Listeners made location response to tones presented in two-clause sentences as well as to sentences in which no tone was present. Listners were always required to be ready to write out the sentence from memory after hearing it. However the critical responses were made on a prepared script of each sentence presented after each trial. Location guessing patterns do not account for either the relatively high correct location of tones objectively between clauses nor for the frequent mislocation of errors into that position. The analysis of correct location responses, location error patterns in relation to location guesses support a clause by clause model of speech processing. (20pp.)