Reading multiple documents on a health-related issue: the roles of a text-highlighting tool and re-reading behaviour in integrated understanding

ABSTRACT

This study aimed to investigate the roles of a text-highlighting tool and readers’ re-reading behaviour in their integrated understanding of multiple documents. University students (N = 95) read five partly conflicting documents on a health-related issue on a touch display with or without a text-highlighting tool. Integrated understanding of documents was assessed by the number of intertextual connections in essays written after reading and by a source-content mapping task. The provision of the text-highlighting tool resulted in longer initial reading times even when subtracting the time taken for highlighting, but shorter re-reading times, particularly for participants with a high number of re-readings. Further, only for participants with a high number of re-readings, the provision of the text-highlighting tool resulted in more intertextual connections than when no text-highlighting tool was provided. Participants’ source-content integration was positively related to the number of re-readings, regardless of whether the text-highlighting tool was provided. Finally, additional exploratory eye-tracking analyses revealed that for two out of the five documents, participants in the with-highlighting condition focused on significantly smaller parts of the documents during re-reading than controls.

KEYWORDS:

• Text-highlighting tool
• multiple documents
• multiple document integration
• re-reading behaviour
• eye-tracking

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Please note that in the study by Kobayashi (2009) participants received one of two reading instructions, namely, to form an opinion or to find relations between texts. However, Kobayashi (2009) only found differences in the relation-finding group. These are the results that are reported here.

2 Based on the regression coefficients provided by Leroy et al. (2021) for the interaction between text-highlighting and the number of revisits on the number of intertextual connections in essays, we conducted a power analysis using R-package interactionPoweR (Baranger 2022). More specifically, we based our power analysis on five simulated data sets generated via the function generate_interaction by using the reported regression coefficients by Leroy et al. (2021). From these data sets we then assessed the correlation coefficients in question using the function test_interaction, and used these for the actual power simulation via the function power_interaction (with 1000 iterations each, and an alpha-level of 0.05) for N = 30 up to N = 100 in steps of 10. These analyses revealed an average power of at least .80 for 50 participants (.59 for N = 30, .74 for N = 40, .82 for N = 50, .89 for N = 60, .92 for N = 70, .94 for N = 80, .97 for N = 90, .98 for N = 100). Please note, however, that due to the exploratory nature of the analyses regarding our second measure of readers’ integrated understanding, we decided for a larger sample size than suggested by the power analysis.

3 On average, participants took M = 3.19 min (SD = 1.46 min.) to write their prior topic knowledge essays that comprised an average of M = 64.66 words (SD = 32.24 words).

4 This interaction was marginally significant for the 2 sec threshold, and significant for the 4 and 5 sec thresholds (see Table S4 of the Supplemental Material for all analyses).

5 Note that these analyses are reported for the 3 sec threshold here. Yet, this pattern of results was also the same for the 2, 4, and 5 sec thresholds (see Table S5 of the Supplemental Material for all analyses).

6 For the 2, 4, and 5 sec thresholds the effect was also marginally significant (see Table S4 of the Supplemental Material).

7 This interaction was also significant for the 1, 2, and 4 sec thresholds, and marginally significant for the 5 sec threshold (see Table S7 of the Supplemental Material).

8 Note that these analyses are reported for the 3 sec threshold here. Yet, this pattern of results was also the same for the 1, 2, 4, and 5 sec thresholds (see Table S7 of the Supplemental Material for all analyses).

Additional information

Funding

This work was partly funded by the Leibniz ScienceCampus Tübingen “Cognitive Interfaces”.