Separate neural processes for retrieval of voice identity and word content in working memory

Transcription

1 available at Research Report Separate neural processes for retrieval of voice identity and word content in working memory Kristiina Relander a,, Pia Rämä a,b,c a Cognitive Brain Research Unit, Department of Psychology, P.O. Box 9, University of Helsinki, Finland b Neuroscience Unit, Institute of Biomedicine/Physiology, University of Helsinki, Finland c Advanced Magnetic Imaging Centre, Helsinki University of Technology, Finland ARTICLE INFO Article history: Accepted 13 November 2008 Available online 27 November 2008 Keywords: Voice identity Word content Speech Working memory Recognition fmri ABSTRACT Working memory for voice identity and words was studied to investigate whether the neural system underlying extralinguistic and linguistic information processing is dissociated and whether the possible differences in the distribution of activity are related to specific periods of working memory tasks. Separate analyses of task-related activations evoked during the encoding, maintenance, and recognition periods of the memory tasks were performed. During the voice task, the superior temporal, ventral prefrontal and medial frontal cortices were activated in comparison with the control task whereas the word task produced activation in the occipital, parietal, and dorsal prefrontal areas. Direct contrasts between different periods of the tasks indicated that the ventral prefrontal cortex and the right superior temporal sulcus/gyrus were more activated during recognition than encoding and maintenance periods in the voice compared with the word task. In contrast, the right supramarginal gyrus was more active during the recognition than encoding period in the word compared with the voice task. The results suggest that dissociable neural substrates are recruited for processing of linguistic and extralinguistic information during the recognition period of a working memory task Elsevier B.V. All rights reserved. 1. Introduction Human speech conveys both verbal and nonverbal information. The voice provides information about the speaker's identity, age, gender, and emotional state. Recognition of voice identity is based on differences in fundamental frequency (pitch) and spectral formant frequencies (timbre) of individual voice (Lavner et al., 2000; Shearme and Holmes, 1959; van Dommelen, 1990). In neuroimaging studies, the superior temporal sulcus/gyrus (STS/STG) has been shown to be responsive for human voices during passive (Belin et al., 2002, 2000) and active (Binder et al., 2000; Fecteau et al., 2004; Meyer et al., 2005; Rämä and Courtney, 2005; Rämä et al., 2004; Shah et al., 2001; Stevens, 2004; Uppenkamp et al., 2006; von Kriegstein et al., 2003; von Kriegstein and Giraud, 2004; von Kriegstein et al., 2005) listening tasks. The STS has also been shown to respond better to speech than nonspeech vocalizations such as laughs and cries (Belin et al., 2002). Some recent studies have compared voice and word processing directly with each other during mismatch negativity (Knösche et al., 2002) and working memory (Stevens, 2004; von Kriegstein et al., 2003, 2005; von Kriegstein and Giraud, Corresponding author. Fax: address: kristiina.relander@helsinki.fi (K. Relander) /$ see front matter 2008 Elsevier B.V. All rights reserved. doi: /j.brainres

2 144 BRAIN RESEARCH 1252 (2009) ) experiments and suggested that voice and word information is processed in parallel fashion during an early pre-attentive stage whereas a dissociation exists during mnemonic processing of voices and words. The dissociation is particularly consistent in the STS/STG, which was recruited during voice versus word processing in all the above mentioned working memory tasks. In addition, using a two-back task, Stevens (2004) found that memory for voices relative to words produced activation in the right superior and middle frontal gyri, posterior cingulate and right angular gyrus, whereas memory for words compared with voices recruited the left inferior frontal gyrus and bilateral supramarginal gyri. von Kriegstein and colleagues (2003, 2005) as well as von Kriegstein and Giraud (2004) found that during voice compared with word recognition tasks, dorsolateral (von Kriegstein and Giraud, 2004), orbital (von Kriegstein and Giraud, 2004; von Kriegstein et al., 2005) and preorbital (von Kriegstein et al., 2005) frontal regions, parietal regions (von Kriegstein et al., 2003, 2005; von Kriegstein and Giraud, 2004) and the cerebellum (von Kriegstein and Giraud, 2004) were recruited in addition to the superior temporal areas. Word compared with voice processing, in turn, produced activation in the left middle temporal and lingual cortices (von Kriegstein et al., 2003). However, von Kriegstein and colleagues (2003, 2005), von Kriegstein & Giraud (2004) and Stevens (2004) did not investigate the activity between different periods of the memory tasks, such as encoding, maintenance, and recognition. Earlier, the dissociation of activity between two kinds of memory task performances has been shown to be dependent on the cognitive demands required during the different periods of the memory tasks (Arnott et al., 2005; Rämä and Courtney, 2005; Rämä et al., 2004). Activation differences for the processing of sound locations in comparison with sound identities was observed predominantly during the recognition and comparison period, whereas specific activity for sound identities occurred during the encoding and maintenance periods of the memory task (Arnott et al., 2005). In two other studies, the dissociation was observed between mnemonic processing of voices and faces (Rämä and Courtney, 2005) and voices and voice locations (Rämä et al., 2004) but the areas of activation were distinct during the different periods of the tasks. Thus, earlier research shows that that it is fruitful to investigate working memory task periods separately in order to find out how different stimulus features are processed depending on cognitive demand. The aim of the present study was to investigate working memory processing of voice identities and words. Specifically, we wanted to find out how neural activation differs between specific periods (encoding, maintenance, and recognition) of working memory tasks. The subjects performed a delayed recognition task in which they were to remember either the speaker's voice or the word spoken, and a sensorimotor control task that required the same motor activity and attention towards stimuli as the memory tasks, but that had no mnemonic demands. Activations were first compared between the memory tasks and the control condition and then between the two memory tasks separately during each task period in order to specify brain areas associated with a certain task and task period. In order to ensure that these activations were specific to a certain task and task period, also the interactions between tasks and task periods were calculated. 2. Results 2.1. Behavioral data Although the subjects were accurate in both tasks during the brain scanning, they made more incorrect responses during the voice (mean accuracy 82%) than word (mean accuracy 98%) trials (pb0.001). Also the reaction times were significantly longer for the voice (1942 ms) than word (1645 ms) task (pb0.001), and the subjects subjectively rated the voice task as Table 1 Sample activity for Voices and Words versus Control Brain area VoiceNControl WordNControl Spatial extent Spatial extent (in mm 3 ) (in mm 3 ) IFG/Insula IFG/MFG MFGa MFGp SFGm SFGm/CinG PreCG PostCinG STS/STG SMG/PostCS LG LG/Cun Note. Areas of significant activity, the peak Talairach coordinates, the peak and mean values and the spatial extent of a given activity.

3 145 Table 2 Delay activity for Voices and Words versus Control Brain area VoiceNControl WordNControl Spatial extent (in mm 3 ) Spatial extent (in mm 3 ) IFG/Insula IFG/MFG a MFGa SFGm a SFS/PreCG STS/STG SMG Note. Areas of significant activity, the peak Talairach coordinates, the peak and mean values and the spatial extent of a given activity. a The cluster included two separate activation loci with local maxima in the left IFG/MFG and the right SFGm. more difficult than the word task. The mnemonic strategies used in the voice task were visual imagination of the speaker (6 subjects), naming the speaker (1 subject) and attending to specific features of the voices, e.g. intonation and timbre (3 subjects). In the word task, subjects simply tried to keep the word in mind, for example by silently repeating it fmri data Word and Voice Task activations relative to control task Sample period (Table 1). During the sample period, both voice and word tasks produced activation in the left inferior frontal gyrus/insula (IFG/Insula), the left anterior middle frontal gyrus (MFGa), the medial part of the superior frontal gyrus (SFGm), bilateral superior temporal sulcus/gyrus (STS/STG), the left supramarginal gyrus/postcentral sulcus (SMG/PostCS), and the left lingual gyrus/cuneus (LG/Cun). The voice task additionally activated the bilateral IFG/MFG, the left precentral gyrus (PreCG), and posterior cingulate gyrus (post- CinG), whereas the word task resulted in activity in the right IFG/Insula, the left posterior MFG (MFGp), and the SFGm/ Cingulate gyrus (Table 1) Delay period (Table 2). During the delay period, activation was detected for both voice and word tasks in the left IFG/Insula, bilateral IFG/MFG, SFGm, and the left SMG. The voice task also produced activity in the right IFG/Insula and the left superior frontal sulcus/precentral gyrus (SFS/PreCG), and word task in the right MFGa and the STS/STG bilaterally (Table 2). Table 3 Test activity for Voices and Words versus Control Brain area VoiceNControl WordNControl Spatial extent Spatial extent (in mm 3 ) (in mm 3 ) IFG/Insula a b a , b IFG/MFG a b a b SFGm a b STS/STG a b a b ,748 SMG a b a b Cerebellum Cun PreCun , Note. Areas of significant activity, the peak Talairach coordinates, the peak and mean values and the spatial extent of a given activity. a The cluster included 9 separate activation loci with local maxima in bilateral IFG/Insula, bilateral IFG/MFG, the SFGm, bilateral STS/TG and bilateral SMG. b The cluster included 9 separate activation loci with local maxima in bilateral IFG/Insula, bilateral IFG/MFG, the SFGm, bilateral STS/TG and bilateral SMG.

4 146 BRAIN RESEARCH 1252 (2009) Test period (Table 3). During the test period, activation for both voice and word tasks occurred in the IFG/Insula and IFG/MFG bilaterally, SFGm, bilateral STS/STG, bilateral SMG, cerebellum, and the precuneus (PreCun). In addition, the word task activated the cuneus (Cun) (Table 3) Direct voxelwise comparison: Voice N Word and WordNVoice (Table 4) The voice task in comparison with the word task activated the right IFG/MFG during the sample period. During the delay period, the SFGm was more active for the voice task than the word task. During the test period, the ventral prefrontal cortex (bilateral IFG/Insula and IFG/MFG), SFGm, and bilateral STS/ STG were more active for the voice task than for the word task (Table 4). The word task compared with the voice task, in turn, produced activity in bilateral LG during the sample period. During the delay period no regions exhibited significantly more activation for the word task than the voice task. During the test period, the right MFG, bilateral SMG and precuneus were more activated for the word task than the voice task (Fig. 1) Direct voxelwise comparisons between task periods (Table 5) The ventral (IFG/Insula, IFG, IFG/MFG) and medial prefrontal cortices (SFGm/CinG) as well as the cuneus were more activated during the test than during the sample period in the voice compared with the word task. In addition to the ventral and medial prefrontal cortices, the right STS/STG was more active for the test than for the delay period in the voice compared with the word task. In contrast, the word task relative to the voice task produced more activity in the right SMG during the test period compared with the sample period (Table 5). 3. Discussion The present results provide evidence for a functional dissociation between the processing of voice identity and spoken words during a delayed recognition task. The superior temporal sulcus/gyrus, medial frontal and ventral prefrontal regions were more active for voice than word processing whereas occipital, parietal, and dorsal prefrontal regions were more activate for word than voice processing. The results are in accordance with an idea that dissociable neural substrates are recruited by working memory processing of different components of speech (Stevens, 2004; von Kriegstein et al., 2003, 2005; von Kriegstein and Giraud, 2004) although the pattern of activity during both voice and word tasks was partly different from the previous studies. In addition, these previous studies did not investigate whether the dissociation was dependent on the cognitive demands required during the different periods of the memory task. The present results further indicate that a double dissociation in the processing of linguistic and extralinguistic information occurs during the recognition period of a working memory task although differences in task difficulty partly limit the interpretation of the results. The ventral prefrontal cortex and superior temporal sulcus/gyrus were activated more during the test than sample and delay periods in the voice compared with the word task. In contrast, the parietal lobe was more active during the test than sample period in the word compared with the voice task. Table 4 Direct comparisons between the memory tasks Brain area VoiceNWord WordNVoice Spatial extent Spatial extent (in mm 3 ) (in mm 3 ) Sample IFG/MFG LG Delay SFGm Test IFG/Insula a ,689 IFG/MFG a MFG SFGm STS/STG SMG PreCun Note. Areas of significant activity, the peak Talairach coordinates, the peak and mean values and the spatial extent of a given activity. a The cluster included two separate activation loci with local maxima in the right IFG/Insula and the right IFG/MFG.

5 147 Fig. 1 Cross-subject average statistical maps of direct comparisons between activations during the sample and test periods of voice and word tasks overlayed on a Talairach normalized anatomical image. Greater activation was detected in the STS/STG for voice than word processing during stimulus recognition compared with stimulus maintenance but not with stimulus encoding. The STS/STG activation for voice versus word recognition was more pronounced in the right than in the left STS/STG (see Table 4). It is largely agreed that that the fast-changing information of speech is predominantly processed in the left hemisphere showing a finer temporal resolution than the right hemisphere which, in turn, is involved in pitch and prosody processing. This temporal processing difference in the two hemispheres has been shown to exist at least in the STS (Bohemio et al., 2005). The right STS has also been found to activate for emotional compared with neutral prosody irrespective of the listener's focus of attention, suggesting that the STS may have a role in orienting towards social and affective aspects of speech (Grandjean et al., 2005). The present results further suggest that there is a modulation of activity for voices compared with word content in the superior temporal cortex during stimulus presentation in a working memory task, and thus this area functions also in controlled retrieval of social auditory information. The ventral prefrontal regions were activated during retrieval, rather than encoding or maintenance, of speaker identity compared with that of word content in the present study. In a recent study, activation in the anterior IFG was found during retrieval of previously learned words when the task was to remember whether the words were in the previously studied list and spoken by the same speaker, compared with a task requiring only word retrieval. The prefrontal activation was suggested to reflect the engagement of monitoring processes during memory retrieval in general, although the authors correctly pointed out that the results could also be accounted by processes directly related to voice retrieval or, on the other hand, more general control processes recruited because of task difficulty (Ranganath et al., 2007). It should, indeed, be taken into account that also in the present task the voice and word tasks differed in both response accuracy and used strategies, both known to affect cognitive functions that are largely mediated by the prefrontal cortex. In the voice task, several subjects used visual imagination as a memory strategy whereas in the word task, no specific strategies were used, presumably because the words were abstract and thus hard to imagine. It has been shown that the Table 5 Interactions between tasks and periods VoiceNWord WordNVoice Spatial extent Spatial extent (in mm 3 ) (in mm 3 ) Test NSample IFG/Insula IFG IFG/MFG CinG/SFGm Cuneus SMG Test NDelay IFG/Insula IFG/MFG STS/STG CinG/SFGm Note. Areas of significant activity, the peak Talairach coordinates, the peak and mean values and the spatial extent of a given activity.

6 148 BRAIN RESEARCH 1252 (2009) processing of concrete, highly imaginable words produces differential brain activations than abstract word processing in various brain regions, including parts of the inferior frontal cortex (Sabsevitz et al., 2005), indicating that the use of imagery as a memory strategy may also contribute to the prefrontal activation for voice versus word processing in the present study. It is also known that difficulty level influences prefrontal activation during memory tasks. Prefrontal activation related to task difficulty has been suggested to reflect e.g. the amount of time that information is held in working memory during task performance (Sabsevitz et al., 2005) or more intense effort during memory task performance (Menon et al., 2000; Schacter et al., 1996; Tulving et al., 1999). In one of the previous studies investigating working memory for voices and words, significant task performance differences of approximately the same magnitude as in the present study were reported (von Kriegstein et al., 2003). Right inferior frontal activity was not found in that study but, instead, it was observed for voice relative to word processing in a study in which the stimuli were concrete and the tasks equally difficult (Stevens, 2004), suggesting a role for the right inferior frontal cortex in extralinguistic speech processing. Further, the present results indicate that the right prefrontal cortex takes part in the recognition, rather than encoding or maintenance of voices. It should be noted, however, that differences in task difficulty may have contributed to the fact that no ventral frontal regions responded to word contrasted with voice processing in the present study and in the study by von Kriegstein et al. (2003).In contrast, activation of the left prefrontal cortex was found for word compared with voice (Stevens, 2004) processing during tasks with equal difficulty levels. It is possible that increased difficulty of the voice task in the present study amplified activation in the left ventral prefrontal regions, thus decreasing the differences between task-related activations. The dissociation between task periods was also present during the recognition period of the word task in comparison with the voice task. The SMG was more active for word than voice processing bilaterally during the test period. In the direct comparison between the task periods, the right SMG was more active during the test than sample period in the word task compared with the voice task, suggesting that the right parietal lobe has a role in recognition and retrieval of verbal information. Earlier, working memory for visually presented letters has been found to produce activity in the SMG in the left hemisphere, and therefore it has been suggested to be the site of the phonological store (Paulesu et al., 1993). Likewise, patients with impairments in acoustic-phonetic processing often have lesions in the left posterior SMG (Caplan et al., 1995). However, in several other studies verbal information has recruited the SMG bilaterally (Clark et al., 2000; Crottaz- Herbette et al., 2004; Honey et al., 2000; Stevens, 2004). Some verbal working memory studies have also explored the pattern of activity during different subcomponents (encoding, maintenance, and retrieval) of memory tasks. The results have confirmed earlier findings on the role of the parietal cortex in storage and rehearsal processes (Habeck et al., 2005; Jonides et al., 1998). However, in one of the studies, the parietal lobe was active also during the retrieval period of the working memory task (Jonides et al., 1998), and in another study, it was shown to exhibit greater activity during retrieval than encoding (Veltman et al., 2003) suggesting that storage and retrieval processes may share common neuronal networks (Jonides et al. 1998). The present results provide further evidence that the right parietal cortex is involved in retrieval of verbal information during speech recognition. In conclusion, the results of the present study confirm earlier findings on a dichotomy in the mnemonic processing of voice identity and word content of speech. Furthermore, the results suggest that a double dissociation in the processing of voices and words occurs during the recognition period of a working memory task. The superior temporal cortex and, possibly also the inferior frontal cortex are involved in the recognition of extralinguistic information of speech, whereas the parietal lobe takes part in the recognition of word content of speech. 4. Experimental procedures 4.1. Subjects 10 right-handed subjects (5 females) between the ages of 22 and 28 (mean 24 years) participated in the study. The subjects were native Finnish speakers. They gave written informed consent, and were paid 20 for participating in the experiment. The experimental protocol was approved by the Ethical Committee of the Helsinki University Hospital Stimuli The voice samples consisted of abstract four-syllable Finnish nouns recorded in a sound-proof room using a DAT tape recorder. The sampling rate was 44.1 khz. Twelve native Finnish female voices were recorded. The speakers were instructed to read the chosen words in a neutral tone. Eight words (tiivistelmä [abstract], vaikutelma [impression], valikoima [selection], olettamus [assumption], suunnitelma [plan], valmistelu [preparation], järjestelmä [system], suhteellisuus [relativity]) were recorded. For the final set of stimuli, eight speakers were selected based on the quality of the recordings, thus resulting in eight words spoken by eight different speakers. The mean durations of the words were 934 ms, 937 ms, 905 ms, 933 ms, 1015 ms, 939 ms, 1033 ms and 1210 ms, respectively. The energy levels (db) of the voice samples were normalized to peak. Control stimuli, that were similar with the memory stimuli regarding their basic physical stimulus properties (frequency content and amplitude) but unidentifiable as speech or words, were obtained by phase-scrambling the voices in the Fourier domain, maintaining frequency information and stimulus amplitude envelopes equal to those in the memory tasks (see Rämä and Courtney, 2005). Before the experiment, subjects heard each voice and word twice to gain familiarity with the stimuli, and once or twice more to practice the memory tasks. The auditory stimuli were delivered through Avotec Silent Scan pneumatic (frequency response 150 Hz 4.5 khz) headphones (Avotec Inc., USA) and the presentation level was individually adjusted well-above hearing level before the imaging session. The visual stimuli (trial instructions and fixation cross) were presented using a projector, located outside of the scanning room, connected to a

7 149 computer running Presentation software (Neurobehavioral Systems Inc., USA). The stimuli were projected on a rear projection screen mounted inside the bore of the magnet, behind the subject's head. Subjects viewed the stimuli by means of a mirror mounted at the top of the head coil Tasks Subjects were instructed to remember either voices, spoken words, or neither in the delayed recognition and control tasks. Three seconds before each trial, subjects were presented with an instruction image (for 1.5 s) consisting of the Finnish word Puhuja ( Speaker ) in the voice task, Sana ( Word ) in the word task, or Kontrolli ( Control ) in the control task, indicating the task to be performed. In the voice task, the subjects were to memorize the speaker independent of the word spoken, and in the word task, the word independent of the speaker. The sample stimulus (1.5 s) was followed by a memory delay of 4.5 s during which the subjects saw a blank screen with a fixation cross. Then, a test stimulus (1.5 s) was presented during which time the subject indicated with a left (index finger) or right (middle finger) button press whether or not the test stimulus was the same as the sample stimulus. Each subject used his right hand for responding. Following each test stimulus, there was an intertrial interval of 3.0 s. In the voice task, the word presented during the test period never matched the word presented during the sample period. Similarly, in the word task, the voice presented during the test period never matched the voice presented during the sample period. In order to exclude false positivities (e.g. activations related to sensory input, attention towards any stimuli or motor response instead of working memory processing per se) in further analyses, the subjects also performed a sensorimotor control task that required the same motor activity and attention towards stimuli as the memory tasks, but that had no mnemonic demands. In this task, the scrambled stimuli were presented with the same timing as the stimuli in the memory tasks but the subjects were instructed that they need not remember the words or voices, but simply press the right button when the test stimulus was played. During the scanning, six runs were conducted. In each run, both memory task conditions were presented in four alternating blocks of four trials each. Each block of four memory task trials was preceded and followed by one control trial. Thus, in each run, there were 8 memory test trials of each information type and eight control trials. The order of tasks was counterbalanced across runs within each subject, and the order of runs was counterbalanced across subjects. The match/no-match responses were recorded during the scanning. After the scanning, each subject was asked to rate the difficulty of each task and describe the mnemonic strategies used in the task performance. The number of incorrect responses and reaction times were calculated and analyzed using a paired t-test FMR imaging and data analysis MR-images were acquired with a 3 T MR scanner (Signa VH/i, General Electric Inc.). A T1-weighted structural image (124 axial slices, 1.0 mm, no gap, TR=8.5 ms, TE=1.8 ms, flip angle=15, matrix , FOV=240 mm) was obtained before the functional scanning. During the performance of the tasks, subjects underwent T2 -weighted interleaved gradient-echo, echo-planar imaging (20 axial slices, 5 mm thickness, no gap, TR=1500 ms, TE=30 ms, flip angle=70, matrix 64 64, FOV =240 mm). The images were phaseshifted using Fourier transformation to correct for slice acquisition time, then motion-corrected using automatic image registration (AIR) software (Woods et al., 1998), and analyzed separately for each subject using multiple regression (Friston et al., 1995) with Analysis of Functional NeuroImages (AFNI) software (Cox, 1996). Changes in neural activity were modelled as square-wave functions matching the time course of periods of experimental tasks. These square-waves were convolved with a gamma function model of the hemodynamic response using the following values: 2.0 s for lag, 3.0 s for rise time, and 5.0 s for fall time to create the regressors of interest in the multiple regression analysis. Additional regressors were included to model sources of variance not related to the experimental manipulations (mean intensity between and linear drift within time series). Different periods of the tasks were each modelled with separate regressors. Both memory task conditions (voice and word) were separately contrasted with the control task, and with each other, for each of the three main periods of the tasks (encoding, maintenance, and recognition). For direct comparisons between the memory tasks (voice vs. word), in order to capture memory-related activations, the analysis was restricted only to the voxels showing significantly greater activity for any of the memory versus control task (voice vs. control task, word vs. control task, both tasks vs. control task) in a given main period of the tasks. Direct contrasts between the different periods of the tasks were also performed. The contrasts between the tasks (voice, word) and task periods (sample, delay and test) were analyzed in terms of interactions. To find out brain regions specific to processing of e.g. voice tests, the comparison was the following: Voice testnvoice sample vs. Word testnword sample. In each interaction, the analysis was restricted to the voxels activated during the compared periods (e.g. the analysis Voice test NVoice sample vs. Word test NWord sample was restricted to the voxels activated for either Voice testnvoice sample or Word testnword sample). Because the delay period had a length (4.5 s) different from that of sample and test periods (1.5 s), only the middle part of the delay period (1.5 s) was included in the interaction analyses. Each of these contrasts resulted in a -map for each subject. -maps were registered in the Talairach coordinate system (Talairach and Tournoux, 1988) and resampled to 1 mm 3. Average -maps were computed by dividing the sum of -values by the square root of the sample size. The tests of voxelwise significance were kept at a threshold of 2.33, corresponding to a pb0.01, and corrected for multiple comparisons (experiment-wise pb0.05) applying a measure of probability that uses the individual voxel score threshold and the number of contiguous significant voxels. Based upon a Monte Carlo simulation run via AFNI (Ward, 2000) for the volume of the entire brain, it was estimated

8 150 BRAIN RESEARCH 1252 (2009) that a 422-mm 3 contiguous volume (six voxels, each measuring 3.75 mm 3.75 mm 5.00 mm) would meet the pb0.05 threshold. For the direct comparison between memory tasks within the restricted number of voxels, depending on the specific comparison, a cluster size varying between 211 and 422 mm 3 (3 to 6 voxels) satisfied a 0.05 experimentwise probability. Activations were anatomically localized in the averaged maps using a single subject T1-weighted image. Acknowledgments This work, as part of the European Science Foundation EUROCORES Programme OMLL, was supported by funds from the Academy of Finland (grant ). Pia Rämä is supported by the Academy of Finland (75790). We wish to thank Marita Kattelus and Dr. Antti Tarkiainen for helping in data acquisition in the Advanced Magnetic Imaging (AMI) center and Alexander Degerman for programming the task files. REFERENCES Arnott, S.R., Grady, C.L., Hevenor, S.J., Graham, S., Alain, C., The functional organization of auditory working memory as revealed by fmri. J. Cogn. Neurosci. 17, Belin, P., atorre, R.J., Lafaille, P., Ahad, P., Pike, B., Voice-selective areas in human auditory cortex. Nature 403, Belin, P., atorre, R.J., Ahad, P., Human temporal-lobe response to vocal sounds. Cogn. Brain Res. 13, Binder, J.R., Frost, J.A., Hammeke, T.A., Bellgowan, P.S.F., Springer, J.A., Kaufman, J.N., Possing, E.T., Human temporal lobe activation by speech and nonspeech sounds. Cereb. Cortex 10, Bohemio, A., Fromm, S., Braun, A., Poeppel, D., Hierarchical and asymmetric temporal sensitivity in human auditory cortices. Nat. Neurosci. 8, Caplan, D., Gow, D., Makris, N., Analysis of lesions by MRI in stroke patients with acoustic-phonetic processing deficits. Neurology 45, Clark, C.R., Egan, G.F., McFarlane, A.C., Morris, P., Weber, D., Sonkkilla, C., Marcina, J., Tochon-Danguy, H.J., Updating working memory for words: a PET activation study. Hum. Brain Mapp. 9, Cox, R.W., AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput. Biomed. Res. 29, Crottaz-Herbette, S., Anagnoson, R.T., Menon, V., Modality effects in verbal working memory: differential prefrontal and parietal responses to auditory and visual stimuli. NeuroImage 21, Fecteau, S., Armony, J.L., Joanette, Y., Belin, P., Is voice processing species-specific in human auditory cortex? NeuroImage 23, Friston, K.J., Holmes, A.P., Poline, J.-B., Grasby, P.J., Williams, C.R., Frackowiak, R.S.J., Analysis of fmri time-series revisited. NeuroImage 2, Grandjean, D., Sander, D., Pourtois, G., Schwartz, S., Seghier, M.L., Scherer, K.R., Vuilleumier, P., The voices of wrath: brain responses to angry prosody in meaningless speech. Nat. Neurosci. 8, Habeck, C., Rakitin, B.C., Moeller, J., Scarmeas, N., arahn, E., Brown, T., Stern, Y., An event-related fmri study of the neural networks underlying the encoding, maintenance, and retrieval phase in a delayed-match-to-sample task. Cogn. Brain Res. 23, Honey, G.D., Bullmore, E.T., Sharma, T., Prolonged reaction time to a verbal working memory task predicts increased power of posterior parietal cortical activation. NeuroImage 12, Jonides, J., Schumacher, E.H., Smith, E.E., Koepp, R.A., Awh, E., Reuter-Lorenz, P.A., Marshuetz, C., Willis, C.R., The role of parietal cortex in verbal working memory. J. Neurosci. 18, Knösche, T.R., Lattner, S., Maess, B., Schauer, M., Friederici, A.D., Early parallel processing of auditory word and voice information. NeuroImage 3, Lavner, Y., Gath, I., Rosenhouse, J., The effects of acoustic modifications on the identification of familiar voices speaking isolated vowels. Speech Commun. 30, Menon, V., White, C.D., Eliez, S., Glover, G.H., Reiss, A.L., Analysis of a distributed neural system involved in spatial information, novelty, and memory processing. Hum. Brain Mapp. 11, Meyer, M., ysset, S., von Cramon, D.Y., Alter, K., Distinct fmri responses to laghter, speech, and sounds along the human peri-sylvian cortex. Cogn. Brain Res. 24, Paulesu, E., Frith, C.D., Frackowiak, R.S.J., The neural correlates of the verbal component of working memory. Nature 362, Rämä, P., Courtney, S.M., Functional topography of memory for face or voice identity. NeuroImage 24, Rämä, P., Poremba, A., Sala, J.B., Yee, L., Malloy, M., Mishkin, M., Courtney, S.M., Dissociable functional cortical topographies for working memory maintenance of voice identity and location. Cereb. Cortex 14, Ranganath, C., Heller, A.S., Wilding, E.L., Dissociable correlates of retrieval processing in prefrontal cortex. NeuroImage 35, Sabsevitz, D.S., Medler, D.A., Seidenberg, M., Binder, J.R., Modulation of the semantic system by word imageability. NeuroImage 27, Schacter, D.L., Alpert, N.M., Savage, C.R., Rauch, S.L., Albert, M.S., Conscious recollection and the human hippocampal formation: evidence from positron emission tomography. Proc. Natl. Acad. Sci. U. S. A. 93, Shah, N.J., Marshall, J.C., afiris, O., Schwab, A., illes, K., Markowitsch, H.J., Fink, G.R., The neural correlates of person familiarity. A functional magnetic resonance imaging study with clinical implications. Brain 124, Shearme, J.N., Holmes, J.N., An experiment concerning the recognition of voices. Lang. Speech 2, Stevens, A.A., Dissociating the cortical basis of memory for voices, words and tones. Cogn. Brain Res. 18, Talairach, J., Tournoux, P., Co-Planar Stereotaxic Atlas Of The Human Brain. Thieme, New York. Tulving, E., Habib, R., Nyberg, L., Lepage, M., McIntosh, A.R., Positron emission tomography correlations in and beyond medial temporal lobes. Hippocampus 9, Uppenkamp, S., Johnsrude, I.S., Norris, D., Marslen-Wilson, W., Patterson, R.D., Locating the initial stages of speechsound processing in human temporal cortex. NeuroImage 31, van Dommelen, W.A., Acoustic parameters in human speaker recognition. Lang. Speech 33, Ward, B.D., Simultaneous Inference for fmri Data. afni.nimh.nih.gov/pub/dist/doc/manual/alphasim.pdf. Veltman, D.J., Rombouts, S.A.R.B., Dolan, R.J., Maintenance versus manipulation in verbal working memory revisited: an fmri study. NeuroImage 18,

9 151 von Kriegstein, K., Giraud, A.L., Distinct functional substrates along the right superior temporal sulcus for the processing of voices. NeuroImage 22, von Kriegstein, K., Eger, E., Kleinschmidt, A., Giraud, A.L., Modulation of neural responses to speech by directing attention to voices or verbal content. Cogn. Brain Res. 17, von Kriegstein, K., Kleinschmidt, A., Sterzer, P., Giraud, A.L., Interaction of face and voice areas during speaker recognition. J. Cogn. Neurosci. 17, Woods, R.P., Grafton, S.T., Holmes, C.J., Cherry, S.R., Mazziotta, J.C., Automated image registration: I. General methods and intrasubject, intramodality validation. J. Comput. Assist. Tomo. 22,