First, the question that Simpson (1994) is examining is whether or not context (i.e., top-down processing) affects basic word recognition. Why would context-dependence be advantageous/disadvantageous? Second, two of the assumptions that seem to come up in much the literature that we're reading are that (a) words have discrete representations (and that we have a good idea of what a word is) and (b) there is a discrete moment at which a word is accessed (i.e., word recognition is an all-or-nothing event). Alternatively, how does thinking about distributed representations and partial activation change the nature of the contextual influence debate?