Methodologies of conversational, discourse and corpus analysis are used together to verify the hypothesis are there differences in mutual conversational interaction of participant in television interviews considering their gender and discourse roles. The issue of this work is a three leveled corpus made of 16 interviews (approximately 73.000 words). The specifics of the corpus is that it includes the same participants of conversation. The number of such interviews is conditioned by availability in media according to the criteria that the participants are the same persons. The corpus core is consistent of ten interviews. Two presenters, male and female, in their interviews host three same male guests, and two same female guests. The analysis of such interviews is complex and two-way related because it examines the differences and similarities in their conversational interaction which are recognized from both positions in asymmetric hierarchic structure of conversational pairs in media discourse. The corpus core attachment with three additional interviews brings in a new female guest in order to make balance between male and female guests. Also, the subject matter are the additional presenters, male and female, from those interviews. The third part of the corpus is the corpus attachment which is also incorporated in additional three interviews where presenters, male and female from the corpus core, host other guests. Short examples from the 17th interview is presented in the analysis as well. The computer program Elan – Linguistic Annotator is used for segmentation and transcription of videos and to conduct the analysis on three levels: organization of speech turns (the issue of the analysis are overlaps, simultaneous speech and interruptions), directness and politeness (type of questions and forms of address are being analyzed in line with the face threatening act (FTA)), time of participating in conversation and indicator of phoric features of TCU (which comprehends time of participation in conversation, speed and amount of talk, interspace between TCUs, and phoric features of TCUs). Comparing to similar works that are analyzing the category of gender from the essential aspect in order to present statistic score of gender marked language resources, this work is more focused on analysis of speech in interaction in local conditioned context. The results of this work can be understood as a relevant start point for more complex researches based on much larger corpus.