The analysis of several sets of data from five-man laboratory groups tends to support the hypothesis that both centrality of seating position and distance between members can be used to predict the interaction pattern. This pattern only appears in a "task" session. In a "social" session for the same type of group, members tend to talk more to the person next to them as they turn away from the group for a more intimate conversation. Personality varibles are also related to seating choice and to interaction rate. More dominant subjects tend to choose the central seats and to do the most talking.