Neural Response Suppression Predicts Repetition Priming of Spoken Words and Pseudowords Eleni Orfanidou, William D. Marslen-Wilson, and Matthew H. Davis Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/18/8/1237/1756283/jocn.2006.18.8.1237.pdf by guest on 18 May 2021 Abstract & An important method for studying how the brain processes response to words compared to pseudowords in both posterior familiar stimuli is to present the same item on more than one and anterior temporal regions, suggesting that both contribute occasion and measure how responses change with repetition. to word recognition. Both reduced and elevated activation for Here we use repetition priming in a sparse functional magnetic second presentations (repetition suppression and enhancement) resonance imaging (fMRI) study to probe the neuroanatomical were observed in frontal and posterior regions. Correlations be- basis of spoken word recognition and the representations of tween behavioral priming and neural repetition suppression were spoken words that mediate repetition priming effects. Partic- observed in frontal regions, suggesting that repetition priming ipants made lexical decisions to words and pseudowords spoken effects for spoken words reflect changes within systems involved by a male or female voice that were presented twice, with half of in generating behavioral responses. Based on the current results, the repetitions in a different voice. Behavioral and neural priming these processes are sufficiently abstract to display priming de- was observed for both words and pseudowords and was not spite changes in the physical form of the stimulus and operate affected by voice changes. The fMRI data revealed an elevated equivalently for words and pseudowords. & INTRODUCTION Naccache, Cohen, et al., 2001), then neural processes in An important method for studying how the brain pro- this region can be inferred to operate over more abstract cesses familiar stimuli is to present the same object or (view independent or case independent) representa- word on more than one occasion and observe how tions. Recent priming studies have characterized a suc- behavioral and neural responses are changed by repeti- cession of neural processing stages by which an abstract tion. This repetition priming technique has produced representation of printed words is generated in the fusi- countless demonstrations of faster and more accurate form gyrus, independent of the case and retinal position behavioral responses for items that have been recently of the constituent letters (Dehaene, Naccache, Ciuciu, encountered. At the neural level, a common finding is et al., 2004; Dehaene, Naccache, Cohen, et al., 2001). that repeated stimulus presentation results in decreased Similarly, for visually presented objects and faces, fMRI neural activity as measured using positron emission studies using repetition priming have revealed a series tomography (PET) or functional magnetic resonance of dissociable processing stages by which the sensory imaging (fMRI). Brain regions that are activated during properties of visual stimuli are converted into more the first presentation of a particular stimulus show a abstract representations, independent of viewpoint, reduced response to second presentations (for a review, surface texture and orientation ( James, Humphrey, see Henson, 2003; Schacter & Buckner, 1998). Gati, Menon, & Goodale, 2002; Vuilleumier et al., 2002; Repetition priming can provide a direct means of Koutsaal et al., 2001; Kourtzi & Kanwisher 2000; Grill- testing how the brain converts sensory impressions into Spector et al., 1999). a stable representation of the important entities in the In the current work we apply a similar fMRI repeti- outside world. The rationale behind these studies is that tion priming technique to probe neural processes in- if a brain region shows the same priming effect irre- volved in the identification of spoken words. When spective of changes to physical characteristics of the applied to spoken word recognition, the approach stimuli, such as size differences for visually presented adopted in neuroimaging studies of visual priming is objects (Vuilleumier, Henson, Driver, & Dolan, 2002) or challenged by a controversy concerning the role of case differences for visually presented words (Dehaene, abstract representations in spoken word recognition. Existing visual priming research is predicated on the assumption that progressively more abstract perceptual representations are generated by discarding irrelevant MRC Cognition and Brain Sciences Unit, Cambridge, UK detail in the sensory input. However, research using D 2006 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 18:8, pp. 1237–1252 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.2006.18.8.1237 by guest on 25 September 2021 spoken stimuli has revealed a number of situations in properties of speech at higher levels (Davis & Johnsrude, which specific, surface details of speech are retained and 2003). However, the neural basis of auditory word recog- influence the processing of subsequent repetitions. For nition in humans has not been clearly established. Some instance, behavioral studies have shown that changes researchers propose that a posterior, inferior temporal to the voice used to produce first and second presen- system plays a crucial role in word recognition (Hickok tations can reduce the magnitude of repetition prim- & Poeppel, 2000, 2004), drawing on evidence from apha- ing observed with spoken materials (McLennan & Luce, sic stroke patients that suggests a role for the poste- 2005; Pilotti et al., 2000; Goldinger, 1996). These and rior inferior temporal cortex in assigning meaning to other results showing the retention of surface detail for speech (Bates et al., 2003). Others emphasize a role for Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/18/8/1237/1756283/jocn.2006.18.8.1237.pdf by guest on 18 May 2021 speech in episodic memory (Goldinger, 1998; Sheffert, anterior temporal regions in word identification (Scott & 1998; Church & Schacter, 1994; Palmeri, Goldinger, Johnsrude, 2003), consistent with the location of auditory & Pisoni, 1993; Schachter & Church, 1992) challenge association cortex in macaques (Romanski et al., 1999) the idea that behavioral or neural priming of spo- and supported by the semantic deficits that follow ante- ken words depends on representations that are suffi- rior temporal lobe damage in humans with semantic ciently abstract as to exclude the surface details of dementia (e.g., Chan et al., 2001; Galton et al., 2001; speech (cf. ‘‘episodic’’ accounts of spoken word recog- Mummery et al., 1999). In the light of these disagree- nition [Goldinger 1996, 1998] that deny any role for ab- ments, a further goal of the present study was therefore stract representations). to use neural repetition priming to explore the func- We can contrast these episodic accounts with more tional role of anterior and posterior temporal systems in traditional, abstractionist accounts in which word recog- spoken word recognition. nition and repetition priming are both mediated by One illustration of how neural repetition priming abstract representations that exclude the surface details can be used to identify systems critical for the recogni- of spoken words (e.g., Luce & Pisoni, 1998; Norris, tion of familiar stimuli comes from the literature on 1994). In assessing the neural basis of these diverse face perception. Neuroimaging studies from Henson accounts of spoken word recognition, neuroimaging and colleagues (Eger, Schweinberger, Dolan, & Henson, evidence concerning the systems that mediate repetition 2005; Rothstein, Henson, Treves, Driver, & Dolan, 2004; priming is highly relevant. For instance, if explicit mem- Henson, Price, Rugg, Turner & Friston, 2002; Henson, ory, but not lexical/semantic representations, encode Shallice, & Dolan, 2000) have highlighted a number of the surface detail of speech (see Luce & Lyons, 1998) regions of the temporal lobe that show neural priming then we might observe different repetition priming for repetitions of familiar faces but not for unfamiliar profiles within neural systems involved in episodic faces, particularly when different views of the same face memory and lexical processing. are repeated (Eger et al., 2005; Rothstein et al., 2004). A further issue in studying neural repetition priming These findings suggest that abstract representations of for spoken words concerns the anatomical complexity facial identity, independent of the physical stimulus, of the auditory system. The ventral, object processing are encoded in ventral, anterior temporal regions that pathway in vision is well studied in macaques, and a show interactions between familiarity and priming. A hierarchy of processing stages has long been established similar interaction has recently been shown for written for visual object processing in ventral occipital and tem- words, with priming observed for familiar words but not poral lobe regions (Ungerleider & Mishkin, 1982). Evi- for unfamiliar pseudowords in the left fusiform gyrus dence for hierarchical organization in fMRI studies (Fiebach, Gruber, & Supp, 2005). In the context of of visual priming in humans therefore builds on, and spoken word recognition, these findings suggest that is supported by, evidence from single-cell recordings in comparing neural priming for familiar words and unfa- primates. However, similar homologies between ma- miliar pseudowords may