6533b82dfe1ef96bd129201e

RESEARCH PRODUCT

The representation of segmental information: an fMRI investigation of the consonant-vowel distinction

subject

description

Harvard University, Cambridge, MA, USAAvailable online 23 July 2004IntroductionRecent studies suggest that consonants and vowels are repre-sented separately in cognitive/neural space. Much of the evidencecomes from research on dysgraphia (for review, see Miceli & Cap-asso, submitted). In the first place, letter substitution errors preservethe consonant/vowel (CV) status of the target (e.g., cinema fi ciremaor cinoma, but not cintma). Second, there are reports of selectiveimpairment for consonants or vowels. Additional evidence comesfrom disorders of phonology, demonstrating the dissociability be-tween consonants and vowels (Caramazza, Chialant, Capasso, Mthe ISI was variable (mean 6.75 s). The experimental task consistedof four conditions in which the target letter could be a vowel inposition 2 or 4 (e.g., cinema) or a consonant in position 3 or 5 (e.g.,cinema). Behavioral measures (error rate and RT) were obtainedduring fMRI scanning. In the control task, subjects saw a squiggleand pressed a button, depending on the direction of an arrow shownon the screen. As a baseline measure, subjects fixated the blankscreen. Stimulus presentation and response recording were imple-mented with Psyscope software (Cohen et al., 1993) running on aPower Macintosh computer (Apple, Cupertino, CA). The experimen-tal, control, and filler conditions were randomised in an event-relatedfMRI design and administered in three scanning sessions lasting eachabout 9 min.Image acquisition and analysisMR data were obtained on a 1.5 T Siemens Vision MR-scannerusing a whole-head coil. Blood oxygenlevel dependent (BOLD) imageswere acquired with an echo-planar imaging sequence (TR = 3.5 s,TE = 40 ms, flip angle = 90 ) in the axial plane (40 slices, 64 · 64matrix, 3 mm isotropic voxel size). For each subject, a high-resolution(1 · 1 · 1 mm) T1-weighted image of the whole brain was also ac-quired (TR = 11.4 ms, TE = 4.4 ms, flip angle = 10 , 256 · 256 ma-trix, 220 coronal slices).Functional images were processed in Matlab, using SPM99(Wellcome Department of Cognitive Neurology, London, UK) and in-house software (BrainShow). Images were corrected for slice timingand head movement, and then normalised to Montreal NeurologicalInstitute (MNI) coordinates and spatially smoothed (FWHM =6 mm). Images were analysed using a two-stage random effect ap-proach (Friston, Holmes, & Worsley, 1999).ResultsCompared to rest, the experimental task resulted in extensive ac-tivation of both occipital and temporal–mesial cortices, and of thefrontal and parietal lobes and the motor cortex on the left. Table 1shows the regions activated during: 1, the consonant verification taskvs the control task (C vs SQG); 2, the vowel verification task vs thecontrol task (V vs SQG); and 3, the consonant verification task vs thevowel verification task (C vs V). For each anatomical label and foreach significant comparison, coordinates in MNI space and Brodmannareas (BA) (Talairach & Tournoux, 1988) are given for voxels repre-senting local maxima of activation.Brain and Language 91 (2004) 35–37www.elsevier.com/locate/b&l