Conducted 8 experiments that attempted to determine whether information obtained early in the process of recognizing a stimulus can be used to begin preparing keypress responses before recognition of the stimulus has finished. This question is relevant to the recent debate between discrete and continuous models of human information processing. Stimulus sets were chosen so that recognition processes could extract incomplete preliminary information about a stimulus faster than they could extract secondary information needed for stimulus identification. Discriminability of secondary information was manipulated to vary the opportunity for response preparation (RP) based on preliminary information, with difficult secondary discriminations providing more time for RP than easy ones. Precues were given on some trials to allow RP to occur before the stimulus was presented, thereby reducing any difference in RP as a function of discriminability. Continuous models predict that precues should facilitate RP less when the secondary discrimination is difficult than when it is easy, and discrete models predict equal facilitation regardless of secondary discrimination difficulty. Evidence of RP was obtained with some but not all stimulus sets. Results are interpreted as support for the asynchronous discrete coding model.