Automatic phonological priming in visual word recognition

LINDSAY J. EVETT MRC Applied Psychology Unit 0 1 2 Cambridge CB 0 1 2 EF 0 1 2 England 0 1 2 0 David Taylor is now at the Department of Psychology, University of Newcastle upon Tyne , Claremont Place, Newcastle upon Tyne, NEI 7RU, England . The authors would like to thank Clive Frankish, MikeHarris, and Penny Wilcoxfor assistance with computing, and Pippa Lopez for typing the manuscript. We would also like to thank Max Coltheart and two anonymous referees for comments. Address reprint requests to Lindsay Evett, MRC Applied Psychology Unit , 15 Chaucer Road, Cambridge CB2 2EF, England 1 DAVID E. TAYLOR University ofBristol , Bristol BS8IHH, England 2 GLYN W. HUMPHREYS Birkbeck College, University ofLondon , London WCIE 7HX, England Theories of visual word recognition have proposed that a word's phonological properties can be involved in reading visually presented words. Further, it is commonly supposed that this phonological information can be arrived at in at least two ways: (1) by looking it up after identifying the word visually (a lexical route) or (2) by rule-governed translating of the word's orthographic code (a nonlexical route). Four experiments were conducted to examine whether phonological information is automatically accessed in visual word recognition, and, if so, how this occurs. A priming technique was used with a display sequence of mask, prime, target, mask. Subjects were asked to make written responses to any words that they thought were present, and prime identification was minimal. A facilitatory effect of phonological priming on target identification occurred when primes were homophones of targets. However, no similar facilitation occurred when the prime was a nonword homophone of the target. Further, the homophone priming effect was found irrespective of whether targets followed the spelling-to-sound rules of English. The results suggest that automatic access to phonology can occur in visual word recognition and that it operates by means of a lexical route. - reported by Hillinger (1980, Experiment 3). He used a priming technique in which subjects made a lexical deci sion about visual prime- and target-letter strings. In a test that assessed whether phonological information is automatically accessed, performance when targets were primed by phonologically similar but graphemically different words (e.g., EIGHT-MATE) was compared with performance when primes were neutral (e.g., ******-MATE). Hillinger found that, relative to the neutral-prime baseline, lexical decisions were facilitated by phonological priming. This indicates that phonological information from the prime affected target recognition. There are at least two ways in which such a facilita tory priming effect could occur. One is by the prime automatically activating representations common to both stimuli. The other is by subjects using phonological information from the prime to anticipate the target (Posner, 1978; Posner & Snyder, 1975). Such an antici patory strategy would lead to incorrect expectations when primes and targets are unrelated, precipitating an inhibition effect relative to the neutral condition (Mclean & Shulman, 1978; Neely, 1977). Since, in Hillinger's (1980) study, no inhibition effect was found when primes and targets were phonologically unrelated (e.g., VEIL-MATE), it appears that subjects did not anti cipate targets from the phonological properties of primes. This suggests that priming occurred because phonologi cal information was automatically extracted from primes, and this facilitated the recognition of subsequent phonologically related targets. However, there is a problem with this conclusion since phonological coding may only have been under taken because of the nature of the nonwords used. Sub jects in HiIIinger's (1980) study had to discriminate between orthographically regular, pronounceable non words (e.g., TANE) and words to make correct lexical decisions about both primes and targets. Under these circumstances, phonological coding may occur because of the strategy used by subjects to facilitate lexical decision; changing the nature of the nonwords can minimize the occurrence of such coding (see Shulman, Hornak, & Sanders, 1978). Further, any strategically produced phonological code may take some measurable amount of time to decay. Thus, even if subjects do not anticipate targets on the basis of such information from primes, it will influence ("automatically") the process ing of targets presented in close temporal succession. To ensure that priming reflects only that information that is unintentionally accessed in word recognition, there must be stricter control over the strategic process ing of primes and targets. A study that goes some way toward this was reported by Tanenhaus, Flanigan, and Seidenberg (1980). They used a Stroop color-naming task, in which colored target words were preceded by phonologically similar or dissimilar word primes. Color naming latencies to targets were slower when primes were phonologically similar (TOWEL-HOWL) than when they were dissimilar (BEND.HOWL). Since anticipating targets from primes disrupts color naming, it seems likely that this effect reflects the information automati cally accessed by primes (however, see Logan, 1980). Unfortunately, a further problem exists for this study. Tanenhaus et aI. (I980) failed to incorporate a graphemic control condition, in which phonologically different primes and targets matched the letters in com mon between phonologically similar stimuli. Without this control, it is impossible to tell whether the priming effect was graphemic rather than phonological in origin (cf. Evett & Humphreys, 1981; Rayner, 1979). To examine the information automatically accessed by primes, we need to prevent the intentional use of prime information by subjects and to carefully control for all the attributes of the stimuli that could produce priming effects. These were the aims of Experiments I and 2 here. One way to prevent the intentional usage of prime information is to present primes under conditions that minimize their explicit identification but that neverthe less allow priming to occur. Such conditions have recently been demonstrated by Allport (I 977), Evett and Humphreys (1981), Fowler, Wolford, Slade, and Tassinary (I981), and Marcel (in press). These studies have shown that semantic priming effects can occur when primes are backward-pattern masked so that they cannot be identified. Evett and Humphreys (I981) tested for phonologi cal priming under such conditions. They used a four field masking procedure in which the presentation of the prime and target stimuli was preceded and followed by the presentation of a pattern mask. Although subjects were required to identify any words present, there was minimal identification of primes. This shows that subjects could not use prime information intentionally, and, therefore, that any priming effects were occurring automatically (Posner, 1978). Tes (...truncated)