The count of key words indicating the quality of reasoning has been used as a method to observe exploratory talk. However, reasoning in talk does not necessarily contain key words. The analysis of key words leaves unattended other aspects of exploratory talk, such as collaborative processes. The question is: to what extent can the analysis of key words in context be used to observe the presence of exploratory talk? For answering this question we compared the key words analysis with another way of assessing the quality of talk, based on the analysis of elaboration. We audio recorded 26 groups of three or four pupils accomplishing a discussion task. We divided the transcripts in episodes and established for each episode whether pupils elaborated or not, and if they elaborated whether it was individually or collaboratively. Five different types of episodes could be distinguished. We also developed a method to count the relative number of key words. Statistical analyses illustrated that the relative number of key words differed with respect to the type of episode and elaboration. Qualitative analyses demonstrated that the use of particular key words depended on the task and context. From these results we conclude that the analysis of key words is only an appropriate method for observing exploratory talk, if reasoning context, task and group processes are simultaneously taken into account.