Although Norris, McQueen and Cutler have provided convincing evidence that there is no need for contributions from the lexicon to phonetic processing, their simplification of the communication between levels comes at a cost to the processes themselves. While their arrangement may ultimately prove correct, its validity is not due to a successful application of Occam's Razor. The kind of statistical accumulation that they propose also presupposes that there are units of a particular size that are the sites for those accumulations. The units could be segments, but the perceptual effects of units of other sizes would perhaps lead to the expectation that frequencies could accumulate at any of those levels. The evidence from the lexical access literature is lacking on this point, since the evidence for accumulation at the segmental level is quite recent. But we would expect listeners to be sensitive to the statistical properties of any unit that they used perceptually. These considerations can also be contrasted with the evidence that listeners parse the signal into the various phonetic units in a fairly complete way, without strict left-to-right processing. If the results of that parsing are the units at which the statistical properties accumulate, then the parsing must be complete before the statistical likelihoods can have an effect, or the statistical likelihoods are part of the parsing itself. The first possibility seems to require yet another level between speech perception and lexical access; the second possibility requires a far more complex version of speech perception than we are typically used to. At this point, it is only possible to make preliminary suggestions for a resolution.