Estimating the rate of failure to notice function word errors in natural reading
Adrian Staub,Alan Chen,Emily Peck,Natasha Taylor
DOI: https://doi.org/10.3758/s13423-024-02586-1
2024-10-02
Psychonomic Bulletin & Review
Abstract:Skilled readers sometimes fail to notice seemingly obvious errors in text, such as the repetition or omission of a function word or the transposition of two words, suggesting that linguistic knowledge can override bottom-up input at either a perceptual or postperceptual level. The present study investigates the role of this top-down process of error correction in natural reading of extended texts. In previous research, critical sentences have been presented one at a time, and subjects were explicitly tasked with detecting errors. In the present study, each participant read a full newspaper article or pair of articles, with their comprehension tested by multiple choice questions. As a secondary task, participants were also instructed that they should make a mouse click on any errors in the text, without any instruction as to the frequency or nature of any such errors. Each article contained nine intentionally inserted errors involving function words: three repetitions, three omissions, and three transpositions. After removing subjects who did not click on the text at all (leaving n = 165), the median subject made seven clicks, but detected only one of the nine inserted errors. Neither error type nor article type (highly professional vs. amateur) clearly modulated the rate of error detection, though subjects clicked more often overall on the amateur articles. We conclude that previous research has dramatically underestimated the rate at which readers fail to notice these function word errors; in natural reading, they are noticed only rarely. No existing reading model can account for this phenomenon.
psychology, experimental, mathematical