The role of neighbourhood density in transposed-letter priming

Abstract

We investigated how transposed-letter (TL) priming effects are modulated by neighbourhood density in lexical decision task and cross-case same-different task. Robust identity priming and TL priming effects were observed in both tasks. Neighbourhood density facilitated ‘word’ decisions and reduced the size of identity priming effect in the lexical decision task. It had no effect on the same-different task, and did not modulate identity or TL priming. We take these results as indicating that neighbourhood density modulates priming during lexical access. In the lexical decision task, the difference between identity priming and TL priming was reduced for words from high-density neighbourhood. The last finding is at odds with the ‘lexical tuning’ hypothesis which has been used to explain modulatory effects of neighbourhood density on masked form priming effects. We also discuss implication of the results for various letter position coding schemes.

Keywords:

• Lexical decision task
• Lexical tuning hypothesis
• Neighbourhood density
• Same-different match task
• Transposed-letter (TL) priming

Acknowledgements

Thanks are due to Greg McLean, Shaun Greenfield, and Betty Mousikou for research assistance. This research was supported by the Australian Research Council Grant (DP0877084) to SK.

Notes

¹Recently Yarkoni, Balota, and Yap (in press) have developed another measure of neighbourhood density that is not limited to words of the same length. Their measure, Orthographic Levenshtein Distance 20 (OLD20), is based on Levenshtein distance (LD), which is the minimum number of operations (insertion, deletion, and substitution) required to change one string into another (e.g., WIDOW can be changed to WINDOW by a single insertion, hence LD of 1). OLD20 is the mean LD from a word to its 20 closest orthographic neighbours based on LD (its sign is therefore opposite to N – sparse neighbourhood is indicated by low N but high OLD20). It is beyond the scope of the present paper to discuss neighbourhood density effects in terms of OLD20, but we note that OLD20 and Coltheart's N are highly correlated (for monosyllabic words in the ELP Database, r=–.925).

²It should be noted that priming is limited to the Same responses in this task. A formal analysis and simulation of this pattern based on Norris’ (2006) Bayesian Reader model of word recognition is given in Norris and Kinoshita (in press) and the interested readers are referred to that paper. Very briefly, it is because the prime contributes evidence for the target for the decision required by the task, and for Different decisions, it is irrelevant whether the prime is identical or different from the target; what is relevant is whether the prime is similar to the reference (which is the same as the target for the trials requiring the Same response).

³This raises a question about whether reading level interacted with identity priming (for high-N words). Castles et al. (1999, 2007) did not report this specifically for high-N words (although Castles et al., 1999, did report that collapsed across neighbourhood density, identity priming was no greater in the developing readers than the adult readers). Thus the viability of developmental lexical tuning hypothesis is also unclear at this stage.

⁴In this context, one possibility is that lexical decisions may be based on ‘global lexical activation’, rather than the activity of the individual lexical entry, as suggested by MROM (Grainger & Jacobs, 1996). Factors such as the nature of nonword foils are known to guide the relative reliance on global vs. individual lexical activation. Future research may be usefully directed towards investigating the role of these factors. In particular, it is possible that the modulatory effects of N on the difference between the identity prime and a form-related prime and the relative degree of facilitation may go hand in hand.