The present study used the backward masking paradigm to investigate the nature and time course of phonological assembly. Two experiments were performed that tried to specify the extent to which phonological assembly is a serial process, or a process that gives special weight to consonants over vowels. Experiment 1 showed that recognition rates in a backward masking task varied as a function of the serial position of the phonemes that were shared between backward masks and target words. When the backward masks and target words shared final phonemes, recognition rates were higher than when they shared initial phonemes. In addition, the data exhibited a weak difference between consonant and vowel preserving masks. In Experiment 2, nonwords were added to the target words of Experiment 1 to discourage guessing from the backward masks. Although serial effects of phonological assembly were again found, the consonant/vowel differences from the previous experiment disappeared. Overall, the results of both experiments showed strong serial effects of assembled phonology compared to much weaker consonant/vowel effects. Furthermore, the emergence of consonant/vowel effects appeared to be modulated by guessing strategies and the consonant/vowel structure of the words. These guessing strategies appear to be an inherent problem with the backward masking task.