8
Acta Psychologica 42 (1978) 21-28 © North-ltolland Publishing Company EXPERIMENTAL EVIDENCE FOR THE SIMILARITY OF PERSONALITY AND ATTITUDE ITEM RESPONDING* T. B. ROGERS The University of Calgary, Alberta. Canada Received April 1977 The process of responding to attitude items was broken down into a series of cognitive stages and a model offered. To test this model, subjects responded to attitude items varying in Extremity under two or five response alternative formats. By measuring response times, and applying Sternberg's (1969) additive factor method, the model was supported. The results were discussed in terms of previous process work involving personality items and sentence verification tasks. 1. Introduction Rogers (1974a, b) has developed a model of the process of re- sponding to personality items. This model postulates that four serial, modular substages intervene between the respondent's perception of the item and his 'true' or 'false' response to it. The first two stages are related to the reading and understanding of the item. Stage 1, Item Encoding, involves the translation of the physical item properties (i.e., the letters, spaces, etc.) into an internal language which will permit operation by subsequent stages. This stage is immediately followed by the second stage, Item Comprehension, wherein the semantic and linguistic components of Stage 1 output are extracted. At this point in the sequence the reader understands the item. In the third stage (Self Referent Decision) the output of the second stage is compared with an internalized memory store. This store would contain the respondent's view of 'self', based on an abstraction of salient self-referent infor- *This research was supported by a grant from the Canada Council. I wish to thank D. M. Taylor, D.W. Fiske, P. Keevil, M.E. Morf and D. Woodhead for their comments on earlier drafts. The author's present address is: Department of Psychology, University of Calgary, Calgary, Alberta, Canada, T2N 1N4.

Experimental evidence for the similarity of personality and attitude item responding

Embed Size (px)

Citation preview

Page 1: Experimental evidence for the similarity of personality and attitude item responding

Acta Psychologica 42 (1978) 2 1 - 2 8 © North-l tolland Publishing Company

EXPERIMENTAL EVIDENCE FOR THE SIMILARITY OF PERSONALITY AND ATTITUDE ITEM RESPONDING*

T. B. ROGERS The University o f Calgary, Alberta. Canada

Received April 1977

The process of responding to at t i tude items was broken down into a series of cognitive stages and a model offered. To test this model, subjects responded to at t i tude i tems varying in Extremity under two or five response alternative formats. By measuring response times, and applying Sternberg's (1969) additive factor me thod , the model was supported. The results were discussed in terms of previous process work involving personality i tems and sentence verification tasks.

1. Introduction

Rogers (1974a, b) has developed a model of the process of re- sponding to personality items. This model postulates that four serial, modular substages intervene between the respondent 's perception of the item and his ' true' or 'false' response to it. The first two stages are related to the reading and understanding of the item. Stage 1, Item Encoding, involves the translation of the physical item properties (i.e., the letters, spaces, etc.) into an internal language which will permit operation by subsequent stages. This stage is immediately followed by the second stage, Item Comprehension, wherein the semantic and linguistic components of Stage 1 output are extracted. At this point in the sequence the reader understands the item. In the third stage (Self Referent Decision) the output of the second stage is compared with an internalized memory store. This store would contain the respondent 's view of 'self', based on an abstraction of salient self-referent infor-

*This research was supported by a grant from the Canada Council. I wish to thank D. M. Taylor, D.W. Fiske, P. Keevil, M.E . Morf and D. Woodhead for their comment s on earlier drafts. The author ' s present address is: Depar tment of Psychology, University of Calgary, Calgary, Alberta, Canada, T2N 1N4.

Page 2: Experimental evidence for the similarity of personality and attitude item responding

22 7: B. Rogers/Personality and attitude item responding

mation. In a structured assessment setting, the respondent must inap the output of the comparison between the item and the memory store into the available response alternatives. While this comparison occurs in Stage 3, the selection of an appropriate response from among those available takes place in the fourth stage, Response Selection. In this stage the outcome of the I tem/Memory Comparison is mapped into the response options. The response then becomes compatible with the task output requirements. Once Stage4 is completed the respondent initiates his external response. Fig. 1 diagrams this four stage model.

ITEM ITEM ITEM NUMBER OF RESPONSE LENGTH AMBIGUITY EXTREMITY ALTERNATIVES

'TEM'--'~ENIcToEDMINGHCOMPREHENSIONH COMPARISON ~ R E S P O N S E

Fig. 1. Diagram of the proposed model of attitude item responding.

This type of stage model can be tested empirically using Sternberg's (1969) 'additive-factor method ' (Trabasso et al. 1971 ; Clark and Chase 1972). This method involves identifying empirical ( independent) variables that affect the amount of time spent in a specific stage. By experimentally manipulating the levels of these independent variables, in a factorial design, and assessing the response times (RT) to the various items, it becomes possible to discover the nature of the relations among these stages. If each independent variable contributes a unique (noninteracting) component to the overall RT, evidence for a separate, modular stage has been offered. For example, Rogers (1974b) defined Item Extremity as the empirical variable associated with the Self Referent Decision Stage of the model, with low extremity items (i.e., those to which about half of the respondents answer ' t rue ' ) requir ing more time for response than high extremity items (i.e., those answered the same way by most respondents). Further, Number of Response Alternatives was defined to control time in the Response Selection Stage. Subjects responded to high and low Extremity items using two or five response alternatives in a situation that permitted assessment of their RTs. If the Self Referent Decision was finished before the other stage began, one would expect low Extremity items to require more time than high Extremity ones, regardless of the Number of Response Alternatives. Conversely, RT effects associated with Number of

Page 3: Experimental evidence for the similarity of personality and attitude item responding

T. B. Rogers/Personality and attitude item respoudb~ 23

Response Alternatives should not be contingent upon item Extremity if the stages are modular or separate. Modularity of stages would be reflected in a nonsignificant interaction between RTs for Item Extremity and Number of Response Alternatives combined with signifi- cant main effects of each factor. Rogers (1974b) found such results for these two stages. Also, Rogers (1974a) has found additivity for all combinations of the first three stages. Thus, Sternberg's (1969) 'additive-factor method ' permits assessment of the nature of the rela- tionships between a series of stages thought to underlie a complex process.

The present paper is concerned with the generality of the four stage personality item responding model. The hypothesis to be tested is that a similar model (e.g., four serial, modular substages) can account for responding to attitude items as well. While responding to attitude items appears to involve a different memory store, compared to personality items, it is still possible that the basic model outlined above will give a useful account of attitude item responding. The support of this hypoth- esis would begin to suggest a basic, process-oriented model of re- sponding to verbal stimuli in a structured situation. Such a general model may provide a means of bridging between disparate areas of psychological research and assessment methodologies.

The present study restricts consideration to Stages 3 and 4 of the model. That is, subjects responded to two kinds of attitude items (high or low Extremity), under two response format conditions (dichot- omous Agree-Disagree or a five-point Likert scale ranging from 'Strong- ly Disagree' through to 'Strongly Agree'). Support of the hypothesis of similarity between personality and attitude item responding would require: (1) RTs highest for low Extremity items, (2) RTs highest for five response alternatives, and ( 3 ) n o interaction between these vari- ables.

2. The e x p e r i m e n t

2.1. M e t h o d

2.1.1. Design Because of the large be tween-sub jec t variance in responses to a t t i t ude i tems, it

was deemed appropr i a t e to make the compar i son be tween two and five response a l ternat ives repea ted measures. This compl ica tes the design as it is no t possible to

Page 4: Experimental evidence for the similarity of personality and attitude item responding

24 T. B. Rogers/Personality and attitude item responding

use the same i t ems unde r each response fo rmat , since the second response to an i tem is typical ly faster t han the first (Rogers 1973). Two ma tched sets ( ~ and B) of i t ems were developed and used in a c o u n t e r b a l a n c e d design. Tile o rder in which the r e s p o n d e n t s used the two- or f ive-point fo rma t was also coun t e rba l anced , as Rogers ( 1 9 7 4 b ) found several o rder effects . Fo l lowing previous research the E x t r e m i t y variable was also wi th in subjects .

2. 1.2. I t ems

All i t ems came f rom scales in Shaw and Wright (1967) tha t listed empir ical ag reemen t p ropor t i ons , se lected to r ep resen t the en t i re range of this variable. I tems were ed i ted to meet four cr i ter ia: (a) posi t ive wording, (b) r easonab ly s imple s en tence s t ruc tu re , (c) t h i rd -pe r son wording, and ( d ) s t r i c t l eng th cr i ter ion (all w i th in a t en - l e t t e r range). 299 i t ems survived this edi t ing process. New agreement p r o p o r t i o n s were genera ted using 50 univers i ty s t uden t s (25 females) who re- sponded to these i t ems u n d e r usual ag reemen t ins t ruc t ions ( f ive-point response fo rma t ) . A similar sample ra ted , using a f ive-point scale, the ambigu i ty of these i t ems (see Rogers et al. 1975 for p rocedura l details) . F r o m this pool of i t ems three sets were chosen : 30 i t ems were used for warm-up , and two sets of 80 i tems each (Sets A and B) tha t were used in the RT s tudy. I t ems in Sets A and B were m a t c h e d pair-wise for ag reemen t p r o p o r t i o n s and ambigu i ty ratings, y ie lding ident ical d i s t r ibu t ions of these s tat is t ics for the two sets.

I t em E x t r e m i t y was def ined by r ank ing each set of 80 i tems by ag reemen t p r o p o r t i o n . The high level of I tem E x t r e m i t y consis ted of the 20 i tems f rom the top and 20 f rom the b o t t o m quar t i l e of this ranked i tem set. The 40 middle i tems made up the low I tem E x t r e m i t y set.

2.1.3. Procedure The e x p e r i m e n t consis ted of two sessions separa ted by 24 hours . Half of the Ss

r e s p o n d e d using two a l te rna t ives (Disagree or Agree) in the first session, and five a l te rna t ives (S t rongly Disagree t h r o u g h to S t rongly Agree) in the second session. Half of these Ss received i t em Set A in the first, and i tem Set B in the second session. All o the r possible c o m b i n a t i o n s of the two order factors were equal ly r ep resen ted .

I t ems were p resen ted using a Singer-Graf lex rear-screen pro jec tor , wh ich p resen ted the t y p e w r i t t e n i t ems o n t o an opaque screen s i tua ted a b o u t 1 m in f ron t of the S. Ss ind ica ted the i r responses using a m u l t i - b u t t o n e d response panel placed w i t h i n easy reach. The p ro jec to r was con t ro l l ed and responses recorded using a PDP 8/I c o m p u t e r loca ted in an ad jacent room. Details of the RT p rocedure used in this s t udy are presen ted in Rogers ( 1 9 7 4 a ) and will no t be dupl ica ted here. Fo r each i t em the S 's response and an RT were recorded . The RT was the t ime, in msec, f r om the onse t of the i t em to the release of a h o m e key which the S had depressed t h r o u g h o u t his processing of the i t em. The S would release this h o m e key when he moved to ind ica te his response. No specif ic ins t ruc t ions to induce a speed set were inc luded .

Page 5: Experimental evidence for the similarity of personality and attitude item responding

T. B. Rogers/Personality and attitude item responding 25

2.1.4. Subjects The 40 vo lun t ee r Ss came f rom unde rg radua te psychology courses. Each of the

four cells associated wi th the order man ipu l a t i ons con t a ined six females and four males. The mean age of the subjects was 25.6 years. Each was paid $3.00 for his par t ic ipa t ion .

2.2. Resul ts

For each S the mean RT to i tems of each level of I tem E x t r e m i t y , unde r each response fo rma t cond i t ion ( two or five response a l te rnat ives) was calcula ted. These means were sub jec ted to a four-way analysis of variance wi th the factors: (1) I t em E x t r e m i t y (High, Low), (2) N u m b e r o f Response Al te rna t ives (Two, Five), (3) Order of Response F o r m a t ( 2 - 5 , 5 - 2 ) , and (4) Order of I t em Sets (A-B, B-A). Fac tors (1) and ( 2 ) w e r e repea ted measures . I t em E x t r e m i t y showed a s ignif icant effect on RT (F(1 , 3 6 ) = 4 0 . 2 0 , p < 0 .01) wi th low E x t r e m i t y i t ems requi r ing a mean of 7723 msec compared to 7172 msec for high E x t r e m i t y i tems. RT was longer ( F ( I , 3 6 ) = 12.47, p < 0 .01) when five response a l te rnat ives were used ( m e a n = 7872 msec) t han for the two-a l te rna t ive f o r m a t (mean = 7068 msec) . The i n t e r ac t i on be tween I tem E x t r e m i t y and N u m b e r of Response Al te rna t ives was no t s ignif icant (F(1 , 36) < 1) and is d i ag rammed in fig. 2. Of the main effects and

8 5 0 0

8000

E

7500

Z

7000

6500

I - - O Low extremity o - - - - o High extremity

s ~

I I 2 5

NUMBER OF RESPONSE ALTERNATIVES

Fig. 2. Reaction time to attitude items as a function of Item Extremity and Number of Response Alternatives.

i n t e rac t ions involving the order factors, on ly one i n t e r ac t i on was s ignif icant . Tha t was the Order of Response F o r m a t × N u m b e r of Response Al te rna t ives i n t e r ac t i on (F(1 , 3 6 ) = 4 .93, p < 0.05) . This i n t e r ac t i on ind ica ted a greater RT savings in the two-response case for the Ss w ho r e sponded in the f ive-al ternat ive c o n d i t i o n first. This i n t e r ac t i on does not compl ica te i n t e r p r e t a t i o n of the I t em E x t r e m i t y x N u m b e r of Response Al te rna t ives i n t e r ac t i on as the h igher order i n t e rac t ions involving these two variables and Order of Response F o r m a t were no t s ignif icant .

Page 6: Experimental evidence for the similarity of personality and attitude item responding

26 T. B. Rogers/Personalit.v and attitude item responding

As a check on the nature of the dis t r ibut ions under ly ing the RT means, the second order cumulants associated with the means analyzed above were analyzed in a 2 × 2 analysis o f variance ( I tem Ex t remi ty x Number of Response Alternatives) . All three F ratios in this analysis were less than 1, indicating no significant RT disl r ibutional d i f ferences among lhe four cells.

3. Discussion

The present data clearly support all three conditions of the hypo- thesis that the modular model can be used to describe attitude item responding. (1) The main effect of Extremity, which indicates that low Extremity items require more time for response, replicates Rogers' (1973, 1974a, b) previous work with tile model. (2 )The present data clearly indicate that it requires more time to answer atti tude items using five alternatives. This finding replicates Rogers (1974b) and is readily predictable from RT data using much simpler tasks (e.g., Hick 1953; Hyman 1953). The extent of the additional time required for a five-point answer is not as great (mean difference = 759 msec) for the atti tude domain as was fottnd with personality items (mean differ- ence = 1204 msec: Rogers 1974b). However, there is still a meaningful increment in time required to answer in a five- as opposed to a two-point format. Whether the extra time required for a five-point answer is justified by more valid responses (Peabody 1962 would suggest it was not) is a matter for further research and development in the study of attitude-item responding. (3 )The non-significant inter- action diagrammed in fig. 2 indicates that Item Extremity and Number of Response Alternatives each add unique, nonoverlapping components to the overall RT to att i tude items. Following Sternberg (1969) these data can be used to infer that Stages 3 and 4 of the proposed model are modular. Stage 3 is completed before Stage 4 is begun. These data replicate the personality item model (Rogers 1974b).

Given support of the basic structure of the model in atti tude item responding, it is interesting to explore the particular stages as they translate into the att i tude domain. While the present data do not permit statements about Stages 1 and 2, it is unlikely that these are very different from the personality item case. The major difference would be in Stage 3. This Stage is labeled the I tem/Referent Comparison in the atti tude case. Here the item is compared with an internalized memory store. This memory store would contain a summary of the respondent's

Page 7: Experimental evidence for the similarity of personality and attitude item responding

T. B. Rogers/Personality and attitude item responding 27

experiences with the referent issue. For example, for an item related to capital punishment, the memory store would be an abstraction or representation of the respondent 's experiences with the issue of capital punishment. The respondent 's cognition, affects and perceived behav- ioral tendencies toward the referent issue would be summarized in this memory store. The store would be derived from reading, debates, exposure through the media, education, early training, affective reactions and view of social expectation associated with the referent issue. While there would appear to be substantial individual variation in both the nature and size of this memory store, it seems plausible to posit that these aspects of a person's experience would be coded, and somehow collated, under the general heading of capital punishment. The nature of the development and organization of this store is beyond the scope of this paper. For the purposes at hand, the existence of the memory store merits emphasis. The output of the comparison between this memory store and the internalized item would trigger the Response Selection Stage, which could be identical for both the attitude and personality stimuli, particularly in light of the present data indicating modulari ty of Stages 3 and 4.

The difference between Stage 3 of the personality and atti tude models has been elaborated in a series of recognition memory studies (Rogers, in press). In these studies it was found that instructions to "read the item, decide if it describes you, and use this to help your memory" induced differential effects for attitude compared to per- sonality items. The 'self-referent instructions' resulted in a performance increment for personality, but not for attitude items, indicating that the comparison processes (i.e., Stage 3) for the attitude and personality domains are different. However, the present data do indicate that the general format of the modular model is useful for att i tude item re- sponding.

The present data indicate that the four' stage model is useful in the attitude domain. This in combination with the personality data suggests that this format of model may have a degree of generality. This generality is underscored by a series of models that have evolved in experimental psycholinguistics. The comprehension of sentences is often researched using a task called verification. Subjects are shown a stimulus (e.g., a black dot) and asked to respond 'true' or 'false' to sentences such as 'The dot is not red'. Clark and Chase (1972) and Trabasso et al. ( 1971 ) using the additive factors method, independently

Page 8: Experimental evidence for the similarity of personality and attitude item responding

28 T. B. Rogers/Personalio, attd attitude item responding

generated four stage models for this task. These models are very similar to the present formulation, particularly since contact with memory occurs in Stage 3 of all o f the models. Thus, it appears that the present inodel can be used to help understand a diversity of tasks involving structural responses to verbal stimuli.

The present data support a model of the psychological processes attending the response to att i tude items, which emphasizes a series of information transformations mediating itetn presentation and response. The hope of presenting such a tnodel is that it will offer a viable means of coming to grips with what the respondent is doing when we measure his attitudes. To the extent one can argue that the respondent 's cog- nitions, expectat ions and emotional valences to the domain being assessed affect the observed measure (cf. Bem and Allen 1974), an understanding of the item responding process cata be seen as a pre- requisite of useful measurement.

References

Bern, D. G. and A. Allen, 1974. On predicting some of tile people some of tire time: Tile search for cross-situational consistencies in behavior. Psychological Review 81,506 520.

Clark, H. tC and W.G. Chase, 1972. On the process of comparing sentences against pictures. Cognitive Psychology 3 ,472 -517 .

Hick, W.E., 1953. On the rate of gain of information. Quarterly Journal of Experimental Psychology 4, 11-26.

Hyman, R., 1953. Stimulus information as a determinant of reaction time. Journal of Experimental Psychology 45, 188-196.

Peabody, D., 1962. Two components of bipolar scales. Psychological Review 69, 65-73 . Rogers, T. B., 1973. Toward a definition of the difficulty of a personality item. Psychological

Reports 33, 159-166. Rogers, T. B., 1974a. An analysis of tile stages underlying the process of responding to per-

sonality items. Acta Psychologica 38 (3), 205 214. Rogers, T. B., 1974b. An analysis of two central stages underlying responding to personality

items: Tile self referent decision and response selection. Journal of Research in Personality 8, 128-138.

Rogers, T. B., in press. Sell-reference in memory: Recognition of personality items. Journal of Research in Personality.

Rogers, T. B., J. C. van der Lee and A. R. Kasper, 1975. Response statistics, desirability and ambiguity ratings for a pool of 401 personality items controlled for length. Catalogue of Selected Documents in Psychology 5,235 236.

Shaw, M. and J. Wright, 1967. Scales for tire measurement of attit6des. New York: McGraw- ttill.

Sternberg, S., 1969. The discovery of processing stages: Extensions of Donder's methods. Acta Psychologica 30, 276 315.

Trabasso, T., H. Rollins and E. Shaughnessy, 1971. Storage and verification in processing concepts. Cognitive Psychology 2, 239-289.