1
Neuroimaging evidence for effects of lexical status on phonetic categorization Emily B. Myers and Sheila E. Blumstein Brown University, Department of Cognitive and Linguistic Sciences Introduction Subjects show sensitivity to high-level information when making lower-level decisions in language tasks. For instance, subjects performing a phonetic categorization task for stimuli varying along an acoustic-phonetic continuum exhibit lexically-biased shifts in the phonetic boundary, a phenomenon known as the lexical effect (Ganong, 1980). Psycholinguists disagree about the source of this effect, arguing that it is either: A. A post-perceptual effect which is an artifact of the decision process B. A perceptual effect inherent in the phonetic processing of the stimulus Previous work (Blumstein, et al., 2005) has shown sensitivity to phonetic category structure in perceptual areas such as the superior temporal gyrus (STG) as well as in areas linked to decision-making, such as the anterior cingulate. In this study, this sensitivity is exploited to determine whether lexically-biased shifts in phonetic category boundary are perceptual or post-perceptual in nature. In general, we expect to see more activation for the boundary-value stimulus influenced by the lexical effect, compared to that same VOT value stimulus in the control continuum, which is not influenced by the lexical effect. If the lexical effect is perceptual, we expect to see this difference in perceptual areas such as the STG, as well as in decision-making/executive function areas such as the anterior cingulate. If the lexical effect is post-perceptual, we expect to see this difference exclusively in decision-making areas. Methods Results Conclusion The superior temporal gyri are sensitive to a shift in phonetic boundary which is mediated by the lexical status of the endpoints of the continuum. The involvement of the STG in the mediation of the lexical effect provides strong evidence that the lexical effect itself is perceptual in nature. Data from aphasic populations also supports the view that the lexical effect is mediated by posterior structures. Blumstein et al. (1994) showed that Broca's aphasics, who have damage to left anterior structures, showed a larger lexical effect than normals, whereas Wernicke's aphasics, who have damage to left posterior temporal and parietal structures, showed no lexical effect. In addition, this study shows that activation of the superior temporal gyri in response to phonetic stimuli is influenced by higher-level linguistic information, in this case, information about the lexical status of the word. This activation may arise as a function of: A. Involvement of the STG in both lexical and phonetic processing. B. Feedback to the STG from separate lexical-semantic areas. C. Feedback to the STG from decision-making areas. Given the relative insensitivity of the STG to phonetic decision difficulty (c.f. Binder, et al., 2004, Blumstein, et al., 2005), it seems unlikely the activation pattern in the STG is solely a result of feedback from executive areas. However, further research is necessary to determine how lexical and phonetic information interact in the STG. The cingulate and the left middle frontal gyrus also show sensitivity to the lexical effect. The executive component of this explicit phonetic categorization task is reflected by activation in frontal areas. Activation in frontal areas is greater for stimuli which fall near the phonetic boundary. Such activation may reflect a role for these areas in resolving competition between phonetic categories. References Acknowledgments The authors would like to thank Emmette Hutchison for help with the design and execution of the study. This work was supported by NIH grant 1 F31 DC006520- 01 to EBM, NIH grant F31006520 to SEB, and the Ittleson Foundation. Functional Behavioral Figure 3. Histogram of phonetic boundary placement for fMRI subjects (n=17) 0 2 4 6 8 10 12 10 15 20 25 30 35 40 45 50 55 VOT # of subjects GIFT <--> kift giss <--> KISS Figure 4 DISCUSSION The left and right posterior STG show sensitivity to the phonetic boundary, with increased activation for the same VOT value when it is perceived as the boundary due to the lexical context (Figure 4). No difference in activation is seen in the STG when comparing the baseline boundary, Stimulus #4 (the giss-KISS boundary stimulus) across the two continua. More activation is also seen in the cingulate and left middle frontal gyrus for the lexical effect boundary stimulus than for the same stimulus in the control continuum. Taken together, these results suggest that the baseline response of the STG to boundary stimuli is boosted when the lexical effect shifts the phonetic boundary. While there seems to be an executive component to this process, as indicated by activation in the cingulate and MFG, the lexical effect itself seems to be mediated by the STG. Pilot Data: Word--Non-word continua (n=9) contrasted with Non-word--Non-word continuum (n=10) 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 10 15 20 25 30 35 40 45 50 55 VOT % G responses GIFT<--> kif t giss <--> KISS gish <--> kish Figure 1. Pilot Data Lexical Effect /g/ /k/ TASK 1: Phonetic categorization (/g/ or /k/?) Stimuli: Modified natural speech ranging from /g/ to /k/ in VOT in ~7 msec steps 2 continua: GIFT--kift (lexical effect continuum) giss--KISS (control continuum) The giss-KISS continuum was established as a control via the pilot study pictured at right (Figure 1) because it does not show a boundary shift in comparison to a non-word to non-word continuum. TASK 2: Tone categorization task PARTICIPANTS: 19 adult native speakers of English, two subjects excluded because they did not show the lexical effect. EXPERIMENTAL DESIGN: Rapid event-related fMRI using a design where stimuli were presented in silent periods between scans. Scans: --EPI functional scan: 15 3x3x5mm slices, covering perisylvian areas --MPRAGE anatomical scan for coregistration Data Analysis: Used AFNI for all data analysis. Planned comparisons: statistical contrast across test and lexical effect continua for each VOT value. The critical comparison is the boundary-value comparison. Data thresholded at a voxel level p<0.05, cluster size of 62 voxels, cluster-level threshold of p<0.025 Significant main effect of VOT on phonetic categorization type (/g/ or /k/) Subjects show the typical 'categorical perception' identification function in both continua (Figure 2). Significant main effect of 'Continuum' on the location of the phonetic category boundary (Figure 3). The presence of a word (GIFT) on one end of the continuum shifts the categorization function approximately 7 msec towards the voiceless end of the continuum. --> Subjects show the lexical effect in the scanner. Lexical Effect Left STG (374 voxels) Right STG (316 voxels) Stimulus #5 for GIFT <--> kift = BOUNDARY > Stimulus #5 for giss <--> KISS = NON-BOUNDARY 0 0.05 0.1 0.15 0.2 0.25 0.3 stim4 stim5 % signal change GIFT--kift giss-KISS 0 0.05 0.1 0.15 0.2 0.25 stim4 stim5 % signal change GIFT-kift giss-KISS Cingulate (960 voxels) Left Middle Frontal Gyrus/ Prefrontal (85 voxels) DISCUSSION As in Blumstein, et al. (2005), the overall pattern of results indicates that both anterior (cingulate and left frontal) and posterior (bilateral superior temporal) structures show sensitivity to phonetic category structure, with more activation emerging for stimuli which are close to the phonetic boundary. These results are consistent with the view that posterior areas have a role in early mapping of acoustic input onto a phonetic category, while frontal areas are involved in executive processing of phonetic stimuli, one aspect of which is resolving competition between the phonetic categories. Figure 5 ROI analysis: Continua aligned to the boundary stimulus for each subject individually Binder, J. R., Liebenthal, E., Possing, E.T., Medler, D.A., & Ward, B.D. (2004). Neural correlates of sensory and decision processes in auditory object identification. Nature Neuroscience (7), 295-301. Blumstein, S.E., Burton, M., Baum, S., Waldstein, R., & Katz, D. (1994). The role of lexical status on the phonetic categorization of speech in aphasia. Brain and Language, (46), 181-197. Blumstein, S.E., Myers, E.B., & Rissman, J. (2005). The perception of voice- onset time: An fMRI investigation of phonetic category structure. Journal of Cognitive Neuroscience, (in press). Ganong, W. F. (1980). Phonetic categorization in auditory word perception. Journal of Experimental Psychology: Human Perception and Performance (6), 110-125. Figure 2. Mean % /g/ responses for fMRI subjects (n=17) 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 10 15 20 25 30 35 40 45 50 55 VOT % /g/ responses GIFT <--> kift giss <--> KISS LIFG n=17 0 0.01 0.02 0.03 0.04 0.05 0.06 kift/giss +1 Boundary GIFT/KISS +1 GIFT/KISS +2 % signal change Cingulate n=17 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 kift/giss +1 Boundary GIFT/KISS +1 GIFT/KISS +2 % signal change
