Effects of Parafoveal Plausibility During Reading

Effects of Parafoveal Plausibility During Reading Laura Wakeford (l.j.wakeford@dundee.ac.uk) School of Psychology, University of Dundee Dundee, DD1 4HN Wayne Murray (w.s.murray@dundee.ac.uk) School of Psychology, University of Dundee Dundee, DD1 4HN Abstract There is controversy concerning the question of whether meaning can be extracted from a parafoveal word during reading and whether this might occur in an overlapping fashion with the lexical processing of the currently-fixated word. We suggest that previous attempts to investigate this have been bedevilled by problems associated with the use of priming methodology. Instead, we used an eye movement contingent change technique and manipulated the plausibility of the parafoveal preview, resulting in it being either valid, a plausible alternative, anomalous, or an illegal letter string. The results showed (a) a meaning-based parafoveal-on-foveal effect, (b) preview benefits driven by both orthographic and semantic influences, and (c) continuing disruption associated with orthographically dissimilar previews. We suggest that this pattern is most consistent with models of eye movement control that allow for distributed attention during reading. Keywords: Eye movements; preview benefit; plausibility; parafoveal-on-foveal effects; boundary technique; reading. Introduction The nature of Preview Benefit (PB) – the advantage accruing to the reader from an accurate parafoveal preview of the following word – critically informs our understanding of the reading process, indicating those features which are, and are not, extracted from an as yet unfixated word. Using the boundary paradigm (Rayner, 1975), many studies have shown that both orthographic and phonological features appear to be extracted from parafoveal words (see Schotter, Angele & Rayner, 2012, for a review); however, evidence for a semantic PB remains controversial (see Radach & Kennedy, 2013; Rayner, White, Kambe, Miller & Liversedge, 2003). While both serial (e.g., E-Z Reader; Reichle, Warren, & McConnell, 2009) and parallel (e.g., SWIFT; Engbert, Nuthmann, Richter & Kliegl, 2005) models of eye movement control during reading provide accounts of orthographic and phonological PB, only parallel models appear capable of accounting for semantic PB. In serial models, lexical processing is restricted to one word at a time, with attention moving to the parafoveal word only when the currently fixated word has been fully identified. Serial models therefore typically only accommodate very early stages of word recognition occurring on parafoveal words before a saccade remarries fixation location with attention. In contrast, in parallel models, all words within the perceptual span can be processed simultaneously, up to and including the level of semantic processing. Studies investigating semantic PB have typically manipulated the sematic relatedness of the preview and the target word, on the basis that responses to semantically related word pairs are facilitated compared to unrelated pairs (Meyer & Schvaneveldt, 1971). By extension, it is suggested that semantically related previews should facilitate target viewing compared to unrelated previews. Using the boundary paradigm, Rayner Balota and Pollatsek (1986) asked participants to read sentences such as “My younger brother has brilliantly composed a new song for the school play”, in which the pre-fixation preview of “song” was either “song” (valid), “tune” (related), “door” (unrelated), or “sorp” (a visually similar nonword). Only once the eye passed an invisible boundary, located before the critical word, did the target word “song” appear. Despite showing that their critical words produced facilitation in a classic priming experiment, Rayner et al found no evidence for a semantic PB during reading. However, in this example sentence, we see that the word to the left of the target contains only three letters, and as short words are frequently skipped (Rayner & McConkie, 1976), the prior fixation may in fact have fallen two words to the left of the target, seriously reducing the chance of it eliciting a semantic PB. A more general problem with experiments investigating semantic PB using associative previews is that while there may be semantic facilitation, there is also a word change that might be expected to give rise to some form of inhibition. Semantically related word pairs, such as north– south, rattle-bottle and arms–legs, (from Rayner et al, 1986), have very different meanings, and this could exert an inhibitory effect on on-going sentence interpretation. Rayner et al (1986) attempted to test this possibility by asking participants to rate their sentence pairs for similarity of meaning and reanalysing the results from only the 20 sentence pairs rated as most similar in meaning. Since this analysis again failed to show a semantic PB, they dismissed this as an explanation for their null result. However, a measure of overall sentence meaning does not necessarily capture the extent to which a local change in word meaning might have disrupted the reading process at the point at which it occurred. We conclude, therefore, that interference resulting from word change remains a possibility.