Comprehension of Online Cancer Information: A Propositional Assessment of Readability, Inferences, and Coherence

Abstract

This study was designed to examine the comprehension processes used when reading cancer information found on the Internet and its relation to readability. The focus of the research was on the utilization of an alternative methodology, propositional analysis, to provide indices of textual difficulty that complement standard readability formulas. Kintsch's (1998) framework about discourse comprehension was used to distinguish between textbase and situational factors influencing comprehension. <br /><br /> This study analysed the verbal protocols of 16 community dwelling older adults. They each read a pair of either breast, prostate, or colorectal cancer web pages at low and high readability levels. Propositional density and coherence were measured for the web pages; recall, concepts, inferences, and coherence were measured for the protocols. Coherence was also captured using network representations. The readability formulas used were the Flesch-Kincaid (F-K), Flesch Reading Ease (FRE) and SMOG. Results showed that propositional density did not agree with readability scores, and that the readability formulas did not even produce consistent results among themselves. Results also showed that readability was not associated with web page coherence. Analysis of the protocols revealed marked individual differences for the kinds of information recalled, the types of inferences made, and the coherence of mental models. Variations in background and personal interest appeared to influence whether superordinate or subordinate propositions were recalled. Dependence on prompted versus non-prompted interview questions also varied by individual. <br /><br /> The findings of this study suggest that propositional analysis should be considered as a complementary methodology to readability formulas. Relying solely on these formulas as an indictor of comprehension may mislead online health providers that their information will be understood. The findings also highlight that that individual create distinct and personalized mental models when presented with web pages that are influenced by text and situation based factors.