![Sample our Behavioral Sciences journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/110801/TOC-Subject-Sample-2021-Behavioral-Sciences.jpg"})
ABSTRACT
Prior research using the boundary paradigm suggests that Chinese readers only process word n + 2 in the parafovea when word n + 1 is a single character, high-frequency word. We attempted to replicate these findings (Experiment 1), and investigated whether greater n + 2 preview effects are observed when word n + 1 and n + 2 form an idiom rather than a phrase (Experiment 2). Experiment 1 replicated prior findings, although additional analyses of word n + 1 and n + 2 as a single region revealed significant preview effects regardless of word n + 1 frequency. In Experiment 2 there was a main effect of phrase type, such that idioms were read more quickly than phrases, and significant n + 2 preview effects. There was no interaction between these variables, suggesting that idioms are not parafoveally processed to a greater extent than phrases. These results suggest that n + 2 preview effects in Chinese occur under several circumstances. Factors influencing the observation of these effects are discussed.
Acknowledgements
We would like to thank Jinmian Yang and Keith Rayner for supplying the stimuli used in Experiment 1.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. Terminology differs between the idiom and compound literature for this factor. Given the focus of this paper we adopt the terminology traditionally used for idioms (i.e. decomposable vs. non-decomposable) rather than compounds (i.e. transparent vs. opaque).
2. We were forced to include different participant groups due to where the first author was based at the different times of testing, rather than for any theoretical reason. It is worth noting that all participants were exposed to all experimental conditions, making this a within subject design.
3. It should be noted that effects appearing on word n + 1 and the composite region could, of course, also be due to delayed parafoveal-on-foveal effects (Risse & Kliegl, Citation2012), as well as a preview benefit appearing in mislocated fixations. Regardless of the exact reason why effects may be distributed over a larger region, at a theoretical level this analysis still demonstrates an effect of upcoming parafoveal information having been processed prior to the eyes crossing the invisible boundary.
4. The probability of the second character after the boundary forming a word with other characters, as opposed to a single-character word, was 0.78. This was calculated by dividing the frequency that this character appeared as part of a multi-character word by the total frequency of the character.
5. While no significant interactions were observed on word n + 2 or in our composite region, there were clearly numerical differences in the size of preview benefit effects for idioms and phrases, albeit of inconsistent directions across measures. In order to affirm that we had obtained a null effect, rather than failed to obtain an effect due to a lack of statistical power, we undertook Bayes factor analysis. Bayes factors provide a value indicating the extent to which a dataset provides support for one of two competing hypotheses. Decreasing Bayes values (below 1) provide increasing support for one hypothesis, while increasing Bayes values (above 1) provide support for the alternative hypothesis (see Kass & Raftery, Citation1995; Rouder, Morey, Speckman, & Province, Citation2012 for a more in-depth discussion). We calculated Bayes factors using the BayesFactor package (Morey, Rouder, & Jamil, Citation2015) in R to ascertain whether our results for word n + 2, and for our composite region were more adequately reflected by a model including an additive, or interactive relationship between phrase type and n + 2 preview type. For all reading time measures on word n + 2 Bayes factors were between 0.1 and 0.15, representing substantial support for an additive model. For first fixation, gaze duration, go past time, and single fixation duration on the composite region, Bayes factors were 0.07, 0.07, 0.12, and 0.38. Three of these values also represent at least substantial support for the additive effect over the interactive effect, while one represents anecdotal support. These values suggest that the lack of a significant interaction in our main analyses were due to the null effect being supported, rather than a lack of statistical power.
6. Although the predictability from the text prior to word n + 2 was slightly higher in the phrases used in Yang et al.’s (Citation2012) study than the phrases in the current Experiment 2 (t(19) = −1.79, p = 0.089), the difference favored Yang et al.’s stimuli producing larger word n + 2 preview effects, which was obviously not the case in the current results. Therefore, in the current study the predictability of word n + 2 could not be the cause of larger word n + 2 preview effect.
7. It could be argued that a ceiling effect might exist for the word n + 2 preview effect observed in the current study, and this ceiling effect might prevent a larger word n + 2 preview effect being observed in idioms. Although we do not reject this possibility, we do consider it unlikely. Indeed, our initial question was whether there would be any n + 2 preview effects at all given the frequency of word n + 1.
8. We thank an anonymous reviewer for this insight.