Misperception of orthographic neighbors during silent and oral reading.

Abstract

The study examined whether words are misperceived during natural fluent reading and the extent to which contextual and lexical properties bias perception. Target words were pairs of orthographic neighbors that differed in frequency. Pretarget context was neutral (Experiment 1) or biased toward the higher frequency member of the pair (Experiments 2 and 3), and posttarget context was neutral, congruent, or incongruent. Critically, incongruent context was constructed so that it was congruent with the target's neighbor. First-pass viewing showed only effects of target frequency. During silent reading (Experiments 1 and 2), rereading measures showed that the target frequency effect was smaller in the incongruent posttarget context condition than in the neutral and congruent conditions, and this occurred irrespective of prior context. Presumably, lower frequency words were less impeded by incongruent context because they were often misperceived as a congruent higher frequency neighbor. An oral reading task (Experiment 3) showed that the lower frequency target was more often misread than the higher frequency neighbor, and this proneness to error was influenced by posttarget context. Although target frequency influenced proneness to error, biased prior sentence context appeared to influence the construal of sentence meaning to accommodate incongruent targets and posttarget context. (PsycINFO Database Record

Regressions during Reading

“…Using eye-movement-contingent display changes to manipulate the visibility of text to the left of a fixation, so that it was either masked or re-readable after a regression, Schotter et al [27] showed that sentences with garden path constructions were understood more successfully when readers re-read text after a regression. Readers also appear to use regressions to correct word identification errors [28], and to fill in “missed” sentence and story parts that were viewed during episodes of mind wandering and attentional lapses [29].…”

“…Using a similar approach, we examined the rate of regressions that moved the eyes onto the source of a presumed processing difficulty [28]. The materials were constructed so that controlled target words would be the source of a processing difficulty in an incongruent condition but not in a congruent condition.…”

“…For this, we re-analyzed the results of two experiments [28] in which the identity of a target word was manipulated in one of the experimental conditions so that the execution of a regression to it resolved a processing difficulty when a target was incongruent with prior context. As noted earlier, targets were followed by different sentence context: One in which the target and its subsequent context were congruent, one in which they were incongruent, and one in which their relationship was “neutral”.…”

“…No change in regression frequency, or a lesser change, was expected for the congruent and the neutral conditions. Figure 4 shows regression rates to targets as a function of trial sequence (beginning vs. ending half) and sentence context (74 participants who participated in Experiments 1 and 2 [28]). As can be seen, regression rates did not increase over the course of the experiment in the incongruent condition, and the interaction of trial sequence with context was negligible ( p > 0.3).…”

“…Why did readers’ use of regressions not increase over the course of the experiments in the incongruent condition? Some readers may have under-used large regressions because they were content with a somewhat odd—but “good-enough”—representation of sentence meaning [28]. Other readers may not have learned to use regressions strategically.…”