A model of relational competence is elaborated and operationalized in which person A's motivation, knowledge, and skills in conversing with B in a given episode are expected to predict both participants' satisfaction with self, other, and the communication. Naturally occurring dyadic conversations were interrupted and questionnaires were distributed referencing the preceding dialogues (n = 180 dyads). The o verall model explained significant proportions of variance in participants' satisfaction with self (R2= 16), other (R2= .26), and the communication (R2= .27). Results were similar when predicting others' satisfaction with self (R2= .25), other (R2= .29) and the communication (R2= .60). Despite these results, the knowledge component fared poorly, relative to the other predictors. Reasons for this finding, along with other theoretical and methodological implications, are examined.