Attention and lexical decomposition in chinese word recognition: Conjunctions of form and position guide selective attention

Abstract

A series of experiments on skilled readers of Chinese demonstrated selective attention to different orthographic components within compound, single-character words, as a function of different reading tasks (pronunciation vs. meaning).

Chinese orthography provides a number of contrasts with alphabetic writing systems. The majority (over 90%) of Chinese single-character words are compound, consisting of a lexical radical (LR) and one or more other constituents. which together form the “non-radical component” (NR). The NR component as a whole specifies the syllabic pronunciation of the whole word (except in the case of “irregular” words). In contrast, the LR specifies (an aspect of) its meaning. Thus in order to take advantage of these regularities in pronunciation tasks, the skilled reader should attend selectively to the NR component. To do so, however, the reader must first locate the LR, which can occur in practically any relative location (e.g. top, bottom, left, or right of the character), in order to segment, and so attend to, the NR component. Moreover, the same physical constituent can function as the LR in some characters, but as (part of) the NR component in others, depending on its relative location in a character. Thus neither a strategy of attending to a fixed spatial location in a character nor one of attending to a fixed set of physical constituents is sufficient. Accordingly, selective attention to the NR component must be guided by preselective encoding of the conjunction of figural identity and relative (within-word) location.

We describe a series of six experiments using a simultaneous same-different comparison paradigm, with two contrasting criteria for the comparison: (a) same pronunciation, (b) same meaning. Skilled readers of Chinese showed a strongly selective attentional bias towards the NR component in the pronunciation task, and a complementary selective bias towards the LR in the semantic task. The same set of stimuli, in the critical conditions, were used for both tasks. The same (crossover) pattern of results was found both in comparisons of familiar words and of orthographically regular pseudowords, and in a speeded-response paradigm as well as in tachistoscopic accuracy measures. These results provide striking evidence of linguistic-knowledge-based segmentation, requiring the conjoint encoding of complex figural identity and within-word position, in the control of spatially selective attention.