Harris, C.L., & Morris,A.L. (1998) Orthographic repetition blindness. Manuscript under review at The Quarterly Journal of Experimental Psychology.
Morris, A.L. & Harris, C.L. (1998) A sublexical locus for repetition blindness: Evidence from illusory words. In press, Journal of Experimental Psychology: Human Perception and Performance.
When two orthographically similar words are briefly sequentially displayed (such as career AREA), the repeated letters cause interference, resulting in failure to report the second word. This talk describes how manipulations of orthographic relatedness can be used to test models of orthography, top-down effects on lexical access, and whether abstract letter identity is represented in both hemispheres. Orthographic RB is of use to all researchers who use words as stimuli or who study the perception-cognition interface.
Two experiments failed to replicate Chialant and Caramazza's (1997) finding that identity RB decreases, but orthographic RB increases, as a function of lag between two critical words. We explain their findings as due to guessing strategies, combined with research participants being affected by repetition priming at long lags and repetition blindness at short lags, with non-identical words benefiting less from repetition priming. We present additional experiments demonstrating that orthographic RB is due to tokenization problems at the level of sequential letter clusters.
Repetition blindness (RB) in non-identical words depends on number of shared letter clusters, coarsely coded for position. Strong RB occurs for non-neighbors (orchestra, chest; bridesmaid, rides) and words sharing 3 alternating letters (itself insult) and 3 middle letters (ensure insult). RB was found for words sharing even a single initial letter (about aware). ŠLetter stealingÆ is suggested by instances like misreport of flower as flow (given career flower). The data are modeled using letter co-occurrence statistics.
We introduce a novel paradigm, called "misreading-RB", for studying the effects of sentence context on word recognition. Repetition blindness (RB) is the failure to report the second of two identical words presented in RSVP sentences (Kanwisher, 1987). RB has also been observed between orthographically similar words, such as candy and can (as in candy bars can make you fat). RB is a strong effect and occurs even when the resulting sentence is ungrammatical or otherwise anomalous.
Previous studies showed that biasing context causes subjects to misread words in RSVP (Potter, Moryadas, Abrams, and Noel, 1993). In our first experiment, we created leading context designed to induce misreading of the first target word. When subjects misread we can prevent forest hires as prevent forest fires, they were protected from repetition deficits; RB for a subsequent occurrence of hires was low, and of a similar magnitude to that obtained from input actually containing prevent forest fires. A correct reading of the sentence resulted in significant RB for the repeated hires.
Our second experiment showed that misreading can also cause a repetition deficit. When subjects misread chocolate dandy bars as chocolate candy bars, RB deficits were as large as when the word candy was actually present in the visual input. This indicates that words activated by context have the same ability to cause repetition blindness for orthographically similar words as words which are activated by visual input.
We will discuss two implications of these experiments for theories of sentence processing. (1) The activation status of words activated via context is indistinguishable from that of words activated by visual input. (2) Misreading under conditions of constraining context happens within a few hundred milliseconds of stimulus input, rather than resulting from reconstructive processes occurring at a later stage of sentence comprehension.
When two orthographically similar words are briefly and successively displayed, the second word is often difficult to detect or recall, a deficit known as repetition blindness, or RB (Kanwisher, 1987). Two experiments used word-nonword pairs to test predictions of a computational model based on similarity inhibition (Bavelier & Jordan, 1992) vs. predictions of a sublexical model (Harris & Morris, 1996, 1997; Morris & Harris, 1997). One striking finding was of strong RB even for a single repeated letter (cope carn; hot hix). Results generally supported a sublexical model where only the shared letters are affected by RB, and each shared letter can be differentially affected in a probabilistic manner.
Repetition blindness (RB) can be influenced by letter clusters activated by context rather than by visual inputs. When subjects misread coffee nups (in RSVP) as coffee cups, RB for a subsequent word nuts was reduced. Letter clusters left over from an RB'ed word can fuse with a subsequent word fragment; the stimulus hand grand avy resulted in report of the illusory word gravy, but the no-RB stimulus hand gr avy did not result in illusory words.