This paper describes our study of prosodic differences between read and spontaneous speech taken from human/machine problem-solving dialogues, as captured by F0. Our research had two goals: first, to show that significant intonational differences exist between speech from these two styles and that these differences can be expressed quantitatively, and second, to demonstrate that the encoding of prosodic information is more salient in spontaneous speech than in read speech. Our analysis of over 4000 read/spontaneous utterance pairs from many speakers indicates that the mean of F0 is statistically significantly higher for spontaneous speech than for read, but that F0 is about equally variable in the two styles. In addition, we found that the encoding of final boundary tone information is more easily obtained from spontaneous speech than from read speech. Over 80% of final boundary tones from spontaneous speech were correctly classified, as opposed to less than 70% of those from read speech.