Wrapping up Sentence Comprehension: The Role of Task Demands and Individual Differences

ABSTRACT

This study used wrap-up effects on eye movements to assess the relationship between online reading behavior and comprehension. Participants, assessed on measures of reading, vocabulary, and spelling, read short passages that manipulated whether a syntactic boundary was unmarked by punctuation, weakly marked by a comma, or strongly marked by a period. Comprehension demands were manipulated by presenting questions after either 25% or 100% of passages. Wrap-up effects at punctuation boundaries manifested principally in rereading of earlier text and were more marked in lower proficiency readers. High comprehension load was associated with longer total reading time but had little impact on wrap-up effects. The relationship between eye movements and comprehension accuracy suggested that poor comprehension was associated with a shallower reading strategy under low comprehension demands. The implications of these findings for understanding how the processes involved in self-regulating comprehension are modulated by reading proficiency and comprehension goals are discussed.

Acknowledgments

We thank Kelly Dann and Justine Greenaway for their assistance with data collection. Portions of these data were presented at the 58th Meeting of the Psychonomic Society, Vancouver, Canada.

Conflict of interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.

Notes

1. An additional five participants were excluded because of eye-tracker calibration failure.

2. The standard time limits for the vocabulary and passage comprehension tests were reduced by half (to 7.5 and 10 minutes, respectively) because these shorter time limits yield a more normal distribution of test scores in university student samples like that tested here (Andrews et al., 2020).

3. Eight items which were answered correctly by fewer than 50% of participants were excluded from the calculation of mean accuracy.

4. Models that included more complex random-effects structures failed to converge.

5. Models that included item random slopes for reading proficiency failed to converge. The total duration model failed to converge with the item random slope for comprehension load.

6. Further evidence of the pay-now-or-pay-later tradeoff was provided by the distribution of regressions into each of the targets. A significant interaction between the non-boundary/T1 difference and boundary marking (b = 0.86, SE = 0.09, z = 9.67) occurred because regressions to unmarked T1 words were almost twice as frequent as for marked T1 words (18.5% vs 9.8%), while non-boundary words showed a small difference in the opposite direction (10.6% vs 11.6%). Like the effects of boundary marking effects on go-past duration, this interaction was more marked in lower proficiency readers (b = −0.07, SE = 0.03, z = −2.46). These effects support Hirotani et al.’s (2006) prediction that readers are less likely regress across a punctuation-marked boundary at which they implemented wrap-up than within a clause. We thank Kiel Christianson for recommending that we analyze the regression data.

Additional information

Funding

This research was supported under Australian Research Council’s Discovery Projects funding scheme (project numbers DP16203224, DP18102705).