappropriate (anticue) and appropriate (procue)

exogenous control usage of the cue was minimized in the central cue condition.

To perform adequately subjects needed to voluntarily and endogenously direct

attention to the appropriate target location. Clearly, with exception of the 50’s

group, at long SOA, performance in the central cue condition was equivalent to

performance in the procue condition for younger as well as for older adults (see

figure 4.3). Performance in the anticue condition at long SOA did not reach the

performance level of the procue condition.

This pattern of results strongly suggests that the degraded performance of older

adults in the anticue condition is due to the difficulty with suppressing a prepotent

inaccurate response to direct attention to the imperative cue. Furthermore, in the

present experiment it is not plausible that the difference in performance at long

SOA between pro- and anticue is due to the adoption of a strategy that uses the

onset of the target stimulus as an exogenous control of attention. If such a strategy

would have been adopted in the anticue condition, it should also be expected to

be adopted in the central cue condition, which in turn would have resulted in a

difference in performance between the central cue and pro cue condition.

Switch task

The results of the switch task are presented in table 4.12 and figure 4.4. One

subject was exluded from analyses because the data showed a lack of following

task-instructions resulting for example in overall negative global switch costs of

279ms.

The reaction time data were analyzed in a 3 (trialtype) X 3 (rsi) X 4 (age group)

GLM. There was a significant main effect of trial type (F(2,146)=384.9, MSE=28718,

p<.001). To distinguish between global switch-costs (differences in reaction times

between task-repetitions trials in mixed blocks on one hand and trials in fixed-

task blocks (fixed task trials) on the other) and local switch costs (differences in

reaction times between task-repetitions trials and switch trials in mixed blocks),

contrast analysis were used. These contrasts revealed global and local switch costs

in reaction times (resp. F(1,73)=129.4, MSE=37969.2, p<.001 and F(1,73)=330.3,

MSE=56520.2, p<.001).

chapteR 4

109


1300

1200

1100

1000

900

800

700

600

500

20-30 yrsRT

(m

s)50-60 yrs

RT (

ms)

PI (ms) 100 600 1500

PI (ms) 100 600 1500

1300

1200

1100

1000

900

800

700

600

500

60-70 yrs 70+ yrs

Figure 4.4: Graphs per age group of the mean reaction times (RT) on no-switch trials (dotted

lines), switch trials (striped lines) and trials in fixed task blocks (filled lines) with short, middle

and long response-stimulus intervals (RSI).

There was a main effect of RSI (F(2,146)=16.0, MSE=6903.5, p<.001). Responses

were slower on trials with an RSI of 100 milliseconds than on trials with an RSI of

600 milliseconds, while there was little to no difference between reaction times on

trials with an RSI of 600 milliseconds and on trials with an RSI of 1500 milliseconds.

An interaction effect of RSI and trial-type was found (F(4,292)=50.0, MSE=2253.1,

p<.001) and contrast analysis revealed that both global and local switch costs

contrasts interacted significantly with RSI (resp. F(1,73)=46.1, MSE=5197.0, p<.001

and F(1,73)=151.6, MSE=5924.4, p<.001). Figure 4.4 shows that RT’s on switch

trials were faster with longer RSI, RT’s on no-switch trials were slower with longer

110


RSI’s, and RT’s on fixed task trials were relatively invariant to RSI. Consequently

local switch costs decreased with longer RSI’s, and global switch costs were larger

for longer RSI’s.

There was a significant main effect of age group on response speed (F(3,73)=12.9,

MSE=64242.7, p<.001) and in table 4.13 it is shown that the youngest group

was faster than the other three age groups and the 50-60 years age group was

significantly faster than the 70+ age group. The interaction of age group by trial-

type was significant (F(6,146)=3.8, MSE=28719.1, p<.01) and contrast analysis

showed that the global switch costs contrast did not significantly interact with

age group (F(3,73)=1.9, MSE=37964, p>.1), while the local switch costs contrast

did significantly interact with age group (F(3,73)=2.8, MSE=56529.5, p<.05). Pair-

wise comparisons showed (see table 4.13) the youngest and oldest age group to

significantly differ in local switch costs. The interaction between age-group and RSI

was significant (F(6,146)=4.9, MSE=6903.5, p<.01). The youngest group showed

a decrease in RT at longer RSI, while the older groups showed no decrease in

mean RT at the longest RSI as compared with the middle long RSI. The three way

interaction of age-group, RSI and trial type was not significant (F(12,292)<1).

With respect to accuracy, significant local switch costs were found (F(1,73)=96.9,

MSE=.001, p<.001), while global switch costs were not present (F(1,73)<1). The

effect of local switch costs interacted with RSI (F(1,73)=9.0, MSE=.0008, p<.01).

Local switch costs in accuracy were smaller at trials with the longest RSI as compared

with the shorter RSI’s. No significant main effect of RSI (F(1,73)<1) no significant

main effect of age-group was found on accuracy (F(3,73)=1.3, MSE=.003, p>.2).

chapteR 4

111


Table 4.12a: Means (standard deviations) of reaction times and switch costs on the switch task

for the age-groups per trial type per response-stimulus-interval (in ms).

Table 4.12b: Accuracy on the different trial types and switch costs on the switch task for the age-

groups per trial type per response-stimulus-interval (in proportion of trials correct).

RSI 20-30 50-60 60-70 70+

Fixed task trials 100 541 (53) 638 (80) 650 (91) 770 (195)

600 527 (56) 594 (58) 624 (85) 679 (156)

1500 511 (59) 630 (109) 658 (141) 731 (202)

Non-switch trials 100 599 (76) 744 (117) 786 (178) 908 (181)

600 626 (114) 750 (163) 796 (180) 856 (153)

1500 640 (128) 827 (197) 898 (219) 913 (184)

Switch trials 100 900 (119) 1124 (173) 1160 (194) 1336 (305)

600 819 (168) 1030 (175) 1104 (246) 1233 (282)

1500 783 (198) 1038 (241) 1116 (300) 1230 (365)

Global switch costs Pooled 95 (73) 153 (124) 176 (129) 160 (117)

Local switchcosts Pooled 212 (93) 291 (138) 300 (136) 374 (226)

RSI 20-30 50-60 60-70 70+

Fixed task trials 100 .99 (.02) .97 (.03) .97 (.04) .95 (.06)

600 .97 (.02) .98 (.02) .97 (.03) .96 (.04)

1500 .96 (.04) .98 (.02) .97 (.03) .96 (.04)

Non-switch trials 100 .98 (.02) .98 (.02) .97 (.04) .97 (.05)

600 .97 (.02) .98 (.02) .98 (.02) .96 (.07)

1500 .96 (.03) .97 (.03) .97 (.02) .96 (.05)

Switch trials 100 .94 (.03) .95 (.04) .95 (.04) .93 (.08)

600 .94 (.04) .95 (.02) .94 (.04) .92 (.08)

1500 .95 (.03) .95 (.04) .96 (.02) .93 (.08)

Global switch costs Pooled .00 (.01) .00 (.02) .00 (.03) -.01 (.06)

Local switchcosts Pooled .03 (.02) .03 (.02) .02 (.02) .03 (.03)

112


The present results with respect to age-related differences in switch-costs seem

at odds with the bulk of existing literature (see also Kray, 2005). Significant larger

global switch-costs are mostly found for older adults than for young adults. The

analysis of the present study, incorporating three older groups, deviates from most

studies which contrast young and old adults. Indeed if subject from the three older

groups are pooled and contrasted to the youngest group the effect of age-group is

significant (F(1,76)=5.4, p<.05).

Table 4.13: Results from pairwise comparisons between age groups on relevant task-variables

using Tukey HSD method. P values are labeled as follows: *<.05; **<.01; ***<.001,and

borderline significant comparisons (p-values below .060 are reported explicitly.

Relationships between tasks

Prior to structural equation analysis, a principal components analysis (PCA) was

conducted on measures of all 11 tasks that were administered in this study (see

table 4.13). The choice of measures per task was determined by sensitivity to aging,

and if no age-related differentiation between different measures of the task were

apparent, an aggregate performance score was used instead. For the switch task

Main Effect 20-30 vs. 50-60 vs. 60-70 vs.

50-60 60-70 70+ 60-70 70+ 70+

TMT B-A *** *** **

DSST *** ** *** *** .056

raven SPM *** ** *** ***

Zoomap ** **

SART *

adaptation factor tracking *** *** ** ***

div att costs tracking ** * **

n-back * **

MODS-R .058

Cue task ** ** *

local switch costs * **

chapteR 4

113


the included measure was mean local switch cost in RT, which showed stronger

age-related differences than global switch costs. For the forward backward span

the mean score on both conditions was used, because of a lack of differentiation

in age effects. The difference between time to complete version B and A of

the trailmaking test were used. From the MODS-R the overall mean accuracy

performance was included as measure. The overall profile score of the Zoomap

was included as indicator for planning. Suppression of prepotent processes was

reflected by accuracy on the PAC-task in the anticue condition. Finally, dual task

costs on the tracking task were used as an indicator of divided attention.

The purpose of PCA is to reduce large sets of data into a few meaningful components

based on the intercorrelations among the items. The rule of eigenvalues greater

than one (Johnsons & Wichern, 1992) was used to determine which components

to retain. The Varimax rotation method using generalized least squares was utilized

to determine the best factor solution. This method maximizes the variances of the

factors without changing the mathematical properties of the solution (Tabachnick

& Fidell, 1996). Sample size restricted the number of variables that could be

subjected to the PCA. Using the rule of thumb of a minimum of 10 participants for

each variable subjected to the component analysis (Bryant & Arnold, 1995), the

current data set allowed for only 8 variables to be used in the PCA. Variables to be

eliminated from the PCA were determined by examining the communalities that

indicate the amount of variance in each variable that is accounted for.

114


Table 4.14: Communalities of the measures in the initial PCA.

Measure Sumofsq/communalities

Local switch costs .436

Forward-backward span .469

Trailmaking B-A .510

MODS-R .523

Zoomap .553

Anti-cue .584

DSST .662

SART .675

n-back .697

Raven .700

Divided attention costs .757

As seen in table 4.14 the three variables in which the least variance in accounted for

(all three below .52) were the local switch costs, the forward-backward span and the

trailmaking measure. The three measures were eliminated and the remaining 8 measures

were subjected to PCA. The resulting factor structure is shown in table 4.15. Eigenvalues

for the two factor solution were 3.3 and 1.2 and accounted for 56.3% of the variance.

All items loaded on at least one of the two factors with loading coefficients ranging from

.41 to .72. The first factor, accounting for 29.7% of the variance, was comprised of the

performance measures of the N-back task, the MODS-R task, SART and the Raven SPM.

The measures comprising this factor, share the need for keeping information and task-

goals active in working memory while processing other information. The second factor

was comprised of performance measures of the anti-cue task, Zoomap planning task,

divided attention task and de digit-symbol substitution task and accounted for 26.6% of

the variance. These measures with the arguable exception of the DSST, are conceptually

strongly related to executive function.

The loading coefficients of both DSST and the Raven test showed lack of differentiation

between loading coefficients on the two factors. Interestingly, both measures have been

chapteR 4

115


used as indicators of general aspects of cognition (respectively general perceptual speed

and general fluid intelligence) by proponents of general factor accounts of cognitive

aging (see also Rabbit, 1997).

Table 4.15 Factor loading coefficients from the principal component analysis (PCA).

Measure Factor 1 Factor 2

n-back 0.838 0.122

MODS-R 0.738 0.185

SART 0.664 0.019

Raven 0.579 0.504

anti-cue 0.081 0.786

Zoomap 0.252 0.641

Divided attention costs 0.049 -0.641

DSST 0.526 0.623

Structural Equation Modeling

The role of age in the relation between the two factors was evaluated by structural

equation modeling. Structural equation analyses were conducted with EQS. A number

of fit statistics were considered when evaluating model fit, including the chi-square test,

which assesses discrepancy between the model and the data, and the nonnormed fit

index (NNFI), comparative fit index, and root-mean-square error of approximation

(RMSEA). The CFI and the NNFI statistics depict improvement compared to an

alternative model, consequently, values closer to 1 correspond to a better fit. The RMSEA

is a measure of discrepancy between predicted and observed values, therefore, values

closer to 0 correspond to better fit.

Structural equation analysis was performed to investigate the role of age within the factor

structure of cognitive performance measures. Four models were examined with respect

to their relative accuracy in fitting the data and are presented in figure 4.5. Three of

these models postulate the two latent factors as described above and possible relations

between these two factors and age. The fourth model postulates the relation between

age and the measures mediated by only one factor.

116


Figure 4.5: Structural equation models. See text for fit-parameters. SART=sustained attention to response

task, mods-r =modified digit working memory span task, SPM=raven standard progressive matrices,

n-back=2-back performance, DSST=digit symbol substitution test, anti-cue=anti-cue condition of PAC-

task, div.att.= divided attention costs on tracking task, zoomap=profile-score on Zoomap test.

SART

n-back

SPM

mods-r

DSST

zooma p

div .att.

ant icue

EF

WM

AGE

.41 *

.62

.73 *

.54 *

.83

.56 *

-.33 *

.53 *

-.77*

model 2

.96 *

SART

n-back

SPM

mods-r

DSST

zooma p

div .att.

ant icue

EF

WM

AGE

.48 *

.77

.71 *

.59 *

.80

.54 *

-.38 *

.57 *

-.62*

-.82*

model 1

chapteR 4

117


SART

n-back

SPM

mods-r

DSST

zooma p

div .att.

ant icue

EF

WM

AGE

.45 *

.72

.75 *

.60 *

.83

.55 *

-.34 *

.55 *

-.80*

model 3

.81*

SART

n-back

SPM

mods-r

DSST

zooma p

div.att.

ant icue

Gen. F AGE

.54 *

.61*

.72 *

.54 *

.82

.55 *

-.32 *

.53 *

-.71*

model 4

118


Model 1 assumes significant effects of age on the executive functions (EF) and

the WM factors while these factors are independent. In other words, in model 1

effects of age on measures of WM are not mediated by the EF factor, and effects on

measure of EF are independent of WM. Model 2 and model 3 assumes dependency

between the two factors. In model 2, WM is assumed to be directly affected by

age. Further more in model 2, the relationship between measures of EF and age is

mediated by the WM factor. In model 3, in contrast, EF is directly modeled to be

directly affected by age, and the relationship between measures of WM and age is

mediated by the EF factor. Model 4 postulates the relation between all measures

and age to be mediated by one general factor.

Table 4.16: Factor loading coefficients from the principal component analysis (PCA).

Model χ² df p CFI NNFI RMSEA

Model 1 41.44 26 <.05 .92 .89 .09

Model 2 37.56 26 <.10 .94 .91 .08

Model 3 29.28 26 <.50 .98 .98 .04

Model 4 37.99 27 <.10 .94 .92 .07

Table 14 shows the fit indices of the models. It shows that model 1, that assumed

independent effects of age (model 1) on the two factors, fitted worse than model

2 and model 3 that both postulated the relations between age and one factor to

be mediated by the other factor. The model that fits the data best was model 3

which assumed the effects of age on WM measures to be completely mediated

by the EF factor. Model 3 can be contrasted with model 2 in which age effects on

EF measures were postulated to be mediated by the WM factor. Model 2 proved

fit the data worse than model 3. Although model 4, assuming one general factor

mediating effects of age on all measures, provided a good fit relative to model 2, it

provided a worse fit than model 3.

Two measures, the Raven score and the DSST score, differentiated only minimally

between the two factors. We therefore conducted a more detailed inspection

on these two measures. A consequence of the lack of differentiation in loading

coefficients on the two factors, is that it should be rather arbitrary on which of

chapteR 4

119


the two factors the Raven and DSST are postulated to load. Also, measures of

fluid intelligence (Raven) and DSST have been found to explain much variance of

performance on other cognitive tasks (e.g. Salhouse, 1993; Rabbitt, 1997). It is

possible that, in the current study, these characteristics of the Raven and the DSST

had a dominant influence on the superiority of model 3 over the other models in

terms of fit indices. To investigate the robustness of the results of the Raven en

the DSST measures were exchanged in the SEM. Although this resulted in a slight

worsening of the fit indices, the relative superiority of model 3 over the other

models remained. This result indicates that the direction of the effects of age in this

study (as evidenced by a better fit of model 3 than model 2) can be attributed to

the other measures of EF and WM measures that the two factors are comprised of.

4.4 General Discusion

Patterns of aging across tasks

Executive control of cognition and aging

Based on the results of the structural equation modeling some concluding remarks

can be made. Factor analysis, applied prior to structural equation modeling,

resulted in the identification of two factors. One factor was comprised of three

measures reflecting executive control function; dual-task costs reflecting the ability

to divide attention between two tasks, anticue performance reflecting the ability

to consistently suppress prepotent responses and the Zoomap score, reflecting

planning behavior. These three measures have in common that all are indices of

executive or frontal functions. The fourth measure that loads highly on this factor

was performance on the digit-symbol-substitution test (DSST). The interpretation

of the DSST is not straightforward. It has been frequently used as measure for

perceptual speed (Salthouse 1993). Parkin and Java (1999, 2000) argue that the

performance on DSST is not exclusively related to perceptual speed, and that

the resulting measure is confounded with cognition such as memory-ability and

intelligence. Thus, from a conceptual perspective, the interpretation of performance

on DSST is not straightforward. Also, in the current study, the DSST shows low

differentiation in loading coefficients between the two factors found in the current

study. Since the other three measures loading highly on the factor, on which the

DSST loads highest, are more clearly related to executive control functions, the

factor was labeled as reflecting executive function (EF).

120


The second factor consisted of two working memory measures (performance on

n-back and MODS-R), a measure for fluid intelligence (Raven SPM) and a measure

for sustained attention (SART). The commonality between these measures is to keep

information active in working memory while processing other information. Hence

this factor was labeled working memory (WM). Support for a WM interpretation

of the Raven comes from Carpenter, ust and Shell (1990), who analyzed individual

differences in performance on the Raven and used computational modeling to

pinpoint the processes that distinguish among individuals. They found that

individual differences in performance on the Raven were largely due to differences

in the ability to manage a large set of problem-solving goals in working memory.

These results can be compared to the factor structure of executive functions

that Myiake et al. (2000) postulated by means of confirmatory factor analysis.

Myiake et al. (2000) postulated three factors representing the executive functions

of shifting between tasks or mental sets, updating and monitoring of working

memory representations, and inhibition of dominant or prepotent responses.

The WM factor in the current study is quite similar to the updating ability of the

Myiake study. The EF factor is partly compatible with the inhibition ability and

partly compatible with the shifting ability of Myiake et al. (2000). They found the

inhibition ability to play an important role in solving the Tower of Hanoi puzzle

(which essentially is a planning task as is the zoomap). Moreover the anti-cue

measure was included in the current study to measure the concept of inhibition

of prepotent responses in the way Myiake and colleagues defined it and is similar

to the version of the anti-saccade task they used as one of the indicators for factor

inhibition. Furthermore, executive control by dividing attention between several

tasks is related to the shifting concept of Myiake et al. (2000). To summarize, the

WM factor in the current study is quite similar to the updating factor postulated

by Myiake et al. (2000) and the factor labeled executive function in the current

study appears to be compatible with a mixture of the factors shifting and inhibition

found in the Myiake et al. (2000) study.

Using structural equation modeling, four models postulating different relationships

between age and latent factors and cognitive measures were compared. The

pattern of age-effects on performance on cognitive tasks was most accurately

represented by the model that postulated a direct effect of age on the EF factor and

an effect of the EF factor on the WM factor; thus according to this model the age-

effects on WM-tasks performance are mediated strictly by age-effects on executive

functioning. The model was superior to models that postulated other patterns of

chapteR 4

121


age-effects on the two factors. This result is consistent with work that has shown

that the dysfunction of the frontal cortex, characterized by aging, underlies much

of the memory impairment that also is characterized by aging (see e.g. Parkin,

1997).

Moreover the model also provided a better fit to the data than a model that

postulated age-effects to be mediated by only one general factor. These results are

compatible with the results of a study of Span (2002). Using structural equation

modeling Span (2002) tested whether a model without an executive factor would

describe the performance data of the children, adults and older adults in their study

significantly worse than a model in which both a speed factor and an executive

factor were postulated. Span (2002) found that the EF factor was necessary only

to be able to account for older adults’ response latency results on the executive

function tasks and they found that older adults differed from young adults respect

to measure of both the speed and executive factor.

Differentiation between the young and the older groups and between the older groups

From the results of the pairwise comparisons between age groups, sensitivity and

the course of aging of a particular measure can be described in two ways. Firstly,

sensitivity of a measure to aging can be conceptualized by using the youngest

group as a reference and examine whether the second youngest group can be

differentiated from this reference with respect to performance on task. Table 4.13

shows that DSST and the adaptation factor of the tracking task differentiated

between the 50-60 group and the 20-30 group. The underlying mechanisms

that lead to performance differences in the DSST are complicated. It has been

used often as a measure of speed and has been shown to be related to executive

function. The adaptation factor of the tracking task can be seen as a more pure

indicator of perceptual-motor speed.

Secondly, sensitivity to aging can be conceptualized by determining how well

a measure differentiates between the three groups of older adults. The pairwise

comparisons showed no significant differentiations between the 70+ group and

the 60-70 age group, but three measures did significantly differ between the 50-60

group and the 70+ group. These measures were the difference between versions

A and B of the trailmaking test, the raven SPM and, almost significantly, the DSST.

In the context of the Berlin Aging Study, Lindenberger and Baltes (1997) extensively

examined differences between young-old (people aged from 70 to 85 years) and

122


the old-old (people aged older than 85 years). One of the aspects studied was the

difference between fluid mechanics of cognition (marked by perceptual speed,

memory and fluency) and pragmatic intelligence (marked by knowledge) in a

longitudinal sample of the study. Singer et al. (2003) found pragmatic intelligence

to remain relatively stable for participants aged under 95, while a considerable

decline in the markers of fluid mechanics was already found in the young-old.

The aim of the present study was not exploring cognitive aging within the group

of people age 70+, but within the ‘young-old’ group as referred to by Baltes and

colleagues. It is possible to further scrutinize the different patterns of (within the

young-old) age-related sensitivity of the measures in the present study.

Most salient here is the difference between the patterns of age-sensitivity. The

measures related to general fluid intelligence and executive functions, the raven

SPM and the trailmaking, significantly differentiated between the three older

groups, but did not significantly differentiate between the 20-30 group from

the 50-60 group (the youngest of the older groups). DSST, the measure that has

conceptually been related to general speed and to executive functions, showed

differentiation between the 20-30 and 50-60 group and differentiation between

the three older groups. Finally, the measure that was clearly related to perceptual-

motor speed and not related to executive function (adaptation factor of the tracking

task) was found to significantly differentiate only between the youngest group and

the 50-60 group and not between the three older groups. This pattern suggests

that cognitive performance that is highly time-critical is sensitive to aging in the

sense that consequences are already and especially manifest at the population

aged between 50 and 60, while in cognitive performance that is more dependent

on executive functions, age-related decrements are more pronounced later in life.

This pattern is in accordance with Ponds et al. (1988) who found no difference

between young and middle-aged adults in divided attention, while elderly adults

did show a decreased ability to divide attention.

Patterns of aging within tasks

Several statistical approaches were employed in the current study. Indices of

cognitive processes were compared with respect to their pattern of sensitivity

to aging. Furthermore, using structural equation modeling patterns of aging

were examined on two factors, a factor reflecting executive control and a factor

reflecting working memory. This analysis was conducted on a selection of

performance indices of the administered task-battery. It was found that the model

chapteR 4

123


that postulated that age affects executive control directly, and that postulated that

the relation between performance on working memory tasks and aging was be

mediated by the executive control factor, provided a better fit than the model that

postulated mediation of age effects in the other direction. Complementary results

of patterns of cognitive aging were found by analysis of performance between and

within tasks.

Speed and executive control

Performance on the DSST showed very clear age-related differences. With structural

equation modeling we showed that the DSST did relate to other performance

measures in the same manner as the Raven SPM test did, supporting Parkin and

Java’s (2000) argument that the DSST performance measure is not best interpreted

as a ‘pure’ measure for perceptual speed and is partly related to intelligence and

to working memory. This argument is in compliance with the suggestion made in

chapter 2 of this thesis (Vogels et al., 2002). In the paper of Vogels et al. (2002)

reaction times on the background task of a prospective memory task were argued

not to reflect “basic mental speed”, but rather to reflect the capacity or ability to

keep various relevant stimulus-response associations activated so that the requisite

response can be quickly and effectively triggered by an imperative stimulus. This

argument led Vogels et al. (2002) to suggest that correlations between prospective

memory performance and speed of reaction-time performance are mediated by

individual differences in the capacity or ability to keep various relevant stimulus-

response associations activated rather than by individual differences in basic mental

speed.

Central cue condition in cue-task

The inclusion of a central cue condition in the pro- vs. anticue paradigm provided

additional insight in the relation between aging and performance decrements

on the anticue condition at long SOA’s compared to performance in the procue

condition. Performance differences between the pro- and anti-cue conditions can

be due to difficulty with suppressing the prepotent response to direct attention

to the cue, or to a failure to endogenously initiate direction of attention to the

location opposite to the cue. The inclusion of the central (symbolic) cue condition

made it possible to distinguish between the relative contribution of the two

potential causes of the decrement of performance in the anti-cue condition. Since

no difference in performance at long SOA’s between the pro-cue and central cue

condition was found, we concluded that the difference at long SOA between the

anticue and procue condition could be attributed to difficulty with suppressing

124


the prepotent response. These results suggest that the central-cue condition is a

promising extension to the pro- vs anticue paradigm to be utilized in future studies

on suppression of prepotent responses.

Working memory and interference

To investigate age-related differences in working memory, two span tasks (forward

and backward span test and MODS-R task) and an n-back task were used. Both span

paradigms consisted of conditions that are commonly believed to differentiate between

the demands that were placed on executive control of memory. Effects of age group

on overall performance on these span tasks were found to be non-significant. On the

n-back, performance differences between age groups were significant. The 1-back and

2-back condition of the n-back working memory task differs from the span working

memory task. The information that is to be kept activated in working memory changes

on every trial in these two conditions of the n-back task. Essentially, the information that

is relevant one trial, is irrelevant on the next. In the span tasks, on every trial relevant

information to be remembered is added to working memory (nothing that was to be

remembered first needs to be intentionally ‘forgotten’ later) until it is has been recalled

at the end of the trial. Thus, in span tasks, information is only to be added to working

memory, while in the 1-back and 2-back conditions of the n-back task information is

to be added to working memory and information is to be eliminated at the same time.

Arguably, this difference between the span tasks and the n-back task could be a reason

for absence of a significant age-group effect on performance on the span tasks while

on the n-back task this effect did reach significance. From this perspective we should

expect an interaction between condition on the n-back task and age-group, because

the 0-back condition does not imply the dynamic demands that are implied by the 1-

and 2-back conditions. Though the data show a trend in this direction, the interaction

was not significant. The line of reasoning is compliant, however, with results from a

study by Keys et al. (2002) who found the 2-back condition of the n-back task to be

most sensitive to aging when compared to the other conditions of the n-back task and

compared to other working memory task.

In the MODS-R task, irrelevant information also needs to be suppressed. The difference

between the irrelevant information of the MODS-R task and the n-back task though,

is that the irrelevant information in the n-back task was relevant when presented

(encoded), while irrelevant information in the MODS-R task needed was always

irrelevant, also at the moment of presentation. This line of reasoning would suggest

that older adults have difficulty with keeping an active representation of task-relevant

information while needing to overcome or prevent the interference of information that

is irrelevant now, but was encoded and relevant (and part of the task-set) some time

chapteR 4

125


earlier in the task. The modeling and empirical work by Braver and colleagues (Braver

et al., 2001; Braver and Barch, 2002) on cognitive control and aging postulates the

same kind of relationship between working memory and inhibition or suppression

of prepotent responses explicitly with respect results of performance on the AX-CPT

task. Their theory postulates that cognitive aging is characterized by impairment in

context processing. The present results are in compliance with the theory of Braver

et al. (2001).

Variability

The most salient outcome of the trailmaking test was the greater spread of performance

in the older groups than in the young group in the B-version, while this difference in

spread of performance was not apparent in the A-version. Similarly, larger individual

differences for the older groups than for the young group were found in the Raven test,

the MODS-R working memory tasks, the switch task and the tracking task in the dual-

task paradigm. An implication of these results is that when studying cognitive aging,

limiting examination to group means of performance over tasks does not suffice to gain

a complete picture of cognitive aging. Apparently, much information lies in individual

differences (within group variability) and within task variability. This outcome is in line

with findings of Hultsch, Mc Donald and Dixon (2002) who found that measures of

intra-individual variability across tasks (dispersion) and across time (inconsistency) as

well as variability between individuals (diversity) were greater in older compared with

younger participants even when group differences in speed were statistically controlled.

Hultsch et al. (2002) concluded that variability of performance is an important indicator

of cognitive functioning and aging. With respect to diversity, several other studies have

also found evidence for multiple qualitatively different cognitive profiles of aging (e.g.

Ylikoski et al., 1999). Ylikosky et al. (1999) identified several subgroups, among a sample

of normal-aged, neurologically healthy adults, which were differentiated by the level

of performance on cognitive tasks. Some of the subjects of their oldest group fell in

the cluster of highest performance. The findings of Ylikosky et al. (1999) also caution

against treating samples of elderly individuals as homogeneous. Gunstad (2006) also

found differential patterns of cognitive aging. The variability of results within age groups

and between age groups has theoretical implications regarding the process of normal

cognitive aging. In the current study it was shown that patterns of effects of aging are

differential with respect to different cognitive functions.

126

chapteR 4

127


Chapter 5 Summary and general discussion

128

The global speed hypothesis of aging states that age-related cognitive decline can be at-

tributed to a decrease in the speed with which elementary cognitive operations are carried

out. The decrease in information processing speed places limits on the level of performance

that can be reached on most cognitive tasks (Eearles, Connor, Smith & Park, 1997; Salthouse,

1996). This hypothesis can be contrasted to hypotheses that attribute age-related effects on

performance on executive control tasks to specific control deficits.

Cognitive control refers to organizing and monitoring cognitive processes. It is crucial for ma-

king behavior adaptive and controlling behavior according to intentions and internal goals.

The concept of cognitive control implies endogenous control of processes and, in that sense,

can be contrasted with situations in which behavior is cued, triggered or prompted explicitly

from or by the environment. Cognitive control has been conceptualized as a unitary process

that is the same in the different situations. It has also been conceptualized as a composite

of fractionated control processes. In case of the second conceptualization, cognitive control

processes in one situation can be different and distinct from processes in another situation.

SummaryChapter 1 reviews some of the theories that posit that age-related differences can be accoun-

ted for solely by one general aspect of cognition (e.g. Birren, 1956; Cerella, 1985; Eearles

et al., 1997; Salthouse, 1996), and other theories that hold that effects of age on cognition

are more specifically limited to executive control functions (e.g. Bryan & Luszcz, 2001; Mayr,

Spieler & Kliegl, 2001; Rabbitt, 1997; Wecker, Kramer, Wisniewski, Delis & Kaplan, 2000).

Also, some of the most influential models of cognitive control are reviewed in chapter 1. In

chapter 2 and 3 studies on cognitive aging are reported focussing on a specific function of

cognitive control.

In chapter 2, a study is presented in which the sensitivity and reliability of four different pros-

pective memory tasks for assessing effects of normal aging were tested, using a group of older

and a group of younger adults. From the perspective of the frontal lobe hypothesis of cognitive

aging, prospective memory tasks are particularly interesting and relevant, because performance

on prospective memory tasks requires both planning and intention-activation (keeping a pros-

pective intention activated during performance on another task). These two functions are gene-

rally believed to involve the frontal lobes. Thus, according to the frontal lobe hypothesis of cog-

nitive aging, robust age-related performance differences on prospective memory tasks should

be expected. The tasks were differentiated on various dimensions such as perceptual saliency

of prospective target events, frequency of occurrence of prospective targets, complexity of

prospective-memory instructions, and provision of feedback after prospective-memory errors.

The role of goal maintenance (or maintaining prospective intentions) and basic mental speed

as mediators for age effects on prospective memory performance were discussed.

chapteR 5

129

Summary and general discussion

In three of the four prospective memory tasks some participants failed to comply with the

task instructions and showed little to no responses to the prospective cues, sometimes toge-

ther with failures to reproduce the relevant instructions during task debriefing. Some of the

‘prospective memory non-performers’ performed well within normal range on the associated

background tasks. This suggests some degree of specificity of their prospective memory pro-

blems in these tasks. Other ‘prospective memory non-performers’ performed also very poorly

with respect to accuracy in background task performance, suggesting a profound misun-

derstanding of task instructions. All but one non-performers were from the older group of

participants. Thus, complete failures to follow prospective memory instructions were limited

almost entirely to the group of older adults.

Two tasks were found to provide sensitive and reliable tools for assessing effects of normal

aging on prospective memory abilities. Both tasks, in different ways, focussed on the partici-

pant’s ability to maintain prospective intentions properly activated and accessible. Consistent

with this perspective, prospective memory accuracy in the two tasks was substantially cor-

related. This correlation was mediated largely by age.

In chapter 3 effects of age on task-switching were reported. Consistent with previous research

local switch costs were larger for old than for young adults, but not disproportionally so. Also

consistent with the existing literature, global switch-costs were disproportional larger for old

than for young adults (see e.g. Kray, 2005). Specifically, the study reported in chapter 3 was

aimed at whether or not residual switch-costs, which are switch-costs that still exist when

ample time to prepare in advance is provided, in old and young adults were attributable to

the same factors. At the level of mean reaction times (RT’s) no significant age-related differen-

ces were found in residual switch-costs. But in the present study, residual switch-costs were

examined in a more detailed way, using analysis of RT distributions. The distributional analysis

revealed residual switch-costs in younger adults to be attributable completely to inconsisten-

cies of engaging in advance preparation, replicating results reported by De Jong (2001) and

by Nieuwenhuis and Monsell (2002). Moreover, the older and younger groups did not signi-

ficantly differ with respect to advance preparation. For the group of old adults, though, this

failure to engage (FTE) account of residual switch-costs did not suffice. Evidently, engaging

in advance preparation did, for old adults, not suffice to attain similar levels of preparation

on switch trials as on no-switch trials. Thus, for older adults extra time after the onset of the

imperative stimulus was needed. A possible explanation for the time it took to respond (even

when engaged in advance preparation during PI) for old adults in the present study is that is

could reflect a decrement of quality of retrieval of the task set information from LTM, which

would be consistent with the “LTM retrieval” interpretation of local switch costs proposed by

Mayr and Kliegl (2000). Mayr and Kliegl (2000) hypothesized that local switch-costs reflect a

process of actively retrieving task-set related information from LTM.

130

chapteR 5

Related to the task switching study is research studying prospective memory, which typi-

cally measures whether a retrieval of an intention can be maintained in the face of high

interference situations. Several studies have dealt with aging and prospective memory, and

old adults are often found in these studies to show worse performance than young adults

under conditions of relatively low environmental support and relatively high interference

(Einstein, Smith, McDaniel and Shaw, 1997; Maylor, 1996; Vogels, Dekker, Brouwer and De

Jong, 2002, chapter 2 this thesis).

In Chapter 4, a more general approach was taken to study the relations between cognitive

control functions and aging. Firstly, instead of focusing on one cognitive control function,

several aspects of cognition were taken into account. Speed of processing, intelligence and

several cognitive control functions were measured by means of different cognitive tasks. Se-

condly, another approach was taken than in the former two chapters with respect to the age

cohorts. Instead of two age-groups, one older group and one younger group, participants

were recruited from four age-groups, one young group (ages between 20 and 30 years) and

three older (from 50 years old) groups. Factor analysis, applied prior to structural equation

modeling, resulted in the identification of two factors which were comprised of variables

indicating, respectively, executive (control) functions and working memory. These results can

be compared to the factor structure of executive functions that Myiake et al. (2000) postu-

lated by means of confirmatory factor analysis. Myiake et al. (2000) postulated three factors

representing the executive functions of shifting between tasks or mental sets, updating and

monitoring of working memory representations, and inhibition of dominant or prepotent

responses. The WM factor in the current study is quite similar to the updating factor postu-

lated by Myiake et al. (2000) and the factor labeled executive function in the current study

appears to be compatible with a mixture of the factors shifting and inhibition found in the

Myiake et al. (2000) study.

Using structural equation modeling, four models postulating different relationships between

age and latent factors and cognitive measures were compared. The pattern of age-effects on

performance on cognitive tasks was most accurately represented by the model that postu-

lated a direct effect of age on the EF factor and an effect of the EF factor on the WM factor;

thus according to this model the age-effects on WM-tasks performance are mediated strictly

by age-effects on executive functioning.

These results are compatible with the results of a study of Span (2002). Using structural equa-

tion modeling Span (2002) tested whether a model without an executive factor would des-

cribe the performance data of the children, adults and older adults in their study significantly

worse than a model in which both a speed factor and an executive factor were postulated.

Span (2002) found that the EF factor was necessary only to account for older adults’ response

131


latency data of the executive function tasks and that older adults differed from young adults

on both the speed and executive factor.

Comparing the age groups revealed different patterns of age sensitivity of the respective

cognitive variables. These patterns suggest that cognitive performance that is highly depen-

dent on speed is sensitive to aging in the sense that consequences are already manifest at the

population aged between 50 and 60, while in cognitive performance that is more dependent

on executive functions, age-related decrements are more pronounced later in life.

Validity of speed indices

Concerning the issue of the relation between cognitive aging and speed of information pro-

cessing, the studies reported in this thesis contribute by repeatedly finding that, at least for

some purported and often used, measures of speed it is evident that it is not a pure measure

of speed and that performance on such tests are not independent on executive function.

This, for example concerns the DSST. In chapter 4 performance on the DSST showed very

clear age-related differences. With structural equation modeling we showed that the DSST

did relate to other performance measures in the same manner as the Raven SPM test did, sup-

porting Parkin and Java’s (2000) argument that the DSST performance measure should not

be interpreted as a ‘pure’ measure for perceptual speed and is partly related to intelligence

and to working memory. This argument is in agreement with the suggestion made in chapter

2 of this thesis (Vogels et al., 2002). In Vogels et al. (2002) reaction times on the background

task of a prospective memory task were argued not to reflect “basic mental speed”, but ra-

ther to reflect the capacity or ability to keep various relevant stimulus-response associations

activated so that the requisite response can be quickly and effectively triggered by an impe-

rative stimulus. This argument led Vogels et al. (2002) to suggest that correlations between

prospective memory performance and speed of reaction-time performance are mediated

by individual differences in the capacity or ability to keep various relevant stimulus-response

associations activated rather than by individual differences in basic mental speed. Thus, in

general, some of the measures that are often used as indicators for speed of information

processing or perceptual speed are also dependent on cognitive control processes. This impli-

cates a nuance in confirming a global speed hypothesis of on basis of age-related differences

in performance on these tasks.

Heterogeneity

One of the salient outcomes of the three studies and previous findings by others (e.g. Ylikos-

ky, 1999, Gunstad, 2006) concerns the heterogeneity of the cognitive performance decline

with aging. As reviewed above, the study in chapter 2 on prospective memory showed that,

although as a group the older adults performed worse than the young adults with respect

to prospective memory, some older subjects showed performance similar to the young sub-

132

chapteR 5

jects, while some older participants performance much worse on the prospective memory

indicators. With respect to task-switching, as evidenced in chapter 3, a related observation

was made. Although the distribution functions of RT’s and the interpretation of the distri-

bution modeling were straightforward when pooled over groups, it was stressed that there

were notable individual differences especially in the older age-group. Finally, in chapter 4,

several instances were reported of heterogeneous nature of effects of aging on performance

on cognitive tasks. The most salient outcome of the trailmaking test was the greater spread

of performance in the older groups than in the young group in the B-version, while this

difference in spread of performance was not apparent in the A-version. Larger individual

differences for the older groups than for the young group were also found in the Raven

standard progressive matrices test, the MODS-R working memory tasks, the switch task and

the dual-task paradigm.

These results concerning heterogeneity suggest that when studying cognitive aging, limiting

examination of group means of performance over tasks does not suffice. Much information

lies in individual differences (within group variability) and within task variability (Gunstad et

al., 2006). The theoretical implications of these results concern the universality of cognitive

aging. This argument has been eloquently made by Rabbitt, Lowe and Shilling (2001). They

argue that one of the most common causes in the central nervous system in old age are ce-

rebrovascular accidents or insufficiencies of cerebrovascular circulation. These changes occur

more often with increasing age and more often in the frontal lobes than in other areas of the

brain (Shaw et al., 1984). Consequently, Rabbitt et al. (2001) argue, the increasing incidence

of focal brain changes, particularly in the frontal lobes, and the resulting local deficits, is more

likely to underlie effects of aging on cognition than a continuous and progressive unfolding

of particular patterns of deficits in all, or most members of a population.

In addition to the theoretical implications of the heterogeneous nature of effects of age on

cognition, there are also more practical implications. For example, in light of the labor market

and the rising retirement ages and low birth rates, a practical implication of the heterogen-

eity of effects of aging on cognitive performance is that the age of an individual solely is not

a reliable indicator of (worsened) quality of work potential. A major source of employees is

overlooked when policy of this subject is based on the age of individuals.

133


134

135

Bibliography

Bibliography

136

Allport, A., Wyllie, G. (2000). Task switching: Stimulus-response bindings, and negative priming. In S. Monsell & J. Driver (Eds.), Control of cognitive processes: Attention and performance, XVIII. Cambridge, MA: MIT Press.

Allport, D. A., Styles, E. A., & Hsieh, S. (1994). Shifting intentional set: Exploring the dynamic control of tasks. In C.Umilta & M. Moscovitch (Eds.), Attention and performance XIX: Conscious and nonconscious information processing.Cambridge, MA, USA: The Mit Press.

Anderson, S.W., Damasio, H., Dallas Jones, R., Tranel, D (1991). Wisconsin card sorting test performance as a measure of frontal lobe damage. Journal of clinical and experimental neuropsychology, 13, 909-922.

Babcock, R.L., Salthouse, T.A. (1990). Effects of increased processing demands on age differences in working memory. Psychology and aging, 5, 421-428.

Backman, L., Ginovart, N., Dixon, R.A. (2000). Age related cognitive deficits medicated by changes in the striatal dopamine system. American Journal of Psychiatry, 157, 635-637.

Baddeley, A.D. & Hitch, G.J. (1974). Working memory. In G.A. Bower (ed.), Recent Advances inLearning and Motivation, Vol. 8, 47–89. New York: Academic Press.

Baddeley, A.D. (1996). Exploring the Central Executive. Quarterly Journal of Experimental Psychol-ogy, 49A(1), 5–28.

Baddeley, A.D. (1998). Recent development in working memory. Current Opinion in Neurobiology, 8:234–238

Baddeley, A.D. (2001). Comment on Cowan: The magic numberand the episodic buffer. Behavioral and Brain Sciences, 24, 117–118.

Baddeley, A.D. (1986). Working memory. Oxford, England: Oxford University Press.

Baddeley, A., Della Sala, S., Gray, C., Papagno, C., & Spinnler, H. (1997). Testing central executive functioning with a pencil-and-paper test. In P. Rabbitt (Ed.), Methodology of frontal and executive function, 61–80, Hove: Erlbaum.

Band, G.P.H., Ridderinkhof, K.R., & Segalowitz, S. (2002). Explaining neurocognitive aging: Is one factor enough? Brain and Cognition, 49 (3), 259-267

Barch, D. M., Braver, T. S., Racine,C., & Satpute, A. B. (2001). Cognitivecontrol deficits in healthy aging: Neuroimaging investigations. Neuro-Image, 13, SI025

Bashore, T.R., Smulders, F.T.Y., 1995. Do general slowing functions mask local slowing effects? Achronopsychophysiological perspective. In: Allen, P., Bashore, T. (Eds.), Age Differences in Wordand Language Processing. Elsevier, Amsterdam, pp. 390–425

Battig, W.F., & Montague, W.E. (1969). Category norms for verbal items in 56 categories: A replication and extension of the Connecticut category norms. Journal of Experimental Psychology: Monograph, 80, 1-46.

Berardi, A., Parasuraman, R., & Haxby, J. V. (2001). Overall vigilance andsustained attention decrements in healthy aging. Experimental Aging Research, 27, 19–39

137

Bibliography

Birren, J. E. (1956). The significance of age changes in speed of perception and psychomotor skills. In J. E. Anderson (Ed.), Psychological aspects of aging, 97-104. Washington, DC: American Psychological Association

Bopp, K. L., & Verhaeghen, P. (2005). Agingand verbal memory span: A meta-analysis.Journal of Gerontology: PsychologicalSciences, 60B, P223–P233

Botwinick, J., Brinley, J. F., & Robbin, J. S. (1958). Task alternation time in relation to problem difficulty and age. Journal of Gerontology, 13,414-417

Brandimonte, M., Einstein, G.O., & McDaniel, M.A. (Eds.). (1996). Prospective memory: Theory and applications. Hillsdale, NJ: Erlbaum.

Braver, T. S., & Barch, D. M. (2002). A theory of cognitive control, aging cognition, and neuromodulation. Neuroscience and Biobehavioral Reviews, 26, 809–817

Braver, T. S., Reynolds, J. R., & Donaldson, D. I. (2004). Neuralmechanisms of transient and sustained cognitive control duringtask switching. Neuron, 39, 713–726

Braver, T.S., Barch, D.M., Keys, B.A., Carter, C.S., Cohen, J.D., Kaye, J.A., Janowsky, J.S., Taylor, S.F., Yesavage, J.A., Mumenthaler, M.S., Jagust, W.J., Reed, B.R. (2001). Context processing in older adults: Evidence for a theory relating cognitive control to neurobiology in healthy aging. Journal of Experimental Psychology: General, 130, 746-763.

Broadbent DE, Cooper PF, FitzGerald P, Parkes KR. (1982) The CognitiveFailures Questionnaire (CFQ) and its correlates. British Journal Clinical Psychology, 21, 1–16

Brouwer, W. & Fasotti, L. (1997). Planning en regulatie. In: Deelman et al (red.). Klinische neuropsychologie. Amsterdam: Uitgeverij Boom

Brouwer, W. H., Waterink, W., Van Wolffelaar, P. C., & Rothengatter, T.(1991). Divided attention in experienced young and older drivers: Lanetracking and visual analysis in a dynamic driving simulator. Human Factors, 33, 573-582

Bryan J.; Luszcz M.A. (2001). Adult Age Differences in Self-Ordered Pointing Task Performance: Contributions From Working Memory, Executive Function and Speed of Information Processing. Journal of Clinical and Experimental Neuropsychology, 23, 608-619

Bryan, J., & Luszcz, M. A. (1996). Speed of information processing as a mediator between age and free recall performance. Psychology and Aging, 11, 3-9

Burgess, P. W., Alderman, N., Evans, J., Emslie, H., and Wilson,B. A. 1998. The ecological validity of tests of executive function. Journal of International Neuropsychological. Society, 4, 547–558

Burgess, P.W., & Shallice, T. (1997). The relationship between prospective and retrospective memory: Neuropsychological evidence. In M.A. Conway (Ed.), Cognitive models of memory. Studies in cognition (pp. 247-272). Cambridge, MA: Mit Press.

Cahn-Weiner, D.A., Boyle, P.A., Malloy, P.F. (2002). Tests of Executive Function Predict Instrumental Activities of Daily Living in Community-Dwelling Older Individuals. Applied Neuropsychology, 9, 187-191.

138

Bibliography

Carpenter, P.A., Just, M.A., Shell, P. (1990). What one intelligence test measures: a theoretical account of the processing in the Raven Progressive Matrices Test. Psychological Review,97, 404-431

Cattell, R. B., & Cattell, A. K. S. (1960). Culture fair intelligence test. Scale 2. Forms A and B. Handbook for the individual or group. Champaign, IL: Institute of Personality and Ability Testing

Cepeda, N. J., Kramer, A. F., & Gonzales de Sather, J. C. M. (2001). Changes in executive control across the life span: Examination of task-switch-ing performance. Developmental Psychology, 37, 715-730.

Cerella, J. (1985). Information processing rates in the elderly. Psychological Bulletin, 98, 67-83

Chao, L. L., & Knight, R. T. (1997). Prefrontal deficits inattention and inhibitory control with aging. Cerebral Cortex, 7, 63–69

Cherry, K.E., & LeCompte, D.C. (1999). Age and individual differences influence prospective memory. Psychology and Aging, 14 (1), 60-76.

Cohen, J. D., Perlstein, W. M., Braver, T. S., Noll, D. C., Jonides, J., & Smith, E. E. (1997). Temporal dynamics of brain activation during a working memory task. Nature, 386, 604-608

Craik, F. I. M., & Grady, C. L. (2002). Aging, memory, and frontal lobefunctioning. In: D. T. Stuss, R. T. Knight (Eds.), Principles of frontallobe function,528–540. London: Oxford University Press.

Craik, F.I.M. (1986). A functional account of age differences in memory. In F. Klix & H. Hagendorf (Eds.), Human memory and cognitive capabilities, mechanisms and performances,409-422. Amsterdam: Elsevier.

Craik, F.I.M., & Kerr, S.A. (1996). Commentary: Prospective memory, aging, and lapses of intention. In M. Brandimonte, G.O. Einstein, & M.A. McDaniel (Eds.), Prospective memory: Theory and applications (pp. 227-238). Hillsdale, NJ: Erlbaum.

Cycowicz, Y. M., Friedman, D., Rothstein, M., & Snodgrass, J.G. (1997). Picture naming by young children: Norms for name agreement, familiarity, and visual complexity. Journal of experimental child psychology, 65 (3), 171-237.

Daily, L. Z., Lovett, M. C., & Reder, L. M. (2001). Modeling individualdifferences in working memory performance: A source activation account. Cognitive Science, 25, 315–353

De Graaf, A., & Deelman, B.G. (1991). Cognitieve Screening Test [Cognitive Screening Test]. Lisse: Swets & Zeitlinger.

De Jong, R. (2000). An intention-activation account of residual switch costs. In S. Monsell & J. Driver (Eds.), Control of cognitive processes: Attention and performance, XVIII. Cambridge, MA: MIT Press.

De Jong, R. (2001). Adult age differences in goal activation and goal maintenance. European Journal of Cognitive Psychology, in press.De Jong, R., Berendsen, E., & Cools, R. (1999). Goal neglect and inhibitory limitations:

139

Bibliography

Dissociable causes of interference effects in conflict situations. Acta Psychologica, 101, 379-394.

Dove, A., Pollman, T., Schubert, T., Wiggens, C. J., & Yves von Cramon, D. (2000). Prefrontal cortex activation in task switching: an event-related fMRI study. Cognitive Brain Research, 9, 103-109.

Dreher, J. C., Koechlin, E., Ali, S. O., & Grafman, J. (2002). The roles of timing and task order during task switching. Neuroimage, 17, 95–109.

Duncan, J. (1990). Goal weighting and the choice of behaviour in a complex world. Ergonomics, 33, 1265-1279.

Duncan, J., Burgess, P., & Emslie, H. (1995). Fluid intelligence after frontal lobe lesions. Neuropsychologia, 33, 261–268

Duncan, J. (1986). Disorganisation of behaviour after frontal lobe damage. Cognitive Neuropsychology, 3, 271-290

Duncan, J., Emslie, H., Williams, P., Johnson, R., & Freer, C. (1996). Intelligence and the frontal lobe: The organization of goal-directed behaviour. Cognitive psychology, 30, 257-303.

Earles JL, Connor LT, Smith AD, Park DC. (1997) Interrelations of age,self-reported health, speed, and memory. Psychology Aging, 12, 675–83.

Einstein, G. O., & McDaniel, M. A. (1996). Retrieval processes inprospective memory: Theoretical approaches and some newempirical findings. In M. Brandimonte, G. O. Einstein, & M. A.McDaniel (Eds.), Prospective memory: Theory and applications, 115-141. Hillsdale, NJ: Erlbaum

Einstein, G. O., Smith, R. E., McDaniel, M. A., & Shaw, P. (1997). Aging and prospective memory: The influence of increased task demands at encoding and retrieval. Psychology and Aging, 12, 479-488.

Einstein, G.O., & McDaniel, M.A. (1990). Normal aging and prospective memory. Journal of Experimental Psychology: Learning, Memory, & Cognition, 16, 717-726.Einstein, G.O., McDaniel, M.A., Richardson, M.A., Guynn, S.L., Cunfer, A.R. (1995). Aging and prospective memory: Examining the influences of self-initiated retrieval processes. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21 (4), 996-1007.

Einstein, G. O., Smith, R. E., McDaniel, M. A., & Shaw, P. (1997).Aging and prospective memory: The influence of increased taskdemands at encoding and retrieval. Psychology and Aging, 12, 479-488

Eysenck, H.J. (1986). The theory of intelligence and the psychophysiology of cognition. Advances in the Psychology of Human Intelligence, 3, 1-34

Filley, C. M., & Cullum, C. M. (1994). Attention and vigilance functionsin normal aging. Applied Neuropsychology, I, 29-32.

Fischer, B., Biscaldi, M., & Gezeck, S. (1997). On the development ofvoluntary and reflexivee components in human saccade generation. Brain Research, 754, 285–297

Fuster JM (1997) The prefrontal cortex: anatomy, physiology, and neu-rophysiology of the

140

Bibliography

frontal lobe, Ed 3. Philadelphia: Lippincott-Raven

Fuster, J.M. (1989). The prefrontal cortex. New York: Raven.

Glisky, E.L. (1996). Prospective memory and the frontal lobes. In M. Brandimonte, G.O. Einstein, & M.A. McDaniel (Eds.), Prospective memory: Theory and application, 249-266. Hillsdale, NJ: Lawrence Erlbaum Associates.

Godefroy O, Cabaret M, Petit-Chenal, V., Pruvo, J., Rousseaux, M. (1999), Control functions of the frontal lobes. Modularity of the central-supervisory system? Cortex 35(1), 1-20

Goldman-Rakic, P. S., & Brown, R. M. (1981). Regional changes ofmonoamines in cerebral cortex and subcortical structures of aging rhesusmonkeys. Neuroscience, 6, 177-187

Gopher, D. (1996). Attention control: Explorations of the work of an executive controller. Cognitive Brain Research, 5, 23-38

Gopher, D., & Kahneman, D. (1971). Individual differences in attention and the prediction of flight criteria. Perceptual and Motor Skills, 33, 1335-1342.

Goschke T. (2000). Intentional reconfiguration and involuntary persistence in task-set switching. In S. Monsell & J. Driver (Eds.), Control of cognitive processes: Attention and performance, XVIII. Cambridge, MA: MIT Press.

Gottsdanker, R. (1980). Aging and the maintaining of preparation. Experimental Aging Research, 6 (1), 13-27.

Greenwood, P. M. (2000). The frontal aging hypothesis evaluated. Journal of the International Society, 6, 705–726Gregoire, J., & Van der Linden, M. (1997). Effect of age on forward andbackward digit spans. Aging, Neuropsychology, and Cognition, 4, 140–149Grigsby, J., Kaye, K., Baxter, J., Shetterly, S. M. & Hamman, R. F. 1998. Executivecognitive abilities and functional status among community dwelling older personsin the San Luis Valley Health and Aging Study. Journal of the American Geriatric Society 46, 590-596

Gunstad J, Paul RH, Brickman AM, Cohen RA, Arns M, Roe D, Lawrence JJ, Gordon E. (2006). Patterns of cognitive performance in middle-aged and older adults: A cluster analytic examination. Journal of Geriatric Psychiatry Neurology. 2006 Jun;19(2):59-64

Hartley, A. A. (1992). Attention. In F. I. M. Craik & T. A. Salthouse (Eds.),The handbook of aging and cognition, 3-49. Hillsdale, NJ: Erl-baum

Hartley, J. T. (1993). Aging and prose memory: Tests of the resource-deficit hypothesis. Psychology & Aging, 8, 538-551

Hartley, A. A., Kieley, J. M., & Slabach, E. H. (1990). Age differences and similarities in the effects of cues and prompts. Journal of Experimental Psychology: Human Perception and Performance, 16, 523-537

Hasher, L., Zacks, R. T. , & May, C. P. (1999). Inhibitory control, circadian arousal, and age. In D. Gopher & A. Koriat (Eds.), Attention and performance XVII,653-675. Cambridge, MA: MIT Press

141

Bibliography

Heathcote, A., Popiel, S. J., & Mewhort, D. J. K. (1991). Analysis of response time distributions: An example using the Stroop task. Psychological Bulletin, 109, 340-347

Hester, R. L., Kinsella, G. J., & Ong, B. (2004). Effect of age on forward and backwardspan tasks. Journal of the International Neuropsychological Society, 10, 475-481

Holland, C. A., & Rabbitt, P. M. A. (1990). Autobiographical andtext recall in the elderly: An investigation of a processing resourcedeficit. Quarterly Journal of Experimental Psychology, 42A, 441-470

Holtzer, R., Stern, Y., & Rakitin, B. C. (2004). Age-related differences in executive control of working memory. Memory & Cognition, 32, 1333– 1345

Horn, J. L., & Cattell, R. B. (1966). Refinement and test of the theory offluid and crystallized intelligence. Journal of Educational Psychology, 57, 253–270

Horn, J.L. & Masunaga, H. (2000). New Directions for research on aging and intelligence: the development of expertise. In T. J. Perfect & E. A. Maylor (Eds.), Models of Cognitive Aging. (pp. 125 -159). Oxford: Oxford University Press

Horn, J.L. (1982) The theory of fluid andcrystallized intelligence in relation to concepts ofaging in adulthood. In: Craik, F.I.M. and Trehub, S., (eds), Aging and CognitiveProcesses, 237–278, Plenum Press

Horn, J.L., & Donaldson, G. (1976). On the myth of intellectual decline in adulthood. American Psychologist, 31, 701-719.

Hudson, P.T.W. (1982). Preliminary category norms for verbal items in 51 categories in Dutch. Internal report, Katholieke Universiteit Nijmegen.

Jensen, A.R. (1987). Process differences and individual differences in some cognitive tasks. Intelligence, II, 107-136

Jensen, A. R. (1985). The nature of the BlackWhite difference on various psychometric tests: Spearman’s hypothesis. Behavioral and Brain Sciences, 8, 193-263

Jonides, J., Schumacher, E.H., Smith, E.E., Lauber, E.J., Awh, E., Misnoshima, S., Koeppe, R.A., 1997.Verbal memory load affects regional brain activation as measured by PET. J. Cognitive Neuroscience. 9,462–475

Just, M. A., & Carpenter, P. A. (1992). A capacity theory of compre-hension: Individual differences in working memory. Psychological Review, 99, 122-149

Carpenter, P. A., Just, M. A., & Shell, P. (1990). What one intelligence testmeasures: A theoretical account of processingin the RavenProgressiveMatrices test. Psychological Review, 97, 404-431

Kane, M. J., & Engle, R. W. (2003). Working-memory capacity and thecontrol of attention: The contributions of goal neglect, response com-petition, and task set to Stroop interference. Journal of Experimental Psychology: General, 132, 47–70

Kane, M. J., Hasher, L., Stoltzfus, E. R., Zacks, R. T., & Connelly, S. L.(1994). Inhibitory attentional mechanisms and aging. Psychology and Aging, 9, 103–112

142

Bibliography

Keys, B. A., & White, D. A. (2000). Exploring the relationship betweenage, executive abilities, and psychomotor speed. Journal of the International Society, 6, 76–82

Kimberg, D. Y., Farah, M. J. (1993). A unified account of cognitive impairments following frontal lobe damage: The role of working memory in complex, organized behavior. Journal of Experimental Psychology: General, 122, 41 1-428

Kliegl, R., Mayr, U., & Krampe, R. (1994). Time–accuracy functions fordetermining process and person differences: An application to cognitiveaging. Cognitive Psychology, 26, 134–164

Kluwe, R.H. (1997). Effects of type of learning on control performance. Proceedings of the International Conference Engineering psychology and cognitive ergonomics, 2, 81-88

Korteling, J. E. (1993). Effects of age and task similarity on dual-taskperformance. Human Factors, 35, 99-113

Kramer, A. F., Humphrey, D. G., Larish, J. F., Logan, G. D., & Strayer, D. L. (1994). Aging andinhibition: beyond a unitary view of inhibitory processing in attention. Psychology and Aging, 9(4),491–512

Kramer, A. F., Hahn, S., & Gopher, D. (1999). Task coordination and aging: Explorations of executive control processes in the task switching paradigm. Acta Psychologica, 101, 339-378.

Kramer, A. F., Larish, J. L., Weber, T. A., & Bardell, L. (1999). Training for executive control: Task coordination strategies and aging. In D.Gopher & A. Koriat (Eds.), Attention and performance XVII: Cognitive regulation of performance: Interaction of theory and application. Attention and performance (xiii ed., pp. 617-652). Cambridge, MA, USA: The Mit Press.

Kramer, A. F., Larish, J. F., & Strayer, D. L. (1995). Training for atten-tional control in dual task settings: A comparison of young and oldadults. Journal of Experimental Psychology: Applied, 1, 50-76

Kray, J. and Lindenberger, U. (2000) Adult age differences in taskswitching. Psychology and Aging, 15, 126–147

Kray, J. (2005). Task-set switching under cue-based versus memory-based switching conditions in younger and older adults. Brain Research, 1105, 83-92.

Kray, J., Li, K. Z. H., & Lindenberger, U. (2002). Age-related changes intask switching components: The role of uncertainty. Brain & Cognition, 49, 363–381

Kuhlman, A., Little, D., Sekuler, R. (2006). An Interactive Test of Serial Behavior: Ageand Practice Alter Executive Function. Journal of Clinical and Experimental Neuropsychology, 28, 126-144.

Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (littlemore than) working-memory capacity?! Intelligence, 14, 389-433

Lehto, J. (1996). Are executive function tests dependent on workingmemory capacity? Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 49(A), 29–50

143

Bibliography

Li S-C, Lindenberger U, Sikström (2001). Aging cognition: fromneuromodulation to representation to cognition. Trends in Cognitive Science, 5, 479–86

Li, S-C. and Lindenberger, U. (1999) Cross-level unification: a computationalexploration of the link between deterioration ofneurotransmitter systems and dedifferentiationof cognitive abilities in old age. In: Nilsson, L,G., Markowitsch, H., (eds), Cognitive Neuroscience of Memory, 103–146, Hogrefe and Huber

Light, L.L. (1991). Memory and aging: Four hypotheses in search of data. Annual Review of Psychology, 42, 333-376.

Lindenberger, U., & Baltes, P. B. (1997). Intellectual functioning in the oldand very old: Cross-sectional results from the Berlin Aging Study.Psychology and Aging, 12, 410–432

Logan, G. D., & Zbrodoff, N. J. (1979). When it helps to be misled:Facilitative effects of increasing the frequency of conflicting stimuli ina Stroop-like task. Memory and Cognition, 7, 166-174

Lovett, M., Reder, L. M., Lebiere, C. (1999) Modeling workingmemory in a unified architecture:An ACT-R Perspective. In Miyake, A. and Shah, P. (Eds). Models of Working Memory. Oxford University Press, pp.135–182

Lovett, M. C., Daily, L. Z., & Reder, L. M. (2000). A source activation theory of working memory: Cross-task prediction of performance in ACT-R. Cognitive Systems Research, 1, 99-118.

Lowe, C., & Rabbitt, P. (1997). Cognitive models of aging and frontal lobe deficits. In P. Rabbitt (Ed.), Methodology of frontal and executive functions,39-59). Hove, UK: Psychology Press

Manly, T., Robertson, I. H., Galloway, M., & Hawkins, K. (1999). The absent mind: further investigations of sustained attention to response. Neuropsychologia, 37, 661-670

Mäntylä, T. (1996). Activating actions and interrupting intentions: Mechanisms of retrieval sensitization in prospective memory. In M.A. Brandimonte, G.O. Einstein, & M.A. McDaniel (Eds.), Prospective memory: Theory and applications, 241-275. Hillsdale, NJ: Erlbaum.

Maylor, E.A. (1990). Age and prospective memory. The Quarterly Journal of Experimental Psychology, 42A, 471-493.

Maylor, E.A. (1996). Age-related impairment in an event-based prospective memory task. Psychology and Aging, 11 (1), 74-78.

Maylor, E.A. (1998). Changes in event-based prospective memory across adulthood. Aging, Neuropsychology, and Cognition, 5 (2), 107-128.

Mayr, U., & Keele, S. W. (2000). Changing internal constraints on action: The role of backwardinhibition. Journal of Experimental Psychology: General, 129, 4–26

Mayr, U., & Kliegl, R. (2000). Task-set switching and long-term memory retrieval. Journal of Experimental Psychology: Learning, Memory, & Cognition, 26, 1124-1140

Mayr, U. (2001). Age differences in selection of mental sets: The role of inhibition, stimulus

144

Bibliography

ambiguity, and response-set overlap. Psychology and Aging, 16(1), 96-109.

Mayr, U., Kliegl, R. (1993). Task-set switching and long-term memory retrieval. Journal of Experimental Psychology: Learning, memory and cognition, 26, 1124-1140.

Mayr, U., Liebscher, T. (2001). Is there an age deficit in the selection of mental sets?. European Journal of Cognitive Psychology, 13 (1/2), 47-69. McDowd, J.M. and Shaw, R.J. (2000) Attentionand aging: a functional perspective. In: Craik, F.I.M., Salthouse, T.A., (eds), The Handbook of Aging and Cognition,221–292, Erlbaum

McDowd, J., Vercruyssen, M., & Birren, J. E. (1991). Aging, divided attention, and dualtask performance. In D. L. Damos (Ed.), Multiple-task performance, 387-414. London: Taylor & Francis

Meiran, N. (1996). Reconfiguration of processing mode prior to task performance. Journal of Experimental Psychology: Learning, Memory and Cognition, 22, 1423-1442.

Meiran, N., Gotler, A., Perlman, A. (2001). Old age is associated with a pattern of relatively intact and relatively impaired task-set switching abilities. Journal of Gerontology: Psychological Sciences, 56B(2), P88-102.

Meuter, R. F. I., & Allport, A. (1999). Bilingual language switching in naming: Asymmetrical costs oflanguage selection. Journal of Memory and Language, 40, 25–40

Mewhort, D. J. K., Braun, J. G., & Heathcote, A. (1992). Responsetime distributions and the Stroop task: A test of the Cohen, Dunbar,and McClelland (1990) model. Journal of Experimental Psychol-ogy: Human Perception & Performance, 18, 872-882

Meyer, D. E. & Kieras, D. E. (1997). A computational theory of executive cognitive processes and multiple-task performance. Psychological Review, 104, 2-65

Monsell, S. (1996). Control of mental processes. In V. Bruce (Ed.), Unsolved mysteries of the mind: Tutorialessays in cognition, 93–148. Hove, UK: Erlbaum (UK) Taylor & Francis Publishers

Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7,134–140

Moscovitch, M., & Winocur, G. (1992). The neuropsychology of memory and aging. In F.I.M. Craik & T.A. Salthouse (Eds.), The handbook of aging and cognition (pp. 315-372). Hilssdale, NJ: Lawrence Erlbaum Associates.

Munoz, D. P.; Broughton, J. R.; Goldring, J. E.; Armstrong, I. T.Age-related performance of human subjects on saccadic eye move-ment tasks. Experimental Brain Research,121, 391–400

Miyake, A., Emerson, M. J., Padilla, F., & Ahn, J. (2004). Inner speech as aretrieval aid for task goals: The effects of cue type and articulatory sup-pression in the random task cuing paradigm. Acta Psychologica, 115,123–142.

Miyake, A., Friedman, N. P., Rettinger, D. A., Shah, P. &Heggarty, M. (2001) How are visuo-spatial working memory,executive functioning, and spatial abilities related: latent-variable analysis. Journal of Experimental Psychology: General, 130, 621–640.

145

Bibliography

Nieuwenhuis, S., Broerse, A., Nielen, M. M. A., & de Jong, R. (2004).A goal activation approach to the study of executive function: Anapplication to antisaccade tasks. Brain and Cognition, 56, 198–214

Nieuwenhuis, S., Ridderinkhof, K. R., Blom, J., Band, G. P. H. & Kok, A. (2001). Error-related potentials are differentially related to awareness of response errors: Evidence from an antisaccade task. Psychophysiology 38, 752-760

Nieuwenhuis, S., and Monsell, S. (2002). Residual costs in task-switching: Testing the failure-to-engage hypothesis. Psychonomic Bulletin & Review, 9(1), 86-92.

Norman, D. A., & Shallice, T. (1986). Attention to action: willed and automatic control of behavior. In:R. J. Davidson, G. E. Schwartz, & D. Shapiro (Eds.), Consciousness and self-regulation: advances inresearch and theory, 4, 1–18, New York: Plenum

Park, D.C., Hertzog, C., Kidder, D.P., Morrell, R.W., Mayhorn, C.H. (1997). Effect of age on event-based and time-based prospective memory. Psychology and Aging, 12 (2), 314-327.

Park, D. C., Lautenschalger, G., Hedden, T., Davidson, N. S., Smith, A. D.,& Smith, P. K. (2002). Models of visuospatial and verbal memory acrossthe adult life span. Psychology and Aging, 17, 299–320

Parkin, A.J.,Java. R.I. (2000). Determinants of age-related memory loss. In: Perfect, T.J., Maylor, E.A. (Eds), Models of cognitive aging,188-203 Oxford: University Press.

Parkin, A. J., & Java, R. I. (1999). Deterioration of frontal lobe function innormal aging: Influences of fluid intelligence versus perceptual speed. Neuropsychology, 13, 539–545

Parkin A.J. (1997). Normal age-related memory loss and its relation to frontal lobe dysfunction. In Rabbitt P.M.A. (Ed), Methodology of Frontal and Executive Function, 177–190, Hove: Psychology Press.

Perfect, T. (1997) Memory aging as frontal lobe dysfunction. In M.A. Conway (Ed.), Cognitive models of memory. Studies in cognition, 315-339, Cambridge, MA: Mit Press.

Phillips, L. H., and Della Sala, S. 1999. Aging, intelligence, andanatomical segregation in the frontal lobes. Learning and Individual Differences, 10, 217–243

Phillips, L. H. (1997). Do “frontal tests” measure executive function?Issues of assessment and evidence from fluency tests. In: Rabbitt, P.M.A. (Ed.), Methodology of frontal and executive function, 191–213; Hove, England: Psychology Press

Ponds, R.W.H.M., Brouwer, W.H., & van Wolffelaar, P.C. (1988). Age differences in dividedattention in a simulated driving test. Journal of Gerontology: Psychological Science, 43,151-156

Rabbit, P. (1990). Applied cognitive gerontology: Some problems, methodologies and data. Applied Cognitive Psychology, 4 (4), 225-246

Rabbitt P.M.A. (1997) Methodology of frontal and executive function. Hove, UK:Psychology Press.

Rabbitt, P., Lowe, C., & Shilling, V. (2001). Frontal tests and models for cognitiveageing.

146

Bibliography

European Journal of Cognitive Psychology, 13, 5 – 28

Raz, N. (2000). Aging of the brain and its impact on cognitive perfor-mance: Integration of structural and functional findings. In F. I. M. Craik& T. A. Salthouse (Eds.), Handbook of aging and cognition (2nd ed., pp.1–90). Mahwah, NJ: Erlbaum.

Raz, N., Gunning-Dixon, F. M., Head, D., Dupuis, J. H., & Acker, J. D. (1998). Neuroanatomical correlates of cognitive aging: Evidence from structural magnetic resonance imaging. Neuropsychology, 12, 95-114

Reitan, R. M., & Wolfson, D. (1994). A selective and critical review ofneuropsychological deficits and the frontal lobes. Neuropsychology Review, 4, 161–197

Reitan R.M. (1958). Validity of the Trail Making Test as an indicator of organicbrain damage. Perceptual and Motor Skills, 8, 271–76

Robbins T.W., James M., Owen A.M., Sahakian, B.J., Lawrence, A.D., McInnes, L.,Rabbitt, P.M.A. (1998). A study of performance on tests from theCANTAB battery sensitive to frontal lobe dysfunction in a large sample of normalvolunteers: implications for theories of executive functioning and cognitive aging.Cambridge Neuropsychological Test Automated Battery. Journal of the International Neuropsychological Society, 4:474–90

Robertson, I. H., Manly, T., Andrade, J., Baddeley, B. T., & Yiend, J.(1997). Oops!: Performance correlates of everydayattentionalfailures intraumatic brain injured and normal subjects: The Sustained Attention toResponse Task (SART). Neuropsychologia, 35, 747-758

Rogers, R. D. & Monsell, S. (1995). Costs of a predictible switch between simple cognitive tasks. Journal of Experimental Psychology: General, 124, 207-231.

Rubinstein, J., Meyer, D. E., & Evans, J. E. (2001). Executive control of cognitive processes in task switching. Journal of Experimental Psychology: Human Perception and Performance, 27, 763–797

Salthouse, T.A. & Fristoe, N. (1995). Process analysis of adult age effects on a computer-administered trail making test. Neuropsychology, 9, 518-528

Salthouse, T. A. (1985). A theory of cognitive aging. Amsterdam: North-Holland

Salthouse, T. A., & Babcock, R. L. (1991). Decomposing adult agedifferences in working memory. Developmental Psychology, 27,763-776

Salthouse, T. A., & Meinz, E. J. (1995). Aging, inhibition, workingmemory, and speed. Journals of Gerontology: Psychological Sciences,50B, P297–P306

Salthouse, T. A., & Ferrer-Caja, E. (2003). What needs to be explained to account for age-related effects on multiple cognitive variables. Psychology and Aging, 18, 91–110

Salthouse, T. A., Hancock, H. E., Meinz, E. J., Hambrick, D. Z. (1996).Interrelations of age, visual acuity, and cognitive functioning. Journal of Gerontology: Psychological Sciences, 51B, 317–330

Salthouse, T. A. (1991). Theoretical perspectives on cognitive aging. (Hillsdale, NJ: Erlbaum)

147

Bibliography

Salthouse, T. A. (1993). Speed mediation of adult age differences incognition. Developmental Psychology, 29, 722-738

Salthouse, T. A. (1994). Aging associations: Influence of speed onadult age differences in associative learning. Journal of Experimental Psychology: Learning, Memory, & Cognition, 20, 1486-1503

Salthouse, T. A., Fristoe, N., McGuthry, K. E., Hambrick, D. Z. (1998). Relation oftask switching to speed, age, and fluid intelligence. Psychology and Aging, 13, 445-461

Salthouse, T. A., Rogan, J. D., Prill, K. (1984). Division of at-tention: Age differences on a visually presented memory task. Memory & Cognition, 12, 613-620.

Salthouse, T. A., Fristoe, N., McGuthry, K. E., Hambrick, D. Z. (1998). Relation of task switching to speed, age, and fluid intelligence. Psychology and Aging, 13, 445-461.

Salthouse, T. A., & Somberg, B. L. (1982). Isolating the age deficit in speeded performance. Journal of Gerontology 37, 59-63

Salthouse, T.A. (1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103, 403-428.

Schaie, K. W. (1994). The Course of Adult Intellectual Development. American Psychologist, 49, 304-313

Schneider, W. (1988). MEL Experimental Laboratory: An integratedsystem for IBM PC compatibles. Behavior Research Methods, Instruments and Computers, 20, 206-217

Shallice, T. & Burgess, P. W. Deficits in strategy applicationfollowing frontal lobe damage in man. Brain, 114, 727–741(1991).

Shallice, T. (1982). Specific impairment of planning. Philosophical Transactions of the Royal Society of London (Biological Science), B298, 199-209

Shallice, T. (1988). From Neuropsychology to Mental Structure. Cambridge: Cambridge UniversityPress

Shaw, T. G., Mortel, K. F., Meyer, J. S., Rogers, R. L., Hardenberg, J., &Cutaia, M. M. (1984). Cerebral blood flow changes in benign aging andcerebrovascular disease. Neurology, 34, 855-862

Shilling, V. M., Chetwynd, A., & Rabbitt, P. M. A. (2002). Individual inconsistency across measures of inhibition: An investigation of the construct validity of inhibition in older adults. Neuropsychologia, 40, 605-619

Singer, T., Verhaeghen, P., Ghisletta, P., Lindenberger, U., & Baltes, P. B.(2003). The fate of cognition in very old age: Six-year longitudinalfindings in the Berlin Aging Study (BASE). Psychology and Aging, 18,318–331

Snodgrass, J. G., & Vanderwart, M. (1980). A standardized set of 260 pictures: Norms for name agreement, image agreement, familiarity, and visual complexity. Journal of Experimental Psychology: Human Learning and Memory, 6 (2), 174-215.

Span, M.M. (2002). Age related changes in executive control. Chapter 4. Doctoral thesis, UvA.

148

Bibliography

Stuss, D. T., & Benson, D. F. (1984). Neuropsychological studies of thefrontal lobes. Psychological Bulletin, 95, 3–28

Stuss, D. T. (1992). Biological and Psychological Development of Executive Functions. Brain andCognition, 20, 8-23

Stuss DT, Shallice T, Alexander MP, Picton TW. A multi-disciplinary approach to anterior attentional functions. Annals of the New York Academy of Sciences 1995;769:191-212

Tabachnick BG and Fidell LS (1996): Using Multivariate Statistics, 441-505. New York: Harper Collins.

Theeuwes, J., Atchley, P. & Kramer, A.F.(2000). On the time course of top-down and bottom-upcontrol of visual attention. In: Monsell, S., Driver, J. (Eds.). Attention and Performance,18; Cambridge: MIT Press

Tranel, D., Anderson, S. W., & Benton, A. (1994). Development of theconcept of executive function and its relationship to the frontallobes. In: Boller, F., Grafman, J. (Eds.), Handbook of neuropsychology, 9, 125–148. Amsterdam: Elsevier

Tun, P. A., & Wingfield, A. (1999). One voice too many: Adult agedifferences in language processing with different types of distractingsounds. Journals of Gerontology, Psychological Sciences and Social Sciences, 54, 317–327

Uylings, H. B. M., & de Brabander, J. M. (2002). Neuronal changes innormal aging and Alzheimer’s disease. Brain and Cognition, 49, 268–276

Van der Molen, M. W., Ridderinkhof, K. R. (1998). The growingand aging brain: Life-span changes in brain and cognitive functioning.In: Demetriou, A., Doise, W., Van Lieshout, C.F.M., (Eds.), Lifespan developmental psychology: A European perspective, 35-99. Chichester, U.K.: Wiley

Verhaeghen P, De Meersman L. (1998). Aging and the Stroop effect: a meta-analysis. Psychology and Aging,13, 120–126

Verhaegen, P., Salthouse, T.A. (1997). Meta-Analyses of Age-Cognition Relationsin Adulthood. Estimates of Linear and Nonlinear Age Effects and StructuralModels. Psychological Bulletin, 122, 231-249.

Verhaeghen, P., Marcoen, A., & Goosens, L. (1993). Facts and fiction about memory aging: A quantitative integration of research findings. Journal of Gerontology: Psychological Sci ences, 48, P157-P171

Verhaeghen, P., Steitz, D. W., Sliwinski, M. J., Cerella, J.(2003). Aging and dual-task performance: A meta-analysis. Psychology & Aging, 18, 443-460.

Verhage, F. (1964). Intelligentie en leeftijd. Dissertation. Assen, The Netherlands

Vogels, W. W. A., Dekker, M. R., Brouwer, W. H., & de Jong, R. (2002).Age-related changes in event-based prospective memory performance:A comparison of four prospective memory tasks. Brain and Cognition, 49, 341–362

149

Bibliography

Wechsler, D. (1987). Wechsler Memory Scale Revised. San Antonio, TX: The PsychologicalCorporation

Wecker, N. S., Kramer, J. H., Wisniewski, A., Delis, D. C., Kaplan, E. (2000). Age effects on executive ability. Neuropsychology, 14, 409–414

West, R. L. (1988). Prospective memory and aging. In: Gruneberg, M.M., Morris, P.E., Sykes, R.N. (Eds.) Practical aspects of memory: Current research and issues,2, 119-125. Chichester, England: Wiley.

West, R.L. (1996). An application of prefrontal cortex function theory to cognitive aging. Psychological Bulletin, 120, 272-292.

West, R.L., Craik, F.I.M. (1999). Age-related decline in prospective memory: The roles of cue accessibility and cue sensitivity. Psychology and Aging, 14 (2), 264-272.

Wickens, C. D., Braune, R., Stokes, A. (1987). Age differences inspeed and capacity of information processing: A dual-task approach.Psychology and Aging, 2, 70-78

Wilson, B.A., Alderman, N., Burgess, P., Emslie, H., Evans, J.J. (1996). Behavioural assessment of the dysexecutivesyndrome. Bury St Edmunds: Thames Valley Test Company

Wingfield, A., Lindfield, L.C., Kahana, M.J. (1998). Adult age differences in the temporal characteristics of category free recall. Psychology and Aging, 13, 256-266.

Withaar, F.K. (2000). Divided attention and driving: The Effects of Aging and BrainInjury. Doctoral Dissertation. University of Groningen,

Yantis, S. (1998). Control of visual attention. In: Pashler, H. (Ed.), Attention, 223-256. Hove, East Sussex, UK: Psychology Press

Yantis, S., Meyer, D.E., Smith, J.E.K. (1991). Analysis of multinomial mixture distributions: New tests for stochastic models of cognition and action. Psychological Bulletin, 110, 350-374.

Ylikoski R, Ylikoski A, Keskivaara P, Tilvis R, Sulkava R, Erkinjuntti T. (1999). Heterogeneity of cognitive profiles in aging: successful aging, normal aging, and individuals at riskfor cognitive decline. European Journal of Neurology, 6, 645-652.

150

151

Samenvatting

Samenvatting

152

De ‘globale snelheid van informatie verwerking’ hypothese van cognitieve veroudering stelt

dat cognitieve achteruitgang gerelateerd aan ouderdom verklaard kan worden door een ver-

mindering van de snelheid waarmee elementaire cognitieve processen worden uitgevoerd.

Deze vermindering in verwerkingssnelheid begrenst het prestatieniveau dat kan worden be-

haald in de meeste cognitieve taken (Eearles, Connor, Smith & Park, 1997; Salthouse, 1996).

Deze hypothese staat in contrast met de hypothese die ouderdomseffecten op prestaties op

cognitieve taken toewijst aan specifieke beperkingen in cognitieve controle processen.

Cognitieve controle refereert aan de organisatie en monitoring van cognitieve processen.

Het is cruciaal voor het aanpassen van gedrag aan intenties en interne doelen. Het concept

cognitieve controle impliceert endogene controle van cognitieve processen en kan daarmee

gecontrasteerd worden met situaties waarin gedrag expliciet wordt geïnitieerd of gecued

vanuit de omgeving. Cognitieve controle wordt soms geconceptualiseerd als een unitair pro-

ces dat hetzelfde is in verschillende situaties. Een andere conceptualisatie is dat cognitieve

controle een samenstelling is van meerdere gefractioneerde controle processen. Volgens deze

laatste conceptualisatie kunnen cognitieve controle processen in de ene situatie anders zijn

dan cognitieve controle processen in een andere situatie.

In dit proefschrift worden drie studies beschreven waarbij de relatie tussen veroudering en ver-

schillende cognitieve controle functies en snelheid van informatieverwerking zijn onderzocht,

Hoofdstuk één beschrijft enkele theorieën die poneren dat leeftijdsgerelateerde verschillen

verklaard kunnen worden door louter een algemeen aspect van cognitie (e.g. Birren, 1956;

Cerella, 1985; Eearles et al., 1997; Salthouse, 1996), en enkele andere theorieën die inhou-

den dat leeftijdseffecten op cognitie specifiek begrensd zijn tot executieve controle functies

(e.g. Bryan & Luszcz, 2001; Mayr, Spieler & Kliegl, 2001; Rabbitt, 1997; Wecker, Kramer,

Wisniewski, Delis & Kaplan, 2000). Daarbij worden de meest invloedrijke modellen van cog-

nitieve controle beschreven in hoofdstuk één.

In hoofdstukken twee en drie worden studies naar cognitieve veroudering gepresenteerd

die zich richten op een specifieke functie van cognitieve veroudering. In hoofdstuk twee

wordt een studie gepresenteerd waarin de sensitiviteit en betrouwbaarheid in het bepalen

van effecten van normale veroudering van vier verschillende prospectief geheugen taken

wordt getest. Dit gebeurt door een groep ouderen met een jongere groep te vergelijken.

Vanuit het perspectief van de ‘frontal lobe’ hypothese van cognitieve veroudering zijn pros-

pectief geheugen taken bijzonder relevant, omdat de prestatie op deze taken zowel planning

als intentie-activatie (een prospectieve intentie actief houden tijdens het uitvoeren van een

andere taken) vereist. In het algemeen wordt van deze twee functies aangenomen dat de

frontale kwab erbij betrokken is. Gedacht vanuit de frontale kwab hypothese van cognitieve

153

Samenvatting

veroudering werden robuuste leeftijdsgerelateerde verschillen in prestaties op prospectief

geheugen taken verwacht. De vier taken verschilden op verschillende dimensies, zoals de

mate waarin de prospectieve gebeurtenis opvalt, de frequentie van voorkomen van de pros-

pectieve gebeurtenissen, de complexiteit van de instructies en het voorleggen van feedback

na een prospectief geheugen fout. De rol van het in het geheugen vasthouden en de rol van

basissnelheid van informatieverwerking als medierende factoren van het effect van leeftijd

op prospectief geheugen werden besproken. In drie van de vier prospectief geheugen taken

lukte het sommige proefpersonen niet om de taak volgens instructie uit te voeren en lieten

zij weinig tot geen reacties zien op de prospectieve cues. Soms kon de relevante instructie na

afloop zelfs niet worden gereproduceerd. Sommige van deze ‘non-performers’ op het gebied

van prospectief geheugen presteerden normaal op de achtergrondtaak. Dit suggereert spe-

cificiteit van prospectief geheugen problemen van deze proefpersonen. Andere ‘non-perfor-

mers’ presteerden wat betreft nauwkeurigheid ook erg slecht op de achtergrondtaken, wat

duidt op het niet begrijpen van de taak instructies. Met uitzondering van één proefpersoon

zaten alle ‘non-performers’ in de groep oudere proefpersonen. Met andere woorden, het fa-

len in het navolgen van de prospectief geheugen instructies was bijna volledig beperkt tot de

groep van oudere volwassenen. Twee verschillende taken bleken gevoelige en betrouwbare

instrumenten te zijn voor het inschatten van effecten van normale veroudering op prospectief

geheugen. Deze taken richtten zich op het vermogen van de proefpersoon om prospectieve

intenties naar behoren geactiveerd en toegankelijk te houden. Consistent hiermee is dat een

substantiële correlatie werd gevonden tussen prestaties op deze twee taken. Bij deze correla-

tie speelde leeftijd een mediërende rol.

In hoofdstuk drie worden effecten van leeftijd op taak-switchen gerapporteerd. Er werden

grotere ‘lokale switch kosten’ gevonden voor ouderen dan voor jong volwassenen, maar niet

disproportioneel groter. In lijn met vorig onderzoek bleken ‘globale switch kosten’ wel dispro-

portioneel groter voor ouderen dan voor jongeren (zie bijv. Kray, 2005). Meer specifiek was

het onderzoek in hoofdstuk drie gericht op het nagaan of ‘residuele switch’ kosten bij oude-

ren en jongeren toewijsbaar zijn aan de zelfde factoren. Residuele switch kosten zijn de switch

kosten die aanwezig blijven zelfs als er voldoende tijd wordt gegeven om voor te bereiden

op de taakswitch, Op het niveau van gemiddelde reactietijden (RT) werden geen significante

leeftijdsgerelateerde verschillen in residuele switch kosten gevonden. In het huidige onder-

zoek is echter meer gedetailleerd naar de residuele switch kosten gekeken, gebruikmakend

van analyse van RT distributies. Een distributionele analyse van de RT’s legde bloot dat de

residuele switch kosten bij jongeren volledig toe te schrijven waren aan inconsistenties bij

voorbereiding van de taak. Dit is een replicatie van de uitkomsten van studies van De Jong

(2001) en van Nieuwenhuis en Monsell (2002). Ouderen bleken op dit aspect van voorberei-

ding van de taakswitch niet te verschillen van jongeren. Echter, voor de oudere groep bleek

dit af en toe falen in bij voorbereiding van de taak niet voldoende verklaring voor de residuele

154

Samenvatting

switch kosten. Klaarblijkelijk bereikten ouderen met voorbereiding van de taak niet het zelfde

niveau van voorbereiding op taakswitch als op taakherhaling. Ouderen hadden blijkbaar nog

extra tijd nodig nadat de stimulus was gepresenteerd. Een mogelijke verklaring voor de tijd

die ouderen nodig hadden om te reageren op een taakswitch (zelfs als ze voorbereid waren)

is dat het een achtergang in de kwaliteit van het terughalen van taak informatie uit het lange

termijn geheugen reflecteert. Dit zou consistent zijn met de interpretatie van locale switch

kosten die Mayr en Kliegl (2000) voorstellen. Mayr en Kliegl (2000) hypothetiseren dat lo-

cale switch kosten een proces van actief terughalen van taak gerelateerde informatie uit het

lange termijn geheugen reflecteren. Gerelateerd aan de taak-switch studie is onderzoek naar

prospectief geheugen, dat veelal gericht is op actief houden van een intentie bij sterk inter-

fererende omstandigheden. Meerdere studies, zoals in hoofdstuk twee van dit proefschrift,

hebben zich bezig gehouden met veroudering en prospectief geheugen, en ouderen blijken

in deze studies vaak slechter te presteren in situaties waarbij relatief weinig omgevingssteun

relatief hoge interferentie is (zie ook Einstein, Smith, McDaniel en Shaw, 1997; Maylor, 1996).

In hoofdstuk vier is een meer algemene benadering toegepast om de relaties tussen de cog-

nitieve controle functie en veroudering te bestuderen. Allereerst is er rekening gehouden

met meerdere aspecten van cognitie, in plaats van alleen te richten op een enkele cognitieve

controle functie,. Snelheid van informatieverwerking, intelligentie en meerdere cognitieve

controle functies werden gemeten met verschillende cognitieve taken. Ten tweede is er voor

een andere benadering gekozen met betrekking tot de leeftijdscohorten. In plaats van twee

leeftijdsgroepen zijn er proefpersonen gerekruteerd uit vier leeftijdsgroepen: een groep jon-

geren (tussen 20 en 30 jaar) en drie groepen met ouderen vanaf 50 jaar. Voorafgaand aan

‘structural equation modeling’ is een factor analyse gedaan waarbij twee factoren werden ge-

identificeerd die bestonden uit variabelen die respectievelijk executieve controle functies en

werkgeheugen meten. Deze uitkomsten kunnen worden vergeleken met de factor structuur

van executieve functies die Myiake en collega’s (2000) vonden door middel van ‘confimatory

factor analyse’. De drie factoren die Myiake et al. (2000) vonden, representeren executieve

functies betrokken bij taak-switchen, bijwerken en actief houden van werkgeheugen repre-

sentaties (WM) en inhibitie van dominante of prepotente responsen. De WM factor in de

studie van hoofdstuk vier is vergelijkbaar met de WM factor van Myiake’s studie en de ‘execu-

tieve controle functies’ factor van de huidige studie lijkt een mix van de andere twee factoren

die door Myiake et al. werden gevonden. Door middel van ‘structural equation modeling’

zijn vier modellen vergeleken die verschillende relaties tussen leeftijd, de latente factoren

en cognitieve maten veronderstellen. Het patroon van gedrag op de cognitieve taken bleek

het meest accuraat gerepresenteerd door het model dat een direct effect van leeftijd op

de executieve functies (EF) factor veronderstelt. Volgens dat model worden leeftijdseffecten

op werkgeheugen taken gemedieerd door de directe leeftijdseffecten op executief functio-

neren. Deze resultaten komen overeen met de resultaten van een studie van Span (2002).

155

Samenvatting

Span (2002) testte door middel van ‘structural equation modeling’ of een model zonder

executieve functies prestaties van verschillende leeftijdsgroepen slechter kan beschrijven dan

wanneer er een snelheid en een executieve functies factor wordt verondersteld. Span vond

dat de EF factor alleen nodig was om de response snelheid van ouderen te kunnen verklaren

en dat de ouderen van jongeren verschilden op zowel de snelheid als op de EF factor. Een

vergelijking van de vier leeftijdsgroepen in de studie van hoofdstuk vier legt verschillende

patronen van leeftijdsgevoeligheid van verschillende cognitieve variabelen bloot. Deze patro-

nen suggereren dat cognitieve prestaties die sterk afhankelijk zijn van snelheid, gevoelig zijn

voor veroudering in die zin dat de gevolgen zich al manifesteren in de groep met de leeftijd

tussen 50 en 60 jaar. Aan de andere kant gaan cognitieve prestaties die meer afhankelijk zijn

van adequaat executief functioneren pas op latere leeftijd achteruit.

De belangrijkste implicaties van de bevindingen van de studies in dit proefschrift hebben

betrekking op validiteit van snelheidsmaten en op heterogeniteit van verouderingseffecten

op cognitie. Met betrekking tot de relatie tussen cognitieve veroudering en snelheid van

informatieverwerking werd herhaaldelijk gevonden dat bij veel gebruikte maten voor snel-

heid er niet sprake is van een pure maat voor snelheid en dat prestaties op dat soort testen

afhankelijk zijn van cognitieve controle functies. Dit impliceert een noodzaak tot nuancering

van het bevestigen van de ‘globale snelheid van informatieverwerking’ hypothese op basis

van leeftijdsgerelateerde verschillen in prestaties op die testen.

Eén van de in het oog vallende uitkomsten van de drie studies en bevindingen van anderen

(bijv. Ylikosky, 1999, Gunstad, 2006) betreft de heterogeniteit in achteruitgang van cognitie-

ve prestaties bij veroudering. De studie in hoofdstuk twee liet zien dat, hoewel de ouderen als

groep slechter presteerden dan de jongere groep met betrekking tot prospectief geheugen,

sommige oudere proefpersonen juist vergelijkbaar presteerden als de jongere proefpersonen,

terwijl andere oudere proefpersonen veel slechter presteerden op de prospectief geheugen

indices. In de taakswitch studie van hoofdstuk drie bleek de interpretatie van de groepsana-

lyses van reactietijden simpel, maar werd benadrukt dat er met name in de oudere groep

proefpersonen grote individuele verschillen bestonden. Ook in hoofdstuk vier werden bij

verscheidene cognitieve taken het heterogene karakter van effecten van veroudering op pres-

taties gerapporteerd. Deze resultaten impliceren dat bij studies naar cognitieve veroudering

het onderzoeken van groepsgemiddelden niet volstaat. Veel informatie ligt in de individuele

verschillen en binnen variabiliteit van gedrag binnen een taak (zie ook Gunstad et al., 2006).

Een meer praktische implicatie in het licht van de arbeidsmarkt, pensioen leeftijden en lage

geboorte cijfers, is dat, als het om cognitief werk gaat, louter de leeftijd van een individu niet

een betrouwbare indicator is van (achteruitgang in) kwaliteit van arbeidspotentieel.

156

157

Dankwoord

dankwoord

158

Het onderzoek dat wordt beschreven in dit proefschrift was niet mogelijk geweest zonder de

inzet van een flink aantal mensen. Het daadwerkelijk schrijven van het proefschrift was een

extensieve bezigheid en is daarmee een symbool van de steun en het geduld van anderen.

Hier wil ik deze mensen noemen en daarmee mijn dank uitspreken.

Allereerst veel dank aan de mensen die als proefpersoon hebben deelgenomen aan de ex-

perimenten. Dit leverde niet alleen de gegevens voor het onderzoek op, maar ook prachtige

gesprekken en verhalen tijdens de pauzes. Veel dank aan mijn promotores, Ritske de Jong

en Wiebo Brouwer, voor hun begeleiding. Ritske, van het begin tot eind van dit onderzoeks-

project ben je altijd beschikbaar geweest voor overleg en begeleiding. Jouw enthousiasme

voor en kennis van functieleer psychologie gaf telkens inspiratie om door te zetten met het

onderzoek en het schrijven van het proefschrift. Bedankt voor je gedrevenheid en geduld. Ik

heb ook erg genoten van de nodige gesprekken over andere zaken, met name over onze ge-

deelde passie voor duursporten. Wiebo, jouw rol was voor mij erg belangrijk bij het opzetten

van het onderzoek en bij het tot stand komen van de finale versie van het proefschrift. Veel

dank voor je inzet, je ideeën en voor je creativiteit. Dank aan de leden van de beoordelings-

commissie: Anke Bouma, Natasha Maurits en Richard Ridderinkhof voor het beoordelen van

dit proefschrift. Werner Vogels was mijn directe collega-AIO en kamergenoot en we hebben

samen het tweede hoofdstuk geschreven. Werner, bedankt voor je collegialiteit, samenwer-

king en tolerantie ten aanzien van mijn rookgedrag in die tijd. Haha, je moet me nu eens

zien! De samenwerking met en inzet van studenten aan het onderzoek was onmisbaar. Dank

hiervoor aan Joost, Jessica, Pines, Harro, Marieke en Marten. Technische ondersteuning was

altijd voorhanden bij het Instrumentatiedienst psychologie. In het bijzonder dank aan Jan

Smit, Pieter Zandbergen en Hans Veldman. Dank aan Frederike Jörg, Ernestine Gordijn en

Simon Hulshoff voor het meelezen en becommentariëren van verschillende onderdelen van

het proefschrift.

Door de collega’s van de afdeling experimentele en arbeidspsychologie heb ik in een prettige

werksfeer aan mijn onderzoek kunnen werken en was hulp of een praatje nooit ver te zoeken.

Ans van Rijsbergen, bedankt voor je hulp en je persoonlijke betrokkenheid. De belangstelling

van Ben Mulder, Dick de Waard en Karel Brookhuis heb ik erg gewaardeerd, dank daarvoor!

Monique Lorist en Harold Bekkering, bedankt voor de inspirerende vakinhoudelijke discus-

sies. De AIO/postdoc-groep bepaalt vooral het beeld van mijn herinnering aan die tijd. Dank

daarvoor aan Sander Martens, Alan White, Janneke Brouwer, Linda Jongman, Klaas Jan Bruin,

Pines Nuku, Lennart Quispel, Jurjen van der Helden, Maarten Boksem, Wokje Abrahamse,

Geertje Schuitema, Paco Guzmán Muňoz en in het bijzonder Judith de Groot en Matthijs

Dicke, tevens mijn klimmaatjes van het eerste uur!

159

Dankwoord

Mijn collega’s bij GGZ Friesland hebben veel support gegeven. Rob Smeets, bedankt voor

jouw aanhoudende belangstelling voor de voortgang van het schrijfproces. Jouw peptalks

waren effectief en ik waardeer het bijzonder. Ik heb de afgelopen jaren ook veel gehad aan de

betrokkenheid en collegialiteit van Harry Tijssen, Margriet Verloop, Marty Kruizenga, Veronica

Kooy, Katja Westra en Sally Schmitz. Hartelijk dank! Sally, de gedachte aan jullie in die bergen

geeft op voorhand al rust. Louwra Weisfelt, bedankt voor je betrokkenheid en energie in het

algemeen en jouw inzet bij de vormgeving en het drukken van dit proefschrift in het bijzon-

der, je bent top! Veel dank aan mijn fiets- en klimmaatjes voor de optimale ontspanning die

naast werk en schrijven onontbeerlijk is geweest voor mijn fysieke en psychische gezondheid.

In het bijzonder dank aan Donovan Rabs en Simon Hulshoff. Super dat jullie mijn paranim-

fen willen zijn. Tenslotte wil ik uiteraard mijn vader en Hilma en Holger bedanken voor hun

belangstelling en steun.

160

Documents

University of Groningen Aging and cognitive control Dekker ... · of performance measures on a battery of executive tasks, rather than a single (unitary) factor. Nevertheless, there