Multiple Session Masked Priming: Individual differences in orthographic neighbourhood effects

Multiple Session Masked Priming was used to investigate differences between individuals in the fine-tuning of their lexical representations. The form-priming effects were determined separately for each of the 50 participants, from the patterning of masked priming effects over three neighbourhood (N) levels. At each N level, primes varied in their orthographic similarity to the target Identity, OneLetter-Different and All-Letters-Different. An analysis of the pooled data showed patterns of masked priming consistent with other group studies, but considered separately, an array of individual differences in the tuning of lexical representations was observed. Inconsistent findings and conflicting evidence characterise much of the word recognition literature, and have lead to a plethora of theories on many aspects of word recognition. One explanation for the conflicting evidence regarding theories of word recognition, is the possibility that it reflects individual differences. The problem here lies in the fact that support for one model or another has generally been drawn from standard group studies, where outcomes are averaged over the sample of participants. Averaged outcomes may show what is generally true for that sample, but fail to capture the critical variations between individuals. In order to study individual differences a task must be found that will provide reliable data concerning the processing capacities of each participant. There are two issues here. First, the task used must reflect the automatic processes of lexical access, and second, the task must generate sufficient data on each participant to enable stable conclusions to be drawn about their individual performance; that is, we need to be able to generate reliable individual profiles of lexical processing. Although there is some debate (e.g. Bodner & Masson, 1997; Bodner & Masson, 2001), Forster and Davis’ (1984) Masked Priming paradigm is widely considered to be an ideal tool for examining automatic lexical processing. In masked priming, a prime word/nonword is presented very briefly (50-60ms), forward masked (for 500ms) by a row of hash marks (######) and backward masked by an upper case target (500ms), to which a timed lexical decision response is made. The relationship between prime and target is manipulated in some way and if performance on the target is enhanced by a particular prime-target relationship, relative to a control condition, this relationship is taken to reflect the properties of written words important for lexical access. Masked priming has been argued to enable the investigation of lexical access processes free from any influence of more central and strategic processes (e.g. Forster, 1998). This is achieved in the way the prime is presented, making it typically unavailable for conscious report by participants. Critically, while the masked prime is not available for conscious report, a variety of priming effects are consistently observed. Such facilitatory priming effects include repetition-priming (e.g. Forster & Davis, 1984) where prime and target are the same word (e.g. farm – FARM) and form-priming (e.g. Forster, Davis, Schocknect & Carter, 1987; Forster & Veres, 1998) where the prime is of similar orthographic form (e.g. firm – FARM). The absence of both expectancy effects (e.g. Forster, 1998) and priming for nonwords (e.g. Forster & Davis, 1984; Forster et al., 1987; cf. Bodner & Masson, 1997), in addition to the existence of semantic(e.g. Perea & Gotor, 1997) and crosslanguage translation-priming (e.g. Kim & Davis, 2003) indicate that masked priming effects reflect lexical level processing. While the masked priming task provides a means of examining the automatic processes of lexical access, a new variant of this task was required to enable the collection of sufficient data from each participant that they could stand as an experiment in their own right. The Multiple Session Masked Priming paradigm (e.g. Byrne, Yelland, Johnston & Pratt, 2000; Yelland & Byrne, 2001) achieves this by repeatedly testing each participant on the same experiment. Since participants are unaware of the primes, the primetarget relationship cannot be realised, even over repeated test sessions. Thus, in the Multiple Session Masked Priming paradigm, the participant’s experience is simply one of repeated exposure to the target items. Used previously to reveal marked individual variation in the use of orthographic and phonological input codes for lexical access (e.g. Byrne, Yelland, Johnston & Pratt, 2000; Yelland & Byrne, 2001), the Multiple Session Masked Priming paradigm has proved useful for investigating individual differences in lexical processing. This study aims to look at the use of this new technique in other areas of the word recognition literature, namely orthographic neighbourhood effects. The neighbourhood effect was first investigated by Coltheart, Davelaar, Jonasson and Besner (1977), who coined the term neighbourhood (N) density. A word’s N density is simply the number of words that can be made from it by changing one letter. For instance, a high N word like torn has many neighbours – corn, born, turn, town, tore etc. In this way, N provides a rough index of a word’s similarity to other words. The effects of such orthographic similarity have been used by researchers to model constraints on, and organisation of, the internal wordrecognition system. Previous experiments looking at the effects of N density have shown form-priming effects when using word or nonword primes that are one-letter-different from their word targets, however, such effects are present for low N density words only, not high N (e.g. Forster et al., 1987). The form-priming effects seen at the low N density are reported consistently, albeit are smaller in magnitude than those observed in repetition-priming (e.g. Andrews, 1997). The lack of form-priming effects in high N words is usually described as an increased tuning of these lexical representations. Forster et al. (1987) first proposed a now commonly used explanation for this change in tuning, whereby a word’s lexical representation becomes more finely tuned in response to increasing N density. A word with many neighbours needs to ensure only exact matches can activate its representation, otherwise a large number of these neighbours will need to be considered and errors in activation are more likely. In contrast, a word with only few neighbours can afford to be less stringent (i.e. it can be activated by a one-letter-different neighbour) as there are only a small number of possible lexical candidates. A range of word recognition theories and models have been provided for these results. Such accounts, however, will only prove useful if either, all readers show the same pattern of priming effects as a function of N density, or they are able to account for individual variation in sensitivity to N levels . The question addressed in this study is whether there are individual differences in the priming effects for similar form, indicating differences in the fine-tuning of lexical representations.