Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 Jan 12;12(1):327.
doi: 10.1038/s41467-020-20037-y.

Experimental evidence for scale-induced category convergence across populations

Douglas Guilbeault  1   2 , Andrea Baronchelli  3   4 , Damon Centola  5   6   7   8
Affiliations

Experimental evidence for scale-induced category convergence across populations

Douglas Guilbeault et al. Nat Commun. .

Abstract

Individuals vary widely in how they categorize novel and ambiguous phenomena. This individual variation has led influential theories in cognitive and social science to suggest that communication in large social groups introduces path dependence in category formation, which is expected to lead separate populations toward divergent cultural trajectories. Yet, anthropological data indicates that large, independent societies consistently arrive at highly similar category systems across a range of topics. How is it possible for diverse populations, consisting of individuals with significant variation in how they categorize the world, to independently construct similar category systems? Here, we investigate this puzzle experimentally by creating an online "Grouping Game" in which we observe how people in small and large populations collaboratively construct category systems for a continuum of ambiguous stimuli. We find that solitary individuals and small groups produce highly divergent category systems; however, across independent trials with unique participants, large populations consistently converge on highly similar category systems. A formal model of critical mass dynamics in social networks accurately predicts this process of scale-induced category convergence. Our findings show how large communication networks can filter lexical diversity among individuals to produce replicable society-level patterns, yielding unexpected implications for cultural evolution.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1
Fig. 1. Larger populations promote category convergence across populations.
Comparing the level of convergence in category systems that emerged in small (N = 2) (a) and large (N = 50) (b) populations. Each row displays the category system constructed by a single unique population in each condition after 100 rounds of interaction. The horizontal axis displays the image continuum of shapes, consisting of 1500 slices. Density distributions display the frequency of successful coordination for each label, as well as the region of the continuum to which each label referred. Each color indicates a unique label. Similarity in the category systems across independent populations indicates convergence.
Fig. 2
Fig. 2. Convergence in the vocabularies that emerged in populations of different sizes, for N = 2 (black dots), N = 6 (blue diamonds), N = 8 (purple squares), N = 24 (green triangles), and N = 50 (yellow circles).
Vertical axis reports the average similarity in vocabulary (average Jaccard Index) between each network trial and all other networks of the same population size. Horizontal axis displays category diversity, measured as the average number of unique labels encountered by subjects in a population. Data points represent experimental results (80 dyads and 15 social networks of each size). Black trend line shows model predictions (averaged over 50 simulated trials; 100 rounds each trial; dmin = 0.01; |L| = 5000; b = 1); see Supplementary Information section 1.2 for model specification (equation S1). The measure of center indicated by the model trend line is the mean Jaccard Index among simulated trials of the same population size, ordered by the average category diversity in each trial (Fig. S1). Error bands show 95% confidence intervals.
Fig. 3
Fig. 3. Larger populations amplify the spread of initially frequent labels.
a Using the Zipf distribution to model the initial frequency of labels (including data from all conditions; N = 2, N = 6, N = 8, N = 24, and N = 50), where initial frequency refers to the number of individuals who introduced a label without any prior exposure to the label in the task. Vertical axis displays the log of each label’s initial frequency. Horizontal axis displays the log of each label’s frequency rank. b Displaying the mean effect of population size on the ability for labels to reach critical mass (when at least 25% of subjects in a network independently introduce a label). Common labels are identified as outliers with high initial frequency (Supplementary Information section 1.3). Data display the proportion of experimental trials in each condition for which each label type reached critical mass. Error bars display 95% confidence intervals. c The correlation between the initial frequency of a label in a population and the proportion of subjects in a population who adopted the label (vertical axis), where adopting a label entails that a subject produced a label after being exposed to it. Horizontal axis displays the diversity of categories in each trial, indicated as the average number of unique labels encountered by each subject in a network. Error bands display 95% confidence intervals. All observations are independent and at the network-level. All panels represent data from 80 unique dyads and 15 unique social networks of each size.
Fig. 4
Fig. 4. Time series showing the adoption of the confederates’ rare label (“sumo”) by experimental subjects (i.e., nonconfederate subjects).
Pink lines indicate the cumulative number of successful uses among experimental subjects of the label “crab”. Black lines indicate the cumulative number of successful uses among experimental subjects of the label “sumo”. Each round is measured as N/2 pairwise interactions, such that each player has one interaction per round. The data displayed exclude all interactions between confederates.
Fig. 5
Fig. 5. The Grouping Game.
Screenshots of “The Grouping Game” interface from the view of a speaker (a) and the view of a hearer (b) on a given round. c A sample of the continuum of novel shapes used as stimuli in the experiment.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Shepard RN, Cermak GW. Perceptual-cognitive explorations of a toroidal set of free-form stimuli. Cogn. Psychol. 1973;4:351–377. doi: 10.1016/0010-0285(73)90018-2. - DOI
    1. Spalding T, Gregory M. Effects of background knowledge on category construction. J. Exp. Psychol. 1996;22:525–538.
    1. Johnson KE, Mervis CB. Effects of varying levels of expertise on the basic level of categorization. J. Exp. Psychol. 1997;126:248–277. doi: 10.1037/0096-3445.126.3.248. - DOI - PubMed
    1. Levinson, S. C. & Wilkins, D. P. Grammars of Space: Explorations in Cognitive Diversity (Cambridge University Press, Cambridge, 2006).
    1. Ranjan A, Srinivasan N. Dissimilarity in creative categorization. J. Creat. Behav. 2010;44:71–83. doi: 10.1002/j.2162-6057.2010.tb01326.x. - DOI

Publication types