Although it is assumed that semantics is a critical component of visual word recognition, there is still much that we do not understand. One recent way of studying semantic processing has been in terms of semantic neighbourhood (SN) density, and this research has shown that semantic neighbours facilitate lexical decisions. However, it is not clear if this facilitation reflects actual word recognition processes or is instead due to participant strategies used during the lexical decision task. To address this, the current research used college students as participants and tested the effect of SN density using the semantic categorisation and progressive demasking tasks. Both of these tasks require word identification and are not susceptible to the participant strategies that are seen when using the lexical decision task. The results show that SN facilitates responding in both tasks, indicating that SN effects are not due to task-specific strategies.Most visual word recognition research has been concerned with understanding how orthography, phonology and semantics interact during processing. Of this research, the influence of orthography and phonology has received the greatest attention, and because of this research, many current computational models contain very detailed and wellspecified orthographic and phonological systems (e.g. Coltheart, Rastle, Perry, Langdon these models, the semantic representation tends to be less well specified or completely absent (although see Harm and Seidenberg [2004] for a model that does include a strong semantic component). A potential reason for this may be that it is not clear exactly how semantics should be defined. For example, in terms of phonology, it is apparent that phonemes are an integral part of the phonological representation. Likewise, letters and graphemes are obviously important to coding the orthographic representation. Unfortunately, what defines the semantic representation of a word is not as clear. Thus, an important avenue for research is to identify what semantic variables are important to word recognition, and thereby gain a better understanding of how to define semantics.One of the first and most studied semantic-level variables is semantic ambiguity (i.e. lexical ambiguity). Previous research has shown that lexical decisions and naming latencies are facilitated for words with multiple meanings, but semantic categorisations are inhibited (Hino, Lupker & Pexman, 2002). In terms of a distributed model, when a task relies primarily on orthography (e.g. lexical decision) or phonology (e.g. naming)