Shadow

Detecting Inconsistencies While Reading

Robert F. Lorch, Jr.& Clarese Lemberger, University of Kentucky (rlorch@pop.uky.edu)

College students read 8 experimental texts and 8 filler texts, each of which fit on a single screen of a 22� computer monitor when triple-spaced.  The structure of each experimental text was the same:  The second, �context sentence� established information (e.g., mutual funds produce a high rate of return) that was referred to later in the text.  The last clause of the sixth sentence (�target clause�) established information that was potentially consistent (i.e., the popularity of mutual funds was rising) or inconsistent (i.e., the popularity of mutual funds was falling) with the context sentence.  Finally, the seventh, �connector sentence� explicitly related the target clause to the context sentence (i.e., �This is because of their rate of return.�).  It was anticipated that reading would be generally disrupted when the relationship between the context and target information was inconsistent as opposed to consistent.  The magnitude of the disruption was taken as an indicator of the extent to which processing of the target clause and connector sentence involved accessing the context information.  Our interest was in whether access to the context information would be influenced by a second manipulation: In half of the texts, the topic established in the context sentence was maintained throughout the text; in the remaining texts, the topic switched in the third sentence then switched back at the start of the sixth sentence.

            Each experimental text was analyzed into five regions: target clause, connector sentence, the final sentence of the text, context sentence, and all other sentences (all of which preceded the target clause.  Gaze durations were computed for each region beginning with the first fixation on the target clause.  This means that all gaze durations on the context sentence and other sentences represent looks back to those regions following processing of the target clause.  Overall, gaze durations were longer for consistent than for inconsistent texts in all regions except the final sentence of the text.  Separate analyses of the effects of inconsistency were conducted for the conditions where the topic switched or remained constant.  When the topic switched, the effect of inconsistency was not reliable in any region of the text.  In contrast, when the topic remained constant, gaze durations were significantly longer in the inconsistent condition on the target clause, the connector sentence, and the other sentences.  This pattern of results demonstrates that a switch of topics interfered with access to information established before the topic switch.