We present results of a perceptual test of the output of a speech synthesiser that uses a concise encoding of prosody. The synthesis paradigm assumes that two parameters of prominence and boundary strength are adequate to determine the acoustic correlates of systematic prosodic variation. The synthesis system utilises a large database of continuous speech as source units for concatenation, selecting on the basis of both segmental and suprasegmental characteristics. We show that the synthesiser is capable of conveying the information that listeners need in order to differentiate phonetically similar, syntactically ambiguous sentences at levels significantly better than chance.