DIVISION OF COMPUTATIONAL LABOR BETWEEN BROCA'S AREA AND WERNICKE'S AREA IN READING VISUALLY PRESENTED WORDS

Jie Huang¹, Lucia Colombo², Thomas H. Carr¹, and Yue Cao¹
Michigan State University¹, East Lansing, MI
Universita’ di Padova, Padova, Italia²

INTRODUCTION

Considerable effort has been devoted to identifying the neural substrate of word reading, and in particular how the brain represents and coordinates rule-governed spelling-to-pronunciation translation based on something like “phonics” versus associative translation based on retrieval from lexical memory. Recent work indicates that bilateral temporo-parietal cortex including Wernicke’s area and its right hemisphere homologue, as well as bilateral inferior frontal cortex including Broca’s area and its right hemisphere homologue, make critical contributions to these processes and to individual differences in the relative dominance of rule-governed versus retrieval-based pronunciation [1]. The present work pursued the computational roles of these regions of tissue, using regularity of spelling-to-pronunciation translation as a marker for involvement in generating phonology and priming from just having read a particular word as a marker for operating via retrieval from associative memory. Any given brain region could be sensitive to one of these markers, to both, or to neither.

METHODS AND MATERIALS

Subjects: Seven neurologically normal right-handed native English speakers (3 female, 4 male, aged 20-56 years) participated in the study.

Language protocol: Stimuli consisted of 80 words, all nouns, monosyllabic and high in frequency [2]. Forty were regular-consistent words that follow the default rules for English spelling to pronunciation conversion, and 40 were irregular words that are inconsistent with the default rules of English. The regular and irregular words were each divided into lists of 20, each of which to serve as the stimuli for one block of imaging trials. The resulting four lists of words were closely matched for initial sounds and frequency. Prior to imaging, subjects read one list of regulars and one list of irregulars selected at random. Each subject then underwent four functional scans representing the factorial combination of regular versus irregular and just read (primed) versus unread (unprimed).

fMRI: Sagittal T2*-weighted images of the whole head were acquired on a GE 1.5 T clinical scanner using a gradient echo Echo-Planar-Imaging pulse sequence (field of view 24 cm, TE/TR = 50/2000 ms, flip angle 90º, matrix size 64x64, slice thickness 7 mm). During each scan, five 12 s periods of task activity were interleaved with five 12 s periods of rest, resulting in a 2 minute scan. Within each task period, four written words were presented sequentially on a LCD display for 3 s each. Subjects were instructed to pronounce each word once, silently. The presentation of the stimuli was synchronized with image acquisition using IFIS (Psychology Software Tools, Inc.). A total of 60 images per anatomic section were acquired during each scan.

Figure shows activation of a typical subject in the left (two left panels) and right (two right panels) hemispheres during reading regular (two top panels) and irregular words (two bottom panels). The subject read the words once prior to the scans.

DATA ANALYSIS

Data pre-processing: Images were assessed and corrected for possible in-plane translation and rotation of the head [3]. The signal intensity time course was corrected for possible slow baseline drifts by fitting 0, 1^st, and 2^nd order polynomials to each voxel’s time series.

Statistical analysis of activation: Time courses of images were cross-correlated [4] with sine and cosine reference functions to obtain a pair of complex cross-correlation coefficients (ccc) voxel by voxel [5]. Magnitude and phase of ccc were further calculated [5]. Activated voxels were thresholded at the magnitude of ccc > 0.33 (estimated p < 0.01) and with phase range of [-120, 0] that was corresponding to a delay in the hemodynamic response up to eight seconds. The activated volumes in two integrated cortical regions, one involving the inferior frontal gyrus (extended Broca’s area) and another the supramarginal gyrus, angular gyrus, and posterior superior temporal gyrus (extended Wernicke’s area), were compared between regular and irregular words.

RESULTS AND DISCUSSION

The extended Broca’s area in left hemisphere was activated to a greater extent than its right hemisphere homologue (p < 0.053). No significant difference in extended Broca’s area activation was observed between regulars and irregulars or between primed and unprimed conditions. Thus the processing done by extended Broca’s area is involved in internal production but is not sensitive to the factors that might affect accessibility of the phonological-articulatory code being produced. This suggests that these frontal regions are not the source of the code despite their importance for its execution as internal speech [6].

For extended Wernicke’s area, total activation was significantly greater for irregulars than regulars (p < 0.009), but did not differ between hemispheres and did not vary with whether the words were primed or unprimed. Thus the processing done by extended Wernicke’s area is sensitive to the regularity of the match between orthography and the phonological-articulatory code being produced, suggesting involvement in spelling-to-pronunciation translation, but is not sensitive to priming, suggesting that this region’s operations do not implement or depend directly on retrieval from associative memory. This in turns suggests a role for bilateral extended Wernicke’s area in rule-governed code generation, or in reconciling such generation with retrieval, or both, but not in retrieval-based generation itself.

In summary, Broca’s area appears to be involved in turning phonological codes into internal speech for all stimulus types, independently of whether the codes are generated by rules or retrieved from lexical memory. Wernicke’s area and its right-hemisphere homologue are involved in generating phonology, as indicated by sensitivity to regularity.

REFERENCES

1. K. R. Pugh, et al., J. Exp. Psy. Human Per. Perf., 23, 299 (1997).
2. W. N. Francis and H. Kucera, Frequency analysis of Engllish usage, Boston: Houghton Mifflin, (1987).
3. Y. Cao, et al., J. Magn. Reson. Med., 3, 869 (1993).
4. P. A. Bandettini, et al., Magn. Reson. Med. 30, 161 (1993).
5. A. T. Lee, et al., Magn. Reson. Med. 33, 745 (1995).
6. J. Huang, et al., Human Brain Mapping, 15, 39 (2001).