ABSTRACT
Little is known about conceptual integration processes and their contributions to memory representations that are constructed in sentence comprehension among adults with underdeveloped literacy skills. We measured word-by-word reading time to examine variation of responsiveness to demands of lexical and conceptual processing during sentence comprehension in a sample of adults with reading proficiency ranging from 4th grade through college-level. Relative to those of their more skilled counterparts, the reading times of adults with lower levels of literacy skill were more sensitive to word-level features, but showed a reduced lengthening at sentence-final words (“wrap up”), which has been argued to reflect conceptual integration processing. They also showed poorer sentence memory. However, regardless of literacy level, readers with better overall sentence memory engaged in a reading strategy marked by a larger sentence wrap-up effect. These findings suggest a pathway to intervention for struggling adult readers.
Conflict of interest
The authors declare no conflict of interest.
Informed Consent
All participants provided informed consent.
Supplementary materials
Supplemented details of this article can be accessed here.
Notes
1. For our target words, the correlation between log frequencies based on the HAL and SUBTL corpora was .956, so the selection of corpus was arbitrary.
2. It should be noted that clause boundary has also been used as a predictor for reading times to examine organizational processing within sentences (e.g., Haberlandt & Graesser, Citation1989b; Payne et al., Citation2012). However, in the current materials set, sentences were relatively short and simple, affording little opportunity to separate out the effects of intra-sentence wrap-up. Additionally, while resource allocation coefficients in sentence processing show fairly reliable individual differences, clause boundary wrap-up tends to be less so (Stine-Morrow et al., Citation2001). Thus, even though clause wrap-up has been shown to be differentially preferred by some readers with reduced WM capacity (e.g., older adults), its effects were not evaluated in the current study.
3. We only report recall for ideas in the target sentence, for which we analyzed reading times. Propositional recall from the target sentence and whole passage was highly correlated, r = .95, suggesting that recall accuracy on the target sentence was representative of recall of the whole passage.
4. As shown in , the variance estimates for the by-subject random slope of Word Frequency in each model are zero. Removing this random effect from all models did not change the overall findings. For the sake of consistency, we kept them in the models.
5. The parameter estimates did not change if these two individual difference variables were excluded from the model.
6. Adults with lower levels of reading skill also showed lower scores on working memory (WM) span (r = 0.57) and fluid ability (r = 0.45). When we fit an additional model including these factors, WM did not predict reading time, but fluid ability did; the effect of reading level remained robust even with these variables in the model. However, in subsequent models with interaction terms, neither fluid ability nor its interactions reached significance.
7. It appears in that the low-proficiency readers may have been at floor for the least memorable propositions. To confirm that it was not just a floor effect that was driving the reduced slope, we analyzed the data including only the propositions that were recalled by more than 40% of the proficient readers. The regression line was y = − 13.4 + 0.87x, with a slope still less than 1, t(147) = 2.03, p < .05.