The methodology used in COSER has been that of the semi-directed interview developed by sociolinguistics, always channelled by the interviewers towards certain topics of traditional life in the countryside. The fact that the interview focuses on these thematic modules does not prevent that, after some time and having gained the informant's trust, the interest is shifted to others, such as education, desires, personal experiences, life itself or the family, depending on the degree of comfort and spontaneity shown by the informant. The decision to focus the interview on thematic modules related to rural life "in the past" has much to do with the fact that, in order to accept the interview, the potential informant has to recognise that he or she is in possession of certain knowledge about a system of life in decay, knowledge that is the product of his personal experience and age and that confers him informative "authority" on the urban interviewer. The informant accepts the interview because he understands our interest in the testimony of a system of life in decline that few people remember and of which he knows himself to be an expert. We believe that the informant's spontaneous collaboration would be much more difficult to achieve if, from the start, he or she were asked to be interviewed about personal opinions or experiences, or about issues unrelated to rural life. In fact, frequently informants have mentioned the university status of the interviewers to evade the interview, claiming "you know everything better than we do!" The survey team's insistence on the precise interest of the strictly local tradition, in contrast to that of other rural localities, and on the informant's unique status as a repository of that tradition, has often been a determining factor in the informant's acceptance of the interview.

Contact with informants is always made at random, without prior arrangements, among local individuals who meet the aforementioned requirements. The unsatisfactory experience of some interviews due to the informant's poor informative capacity (people who speak little, answer in very short sentences or monosyllables) led to the subsequent addition of the condition of loquacity ("who likes to talk") to the informant selection protocol. As should be obvious to anyone who has carried out fieldwork, success is never guaranteed, and an interview can be good or bad with the same starting conditions. Thus, not all interviews are equally suitable or informative, depending on the disposition of the informant, the skill of the interviewers and the interaction between the two, but we do not believe that any testimony should be disregarded.

Regarding the number of informants in each locality, COSER has generally preferred to interview only one person in depth, whether male or female. However, the conditions of the recording sometimes make it impossible to avoid interruption by other individuals (usually family members or acquaintances who, attracted by the extraordinary circumstance of the interview, are tempted to intervene and give their testimony). Thus, although up to 2,910 informants are registered in COSER, most of the time only one informant per locality has been interviewed in the desired depth.

Initially, the interview protocol was designed with the aim of documenting certain linguistic phenomena (specifically, the use of unstressed pronouns), but it soon became clear that information was always obtained on many other aspects apart from the expected ones. From that moment on, the development of the conversation has sought to create contexts which would favour the appearance of dialectal data of all kinds, some of which have been mentioned in the specialised bibliography and others which had scarcely attracted attention. In this regard, it should be noted that the COSER interviews have been particularly useful for documenting dialectal phenomena relating to grammar, an aspect traditionally under-represented in dialect monographs and in the questionnaires of linguistic atlases.

The selection of the localities has followed various criteria over time: initially, in the years 1990-1993, the aim was to delimit the isogloss of the referential system of unstressed third person pronouns ("Pronombres 90", "Pronombres 91 Primavera", "Pronombres 91 Verano", "Pronombres 92 Primavera", "Pronombres Julio 92", "Pronombres Marzo 93", "Pronombres Julio 93" survey campaigns). Later, in 1994 and 1995, the coverage of the inland area framed by this isogloss was addressed ("Castellano 94", "Castellano 95" y "Castellano 95 Primavera"). Since 1996, survey campaigns have continued in other areas, generally focusing on a provincial, island or regional demarcation ("Navarra 96", "La Rioja 97", "La Mancha-Castellano 98", "Castellano 99", "País Vasco 2000", "Teruel 2001", "Jaén 2002", "Cuenca 2003", "León 004", "Asturias 2005", "Extremadura 2006", "Huesca 2007", "Zaragoza 2008", "Albacete 2009", "Badajoz 2010", "Córdoba 2011", "Huelva 2012", "Andalucía 2012" [Cádiz, Sevilla and part of Málaga], "Andalucía 2013" [Málaga, Granada and Almería, "Oriente 2013" (Castellón, Valencia, Alicante and Murcia], "Lérida 2014", "Asturias 2014", "País Vasco 2015", "Salamanca 2015", "Tenerife 2016", "Oriente 2016", "Gran Canaria 2017", "Mallorca 2017", "Galicia 2017", "Fuerteventura-Lanzarote 2018", "Cataluña 2018", "La Gomera 2018", "La Palma 2019", "Zamora-Salamanca 2019", "El Hierro 2019", "Gerona 2021", and "Galicia 2022" campaigns).

The average duration of the recordings is one hour and four minutes per locality but can range from only half an hour to more than two and a half hours. The quality of the recordings is not directly proportional to the duration, as there are excellent and very informative recordings of only half an hour, the results of which are comparable to those obtained in a longer session.

Since 2010, some recordings have been made in both audio and video, a practice that has become widespread since "Andalucía 2012".