12 pages

Aging affects both perceptual and lexical/semantic components of word stem priming: An event-related fMRI study

of 12
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Aging affects both perceptual and lexical/semantic components of word stem priming: An event-related fMRI study
  Neurobiology of Learning and Memory 83 (2005) 251–$ - see front matter ©  2005 Elsevier Inc. All rights reserved.doi:10.1016/j.nlm.2005.01.005 Aging a V  ects both perceptual and lexical/semantic components of word stem priming: An event-related fMRI study Sander M. Daselaar a, ¤ , Dick J. Veltman b , Serge A.R.B. Rombouts c , Jeroen G.W. Raaijmakers d , Cees Jonker e a Center for Cognitive Neuroscience, Duke University, Box 90999, LSRC Bldg., Rm B243N, Durham, NC 27708, USA b Department of Psychiatry, Clinical PET-center, “Vrije Universiteit” Medical Center, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands c Department of Physics and Medical Technology/Neurology/Alzheimer Center VUMC, “Vrije Universiteit” Medical Center, De Boelelaan 1117, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands d Department of Psychology, University of Amsterdam, Roetersstraat 15, 1018 WB Amsterdam, The Netherlands e Institute for Research in Extramural Medicine, “Vrije Universiteit” Medical Center, vd Boechorstraat 7, 1081 BT Amsterdam, The Netherlands Received 13 January 2004; revised 4 August 2004; accepted 26 January 2005 Abstract In this event-related fMRI study, brain activity patterns were compared in extensive groups of young ( N  D 25) and older ( N  D 38)adults, while they were performing a word stem completion priming task. Based on behavioral W ndings, we tested the hypothesis thataging a V  ects only the lexical/semantic, but not the perceptual component of word stem priming. To this end, we distinguishedbetween priming-related activity reductions in posterior regions involved in visual processing, and regions associated with lexical/semantic retrieval processes, i.e., left lateral temporal and left prefrontal regions. Both groups revealed signi W cant priming-relatedresponse time reductions. However, in accordance with earlier W ndings, a larger priming e V  ect was found in the group of young par-ticipants. In line with previous imaging studies, the groups showed common priming-related activity reductions in the anterior cingu-late, and the left inferior prefrontal cortex extending into the anterior portion of the left superior temporal gyrus, and at lowerthresholds also in the right occipital lobe. However, when directly comparing the groups, greater priming-related reductions werefound for the young group in the left anterior superior temporal gyrus and the right posterior occipital lobe. These W ndings suggestthat, converse to current psychological views, aging a V  ects both perceptual and lexical/semantic components of repetition priming. ©  2005 Elsevier Inc. All rights reserved. 1. Introduction Aging is accompanied by marked de W cits in declara-tive memory, the conscious and intentional recollectionof episodes and facts (Cabeza, Nyberg, & Park, 2005).Although the precise relationship between brain atrophyand functional decline remains unclear (e.g., Johnsonetal., 2000), these de W cits are thought to be related toage-related volume reductions in structures that areassumed to be critical to declarative memory function,including the medial temporal and frontal lobes(Golomb et al., 1993; Haug et al., 1983; Raz et al., 1997; Tisserand, Visser, van Boxtel, & Jolles, 2000). Interest-ingly, aging e V  ects on nondeclarative types of learningand memory, such as repetition priming, appear to beless pronounced (Cabeza et al., 2005). This suggests thatthe structures responsible for these types of learning arerelatively spared by the aging process, and that brainregions that mediate declarative memory are not con-tributing to priming e V  ects.Repetition priming refers to the phenomenon thatresponse to a stimulus is faster, more accurate, and/orbiased when this stimulus has been processed previously.The most widely used priming test has been the wordstem completion (WSC) task, which is composed of a * Corresponding author. Fax: +919 681 0815. E-mail address: (S.M. Daselaar).  252 S.M. Daselaar et al. / Neurobiology of Learning and Memory 83 (2005) 251–262 study and a test phase. At study, participants are pre-sented with a list of target words, usually incorporated ina particular encoding task to be performed. Then, at test,the participants are visually presented with two or threeletter word stems under the instruction to complete thesestems with the W rst word that comes to mind. In general,half of the word stems are taken from the study list, theother half are extracted from a control list not presentedat study. Implicit learning is indicated by the di V  erencebetween the number of stems completed with targetwords relative to control words.Neuroimaging studies on word stem priming havereported priming-related activity changes, typicallyreductions, in visual brain regions (Backman, Almkvist,Andersson, & Nordberg, 1997; Badgaiyan & Posner,1997; Buckner, Koutstaal, Schacter, & Rosen, 2000;Buckner & Miezin, 1995; Thiel, Henson, Morris, Friston,& Dolan, 2001), and in regions generally assumed to beinvolved in lexical/semantic retrieval processes, i.e., theleft lateral temporal and left prefrontal cortex (Buckneret al., 2000; Thiel et al., 2001). These activity reductions are assumed to re X ect enhanced e Y ciency of repeatedprocessing resulting in diminished computationaldemands. Interestingly, Buckner and colleagues did notobserve reductions in posterior visual areas (BA 17/18)when the word stems were aurally cued, whereas thepriming reductions in the temporal and prefrontal cortexwere still present. These W ndings suggest that word stemcompletion priming consists of a perceptual, modality-speci W c component and a lexical/semantic component,independent of stimulus modality (Buckner et al., 2000). Other evidence for the involvement of lexical/semanticprocesses in word stem priming has come from studiesshowing a greater priming e V  ect with additional lexicalor semantic processing during the study phase (Challis &Brodbeck, 1992).Behavioral studies have generally indicated a modestage-related reduction in WSC priming (Chiarello &Hoyer, 1988; Hultsch, Masson, & Small, 1991; Winocur, Moscovitch, & Stuss, 1996). It has been suggested thataging a V  ects only the lexical/semantic component of word stem priming and that perceptual priming is stillintact (Rybash, 1996). Several W ndings concur with thishypothesis. Winocur et al. found a correlation betweenpriming magnitude on the WSC task and frontal lobetest performance in both community-dwelling and insti-tutionalized elderly adults. This W nding indicates thatfrontal lobe dysfunction is a factor in reduced primingon the WSC test (Winocur et al., 1996). In addition, age- related volume decreases are generally greater in frontaland temporal lobes compared to perceptual regions(Golomb et al., 1993; Good et al., 2001; Haug et al., 1983; Tisserand et al., 2000). Finally, age di V  erences in primingare more reliably observed on priming tests that areviewed as mainly conceptual (e.g., word associationpriming) than on tests that are regarded as mostly per-ceptual (e.g., perceptual identi W cation priming) (Rybash,1996). Still, it has been di Y cult to distinguish betweenperceptual and lexical/semantic priming componentsusing behavioral paradigms.In the present functional MRI study (fMRI), weexamined to what extent aging a V  ects perceptual andlexical/semantic components of WSC priming by com-paring brain activity patterns in extensive groups of young and elderly adults while they were performingan adapted version of the WSC task. We used (self-paced) event-related fMRI, which permits randomiza-tion of the presentation order of primed and controlword stems, thereby preserving the implicit nature of the task. We tested the hypothesis, based on behavioral W ndings, that young compared to old participants willshow greater priming-related activity reductions inregions associated with lexical/semantic retrieval, but asimilar reduction in modality-speci W c visual regions,re X ecting spared perceptual priming in the elderlygroup. 2. Methods  2.1. Participants Twenty- W ve right-handed males between the ages of 30 and 35 and 38 right-handed males between the ages of 63 and 71 participated. They were recruited by means of advertisements in local newspapers. None of the partici-pants were taking psychoactive medication and they didnot report any neurological or psychiatric impairmenton a general health questionnaire. We used a cut-o V   of 25 (out of 30) or higher on the Mini Mental StatusExamination (MMSE; (Folstein, Folstein, & McHugh,1975), a common test for assessing cognitive compe-tence. Mean MMSE score was 27.8 [ SD D 1.5]) with onlytwo older adults having a score of 25. In addition, struc-tural MR images (MPRAGE: inversion time: 300ms,TR D 15ms, TE D 7ms, X ip angle D 8°), which wereacquired previous to this study, did not reveal indica-tions for anatomical deviations that were uncharacteris-tic for this age group. All participants had completed atleast the lowest level of high school in the Netherlands,which corresponds to level 4 on a 7-point Dutch educa-tional scale (1 D “no education,” 7 D “university level”).All elderly participants, and most of the young partici-pants (19 out of 25) had participated in a previous fMRIstudy of memory (Daselaar, Rombouts, Veltman, Raaij-makers, & Jonker, 2003a) and visited the outpatientclinic for the second time. The participants’ informedconsent was obtained according to the declaration of Helsinki and approved by the ethical committee of the“Vrije Universiteit” Medical Center. Demographic dataand self-rated health measures are shown in Table 1 forthe di V  erent groups.  S.M. Daselaar et al. / Neurobiology of Learning and Memory 83 (2005) 251–262 253  2.2. Magnetic resonance procedures Imaging was performed on a 1.5T Siemens Sonata(Siemens, Erlangen, Germany) scanner using a standardcircularly polarized head coil. Stimuli were generated bya Pentium PC and projected on a screen at the back endof the scanner table. The projected image was seen via amirror placed above the subject’s head. Two magnet-compatible four-key response boxes were used to recordthe subject’s responses and reaction times. The subject’shead was immobilized using foam pads to reduce motionartifact and earplugs were used to moderate scannernoise. For each subject, a series of echo planar images(EPI) was acquired sensitive to BOLD contrast, involv-ing a T2*-weighted gradient echo sequence (RepetitionTime D 2.1s, Echo Time D 60ms, X ip angle D 90°) con-sisting of 293 transversal whole-brain acquisitions (20slices, 3 £ 3mm 2  in-plane resolution, 5mm slice thick-ness, 1mm interslice gap).  2.3. Word stem completion task  Two lists of 80 nouns were drawn from a standardDutch dictionary with word lengths ranging from W ve tonine letters. One set of nouns represented animateobjects (e.g., “chicken,” “barber,” “begonia”), the otherset referred to inanimate objects (e.g., “pencil,” “cos-tume,” “bumper”). Word stems were constructed byextracting the W rst three letters of these nouns. We madesure that at least W ve and at most 10 alternative wordscould be created by completing the stems as indicated inthe dictionary. Next, the nouns from the two lists wererandomly assigned to be used either as primes or controlwords, yielding di V  erent stimulus sets for each individualsubject (see below).In view of possible motion artifacts, we used anadapted version of the WSC task in which the wordstems were completed in silence followed by a button-press. The WSC task included eight study blocks, con-sisting of 10 trials, and eight WSC blocks, consisting of 20 trials, which alternated throughout the scan session.At study, participants were presented with a noun in themiddle of the screen. They were instructed to indicatewhether the word represented a living (left-handed press)or a nonliving object (right-handed press) using theirindex W ngers. On each trial, response options were indi-cated at the bottom of the screen by two cursors point-ing to the left (“living”) and right (“nonliving”).During a subsequent word stem completion block,participants were randomly presented with three letterword stems. They were instructed to silently  completethe stems with the W rst word that came to mind, and sub-sequently, to press a right-hand button as fast as possibleusing their index W nger as soon as the word stem wassuccessfully completed. Half of the word stems camefrom the previous study block; the other half were takenfrom the control list. A measure of implicit learning wasobtained by comparing the response times for primedand control word stems, expecting the W rst to be reducedrelative to the latter. However, as an additional manipu-lation check, the group of young participants wasrequired to complete the entire list of word stems againin writing immediately after the scan session had ended.The study and test blocks were separated by a 2-sinstruction screen (i.e., “LIVING/NONLIVING”;“WORD STEM COMPLETION”). Stimuli were pre-sented in a self-paced fashion, although a time limit of 5swas applied in case of nonresponses. For the studyblocks, the response-stimulus interval (RSI) was W xed at1s, for the WSC blocks, at 2s. Hence, the intertrial inter-vals for the WSC blocks constituted 2s RSI+theresponse time. Trials were only included when partici-pants responded within the 5-s time limit. The number of functional scans was W xed at 293 in order to meet theequal variance assumption intrinsic to random e V  ectsanalyses (see the following section).  2.4. Data analysis Data were analyzed using SPM99 (Wellcome Depart-ment of Cognitive Neurology, http://www. W After discarding the W rst three volumes, time-series were corrected for di V  erences in slice acquisitiontimes, and realigned. Next, the EPI volumes were spa-tially normalized into approximate Talairach and Tour-noux space (Talairach & Tournoux, 1988) de W ned by astandard SPM EPI template and resliced to a resolutionof 3 £ 3 £ 3mm voxels. Data were spatially smoothedusing a Gaussian kernel of 8mm and normalized forglobal e V  ects using proportional scaling.The event-related fMRI analysis was based upon theassumption that individual hemodynamic responsessummate in a practically linear fashion over time (Boyn-ton, Engel, Glover, & Heeger, 1996; Buckner, 1998; Dale & Buckner, 1997). First, the response times (RT) of eachindividual subject were checked for outliers. The WSCtrials which resulted in a RT exceeding the set cut-o V  ( D percentile[75] ¡ 1.5 £ percentile[25]+percentile[50]),were excluded from the analyses. In the young partici-pants, this resulted on average in exclusion of 9.97%( SD D 4.50) and 9.70% ( SD D 3.33) of the primed and ae Demographic data and self-rated health[ SD ]Young ( N  D 25)Elderly ( N  D 38)Age32.3 [1.8]66.4 [2.0]Education (7-point scale)6.0 [1.0]5.5 [0.8]Self-rated physical health (1 D bad, 5 D excellent)4.0 [0.6]4.0 [0.6]Self-rated psychological health (1 D bad, 5 D excellent)4.1 [0.7]4.3 [0.7]  254 S.M. Daselaar et al. / Neurobiology of Learning and Memory 83 (2005) 251–262 control trials, respectively. In the older group, thisresulted in exclusion of 9.98% ( SD D 4.39) of the primedtrials and 8.09% ( SD D 3.91) of the control trials. Next,after applying a 40-s high-pass W lter, evoked hemody-namic responses to the WSC events and the study blockswere modeled, respectively, as delta functions and boxcar functions (mixed epoch/event-related design), whichwere convolved with a synthetic hemodynamic responsefunction in the context of the general linear model(Josephs, Turner, & Friston, 1997). We assessed averageactivations across participants by carrying out a two-step random e V  ects analysis (Woods, 1996). The numberof scans, and thereby the number of degrees of freedom,was held equal across participants, because this type of analysis assumes approximately equal within-subjectvariances. Hence, the number of trials completed variedacross participants due to the self-paced design. The fea-sibility of using self-paced event-related fMRI designshas been discussed in previous papers (Daselaar et al.,2001; Josephs et al., 1997; Maccotta, Zacks, & Buckner,2001).As a W rst step, speci W c e V  ects were tested by applyingappropriate contrasts to the parameter estimates foreach event, giving a t  statistic for every voxel. In thisway, SPMs (Statistical Parametric Maps; Friston, Frith,Frackowiak, & Turner, 1995) were assessed for eachindividual subject. During the second step, we carriedout a one-sample t  test upon the resulting contrastimages in order to assess the main e V  ects for each group.Group comparisons were carried out by means of a two-sample t  test without constant term, which allows inclu-sive masking of the main e V  ects of either one of thegroups, thereby controlling for possible deactivationsassociated with the opposite main e V  ects (see below).Based on the results of previous neuroimaging studies(Backman et al., 1997), we only tested for signi W cantpriming-related activity reductions, i.e., control vs.primed word stems. Furthermore, the groups weredirectly compared to each other on this contrast. Weapplied an FDR-corrected (Genovese, Nichols, & Lazar,2002) threshold of  p <.05 with a minimum cluster size of 5 voxels for the assessment of group averages. The areassurviving this threshold in either one of the groups weresubsequently identi W ed as regions of interest for theregression analysis described below. In order to increasethe sensitivity for detecting group di V  erences, we applieda slightly lower threshold (  p <.001 uncorrected, with acluster size D 10) for the assessment of group interac-tions, which were masked inclusively by the main e V  ect(control vs. primed;  p <.005, uncorrected) of the respec-tive groups.Additionally, we applied a simple regression model tothe data for each group using percent response timereductions of the individual participants as a covariate.This analysis was also set at a lower threshold of  p <.001,uncorrected with a minimum cluster size of 10 adjacentvoxels. The rationale for the regression analysis was two-fold; W rst, we wanted to examine the extent to whichbehavioral and neurofunctional measures were correlat-ing in general, and second, whether perceptual and lexi-cal/semantic regions were di V  erentially contributing tothe magnitude of the priming e V  ect in the young andelderly groups. 3. Results 3.1. Behavioral results The performance measures on the word stem comple-tion task are summarized in Table 2. Both the young( t (24) D 8.5;  p <.001) and the elderly group ( t (37) D 7.9;  p <.001) showed a signi W cant priming e V  ect as re X ectedin the percent reduction in RTs for primed vs. controlword stems. However, the priming e V  ect was larger forthe young participants ( t (61) D 4.7;  p <.001). Surpris-ingly, the elderly participants were signi W cantly fasterthan the young participants in completing the controlword stems ( t (61) D 2.124,  p D .038). Consequently, thenumber of completed word stems was signi W cantlyhigher (primed: t (61) D 2.995,  p D .004; control: t (61) D 2.482,  p D .016) for the old (primed D 64.0, SD D 6.6; control D 64.0, SD D 6.7) than for the youngadults (primed D 59.2, SD D 5.6; control D 59.7; SD D 5.7).The faster performance in the older participants mayindicate that they had a larger vocabulary at their dis-posal. Alternatively, though, this pattern of results couldimply that the elderly were performing the covert com-pletion task di V  erently than the young participants,which could also have contributed to the observed groupdi V  erence in the priming e V  ect. For instance, someelderly may have simply pressed a button without gener-ating a word on some trials. However, such a responsepattern would predict either greater variability in the oldgroup, in the case that only a subset of the older partici-pants were button-pressing without generation, orgreater within-subject variance in the older adults in thecase that in most participants only some of the trialswere button-pressed without generation.In order to test the W rst possibility, we calculated thevariance ratio, which indicated signi W cantly greater vari-ance in the priming e V  ect (i.e., % RT reductions) for theyoung participants (variance ratio [young/old] D 2.97, F  (24,37) D 3.09;  p D .004), indicating greater similarity in Table 2Behavioral results[ SD ]YoungElderlyRT primed word stems (s)1.31 [.40]1.22 [.44]RT control word stems (s)1.61 [.43]1.35 [.49]Priming e V  ect: % RT18.7 [11.2]8.3 [6.5]  S.M. Daselaar et al. / Neurobiology of Learning and Memory 83 (2005) 251–262 255 the response patterns of the older adults. In order to testthe second alternative, we assessed the coe Y cient of vari-ance (COV D average/standard deviation) in the RTs foreach subject, which is a measure of within-subject vari-ability. Again, the COV was found to be greater for theyoung participants both for the primed (COV-young D 0.430, SD D 0.099; COV-old D 0.344, SD D 0.131; t (61) D 2.961;  p D .004) and the control wordstems (COV-young D 0.459, SD D 0.339; COV-old D 0.344, SD D 0.131; t (61) D 3.994;  p <.001), indicat-ing that the elderly actually showed a more consistent,rather than a more variable, response pattern. In sum,the fact that the older adults showed clear priming e V  ectsas re X ected in a clear reduction in the RTs, together withthe W nding of smaller variability in the old group’sresponse patterns, does not indicate that the older adultswere performing the task inappropriately.Still, the di V  erence in baseline completion times com-bined with the observed di V  erence in RT reductionscould imply that the priming e V  ects were nonlinear, forinstance, as a result of a ceiling e V  ect. However, inthat case, one would expect the slowest participants toshow the greatest priming e V  ects and vice versa. Wetested this possibility by calculating the correlationbetween baseline completion times and percentage RTreductions across all participants. The results wereclearly nonsigni W cant ( R (rt[baseline],rt[%reduction]) D 0.079;  p D .542). As a W nal check, we reanalyzed thebehavioral and imaging data after creating an old and ayoung group with similar baseline completion timesusing the young participants’ normal score(means § SD ) as a cut-o V   (RT between 1.18 and 2.04s).The resulting groups (17 young, 16 old participants) didnot di V  er signi W cantly ( t (31) D 1.075;  p D .29) in comple-tion times for control word stems (young D 1.59s, SD D 0.26; old D 1.49s, SD D 0.27). Comparison of thesetwo groups revealed similar age di V  erences in primingmagnitude (young D 18.8%, SD D 10.4; old D 10.0%, SD D 6.7; t D 2.86;  p D .008) and priming-related activity(see Section 3.2). Hence, collectively, the current W ndingsindicate that the age di V  erences we observed cannot beaccounted for by di V  erences in task execution or baselinecompletion times. Instead, results suggest that, althoughpriming was reduced in the older adults, they were morepro W cient in the completion of control word stems thanthe young participants.As an additional manipulation check, the entire list of word stems that had been presented in the scanner wascompleted again in writing by the young participants.The primed word stems were completed with studywords in 42.1% [ SD D 4.8] of the cases, and the controlword stems with control words in 22.1% [ SD D 10.0] of the cases, which was a highly signi W cant di V  erence[ t (24) D 10.2;  p <.001]. In addition, a signi W cant correla-tion was found between the primed vs. control ratio andthe percent reduction in response times ( r D .44;  p D .026), indicating the validity of the use of responsetime reductions as a measure of priming. 3.2. Imaging results Both groups showed priming-related activity reduc-tions in the left inferior frontal cortex, and the anteriorcingulate. However, the young group revealed additionalreductions in the anterior part of the left superior tempo-ral gyrus and the cerebellum (Table 3, Fig. 1A). These di V  erences were found to be signi W cant as determined bya direct group comparison (Table 4). The older partici-pants revealed an additional reduction in the right infe-rior prefrontal cortex, whereas the young participantsdid not. However, this di V  erence was not signi W cant asindicated by a direct group comparison, which wasrelated to the fact that the young participants showedsimilar reductions at a slightly lower threshold (Table 3,Fig. 1A).An additional region in the right posterior occipitallobe (BA 17/18) also showed a larger activity decrease inyoung relative to old participants. However, at the setthreshold ( D  p <.05; FDR-corrected), no evidence wasfound for activity reductions in this area in the youngparticipants’ group average (Table 3). Still, at a slightlylower threshold (  p <.001, uncorrected) we did observereduced activity in this region (Figs. 1A and B; Talairachcoordinates [ x ,  y , z ]: 27, ¡ 96,5; z D 3.68). Only when we Table 3Maxima of regions showing signi W cant BOLD signal reductions (  p <.05, FDR-corrected; cluster D 5; unless indicated otherwise by asterisk) in com-parison of primed vs. control word stemsRegion of activationLeft/rightBrodmann areaTalairach coordinates x ,  y , z { mm} Z   value Young  Inferior frontal gyrusL44/45  ¡ 484115.26R475312103.40 * Anterior cingulate gyrusL32  ¡ 325404.30Superior temporal gyrusL22  ¡ 531554.97CerebellumL—   ¡ 9  ¡ 45  ¡ 234.97 Elderly Inferior frontal gyrusL47  ¡ 4517  ¡ 65.33R474520  ¡ 64.36Anterior cingulate gyrusR24/32319325.11
Related Documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks