The divergence of actual spoken usage from the prescriptive Croatian accentual norm has been widely noted, but such observations are largely impressionistic. Relatively little acoustic data is available for the realization of lexical prosodic features specifically in Croatian, as opposed to other closely related varieties, and previous studies have focused mainly on measurements of isolated forms produced by "model" speakers, chosen specifically for their ability to reproduce the standard accentuation. The current study analyzes samples of connected speech taken from recordings of the program <i>Govorimo hrvatski</i> on Croatian Radio 1, comparing the results to those in previous acoustic studies of Croatian or Serbian accentuation. The implications of these findings for the viability of the current prescriptive norm are considered within the Croatian sociolinguistic context.