This paper discusses the issue of whether and how data from dialect surveys provide insights into women's language. The Linguistic Atlas projects of the United States and Canada, the Dictionary of American English project, the Arkansas Language Survey and smaller projects are considered; and it is stated that in order to get at conversational interaction, only interviews which are truly conversational in nature should be used. The research reported is based on twenty tapes made in Newton County, Arkansas in l970, and specifically on interviews with two elderly married couples. The working hypotheses, similar to those of Hirschman (1972) and Eubanks (1975), are: (1) males talk more than females in conversational situations, (2) males tend to control conversations by signalling beginnings and ends of conversations, (3) males make more judgmental, analytical statements, (4) males and females signal their perceived roles by the use of standard and non-standard verbal forms, and (5) females make more rewarding and encouraging remarks, or show agreement or indecision. Results at this stage of the research indicate that rules for conversation in interview situations should be determined before differences on male-female conversational interaction cues are defined, since the interviewee's perception of his or her role in the interview as the primary or supplementary informant affects the interview, independent of sex. (CLK)