Our study investigates prosodic phrasing in a corpus of French conversational speech. We looked at statistical and temporal properties of prosodic constituents, which were previously identified within laboratory phonology paradigm. Prosodic annotation of our corpus implements two-level hierarchical model distinguishing major prosodic units (Intonational Phrases, IPs) and minor prosodic units (Accentual Phrases, AP). Both temporal data and distribution of the number of APs in an IP evidence the global tendency to produce shorter units in conversation. Moreover, Intonational phrases containing no more than two Accentual phrases cover 80% of the data. We discuss the implication of these results for both phonological studies of the constraints on prosodic phrasing and oral document tagging.