Two experiments using a probe-RT paradigrn showed that partial information about the size of a stimulus can influence response processes before complete size information is available. Contrary to the asynchronous discrete coding model, these results suggest that the perceptual system may transmit to the response system information that only partially specifies a single stimulus attribute. In combination with previous findings, these results also suggest that there are at least two dissociable forms of selective preparation for a given response.One of the most common assumptions about human information processing is that the flow of information through the nervous system can be decomposed into distinct stages (e.g., Donders, 1868Donders, /1969Sternberg, 1969). For example, when the presentation of an object requires the performance ofan action, the hypothetical stages may include stimulus detection, stimulus identification, selection ofa response, response preparation, and response execution. Sternberg (1969) suggested that such stages occur in strict succession (i.e., in discrete stages), so that each stage can begin only after the previous stage has finished. Because the stages do not overlap, this conception predicts that the reaction time (RT) to a stimulus can simply be interpreted as the sum of the durations of the individual stages.In contrast, other theorists have proposed continuous models in which stages' may overlap in time (e.g., Eriksen & Schultz, 1979;McClelland, 1979). For example, suppose that the perceptual stage carries out a number ofseparate analyses on a given stimulus, and that as soon as the first analysis has been completed, its result istransmitted to the next stage. This next stage could then start processing based on the initial output from the perceptual stage, even though it would continue to receive more perceptual information as further discriminations were made. Ifthese models are correct, response-level processing could begin before perceptual analysis completed, and RT could not be understood as simply the sum ofthe times for each of the processes required by the task.As has been noted previously (e.g., Miller & Hackley, 1992), neither discrete nor continuous models need be applicable under all circumstances, so the goal in studying these models is not so much to decide which one is true as to understand the conditions under which each This research was supported by Otago Research Grant MFU-B75. The authors thank Michael Coles, Patricia Haden, Arthur Kramer, Robert Proctor, Rolf Ulrich, and an anonymous reviewer for helpful comments on earlier drafts ofthe article. Correspondence should be addressed to J. Miller, Department ofPsychology, University ofOtago, Dunedin, New Zealand (e-mail: miller@otago.ac.nz).model best describes the way we process information. For example, Meyer, Yantis, Osman, and Smith (1985) found evidence that processing tends to be discrete in tasks with few responses and compatible stirnulus-response (S-R) mappings but tends to be continuous when there are ma...