Abstract
In this investigation, we examine the contribution of intrinsic content density (ICD) to measures of expository text processing. In Studies 1 and 2, the factor structure of select text density metrics was examined and refined using two text samples (Ns = 150) randomly selected from an expository text corpus. Scores on the ICD measure based on the entire text sample (N = 300) explained unique variance in readability and text easability. In Study 3, ICD predicted adults’ text ratings of interest and ease of comprehension above and beyond established easability measures. Participants’ text familiarity moderated the relation between ICD and ease of comprehension, revealing a density-facilitative effect for participants more familiar with the text content. Finally, in Study 4, measures of text difficulty, processing, and comprehension were obtained from adult readers using 10 researcher-constructed science texts; evidence of descriptive density effects on each measure was obtained. Implications for future research are discussed.
Acknowledgment
Any opinions expressed in this article are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Disclosure statement
The authors declare that they have no conflicts of interest.
Notes
1 A sample of 150 texts was selected because this text sample size was sufficient to ground the factor analytic approach conducted. This text sample size is also consistent with recent published work evaluating text processing, familiarity, and comprehension among adult readers (see Crossley et al., Citation2017).
2 While one might expect a strong degree of overlap between the metrics representing the incidence of causal verbs and participles and the incidence of verb phrases, the Pearson correlation coefficient between scores on the two metrics was weak to moderate in magnitude, r = 0.27.
3 Communalities describe the amount (i.e., percent) of variance in specific indicators that is explained by the common factor. Communalities greater than 0.20 indicate that the indicator is well accounted for by the factor solution (Kline, Citation2005).
4 This method is also referred to as the maximum a posteriori method.
5 This procedure was used in lieu of structural equation modeling because a model with five continuous outcomes would not have been appropriately identified.
6 This model likewise excluded incidence of causal connectives.
7 Because the text familiarity rating relied on one score, an estimate of reliability for the ratings could not be computed.
8 The compensation amount was determined by Prolific’s pricing estimator and was based in part on estimated study length.
9 We aimed to develop science texts that were above the 50 percentile in each of these metrics. Thus, texts were not matched on these characteristics and there existed variability in the measured degree of these easability metrics on the texts constructed.
10 As noted above, this approach was taken in lieu of a correlational approach based on the size of the text sample administered and the within-subjects approach taken to the examination of RCD in Study 3.