Different neural correlates of conceptual processing of Chinese nouns and verbs

Abstract

INTRODUCTION: Early evidence for different neural correlates of nouns and verbs were based on tasks with action and object names in brain-injured and normal individuals (Miceli et al., 1984). This was challenged by the findings that grammatical class effects disappeared once imageability was controlled, and only remained when morpho-syntactic processing was required (Tyler et al., 2004, 2008). However, Bedny et al. (2008) reported greater verb than noun activation in the PLTC using semantic judgments, which were not attributable to visual-motion features, but plausibly to event concepts or form class. Languages with little inflectional morphology, e.g. Chinese, constitute an ideal case to test neural differences of nouns and verbs. This was assessed in Li et al. (2004) using a lexical decision task. No regions specifically activated for either class were found. Nonetheless, the results are not conclusive as lexical decisions do not necessarily involve conceptual or morpho-syntactic processing. METHODS: Fifteen subjects (four men) participated in the study. The task during scanning was to judge mentally whether two words were semantically related on each trial. Experimental materials included 96 nouns and 96 verbs, half with high imageability (Himg) and half with low (Limg) in each grammatical class. Imageability ratings for nouns and verbs were matched in the Limg case. Variables including frequency, age-of-acquisition, and visual complexity were matched between nouns and verbs. Words of the same conditions were selected to form semantically related and unrelated pairs. A block design was adopted with experimental blocks separated by fixation blocks. Scanning was carried out on a 3T Siemens scanner using a gradient-echo EPI sequence (flip angle 90°, TE = 30ms, TR = 2000ms, in-plane resolution = 3.125*3.125, slice thickness = 4mm, slice gap = 0.8mm). Data pre-processing was performed using SPM5, including realign, normalization and smoothing. Three models were used. In Model 1 fMRI signal change was modeled by three event types (nouns vs. verbs vs. fixation) to compare with findings in Li et al. (2004). In Model 2 two additional covariates of interests were included (mean-corrected noun imageability and verb imageability) to evaluate the effects of grammatical class, imageability and potential interactions. Model 3 employed a factorial design with four conditions (Himg nouns; Himg verbs; Limg nouns; Limg verbs), with the comparison of interest being that between Limg nouns and verbs as imageability values for these two groups were matched. Significance threshold was puncorrected < 0.001 (k=10) unless otherwise specified. We further carried out region of interests (ROI) analyses with Models 2 and 3 in 13 areas which had been reported to be sensitive to grammatical classes and/or imageability (reviewed in Bedny & Thompson-Schill, 2006). All anatomical ROIs were created by building spheres of 33 voxels centered on the reported activation peaks (ROI radium: 6mm). RESULTS: Using Model 1, which was identical to that in Li et al. (2004), results of whole-brain analyses show higher activation for nouns than verbs in the left middle temporal and inferior temporal (IT) gyri, the right superior temporal (ST) and middle temporal (MT) gyri, right supramarginal gyrus, and the left medial superior (MS) and middle frontal (MF) gyri, and higher verb than noun activity in the left STG and MTG and left inferior frontal opercularis. Using Model 2, with covariates of interest including noun, verb, noun and verb imageability, no regions were significant using the criterion in the first analysis. Applying a more lenient threshold (puncorrected < 0.005, k = 10) resulted in regions that were more activated for nouns than verbs in the left ST pole, the left medial superior frontal (SF) gyrus and left inferior frontal opercularis, and the left and right insula. Conversely, greater verb than noun activation was found in the left medial SFG and the right MFG. ROI analyses in the 13 ROIs described earlier revealed that imageability was positively correlated with activity in the left fusiform and negatively related to activation in the left frontal area (BA 44, 45, 46, 47), the left MTG, and the left posterior cingulated gyrus (p < 0.05). A marginally significant interaction with greater response for verb than noun imageability was observed in the left IFG BA 47post (p = 0.073). Finally, in Model 3 where the effect of word class was examined controlling for imageability, higher activation for verbs than nouns of low imageability was found in the left IFG BA44 (p = 0.019). CONCLUSIONS: Li et al. (2004) did not find brain regions with bold signal change as a function of grammatical word class using a lexical decision task. In contrast, when participants in this study were asked to make judgments about semantic relatedness of words, different cortical areas were differentially responsive to nouns and verbs. Not incompatible with previous findings, when imageability was also considered, no form class effects were observed in whole-brain analyses unless a lenient threshold was applied. However, results of ROI analyses not only showed effects of imageability but also suggested an interaction between grammatical class and imageability in areas previously reported to be associated with these effects. Further contrast with a focus on low imageability words revealed greater activation for verbs than nouns in LIFG BA44. We conclude that conceptual processing of nouns and verbs in a language with few morpho-syntactic morphemes still have different cortical representations.