These pages contain a number of audio clips illustrating the uses of non-lexical utterances in casual English dialog.
The data and its significance is described in more detail in Non-Lexical Conversational Sounds in American English, an article in Pragmatics and Cognition (abstract, draft in pdf, draft in html)
In the diagrams, each strip includes two tracks and a timeline. In each strip the top track is one speaker and the bottom track the other. Each track includes: a transcription, the signal, the pitch in red, and the 26th percentile pitch level (horizontal blue dotted line).
The recording conditions for the various samples were not optimal. In particular, the record levels vary across conversations and across tracks, for some conversations exhalations sound very loud, and for some conversations there is bleeding of sound from one track to the other.
To listen to the data on your local system may require some persistance and luck. On some systems Netscape does a better job with audio, on other systems Explorer does better, and some installations you may have to download the audio files and play them by hand. In particular, if some audio sounds abnormally slow, your browser is probably failing to handle stereo correctly. Also, you will probably also do better with headphones when using a desktop with a proper soundcard, but with external powered speakers when using a laptop.
The Switchboard Transcription FAQ at Mississippi State, with more examples of grunts found in conversation.
Melissa Wright's page of Clicks as Markers of New Sequences in English Conversation.
The ICSI Switchboard Transcription Project, with papers and links about related issues.
The Responsive Systems Project page, which describes why such utterances are of practical importance.
Margaret Magnus' Bibliography of Phonosemantics.