SUMMARY The term subitizing was coined by Kaufman, Lord, Reese, and Volkmann in 1949 to describe the rapid, confident, and accurate report of the numerosity of arrays of elements presented for short durations. They noted that this process, different from counting and estimating, was restricted to arrays with 6 or fewer elements. Ever since the general awareness of some such process in the nineteenth century, the phenomenon has been a benchmark for the limited capacity of human consciousness, Previous research, as well as the data in these experiments, shows that the reaction time function for different arrays (ranging in size from 1 to 15) has a shallow slope for arrays with 1 to 3 elements followed by a straight line slope for arrays of 4 to 6 or 7, at which point the reaction time discontinuity typically occurs; reaction times then stay fairly constant as the array size increases, and the numerosity response becomes much less accurate. When the array is exposed for unlimited presentation time, the reaction time slope is a straight line from size 4 to as large as 30 or 40. In the latter case, subjects are clearly counting the array. Apart from replicating the previous findings under a variety of conditions, we have shown that subjects can make a very fast countability judgment indicating whether or not they could, if requested, give an accurate numerosity response. These judgments are fast and produce a yes response within the subitizing range, and a no response thereafter. Developmental data had indicated that children count arrays as small as 2 and 3; adults seem to give a more automatic response, shown ajso in the fast reaction times to those array sizes. The suggestion that this response is an acquired one to certain frequently appearing canonical patterns of two and three events (pairs/lines and triples/ triangles) was explored in another experiment in which subjects were given canonical patterns for arrays of up to 10. These data showed that within very few trials the response to these canonical patterns was just as fast and accurate as the response to the smaller (1 to 3) array sizes. Finally, the data demonstrated that the slope for array sizes from 4 to 6 with short exposure time was indistinguishable from the slope for array sizes from 4 to 15 under an unlimited exposure condition. We concluded that the reaction time function found in subitizing consists of three processes: a response to arrays of 1 to 3 that is fast and accurate and is based on acquired canonical patterns; a,response to arrays of 4 to 6 or 7 that is based on mental counting, that is, the counting of arrays that can be kept in consciousness (attention); and an estimating response for arrays larger than 6 that cannot be held in consciousness for mental counting.