Evidence.

Cultural consensus

Does a group share one understanding?

• Bennardo, de Munck and Chrisomalis, editors (2024). Cognition In and Out of the Mind: Advances in Cultural Model Theory. Palgrave Macmillan.A 2024 snapshot of how important scholars think about this work today.
• Borgatti (2002). A statistical method for comparing aggregate data across a priori groups. Field Methods.The test behind comparing groups, such as the people applying and the teams serving them.
• Maltseva (2024). Measuring shared collective knowledge and belief systems. In the Bennardo volume above.Practical strategies for measuring what a group shares, including with smaller samples.
• Romney, Weller and Batchelder (1986). Culture as consensus. American Anthropologist.The founding paper: the math that tells whether a group shares one understanding.
• Weller, Johnson and Dressler (2023). Validating cultural models with cultural consensus theory. Qualitative Psychology.How to verify a model honestly: a fresh group answers and some questions run in reverse.
• Weller (2007). Cultural consensus theory: applications and frequently asked questions. Field Methods.The working guide: how many people to ask, what data works and what to check.

Freelisting

One good question and what lists reveal.

• Bernard, Wutich and Ryan (2016). Analyzing Qualitative Data: Systematic Approaches, second edition. SAGE.Box 18.3, page 411, in the authors' own words: "Free listing, however, produces lists of varying length. It's one thing to name elephants fifth in a list of 30 animals and quite another to name elephants fifth in a list of 10 animals." The same box quotes Borgatti (1999, page 149) for Smith's S being "highly correlated with simple frequency"; that sentence is Borgatti's, not theirs.
• Brewer (2002). Supplementary interviewing techniques to maximize output in free listing tasks. Field Methods.Brewer tests three techniques: prompting without naming anything specific, reading the list back and using the items a person already named to cue more. We run all three. He also tested cueing letter by letter through the alphabet. We run that one after the others, which is where he suggests it might add a little. He measured it as weaker than cueing from a person's own items. Brewer traces the "people remember more on a second pass" finding to Brown in 1923, long before anyone called this freelisting.
• Cheney and colleagues (2018). Veteran-centered barriers to VA mental healthcare services use. BMC Health Services Research.A full worked example of this pathway in government healthcare: a freelist question about barriers, Brewer's probes, then grouping, in veterans' own words.
• Keddem and colleagues (2021). Practical guidance for studies using freelisting interviews. Preventing Chronic Disease (CDC).A how-to for freelisting in public health practice, published in the CDC's own journal.
• Maltseva (2016). Using correspondence analysis of scales in mixed methods design. Journal of Mixed Methods Research.Mixing interviews, freelists and statistics in one design.
• Quinlan (2005). Considerations for collecting freelists in the field. Field Methods, 17(3), 219 to 234.An accessible field guide to doing freelists well, tracing the method back to its 1960s roots. Page 226: "Determining which items are salient is not standardized. Drawing this boundary is a matter of judgment." Page 231: "Omission and clustering of terms may reduce precision of salience estimates."
• Ryan, Nolan and Yoder (2000). Successive free listing. Field Methods.Using several lists in one interview to build an early explanatory model.
• Smith and Borgatti (1997). Salience counts and so does accuracy. Journal of Linguistic Anthropology.Their own correction and update of the salience measure (the words are the paper's own subtitle). Cultural Consensus Theory for All is being built to report this version.
• Smith, Furbee, Maynard, Quick and Ross (1995). Salience counts. Journal of Linguistic Anthropology.Why the order of a person's list carries meaning.
• Weller and Romney (1988). Systematic Data Collection. Sage, Qualitative Research Methods, volume 10.Page 11 names two measures of saliency: "the position of an item on a list" and "the proportion of the lists on which the item appears". Page 15: "Frequencies or percentages may be used as estimates of how salient or important each item is to the sample of informants." Page 16, written in 1988, five years before Smith's index and long before the range around it: "Finally, there are no generally recognized ways to check the statistical reliability of the free listing task." Page 14: "Usually with a coherent domain, 20 to 30 informants are sufficient. ... If one keeps track of the frequencies in a sequential way it is possible to tell when stability in order is reached and use this as a guide for how many informants are necessary."
• Weller (2014). Structured interviewing and questionnaire construction. In Bernard and Gravlee, editors, Handbook of Methods in Cultural Anthropology, second edition. Rowman and Littlefield.Page 351: "As the number of interviewed informants increases, say in increments of five; from 5 to 10, 10 to 15, and so forth, there will reach a point where little new information is added to the content and order of tabulated items. This is sometimes referred to as the point of saturation. Thus, the sample size is adequate when the addition of new people or groups does not alter the frequency distribution of items and few new items are added."
• Weller and colleagues (2018). Open-ended interview questions and saturation. PLOS ONE, 13(6), e0198606.Page 15: "Empirically observed stabilization of item salience may indicate an adequate sample size." Page 11: the Smith index and the plain sample proportions correlate at 0.89 across 28 examples. Probing matters more than the number of interviews: ten interviews with exhaustive listing captured 95% of the widely shared ideas, while ten casual ones captured about half. Tracking item salience until it settles tells you when your sample is enough.
• Wutich, Beresford and Bernard (2024). Sample sizes for 10 types of qualitative data analysis. International Journal of Qualitative Methods.Empirical sample-size guidance across qualitative methods, from the field's leading methodologists. Their floor for free lists is ten, on one condition: the probing has to be thorough. They warn that light probing can seem to reach the end of the list and stop a team too early.
• Major-Smith and Purzycki (2026). Modeling uncertainty around free-list cultural salience scores. Field Methods, 38(1), 62 to 75.Page 2 of the authors' preprint (DOI 10.31219/osf.io/k5ef4_v1): "A Smith's S score of 0.3 from one sample could change to, say, 0.21, 0.17, 0.32, or 0.38 with other samples from the same population. Thus, these point estimates give us a false sense of precision." Page 10: "Researchers using free-list data in any way beyond describing the sample and data should consider calculating uncertainty, particularly for cultural salience estimates."
• Chaves, Nascimento and Albuquerque (2019). What matters in free listing? A probabilistic interpretation of the salience index. Acta Botanica Brasilica, 33(2), 360 to 369.Page 361: "The interpretation of these values, however, is quite subjective." Salience blends how often and how early an item is named, which can pull in opposite directions, so it is not a clean importance ranking.
• Meireles, de Albuquerque and de Medeiros (2021). What interferes with conducting free lists? Journal of Ethnobiology and Ethnomedicine, 17, article 4.A comparison found repeat-lists matched poorly, at a mean similarity of 0.26, pointing to memory load rather than the room as the main driver.

Tools ours builds upon

Scholars built these to do this work. We honor them.

We apply the ideas from their writings, including the math their tools validated. Scholars designed their software for scholars. We are practitioners, building on their work. We want a much broader audience to apply their ideas to understand and respond to complex public and organizational problems. If you are one of these scholars or their student: we would welcome your feedback and collaboration so more people can apply your ideas.

• AnthroTools, an R package, by Benjamin Grant Purzycki and Alastair Jamieson-Lane.Freelist salience and consensus analysis for researchers who work in R.
• ANTHROPAC, by Stephen Borgatti.The classic freelisting and consensus program, built for DOS and no longer updated.
• CCTpack, an R package, by Royce Anders.Bayesian cultural consensus models; removed from the main R archive in 2025 and no longer maintained.
• UCINET, by Stephen Borgatti, Martin Everett and Linton Freeman.The long-running social network analysis package from the same research tradition.