The Austronesian Game Taxonomy: A cross-cultural dataset of historical games

This study addresses longstanding limitations in cross-cultural research on rule-based games. Prior work has often defined games as inherently competitive, which excludes cooperative and solitary forms and narrows understanding of cross-cultural variability. The authors assemble and share a large, codeable dataset of historical game descriptions from Austronesian-speaking cultural groups to enable transparent, replicable research. They also introduce a typology of game goal structures to capture variation in cooperativeness and competitiveness, aiming to support investigations into how games relate to broader cultural contexts, such as social complexity, stratification, cooperation, and child development.

The paper situates its contribution within several strands of literature: (1) classic cross-cultural studies on games (e.g., Roberts et al., 1959; Roberts and Sutton-Smith, 1962, 1966; Silver, 1978; Chick, 1998, 2015) that often adopt a competitive definition, emphasizing winners and losers; (2) developmental research showing children’s early understanding of and enforcement of game rules (e.g., Rakoczy, 2007; Rakoczy et al., 2008, 2009; Hardecker et al., 2017); (3) findings linking game types to cultural variables such as political integration and social classes (Roberts et al., 1959) and child-rearing practices (Roberts and Sutton-Smith, 1962); and (4) rare studies examining cooperativeness in games (e.g., Eifermann, 1970) but with very limited cultural samples. The authors argue that the field lacks accessible raw data and comprehensive inclusion of non-competitive games, motivating the creation of a transparent, codeable dataset and a goal-structure typology that captures cooperative, competitive, and solitary play.

Definition of a game: Adopting and clarifying criteria from Whittaker (2012), a game is an activity with (1) explicit rules accepted by players, (2) undetermined outcomes or actions, (3) a contest or challenge, and (4) non-utilitarian value. Explicit rules are constitutive and specify allowed/prohibited actions; outcomes/actions are not predetermined; contests may involve competition between teams or challenges against chance, time, or one’s abilities; and non-utilitarian value indicates activities chosen freely rather than imposed.

Goal-structure typology and coding: Building on interdependence theory (Deutsch, 1949; Johnson and Johnson, 1974, 2011), the authors code the cooperativeness of games’ goal structures. Types include: Solitary (no interdependence), Competitive (mutually exclusive individual goals), Competitive vs. Solitary (one competitor vs multiple non-interdependent individuals), Competitive vs. Cooperative group (one competitor vs a cooperative group), Cooperative group vs. Cooperative group (teams compete; within-team positive interdependence), and Cooperative group (all players share a goal; no competition). Examples from the dataset illustrate each type. The number of players often does not alter goal structure; referees and non-players are not included in coding.

Data sources and search strategy: Four primary sources were systematically searched for Austronesian games: eHRAF (Oceania excluding Australia; subjects: Games 524, Athletic Sports 526, Childhood Activities 857), Pulotu resources, American Anthropologist, and The Journal of the Polynesian Society. An additional 12 sources were obtained opportunistically. Totals: 1,738 sources searched, 219 yielded game information. In eHRAF, 2,408 paragraphs from 196 sources were reviewed (final search Aug 2017). Pulotu listed 743 resources; books were screened via keywords and table-of-contents cues; PDFs were searched for terms (e.g., game, play, child, amusement, fun, sport; German equivalents). American Anthropologist search (Oct 2017): 413 sources with terms for game and regional keywords. Journal of the Polynesian Society search (Jan 2018): 374 sources via the term “game.” Only English and German sources were included. Geographic and linguistic metadata were gathered and matched to Austronesian Basic Vocabulary Database (ABVD) codes when possible.

Cultural group identifiers and cross-referencing: Cultural groups were treated as ethnolinguistic groups (Pulotu). Language codes from ABVD, Glottolog, and ISO 639-3 were assigned using geographic information in sources; multiple codes were separated by semicolons. ABVD codes were cross-referenced with Pulotu, ABVD, Glottolog, eHRAF, and D-Place. ABVD codes were matched to languages on the Austronesian language phylogeny (Gray et al., 2009) when possible.

Record linkage: Multiple descriptions within a cultural group were linked into a single game when evidence (local name, common name, rules, details of play, location, group identifiers) indicated they described the same game. If uncertain, descriptions were not linked. Games may appear across multiple ethnolinguistic groups but only once per group unless rules differ substantially.

Filtering and variables: The dataset includes four CSVs (Sources, Descriptions, Games, Cultures) and a JSON metadata file supporting CLDF. Variables include identifiers, names, geographic location, whether the description qualifies as a game, ABVD codes, goal structure and uncertainty/comments, introduced origin coding and uncertainty, Pulotu time-frame matches (±0 or ±50 years), and culture cross-references. Optional filters (implemented in an R package) include: (1) qualifies as a game, (2) assignable ABVD code, (3) goal structure code available, (4) local/non-local origin inclusion, (5) ABVD corresponding to Pulotu group, (6) time-frame matched to Pulotu (±0 or ±50 years), (7) ABVD on the language phylogeny.

Introduced origin coding: Descriptions were scanned for origin-related keywords (e.g., origin, introduced by, tradition, missionary, foreign, American, Japanese, etc.). If at least one keyword occurred, coders determined origin as non-local, local, or undetermined; descriptions without keywords were coded NA for origin.

Reliability: Goal-structure coding reliability (Cohen’s kappa) improved across rounds, reaching 0.94 in round 3 (25% of 702 games coded by a reliability coder vs the first author). Introduced origin coding showed moderate reliability (κ = 0.487), but binary recoding to “exclude non-local” vs “keep local/undetermined” yielded good reliability (κ = 0.808). Uncertainties and disagreements are recorded in comments/uncertainty fields.

Time-frame matching: Field dates or focus dates were extracted from sources; if absent, publication date was used. Pulotu’s traditional time foci can be used to filter games to match cultural variable time frames exactly or within ±50 years.

Data access: Processed data (CSV), metadata (JSON for CLDF), and an R package for loading/filtering are on GitHub and Zenodo. Raw game descriptions are available upon request due to copyright.

• Dataset scope: 907 distinct games documented across Austronesian ethnolinguistic groups; 219 of 1,738 searched sources yielded game information.
• Filtering outcomes (Table 5): 952 linked game descriptions before filtering; 907 coded as games; 764 games with assignable ABVD codes (79 cultural groups); 521 games with goal structure coding (68 groups); 694 games with ABVD codes present on the phylogeny (63 groups). Example combined filters yield: Filters 1,2,3,4a,5a,6 -> 53 games across 10 groups; Filters 1,2,3,4a,5b,6 -> 172 games across 27 groups.
• Phylogenetically pruned sample: Applying selected filters (1,2,3,6) produced 452 games across 55 ethnolinguistic groups, used for visualizing distributions on the Austronesian language phylogeny.
• Distribution of goal structures (in the filtered n = 452): Competitive (n = 228) and Cooperative group vs. Cooperative group (n = 121) were the most common types; distributions varied markedly across cultural groups (as shown in Fig. 2).
• Reliability: Goal-structure coding achieved very good inter-rater reliability in the final round (κ = 0.94). Introduced origin coding showed good reliability when treated as a binary exclude-nonlocal vs keep-local/undetermined (κ = 0.808).
• Illustrative cross-referencing: Multiple independent descriptions of the same game (e.g., Yap’s văt) were linked and coded; ABVD coding and cross-database identifiers enable matching to Pulotu, Glottolog, D-Place, and phylogenetic trees for comparative analyses.

By broadening the definition of games beyond competitive formats and introducing a structured typology of goal interdependence, this work reveals substantial cross-cultural variation in the cooperativeness of games among Austronesian groups. The descriptive results show that both competitive and team-based competitive formats are prevalent, but the relative frequencies vary across groups, suggesting cultural patterning rather than uniformity. The curated, codeable dataset and alignment with linguistic phylogenies enable researchers to address questions about how game structures relate to cultural features (e.g., social stratification, intergroup conflict, child socialization) while controlling for shared ancestry. The open resources promote transparency and replicability in cross-cultural game research, fostering broader tests of hypotheses about the cultural evolution and functions of play.

This paper contributes an openly available, codeable dataset of 907 Austronesian games with a novel, implementable typology of goal structures that captures cooperative, competitive, and solitary forms. It standardizes identifiers for cross-referencing with major cultural and linguistic databases and provides tools for filtering by origin, time frame, and phylogenetic coverage. The resource facilitates phylogenetically informed analyses of the distribution and evolution of game types and their cultural correlates. Future research can build on this dataset by coding additional game dimensions (e.g., skill vs. chance, age and sex of players, objects used), integrating data with cultural variables (e.g., social stratification, interaction patterns), examining cultural transmission and introduced games, and conducting ancestral state reconstructions and co-evolutionary analyses across the Austronesian phylogeny.

• Source limitations: Only English and German sources were systematically included; raw descriptions are subject to copyright and are available upon request, which may limit immediate reinspection by all users.
• Coverage and completeness: The dataset is not an exhaustive inventory of all games in the included ethnolinguistic groups; many groups and time periods remain under-documented.
• Coding constraints: Some game descriptions lacked sufficient detail to code goal structures or origins; vocabulary varied across periods and languages, increasing ambiguity. Inter-rater reliability for introduced origin coding was moderate in three-way coding, with better agreement only after binary recoding.
• Phylogenetic and identifier matching: Not all ABVD codes map to languages on the Austronesian phylogeny used; thus, phylogenetically filtered samples are smaller. Some cultural groups could not be matched to ABVD codes, limiting inclusion in comparative analyses.
• Temporal alignment: Matching games to Pulotu’s traditional time foci may exclude some relevant games; conversely, without matching, temporal mismatches may confound comparisons with cultural variables.
• Scope: The study is focused on Austronesian-speaking groups and may not generalize to other language families or regions without further data collection.