NeuroImage 61 (2012) 832–840

Contents lists available at SciVerse ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r .com/ locate /yn img

Neural processing of overt word generation in healthy individuals: The effect of ageand word knowledge

Arne Nagels, Tilo Kircher, Bruno Dietsche, Heidelore Backes, Justus Marquetand, Axel Krug ⁎Department of Psychiatry and Psychotherapy, Philipps-University Marburg, Rudolf-Bultmann-Str. 8, 35039 Marburg, Germany

⁎ Corresponding author. Fax: +49 6421 5868939.E-mail address: Axel.Krug@med.uni-marburg.de (A.

1053-8119/$ – see front matter © 2012 Elsevier Inc. Alldoi:10.1016/j.neuroimage.2012.04.019

a b s t r a c t

a r t i c l e i n f o

Article history:Accepted 7 April 2012Available online 13 April 2012

Keywords:Semantic verbal fluencyfMRIInferior frontal gyrusSuperior temporal gyrusACCAgeWord knowledge

Verbal fluency is a classical and widely used neuropsychological instrument to assess cognitive abilities. Re-sults of previous studies indicate an influence on verbal fluency performance of both, age and word knowl-edge. So far, no imaging study has investigated the neural mechanisms underlying an age and wordknowledge related decline on the quantitative verbal output in a highly demanding overt and continuous se-mantic fluency task.Fifty healthy volunteers (age 22–56 years, verbal IQ 95–143) overtly and continuously articulated words inresponse to ten visually presented semantic categories while BOLD signal was measured with fMRI. Verbalresponses were recorded with an MRI compatible microphone and transcribed after the scanning session.The number of produced words as well as age, word knowledge and level of education was implementedin the design matrix enabling a separate analysis of these factors on both, neural responses and behavioraldifferences.There was a significant correlation of level of education and number of generated words, but no significantcorrelations of generated words and age or word knowledge were observed. On the neural level, a wide-spread network was found for the word production task as contrasted with the resting condition, encompass-ing the bilateral superior temporal gyri, the cerebellum and the SMA. An age related positive correlation wasfound in the bilateral inferior and middle frontal gyri, the anterior cingulate gyrus, the left precentral gyrusand the right insula. A lower word knowledge resulted in enhanced BOLD responses in the right superiortemporal gyrus and the left superior frontal gyrus.Results are interpreted in terms of compensation mechanisms countervailing potential age and word knowl-edge related effects.

© 2012 Elsevier Inc. All rights reserved.

Introduction

Verbal fluency tasks represent a classical and sensitive neuropsy-chological tool to assess speech-related cognitive abilities (Burt,1917; Thurstone and Thurstone, 1941). Subjects are required toovertly articulate as many words that come to their mind accordingto given criteria and within a predetermined time window (Lezak,1995), either with regard to a semantic category (“animals”; semanticverbal fluency) or an initial letter (“initial letter A”; lexical/phonolog-ical verbal fluency). In the clinical context verbal fluency deficits arereported in neurodegenerative and psychiatric patient groups(Henry and Crawford, 2005; Kircher et al., 2008; Szoke et al., 2008;Vogel et al., 2009). For this reason it is of considerable interest to fur-ther elucidate the neural mechanisms underlying continuous overtverbal fluency (Birn et al., 2010).

Krug).

rights reserved.

Different investigations of age-related and intelligence-related ef-fects on verbal fluency performance have yielded no consistent re-sults. An age-related decline in verbal fluency performance wasreported (Brickman et al., 2005; Chan and Poon, 1999; Kempler etal., 1998; Kosmidis et al., 2004; Ratcliff et al., 1998; Van der Elst etal., 2006), particularly in semantic fluency tasks (Benito-Cuadradoet al., 2002; Tomer and Levin, 1993). Bolla and colleagues, however,found no association with age but strong effects with regard to verbalintelligence on verbal fluency performance (Bolla et al., 1990). Thereis convincing evidence that the general ability to quickly organize se-mantically related information clusters and formulate effective recallstrategies plays a critical role in verbal fluency performance. More-over, a high level of word knowledge and verbal intelligence positive-ly influence task performance, since a larger vocabulary allows for alarger pool from which potential words can be selected (Ruff et al.,1997). On the other hand, efficiently organizing and structuring alarger pool of semantically related lexical information according tospecified rules, (Lezak, 1995), additionally demands verbal workingmemory resources and enhanced control processes. In conclusion,there are no common findings with regard to socio-demographic

Table 1Mean number and standard deviations of words produced for each block.

Category Mean SD

Animals 8.84 2.37Sports 7.56 1.57Clothes 8.32 1.86Professions 6.64 1.33Fruit 7.74 1.69Vehicles 6.08 1.75Furniture 6.26 1.67Flowers 5.52 1.94Hobbies 6.24 1.78Spices 6.62 1.95

833A. Nagels et al. / NeuroImage 61 (2012) 832–840

variables and their influence on the quantitative verbal output of se-mantic word generation.

Word generation tasks were among the first paradigms to be inves-tigated with brain imaging techniques (Friston et al., 1991; Frith et al.,1991). Cuenod and colleagues utilized an imaging task asking nine par-ticipants to rest or to covertly produce words (Cuenod et al., 1995). Theauthors report task-related signal changes predominantly not only inthe left fronto-temporal regions, but also in the corresponding areasin the right hemisphere. This finding was replicated in terms of a cogni-tive model of word generation, in particular involving three cortical re-gions: the left dorsolateral frontal cortex, the temporal cortex and thestriate/extrastriate cortex (Friedman et al., 1998).

Recent studies have investigated different aspects of verbal fluen-cy, using different types of word generation tasks, e.g. lexical fluency(Fu et al., 2002, 2005a, 2005b; Kircher et al., 2011; Nagels et al., 2011)and semantic fluency (Heim et al., 2008; Kircher et al., 2011; Krug etal., 2010; Nagels et al., 2011), the majority of imaging studies investi-gated silent word production paradigms (Brazdil et al., 2005; Cuenodet al., 1995; Pujol et al., 1999; Sabbah et al., 2003). Since behavioralresponses represent an important factor to evaluate the quality ofproduced words some recent imaging studies investigated the neuralcorrelates of overt production (Birn et al., 2010; Fu et al., 2002, 2005a,2005b; Heim et al., 2008; Kircher et al., 2009; Krug et al., 2010; Nagelset al., 2011; Whitney et al., 2009). This approach more closely mirrorsthe standard behavioral test demands (Birn et al., 2010) and has beenshown to be suitable for fMRI research (Birn et al., 2010).

With regard to maturational changes in the functional organiza-tion of the cortex, age related changes were found in additionallyrecruited frontal and parietal regions during a word generation task(Brown et al., 2005). The authors argue that these ‘new’ regions canbe recruited over the course of learning, practice and skill acquisitionto support specialized control systems. A recent imaging studyrevealed increased activations in the bilateral middle temporal gyri,medial frontal gyri, middle frontal gyri as well as in the inferior fron-tal gyri in 14 older (mean age: 71 years) as compared to 14 youngerindividuals (mean age: 25 years) during a word generation task(Soros et al., 2011). Soros and colleagues interpret their imaging re-sults in terms of compensation mechanisms of age-related functionaldecline, in particular with regard to inferior frontal activations.Meinzer et al. (2009) used an overt verbal fluency design to comparedifferences in brain activation of healthy older (n=16) and youngeradults (n=16) (Meinzer et al., 2009). For the semantic word produc-tion condition the authors report an additional right inferior and mid-dle frontal activity for the older group. Wierenga and colleaguesfound a pattern of activation in the right inferior frontal region (Bro-ca's homologue) for older adults relative to younger adults during apicture naming task (Wierenga et al., 2008). The authors argue thatword-finding problems and the compromise in word retrieval arelikely due to difficulty in accessing and selecting lexical-semanticinformation.

The concrete influence of word knowledge on overt verbal outputhas not directly been investigated with fMRI, so far. Based on behav-ioral study results it can be assumed that a low verbal IQ negativelyaffects the number of produced words (Bolla et al., 1990; Ruff et al.,1997). On the neural level the inefficacy to structure and recall se-mantic information may result in the recruitment of the language-related fronto-temporal network.

In the present fMRI study, a novel continuous overt fluency taskwas applied encompassing ten different semantic categories alternat-ing with a low level resting baseline. The methodological approach al-lows for an individual a posteriori evaluation of the obtained verbalfluency data as well as a simultaneous acquisition of BOLD signalchanges. The power to detect speech-related brain activation in thecurrent task is assumed to be higher as compared to event-relatedsingle word production paradigms. The task moreover allows for fi-nite impulse response (FIR) analysis so that not only differential

task effects are averaged across a block, but also neural responsescan be analyzed individually with regard to time courses within spe-cific brain regions related to age and word knowledge. The novel se-mantic generation task design was applied in a large group ofhealthy subjects to investigate the influence of age and word knowl-edge on both, neural activations and behavioral measures using cor-relation analyses. In line with previous imaging studies enhancedneural responses in the left inferior frontal cortex are expected in el-derly individuals. With regard to word knowledge, a lower amount ofsemantic information results in the recruitment of language-relatedcore regions being generally involved in the search and recall process.

Material and methods

Subjects

Fifty-six native German speakers were recruited in Marburg, Ger-many. Exclusion criteria were neurological or psychiatric disorders, orany condition that might have an effect on cerebral metabolism or MRsafety. All subjects were tested with the SCID-I in order to exclude anypsychiatric disorders. After a complete description of the proceduresubjects provided written informed consent to participate in thestudy. The protocol was approved by the local ethics committeeaccording to the declaration of Helsinki. After analyzing MRI and be-havioral data, six subjects had to be excluded from further analyses:two subjects showed symptoms in the SCID interview, two subjectshad anatomical anomalies and for two subjects, performance datacould not be recorded due to technical problems. The resulting sub-jects were 36 men and 14 women who had a mean age of35.22 years, standard deviation (SD) 9.87, range 22–56, and an esti-mated mean verbal IQ of 121.48 (SD=13.86, range 95–143).

Behavioral tasks

In addition to the scanning procedure, all subjects underwent abrief word knowledge screening with the MWT-B (Lehrl et al.,1995). Scores were also converted into verbal IQ scores.

fMRI task

Task and stimuliStimuli were presentedwith Presentation software package (Neuro-

behavioral Systems Inc., San Francisco, CA). This semantic verbal fluen-cy task used a block designwith two alternating conditions: In responseto a German noun, subjects had to name as manymembers of the cate-gory this noun represented (e.g. say “dog”, “cat”, “eagle” … when theword “animal” was presented; word-generation condition), or theywere required to be silent when the word “silence” was presented(baseline condition). There were ten blocks for each condition. The cat-egories for the word-generation condition are displayed in Table 1. Atthe beginning of each block, an instruction slide with the categoryname was shown for 3000ms, followed by either a “+” sign in the

834 A. Nagels et al. / NeuroImage 61 (2012) 832–840

semantic verbal fluency condition or a “#” symbol in the silence condi-tion. These symbols were presented for 12 s and represented the periodof time in which either words were to be produced or subjects were tobe silent. Incorrect responses as repetitions or different grammaticalforms of previously produced words were not allowed (Lezak, 1995).Words were presented in white color on a black background and sub-jects' responses were recorded via a scanner-compatible microphoneand analyzed using Audacity 1.2.6 software. In order to estimate theparadigm's reliability individual responses for each blockwere summa-rized and Cronbach's alpha was calculated for the 10 blocks based onthe sums of produced words in each block. Prior to the scanning proce-dure a test session was performed to familiarize the participants withthe nature of the task and the rules for word generation. These stimuliwere not part of the subsequent fMRI investigation.

Data acquisition

fMRI image acquisitionImaging was performed on a 3-T Tim Trio MR scanner (Siemens

Medical Systems) in the Department of Psychiatry and Psychotherapyof the Philipps-University Marburg. Functional images were collectedwith echo planar imaging (EPI) sensitive to BOLD contrast (T2*,64×64 matrix, FoV 224 mm×224 mm, 40 slices, 3,5 mm thickness,TR=2.5 s, TE=30 ms, flipangle=90°). Slices covered the whole brainandwere positioned transaxially parallel to the anterior–posterior com-missural line (AC–PC). 130 functional images were collected, and theinitial three images were excluded from further analysis in order to re-move the influence of T1 stabilization effects. While participants lay insupine position, head movement was limited by foam paddings.

Behavioral data acquisition

The overt speech production in the scanner was captured andrecorded with a 40 dB noise-reducing microphone system (FOMRI-II, Optoacoustics Ltd.) allowing for an on-line speech synchronization.A dual adaptive filter system subtracted the reference input (MRInoise) from the source input (speech signal) and filtered the speechproduction instantly while recording the overt output. The opticfiber microphone was mounted on the head coil and wired to thesound filter box, of which the output port was directly wired to theaudio in-line plug of the notebook sound card. The audiofiles (Audac-ity, version 1.2.6, Softonic International S.L.) were saved and tran-scribed into text files. The individual number of produced words foreach block and each subject was used as a parametric regressor inthe first level analysis (cf. fMRI data analysis).

fMRI data analysis

FMRI data analyses were calculated using SPM8. Slice-time cor-rection (to the 20th slice), realignment and stereotaxic normaliza-tion (2 mm×2 mm×2 mm), an 8 mm full-width-at-half-maximum(FWHM) Gaussian smoothing kernel was applied to increase thesignal-to-noise ratio and compensate for inter-subject anatomicalvariation.

Semantic verbal fluency related brain activation was analyzed foreach subject contrasting the word-generation condition with the base-line condition and modulating the word-generation condition witheach subject's word production for each block. Because an overt seman-tic fluency task was used which could lead to increased head move-ment, realignment parameters were entered as a covariate in the first-level analysis. All subjects had tolerable head movement smaller thanone voxel size, similar to previous reports (Kircher et al., 2009; Kruget al., 2011; Nagels et al., 2011).

On the second level, a one-sample t-test was calculatedwith the first-level ‘word-generation’> ‘baseline’ contrasts. Age,word knowledge andlevel of education (ranging from 1 = no graduation to 6 = university

diploma) were entered as covariates. As there were more men thanwomen in the study, gender was entered as a covariate of no interest.Brain activations were plotted on the anatomical MRIcroN template.

In order to explore the blood oxygen level dependent (BOLD) ef-fect at the time resolution of one TR (2.5 s), a finite impulse response(FIR) model was fitted during the word-generation and the silenceconditions. On the individual subject level, 6 bins of the length ofone TR were modeled in a time window of 15 s at the onset of theword-generation and silence instructions, respectively. For furtheranalyses, resulting bins in the baseline condition were subtractedfrom the corresponding contrasts of the word-generation condition.Resulting images for each individual and time bin were entered inrandom-effects GLM with the repeated factor coding the 6 differentlevels in the time domain.

In order to clarify the potential functions of regions correlated withage orword knowledge, seed region analyseswere calculated for signif-icant clusters found in the main analysis in the contrasts for age andword knowledge. On the individual subject level seed time serieswere extracted as the first eigenvariate of the local maximum of theaccording cluster as implemented in SPM8. Task-related variance wasremoved by applying an effects-of-interest correction with the F-contrast set on the six movement parameters (Bedenbender et al.,2011; Esslinger et al., 2009; Paulus et al., 2011). The extracted time se-ries were re-entered in first-level analyses together with the headmovement parameters. Further, effects of age and word knowledgewere assessed in a random-effects second level GLM with age andword knowledge as regressors of interest. Whole brain analyses for allseed regions were conducted. The aim of this procedure was to firstidentify clusters correlating with clusters significantly associated withage or word knowledge and second to further reveal those correlationsthat were linearly affected by age and word knowledge.

For the main effect of the contrast ‘word-generation’> ‘baseline’, athreshold of pb .05 (corrected for multiple comparisons) and a clusterthreshold of 20 voxels were chosen. In order to test for the influenceof the entered covariates (age, word knowledge, level of education)and corresponding seed region analyses, a threshold of pb .001(uncorrected) and a cluster threshold of 30 voxels were chosen.

Results

Behavioral results

During scanning, subjects produced a mean number of 6.98(SD=1.12) words. The number of words for the individual trials isgiven in Table 1. An ANOVA with the ten different blocks as withinsubject factor revealed that word production differed significantly be-tween blocks as soon as the mean difference between any two blocksreached approximately 1 word (pb .05). While there was a significantcorrelation of level of education and number of words produced dur-ing the task (rlevel of education, number of words=.37, pb .01) there was noinfluence of word knowledge or age on number of produced wordsduring the scanning procedure (rword knowledge, number of words=.059,rage, number of words=− .121 all ps>.1). Even when correlating ageand word knowledge with the single blocks, there was only one sig-nificant correlation of age and number of produced words duringthe “professions” block (r=− .378, p=.007).

Based on the number of produced words, the 10 blocks had aCronbach's alpha coefficient of α=.82.

fMRI data

For the main effect ‘word-generation’> ‘baseline’, there were sig-nificant clusters of activation in the bilateral cerebellum, the bilateralsuperior temporal gyrus and the bilateral postcentral gyrus as well asleft SMA and left cuneus (activations are depicted in Fig. 1 andTable 2). FIR-analyses revealed that for the superior temporal gyrus

Fig. 1. Cortical activation for the contrast ‘word-generation’> ‘baseline’. Left side (brainimages) illustrates activations mapped on the standard MRIcroN render template.Right side depicts time course of bilateral superior temporal gyrus activation asextracted by FIR-analyses. Each number corresponds to one TR (2.5 s). Black line de-picts course of activation for the left superior temporal gyrus (STG), orange line depictscourse of activation for the right STG. The shaded gray area depicts the time of instruc-tion, error bars represent 95% CL. The contrast was thresholded at pb .05 (FWE cor-rected), cluster threshold 15 voxels. Brain activations were extracted from a spherewith a 3 mm radius for the left and right superior temporal gyrus at MNI coordinates−64; −16; 2 and 62; −18; 0, respectively.

835A. Nagels et al. / NeuroImage 61 (2012) 832–840

as well as other regions, an activation maximum was reached after 4TRs after which the activation declined.

For the covariate age, therewas a linear influence on the ‘word-gen-eration’> ‘baseline’ contrast: The older the subjects were, the higher ac-tivations in the right inferior frontal gyrus, the left anterior cingulate,left precentral gyrus/inferior frontal gyrus as well as other regions(see Fig. 2 and Table 3). Using the main effect as a small volume correc-tion, it was found that of all significant clusters only 54 voxels in the leftcerebellum (MNI coordinates −26; −58; −50) and 47 voxels in theright cerebellum (MNI coordinates 12; −74; −42) showed an overlapwith the main effect.

For the covariate word knowledge, therewas a linear effect on themain contrast: the lower the word knowledge, the higher the activa-tions in the right superior temporal gyrus and the left superior fron-tal gyrus (see Fig. 3 and Table 3). Using the main effect as a smallvolume correction, it was found that two clusters of 2 and 3 voxels

Table 2Brain areas activated as main effect for the contrast ‘word-generation’> ‘baseline’ (pb .05 F

Main effect Anatomical Region Hem.

‘Word-generation’> ‘baseline’Cerebellum L

RRRRR

Postcentral gyrus LSuperior temporal gyrus LSuperior temporal gyrus LPostcentral gyrus LPostcentral gyrus RSuperior termporal gyrus RSuperior temporal gyrus RSuperior temporal gyrus RSMA LSMA L

in the right superior temporal gyrus showed an overlap with themain effect.

For the covariate level of education, no significant clusters werefound.

For the seed region analyses of the clusters found to be correlatedwith lower word knowledge (left superior frontal gyrus and right su-perior temporal gyrus) no significant correlations with word knowl-edge were found.

The seed region analyses for clusters found to be correlated withhigher age revealed significant oscillations in cortical and cerebellarregions. For the ACC, correlating clusters were found in the bilateralSMA, the caudate nucleus and the right superior medial gyrus. Forthe right IFG, correlating clusters were found in the right precentralgyrus while for the left IFG/precentral gyrus, correlating clusterswere found in the bilateral inferior frontal gyrus, the left postcentralgyrus, the right insula and the cerebellum (see Fig. 4 and Table 4).

Discussion

In this study, the neural correlates of a newly developed overt andcontinuous semantic verbal fluency task were investigated in a largenumber of subjects. The task itself revealed a high internal consisten-cy (Cronbach's α=.82). On a neural level, the continuous overt pro-duction of semantic category members in blocks of 12 s resulted inactivations in a widespread cortical network, comprising of the bilat-eral superior temporal gyrus, the cerebellum, the bilateral postcentralgyrus as well as the SMA. While age and word knowledge had no ef-fect on task performance, higher age led to enhanced neural re-sponses mainly in the bilateral inferior frontal gyri and the left ACC.Word knowledge was negatively correlated with activations in theposterior part of the right superior temporal gyrus and the left supe-rior frontal gyrus.

Behavioral results

The number of generated words differed between the ten catego-ries. This inconsistency with regard to task performance can be as-cribed to unequal levels of difficulty between the conditions. In thisrespect, in overlearned sequences (Amunts et al., 2004) such as “ani-mals” it was comparatively easier to produce words than for the cat-egories such as “flowers”. This difference resembles the standardneuropsychological test conditions usually encompassing differentcategories with varying semantic search requirements (Amunts etal., 2004) or search space, e.g. “animals” as opposed to “professions”(Aschenbrenner et al., 2000).

WE corrected, cluster threshold 15 voxels). Coordinates are listed in MNI atlas space.

Coordinates t-value No. ofvoxels

x y z

−16 −58 −22 10.91 522814 −60 −20 10.926 −64 −50 9.9412 −70 −42 9.7736 −62 −26 9.4844 −68 −28 9.18

−44 −14 36 10.07 3652−64 −16 2 9.82−66 −22 4 9.47−56 −6 48 9.15

48 −10 38 12 300762 −18 0 8.7552 −30 4 7.7866 −28 8 7.3−2 0 68 8.07 868

−10 12 50 6.03

Fig. 2. Cortical activation as neural correlates of age during ‘word-generation’> ‘baseline’. Top row (brain images) illustrates activations mapped on the standard MRIcroN rendertemplate. Lower row depicts the time course of parameter estimates derived from clusters of activation that correlated with age in younger subjects (time courses in red) and oldersubjects (time courses in blue). Groups were established by median-split. Lower left graph shows the time courses in the right inferior frontal gyrus (IFG), middle graph shows thetime courses in the left anterior cingulate (ACC) and the right graph shows time courses of activation in the right IFG. The shaded gray area depicts the time of instruction, error barsrepresent 95% CL. Each number corresponds to one TR (2.5 s). The contrast was thresholded at pb .001 (uncorrected) and a cluster threshold of 30 voxels. Brain activations wereextracted from a sphere with a 3 mm radius at MNI coordinates 50; 36; 14 (right IFG), −10; 42; −2 (left ACC) and −48; 6; 14 (left IFG).

836 A. Nagels et al. / NeuroImage 61 (2012) 832–840

The absence of performance effects due to age or word knowledgemay be attributed to the comparatively small time window as com-pared to classical tests requiring the participants to generate as manywords as possible within a time period of 60 s. In addition, subjectshad an overall high verbal IQ and were under the age of sixty so thatsubjects were able to compensate for subtle age and word knowledgerelated effects on quantitative verbal output (cf. effect of covariates).

Another reason for the behavioral inconsistencies in comparisonto other behavioral results stems from the distinct word stimuliused resulting in varying levels of difficulty across studies (e.g. “ani-mals” (Benito-Cuadrado et al., 2002; Brucki and Rocha, 2004;Kempler et al., 1998; Mathuranath et al., 2003; Ratcliff et al., 1998;Troyer, 2000), “clothes” (Pekkala et al., 2009), “fruit” (Kave, 2005;Ratcliff et al., 1998; Troyer, 2000), “supermarket” (da Silva et al.,2004; Troyer, 2000), “transportation” (Chan and Poon, 1999), “vehi-cles” and “vegetables” (Kave, 2005), or “profession” (Van der Elst etal., 2006)), which makes it difficult to compare or generalize the indi-vidual study results. Another problem is raised by the varying number

Table 3Brain areas that correlated with the covariates “age” and “verbal IQ” during ‘word-generatioin MNI atlas space.

Covariates Anatomical Region Hem. Coo

x

AgeInferior frontal gyrus R 5Inferior frontal gyrus R 4Middle frontal gyrus R 4Middle frontal gyrus R 4Anterior cingulate cortex L −1Cerebellum L −4Anterior cingulate cortex R 1Precentral gyrus L −4Cerebellum R 1Vermis RFusiform gyrus L −4

Verbal IQSuperior temporal gyrus R 6Superior frontal gyrus L −1

of participants, ranging from 84 (Tomer and Levin, 1993) to 1856 sub-jects (Van der Elst et al., 2006) and fluctuations based on age differ-ences ranging from 54 to 99 years (Kempler et al., 1998) ascompared to 18 to 91 years (Troyer, 2000) or 85 years of age (Kave,2005), respectively. Also, aspects of language typology and culturalbackground contribute to the differences in quantitative performance(Kempler et al., 1998), e.g. Finnish words being significantly longerthan English words (Pekkala et al., 2009).

Main effect of task

In line with previous imaging studies investigating overt speechproduction as opposed to rest, a widespread network of cortical acti-vations was found in the motor cortical areas (postcentral gyri, SMA),the bilateral temporal lobes, the cerebellum and the cuneus (Fu et al.,2002; Heim et al., 2008; Kircher et al., 2011; Nagels et al., 2011;Whitney et al., 2009).

n’> ‘baseline’ (pb .001 uncorrected, cluster threshold 30 voxels). Coordinates are listed

rdinates t-value No. ofvoxels

y z

0 36 14 4.42 4492 40 −2 4.250 36 22 3.534 40 28 3.410 42 −2 4.06 2662 −70 −42 3.76 1974 32 11 3.69 948 6 14 3.91 654 −76 −46 3.61 424 −58 −34 3.56 406 −56 −20 3.86 31

6 −36 10 3.72 434 56 20 3.76 41

Fig. 3. Cortical activation as neural correlates of verbal IQ during ‘word-genera-tion’> ‘baseline’. Left side (brain image) illustrates activations mapped on the stan-dard MRIcroN render template. Right side depicts time course of the right superiortemporal gyrus (STG). Red line depicts course of activation for subjects with lowerverbal IQ, blue line depicts course of activation for subjects with higher verbal IQ.Groups were established by median-split. The shaded gray area depicts the time ofinstruction, error bars represent 95% CL. Each number corresponds to one TR(2.5 s). The contrast was thresholded at pb .001 (uncorrected), and a cluster thresh-old of 30 voxels. Brain activations were extracted from a sphere with a 3 mm radiusat MNI coordinates 66, −36, 10.

837A. Nagels et al. / NeuroImage 61 (2012) 832–840

As FIR-analyses revealed, the highest levels of task-related activa-tions were found after the fourth TR (about 10 s, see Fig. 1) afterwhich activations declined again. This pattern suggests that the cho-sen time frame of 12 s for word generation was optimal with regardto the time course of the BOLD signal. It can be hypothesized that ad-ditional time would not have led to more statistical power but rathera decrease in the mean signal. The notion of an efficient design is alsosupported by the fact that performance differed between blocks. Thissuggests that even though 12 s is a short period of time for word

Fig. 4. Results for the seed region analyses of the three prefrontal clusters found active for th(IFG) and the left IFG/precentral gyrus. Seed regions are depicted on the bottom part of the leside illustrates functional connectivity of the anterior cingulate cortex (ACC; depicted on thpart of the right side.

generation (compared to 60 s in the usual outside-the-scannertests), there was enough time to elicit differences in response to thecategories. If this period of time would have been too short, no differ-ences in performance between blocks should have emerged.

Effects of the covariates

As there was no detectable influence of age and word knowledgeon number of produced words, the neural correlates of these vari-ables are interpreted as additional compensatory activations.

AgeIn the case of age, additional brain areas which were not found to

be activated as part of the main effect (‘word generation’> ‘baseline’)were activated. These areas mainly comprised the bilateral inferiorfrontal gyrus and the left anterior cingulate. Seed region analyses ofthese three regions revealed functional coupling with other areas ofthe language network (e.g. bilateral inferior frontal cortex as wasthe case for the left IFG seed region) or motor areas (as was thecase for the ACC and the right IFG seed region). It could be arguedthat this verbal fluency task relies on working memory components:subjects have to maintain the specific category and they also have tomonitor and upkeep the words they have already produced in a spe-cific block. This process is reflected in the higher activations of the bi-lateral inferior frontal gyrus, which has been shown to be activatedhigher in older individuals as compared to younger individuals duringworking memory tasks in a recent meta-analysis (Turner and Spreng,2012). In addition, semantic word generation happened according topredetermined rules requiring the inhibition of potential response er-rors, since no repetitions of the same word or word stem wereallowed (Lezak, 1995). It is hypothesized that this highly controlledword retrieval process called for additional neural activations in fron-tal brain regions. Since semantic knowledge as such does not guaran-tee successful word retrieval, efficient strategies must be appliedallowing for information recall in a limited amount of time. In addi-tion, this process must be monitored and coordinated.

Higher activations of older subjects compared to younger subjectsin medial frontal and bilateral inferior frontal areas were also found ina recent study in which participants had to repeat a vowel or a poly-syllabic utterance (Soros et al., 2011). As these regions subserve pho-nological, semantic and syntactic processing, the authors suggest thatthis higher activation in older individuals is compensational in nature,

e age effect. Left side illustrates functional connectivity of the right inferior frontal gyrusft side, functionally connected regions are depicted on the top part of the left side. Righte bottom part of the right side), functionally connected regions are depicted on the top

Table 4Seed region analyses for the three main activations for the contrast “age” (pb .001 uncorrected, cluster threshold 30 voxels). Coordinates are listed in MNI atlas space. IFG = inferiorfrontal gyrus. STG = superior temporal gyrus. SFG = superior frontal gyrus.

Seed cluster Anatomical Region Hem. Coordinates t-value

No. ofvoxels

x y z

Covariate ageLeft IFG/Precentral gyrus Postcentral gyrus L −50 −14 26 4.58 407

Postcentral gyrus L −58 −2 18 4.53Cerebellum L −6 −46 −8 3.89 371Precentral gyrus R 54 6 26 4.18 264Inferior frontal gyrus R 52 10 22 3.96Inferior frontal gyrus R 38 2 26 3.94Inferior frontal gyrus R 52 20 22 3.78Inferior parietal lobe R 42 −16 32 3.88 52Inferior frontal gyrus L −52 24 16 4.04 42Inferior frontal gyrus L −52 18 20 4.03Insula R 48 0 −4 3.78 33

ACC SMA L −8 −2 78 5.4 671SMA L −6 −6 78 5.27Superior medial gyrus R 4 30 62 4.45SMA R 12 10 74 4.2Caudate R 14 24 2 6.15 493Caudate L −12 22 0 3.8Caudate L −16 24 0 3.76Superior parietal lobe R 22 −44 −32 4.5 304Superior frontal gyrus R 28 60 2 3.92 33Superior medial gyrus R 14 60 4 3.56Middle occipital gyrus L −24 −100 −8 3.93 32Inferior occipital gyrus L −28 −98 −10 3.67

Right IFG Precentral gyrus R 60 6 36 3.68 37Precentral gyrus R 56 −2 38 3.49

Covariate word knowledgeRight STG No significant resultsLeft SFG No significant results

838 A. Nagels et al. / NeuroImage 61 (2012) 832–840

as (in accordance with the present study) there were no behavioraldeficits associated with higher age. In addition, BOLD enhancementsin the contralateral hemisphere have been thought to reflect addi-tional neurocognitive effort in order to maintain the level of perfor-mance (Wierenga et al., 2008). In the absence of an age-relateddecline in verbal fluency performance, compensatory functions arelikely to be involved in the additional recruitment of the right inferiorfrontal region.

In a study comparing free association with semantic and phono-logical verbal fluency, Whitney et al. (2009) found activations of theACC during the latter conditions which were discussed to be elicitedby the relatively more difficult task conditions. In a study with pa-tients after acute ACC stroke, a speech disorder emerged that was at-tributable to difficulties in initiating speech (Chang et al., 2007).Given that the ACC seed region was functionally coupled with the bi-lateral SMA and the caudate nucleus, and the seed region in the rightIFG was also correlated with the right precentral gyrus and that thiscoupling was age-dependent, it can be argued that additional recruit-ment of these two areas serves speech initiation, motor planning andarticulatory processes.

This finding also strengthens the idea that highly complex pro-cesses within the “articulatory network” basically encompass the in-dividual motor phonetics of each word (phonological properties,syllabification, prosody, word stress) and the motor coordination ofappropriate phonation, which finally results in the overt articulationin the vocal apparatus (Hickok et al., 2011).

In a previous study, gray matter densities of healthy subjects werecorrelated with verbal fluency performance. It was found that therewere significant correlations of gray matter densities in the pre-SMA and the caudate and semantic verbal fluency performance(Grogan et al., 2009). Importantly, in the present study age-related ef-fects were found within these two structures. A further imaging studyreported that verbal fluency scores were higher for subjects with in-creased cortical thickness that was also correlated with higher frac-tional anisotropy (FA) scores of the arcuate fasciculus (AF) (Phillips

et al., 2011). Regions of higher cortical thickness, which correlatedwith FA of the AF included the left IFG, precentral gyrus, postcentralgyrus, middle and superior temporal gyrus as well as the supramargi-nal gyrus.

Word knowledgeActivations in the posterior part of the right STG – an area already

activated in the main contrast – and the left superior frontal gyruswere negatively correlated with word knowledge. The data suggestthat in order to compensate for lower word knowledge, it is partlysufficient to rely on the network associated with the task and recruitparts of it more strongly than recruiting additional brain areas. In astudy investigating covert verb generation in healthy children, a neg-ative correlation of IQ and activation in the right STG was also ob-served (Schmithorst and Holland, 2006) which supports the presentinterpretation. In addition to the recruitment of the right STG, activa-tions of the left superior frontal gyrus were found. It could be arguedthat subjects with lower word knowledge rely more heavily on thisregion as it has been shown to be involved in the generation andmonitoring of internally generated responses (such as category mem-bers in the present task) (for an overview see (Ramnani and Owen,2004)). In this case, subjects need to monitor and potentially rehearsethe generated material as well as the category itself more closelywhich thus activates the SFG. The influence of word knowledge onthe neural correlates of continuous overt word generation deservesfurther investigation, possibly taking into account subjects witheven lower word knowledge than included in the current study.

In sum, this new semantic verbal fluency task consisted of tenblocks of semantic verbal fluency with a high internal consistency.The neural correlates of the main effect for word-generation com-prised activations in the bilateral superior temporal gyrus, cerebel-lum, bilateral postcentral gyrus and bilateral SMA. It was found thatphases of word-generation which lasted for 12 s were sufficientwith regard to brain activation and to elicit differences in perfor-mance between categories. On a behavioral level, there were no

839A. Nagels et al. / NeuroImage 61 (2012) 832–840

effects of age or word knowledge on number of produced words (ag-gregated across all ten blocks or with individual blocks with the soleexception of age and profession, see above). On the level of brain ac-tivation, there were correlates of age and word knowledge on theprefrontal cortex (for age) and the right STG (word knowledge). Inthe case of age, new areas not found to be activated as part of themain effect were additionally recruited in order to maintain task per-formance. Seed voxel analyses revealed that these additionallyrecruited areas were also functionally coupled in an age-dependentmanner with pre-motor areas that serve speech initiation, motorplanning and articulatory processes. This hints to potential difficultiesin speech initiation in older subjects. In the case of word knowledge,the right STG and the left SFG were found to be activated stronger.The right STG was part of the main effect and it is suggested thatstronger recruitment of areas already involved in the task is sufficientin order to maintain a high level of task performance as neither theSTG nor the SFG was functionally coupled with other brain areas ina word knowledge dependent fashion. Because of the different effectsof age and word knowledge on brain function it can be argued thatthis task is more sensitive in detecting subtle influences of certainvariables that might not be detectable in overt behavior. In conclu-sion, this task could be useful in investigating clinical populationssuch as neurological or psychiatric patients.

Acknowledgments

The authors report no conflict of interest. This study was sup-ported by the Deutsche Forschungsgemeinschaft (German ResearchFoundation; grant no. KR 3822/2-1).

References

Amunts, K., Weiss, P.H., Mohlberg, H., Pieperhoff, P., Eickhoff, S., Gurd, J.M., Marshall,J.C., Shah, N.J., Fink, G.R., Zilles, K., 2004. Analysis of neural mechanisms underlyingverbal fluency in cytoarchitectonically defined stereotaxic space—the roles ofBrodmann areas 44 and 45. NeuroImage 22, 42–56.

Aschenbrenner, S., Tucha, O., Lange, K., 2000. Regensburger Wortflüssigkeits-Test.Hogrefe, Göttingen.

Bedenbender, J., Paulus, F.M., Krach, S., Pyka, M., Sommer, J., Krug, A., Witt, S.H.,Rietschel, M., Laneri, D., Kircher, T., Jansen, A., 2011. Functional connectivity ana-lyses in imaging genetics: considerations on methods and data interpretation.PLoS One 6, e26354.

Benito-Cuadrado, M.M., Esteba-Castillo, S., Bohm, P., Cejudo-Bolivar, J., Pena-Casanova,J., 2002. Semantic verbal fluency of animals: a normative and predictive study in aSpanish population. J. Clin. Exp. Neuropsychol. 24, 1117–1122.

Birn, R.M., Kenworthy, L., Case, L., Caravella, R., Jones, T.B., Bandettini, P.A., Martin, A.,2010. Neural systems supporting lexical search guided by letter and semantic cat-egory cues: a self-paced overt response fMRI study of verbal fluency. NeuroImage49, 1099–1107.

Bolla, K.I., Lindgren, K.N., Bonaccorsy, C., Bleecker, M.L., 1990. Predictors of verbal flu-ency (FAS) in the healthy elderly. J. Clin. Psychol. 46, 623–628.

Brazdil, M., Chlebus, P., Mikl, M., Pazourkova, M., Krupa, P., Rektor, I., 2005. Reorganiza-tion of language-related neuronal networks in patients with left temporal lobe ep-ilepsy — an fMRI study. Eur. J. Neurol. 12, 268–275.

Brickman, A.M., Paul, R.H., Cohen, R.A., Williams, L.M., MacGregor, K.L., Jefferson, A.L.,Tate, D.F., Gunstad, J., Gordon, E., 2005. Category and letter verbal fluency acrossthe adult lifespan: relationship to EEG theta power. Arch. Clin. Neuropsychol. 20,561–573.

Brown, T.T., Lugar, H.M., Coalson, R.S., Miezin, F.M., Petersen, S.E., Schlaggar, B.L., 2005.Developmental changes in human cerebral functional organization for word gen-eration. Cereb. Cortex 15, 275–290.

Brucki, S.M., Rocha, M.S., 2004. Category fluency test: effects of age, gender and educa-tion on total scores, clustering and switching in Brazilian Portuguese-speakingsubjects. Braz. J. Med. Biol. Res. 37, 1771–1777.

Burt, C., 1917. The Distribution and Relations of Educational Abilities. London CountyCouncil, London.

Chan, A.S., Poon, M.W., 1999. Performance of 7- to 95-year-old individuals in a Chineseversion of the category fluency test. J. Int. Neuropsychol. Soc. 5, 525–533.

Chang, C.C., Lee, Y.C., Lui, C.C., Lai, S.L., 2007. Right anterior cingulate cortex infarctionand transient speech aspontaneity. Arch. Neurol. 64, 442–446.

Cuenod, C.A., Bookheimer, S.Y., Hertz-Pannier, L., Zeffiro, T.A., Theodore, W.H., Le Bihan, D.,1995. Functional MRI during word generation, using conventional equipment: a poten-tial tool for language localization in the clinical environment. Neurology 45, 1821–1827.

da Silva, C.G., Petersson, K.M., Faisca, L., Ingvar, M., Reis, A., 2004. The effects of literacyand education on the quantitative and qualitative aspects of semantic verbal fluen-cy. J. Clin. Exp. Neuropsychol. 26, 266–277.

Esslinger, C., Walter, H., Kirsch, P., Erk, S., Schnell, K., Arnold, C., Haddad, L., Mier, D.,Opitz von Boberfeld, C., Raab, K., Witt, S.H., Rietschel, M., Cichon, S., Meyer-Lindenberg, A., 2009. Neural mechanisms of a genome-wide supported psychosisvariant. Science 324, 605.

Friedman, L., Kenny, J.T., Wise, A.L., Wu, D., Stuve, T.A., Miller, D.A., Jesberger, J.A.,Lewin, J.S., 1998. Brain activation during silent word generation evaluated withfunctional MRI. Brain Lang. 64, 231–256.

Friston, K.J., Frith, C.D., Liddle, P.F., Frackowiak, R.S., 1991. Investigating a network modelof word generationwith positron emission tomography. Proc. Biol. Sci. 244, 101–106.

Frith, C.D., Friston, K.J., Liddle, P.F., Frackowiak, R.S., 1991. A PET study of word finding.Neuropsychologia 29, 1137–1148.

Fu, C.H., Morgan, K., Suckling, J., Williams, S.C., Andrew, C., Vythelingum, G.N., McGuire,P.K., 2002. A functional magnetic resonance imaging study of overt letter verbalfluency using a clustered acquisition sequence: greater anterior cingulate activa-tion with increased task demand. NeuroImage 17, 871–879.

Fu, C.H., Abel, K.M., Allin, M.P., Gasston, D., Costafreda, S.G., Suckling, J., Williams, S.C.,McGuire, P.K., 2005a. Effects of ketamine on prefrontal and striatal regions in anovert verbal fluency task: a functional magnetic resonance imaging study. Psycho-pharmacology (Berl) 183, 92–102.

Fu, C.H., Suckling, J., Williams, S.C., Andrew, C.M., Vythelingum, G.N., McGuire, P.K.,2005b. Effects of psychotic state and task demand on prefrontal function in schizo-phrenia: an fMRI study of overt verbal fluency. Am. J. Psychiatry 162, 485–494.

Grogan, A., Green, D.W., Ali, N., Crinion, J.T., Price, C.J., 2009. Structural correlates of se-mantic and phonemic fluency ability in first and second languages. Cereb. Cortex19, 2690–2698.

Heim, S., Eickhoff, S.B., Amunts, K., 2008. Specialisation in Broca's region for semantic,phonological, and syntactic fluency? NeuroImage 40, 1362–1368.

Henry, J.D., Crawford, J.R., 2005. Ameta-analytic reviewof verbalfluency deficits in schizo-phrenia relative to other neurocognitive deficits. Cogn. Neuropsychiatry 10, 1–33.

Hickok, G., Houde, J., Rong, F., 2011. Sensorimotor integration in speech processing:computational basis and neural organization. Neuron 69, 407–422.

Kave, G., 2005. Phonemic fluency, semantic fluency, and difference scores: normativedata for adult Hebrew speakers. J. Clin. Exp. Neuropsychol. 27, 690–699.

Kempler, D., Teng, E.L., Dick, M., Taussig, I.M., Davis, D.S., 1998. The effects of age, edu-cation, and ethnicity on verbal fluency. J. Int. Neuropsychol. Soc. 4, 531–538.

Kircher, T., Whitney, C., Krings, T., Huber, W., Weis, S., 2008. Hippocampal dysfunctionduring free word association in male patients with schizophrenia. Schizophr. Res.101, 242–255.

Kircher, T., Krug, A., Markov, V., Whitney, C., Krach, S., Zerres, K., Eggermann, T.,Stocker, T., Shah, N.J., Treutlein, J., Nothen, M.M., Becker, T., Rietschel, M., 2009. Ge-netic variation in the schizophrenia-risk gene neuregulin 1 correlates with brainactivation and impaired speech production in a verbal fluency task in healthy indi-viduals. Hum. Brain Mapp. 30, 3406–3416.

Kircher, T., Nagels, A., Kirner-Veselinovic, A., Krach, S., 2011. Neural correlates of rhym-ing vs. lexical and semantic fluency. Brain Res. 1391, 71–80.

Kosmidis, M.H., Vlahou, C.H., Panagiotaki, P., Kiosseoglou, G., 2004. The verbal fluencytask in the Greek population: normative data, and clustering and switching strate-gies. J. Int. Neuropsychol. Soc. 10, 164–172.

Krug, A., Nieratschker, V., Markov, V., Krach, S., Jansen, A., Zerres, K., Eggermann, T.,Stocker, T., Shah, N.J., Treutlein, J., Muhleisen, T.W., Kircher, T., 2010. Effect of CAC-NA1C rs1006737 on neural correlates of verbal fluency in healthy individuals. Neu-roImage 49, 1831–1836.

Krug, A., Markov, V., Krach, S., Jansen, A., Zerres, K., Eggermann, T., Stocker, T., Shah, N.J.,Nothen, M.M., Georgi, A., Strohmaier, J., Rietschel, M., Kircher, T., 2011. Genetic var-iation in G72 correlates with brain activation in the right middle temporal gyrus ina verbal fluency task in healthy individuals. Hum. Brain Mapp. 32, 118–126.

Lehrl, S., Triebig, G., Fischer, B., 1995. Multiple choice vocabulary test MWT as a validand short test to estimate premorbid intelligence. Acta Neurol. Scand. 91, 335–345.

Lezak, M.D., 1995. Neuropsychological Assessment, 3 ed. Oxford University Press.Mathuranath, P.S., George, A., Cherian, P.J., Alexander, A., Sarma, S.G., Sarma, P.S., 2003.

Effects of age, education and gender on verbal fluency. J. Clin. Exp. Neuropsychol.25, 1057–1064.

Meinzer, M., Flaisch, T., Wilser, L., Eulitz, C., Rockstroh, B., Conway, T., Gonzalez-Rothi,L., Crosson, B., 2009. Neural signatures of semantic and phonemic fluency in youngand old adults. J. Cogn. Neurosci. 21, 2007–2018.

Nagels, A., Kirner-Veselinovic, A., Krach, S., Kircher, T., 2011. Neural correlates of S-ketamine induced psychosis during overt continuous verbal fluency. NeuroImage54, 1307–1314.

Paulus, F.M., Krach, S., Bedenbender, J., Pyka, M., Sommer, J., Krug, A., Knake, S., Nöthen,M.M., Witt, S.H., Rietschel, M., Kircher, T., Jansen, A., 2011. Partial support forZNF804A genotype-dependent alterations in prefrontal connectivity. Hum. BrainMapp. http://dx.doi.org/10.1002/hbm.21434 (Electronic publication ahead of print).

Pekkala, S., Goral, M., Hyun, J., Obler, L.K., Erkinjuntti, T., Albert, M.L., 2009. Semanticverbal fluency in two contrasting languages. Clin. Linguist. Phon. 23, 431–445.

Phillips, O.R., Clark, K.A., Woods, R.P., Subotnik, K.L., Asarnow, R.F., Nuechterlein, K.H.,Toga, A.W., Narr, K.L., 2011. Topographical relationships between arcuate fascicu-lus connectivity and cortical thickness. Hum. Brain Mapp. 32, 1788–1801.

Pujol, J., Deus, J., Losilla, J.M., Capdevila, A., 1999. Cerebral lateralization of language innormal left-handed people studied by functional MRI. Neurology 52, 1038–1043.

Ramnani, N., Owen, A.M., 2004. Anterior prefrontal cortex: insights into function fromanatomy and neuroimaging. Nat. Rev. Neurosci. 5, 184–194.

Ratcliff, G., Ganguli, M., Chandra, V., Sharma, S., Belle, S., Seaberg, E., Pandav, R., 1998.Effects of literacy and education on measures of word fluency. Brain Lang. 61,115–122.

Ruff, R.M., Light, R.H., Parker, S.B., Levin, H.S., 1997. The psychological construct of wordfluency. Brain Lang. 57, 394–405.

840 A. Nagels et al. / NeuroImage 61 (2012) 832–840

Sabbah, P., Chassoux, F., Leveque, C., Landre, E., Baudoin-Chial, S., Devaux, B., Mann, M.,Godon-Hardy, S., Nioche, C., Ait-Ameur, A., Sarrazin, J.L., Chodkiewicz, J.P.,Cordoliani, Y.S., 2003. Functional MR imaging in assessment of language domi-nance in epileptic patients. NeuroImage 18, 460–467.

Schmithorst, V.J., Holland, S.K., 2006. Functional MRI evidence for disparate developmen-tal processes underlying intelligence in boys and girls. NeuroImage 31, 1366–1379.

Soros, P., Bose, A., Sokoloff, L.G., Graham, S.J., Stuss, D.T., 2011. Age-related changes inthe functional neuroanatomy of overt speech production. Neurobiol. Aging 32,1505–1513.

Szoke, A., Trandafir, A., Dupont, M.E., Meary, A., Schurhoff, F., Leboyer, M., 2008. Longitudinalstudies of cognition in schizophrenia: meta-analysis. Br. J. Psychiatry 192, 248–257.

Thurstone, L.L., Thurstone, T.G., 1941. Factorial studies of intelligence. PsychometricMonographs.

Tomer, R., Levin, B.E., 1993. Differential effects of aging on two verbal fluency tasks.Percept. Mot. Skills 76, 465–466.

Troyer, A.K., 2000. Normative data for clustering and switching on verbal fluency tasks.J. Clin. Exp. Neuropsychol. 22, 370–378.

Turner, G.R., Spreng, R.N., 2006. Executive functions and neurocognitive aging: disso-ciable patterns of brain activity. Neurobiol. Aging 33 (4), 826.e1–e13 (Electronicpublication ahead of print 2011 Jul 24).

Van der Elst, W., Van Boxtel, M.P., Van Breukelen, G.J., Jolles, J., 2006. Normative datafor the Animal, Profession and Letter M Naming verbal fluency tests for Dutchspeaking participants and the effects of age, education, and sex. J. Int. Neuropsy-chol. Soc. 12, 80–89.

Vogel, A.P., Chenery, H.J., Dart, C.M., Doan, B., Tan, M., Copland, D.A., 2009. Verbal fluen-cy, semantics, context and symptom complexes in schizophrenia. J. Psycholinguist.Res. 38 (5), 459–473 (Electronic publication ahead of print 2009 Mar 4).

Whitney, C., Weis, S., Krings, T., Huber, W., Grossman, M., Kircher, T., 2009. Task-depen-dent modulations of prefrontal and hippocampal activity during intrinsic wordproduction. J. Cogn. Neurosci. 21, 697–712.

Wierenga, C.E., Benjamin, M., Gopinath, K., Perlstein, W.M., Leonard, C.M., Rothi, L.J.,Conway, T., Cato, M.A., Briggs, R., Crosson, B., 2008. Age-related changes in wordretrieval: role of bilateral frontal and subcortical networks. Neurobiol. Aging 29,436–451.