A PET investigation of lexicality and phonotactic frequency in oral language processing

A PET investigation of lexicality and phonotactic frequency in oral language processing
of 18
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.
Related Documents
   A PET  INVESTIGATION OF LEXICALITY ANDPHONOTACTIC FREQUENCY IN ORAL LANGUAGEPROCESSING S. Majerus and F. Collette University of Liège, Belgium M. Van der Linden University of Liège, Belgium and University of Geneva, Switzerland  P. Peigneux, S. Laureys, G. Delfiore, C. Degueldre, A. Luxen, and E. Salmon University of Liège, Belgium Lexicality and phonotactic frequency effects are observed in many cognitive studies on language pro - cessing, but little is known about their underlying neural substrates, especially with regard tophonotactic frequency effects. Here, we conducted a positron emission tomography (PET) study in which11right - handedvolunteershadeithertorepeatortolistentolistsofwords,highphonotacticfre - quency nonwords, and low phonotactic frequency nonwords.Thecomparisonofword versusnonwordprocessing consistently confirmed previous findings of left temporal and prefrontal activations classi - cally ascribed to lexicosemantic processing. Higheractivationwas found in therightposterior superiortemporal gyrus when comparing high phonotactic frequency nonwords to words, but not when com - paring low phonotactic frequency nonwords to words. We propose thatthis region is implicated in theformation of temporary phonological representations for high - probability phonological events, whichmay support processing of high phonotactic frequency nonwords. INTRODUCTION  A number of studies in cognitive psychology andneuropsychology have shown an advantage in pro - cessing words compared to nonwords in speechperception and identification tasks (Pitt & McQueen, 1998; Vitevitch & Luce, 1998;Vitevitch & Luce, 1999; Vitevitch, Luce, Charles - Luce, & Kemmerer, 1997) as well as in verbalshort - term memory tasks (Gathercole, Frankish,Pickering, & Peaker, 1999; Gathercole, Hitch,Service, & Martin, 1997; Hulme, Maughan, & Brown, 1991). This advantage is considered to bedue to the presence of lexical and semantic repre - sentations which underlie perception and short - term storage of words. Furthermore, nonwords,constructedby usingphoneme associations thatarefrequent and highly probable (high phonotactic COGNITIVE NEUROPSYCHOLOGY, 2002, 19 (4), 343–360 Ó  2002 Psychology Press Ltd DOI:10.1080/02643290143000213 343 Requestsfor reprints should be addressed to Steve Majerus, Neuropsychology Unit, Boulevard duRectorat, 3 (B33), 4000Liège,Belgium (Tel: 0032 4 3662399; Fax: 0032 4 3662808; Email: ThisworkwassupportedbytheBelgianNationalFundforScientificResearch(FNRS),theFondationMédicaleReineElisabeth,and the Interuniversity Pole of Attraction Program P4/22, Belgian State, Prime Minister’s Office, Federal Office for Scientific, Technical and Cultural Affairs. Steve Majerus is a Research Fellow and Fabienne Collette is a Postdoctoral Researcher, both for theBelgian National Fund for Scientific Research (FNRS). We thank Dr Eleanor Saffran and two anonymous reviewers for their very constructive and helpful comments on an earlier version of this manuscript.  frequency) in a given natural language, are pro - cessed more quickly in shadowing and same - differentjudgmenttasksaswellasretainedbetterinshort - term memory than nonwords composed of infrequentphonemeassociations(Gathercoleetal.,1999; Vitevitch & Luce, 1998, 1999; Vitevitch etal., 1997). Similarly, 9 - month - old infants preferlistening to monosyllables containing frequently rather than infrequently occurring phonetic pat - terns relative to their native language (Jusczyk,Friederici, Wessels, Svenkerud, & Jusczyk, 1993; Jusczyk, Luce, & Charles - Luce, 1994). The nonword phonotactic frequency effect hasbeen attributed to the existence of sublexical pho - nological knowledge, containing informationabout the statistical regularities and probabilities of phoneme associations in a given language, favour - ingthe processingofnonword stimuli composedof more frequent phoneme sequences (Gathercole etal.,1999;Vitevitch &Luce, 1998).A second inter - pretation for this effect considers that nonwordfamiliarity stems from the similarity of thenonwords to known words, with high phonotacticfrequency nonwords having more lexical neigh - bours (familiar words which share phonemesequences with the nonwords) than low  - frequency nonwords (Greenberg & Jenkins, 1964; Luce & Pisoni, 1998; Newman, Sawusch, & Luce, 1996;Vitevitch, 1997). However, this supposed lexicalinfluence on nonword processing might actually slow down the processing of high phonotactic fre - quency nonwords in certain conditions where lexi - cal information is a determinant for the task to beprocessed. Indeed, Vitevitch and Luce (1999)showed that in lexical decision tasks, highphonotactic frequency nonwords are actually pro - cessed more slowlythanlow phonotactic frequency nonwords. This inversed phonotactic frequency effectfornonwordsisexplained bytheactivationof a greater number of lexical neighbours by highphonotactic frequency nonwords, which slowsdown lexical decision times. However, a normalphonotactic frequency effect for nonwords wasobservedwhenthesamenonwordstimuliwerepre - sented in a speeded identification task. Accordingto Vitevitch and Luce (1999), these data suggestthatthelexicalandsublexicalinfluenceonnonwordprocessing can be dissociated depending on theprocessing level which is required, with facilitatory effects for high phonotactic frequency nonwords when phonotactic, sublexical knowledge is used,and inhibitory effects when lexical knowledge isused.Further, phonotactic frequency effects in theprocessing of words reflect influences of lexicalneighbourhood: indeed, high phonotactic fre - quency words are processed more slowly in aspeededidentificationandinalexicaldecisiontask,suggesting that more lexical competitors are acti -  vated when processing words containing high ver - sus lowphonotacticprobabilitypatterns, inhibitingprocessing of high phonotactic frequency words. Thus effects of phonotactic frequency operate indifferent ways, depending on the level of represen - tationthatdominatesprocessing.Nonwords,whenprocessed in shadowing and same - different judg - ment tasks, benefit from sublexical phonotacticknowledge. Words, and nonwords when processedin lexical decision tasks, are influenced by lexicalcompetition rather than sublexical phonologicalknowledge. Although lexicality and nonword phonotacticfrequency seem to be reliable and well - investigatedeffects in cognitive studies, brain imaging studiesappear less consistent in showing functional differ - ences in brain activation for word versus nonwordprocessing. Several studies observed activations inbrain areas attributed to lexicosemantic processing when comparing words to nonwords. Left - hemispheric activations were found in the anteriorsuperiortemporalsulcus (BA 22/38),anteriormid - dle (BA 21) and posterior inferior temporal gyri(BA 20/37), inferior parietal (BA 39, angulargyrus), and inferior frontal gyrus (BA 46/47) anddorsal prefrontal regions (BA 8/9/10); activations were also found in the righthemisphere in the pos - terior inferior temporal gyrus (BA 20/37) and theright angular gyrus (BA 39) (Binder et al., 1999;Binder, Frost, Hammeke, Rao, & Cox, 1996;Démonet et al., 1992; Démonet, Price, Wise, & Frackowiak, 1994; Howard et al., 1992; Perani etal., 1996; Price et al., 1996; Scott, Blank, Rosen, &  Wise, 2000). These regions are also partially con - sistent with data from brain - damaged patients. 344  COGNITIVE NEUROPSYCHOLOGY, 2002, 19 (4) MAJERUS ET AL.  Indeed, processing of lexicosemantic information,as evidenced by object naming, word repetition,and word comprehension tasks, is impaired inpatients having left - sided or bilateral lesionsincluding the middle temporal cortex, anterior andposteriorinfero - lateraltemporalcortex,thetempo - ral pole, fronto - parietal areas, and the temporo - parieto - occipital junction (Caramazza & Shelton,1998; Damasio, Grabowski, Tranel, Hichwa, & Damasio, 1996; De Renzi & Lucchelli, 1994;Foundas, Daniels, & Vasterling, 1998; Hart & Gordon, 1990; Hillis & Caramazza, 1991; Knott,Patterson, & Hodges, 1997; McCarthy &  Warrington, 2001; Sacchett & Humphreys, 1992;Silveri & Gainotti, 1988; Warrington & McCar - thy, 1983, 1987; Warrington & Shallice, 1984). These patients have a severe impairment in wordprocessing, whereas nonword processing in speechperception and phonological tasks can be in thenormalrange (e.g.,Hart&Gordon,1990;Knottetal., 1997). Lesions ofprefrontal areas are, however,less frequently observed in patients with a lexico - semantic impairment.Prefrontal activation observed in functionalneuroimaging tasks comparing word to nonwordprocessing might be related to retrieval, selection,and monitoring processes performed on semanticinformation, rather than to lexicosemantic repre - sentations per se (Fiez, 1997; Price, Indefrey, & Van Turennout, 1999; Rapcsak & Rubens, 1990;Swick, 1998).Moreover, other functional neuroimaging stud - ies did not always find significant differences inbrain activation when comparing the processing of  words to nonwords (Binder et al., 2000; Hirano etal.,1997;Wiseetal.,1991).Thesedivergentresultscan be at least partially explained by the require - ments of the tasks used to elicit brain activation: The studies that reported a difference between word and nonword processing used rather complexandactivetasks,inwhichsubjectseitherhadtoper - form phonological or semantic judgments (Binderetal.,1999;Démonetetal.,1992,1994)ortolistento complex speech or “nonspeech” sentences andstories (Perani et al., 1996; Scott et al., 2000), or in which repetition of words was compared to passivelistening to reversed words (Howard et al., 1992). Although these tasks may maximise the differencein brainactivationbetween words and nonwords astheyrequireratherelaborateprocessingofthestim - uli that are presented, they are also likely to engageprocesses other than linguistic ones, e.g.,metalinguistic/executive processes in the case of phonological and semantic judgements and motorprocesses when comparing word repetition tononword listening. Yet,methodologicaldifferencesin thestatisticalanalyses might also explain some of the divergingresults between studies. For example, contrary toPrice et al. (1996), Binder et al. (2000) did notreport significant brain activation differencesbetween passive listening to words and nonwords. A noticeable difference between the two studies isthatBinder etal. reported theirresults after correc - tion for multiple comparisons in the brain volume, whereas Price et al. used a region of interest (ROI)strategywhichallows reportingresultswithoutfur - ther corrections. But actually, Binder et al. alsofound left - hemisphere activations at a lower,uncorrected statistical threshold in the inferior andposteriormiddletemporalgyri,in theangulargyrusand in the middle frontal gyrus, which minimisesthe differences between those two studies. Another critical methodological factor is thatthe degree of dissimilarity between words andnonwordswasnotalwayscontrolledinthedifferentstudies. Indeed, when nonwords are very similar to words, it is not really surprising that they seem toactivate similar brain regions. When presentingnonwords that consist of recorded words playedbackwards (Howardetal.,1992),itis notclearhow distantthisnonspeechisfromintelligiblespeech,atleast in psycholinguistic terms. Therefore, itappears important to control the degree of similar - ity between words and nonwords, by controllingthe phonotactic frequency of the nonword pho - neme sequences, in order to assess how likely thephoneme associations of the nonwords are in thenative languageoftheparticipants.Itmightbe thatgreater and more reliable differences in brain acti -  vationbetweenwords andnonwords wouldemergeifnonwordswereverydistantfromfamiliarwords. To our knowledge, no brain imaging study has directly investigated the neural bases of the COGNITIVE NEUROPSYCHOLOGY, 2002, 19 (4)  345 LEXICALITY AND PHONOTACTIC FREQUENCY   phonotactic frequency effect in comparingnonwords with high and low phonotactic frequen - cies. Depending on the lexical or sublexical inter - pretation of the nonword phonotactic frequency effect, activation in brain regions implicated eitherin lexicosemantic processing or in phonologicalsublexical processing might be expected. We haveseen that in lexical decisiontasks, high phonotacticfrequency nonwords are processed more slowly than their low  - frequency counterparts, due to theactivation of a larger number of lexical neighbours.In that case, the same temporal, parietal, andprefrontal regions subtending lexicosemantic rep - resentations for words should also be activated whencomparinghightolowphonotacticfrequency nonwords.However,inthemoretypicalcasewherehigh phonotactic frequency nonwords are pro - cessed faster than low phonotactic frequency nonwords,as is thecase in simplenonword percep - tion, repetition, and verbal short - term memory,sublexical phonological knowledge is supposed tobe involved and thus brain regions implicated insublexical phonological processing should be acti -  vatedwhencomparinghightolowphonotacticfre - quency nonwords. The most likely brain area subtending the pro - cessing of sublexical phonological knowledgeseems to be the superior temporal area. Data frombrainlesionedpatientswith“wordsounddeafness,”a condition characterised by a selective impairmentof speech sound discrimination, show damage tobilateralsuperiortemporallobeareas.However,theimpairment in sound discrimination is not alwaysrestricted to speech sounds, as these patients some - times have additional difficulties in discriminatingnonlinguistic acoustical information, suggesting amore general impairment in auditory discrimina - tion, at least for some patients (Best & Howard,1994; Buchman, Garron, Trost - Cardamone, Wichter, & Schwartz, 1986; Tanaka, Yamadori, & Mori, 1987).Functional neuroimaging studies also suggestthat the superior temporal area is implicated insublexical phonological processing. Scott et al.(2000)found an area in theposteriorpartoftheleftsuperior temporal sulcus that seems specifically torespond to phonetic information. This regionresponded specifically to speech, to spectrally rotatedspeech(whichisunintelligiblebutpreservesphonetic features and some srcinal intonation), tonoise -  vocodedspeech (whichis intelligible,buthas very weak intonation), but not to spectrally rotatednoise -  vocoded speech (which is completely unin - telligible, does not contain any phonetic informa - tion and does not sound like a human voice). Thelowest common denominator of the three condi - tions to which the posterior superior temporalsulcus responded is precisely sublexical phonemicandphoneticinformation.Similarly,Mazoyeretal.(1993) also found left - lateralised posterior superiortemporal gyrus activation when they presented astory spoken in an unfamiliar language (Tamil) toFrench subjects. Listening to an unfamiliar lan - guage like Tamil involves phonological processingof the phonemes that Tamil shares with French.However,sincethe word formsare veryunfamiliar,theyarenotlikelytotriggerlexicosemanticprocess - ing. Hence the left posterior superior temporalgyrus activation might be also related to sublexicalphonological processingin thestudyby Mazoyeretal. Finally, Binder et al. (2000) found a bilateralsuperior temporal sulcus activation for passive lis - tening to words, pseudowords, and reversed speechbut not for frequency  - modulated tones. What iscommonto the three speech conditions is that they contain phonological and phonetic information,but lexical - semantic content is greatly diminishedfor pseudowords and reversed speech. Frequency  - modulated tones only share acoustic features withspeechstimuli,butdonotcontainanyphonologicalor phonetic features. In summary, brain imagingandfunctionalneuroimagingstudiesconcurtosup - port thehypothesis thatidentificationand process - ing of sublexical phonological information mightbe related to activation in the superior temporalgyrus.In the present PET study the neural bases of  word and nonword processing were investigatedusing a paradigm in which the degree of similarity of the nonwords to familiar words was carefully controlled on the basis of phonotactic frequency. Two sets of nonwords were used: nonwords that were very close to existing words and whosephonotactic frequencies were matched to words, MAJERUS ET AL. 346  COGNITIVE NEUROPSYCHOLOGY, 2002, 19 (4)  andnonwordsverydistantfromthewordsandfromthe first nonword set, with very low phonotacticfrequencies. Furthermore, the three stimulus con - ditions were administered both in passive listeningand repetition conditions. A first aim was to investigate the neural basis of the lexicality effect, during repetition and passivelistening, by contrasting words to high and low phonotactic frequency nonwords. We expectedthat words would activate areas implicated inlexicosemantic processing, i.e., middle and inferiortemporal, parietal, and prefrontal regions, espe - ciallywhencomparedtolowphonotacticfrequency nonwords. These differences in brain activationmight be less important when comparing words tophonologically more familiar nonwords, i.e., thenonwords that are matched to the words on thebasis of phonotactic frequency. A second aim of the study was to investigate theneural basis ofthe phonotacticfrequencyeffect. Asthe phonotactic frequency effect is quite complexand as PET studies only allow a limited number of conditions to be compared, we chose to restrict ourstudy to the exploration of sublexical phonologicalinfluences on phonotactic frequency. Vitevitch andLuce (1998, 1999) haveshownthatsublexical pho - nological influences are implicated when usinghigh and low phonotactic frequency nonwords and when comparingthesenonwords in simple percep - tion and repetition tasks. In this case highphonotactic frequency nonwords are processedfasterandmoreaccurately,assublexicalphonologi - cal knowledge is considered to support processingof nonwords containing high probability phonemeassociations. We investigated the neural substrateof sublexical phonological influences on nonwordphonotactic frequency by hypothesising that com - parison of cerebral activation for high and low phonotactic frequency nonwords during repetitionand passive listening conditions should evidenceactivation of left - sided or bilateral superior tempo - ralbrainregionsimplicatedin sublexicalphonolog - ical processing, but not inferior temporal, parietaland prefrontal brain regions implicated in lexico - semantic processing. Thus, the design of this study allowed us toisolate the cerebral areas involved in lexicosemanticprocessing (words versus high phonotactic fre - quency nonwords and words versus low phonotactic frequency nonwords) and in sublexicalphonological processing (high - frequency non -  words versus low  - frequency nonwords), independ - ently of task  - specific requirements (repetition orsimple passive listening). Moreover, this study alsoenabled us to investigate brain regions activated in verbal repetition and passive listening (repetition versus listening) independently of the lexical andphonotactic status (words or nonwords) of thestimuli. MATERIAL AND METHODSSubjects ElevenFrench - speakingright - handedmalevolun - teers (age 18–25 years) gave their written informedconsent to take part in this study, which wasapproved by the Ethics Committee of the Univer - sity of Liège. None had any remarkable past medi - cal history nor did they use any medication. Cognitive tasks  Threelistsofbisyllabicstimuliwerecreated:80 low phonotactic frequency nonwords, 80 highphonotactic frequency nonwords, and 80 words.Each stimulus had the same syllabic structureCVCCVC. The low  - frequency nonwords wereconstructed using CV and VC diphones, whichare quite rare in French (e.g., in the nonword/f  û glo S / , the diphones /f  û /, / û g/, /lo/, and /o S /are not very frequent in French). On the otherhand, the high - frequency nonwords contained CV and VC diphones which are quite frequent (e.g., inthe nonword /pebmyn/, the diphones /pe/, /eb/,/my/, and /yn/ are very frequent in French). Fur - thermore, the diphones of the high - frequency nonwords had the same phonotactic frequency asthe diphones of the words used in this study. Thediphone frequencies were taken from a phoneticdatabase of French by Tubach and Boë ( Un corpus de transcription phonétique  , 1990). This database was developed on the basis of a phonetic transcrip - COGNITIVE NEUROPSYCHOLOGY, 2002, 19 (4)  347 LEXICALITY AND PHONOTACTIC FREQUENCY 
Similar documents
View more...
Related Search
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