Searching for a word in Chinese text: insights from eye movement behaviour

ABSTRACT

Locating relevant information in text is an important aspect of the reading process, however relatively few studies have examined this, especially for logographic languages such as Chinese. The present study examines eye movement behaviour during search for a target word in Chinese sentences, compared with reading the sentences for comprehension. Although there were clear effects of word frequency during reading for comprehension, the study shows no evidence for an influence of the word frequency of non-target words on eye movement behaviour during target word search. The results are in line with previous research undertaken in English (Rayner, K., & Fischer, M. H. (1996). Mindless reading revisited: Eye movements during reading and scanning are different. Perception & Psychophysics, 58, 734–747.), such that during search for a target word, eye movement behaviour for non-target words is largely driven by superficial processing of those words. The study also highlights the prevalence of word skipping, indicating that words are often sampled only in visually degraded parafoveal vision during target word search in Chinese.

KEYWORDS:

• Reading
• search
• Chinese
• word frequency
• eye movements

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Sarah J. White http://orcid.org/0000-0003-3943-7149

Notes

1 Note that the present study employed the same search target word throughout. Future work may examine the different issue of whether the word frequency of the target word itself might modulate search behaviour for both target and non-target words (see: Dampuré et al., 2012; Grabbe, 2014).

2 In Chinese, “zebra” is an infrequent word (3.28 counts per million, SUBTLEX-CH), the first character is composed of twelve strokes and the second character is composed of three strokes. The characteristics of any search target word may affect participants’ search behaviour. For example, search may have been facilitated by the visually distinctive and infrequent letter “z” in Rayner and Fischer’s (1996) study, or by the visually simple second character of the target word in the present study. Given Yu et al.’s (2018) findings, skipping rates during target word search may be lower than reported here if the target word characters are more visually complex.

3 For completeness, for the critical word measures, LMM models including task (read vs. search), word frequency (high vs. low) and the interaction were also undertaken. All four measures produced clear effects of task (t/z > 4.4). The t/z values for the interaction term were as follows: Skipping z = 2.10; first-fixation duration t = −1.05; gaze duration t = −2.45; total time t = −2.73.

4 The experiment was designed such that there were 18 items per participant per condition. However due to skipping and data exclusions on average each participant contributed ∼7 data points for each high and low frequency condition for target word search, and ∼14 data points for each high and low frequency condition for reading. Brysbaert and Stevens (2018) recommend a total of at least 1,600 observations per condition. Therefore, in order to provide a very strong test of whether there are effects of word frequency during target word search, given high skipping rates, a larger number of items would be necessary. Ideally future studies would employ sufficient items such that all participants would be likely to generate enough fixations on the critical word per condition to enable examination of any effects of word frequency on the distribution of fixation durations (see Discussion).

Additional information

Funding

The work was supported by an International Exchanges Scheme grant from the Royal Society (IE141063). We thank Simon Liversedge, Raymond Bertram and Bernhard Angele, as well as anonymous reviewers, for their helpful comments on an earlier version of the manuscript.