This dissertation investigates parsing in segmental perception, or the process by which listeners map the continuous acoustic signal that reaches their ears to the linguistic representations over which phonology operates. It addresses questions of when listeners decide that they have heard acoustic evidence about the identity of one speech sound, versus evidence about the identity of a following sound, and when this linguistic knowledge is applied relative to when it is received during the course of on-line perception and processing. The central argument advanced here is that the beginnings of answers to these questions require the recognition of a domain-general perceptual bias to continue attributing incrementally-received input to a previously-recognized event, rather than posit that first event's completion and the beginning of a second event before it is necessary to do so. An outline of a new model of general segmental perception that includes this bias is then advanced. This approach has implications for our understanding of the evolution of the typology of the world's languages, in particular for the ways that the acoustic qualities of cues to phonological contrast can determine which potential processes are or are not phonologized.