In the present paper, the experimental procedure, the details of sound and video recording set-up as well as the system for speech and gesture transciption and coding used in the Polish Cartoon Narration Corpus (PCNC) project are described. The audio-visual data come from a cartoon narration task performed by both children and adults. The recordings are transcribed orthographically and phonemically, and labelled for selected phenomena on a number of levels, including gesture, lexicon, prosody, and dialogue acts.