This study investigated the ways in which two groups of four adult learners of English as a second language (ESL) responded to unfamiliar words they encountered in four communication tasks and the effect that different levels of engagement with these words (including negotiation of form and meaning) had on subsequent recall of word meaning. Of the four tasks, two were information gap tasks and two were opinion gap tasks. The results showed a strong task type effect on both the amount and type of negotiation, with more negotiation of the form of words (including spelling and pronunciation) in the information gap tasks and, conversely, more negotiation of meaning in the opinion gap tasks. Through the negotiation process, the learners in the study provided accurate information to each other on word meaning. However, only a small proportion of the total number of unfamiliar words in the tasks were actually negotiated for meaning. The prediction that negotiated words would be more likely to be learnt was confirmed, although the learners also showed improved recall of many words that had not been negotiated. Averaged across learners and tasks, the post-test gains approximated to four instances of word learning per 30 minutes of task work, a gain measured three days after the words were met in the tasks. Seen as a proportion of unfamiliar words in the pre-test these are gains of around one in every three unfamiliar words met in the tasks. The finding that much of the improved recall of word meaning was for words that had not been negotiated indicates that the role of negotiation in learning through communication tasks needs to be viewed from a wider perspective. (Contains 12 tables.)