From this article, one doesn't know what the number and variety of factors were that were considered in the research.

For example, were the subjects told to converse about only one subject? Were all participants in the research given the same prescribed subject to discuss? Was the nature of the subjects which were discussed considered for such factors as range of variables - if the participants were supposed to do no more than discuss their activities of the previous day, that would be different if they had been required to discuss the philosophical nature of morality, for example.

A further factor which in my experience frequently determines the length of conversations is that of a game of "one-upmanship" on the part of the person who indicates that they are ending the conversation when they do.