Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Cumulative Cultural Evolution and Transmission Fidelity

Abstract. Cumulative cultural evolution depends on the historical retention, modification, and recombination of socially acquired information. Transmission fidelity is central to this process but varies across actions, products, procedural knowledge, explanatory information, and distributed cultural records. This critical integrative research monograph reconstructs foundational cultural evolutionary theory and compares mathematical models, microsociety experiments, developmental and comparative investigations, demographic and network analyses, archaeological inference, recent animal evidence from chimpanzees, bumblebees, and homing pigeons, and 2026 experiments on selective knowledge communication, causal reasoning, and the cultural retention of inefficient solutions. It separates stable tradition, acquisition of difficult know-how, cumulative functional improvement, and open-ended cultural diversification as distinguishable empirical claims. The synthesis evaluates imitation and emulation, teaching, prosocial learning, skill correction, model availability, network modularity, and information storage while examining causal controls, replication, measurement validity, and historical limits. The resulting research framework treats cumulative culture as a multi-level population process with mechanisms and explanatory bounds, including emerging human–Artificial transmission environments.

Keywords: cumulative cultural evolution; transmission fidelity; cultural ratchet effect; social learning; imitation; emulation; cultural innovation; teaching; cultural evolution; experimental microsocieties; demographic effects; population connectivity; chimpanzee culture; comparative cognition; open-ended cultural evolution; Artificial Era

The Research Object and Its Levels of Explanation

Cumulative cultural evolution describes an intergenerational process in which socially transmitted information is preserved, modified, and recombined so that later learners can inherit functioning practices that embody earlier contributions. Transmission fidelity concerns the probability and precision with which relevant information survives a transmission event. This monograph studies their connection in populations of learners, tools, skills, and communication networks. Its explanatory unit is a lineage of cultural variants rather than the intelligence of an isolated performer. Improvement must be assessed against a task-specific criterion, and continuity must be documented rather than assumed from superficial similarity Lewis & Laland (2012) Caldwell et al. (2020) Morgan & Feldman (2025).

Three separate empirical propositions guide the analysis. First, a behavior can spread socially and become a stable tradition. Second, a succession of learners can retain and improve an inherited variant. Third, accumulated variants can extend the reachable technological or behavioral repertoire beyond what the same individuals would produce by learning alone. These propositions require increasingly demanding designs. A cross-sectional distribution of similar behaviors may establish a plausible tradition; a generational sequence paired with independent baselines supports accumulation; and controlled demonstrations of dependence on preceding improvements support an account of culture as a historical scaffold Caldwell & Millen (2008) Dean et al. (2012) van Leeuwen et al. (2024).

The synthesis integrates formal models, comparative experiments, laboratory microsocieties, field observation, archaeology-informed inference, and recent methodological reviews. It asks which components of fidelity are necessary, which forms of innovation can accumulate, how networks and population turnover change retention, and how animal evidence bears on claims of human uniqueness. The work uses a critical narrative review method with transparent source attribution. No systematic-search protocol, pooled meta-analysis, new participant dataset, or original causal estimate is claimed. Observed effects and theoretical interpretations are identified at their actual level of support Mesoudi et al. (2006) Derex et al. (2024).

What Counts as Cumulative Change

A cumulative sequence exhibits historically dependent differences across transmission events: a later solution benefits from earlier adaptations or combinations. Improved objective performance provides one possible measure, while increasing functional complexity, expanded solution space, and the retention of nonobvious know-how are others. Complexity is not equivalent to the number of visible motions or parts. More elaborate behavior may be wasteful, and simplification can produce genuine functional improvement. Investigators should report the criterion chosen before treating an ascending performance curve as evidence of a ratchet Tennie et al. (2009) Caldwell et al. (2020).

The relevant comparison is between a culturally informed lineage and a justified alternative. Repeated practice by one person can improve performance, and regression toward a better average can make a poor first generation look like progress. A properly designed experiment includes asocial learning, extended individual practice, and where possible a reset or shuffled-history condition. Later generations must inherit usable information from predecessors, and researchers should test whether the quality of available information changes the performance of new learners. These distinctions are central to newer experimental design analyses Caldwell et al. (2020) Derex et al. (2024).

Open-endedness is a further characteristic to be measured independently. A lineage can improve routes or tools for several iterations yet reach a narrow ceiling; another can repeatedly create combinations that establish new tasks, materials, and affordances. Morgan and Feldman argue that cumulative change and high fidelity occur outside humans and that the extraordinary scope of human cultural expansion requires attention to the kinds and range of heritable variation. The position makes the field's comparative categories more precise without settling which social and cognitive mechanisms generate open-ended diversification Morgan & Feldman (2025) Sasaki & Biro (2017).

Foundations of Cultural Evolutionary Theory

Formal cultural evolution emerged from treating socially learned variants as an inheritance system with its own pathways of transmission, innovation, selective copying, and transformation. Boyd and Richerson's 1985 framework described processes such as guided variation and transmission biases and made it possible to investigate cultural distributions without assuming that every shared behavior is directly selected for genetic fitness. Cultural change may also reflect individual invention, copying errors, institutional regulation, and ecological conditions; these mechanisms interact within populations Boyd & Richerson (1985).

The later case for a unified science of cultural evolution connected anthropological history, archaeological sequences, psychological experiments, and mathematical population models. Mesoudi, Whiten, and Laland proposed that transmission, selection, drift, and lineage reconstruction permit explanatory comparisons with evolutionary biology, while cultural inheritance differs in its routes, speed, transformations, and social mediation. This research tradition provides a language for specifying testable predictions about population size, network structure, model choice, and learning opportunities Mesoudi et al. (2006) Henrich (2004).

Psychological theories added mechanisms through which learners acquire action sequences, intentions, goals, and explanations from others. Tomasello, Kruger, and Ratner distinguished imitative, instructed, and collaborative learning, linking transmission to social-cognitive development. Later work challenged any inference that this historical taxonomy identifies an exclusive prerequisite for accumulation on every task. The explanatory project is therefore to determine which information channels achieve sufficient accuracy for particular cultural lineages, and under which ecological and social conditions they do so Tomasello et al. (1993) Caldwell & Millen (2009).

The Ratchet Metaphor and Its Evidentiary Burden

The ratchet metaphor expresses the retention of workable innovations between generations. An improvement becomes the starting condition for future exploration, lowering the need for rediscovery. In a reliable ratchet, loss is sufficiently uncommon and corrective opportunities sufficiently abundant that useful refinements persist, even when individual innovators disappear. This is a probabilistic property of transmission and selection, not a guarantee that all historical sequences improve. Cultural lineages also stagnate, branch, degrade, or disappear Tennie et al. (2009) Lewis & Laland (2012).

One source of confusion is to treat a highly sophisticated final artifact as direct proof of the mechanism that produced it. Complexity can emerge from a single capable inventor, from independent convergence, or from external ecological changes. Stronger evidence reconstructs intermediate states and demonstrates the inheritance of specific innovations, including features whose function would have been difficult to rediscover. Experimental chains can document these transitions; archaeology supplies historical constraints but often cannot observe the exact teachers or social-learning encounters involved Caldwell & Millen (2008) Henrich (2004).

The metaphor also obscures losses and transformations. A transmitted technique may lose a decorative feature yet gain reliability, or preserve a ritual sequence even when functional output remains unchanged. Performance, component fidelity, behavioral resemblance, and functional knowledge are therefore different outcome variables. An analysis that regards high exact-copy fidelity as universally desirable may miss the value of selective copying, experimentation, and recombination. Cultural accumulation depends on a task-specific balance between conservation and productive change Lewis & Laland (2012) Derex et al. (2024).

Methodological Approach and Evidence Classification

Primary laboratory evidence includes controlled tasks in which participants act as successive learners, groups undergo member replacement, or individuals receive social information of experimentally varied quality. Comparative studies add species contrasts and animal transmission opportunities; network approaches observe diffusion under heterogeneous contact. Formal models identify conditions under which accumulation is possible, rather than directly proving that those conditions operated in a particular past population. The monograph evaluates each class according to its design and unit of inference Dean et al. (2012) Muthukrishna et al. (2014) Gunasekaram et al. (2024).

Three evidential questions apply to every experiment: whether the purported cultural trait has been measured reliably, whether observed improvement causally depends on prior social information, and whether plausible alternative mechanisms are excluded. Performance differences should be interpreted with confidence intervals and independent replications where reported. Null findings require consideration of power, engagement, access to relevant demonstrations, prior experience, and ceiling or floor effects. Negative results from one apparatus establish constraints within that apparatus and sampling frame, rather than a species-wide absence of capability Caldwell et al. (2020) van Leeuwen et al. (2024).

Because the present review synthesizes sources chosen for relevance rather than a preregistered exhaustive search, its conclusions are qualitative and bounded. The literature includes mathematical analyses, large and small laboratory samples, ethnographic or archaeological comparisons, and new animal experiments; none can substitute for the others. Research claims are connected to bibliographic source identifiers in this manuscript so the release renderer can produce author–year hyperlinks. Bibliographic metadata are independently checked against accessible publisher or PubMed records for the selected corpus Derex et al. (2024) Morgan & Feldman (2025).

The Difference Between Social Transmission and Cultural Accumulation

Traditions require transmission across individuals or generations, with enough continuity to persist in a population. They may stabilize through conformity, prestige, reward contingencies, habit, or shared environmental demands. A widely distributed behavioral tradition is therefore an important cultural fact without necessarily demonstrating successive improvement. The difference matters acutely in field studies where tool use or foraging techniques have been documented across multiple communities but the sequence of invention and refinement is only partly known Tennie et al. (2009) Gunasekaram et al. (2024).

Evidence of cumulative evolution requires a historical dimension. In a laboratory lineage, record the variant entering each generation, the learner's intervention, the variant offered to the next generation, and functional performance. For animals, pair turnover, diffusion tracing, or experimentally trained demonstrators can show information flow, but different designs establish different aspects of the cumulative claim. A skill adopted after demonstration is evidence of social acquisition; a series of retained improvements needs separate evidence Sasaki & Biro (2017) Bridges et al. (2024) van Leeuwen et al. (2024).

Some practices are preserved with extraordinary accuracy yet never improve, while others achieve functional gains with imperfect copying. This is not paradoxical: selection can retain useful outputs even when action sequences change, and exact repetition can preserve maladaptive procedures. Research on transmission fidelity must therefore ask what is being copied, which features matter for the cultural function, and how the inherited representation supports further improvement. These distinctions prevent conflating long persistence, success of diffusion, and cumulative innovation Caldwell & Millen (2009) Lewis & Laland (2012).

Transmission Fidelity as a Multidimensional Construct

Transmission fidelity is best operationalized as retention of specified information between a model and a learner or between successive cultural states. An experiment may measure action fidelity, product fidelity, sequence fidelity, accuracy of verbal explanation, retention of causal relations, or the persistence of performance-relevant features. Each measure encodes a different theory of what must survive. A learner who recreates a successful object by different actions may transmit the functional design accurately while showing low action imitation. A learner who repeats every movement but misunderstands the mechanism can show high behavioral resemblance with poor functional competence Caldwell & Millen (2009) Lewis & Laland (2012).

The distinction between discrete and continuous traits also affects claims about fidelity. For a binary technique, the estimated probability of successful reproduction may be clear; for a tool with numerous adjustable components, error involves a distribution of deviations, dependencies among components, and tolerances. The relationship between fidelity and improvement is nonlinear when essential steps fail below a threshold, when error is corrected by feedback, or when social learning combines demonstrations. A single universal fidelity coefficient cannot describe all these conditions Lewis & Laland (2012) Muthukrishna et al. (2014).

Fidelity belongs to a transmission process embedded in an environment. Written instructions, stable artifacts, expert correction, durable social networks, and repeated exposure can preserve information that a single observation would not. Researchers should identify fidelity at the individual event, lineage, and institutional levels. Cultural robustness can arise from distributed redundancy even when each learner is imperfect. This system-level perspective connects psychological copying capacities with demographic connectivity and material memory Derex & Boyd (2016) Morgan & Feldman (2025).

Imitation, Emulation, Instruction, and Reconstruction

Imitation is copying aspects of a model's action organization; emulation is learning about results, affordances, or environmental states and independently reconstructing an effective procedure. Instruction can transmit abstract constraints and causal accounts, while direct participation allows guidance through shared tasks. These routes vary in how much detail they transmit and what modifications they invite. Psychological theories of cultural learning emphasized intention understanding and coordinated attention, but the productive question is whether the route carries the information necessary for the focal artifact or skill Tomasello et al. (1993) Tennie et al. (2009).

Caldwell and Millen tested paper-airplane construction using 700 participants in 70 laboratory microsocieties. Their manipulation separated access to imitation, emulation, and teaching, and the authors reported cumulative performance under each information condition. The result shows that reproducing a demonstrated motor process was not necessary for accumulation on that task. It does not establish that all high-dimensional skills can accumulate under product-only access, or that all three mechanisms are interchangeable for the transmission of hidden causal steps Caldwell & Millen (2009).

Different information channels can also substitute for one another across generations. A difficult manual technique may depend on observation early in training, verbal correction for error detection, and artifacts that preserve successful design parameters between experts. Transmission studies should manipulate channels separately and in combination, measuring what each learner has actually acquired. Such designs refine historical claims about imitation and teaching without assuming one universal psychological route to the ratchet effect Dean et al. (2012) Derex et al. (2024).

Innovation, Modification, and Recombination

Innovation may introduce a previously absent trait, alter an existing trait, or join established elements into a new arrangement. These mechanisms have different consequences for cumulative culture. A novel component provides an opportunity, a modification changes local performance, and recombination can expand the design space by making capabilities interact. Whether a change is beneficial depends on evaluation criteria, ecological fit, and the capacity of later learners to preserve the innovation. Large numbers of inventions can coexist with poor cumulative retention Lewis & Laland (2012) Boyd & Richerson (1985).

Lewis and Laland's formal model compared transmission loss with novel invention, modification, and combination. Under the model's assumptions, relatively small changes in fidelity produced pronounced differences in cultural accumulation, and combination contributed importantly to progressive cultural change. These are properties of a specified mathematical representation; empirical investigations must test whether the modeled trait structure, opportunity for combination, and copying errors resemble the cultural systems under study. The model is valuable because it turns an intuitive emphasis on fidelity into a manipulable hypothesis Lewis & Laland (2012).

Recombination creates special methodological problems. A downstream artifact may draw on several unrelated lineages, making simple ancestor–descendant chains inadequate. Researchers need component provenance, records of access to multiple models, and a way to distinguish genuine integration from convergent individual invention. Networked experiments allow variants to diffuse and cross-fertilize; histories of technological design may demand lineage graphs rather than trees. Claims of increased sophistication should still be tied to independent tests of function rather than visual complexity Derex & Boyd (2016) Morgan & Feldman (2025).

Feedback, Causal Knowledge, and Error Correction

An inherited demonstration can transmit an effective procedure without explaining why it works. Subsequent learners may improve it through performance feedback, targeted experimentation, or causal reasoning. Accurate causal knowledge can enable transfer to a changed task and help diagnose failures, while faithful copying can protect opaque solutions long enough for later understanding to develop. These processes operate at different timescales. The central evidence is whether learners can retain inherited components and improve performance when relevant conditions change Dean et al. (2012) Derex et al. (2024).

Feedback can imitate cultural gains in poorly controlled designs: participants exposed to a predecessor often also receive more time, more salient cues, or a clearer reward structure. Factorial designs should balance practice, exposure, and evaluation opportunities between social and asocial conditions. Some tasks reward the exploration of many alternatives, while others punish initial errors so severely that copying is rational. Transmission fidelity is consequently shaped by the payoff distribution and by how costly mistakes are to correct Caldwell et al. (2020) Derex et al. (2024).

Models with stable physical artifacts illustrate how functional knowledge can persist beyond individual memory. A template or standardized procedure narrows reconstructive error; repeated use can reveal overlooked performance advantages; and external records can allow independent groups to test alternatives before integration. The psychological study of fidelity therefore extends beyond mirroring observed movements to the learner's management of evidence, causal constraints, and feedback. This wider approach gives a plausible route from modest individual copying accuracy to robust collective retention Morgan & Feldman (2025) Derex & Boyd (2016).

Teaching, Cooperation, and Prosocial Transmission

Teaching can change transmission by directing attention, structuring practice, adjusting demonstrations to a novice, and providing explicit corrective information. Cooperation can make information sharing worthwhile even when the expert pays an opportunity cost. In human groups, instructional language and social support frequently accompany complex skill learning, but their contribution must be identified experimentally instead of being read into successful performance. The mechanisms may influence the number of successful learners, the variability of replicas, and the probability that knowledge remains available after expert departure Tomasello et al. (1993) Dean et al. (2012).

Dean and colleagues compared groups of children, chimpanzees, and capuchin monkeys on a three-stage puzzlebox providing increasingly attractive rewards. Children more frequently reached higher-level solutions, and their success was associated with teaching by verbal communication, imitation, and prosociality; these features were not observed in equivalent form in the tested nonhuman groups. The association and task differences make the study a central comparative result, while leaving open which components are causally necessary, how much prior educational experience mattered, and whether another apparatus would yield different species rankings Dean et al. (2012).

Teaching should be analyzed as a family of interventions: demonstration, social reinforcement, verbal explanation, scaffolding, and correction differ in information content and accessibility. The presence of language can reduce ambiguity about action order, but reliable apprenticeship also occurs in silent and material forms. A cumulative-culture model that uses teaching as a single binary attribute misses developmental variation and unequal opportunities to practice. Experiments that manipulate instruction quality and learner access offer stronger evidence for specific pedagogical mechanisms Caldwell & Millen (2009) Caldwell et al. (2020).

Developmental Acquisition and Intergenerational Learning

Children's growing capacities for sustained attention, action planning, explicit explanation, and coordination with peers may change which cultural information they can use. Yet a developmental comparison must consider the familiarity of instructions, materials, reward systems, and social expectations. A child who performs poorly on an abstract apparatus may master a culturally familiar practice through guided participation; performance in a single laboratory task is a joint outcome of developmental ability and task design. Cultural learning unfolds within patterned opportunities rather than an age-independent cognitive module Tomasello et al. (1993) Caldwell et al. (2020).

Transmission across generations can involve individuals of similar age, overlapping cohorts, and adult–child relationships. Vertical transmission through families, horizontal transmission among peers, and oblique transmission from unrelated older individuals differ in exposure diversity and the persistence of rare skills. Laboratory chains usually compress these social arrangements into brief episodes and therefore measure a restricted form of generational succession. Their causal control is valuable, but direct extension to multiyear apprenticeship requires ethnographic and longitudinal corroboration Boyd & Richerson (1985) Muthukrishna et al. (2014).

Research also needs to distinguish a young learner's ability to copy a complete unfamiliar solution from the ability to improve a familiar one. Improvements may first depend on expert support, later becoming autonomous; a population can accumulate knowledge even when many individuals have only partial competence. Future developmental experiments should separate initial acquisition, delayed retention, reconstruction from products, and transfer to new tasks, with matched opportunities for social assistance. Such evidence would connect the psychology of learning to the population consequences of fidelity Caldwell et al. (2020) Derex et al. (2024).

Demographic Size, Effective Population, and Cultural Loss

The persistence of complex skills depends partly on who can learn from whom. Total census size differs from the effective population of accessible experts, and even a large population may contain isolated specialist communities. When models are scarce and transmission imperfect, a hard-to-relearn procedure can disappear after demographic disruption. Mathematical treatments connect these outcomes to the distribution of model competence, copying error, and learner selection. Population size is therefore an explanatory variable that must be linked to actual contact and cultural access Henrich (2004) Powell et al. (2009).

Henrich's 2004 model explored how reduced effective learner pools could produce maladaptive technological losses, discussing the Tasmanian case. Powell, Shennan, and Thomas later proposed a demographic account of the late-Pleistocene appearance of modern human behavior. Their contributions established powerful hypotheses about demographic thresholds and accumulation. The archaeological interpretation is more uncertain than the existence of the underlying mathematical possibility, since historical population estimates, technology classification, social networks, and ecological alternatives are imperfectly observed Henrich (2004) Powell et al. (2009).

A useful empirical design measures networks of potential teachers and the redundancy of specialized skills instead of using national or settlement population size as a direct proxy. It also distinguishes maintaining a technique from inventing a better one. Demographic expansion may increase access to rare models, allow parallel experimentation, or distribute risk of knowledge loss; these effects need not follow the same response curve. Laboratory studies with manipulated numbers of demonstrators can isolate some mechanisms, whereas archaeological evidence constrains the plausibility of projected historical scenarios Muthukrishna et al. (2014) Derex & Boyd (2016).

Sociality and Access to Multiple Cultural Models

Learning from several individuals may protect against a poor demonstration, give access to complementary strategies, or allow selection of higher-performing variants. The benefit depends on quality differences among models and on a learner's ability to compare them. If all demonstrators replicate the same error, additional exposure yields little new information. If their solutions provide incompatible pieces, recombination may demand substantial integration skill. Sociality is therefore a measurable architecture of information availability rather than a uniformly beneficial psychological trait Muthukrishna et al. (2014) Lewis & Laland (2012).

Muthukrishna and colleagues tested how access to one versus multiple cultural models affected acquisition and retention of complex tasks in laboratory generations. Their experiments provide controlled evidence that access to models can alter cultural complexity, though the imposed tasks and undergraduate populations constrain generalization. The study separates a concrete contact mechanism from the much broader claim that demographic size alone causes technological development. Model availability can modify loss probabilities without guaranteeing cumulative improvement in every social environment Muthukrishna et al. (2014).

In natural groups, individuals choose whom to observe according to experience, status, kinship, proximity, and prior success. Those choices create information bottlenecks that determine whose discoveries are retained. A large network with little cross-community learning may preserve local traditions but hinder recombination, while a small but connected group may exchange specialized skills efficiently. Causal tests should independently manipulate model number, visibility of outcomes, and opportunities to combine information. Such experiments connect person-level learning decisions with population-level cultural dynamics Derex & Boyd (2016) Gunasekaram et al. (2024).

Network Connectivity, Modularity, and Innovation Diversity

Network architecture determines both the circulation of discoveries and the independence of experimentation. Dense connectivity can rapidly broadcast an effective solution, reducing immediate losses, but it can also make learners converge early on a merely adequate design. Partial connectivity preserves distinct local approaches long enough for some groups to explore alternatives and later exchange successful components. The resulting balance between exploitation and exploration depends on task payoffs, time horizons, mobility, and the extent to which variants can be combined Derex & Boyd (2016).

Derex and Boyd experimentally examined partially connected groups and reported greater cultural accumulation under particular network conditions. The central result concerns the advantage of maintaining some separation between local experimentation and eventual exchange, rather than a universal prescription for sparse networks. More recent comparative work on chimpanzee population connectivity highlights analogous historical possibilities while drawing on observational genetic and behavioral reconstructions. Such evidence supports network-mediated hypotheses but is sensitive to migration proxies, trait classification, and unmeasured ecological history Derex & Boyd (2016) Gunasekaram et al. (2024).

A cultural network may contain several kinds of edges: observation, active instruction, movement of practitioners, exchange of physical artifacts, or communication about successful designs. These links have different capacities to transmit tacit procedures, causal explanations, and complete objects. A simple measure of connectivity obscures those differences. Better studies record the information made available along each edge and estimate the survival of components across generations. Dynamic models can then test whether modularity increases cultural exploration under conditions observed in real systems Morgan & Feldman (2025) Derex et al. (2024).

Mathematical Transmission Models and Their Assumptions

Formal models clarify why cultural information may accumulate without assuming an extraordinary individual learning capacity. A simple lineage can be represented by inherited quality, the probability of retaining relevant features, innovation opportunities, and selection among candidate variants. More detailed models represent multiple components, several potential teachers, changing task demands, and network topology. Their outputs describe conditional trajectories: accumulation under some parameter settings, loss or stasis under others. The parameters require empirical estimation before historical interpretation Lewis & Laland (2012) Henrich (2004).

The transmission-fidelity model of Lewis and Laland shows that changes in copying accuracy can substantially alter the ability of a cultural lineage to maintain and combine refinements. This establishes a mathematically coherent route from modest individual differences to population-level divergence. Model dependence matters: cultural traits can have different error structures, and stable external records can compensate for imperfect psychological memory. A model with an abstract copying coefficient may predict the direction of change while failing to reproduce the component-level mechanisms of a particular technology Lewis & Laland (2012) Morgan & Feldman (2025).

Demographic models additionally depend on assumptions about how learners select competent exemplars. Results can change when imitation is random, conformist, success-biased, or directed toward exceptional individuals. Therefore claims about thresholds should include model sensitivity analysis and realistic estimates of social learning strategies. Experiments that manipulate access to experts, compare competing designs, and record diffusion routes provide the observations necessary to evaluate the model components instead of simply fitting a broad narrative of cultural progress Henrich (2004) Powell et al. (2009) Muthukrishna et al. (2014).

Experimental Microsocieties and Replacement Chains

Laboratory microsocieties place participants into small sequential groups, preserving or replacing individuals according to an experimental schedule. Researchers can measure the inherited artifacts, available information, performance, and innovations at each stage. Such chains make cultural history observable over compressed timescales and allow comparison with groups that repeatedly start afresh. Early experiments developed this approach for artifact design and other tasks, demonstrating that population-level dynamics can be studied with standard psychological interventions Caldwell & Millen (2008).

A key design choice is how much social information is transferred. Passing the best physical product allows reverse engineering but conceals its production process. Supplying a video or direct observation may preserve procedural details; allowing communication can transmit abstract explanations and explicit correction. Each condition tests a distinct theory of learning. The resulting performance curve should be accompanied by records of the intermediate designs, since improvements can be episodic, nonlinear, and dependent on unusual successful individuals Caldwell & Millen (2009) Derex et al. (2024).

Chains also generate dependence between observations: a late-generation outcome inherits the contingencies of earlier members, and treating all participants as independent can exaggerate precision. Suitable inference requires replicate lineages, prespecified generation-level outcomes, and analysis that respects nested or dependent observations. Chain attrition, experimenter intervention, and variable motivation can further bias results. This methodological discipline is essential when comparing relatively small animal groups, young children, and adult laboratory participants Caldwell et al. (2020) Derex et al. (2024).

Controls for Individual Practice and Social Information

Improvement over generations is a descriptive pattern until controls identify why it occurred. Repeated individual learning tests whether equivalent time and experience yield similar gains. Reset conditions compare later generations to fresh starters; scrambled-history controls disrupt access to a genuine predecessor. In tasks with strong practice effects, participants may outperform predecessors simply through longer exposure, altered instructions, or favorable initial starting points. The strongest interpretation compares social and asocial paths with matched opportunities and transparent measures Derex et al. (2024) Caldwell et al. (2020).

Derex, Edmiston, Lupyan, and Mesoudi compared common chain and group designs and alternative control conditions in a 2024 PNAS investigation. Their methodological findings show why experimental architecture can modify the learning trade-offs participants face. A test of cultural evolutionary theory must specify whether it concerns successive copying, collective search, the benefit of social information, or the creation of new solutions. These mechanisms can produce similar curves but require different causal contrasts Derex et al. (2024).

The absence of improvement also requires careful interpretation. If an artifact already approaches a ceiling, its stable reproduction may represent effective conservation, whereas a deliberately opaque task may prevent any learner from accessing its critical operations. Power, reward value, model visibility, and the time allowed for exploration all affect a negative result. Comparative research should ensure that social learning opportunities are meaningful to each species and that control tasks establish basic perceptual and motor competence Dean et al. (2012) Caldwell et al. (2020).

Comparative Evidence from Children, Chimpanzees, and Capuchins

Comparative experiments examine whether different learners can use preceding solutions to reach more complex or rewarding states. Dean and colleagues' three-stage puzzlebox study found that the children's access to later rewards covaried with social processes including instruction, imitation, and helping, while chimpanzees and capuchins under the tested conditions did not show equivalent progression. The experiment supports a conjunction of task performance and social process differences; it cannot by itself identify which component of the social package produced the observed pattern Dean et al. (2012).

Comparative inference requires attention to a species' motor repertoire, attentional ecology, motivation, and social context. A device that is intuitive to school-age humans may contain unusually opaque actions for another primate, and brief captivity-based experiments may misrepresent typical opportunities for observing skilled models. Conversely, failures can be informative when apparatus accessibility and incentives are well controlled. Experiments should ask which required information remained inaccessible and how intervention changes the probability of acquisition Caldwell et al. (2020) Tennie et al. (2009).

The older assumption that humans alone can copy knowledge beyond individual invention has been qualified by subsequent animal evidence. The revised question concerns the extent, reliability, and flexibility with which species can retain, combine, and extend such knowledge across generations. Direct comparisons of these dimensions need common functional metrics, comparable baselines, and procedures tailored to the perceptual capacities of the tested learners. Distinct task affordances must remain visible in the interpretation van Leeuwen et al. (2024) Bridges et al. (2024) Morgan & Feldman (2025).

Chimpanzee Know-How Beyond Individual Reinnovation

Van Leeuwen, DeTroy, Haun, and Call studied chimpanzees' acquisition of a sequential puzzlebox solution in a controlled sanctuary setting. Following an extended baseline in which naive individuals did not independently solve the task, trained demonstrators were introduced to two groups. The researchers reported that 14 naive chimpanzees acquired the skill after social exposure and used network-based diffusion analysis to assess the role of observation. This provides evidence that chimpanzees can use social learning for a skill they did not reinnovate under the baseline conditions van Leeuwen et al. (2024).

The study addresses a fundamental premise of cumulative culture: whether a learner can inherit a successful, relatively complex behavior without independently discovering it. It does not experimentally demonstrate a long succession of independently generated improvements in the same chimpanzee group. Skill acquisition after a seeded demonstration is one prerequisite that could support a ratchet, and its relation to ongoing cultural accumulation remains a distinct hypothesis. The difference between a prerequisite and a completed cumulative lineage matters when assessing claims of evolutionary uniqueness van Leeuwen et al. (2024) Caldwell et al. (2020).

Observational exposure, demonstration frequency, and social structure were part of the empirical explanation, but as with network diffusion analyses, interpretation depends on accurate records of opportunities and alternative routes of acquisition. Repeated baseline failures strengthen the contrast with social learning, while individual variation and behavioral motivation continue to shape generalization. A useful next design would document whether multiple waves of newly trained chimpanzees preserve the behavior and produce replicable functional refinements van Leeuwen et al. (2024) Gunasekaram et al. (2024).

Bumblebee Social Acquisition of a Two-Step Technique

Bridges and colleagues trained bumblebee demonstrators to solve a two-step puzzle involving an initially unrewarded action. Naive observers subsequently acquired the complete sequence after witnessing trained individuals, even though independent acquisition was not observed under the study's baseline opportunities. The authors reported that five of fifteen naive observers learned the technique. The result is especially informative because the behavior required a step whose immediate reward was absent, placing constraints on unguided trial-and-error learning Bridges et al. (2024).

This experimental result challenges the notion that acquisition of complex know-how beyond observed individual reinnovation requires specifically human capacities. It does not show an established lineage of successive bee-generated improvements, and experimental training was essential to seed the initial demonstration. The study therefore informs the comparative distribution of a potential precondition for cumulative culture rather than settling whether free-living bumblebee populations maintain open-ended cultural technologies Bridges et al. (2024) Morgan & Feldman (2025).

Mechanistic claims should specify exactly which cues the observer used. Seeing another bee obtain a reward may change attention to the apparatus, while observation of particular actions may provide sequence-specific information; those processes have different implications for fidelity. Future experiments can vary which stage is demonstrated, restrict access to product information, and document delayed reproduction and generational turnover. Invertebrate systems broaden comparative theory by exposing the difference between basic social acquisition, complex action learning, and cumulative historical change Bridges et al. (2024) Caldwell et al. (2020).

Pigeon Route Optimization and Collective Accumulation

Sasaki and Biro tested homing-pigeon navigation using pairs in which experienced birds were repeatedly replaced with naive partners. Across such turnover, routes became more efficient, supporting the interpretation that collective route knowledge can undergo cumulative improvement. The functional measure—route efficiency—was clearer than a vague impression of greater cultural complexity. Pair replacement created overlapping continuity while allowing new individuals to alter the trajectory. This illustrates how population-level gains can emerge from repeated interaction rather than deliberate teaching Sasaki & Biro (2017).

The mechanism differs from transmission of a manufactured tool. Navigation involves landmark memory, directional preferences, and coordinated motion; a newcomer can alter a path through local interactions without reconstructing the predecessor's internal representation. An efficient route may be cumulative even when no individual encodes a complete symbolic plan. Researchers should avoid equating such improvement with the unlimited combinatorial expansion of human technology. The experiment establishes a tractable case of cumulative functional change under particular navigation conditions Sasaki & Biro (2017) Morgan & Feldman (2025).

Dalmaijer subsequently used minimal artificial navigators with goal direction, memory, social alignment, and movement continuity to examine how similar improvements could emerge. The simulation provides a proof of possibility that relatively simple interaction rules can generate accumulation-like route trends, while it cannot establish that pigeons use exactly those computational rules. Comparing simulations and animal trajectories can sharpen hypotheses about required cognitive mechanisms, but successful imitation of a pattern by a model does not identify the actual causal process Dalmaijer (2024) Sasaki & Biro (2017).

Chimpanzee Cultural Connectivity and Reconstructed Histories

Gunasekaram and colleagues combined genetic markers of population connectivity with comparative cultural trait information across chimpanzee subspecies. Their 2024 analysis reported associations between population connectivity and the distribution and complexity of cultural traits, interpreted as consistent with limited instances of cumulative culture and historical exchange between groups. Such work expands the evidence beyond small experimental apparatuses and asks whether long-run network processes leave detectable traces in natural populations Gunasekaram et al. (2024).

The observational approach has important limits. Genetic resemblance is an indirect record of migration and shared history, not a timestamped record of teaching a particular technology. Trait inventories may be incomplete, and shared ecology can favor similar solutions independently. Cultural trait classification must distinguish component accumulation from coexistence of several independent behaviors. The analysis strengthens a plausible network-based account but does not eliminate all alternative causal histories Gunasekaram et al. (2024) Morgan & Feldman (2025).

Combining field observations, material sequences, genealogy or dispersal markers, and experimental tests would improve historical reconstruction. Such triangulation can identify whether innovations spread preferentially along migration corridors, whether communities retain component variants after exchange, and whether different tool elements have different histories. A mature comparative framework recognizes ecological opportunity, social access, and fidelity as interacting variables instead of assigning cumulative culture to a species as a single undifferentiated trait Gunasekaram et al. (2024) Derex & Boyd (2016).

Human Uniqueness and the Open-Endedness Hypothesis

Human cultural histories display enormous diversity of technologies, institutions, scientific knowledge, and socially maintained skills. The challenge is explaining sustained expansion across distant domains and exceptionally long sequences of dependence. Earlier comparative theories often attributed this phenomenon to uniquely human high-fidelity imitation and cooperation. Recent evidence of accumulation-like processes and complex socially acquired skills in other animals places pressure on an absolute categorical boundary Tennie et al. (2009) van Leeuwen et al. (2024) Bridges et al. (2024).

Morgan and Feldman propose that the most distinctive feature of human culture may be open-endedness rather than cumulativeness as such. Their comparison considers variation available for inheritance, pathways of inheritance, and consequences of culturally generated novelty, moving the question from whether animals accumulate anything to how far cumulative processes can explore new design spaces. The proposal is a theoretical interpretation supported by comparative argument; its predictions require operationalized measures and prospective tests across populations and tasks Morgan & Feldman (2025).

A rigorous research program can compare the number and type of independent innovation dimensions, interactions among components, persistence after expert loss, and transfer across environments. Human advantage might arise from several interacting processes—language, institutions, teaching, population mobility, material archives, and cooperative specialization—rather than a single copying mechanism. Some mechanisms may be widespread, others unusually developed, and their configurations may produce nonlinear historical outcomes. Such interactionist explanations fit the evidence more closely than untested assertions of species-exclusive psychological traits Dean et al. (2012) Morgan & Feldman (2025).

Strategic Selection of Knowledge in Developmental Transmission

Transmission accuracy is partly determined by the sender's choice of information. An expert may preserve the highest-performing component, draw attention to diagnostic failures, or transmit a convenient but ineffective detail. Mackintosh, Atkinson, Renner, and Caldwell tested the developmental emergence of such strategic selection and found that adults prioritized knowledge useful to a subsequent learner. Relative to randomly chosen task information, intentional selection supported the accumulation of beneficial discoveries. This finding concerns the content and selectivity of communication, extending analysis beyond an observer's ability to imitate a demonstration Mackintosh et al. (2026) Caldwell & Millen (2009).

In the child arm of the experiment, competent younger participants did not show an equally strong tendency to transmit strategically useful knowledge. The authors reported that the advantage of intentional over incidental transmission was not apparent until approximately nine to ten years of age. The interpretation is specific: mastery of a task and the capacity to identify which details another learner needs are separable achievements. It would be unwarranted to infer that younger children cannot transmit culture, as they may transmit effectively in other tasks and under substantial social guidance Mackintosh et al. (2026) Tomasello et al. (1993).

The study adds a sender-side dimension to fidelity. Measuring how well learners reproduce an example cannot reveal whether the exemplar was well chosen in the first place. Future experiments should assess the quality of available discoveries, selection of transmitted examples, downstream reproduction, and functional value separately. Developmental comparisons also need to establish whether verbal reasoning, teaching experience, or a changing model of the recipient mediates selective communication. Intentional transmission becomes an empirically decomposable mechanism rather than a synonym for high-fidelity copying Mackintosh et al. (2026) Dean et al. (2012).

Causal Reasoning, Technical Intuition, and Collective Performance

Saral, Ibáñez de Aldecoa, and Derex investigated whether technological improvement across transmission chains requires causal understanding of an artifact's underlying mechanism. Across two experimental chains involving 900 participants, they contrasted tasks allowing technical reasoning, structurally corresponding causal reasoning without ordinary physical cues, and a pure transmission condition without usable causal structure. Cumulative improvement occurred even in the pure transmission condition. This provides strong task-specific evidence that the inheritance and selection of successful variants can generate collective advances beyond any one learner's causal insight Saral et al. (2026).

Causal reasoning accelerated early improvements by directing exploration toward promising areas of the solution space, yet its advantage diminished across generations. Technical reasoning provided no extra benefit beyond the abstract causal condition in the experiments. Such patterns complicate a universal claim that domain-specific physical intuition is a necessary psychological foundation of cumulative technological evolution. They also establish a positive role for causal reasoning under certain temporal conditions: it can change the speed of exploration even when inheritance permits eventual improvement without it Saral et al. (2026) Lewis & Laland (2012).

Interpretation depends on the restricted design space and payoff structure of the tasks. A pure cultural transmission condition does not represent every real-world engineering challenge, particularly where hidden failures make performance hard to evaluate or environmental variation is large. The experimentally supported conclusion is that causal representations were dispensable for accumulation in the tested lineages, while early exploration benefited from them. Combining such manipulations with knowledge-selection studies would test how reasoning by senders and reconstruction by receivers jointly determine downstream fidelity Saral et al. (2026) Mackintosh et al. (2026).

Cognitive Appeal, Content Bias, and Ineffective Cultural Persistence

High fidelity can conserve errors as efficiently as improvements. Saral, Singh, Jacquet, Jang, and Derex experimentally varied the visual appeal of ineffective solutions in a preregistered transmission-chain study. The payoff structure remained constant while visual features made an intuitively appealing strategy more tempting in one condition. The low-appeal treatment involved 240 participants and showed increasing solution efficiency across generations; the high-appeal treatment also involved 240 participants, produced costlier solutions, and showed no statistically significant efficiency improvement over the study's generations Saral et al. (2026).

Participants in the visually appealing condition converged on matching colorful inputs to colorful targets even when the costly matches were not functionally justified. The favored solutions circulated through both observed choices and written advice. Consequently, the experiment demonstrates that the cultural transmission of information can increase stability while reducing useful exploration; an inherited explanation can focus the next generation on a misleading region of the design space. The result is more precise than the general statement that cultural evolution sometimes produces maladaptation, because a specific task-level content bias was manipulated Saral et al. (2026).

The experiment tests a particular visual correspondence heuristic in a virtual resource-allocation task. It does not establish that all ritual practices, medical beliefs, or technologically persistent inefficiencies share that mechanism. Nevertheless, it shows why a theory of accumulation must model subjective appeal, perceived plausibility, and the quality of payoff feedback alongside copying accuracy. More social models, larger communities, and stronger instruction may preserve a solution while also amplifying its errors. Historical success and persistence require independent criteria rather than being inferred from the mere durability of a tradition Saral et al. (2026) Morgan & Feldman (2025).

Archaeological and Historical Inference from Material Sequences

Archaeological assemblages provide long records of changing artifact forms and techniques that laboratory chains cannot reproduce. Researchers compare stratigraphic order, component organization, material availability, likely function, and geographical distribution to infer continuity and transformation. The strongest claims identify inherited design features together with innovations whose sequence is consistent with dependence on predecessors. Because artifacts can be convergent solutions to common constraints, similarity requires contextual corroboration before being interpreted as social transmission Henrich (2004) Powell et al. (2009).

A historical increase in technical complexity can coexist with periods of stasis and abrupt loss. Preservation biases mean that perishable technologies and informal instruction may remain invisible, while stone or durable tools overrepresent certain aspects of cultural change. Models suggesting links between demographic expansion and behavioral innovation make predictions about the survival of complex practices, but archaeology rarely measures expert contact or copying error directly. Cross-validation across independent sites, ecological conditions, and different technologies is needed to strengthen inference Henrich (2004) Powell et al. (2009).

The distinction between evolutionary reconstruction and experimental demonstration should be maintained in public scientific narratives. An experiment may establish that increasing access to models protects a task solution; historical work asks whether a particular ancient community actually had that access and whether the result explains its recorded artifacts. When the necessary variables are unobservable, researchers can compare alternative scenarios and report conditional plausibility rather than converting a plausible simulation into a confirmed history. This form of restraint produces a more cumulative scientific literature in its own right Mesoudi et al. (2006) Derex et al. (2024).

Measuring Fidelity and Cultural Complexity Across Contexts

Measurement starts by defining the unit of inherited information. In paper-airplane construction, outcome may be flight distance, fold sequence, or geometric design. In navigation, it may be route efficiency and landmark use. In multi-step puzzles, action ordering and access to a reward can be separated. Because these units have distinct error distributions, direct numerical comparison across tasks is rarely valid. A comprehensive report should include task validity, model exposure, opportunity to explore, individual learning controls, and the temporal persistence of the acquired variant Caldwell & Millen (2009) Sasaki & Biro (2017) van Leeuwen et al. (2024).

Structural complexity and functional sophistication demand explicit coding schemes. Counting components can misclassify minimalist but high-performing artifacts; counting stages can misclassify arbitrary ritualized behavior as advanced technology. A more informative system reports at least functional outcome, component dependencies, reproducibility, and expansion of the feasible design space. Predefined scoring and blind coding where feasible reduce the risk that researchers retrospectively label successful conditions culturally cumulative. Repeated evaluation after delayed intervals helps distinguish immediate task success from durable transmission Caldwell et al. (2020) Derex et al. (2024).

Cross-cultural and cross-species comparability also involves measurement invariance at the level of task meaning. The same apparatus may afford different actions to learners with different bodies or training histories. Researchers should test whether the task genuinely measures the same hypothesized capacity in each population and report context-specific constraints. In comparative biology, methodological adaptation is often necessary for fair inference; equivalence is achieved through a common theoretical question and credible controls, not necessarily identical hardware Dean et al. (2012) Bridges et al. (2024).

Replication, Generalization, and Competing Explanations

A robust cumulative-cultural finding should persist across independent lineages, suitable controls, and clearly defined performance criteria. Variation between lineages is informative: a few unusually successful founders may drive the average, while most histories stagnate or lose innovations. Replication should therefore examine the distribution and shape of trajectories as well as mean improvement. Dependence of later generations on earlier ones must be represented in the uncertainty estimates and in the description of what was actually repeated Caldwell et al. (2020) Derex et al. (2024).

Competing explanations include individual learning, stimulus enhancement, local environmental constraint, repeated exposure, experimenter scaffolding, group convergence, and social transmission of information without cumulative modification. These are hypotheses that can be separated by systematically changing the information provided and tracking the intermediate states of the cultural variant. In animal experiments, absence of a naive solution over a defined baseline supports a particular counterfactual only under the tested time, reward, and access conditions. The literature is strengthened when those bounds remain explicit van Leeuwen et al. (2024) Bridges et al. (2024).

Formal models can be replicated computationally, but simulation robustness requires parameter variation and comparison with independent empirical patterns. The emergence of efficient routes from simple artificial-agent rules illustrates that one observable trajectory can have several possible mechanistic explanations. Studies should report sensitivity to initial conditions, stochastic variability, and decisions about agent cognition. A good model reduces the set of plausible mechanisms; its resemblance to one empirical curve alone is insufficient to identify the real mechanism Dalmaijer (2024) Sasaki & Biro (2017).

Institutions, External Records, and Technological Scaffolding

Inherited culture is organized through social relations and durable environments as well as individual memory. Institutions can regulate training, credential expertise, standardize procedures, and maintain archives; material artifacts can store successful solutions for individuals who never encountered the original inventor. These resources change the effective fidelity of a system by enabling error checking and reconsultation. The theoretical consequence is that fidelity should sometimes be measured at the level of a distributed learning arrangement rather than exclusively as the accuracy of a single learner Boyd & Richerson (1985) Morgan & Feldman (2025).

External records can enlarge the number of available cultural ancestors: a learner may combine instructions, artifacts, and demonstrations generated at different times and places. This produces overlapping lineages and can reduce dependence on one charismatic or unusually skilled teacher. Yet stable records also codify mistakes, and access inequality may restrict which groups benefit from stored expertise. Research must distinguish the existence of information in an archive from its successful retrieval, interpretation, and adaptive use Derex & Boyd (2016) Muthukrishna et al. (2014).

Technological scaffolding can also alter the direction of exploration. Standardized components may lower error and make combinations feasible, while rigid conventions may discourage experimentation. The resulting trade-off resembles that observed in network designs: innovation benefits from enough stable inherited structure to make experiments meaningful and enough diversity to compare alternatives. This interpretation extends the psychological account of copying to a culture's infrastructure without treating infrastructure as a substitute for empirical learning mechanisms Lewis & Laland (2012) Derex & Boyd (2016).

Cultural Evolution in the Artificial Era

The Artificial Era creates new research settings in which human learners exchange information through search engines, recommendation systems, digital repositories, collaborative platforms, and artificial cognitive systems. These media can preserve text and procedure with high surface accuracy while changing which variants are discovered, selected, recombined, and transmitted. A copied document may be identical at the character level yet misunderstood in use; a reconstructed explanation may differ lexically while maintaining functional knowledge. The established distinction between product, process, and functional fidelity is especially relevant to digital transmission Caldwell & Millen (2009) Morgan & Feldman (2025).

Artificial systems can be modeled as participants in information networks, but the attribution of psychological capacities requires independent operational criteria. A software agent might reproduce a procedure, optimize an artifact, or coordinate with other agents under explicit objectives. Those observations establish performance on designed tasks and may permit experiments on transmission bottlenecks or collective search; they do not by themselves establish consciousness, humanlike motivation, or a particular social-cognitive mechanism. Artificial-agent route simulations illustrate the usefulness of mechanistic baselines and the limits of inferring biological cognition from matching outputs Dalmaijer (2024).

Research at the human–Artificial interface should compare human-only, artificial-only, and mixed chains with matched time, feedback, and information access. It can measure factual retention, procedural reproducibility, novelty, dependence on upstream exemplars, and persistence after tool access changes. Researchers should audit provenance, contamination between experimental conditions, and independent evaluation of claimed improvements. Such designs situate Artificial cognition within cultural transmission research while preserving a clear distinction between empirically observed function and the broader philosophical horizon of Artificial Sapiens Derex et al. (2024) Morgan & Feldman (2025).

Unresolved Problems and a Cumulative Research Agenda

The field's most important unresolved question is how copying accuracy, innovative variation, and network ecology combine to create persistently expanding cultural repertoires. Existing studies have demonstrated retention, social acquisition of difficult skills, and cumulative functional improvement under selected conditions. They have yet to provide a single general model that predicts open-ended technological diversification across species and ecological contexts. A productive program will compare component-level fidelity, innovation distributions, effective teacher pools, and the frequency with which independent lineages merge Lewis & Laland (2012) Morgan & Feldman (2025).

A second challenge is causal scale. Laboratory studies identify mechanisms over minutes or days, whereas historical and ecological claims extend across years, generations, or evolutionary periods. Bridging these scales requires field experiments, longitudinal studies of learning networks, historical artifact databases, and models calibrated to observed contact structures. Carefully replicated comparative research should also revisit apparent species boundaries with accessible tasks and transparent negative-evidence criteria. The 2024 animal studies provide new tests of know-how acquisition, and network analyses provide hypotheses about its longer-run consequences van Leeuwen et al. (2024) Bridges et al. (2024) Gunasekaram et al. (2024).

A third challenge concerns the classification of outcomes. Functional improvement can arise through collective interaction without precise copying of actions, while exact transmission can preserve an unproductive convention. Future work should track both the inherited informational components and the resulting functional trajectories, distinguishing accidental gain, local optimization, and genuinely new design dimensions. These distinctions will improve theory comparison, cross-study synthesis, and the eventual quantitative integration of experimental literatures that currently use overlapping labels for different phenomena Sasaki & Biro (2017) Dalmaijer (2024) Derex et al. (2024).

Conclusion

Cumulative cultural evolution is a historical population process sustained by information transmission, selection among variants, innovation, and the preservation of usable results. Transmission fidelity is a crucial mechanism because a discovery that repeatedly disappears cannot reliably support later refinements. Its effect depends on what information survives, how errors are corrected, and whether learners have access to diverse complementary models. Laboratory chains, formal models, comparative experiments, and natural histories together support this integrated account, each at a different inferential scale Lewis & Laland (2012) Caldwell & Millen (2008) Derex & Boyd (2016).

The comparative evidence now includes socially transmitted complex skills in chimpanzees and bumblebees and cumulative navigation improvements in pigeons. These findings revise simple claims that the prerequisites or all forms of accumulation are confined to humans. They also sharpen the explanatory problem posed by human cultural histories, whose unusually extensive and open-ended technological and institutional development remains a central research target. Human cultural expansion is best investigated as an interaction of learning, network connectivity, cooperative instruction, material preservation, and recombination van Leeuwen et al. (2024) Bridges et al. (2024) Sasaki & Biro (2017) Morgan & Feldman (2025).

The scientific synthesis therefore establishes a methodological program: define the inherited object, measure the relevant dimensions of fidelity, reconstruct the actual lineage of modifications, compare it with controlled individual-learning paths, and test the sensitivity of models to social and demographic assumptions. Where evidence remains associative or simulation-based, its explanatory status must stay explicit. This program links developmental psychology, comparative cognition, evolutionary modeling, and the emerging study of cultural transmission through artificial systems in a common framework of observable, historically dependent knowledge Dean et al. (2012) Derex et al. (2024) Dalmaijer (2024).

References

  • Boyd, R., & Richerson, P. J. (1985). Culture and the evolutionary process. University of Chicago Press. https://press.uchicago.edu/ucp/books/book/chicago/C/bo5970597.html
  • Bridges, A. D., Royka, A., Wilson, T., Lockwood, C., Richter, J., Juusola, M., & Chittka, L. (2024). Bumblebees socially learn behaviour too complex to innovate alone. Nature, 627, 572–578. https://doi.org/10.1038/s41586-024-07126-4
  • Caldwell, C. A., & Millen, A. E. (2008). Studying cumulative cultural evolution in the laboratory. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1509), 3529–3539. https://doi.org/10.1098/rstb.2008.0133
  • Caldwell, C. A., & Millen, A. E. (2009). Social learning mechanisms and cumulative cultural evolution: Is imitation necessary? Psychological Science, 20(12), 1478–1483. https://doi.org/10.1111/j.1467-9280.2009.02469.x
  • Caldwell, C. A., Atkinson, M., Blakey, K. H., Dunstone, J., Kean, D., Mackintosh, G., Renner, E., & Wilks, C. E. H. (2020). Experimental assessment of capacities for cumulative culture: Review and evaluation of methods. WIREs Cognitive Science, 11(1), e1516. https://doi.org/10.1002/wcs.1516
  • Dalmaijer, E. S. (2024). Cumulative route improvements spontaneously emerge in artificial navigators even in the absence of sophisticated communication or thought. PLOS Biology, 22(6), e3002644. https://doi.org/10.1371/journal.pbio.3002644
  • Dean, L. G., Kendal, R. L., Schapiro, S. J., Thierry, B., & Laland, K. N. (2012). Identification of the social and cognitive processes underlying human cumulative culture. Science, 335(6072), 1114–1118. https://doi.org/10.1126/science.1213969
  • Derex, M., & Boyd, R. (2016). Partial connectivity increases cultural accumulation within groups. Proceedings of the National Academy of Sciences, 113(11), 2982–2987. https://doi.org/10.1073/pnas.1518798113
  • Derex, M., Edmiston, P., Lupyan, G., & Mesoudi, A. (2024). Trade-offs, control conditions, and alternative designs in the experimental study of cultural evolution. Proceedings of the National Academy of Sciences, 121(48), e2322886121. https://doi.org/10.1073/pnas.2322886121
  • Gunasekaram, C., Battiston, F., Sadekar, O., Padilla-Iglesias, C., van Noordwijk, M. A., Furrer, R., Manica, A., Bertranpetit, J., Whiten, A., van Schaik, C. P., Vinicius, L., & Migliano, A. B. (2024). Population connectivity shapes the distribution and complexity of chimpanzee cumulative culture. Science, 386(6724), 920–925. https://doi.org/10.1126/science.adk3381
  • Henrich, J. (2004). Demography and cultural evolution: How adaptive cultural processes can produce maladaptive losses—The Tasmanian case. American Antiquity, 69(2), 197–214. https://doi.org/10.2307/4128416
  • Lewis, H. M., & Laland, K. N. (2012). Transmission fidelity is the key to the build-up of cumulative culture. Philosophical Transactions of the Royal Society B: Biological Sciences, 367(1599), 2171–2180. https://doi.org/10.1098/rstb.2012.0119
  • Mackintosh, G., Atkinson, M., Renner, E., & Caldwell, C. A. (2026). Intentional transmission of knowledge optimises cumulative cultural evolution. Scientific Reports. Advance online publication. https://doi.org/10.1038/s41598-026-62734-6
  • Mesoudi, A., Whiten, A., & Laland, K. N. (2006). Towards a unified science of cultural evolution. Behavioral and Brain Sciences, 29(4), 329–347. https://doi.org/10.1017/S0140525X06009083
  • Morgan, T. J. H., & Feldman, M. W. (2025). Human culture is uniquely open-ended rather than uniquely cumulative. Nature Human Behaviour, 9, 28–42. https://doi.org/10.1038/s41562-024-02035-y
  • Muthukrishna, M., Shulman, B. W., Vasilescu, V., & Henrich, J. (2014). Sociality influences cultural complexity. Proceedings of the Royal Society B: Biological Sciences, 281(1774), 20132511. https://doi.org/10.1098/rspb.2013.2511
  • Powell, A., Shennan, S., & Thomas, M. G. (2009). Late Pleistocene demography and the appearance of modern human behavior. Science, 324(5932), 1298–1301. https://doi.org/10.1126/science.1170165
  • Saral, A. S., Ibáñez de Aldecoa, P., & Derex, M. (2026). The roles of cultural transmission and causal reasoning in the cultural evolution of technology. PNAS Nexus, 5(4), pgag090. https://doi.org/10.1093/pnasnexus/pgag090
  • Saral, A. S., Singh, M., Jacquet, A., Jang, H., & Derex, M. (2026). Cognitive appeal promotes the persistence of inefficient solutions and hinders cumulative cultural evolution. Nature Communications, 17, Article 10752. https://doi.org/10.1038/s41467-026-77677-9
  • Sasaki, T., & Biro, D. (2017). Cumulative culture can emerge from collective intelligence in animal groups. Nature Communications, 8, 15049. https://doi.org/10.1038/ncomms15049
  • Tennie, C., Call, J., & Tomasello, M. (2009). Ratcheting up the ratchet: On the evolution of cumulative culture. Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1528), 2405–2415. https://doi.org/10.1098/rstb.2009.0052
  • Tomasello, M., Kruger, A. C., & Ratner, H. H. (1993). Cultural learning. Behavioral and Brain Sciences, 16(3), 495–552. https://doi.org/10.1017/S0140525X0003123X
  • van Leeuwen, E. J. C., DeTroy, S. E., Haun, D. B. M., & Call, J. (2024). Chimpanzees use social information to acquire a skill they fail to innovate. Nature Human Behaviour, 8, 891–902. https://doi.org/10.1038/s41562-024-01836-5