Abstract:

According to the lexical quality hypothesis, high proficient (fast) readers have well-specified lexical representations which enable automatic word identification and less context decoding, while low proficient (slow) readers rely on context for word identification during reading due to their imprecise lexical quality. In contrast, the predictive coding framework assumes that high proficient readers rely more on their reading experience to predict the upcoming context compared to low proficient readers. However, it is still unclear how skilled readers with different levels of reading proficiency rely on context information (e.g., predictability) for word processing during Chinese reading. In two experiments, the present study aimed to investigate individual differences in the use of predictability for word identification by using the eye-tracking technique.

In Experiment 1, eye movements of fast and slow readers were recorded while they were reading sentences containing predictable or unpredictable target words, with the aim to investigate the differences in predictability effects between the two groups. Sixty pairs of predictable-unpredictable target words were selected, each of which was embedded into the same sentence frame. Fifteen fast and 15 slow readers, selected from a group of 66 participants based on their reading rates, participated in Experiment 1. In Experiment 2, parafoveal previews of the 60 predictable target words (identical word, visually similar pseudocharacter, unpredictable word or visually dissimilar pseudocharacter) were manipulated by using the boundary paradigm to explore how parafoveal preview influences processing of predictability information in the fast and slow readers. The eye movements of 20 fast and 20 slow readers, selected from a group of 80 participants on the basis of their reading rates, were recorded while they were reading sentences containing predictable target words with different previews in Experiment 2.

The results showed that fast readers fixated shorter and less on the target words and were more likely to skip the target words than slow readers. In Experiment 1, although reliable predictability effects with shorter fixations for predictable than unpredictable words were found, it did not interact with reading groups (see Table 1). However, results in Experiment 2 showed robust parafoveal preview effects on the target word which interacted with reading groups (see Table 2). In particular, the two groups had the same first-pass fixation times (i.e., FFD, SFD, GD) at the target words under the identical previews, while slow readers made longer fixations than fast readers at the targets with unpredictable previews (slow readers: ts > 4.10, ps < 0.001; fast readers: for FFD and SFD, |t|s < 1.67, ps > 0.05; for GD, b = 0.06, SE = 0.02, t = 2.39, p = 0.017) or unrelated previews (slow readers: ts > 8.87, ps < 0.001; fast readers: ts > 5.16, ps < 0.001). In addition, fast readers skipped target words at a similar probability under both the identical preview and unpredictable preview conditions (b = 0.03, SE = 0.02, z = 1.70, p = 0.088), while slow readers were less likely to skip target words with unpredictable previews than identical previews (b = −0.04, SE = 0.02, z = −2.23, p = 0.025).

The current findings indicate that fast and slow readers rely on context to a similar degree during their foveal lexical processing whereas the two groups show different utilization of previews of the predictable word during their parafoveal processing. To be specific, compared to fast readers, slow readers are inefficient in activating the predictable word with a visually similar preview; moreover, slow readers are disturbed more by the unpredictable preview or the visually dissimilar preview for their lexical processing, which suggests that slow readers are less effective in suppressing unrelated or inappropriate information during reading. Such findings provide evidence for the lexical quality hypothesis and are in support of the linguistic-pro?ciency hypothesis related to individual differences in the E-Z reader model.

Key words: predictability, foveal processing, parafoveal preview, fast reader, slow reader, Chinese reading