Relations between Description and Experimentation in the Metacontingency Enterprise: An Interbehavioral Analysis

Abstract

Despite extensive theoretical development, there is a lack of consensus in the metacontingency enterprise on the extent to which current metacontingency constructs describe experimental happenings. The purpose of this article is to provide an interbehavioral analysis of the metacontingency enterprise that examines relations between description and experimentation in order to facilitate research on cultural selection occurring through metacontingencies. In particular, this article considers how stimulus functions of descriptions of metacontingency constructs participate in metacontingency experiments in terms of specificity, types of analysis, levels of analysis, and procedures. The extent to which experimental findings are able to be described in terms of metacontingency constructs is assessed. Prominent events and relations demonstrated by metacontingency experiments are summarized and discussed, as well as inconsistencies between relations described and relations constructed based on events observed. Recommendations for experimental and descriptive adjustments are offered. Although this analysis may or may not have any bearing on the metacontingency enterprise, it may serve as a template for conducting interbehavioral analyses of activities in other enterprises, if not more analyses of the metacontingency enterprise.

Behavior analysts have long been concerned with understanding and describing cultural events (Glenn, 1986; Hayes & Fryling, 2009; Skinner, 1948, 1961). Much of the culturobehavioral scientific enterprise has focused on metacontingencies (Glenn, 1986; Glenn et al., 2016), hereafter referred to as the “metacontingency enterprise.” There has been an inconsistent consensus on the definition of metacontingency (Glenn et al., 2016; Zilio, 2019), the level of analysis at which metacontingencies operate (Houmanfar & Rodriguez, 2006; Krispin, 2016; Tourinho, 2013), and characteristics of metacontingencies (Tourinho, 2013). In light of these areas of uncertainty, Glenn et al. (2016) formally defined metacontingency and related concepts for the explicit purpose of facilitating functional analyses of cultural events. Given disagreement regarding the extent to which experimental analyses actually demonstrate metacontingencies, cultural selection, or metacontingent control (Baia & Sampaio, 2019; Hunter, 2012; Krispin, 2016; Tourinho, 2013; Velasco et al., 2017; Zilio, 2019), an analysis of the metacontingency enterprise may be useful for constructing and refining procedures, analyses, and descriptions that facilitate the isolation of events of study and constructing relations among them relevant to the enterprise.

An interbehavioral analysis of scientific system building may contribute to the progressive evolution of experimental analyses concerning metacontingencies. Any given scientific enterprise can be characterized by the activities and products of the scientific workers of that enterprise (Kantor, 1953, 1970). Description and experimentation are primary activities, both for scientific systems in general and the metacontingency enterprise in particular. An interbehavioral approach agrees with Glenn et al. (2016) that descriptions of events of study should foster experimental procedures, analyses, and interpretation. However, relations between other activities and products—like specifying pertinent levels of analysis and properties of events of study—are also important to consider. An analysis that considers relations between description and experimentation may help pinpoint discrepancies and ambiguities between events of study and how those events are studied that can be remediated by altering practices under the auspice of the metacontingency enterprise.

The aim of this article is to provide an interbehavioral analysis on how descriptions of relations among events of interest participate in experimentation within the metacontingency enterprise. The purpose is not to provide rationale for the actual existence of metacontingencies, nor is it to elevate one scientific enterprise over another. No scientific enterprise is more or less valid than any other (Kantor, 1958; Parrott, 1983). An enterprise’s only standard is itself, the congruence of scientific activities that compose the enterprise (Hayes et al., 2009; Kantor, 1945, 1958), although significance across enterprises can be assessed (Kantor, 1958). However, the metacontingency enterprise is not only a distinct enterprise but one that is built on constructs from the operant-based behavior-analytic enterprise. In the behavior-analytic enterprise, operant behavior is considered to be maintained through contingencies of reinforcement. In the metacontingency enterprise, operant contingencies themselves are thought to be part of the unit selected within metacontingencies. Because demonstrating cultural selection occurring through metacontingencies involves observation of operant behavior, metacontingency experiments should demonstrate events and relations that are both distinct from and inclusive of those described by operant constructs to be exclusive to metacontingency constructs. Thus, the extent to which experiments demonstrate events and relations that are exclusively described by metacontingency constructs—and cannot be adequately described in terms of related constructs like macrocontingencies or coordinated operant contingencies—is assessed in this analysis. As such, this article may be relevant for those gauging the compatibility of the metacontingency construct with the behavior-analytic enterprise at large, but cohesiveness will not be extensively weighed here. Rather, this article will focus on features of description and experimentation that may be adjusted to increase correspondence between relations among events described and those investigated, measured, and reported.

Analyses of Scientific Enterprises

Before analyzing the metacontingency enterprise, a description of the general features and orientation of the analysis is warranted. A scientific enterprise consists of a myriad of activities related in assorted ways (Kantor, 1953). Of particular importance to the present discussion are contacts between scientific workers and events of study. Scientific workers are those individuals interbehaving in ways that constitute a scientific enterprise, such as observing events of study, experimenting, and theorizing. Events of study1 consist of those that are the focus of scientific activities of a particular enterprise. Understanding how scientific workers contact events of study and construct relations between those events are pivotal to understanding how a scientific enterprise operates and for identifying points at which an enterprise is inconsistent or concerned with events other than those of interest. This may be relevant for the metacontingency enterprise, considering that scientific work within the metacontingency enterprise concerns events that are intertwined with behavioral events that typically are described without reference to metacontingencies.

All contacts between scientific workers and events of study constitute psychological events. Psychological events are analytical constructs that describe particular evolutions of a continuous, ongoing whole organism–environment interaction,2 as depicted in Fig. 1. Organisms are always responding with respect to stimulation of stimulus objects, not because of such stimulation (Kantor, 1924; Muñoz-Blanco & Hayes, 2017). In light of this and the observation that responding and stimulus objects can have multiple functions (Kantor, 1958), stimulus–response functions mutually coordinated through a medium of contact participate in psychological events as a single unit rather than as distinct entities (Muñoz-Blanco & Hayes, 2013). Stimulus–response functions occur within a setting consisting of both organismic factors and stimuli with various functions (Kantor, 1924). Considering that organism–environment interactions are constantly evolving, functions of stimuli and responses and their organization are established across interactions (Kantor, 1958). As such, psychological events can be characterized by the following factors: (1) stimulus functions, (2) response functions, (3) medium of contact, (4) setting factors, and (5) interbehavioral history (Hayes & Fryling, 2018; Kantor, 1958).

Fig. 1.

Generic Psychological Event

Any contact between an organism and a stimulus object requires a physical medium of contact. Media of contact include light, airwaves, and other physical substrates by which stimulation may be coordinated with responding (Kantor, 1924). When contact occurs between respective functions of an organism and a present stimulus through a physical medium, that contact is direct. Indirect contact assumes contact with a stimulus object by virtue of direct contact with another stimulus. Most experiments involve indirect contact with events distinct from but related to events of study that produce permanent products. For example, experimenters concerned with lever pressing in free operant procedures typically track discrete responses through a computerized event record in lieu of watching organisms press levers. Although many recording procedures allow for tracking events in real time, scientific workers often only contact events of study indirectly through stimulus contacts in which multiple events are compounded (e.g., total number of responses in a session).

Whenever events of study are contacted indirectly, stimulus substitution is fundamental to the account. “Stimulus substitution” refers to a stimulus acquiring functions of another (Delgado & Hayes, 2014; Kantor, 1924; Parrott, 1984). Stimuli readily acquire functions of other stimuli through particular evolutions of organism–environment interactions,3 such as in referential episodes (Kantor, 1977), conditional discrimination procedures (Delgado & Hayes, 2007; Hayes, 1992; Hayes et al., 2001; Sidman, 1994; Sidman & Tailby, 1982), and respondent conditioning procedures (Delgado & Hayes, 2014; Valverde et al., 2009). Stimuli can acquire both formal and arbitrary functions of other stimuli (Kantor, 1924) and those functions can be perceptual (i.e., contacting stimulus objects, e.g., seeing or hearing stimulus objects) or consummatory (i.e., interbehaving with respect to contacted stimulus objects; Kantor, 1958). When stimuli acquire arbitrary functions, those functions are said to be verbal or conventional (Delgado & Hayes, 2007; Parrott, 1984). Viewed in this way, objects are never actually “contacted indirectly”—stimulus objects have just acquired functions of other stimulus objects.

The degree to which verbal or linguistic events involve substitutional stimulus functions varies across events. Kantor (1977) identified two major classes of linguistic events: referential and symbolic events. Referential events can be described in terms of bistimulation where a response function of either a referor (i.e., a speaker) or a referee (i.e., a listener) is coordinated with two stimulus functions, one of an adjustment stimulus (i.e., a referent) and one of an auxiliary stimulus (i.e., a referee or referor). In these kind of events, a referor refers a referee to a referent, which may or may not involve any substitutional stimulus functions (e.g., reference to a past event). Symbolic events refer to events in which response functions are coordinated with arbitrary substitutional stimulus functions that, in turn, produce implicit stimulation with functions that subsequent response functions are coordinated with (e.g., recognizing $\sqrt{4}$ as equal to 2 and subsequently responding with respect to 2, to borrow an example from Kantor, 1977). Given the exclusivity and arbitrariness of symbolic stimulus functions, symbolic interbehavior always involves substitutional stimulus functions. Symbolic events are often portrayed as linear sequences, but they also may be considered in terms of embedded stimulus functions in which the actualization of certain stimulus–response functions constitute setting factors for subsequent interactions.

Both referential and symbolic activity are paramount within the metacontingency enterprise. A substantial majority of scientific work within the metacontingency enterprise constitutes constructing descriptions of relations among events, performing experiments to observe events and analyze relations among them, and reading descriptions of theoretical constructs and experiments concerning such relations. When a scientific worker describes an experiment, they usually do so for a particular readership. In that sense, describing experiments constitutes referential activity in which a referor describes a referent (i.e., an experiment) to a specific audience of referees, although that audience may only participate substitutionally in the writing of such descriptions. This activity necessarily involves the production of stimulus objects (i.e., descriptions) that a readership will symbolically interact with (e.g., data points as instances of behavior). Considering that description and experimentation intertwine within scientific activities in various and complex ways, differentiating between description and experimentation is essential to analyzing relations between them.

Description

Description refers to both creating descriptions and stimulus objects that are products of such activity. Descriptions as stimulus objects are ideally verbal constructs with partial substitutional functions (i.e., substitution resulting in a stimulus acquiring some—but not all—functions of another; see Delgado & Hayes, 2007) with respect to events of study and relations among them. Well-refined descriptions exclusively apply to events and relations they describe; they should not acquire or retain substitutional functions of other events or relations (Kantor, 1953). In the metacontingency enterprise, descriptions of metacontingencies and related constructs should specify events and relations that are different from other contingency constructs. Technical conventionality in the metacontingency or any other enterprise is progressively constructed as descriptions of events and their relations are refined, expanded, transformed, adopted, and abandoned. Because most scientific workers do not directly contact every event that the enterprise is concerned with, definitions, laws, principles, graphical representations, categorizations, and other descriptions (Kantor, 1953) become the primary means by which they interbehave with events of an enterprise.

The reliance of substitutional contact with events of study within an enterprise poses the immediate issue of what substitutional functions are acquired by descriptions. Substitution requires close spatiotemporal proximity or formal similarity of stimulus–response functions (Hayes, 1992), not stimulus objects themselves, in order for one stimulus object to acquire functions of another (Kantor, 1977). Given that descriptions cannot readily acquire functions of events when scientific workers do not directly contact those events, a scientific enterprise must engender other activities by which descriptions can acquire such functions. Scientific workers use examples for this purpose. Examples consist of stimulus objects in a particular organization by which descriptions of events can acquire substitutional functions. The effectiveness of an example depends on stimulus functions acquired by an example through an individual’s interbehavioral history of interacting with stimulus objects. As an example of an example, consider reinforcement. Reinforcement, defined as an increase in or maintenance of the probability of behavior due to reinforcing consequences contingent on that behavior, may be illustrated with a description of a rat pressing a lever at a particular rate because pressing the lever has been reinforced by food. Like other constructs that describe relations among events, reinforcement can acquire stimulus functions necessary for general principles or laws by establishing hierarchical, conceptual relations (Delgado & Hayes, 2007) through multiple examples.

Another way in which descriptions acquire substitutional functions is through metaphor. Novel events and relations are often likened to other events and relations (Kantor, 1953). This may be useful for initial orientation, but an enterprise should ultimately abandon describing events as other events to refine exclusive descriptions devoid of extraenterprisal functions. If metaphorical extensions are retained in descriptions of events, problematic stimulus–response functions may interfere with scientific activity. Metaphorical descriptions of events may orient scientific workers to investigate relations that cannot be shown among particular events, or they may lead to describing relations or features of events that have not actually been observed (Delgado, 2012). However, an enterprise may benefit (i.e., increase prolonged productivity) from maintaining such descriptions. Early comparisons between reinforcement and natural selection (Skinner, 1953, 1966, 1981) have been fundamental for integrating the behavior-analytic enterprise into evolutionary sciences (Brown & Hendy, 2001; Wilson et al., 2014). They may also engender effective communication between scientific workers of different enterprises that might not otherwise be aware of the similarity between their events of study. How metaphorical descriptions participate in the evolution of an enterprise and its sustained productivity depends on functions acquired by such descriptions.

Other less effective activities also allow descriptions to acquire functions. Juxtaposing a description with a less technical description may be helpful, but doing so calls into question the utility of having two descriptions. A description may specify that an event belongs to a class, but that class still needs to be differentiated from other classes. Even when such differentiation occurs, constructing classes of events amounts to constructing hierarchical relations among constructs, not events themselves. A description may be augmented by further description or mathematical operations concerning how events are related to other events, but such an orientation may not lead to observation of those events. Not only may each of these activities hinder other scientific activities but none alone bring scientific workers into direct contact with events of study. Experimentation allows for direct contact with events, and descriptions should orient experimental activity toward events and relations of interest. In the metacontingency enterprise, descriptions of the metacontingency construct should orient experimental activity toward events and relations unique to that construct.

Descriptions in the Metacontingency Enterprise

All scientific systems and their various activities—including description—involve adherence to preanalytic assumptions and orientations (Kantor, 1953) that can be described. A thorough analysis of relations between assumptions and descriptions of events and relations of interest is outside the scope of the current analysis, but understanding relations between description and experimentation of events is informed by acknowledging relevant presuppositions. Considered as an enterprise built in part from the behavior-analytic enterprise, certain orientations common within the behavior-analytic enterprise are likely to be conventional within the metacontingency enterprise. These include selectionist (e.g., Catania, 1995; Skinner, 1981), contextualistic (e.g., Hayes et al., 1988; Morris, 1988), monistic (e.g., Baum, 2011), and single-subject (e.g., Sidman, 1960) approaches to scientific activities. Although many of these orientations are only implicitly derived from the products of scientific work on metacontingencies, they are discussed at length in several articles on theory and experimentation concerning metacontingencies (Glenn, 2004, 2010; Hayes & Houmanfar, 2004; Houmanfar & Rodriguez, 2006). Just as many of these orientations are not wholly accepted within the behavior-analytic enterprise (e.g., Hayes & Fryling, 2019; Morris, 1997), consensus on philosophical and theoretical assumptions in the metacontingency enterprise has not been formally solidified. Because an analysis of an enterprise should be conducted in accordance with its own logical and conventional standards (Kantor, 1953), the current analysis is conducted from and with respect to these orientations.

Although there are several descriptions of metacontingencies and related constructs that the current analysis could consider, those included in Glenn et al.’s (2016) article titled, “Toward consistent terminology in a behaviorist approach to cultural analysis,” will serve as a benchmark for two reasons. First, the article represents a concerted effort by leading metacontingency theoreticians and experimenters to agree upon terms pivotal to the metacontingency enterprise, including “metacontingency” itself, to construct, “a functional approach to the analysis of cultural level phenomena” (Glenn et al., 2016, p. 25). Second, not only is their article commonly cited by experimenters studying cultural selection through metacontingencies when stipulating experimental aims (Soares et al., 2018; Soares et al., 2019; Velasco et al., 2017), but descriptions provided of key events and relations are similar to, inclusive of, refine, and build upon older descriptions (e.g., Glenn, 1986, 1988, 1991, 2004, 2010; Glenn & Malott, 2004; Houmanfar & Rodriguez, 2006; Houmanfar et al., 2010; Tourinho, 2013) that are commonly cited in several older experiments (e.g., Costa et al., 2012; Hunter, 2012; Morford & Cihon, 2013; Ortu et al., 2012; Sampaio et al., 2013; Smith et al., 2011; Tadaiesky & Tourinho, 2012). These descriptions may be elaborated (e.g., Houmanfar et al., 2010), but most descriptions of metacontingency experiments specify experimental goals, procedures, and analyses in terms epitomized by Glenn et al.’s (2016) article. There are other notable descriptions of metacontingencies (Baia & Sampaio, 2019; Krispin, 2016), some of which will be discussed as relevant in subsequent sections.

It should be reiterated that descriptions of events are distinct from events observed (Kantor, 1953). Although Glenn’s (1986) original formulation of the metacontingency construct was reportedly based in part upon contacts with both real and hypothetical communities like Walden Two (Skinner, 1948), it was also constructed in a system with other descriptions of behavior change and theoretical constructs. The definition of “metacontingency” has consistently transformed since its introduction (Zilio, 2019), but detailing this transformation should not be seen as a critique of the construct. Descriptions are refined and restructured within a scientific system through various activities, such as experimentation, commentary, and discussion. Chronological variations of descriptions of key relations among events will be given in the current analysis not solely to demonstrate that transformations have occurred but to demonstrate what metacontingencies are—and what they are not—for the sake of constructing experiments of unambiguous events and relations. The definition of metacontingency given by Glenn et al. (2016) is as follows: “A contingent relation between 1) recurring interlocking behavioral contingencies having an aggregate product and 2) selecting environmental events or conditions” (p. 13).

Figure 2 illustrates a metacontingency. Although similar to an operant contingency, the metacontingency construct has acquired institutional functions (i.e., functions shared across individuals and established under group auspices; Kantor, 1982) that orient subsequent interactions toward selection processes and other relations among events distinct from operant accounts. Whereas operant contingencies select operant behavior, metacontingencies select culturants (i.e., IBCs and their aggregate products). Whereas the process by which operant behavior is selected by reinforcing events is referred to as reinforcement, the process by which culturants are selected by selecting events is referred to as cultural selection. In other words, descriptions of metacontingencies should acquire substitutional functions regarding events and relations differentiated from those of operant contingency constructs.

Fig. 2.

A Metacontingency. Note. IBCs = interlocking behavioral contingencies; AP = aggregate product; SE = selecting event; R = operant behavior; S^D = discriminative stimulus/event; S^R+ = reinforcing stimulus/event. The solid rectangle encapsulate IBCs, whereas the dotted rectangle encapsulates a culturant. Single-arrows denote dependency relations. Dots denote evocative or discriminative functions. The ellipsis suggests that IBCs can involve more than two individuals. Note that the IBCs represented here resemble a verbal episode between two individuals, P1 and P2, described by Skinner (1957) with similar nomenclature. These IBCs only constitute one example—not all IBCs involve such relations

Definitions for component constructs within the metacontingency construct can be found elsewhere. Glenn (2004) defined “interlocking behavioral contingencies” or IBCs, as “operant contingencies in which the behavior of two or more people functions as environmental events for the behavior of the others” (p. 144). IBCs do not simply describe operant behavior of multiple individuals but operant behavior of multiple individuals that is functionally related to each other. As such, demonstrating IBCs requires showing changes in the operant behavior of one individual correlated with and contingent on changes in the environment produced by the operant behavior of another. Although it has not been definitively addressed within the enterprise, the metaphorical extension of “culturant” from “operant” (Glenn et al., 2016; Hunter, 2012), earlier writings on metacontingencies (Glenn, 1986, 1988), and the assertion that, “IBCs themselves are made up of interlocking contingencies of reinforcement in which the local behavior of participants is directly reinforced” (Glenn et al., 2016, p. 13) suggest that behavioral events are conceptualized from a molecular perspective within the metacontingency enterprise.

IBCs are similar to—but not the same as—what Skinner referred to as “interlocking systems” in his analyses of social episodes (Skinner,1953, p. 201), “interlocking contingencies” in verbal episodes (Skinner, 1957, p. 432) and elsewhere (Skinner, 1958) in the sense that IBCs not only require substitutional orientation to other constructs—like reinforcing events and discriminative stimuli—that describe how the behavior of multiple individuals can be functionally related but to events dependent on the totality of such interactions. Unlike interlocking systems or contingencies, IBCs are thought to produce aggregate products that together are selected as a singular unit by events external to IBCs, leading to the reoccurrence of those IBCs and the reproduction of specific aggregate products (Glenn et al., 2016). Hunter (2012) offers a definition of “aggregate products”: “The effect on the environment produced by an IBC is an aggregate of the interlocking behavior of all participants, named the aggregate product” (p. 44; italics in original). As Houmanfar and Rodrigues (2006) highlight, aggregate products were described by Glenn (2004) without defining them as such. In describing how IBCs produce aggregate products by using an example in which a meal is prepared at a restaurant, Glenn (2004) states that: “[The meal] is not the cumulative effect of their individual behaviors. It is the outcome of their interrelated behavior” (p. 145). Here, “interrelated behavior” refers to functionally related operant behavior necessary to produce a particular environmental alteration (i.e., IBCs). Although a meal is a tangible, physical outcome of IBCs, aggregate products are not required to be so by Hunter’s (2012) definition. An aggregate product must only be an event dependent on IBCs. Thus, IBCs are commonly considered to be assumable by observation of aggregate products (e.g., Glenn et al., 2016; Ortu et al., 2012; Vichi et al., 2009), leading some to suggest that classes of IBCs can be defined by the aggregate products they produce (Baia & Sampaio, 2019). Given that behavior can constitute stimulation for the behavior of other individuals, aggregate products can amount to no more than stimulation produced by individuals behaving with respect to one another. Thus, aggregate products can include a large range of outcomes limited only by the environments in which they occur (Vasconcelos & Todorov, 2015). The nearly endless variability this construct inheres likely has contributed to the prevalence of juxtaposing the term “aggregate product” with examples—like making meals—to foster the acquisition of orienting stimulus functions toward a myriad of events with common, abstract properties.

The term “culturant” has referred to both IBCs and their aggregate products, as defined by Glenn et al. (2016), as well as “IBCs selected by a metacontingency” when the term was first coined by Hunter (2012, p. 44), who considers a metacontingency to be “The contingency between occurrences of the IBC and resulting aggregate products” (p. 44). This is an important distinction when considering the constant and rapid evolution of metacontingency descriptions. Hunter’s (2012) conceptualization of metacontingencies is largely based on earlier conceptualizations by Glenn (1988, 1991, 2004). Earlier conceptualizations were constructed as metaphorical extensions of two- and three-term operant contingency constructs (Glenn, 2004; Houmanfar & Rodrigues, 2006), which are themselves in part metaphorical extensions of Darwinian selection (Delgado, 2012; Skinner, 1981) and psychological reflex (Skinner, 1930, 1931) constructs. Descriptions of metacontingencies have since transformed (Zilio, 2019), but it is clear that selecting events or conditions were not always distinguished from aggregate products. For example, metacontingencies are described by Glenn (2004) as “[addressing] evolution by selection when the lineages that evolve are not the recurring acts of individuals . . . but rather are recurring interlocking behavioral contingencies (IBCs) that function as an integrated unit and result in an outcome that affects the probability of future recurrences of the IBCs” (p. 144). That same year, Glenn and Malott (2004) cast metacontingencies within organizational settings as “[having] three components: interlocking behavioral contingencies, their aggregate product, and their receiving system. The receiving system is the recipient of the aggregate product and thus functions as the selecting environment of the interlocking behavioral contingencies” (p. 100). Glenn and Malott’s (2004) description of metacontingencies is more similar to Glenn et al.’s (2016) with respect to distinguishing selecting events from aggregate products. Selecting events are also referred to as cultural consequences, as described by Vichi et al. (2009): “Metacontingencies are defined as interlocking behavioral contingencies (IBCs) that produce an aggregate effect (that could not be produced otherwise) on which the action of an external environment is contingent. The action of the external environment is called a cultural consequence” (p. 42). Tadaiesky and Tourinho (2012) explicitly described cultural consequences in terms of selective functions: “an environment change that selects the IBCs and their aggregate product” (p. 134).

To state that earlier descriptions of metacontingencies did not orient attention toward discrete selecting events is not to say that certain events did not have selective functions. Regarding descriptions in Glenn (2004) and earlier works (Glenn, 1988, 1991), dependent outcomes of IBCs were considered selective, similar to how reinforcing events select operant behavior. For example, Glenn (2004) states that “both automatic outcomes and socially mediated outcomes that depend on the features of the automatic outcome” (p. 146) can select IBCs. This description of metacontingencies is different from the definition provided by Glenn et al. (2016) in which selecting events select for IBCs and their aggregate products on the basis of interactions between aggregate products and events external to IBCs. By adopting and refining Glenn and Malott’s (2004) metacontingency construct to describe cultural selection in organizational settings, Glenn et al. (2016) suggest that metacontingencies require a form of mediation not required to reinforce behavior (i.e., behavior does not necessarily have to produce an environmental alteration in order for reinforcement to occur; Vaughan & Michael, 1982). Arranging descriptions so that metacontingencies acquire stimulus functions of descriptions of reinforcement by analogy—which has been an institutional practice (e.g., Glenn, 1986, 2004; Glenn & Malott, 2004; Houmanfar & Rodrigues, 2006)—may have been initially useful for orienting scientific workers toward events of interest but is somewhat misleading when selecting events are considered distinct from aggregate products and do not necessarily have reinforcing functions. However, when reinforcing events are thought to be contingent on behavior that produces environmental alterations, Glenn et al.’s (2016) suggestion that selecting events are contingent on IBCs that produce aggregate products is more analogous.

When scientific workers recognize seemingly incompatible, contradictory stimulus functions acquired by descriptions, those descriptions are often refined and alternative descriptions are constructed for similar but distinct relations among events. Presumably toward this end, Glenn et al. (2016) describe relations that, “might be considered as metacontingencies” (p. 15), such as when IBCs are maintained by aggregate products that function as both selecting and reinforcing events, similar to Glenn’s (2004) conceptualization. Baia and Sampaio (2019) have also addressed this issue by differentiating descriptive culturants (i.e., classes of IBCs that produce specific aggregate products) from functional culturants (i.e., IBCs and dependent aggregate products selected by selecting events).

In this vein and given ambiguity in early descriptions of metacontingencies (Glenn, 1986, 1988), metacontingencies have also been distinguished from macrocontingencies. Glenn (2004) defines macrocontingency as “the relation between a cultural practice and the aggregate sum of consequences of the macrobehavior constituting the practice” (p. 142). A more succinct definition is offered by Glenn et al. (2016): “[A] relation between 1) operant behavior governed by individual contingencies and/or IBCs governed by metacontingencies and 2) a cumulative effect of social significance” (p. 19).

An illustration of a macrocontingency is given in Fig. 3. Macrocontingencies involve multiple individuals behaving with consequential similarity (i.e., their operant behavior contributes to a cumulative effect). This may or may not involve individuals behaving within IBCs, which can also produce cumulative effects. Although most examples of macrocontingencies involve individuals behaving in the same way (Glenn, 2004; Glenn et al., 2016), their behavior is only related by shared contribution to a cumulative effect.

Fig. 3.

A Macrocontingency. Note. The arrow outside the box denotes a dependency relation between the operant behavior of multiple individuals and a cumulative effect produced by their behavior that does not have reinforcing properties. IBCs are not accounted for in this figure, but they can also produce cumulative effects

Major differences between metacontingencies and macrocontingencies as described by Glenn et al. (2016) include the role of selection. Selection is not thought to occur through macrocontingencies as it does through metacontingencies. Macrocontingencies consist of cumulative effects produced by behavior controlled by reinforcement, IBCs controlled by cultural selection, or some combination of the two. Events that contribute to the production of cumulative effects are selected across individuals or IBCs independently; neither operant behavior nor IBCs are selected by the cumulative effects they contribute to (Glenn, 2004). In other words, macrocontingencies only involve dependency relations between operant behavior, IBCs, and cumulative effects whereas metacontingencies involve functional relations between culturants and selecting events. Whereas dependency relations specify the conditionality of one event occurring given the occurrence of another, functional relations specify, “properties of a stimulus or response event with respect to other(s)” (Ribes-Iñesta, 1997, p. 6274). When extended to cultural events, this difference between macrocontingencies and metacontingencies carries important implications for experimentally examining metacontingency events and relations, as it emphasizes the importance of demonstrating both dependency and functional relations between culturants and selecting events.

Metacontingency constructs should also retain institutional orientating stimulus functions toward events different from those of operant contingency constructs. In metacontingencies, operant contingencies must be interlocking so that environment alterations dependent on organisms’ operant behavior are functionally related to the behavior of other organisms. Similar relations have been described using group contingency constructs (Cariveau et al., 2020; Schmitt, 1984) through which selection of operant behavior across individuals occurs contingent on reinforcing events dependent on the behavior of multiple individuals. Some, like Hunter (2012) and Glenn et al. (2016), have pointed out that relations among events described by group contingencies may be interpreted in terms of metacontingencies, but they can also be described in terms of coordinated operant contingencies. Coordinated operant contingencies describe relations in which reinforcing events are dependent on multiple environmental alterations produced by operant behavior. Unlike group contingency constructs, which highlight the participation of multiple individuals, coordinated operant contingency constructs emphasize more generally that reinforcing events are contingent on multiple environmental alterations that can acquire discriminative and reinforcing properties.5 Behavior can be partitioned in terms of discrete responses that produce specific environmental alterations to isolate various functional and dependency relations in accordance with analytical goals, although behavior may always be considered to be abstracted from larger patterns (Baum, 2012; Rachlin, 1992). Thus, events that maintain other events in one setting may only acquire or retain reinforcing functions when correlated with other events (e.g., gaining money is reinforcing because other events are conditional on using money).

Coordinated operant contingencies can take on many configurations, a few of which are shown in Fig. 4. The top two panels show group contingency variants where environmental alterations produced by the operant behavior of individuals do (middle panel) or do not (top panel) necessarily function as discriminative stimuli for the behavior of others. They also demonstrate that contact with the same or different stimulus objects or events can be discriminative of reinforcing events contingent on behavior. The model in the middle panel emphasizes the complexity of situations in which reinforcing events could be produced by only one organism but multiple organisms take part in producing them. Thus, coordinated operant contingencies can be used to describe evolving circumstances as different functional and dependency relations are established, given affordances of previously established relations and other factors.

Fig. 4.

Molecular Coordinated Operant Contingencies. Note. Dotted arrows denote correlations between responding and environmental alterations. The group contingency in the top panel depicts reinforcing events contingent on the operant behavior of three individuals that are not interlocking. The group contingency in the middle panel depicts reinforcing events contingent on the operant behavior of two individuals that are interlocking. The contingency in the bottom panel depicts reinforcing events contingent on two environmental alterations produced by the operant behavior of one individual. It is important to note that, in some settings, events may only have “reinforcing” functions because they are discriminative of other events

Some necessary requirements of metacontingencies—such as multiple individuals behaving with respect to one another—are somewhat arbitrary when considered in terms of coordinated operant contingencies without considering other factors. At a minimum, coordinated operant contingencies stipulate that reinforcing events are contingent on multiple environmental alterations. Thus, coordinated operant contingencies can be used to describe relations that do or do not involve multiple individuals.6 Imagine that three different environmental alterations—A, B, and C—are required to occur in alphabetical order to produce reinforcing outcomes. Three individuals may each respond so that when one individual produces alteration A, A is discriminative for another individual producing B, which then functions as the same for the other producing C. However, one individual may produce A, B, and C, resulting in the same reinforcing events, albeit only for the behavior of one person. From a coordinated operant contingency perspective, the particular configuration of dependency and functional relations that describes events is related to other factors, such as the time to produce each necessary environmental alteration, differential environmental contacts, the number of individuals present, instructions delivered, other contingencies individuals are behaving with respect to, how and when rewards are contacted, and previous contacts with events sharing properties with the current circumstance. When two or more individuals participate in the production of specific reinforcing events, reinforcement coordinates patterns of behavior across multiple individuals. When only one individual participates, reinforcement coordinates patterns of behavior across one individual. Different behavioral patterns may be observed when dependency relations are manipulated in circumstances involving different numbers of individuals, but in all cases those changes can be described in terms of behavioral processes, according to coordinated operant contingency models.

How coordinated operant contingencies should be interpreted and constructed is in part based on preanalytic assumptions. Dependency and functional relations describe relations among events, although those relations can be conceptualized in different ways. Although there are many different molecular (e.g., Hale & Shimp, 1975; Michael, 2004a) and molar (e.g., Baum, 2012; Rachlin, 1992; Timberlake, 1980) approaches to understanding behavior, differences between molecular and molar perspectives largely concern whether reinforcement involves a class of behavior reinforced by contiguous, immediate consequences or patterns of behavior—extended across time—that are reinforced on the basis of correlation with patterns of reinforcing events, respectively (Baum, 2002; Rachlin, 2013; although see Ribes-Iñesta, 2018, for an alternative view). Thus, contingencies can be thought to describe relations by which classes of behavior immediately change or environmental regularity with respect to which patterns of events can become correlated. From either orientation, relations between events necessary to observe in order to construct molecular or molar descriptions can be described in terms of dependency and functional relations.

Reinforcing events of coordinated operant contingencies are similar to selecting events proposed by Glenn et al. (2016) with respect to dependency relations (i.e., events contingent on multiple environmental alterations) but not functional relations (i.e., selection of operant behavior rather than culturants). This is not to say that functional relations between operant responses of multiple individuals cannot be established in a coordinated operant contingency framework; the establishment of discriminative stimulus functions has long been recognized as an outcome of reinforcement (Skinner, 1938). This is to say, though, that such relations are accounted for by operant processes without relying on additional processes occurring on higher levels of selection, emergent or otherwise (Houmanfar et al., 2010). In terms of coordinated operant contingencies, the emergence of a cultural unit of analysis—functional relations between operant behavior across multiple individuals—can be considered in terms of social interactions maintained by reinforcing events contingent on multiple environmental alterations produced by multiple individuals, among other factors.

Although it is difficult to describe the model in the top panel of Fig. 4 in terms of metacontingencies because there are no functional relations depicted directly between the behavior of multiple individuals (e.g., one person’s behavior exerting stimulus control over another’s), the coordinated operant contingency depicted in the middle panel of Fig. 4 may be thought to constitute a certain type of metacontingency. As noted above, Glenn et al. (2016) describe other contingencies that may be considered metacontingencies. In one variation, selecting events also constitute reinforcing events for behavior that composes IBCs but are differentiated from aggregate products. This description is problematic on multiple grounds with respect to constructing a description with stimulus functions that orient scientific workers toward exclusive events and relations. If selecting events can have reinforcing properties, delimiting between the two is largely cosmetic except to imply a difference in the number of events necessary to produce outcomes and how relations of events are described. One may contend that selecting events can have dual functions in which they both reinforce behavior within IBCs and increase the probability or rate of occurrences of culturants (Glenn et al., 2016). However, functional relations that are established between responses across individuals can be described through behavioral processes; processes occurring at a different level of selection are not necessary to account for any dual functions observed because functional relations are established through reinforcement. Selecting events with reinforcing functions may be considered to select both culturants and operant behavior, but they may also be considered to reinforce operant behavior and, in doing so, establish functional relations among responses across multiple individuals. As outlined in Fig. 4, the middle panel does not demonstrate a metacontingency because only reinforcing events are contingent on behavior.

Glenn et al. (2016) also suggest another possible metacontingency variant in which aggregate products function as selecting events with reinforcing functions. To foster the acquisition of functions for their description, the authors outline an example in which several people work together to build a puzzle. In addition to the same exclusivity issues described above, this metacontingency variant does not adhere to the formal definition they provide. Aggregate products are a part of the selected unit (i.e., culturant) and thus cannot also be what selects the selected unit within a metacontingency without appealing to a form of automatic cultural selection analogous to automatic reinforcement. This issue may be circumvented if individuals whose operant behavior composes IBCs interact with aggregate products after producing them (e.g., members of a group recognize that they have completed a puzzle rather than just completing a puzzle). In either case, however, such interactions may simply constitute reinforcing events for the behavior of each group member. Although the authors state that selecting events within metacontingencies have been shown to maintain culturants, “in the absence of direct reinforcement for participant behavior” (p. 13), descriptions of metacontingencies are only delimitative and exclusive when behavior composing IBCs is maintained by reinforcing events that are distinct from selecting events, which should be able to maintain culturants whereas other events maintain operant behavior according to Glenn et al. (2016).

As previously stated, the purpose of distinguishing different descriptions of relations among events is not to discredit any single description. Scientific enterprises are progressive, and precision with respect to descriptions of relations among events is necessary to establish descriptions with exclusive stimulus functions with respect to particular events and relations. Transformations toward more precise descriptions are indicative of a productive enterprise. In doing so, descriptions are constructed that not only specify dependency and functional relations but differences between other events and relations. In the case of the metacontingency enterprise, Glenn et al.’s (2016) description of the metacontingency construct is distinct from both earlier formulations and related constructs, in part because it is juxtaposed by other descriptions of other relations among events. Such refinement and contrast is advantageous to an enterprise, because it fosters experimental contact between scientific workers and events of interest that in turn can be described.

Experimentation

Before discussing how descriptions and experimentation are related in the metacontingency enterprise, attention should be given to how experimentation is conceptualized from an interbehavioral perspective. The primary means by which scientific workers contact events of interest—whether directly or indirectly by reading descriptions—is through experimentation. Experimentation is an activity or reaction system in which circumstances are arranged to observe particular events so relations among them can be constructed. In behavioral science, manipulation is the defining feature of experimentation (Sidman, 1960). Causal response–reinforcing event relations are largely realized by demonstrating behavior change relative to environmental change (Skinner, 1947). Noncausal relations involve observation of the evolution of an ongoing interaction as participating factors are rearranged, modified, included, and excluded (Kantor, 1958). From an interbehavioral perspective, one event (e.g., a stimulus) does not cause another event to change or occur (e.g., a response). Functions participate in events—they do not relate events to themselves.

Experiments necessitate functions of events and relations that are acquired by descriptions in order to construct means by which they may be observed. This is not to say that experiments are based on particular stimulus objects but rather that substitutional functions acquired by descriptions participate in experimentation. An experiment may be devised to observe particular events without being able to describe relations among them at the onset. Unlike in deductive manipulative experimentation where correspondence is gauged between relations among events constructed prior to experimentation and relations constructed based on observed events, inductive manipulative experimentation involves orientation toward dimensions by which events may be related that have been described prior to investigation but specific relations of interest are unknown. In both cases, constructing circumstances to describe relations among events requires substitutional orientation toward events that should be observed. Descriptions can facilitate experiments with respect to this end and may induce such activity.

More than just a means of direct contact with events of study, experiments also provide a strong basis for substitutional contact. Experiments—like any other activity—can be described (Kantor, 1953). Experimental descriptions are often highly detailed and refined composites in which precise and fine-grained descriptions of participating factors are organized. As such, experiments can have multiple functional purposes for an enterprise outside of the rationales for which experiments are conducted (Sidman, 1960). Experiments can bring functions of description into proximity with functions of observed events and constructed relations among them. They also provide an empirical basis for which events and relations can be thoroughly described by detailing methodologies in which they are investigated.

Description and Experimentation as Participating Factors

Scientific activities are historical and, as such, permit a complex array of ways in which descriptions can participate in experimentation. Figure 5 illustrates a simplified coordination of adjustment and auxiliary7 stimulus functions involved in experimentation. Prior to experimenting with respect to particular events of study, scientific workers directly contact descriptions that have acquired substitutional functions pertaining to such events. Within experimental settings in which scientific workers arrange circumstances with respect to particular adjustment stimulus functions (i.e., those of events of study), interactions with descriptions are themselves contacted substitutionally as auxiliary stimulus functions. In other words, although an experimenter may not directly contact written or vocalized descriptions of events within experimental settings, stimulus functions of such descriptions are nonetheless substitutionally present during experimental activity and serve to orient activity toward particular events and relations that may also have or acquire such functions. As a complex, partially symbolic auxiliary reaction system, experimentation may involve many substitutional functions of various descriptions present at different points of experimental activity (e.g., conceptualization, planning, manipulation, observation).

Fig. 5.

Participation of Description in Experimentation as an Auxiliary Stimulus Function. Note. Solid rectangles denote scientific workers or events of study. Dotted rectangles denote substitutional stimulus functions. Solid bidirectional arrows denote functional contact whereas the dotted bidirectional arrow denotes dual contact with formal and substitutional functions of events of study. The large solid black line segregates different interactions, because the scientific worker must first contact descriptions of events for such an interaction to be substitutionally present in experimental situations. Note that the psychological events represented here are simplified but could be more elaborate (see Fig. 1)

Stimulus functions of descriptions participate in experimentation in at least four ways. This is not by any means an exhaustive list of participations, nor does it encapsulate the complexity of such interactions. The extent to which any activity participates in another is determined in part by the enterprise of interest, but several participations are common across enterprises and especially relevant for the metacontingency enterprise. These include (1) specificity (i.e., descriptions of events and relations facilitate orientation toward particular events and relations); (2) levels of analysis (i.e., characterizations of subject matter facilitate orientation toward particular types of events and relations); (3) types of analysis (i.e., descriptions of events and relations facilitate how events are analyzed); and (4) procedures (i.e., descriptions of events and relations facilitate how circumstances are arranged to observe particular events). These participations are largely teleological, which is to say that the stimulus functions of description largely pertain to facilitating orientation toward particular events using methods that allow for constructing relations among them through experimentation. Events observed should correspond to descriptions of those events (Hayes, 1993).

Specificity participations suggest a particular type of rigidity that contributes to the continuation of investigative activities. Not only do descriptions acquire stimulus functions of events of interest, events of interest also acquire arbitrary stimulus functions. In the case of studying behavioral events, functions established between scientific workers and particular evolutions of continuous, ongoing whole organism–environment interactions determine the boundaries of events observed. For example, recognizing “a lever press” as an event when one observes a change in the level of a lever produced by organismic activity is a function acquired by particular evolutions of interactions that share similar properties. Observation of evolutions with similar properties allows scientific workers to contact larger events—such as an increase in the rate of reinforced responding—but only by virtue of responding with respect to arbitrary functions acquired by properties of an interaction as if they were discrete and different events. Thus, many “events” that are observed during experimentation are constructs in the sense that they are observed on the basis of relating an event to itself as different events instead of recognizing past events as previous evolutions or features of such participating in the current evolution of an event (i.e. a continuous, ongoing organism–environment interaction). This is similar to what Baum (2002) suggests in saying that responses are abstracted from patterns of activity, although larger patterns of activity are also susceptible to acquiring functions that orient how events are observed, isolated, organized, and related to other events. Logical issues do not necessarily arise by this type of scientific activity unless factors of events—including interactional history—are considered to cause current evolutions. However, once functions have been established, they persist across experimental settings that pertain to the same subject matter. This poses a significant risk for the metacontingency enterprise. Considering that events that should be recognized through experimentation were identified in nonexperimental settings in accordance with constructs by workers interested in different events, cultural events studied by metacontingency workers may acquire functions that they may have not otherwise acquired if metacontingency constructs had been constructed through more inductive activity.

For each type of participation of description in experimentation there is somewhat of an inverse participation of experimentation in description. Given that experiments involve contact with events of study and can themselves be described, these participations are more reactionary than teleological in that descriptions can acquire stimulus functions through experimentation. In addition, contacts with experimental events may participate in describing events and relations, whether referentially (e.g., describing to a reader what occurred in an experiment) or symbolically (e.g., codifying a mathematical relation between variables given experimental observations). These participations can be described in terms of (1) artifice (i.e., experimenting informs the degree to which described events and relations can be contacted); (2) breadth (i.e., experimenting informs the extent to which events can be described in relation to other events); (3) data (i.e., experimenting brings scientific workers into contact with events that can be codified as data points or other representations); and (4) precision (i.e., experimenting informs the generality of factor participation across events). In other words, experimenting and contacting experimental findings participate in both describing events and relations and when descriptions acquire functions. In turn, future experiments can become more useful, precise, and produce more reliable results when known properties and relations among events accounted for and other events and relations are controlled for. This may produce rigidity in an enterprise with respect to methodology and interpretations that do not conform to the enterprise. Doubt of findings that are not congruent with descriptions of events based on extensive experimentation can be a useful deterrent against transformations of descriptions that have evolved without precise contact with events and relations of study. Although rigidity may deter creative orientations within an enterprise to events of interest, it is also defensive with respect to establishing stimulus functions of descriptions that are incoherent with the enterprise or otherwise lacking.

Logical analyses of either set of these participations within an enterprise with respect to conventions of that enterprise may illuminate enterprisal inconsistencies that can be addressed by altering descriptions or experimental activities. Given that (1) the metacontingency construct was not originally an interpretation of experimental activity; (2) events and relations observed through metacontingency experiments may possibly be described in terms of coordinated operant contingencies or macrocontingencies; and (3) how experimentation participates in alterations of descriptions necessarily involves descriptions already adopted by an enterprise, the current analysis will focus on the participation of description in metacontingency experimentation. First, specificity will be addressed by distinguishing differences in relations among events that compose metacontingencies, macrocontingencies, and coordinated operant contingencies in experimental settings. Second, based on events and relations described, levels and types of analyses appropriate for investigating cultural selection will be considered. Third, procedures that can be utilized to allow for metacontingency events to be observed and analyzed on suitable levels of analysis will be outlined. Contacting such analyses of these participations—on part of the reader—will constitute a partially symbolic situation (Kantor, 1977) in which metacontingency experiments will be analyzed with respect to descriptions of relations among events such experiments are constructed to observe.

Participations of Description in Metacontingency Experiments

Specificity

Before outlining events and relations pertinent to investigating cultural selection occurring through metacontingencies, attention should be drawn to the extent events can be observed and relations can be constructed. As briefly described above, Baia and Sampaio (2019) suggest that the culturant should be delimited as the unit of analysis of metacontingency experiments and metacontingencies should be regarded as procedures used to study culturants. They state that metacontingencies should, “refer to the arrangement of relations between classes of IBCs (culturants) and [cultural consequences]” (p. 210). Conceptualizing contingencies in this way is useful because it allows one to avoid ill-fated attempts at searching for constructs among events. From this perspective, metacontingencies and operant contingencies are differentiated in terms of dependency relations. In metacontingencies, selecting events are made contingent on aggregate products that require responses of multiple individuals whereas, in operant contingencies, reinforcing events are made contingent on the behavior of a single individual. Although this is a valid approach, contingencies will not be characterized in this way in this analysis for two reasons. First, this is not how Glenn et al. (2016) described metacontingencies, and their analysis is based off of their constructs. Second, descriptions of metacontingencies do not only describe dependency relations but also functional relations. Not only are metacontingencies defined in terms of some events (i.e., selecting events) being contingent on other events (i.e., the production of aggregate products), but they are also defined in terms of functional relations. With respect to metacontingencies, both culturants and selecting events are functionally defined in terms of the other by the process of cultural selection. A culturant cannot be described if the probability or rate of occurrences of that culturant are not shown to increase or maintain given contingent selecting events. Likewise, selecting events cannot be considered selective of a culturant if occurrences of that culturant are not shown to be sensitive to changes in selecting events. Until such relations are demonstrated, events observed are only potential occurrences of a culturant and potential selecting events. The same can be said of operant behavior, which also should be demonstrated in metacontingency experiments considering that IBCs involve operant behavior across multiple individuals. For these reasons, metacontingencies will be considered in terms of relations among events instead of procedures.

In a similar vein, some have argued that interlocking behavior rather than IBCs should be considered to be selected by selecting events (Hayes & Houmanfar, 2004; Houmanfar & Rodriguez, 2006; Houmanfar et al., 2010). They argue that contingencies, as constructs, cannot be selected, but behavior that is functionally related to the behavior of others can. This distinction has been adopted by some metacontingency models (Houmanfar & Rodriguez, 2006; Houmanfar et al., 2010), but this terminology was not adopted by Glenn et al. (2016), nor is it commonly used throughout the enterprise. As such, IBCs will be discussed in the following description of metacontingency events and relations rather than interlocking behaviors.

At this point, events and relations that should be observed to demonstrate cultural selection—and thus metacontingencies—in experimental settings can be outlined, as shown in Table 1.

Table 1.

Events and Relations Composing a Metacontingency in Experimental Settings

Cultural selection of culturant	Event 1. Increase in or maintenance of probability (> 0) or rate (> 0) of potential occurrences of culturant
	Events 2. Potential selecting events
	Relation 1. Dependence of potential selecting events on potential occurrences of culturant
	Relation 2. Conditionality of increase in or maintenance of probability or rate of occurrences of culturant given dependent selecting events
(a) Probability or (b) rate of potential occurrences of culturant	Events 1. Potential occurrences of culturant
	Relation 1a. Occurrences across opportunities
	Relation 1b. Occurrences per unit of time
Potential occurrences of culturant	Events 1. Interlocking behavioral contingencies
	Relation 1. Dependence of aggregate products on interlocking behavioral contingencies (provides basis for delimiting occurrences)
Interlocking behavioral contingencies	Events 1. Operant behavior (across n individuals; n > 1)
	Relations 1 (> 0). Determinable probability or rate of occurrences of one individual’s operant behavior given environmental alterations produced by operant behavior of another individual (relation is reflexive; requires operant behavior of multiple individuals composing the group)
Reinforcement of operant behavior	Event 1. Increase in or maintenance of probability (> 0) or rate (> 0) of potential occurrences of operant behavior
	Events 2. Potential reinforcing events
	Relation 1. Dependence of potential reinforcing events on potential occurrences of operant behavior
	Relation 2. Conditionality of increase in or maintenance of probability or rate of occurrences of operant behavior given dependent reinforcing events
(a) Probability or (b) rate of potential occurrences of operant behavior	Events 1. Potential occurrences of operant behavior
	Relation 1a. Occurrences across opportunities
	Relation 1b. Occurrences per unit of time
Potential occurrences of operant behavior1	Event 1. Whole organism-environment interaction
	Relation 1. Dependence of environmental alterations on whole organism-environment interaction (provides basis by which behavior can be divided into occurrences)

¹ Although this may be more typically described in terms of operant responses, this description is used to highlight that (1) patterns of responses may constitute a single unit and (2) operant behavior is measured by environmental change it produces, whether mechanical or just stimulational. However, it is worth noting that operant behavior can also be measured in terms of time allocation (Baum & Rachlin, 1969) rather than probability or rate of occurrences.

As Table 1 illustrates, demonstrating cultural selection of a culturant through a metacontingency requires showing conditionality of an increase in or maintenance of the probability or rate of occurrences of a culturant given a particular dependence of selecting events on such occurrences (i.e., a contingency relation between selecting events and occurrences of a culturant) that describes the probability or rate at which selecting events occur (e.g., one selecting event contingent on three occurrences of a culturant, one selecting event contingent on the first occurrence after 10 min since the last selecting event). Analyses and procedures that allow for this conditionality to be asserted should be integrated into experimentation as appropriate given experimental aims and previous findings. As selecting events should select culturants, occurrences of culturants should only reliably occur when correlated with contingent selecting events; they should not maintain in the absence of selecting events. How altering properties of selecting events affects the probability or rate of occurrences of culturants is an empirical matter, but—given the proposed selective function of selecting events—decreasing the probability or rate of selecting events by virtue of altering dependency relations between selecting events and occurrences of a culturant should decrease the probability or rate of occurrences of that culturant, given consistency of other factors.

In order to observe cultural selection, specific events and relations should be demonstrated in order to delimit occurrences of culturants from just operant behavior and selecting events from reinforcing events occurring within IBCs. These events and relations differ from those composing macrocontingencies and coordinated operant contingencies. For the sake of brevity, events and relations referenced in Table 2 can be found in Table 1. Relations among events specified by macrocontingency and coordinated operant contingency constructs are described in Table 2.

Table 2.

Events and Relations Composing other Contingency Constructs in Experimental Settings

Events and Relations Composing a Macrocontingency
Production of cumulative effect	Events 1. Some combination of operant behavior across x individuals and IBCs of y groups (x + y = n; n > 1)
	Events 2. Cumulative effect contributions
	Event 3. Cumulative effect
	Relation 1. Dependence of cumulative effect contributions on IBCs and/or operant behavior
	Relation 2. Dependence of cumulative effect on cumulative effect contributions
Events and Relations Composing a Coordinated Operant Contingency
Coordinated reinforcement of operant behavior	Event 1. Increase in or maintenance of probability (> 0) or rate (> 0) of potential occurrences of operant behavior (across n individuals; n > 0)
	Events 2. Potential reinforcing events
	Relation 1. Dependence of potential reinforcing events on multiple environmental alterations produced by potential occurrences of operant behavior
	Relation 2. Conditionality of increase in or maintenance of probability or rate of occurrences of operant behavior given dependent reinforcing events
	Relations 3 (≥ 0). Determinable probability or rate of occurrences of one individual’s operant behavior given environmental alterations produced by operant behavior of the same or different individual(s). Only necessary to demonstrate interlocking behavioral contingencies in which environmental alterations are produced by two or more individuals or a pattern of behavior in which an individual responds with respect to environmental alterations produced by their own behavior

See Table 1 for events and relations composing culturants and operant behavior, as applicable.

Although the events and relations composing these various types of contingencies are distinct, they may be observed in conjunction with one another. Coordinated operant contingencies may describe relations in IBCs that produce aggregate products selected for within metacontingencies. Metacontingencies and coordinated operant contingencies can both influence behavior that contributes to a cumulative effect within a macrocontingency. Experimental procedures should be constructed to allow for confident delimitations of events composing discrete relations considering similarities between them. Glenn et al. (2016) suggest that, in some situations, events may be describable in terms of both metacontingencies and operant contingencies simultaneously given that selecting events can have reinforcing functions. In experimental settings, however, such situations should be avoided considering that IBCs should be selected along with aggregate products by events contingent on them, not events within them.

Levels of Analysis

Delimiting the constitution of “behavior” raises concern for the subject matter of the metacontingency enterprise. Enterprise-wide consensus on the level of analysis of cultural events is contested. As indicated above, whereas some contend that metacontingencies operate at a cultural or sociological level where functional relations among responses are selected by environmental events (Glenn, 2010), others have suggested—as previously stated—that interlocking behavior should be considered the unit of selection (Hayes & Houmanfar, 2004; Houmanfar et al., 2010). Whether metacontingencies operate on one level or another, the events and relations outlined above suggest that analyses at both the individual and group level are necessary. Identifying functional relations in which one organism interacts with another is a matter of observation at the psychological level. The behavior of any one organism can be shown to be functionally related to the behavior of another within IBCs, but IBCs cannot be discretely observed when considering only the behavior of one organism. IBCs refer to a specific set of relations between the behavior of multiple organisms that can be selected by environmental events. As such, metacontingency analyses require observation of individual behavior at the psychological level to demonstrate functional relations between the behavior of interacting organisms and observation of relations between individuals’ behavior at the cultural level to demonstrate IBCs unless procedures are arranged to ensure functional relations among responses of multiple individuals and exclude the possibility that selecting events are reinforcing events. Cultural selection certainly involves observation at the cultural level of analysis (e.g., production of aggregate products requires observation of environmental alterations contingent on the operant behavior of multiple individuals), but IBCs should still be analyzed on the psychological level to demonstrate that part of what is supposed to be selected is selected. The subject matter of the metacontingency enterprise is sets of whole organism–environment interactions involving two or more organisms.

Types of Analysis

Types of analysis will differ based on investigative purposes, but demonstration of metacontingency events and relations outlined above requires that certain analyses be performed. These analyses include:

Control of Operant Behavior within IBCs

Differential probabilities or rates of occurrences of operant behavior should be evident across different conditions that are produced by other group members.

Aggregate Products Dependent on IBCs

Different aggregate products should be shown to be dependent on different IBCs, or particular aggregate products should only occur given occurrences of particular IBCs.

Selecting Events Contingent on Aggregate Products

Selecting events should be shown to be contingent on particular aggregate products; particular selecting events should be shown to occur given the production of particular aggregate products.

Establishing a Culturant

Occurrences of a culturant should be shown to increase in probability or rate as a function of selecting events. At a minimum, this involves a comparison in the rate or probability of occurrences of a culturant prior to and after sufficient exposure to a condition defined in part by the dependence of selecting events on occurrences of that culturant.

Maintaining a Culturant

Given a stable probability or rate of selecting events defined by the dependence relation between selecting events and occurrences of a culturant, the rate or probability of occurrences of that culturant should be consistent after sufficient exposure.

Conditionality of a Culturant

Prior to or after demonstrating an increase in or maintenance of the probability or rate of occurrences of a culturant, the dependence of selecting events on occurrences of that culturant should be altered to demonstrate sensitivity of culturants to changes in the rate or selective properties of selecting events. Increasing the rate or selective properties of selecting events should increase the probability or rate of occurrences of a culturant whereas reducing the rate or selective properties of selecting events should decrease the probability or rate of occurrences of a culturant, although conditions may restrict how high or low the probability or rate of occurrences of a culturant may change. Conditions should persist until the probability or rate of occurrences of a culturant are shown to be consistent. In the absence of other contingencies that favor behavioral interactions that differ from IBCs required to produce target selecting events, an established culturant may continue to persist at the same probability or rate despite altering the dependence of selecting events on occurrences of a culturant. If this occurs and variability in observed occurrences of culturants does not increase, demonstrating differential probabilities or rates of occurrences of two or more different culturants may be necessary to demonstrate metacontingent control, or the dependence of selecting events on occurrences of a culturant should be altered again.

Not all dependency relations need to be demonstrated by formal analyses given experimental procedures that necessitate such relations. In this vein, some authors have argued that IBCs can be assumed by observation of aggregate products (Baia & Sampaio, 2019; Costa et al., 2012; Glenn et al., 2016). This may or may not be a valid claim depending on how a functional relation is defined. If a functional relation is defined as a relation among events or relations among events in which changes in one produce changes in another, then IBCs may be assumed under certain conditions (e.g., selecting events only consistently occur at a high probability when they are contingent on IBCs in which one individual’s response determines how another individual should respond). However, if a functional relation is defined in terms of behaving with respect to acquired or natural properties of stimuli, as suggested by Fryling and Hayes (2011) and Ribes-Iñesta (1997), assuming IBCs is problematic. A major difference between these definitions concerns dependency relations (Fryling & Hayes, 2011). The latter definition does not define function in terms of dependence, although stimuli acquire functions when correlated with other stimuli and responses. Correlations between events are produced by dependency relations among events; manipulating dependency relations between events can—and should—alter behavior given functional relations among those events. Operant responses occur, in part, because they are correlated with events with reinforcing properties. In the absence of such a property, reinforcement would not occur regardless of the contingency in place. Similarly, as evident by verbal behavior (Skinner, 1957), reinforcing events do not need to be mechanically produced by operant behavior (i.e., as behavioral events can be evoked given functional relations among them). Not demonstrating IBCs risks assuming functional relations between the behavior of multiple individuals when only dependency relations with common reinforcing events may be accurately described. In light of Glenn’s (2004) definition of IBCs, both definitions should be upheld. When reinforcing events are made contingent on the behavior of multiple individuals, changes in one individual’s behavior may alter that of others. However, if their responding is not shown to be related in any other way, there is no evidence of IBCs being selected by selecting events given that functional relations between responses of multiple individuals is an essential aspect of the unit selected by selecting events. How individuals respond in such circumstances may determine whether or not responses of other individuals are reinforced, but such an arrangement alone does not ensure that any functional relations (e.g., discriminative functions) form between responses required to produce reinforcing events.

To demonstrate IBCs, several different functional relations between the behavior of group members can be established. For example:

Reinforcing Events

For interactions with environmental alterations produced by individual A’s operant behavior to function as reinforcing events for individual B’s operant behavior, individual B’s operant behavior should increase or maintain if followed by individual A’s operant behavior and individual B’s operant behavior should decrease if individual A’s operant behavior no longer follows it.

Discriminative Stimuli

For environmental alterations produced by individual A’s operant behavior to function as discriminative stimuli for individual B’s operant behavior, individual B’s operant behavior should be more likely to occur in the presence of environmental alterations produced by individual A’s operant behavior than in the absence of such.

S-Deltas

For environmental alterations produced by individual A’s operant behavior to function as stimuli delta for individual B’s operant behavior, individual B’s operant behavior should be less likely to occur in the presence of environmental alterations produced by individual A’s operant behavior than in the presence of discriminative stimuli for individual B’s operant behavior.

This is not an exhaustive list of possible functional relations to examine between the operant behavior of multiple individuals, nor is it inclusive of all the features that can be tested for functional relations identified (Michael, 2004b). Although IBCs require only one such functional relation to be demonstrated, not demonstrating functional relations among the behavior of all group members introduces doubt pertaining to the composition of the group.

Several of these analyses require replication to promote confidence. Occurrences of a culturant may not be functionally related to potential selecting events unless conditionality between the probability or rate of occurrences of a culturant and contingent selecting events is shown. Functional relations are examined through patterns of events, not on an event-by-event basis (Baum, 2011; Skinner, 1930). Replication of these patterns may be accomplished through replication of establishment and maintenance of the probability or rate of occurrences the same culturant within the same group or the same culturant across groups. Analyses should be performed that describe the extent to which these patterns have been replicated across opportunities for such patterns to occur.

Procedures

Consideration of events, relations, and properties of metacontingencies described above warrant specific experimental features. In order to demonstrate IBCs, two or more individuals must be interacting with one another in an experiment. This requires that environments with respect to each individual involved are able to be altered by the behavior of other individuals (i.e., one individual’s operant behavior can produce stimulation for another) prior to the occurrence of selecting events contingent on IBCs and their aggregate products. Interactions can involve contact with environmental alterations dependent on the behavior of group members, which may amount to seeing others behave or other changes in the environment produced by their behavior. If the only relation that is shown is a functional relation between operant behavior and selecting events contingent on the behavior of multiple individuals, experimental events may be more accurately described in terms of coordinated operant contingencies than metacontingencies. This is because the behavior of multiple individuals may only be functionally related to each other by virtue of common reinforcing events. In order to demonstrate functional relations between the operant behavior of multiple individuals composing IBCs, experiments should be arranged so that at least one individual can behave with respect to how another behaves prior to the occurrence of programmed selecting events. Although these relations may—and should—only be maintained when correlated with selecting events contingent on occurrences of a culturant, functional relations composing IBCs should be analytically isolatable.

In the case of each functional relation able to be demonstrated within IBCs, the probability or rate of occurrences of operant behavior should differ under different conditions. To demonstrate differential probability or rate of occurrences of one individual’s behavior given another’s, behavior should be observed under different conditions relevant to particular functional relations. Behavior should be observed in each condition until a stable probability or rate of occurrences can be asserted, although stability can be shown across discontinuous conditions. For example, consider a circumstance where selecting events are contingent on two individuals stating numbers in which individual A must say a number 1 greater than a sample presented to them and individual B must say a number 1 greater than what individual A says. If the number presented to individual A is randomized, a consistent rate of selecting events can only be achieved if individual B’s behavior is under stimulus control of individual A’s behavior, given that participant B cannot contact the number presented to participant A. Differential stability of occurrences of behavior can be achieved in either free operant or discrete trial preparations, as is best suited for observing events of interest and other experimental goals.

Demonstrating IBCs by showing how environmental alterations produced by one individual’s behavior reinforce another’s behavior is potentially problematic depending on how dependency relations are arranged. Consider IBCs in which contacting individual A’s behavior serves as reinforcing events for individual B’s behavior and the proposed aggregate products of IBCs composed of A and B’s behavior is a relation between them (e.g., A and B’s responses are identical in form). This may constitute a culturant, but these events could also be described in terms of behavior maintained by reinforcing events in which metacontingency constructs are not necessary. For example, reinforcing events may be considered to be chained if A’s behavior is reinforced by terminal reinforcing events (i.e., programmed selecting events) contingent on aggregate products and A’s behavior produces conditioned reinforcing events—with properties acquired due to their correlation with terminal reinforcing events—for B’s behavior. Procedures should be arranged to avoid such relations among events. If environmental alterations produced by A’s behavior are not perceivable by B, such alterations cannot function as reinforcing events for B’s behavior. However, if both A and B can contact programmed selecting events, their behavior may also be considered to be reinforced by the same reinforcing events without needing to appeal to a reinforcement chain.

Certain issues arise if IBCs do not involve reinforcing events distinct from selecting events. If IBCs do not involve reinforcing events that are distinct from selecting events, selecting events may function as reinforcing events for behavior within IBCs. Because reinforcement can endow stimuli with discriminative properties, reinforcement may then account for occurrences of IBCs. For this reason, metacontingency experiments often program discrete individual and group consequences (e.g., Sampaio et al., 2013; Soares et al., 2018). However, even if multiple consequences are programmed for each type of contingency, if their predicted effects would likely produce the same changes in behavior, cultural selection may not be able to be differentiated from reinforcement through coordinated operant contingencies. This may constitute a possible metacontingency variant proposed by Glenn et al. (2016), but a coordinated operant contingency description seems more accurate and productive given that it would account for events and relations observed without having to relate them to another set of parallel relations among events or assume relations among events that were not observed.

Functionally related responses in IBCs are not necessarily contiguous based on spatiotemporal dimensions alone. Temporally extended events may be functionally related through verbal processes (Hayes & Sanford, 2014). Although demonstration of verbal processes are not necessary to demonstrate IBCs, verbal behavior must still be shown to be functionally related to the behavior of others to precisely demonstrate IBCs in which verbal behavior occurs. If participants in metacontingency studies can talk to one another, relations between verbal responses and nonverbal responses—as well as verbal responses and other verbal responses—should not be assumed or IBCs themselves must be assumed and cannot be described with confidence. For example, if selecting events are made contingent on all participants pressing a blue button, someone may instruct the others to always press the blue button. Following instructions may compose IBCs on one trial, but if participants continue to press the blue button across trials in the absence of instructions, their behavior may only be shown to be under control of reinforcing events and may not be functionally related to the behavior of others. Even if one contended that delivering instructions once functioned as a motivating operation with effects extended across trials, the continued occurrence of reinforcing events in the absence of subsequent interactions between participants may hinder any conclusions about functional relations between the instructions delivered and continued operant responding.

Aggregate products must be observable. Aggregate products must only be dependent on IBCs; they can amount to any environmental alteration produced by multiple individuals behaving with respect to one another. Stimulus objects structured, arranged, and/or transformed by more than one individual (e.g., a meal at a restaurant; Glenn & Malott, 2004) or a particular configuration of responses (i.e., aggregate responding) can constitute aggregate products. If configurations of the same responses or environmental alterations serve as aggregate products (e.g., all participants pressing a blue button), confidence in observation of IBCs and cultural selection become especially important to distinguish metacontingency events from those composing macrocontingencies. A metacontingency in which multiple individuals behaving in the same way is selected by selecting events may not be discriminable from a macrocontingency in which multiple individuals behaving in the same way produces distinct reinforcing events for all responses when both reinforcing and selecting events are programmed to occur (i.e., selecting events may not have selective effects and only aggregate like a cumulative effect). It may also be indistinguishable from a coordinated operant contingency in which the behavior of all individuals is being reinforced by both programmed selecting and reinforcing events. In other words, although different events may be programmed to be contingent on the behavior of one or multiple individuals, all such events may be described as reinforcing events.

Few necessary properties are required of environmental events to specify them as potential selecting events. The event should temporally follow IBCs and their aggregate products. In order to place the event within a contingency, the event should be observable and recordable regardless of any relation. Only by demonstrating a contingent relation between a culturant and environmental events that change the probability or rate of occurrences of that culturant can those environmental events be classified as selecting within a metacontingency. A change in the probability or rate of occurrences of culturants requires differential behavioral outcomes produced when selecting events are made contingent on different aggregate products, the dependence of selecting events on occurrences of a culturant is altered, or properties of selecting events are adjusted. Selection is relative with respect to other outcomes; any selective increase requires a selective decrease. However, as described above, great care should be taken to differentiate selecting events from reinforcing events. If selecting events are shown to have reinforcing functions, they may be described as reinforcing events in terms of coordinated operant contingencies.

Events in at least two conditions should be analyzed: one in which a culturant is established and maintained and another in which it is not or maintained to a lesser degree. Cultural selection should be evident when selecting events that increase or maintain the probability or rate of occurrences of a culturant are contingent on occurrences of those culturants, but demonstrating the function of selecting events requires showing that altering selecting events or their dependence on occurrences of a culturant affects the probability or rate of subsequent occurrences. Once occurrences of a culturant are shown to maintain, metacontingent control may be demonstrated by altering relative parameters of selecting events (e.g., magnitude of selecting events, discriminability of selecting events), altering the dependence of selecting events on culturants (e.g., decreasing the probability or rate of selecting events), or by removing the contingent selecting events. Metacontingent control may also be demonstrated by showing differential maintenance of probabilities or rates of occurrences of different culturants (i.e., those that produce different aggregate products or upon which different selecting events are contingent) under different conditions in which dependency relations between such occurrences and selecting events are altered. Conditionality may be assumed by virtue of previous experimental findings, but any demonstration of change in the probability or rate of occurrences of a culturant produced by change in selecting events or their dependence on such occurrences increases confidence in the functionality of selecting events, especially given potential confounding variables (e.g., delivery of instructions, maintenance of responding solely by reinforcing events).

Metacontingency Experiments

Several experiments have sought to demonstrate metacontingent control of culturants (or IBCs in earlier studies). The purpose of this section is to assess the degree to which metacontingency experiments isolate events and relations required to demonstrate metacontingencies so that relevant events and relations can be described. Given that observation of metacontingency events and relations are dependent on procedures used and analyses performed, participations of descriptions will not be partitioned as they were in the previous section. Rather, after discussing general properties of metacontingency experiments, this section will be organized by events and relations important for the demonstration of cultural selection of culturants through metacontingencies. Although several studies will be considered and described, linguistic limitations of the first author prohibit an extensive analysis of the many metacontingency experiments described in Spanish and Portuguese. Analyses of these experiments can be found elsewhere (Tourinho, 2013; see Zilio, 2019, for more reviews).

It should be noted that the lexicon adopted here to discuss metacontingency experiments is not how experimenters necessarily describe them. Some experiments discuss procedures, data, and findings in terms consistent with Glenn et al. (2016), but others do not (e.g., Smith et al., 2011; Vasconcelos & Todorov, 2015; Velasco et al., 2017), For consistency with the overall analysis, the terms used to describe these experiments will be those conforming with Glenn et al.’s (2016) definition of metacontingency and related constructs.

General Properties

Metacontingency experiments share several qualities. All metacontingency experiments have involved two or more organisms interacting with one another, with group sizes spanning from two (e.g., Vasconcelos & Todorov, 2015) to nine (e.g., Franceschini et al., 2012) individuals. Human participants—and to a lesser extent, pigeons (Velasco et al., 2017) and fish (de Carvalho et al., 2017)—are used as subjects. Group interactions occur in settings where participants are allowed to talk to one another (e.g., Vichi et al., 2009), although talking is restricted in some experiments (e.g., Hunter, 2012) and allowed to occur conditionally in others (e.g., Soares et al., 2018). Most experiments have incorporated replication of experimental procedures across more than one group, albeit differentially in accordance with events observed (e.g., in cases where a specific control group is warranted).

Single-subject methodologies regularly used in behavior-analytic studies are universally employed to demonstrate experimental control in metacontingency experiments. In most studies, events during discrete trials were observed, although “free culturant” preparations have been devised (Toledo et al., 2015) that involve procedures analogous to those involved in free operant arrangements. Metacontingent control is sought by demonstrating differential occurrence of culturants when contingent selecting events are and are not available (Soares et al., 2019) or available for different culturants (e.g., Ortu et al., 2012) across conditions whereas other individual consequences meant to reinforce behavior are kept constant. In some cases, metacontingent effects are observed by comparisons across groups, especially when investigating the influence of other variables (e.g., Costa et al., 2012). Stability criteria (i.e., consistent production of aggregate products) typically determines length of conditions (e.g., Vasconcelos & Todorov, 2015). As an alternative, the number of trials in or a certain amount of time to complete a particular condition are set a priori to any observations (e.g., Smith et al., 2011). If other requirements are not met, a maximum number of trials in a condition is imposed (e.g., Soares et al., 2018).

Occurrences of Operant Behavior

All relations among events necessary to demonstrate cultural selection through metacontingencies require operant behavior of multiple organisms. As such, metacontingency experiments should be constructed to allow for observation and measurement of participants’ responses. This is largely the case across all metacontingency experiments, although not all relevant responses are measured, recorded, or described in data analyses. For example, in Vichi et al.’s (2009) original metacontingency experiment in which participants placed bets of tokens as a group and rewards were made contingent on certain aggregate products (i.e., whether participants evenly distributed winnings among themselves or not), only the cumulative frequency of the production of aggregate products and the amount of winnings earned and bets made across sessions were analyzed and described using single-subject graphical representations to demonstrate relations between aggregate product production and contingent rewards. Some individual responses may be deduced from reported occurrences of equal and unequal distributions of winnings given experimental requirements (i.e., participants contributing tokens toward group bets, participants adding winnings to their own individual winnings), but other likely responses that potentially participated within IBCs (e.g., vocal speaker–listener interactions between participants) were not analyzed. This approach—in which operant behavior is only described as necessary to specify which aggregate products were observed—is common practice.

In other studies, aggregate products are defined in terms of specific responses of all group members, allowing for individual responses and aggregate products to be analyzed simultaneously. This is most apparent in experiments using iterated prisoner’s dilemma games, or IPDGs (e.g., Costa et al., 2012; Ortu et al., 2012) in which the delivery of points is made contingent on whether all, none, or some proportion of participants choose one of two alternatives (i.e., to cooperate or to defect) on a given trial. When aggregate products are defined in such a way, any analysis provided for the occurrence of aggregate products also functions as an analysis of specific responses that occurred. However, descriptions of such aggregate products do not typically denote how each individual responded, only that certain responses occurred. In situations where selecting events were programmed to occur contingent on all participants responding in the same way (e.g., Ortu et al., 2012), continuity of individual responses is possible to analyze in a limited capacity based on observed aggregate products reported.

Methods for measuring and recording responses differ across experiments. In all cases, responses of significance (i.e., those necessary to determine aggregate products) are either recorded by hand by one or more independent observers (e.g., Soares et al., 2018) or by software in cases where participants work at computer terminals (e.g., Hunter, 2012). Experimenters have video- and audio-recorded interactions between participants (e.g., Smith et al., 2011). Data from such recordings is rarely subjected to formal data analyses, although in many cases researchers provide anecdotal descriptions of exemplary interactions as evidence of IBCs.

IBCs

Although some experiments have involved interactions between participants that were allowed to be so complex that IBCs could only be assumed by aggregate products, dependency relations in other experiments have been structured so that potential selecting events could only occur if all necessary events composing IBCs occurred. This difference is observed most notably when comparing group-decision and individual-decision Wiggins’s matrix procedures. Wiggins’s matrix procedures (WMPs) utilize a matrix with rows and columns that have varying characteristics (e.g., color, shapes in each cell). The matrix is checkered with different symbols across rows and columns. In group-decision WMPs (GWMPs), each trial consists of groups of participants that conjointly select a single row on the matrix (e.g., Franceschini et al., 2012; Vichi et al., 2009). In individual-decision WMPs (IWMPs), each trial consists of participants taking turns selecting rows on the matrix (e.g., Soares et al., 2018; Soares et al., 2019; Tadaiesky & Tourinho, 2012). Whereas IBCs are typically assumed in GWMPs because the behavior producing aggregate products is presumably complex (e.g., equally distributing rewards between participants, joint-decision making), IBCs can be made observable in IWMPs—across trials—by requiring rewards to be delivered contingent on behavior that would only reliably produce rewards consistently across trials if participants’ behavior exerted stimulus control over that of other participants.

In Tadaiesky and Tourinho’s (2012) study, several complementary controls were integrated into experimental designs in order to increase confidence that participants’ operant behavior exerted stimulus control over the behavior of others when target aggregate products were produced. In Experiment 1, participants took turns in which they bet tokens of various colors on selected rows of a Wiggins’s matrix. During Condition B across triads (i.e., groups of three participants), participants could only earn group rewards if (1) all participants selected different color tokens to bet trial to trial and (2) the first participant to bet a token bet one that was the same color as the token bet by the last participant to respond in the previous trial. Participants could only bet tokens of one of three colors, and participants alternated the order they responded in across trials. These reward and response requirements constituted a circumstance in which—to consistently produce rewards across trials—(1) the first participant’s response on a current trial should have been under stimulus control of the last participant’s response on the previous trial and (2) the third participant’s response should have been under stimulus control of the second participant’s response on the current trial to avoid selecting a token of a color incompatible with adhering to both reward requirements on the following trial. Although the second participant’s response on any given trial could have been forced to come under stimulus control of other participants’ behavior by requiring that all participants bet a different color token on each trial,8 rotating the order in which participants responded across trials ensured that responses of all participants were functionally related to some extent across trials if the triad was able to consecutively produce rewards. Multiple sessions were necessary to achieve replicability of aggregate products (i.e., behavior of all triad members on a given trial that met both reward requirements), but consistent aggregate product production was observed across multiple triads. As the authors note, because participants were allowed to freely talk to one another, IBCs were likely more complicated than just involving discriminative stimulus control of one participant’s token bets over another’s. Despite such likely intricacies, reward requirements incorporated ensured that IBCs occurred when consistent production of rewarded aggregate products was observed.

Other IWMP studies demonstrate IBCs to a lesser extent. In a study by Soares et al. (2018), participants in different groups took turns selecting rows on a Wiggins’s matrix. Each participant could maximize their own individual earnings by selecting odd-numbered rows rather than even-numbered rows. However, in some conditions and groups, participants could produce programmed selecting events (e.g., items to be donated to a school) if all participants selected even-numbered rows of different colors. IBCs may have been likely given that all participants needed to select different colored even-numbered rows in order to produce programmed selecting events and they could see one another select rows, but these features do not necessarily imply that responses were controlled by anything other than contingent consequences (i.e., once participants each selected rows of different colors, they may have continued selecting those rows because doing so resulted in rewards). A study by Soares et al. (2019) operated with similar mechanics, although they claim to have demonstrated control over the reproduction of IBCs. It should be noted, though, that in neither of these studies did the authors specifically define IBCs, only culturants and aggregate products.

Sampaio et al. (2013) utilized a novel procedure designed to examine IBCs with high precision. Like Tadaiesky and Tourinho’s study (2012), Sampaio et al.’s experiment incorporated specific procedures and analyses to increase the likelihood of observing stimulus control of participants’ operant behavior over one another’s. In their experiment, triad members selected figures with different numbers of arrows pointing in different directions in the presence of sample figures with some configuration of two or more arrows. Participants could always earn points by selecting a figure with one fewer arrow than the sample shown to them. In some conditions, participants could also earn group bonus points if they selected figures pointing in specific directions relative to those selected by other participants. Because both the order in which participants responded and what figures they selected determined what other participants should select to produce bonuses, IBCs could be assumed by aggregate products observed. To more accurately describe IBCs observed, the authors also provided supplementary evidence (i.e., on the variability of response combinations observed, vocal statements made between participants, and reportedly higher number of verbal interactions when participants could—and did—produce bonuses). Given that only triads in which participants were allowed to talk to one another reliably produced bonuses, IBCs observed likely involved verbal discriminative stimuli.

By allowing participants to freely talk to one another and respond in any order they chose, Sampaio et al. (2013) were also reportedly able to observe leader–follower relations in which one or more participants instructed other group members. Other studies have reported similar findings. Franceschini et al. (2012) reported similar relations observed during a GWMP experiment, noting that, “participants followed the orientation of a ‘leader’” (p. 93) despite providing limited data to justify or explain this claim. Hunter (2012) observed, in a study with IPDG like features, that one participant was more likely to respond first when points were contingent on the behavior of both dyad members but less likely to respond first when points were only contingent on individual responses in one dyad. Hunter’s findings are especially convincing, considering that he did not permit participants to talk to one another; participants could only “follow” one another by behaving in the presence of stimuli presented to them dependent on their partner’s responses. Although Hunter demonstrated a high likelihood of one participant responding after another rather than stability of such responses according to some requirement (i.e., consecutive responses in which one participant responded in a certain way after their partner responded in a certain way), not allowing participants to talk to one another provided more clarity of discriminative relations established between operant responses of dyad partners.

In other studies, specific programmed capacities have allowed IBCs to be partially examined. For example, in a series of IPDG experiments by Ortu et al. (2012), participants could prevent other group members from earning points for a single trial if all other participants “kicked” them. Likewise, any participant could “fine” any other participant in Morford and Cihon’s (2013) IPDG study, resulting in that participant losing points they had accrued. These types of interactions allowed participants to effectively punish the behavior of other participants, but such interactions were only likely to occur in certain circumstances (e.g., when one participant’s response resulted in other participants receiving less points than they would have if the participant had responded in a different way on a given trial). This may have established IBCs in which participants’ behavior was under control of avoidance contingencies maintained by other participants, although the experiments described above did not involve systematic removal of such capacities to better demonstrate functional relations between potentially punishing responses and responses of other participants. Similar interactions were reported by Neves et al. (2012) in a study examining dyads in which participants were shown to reliably return hypothetical resources given to them by their partner if they did not need them to survive until the next round. These types of interactions allow for IBCs to be examined across some but not all trials, only those in specific situations during experiments. It is also worth noting that these types of IBCs do not typically produce aggregate products in themselves and may only be relevant to their production across trials.

The extent to which IBCs are observable in experiments with nonhuman animals (e.g., de Carvalho, 2016; de Carvalho et al., 2018; Velasco et al., 2017) is unclear. Although nonhuman experiments offer interesting insights into the effects of consequences contingent on the behavior of multiple organisms, they do not offer conclusive evidence of IBCs maintained by selecting events. For example, in a study by Velasco et al. (2017), each pigeon in a dyad could earn 3-s access to grain if they pecked an illuminated key 10–12 s after the start of a trial, as well as 4-s access to grain if they pecked an illuminated key directly or diagonally across from their partner—depending on the light of the keys—within 0.5 s of their partner’s peck. The authors were able to demonstrate that pigeons’ pecks could be coordinated under such contingencies, but, as they stated, “the behavior of one pigeon did not have to exert stimulus control over the other’s behavior” (p. 543). As in most metacontingency studies, IBCs cannot be isolated because the behavior of participants did not necessarily produce environmental events with discriminative functions for other participants’ behavior except by virtue of altering the probability of reinforcing events contingent on the behavior of all participants. Although Velasco et al. state that consequences “were socially mediated: The length of time of feeder access depended on what the other subject did” (p. 542), social mediation suggests one organism reinforcing or punishing another’s behavior with respect to their behavior. To reiterate an earlier point, consequences contingent on the behavior of multiple individuals produce situations where one organism’s behavior determines whether another’s behavior is reinforced, but this does not necessarily imply that any organisms are behaving with respect to natural or acquired properties—like discriminative functions acquired through reinforcement—of environmental alterations produced by the behavior of other organisms so that IBCs and their aggregate products may be selected.

Occurrences of Culturants

Although IBCs are not always definitively observed, measured, or reported in metacontingency experiments, the production of aggregate products is almost always readily determinable based on descriptions of analyses provided by experimenters if IBCs are assumed. As noted above, aggregate products typically amount to specific configurations of responses of group members. Such configurations produce stimulation for experimenters to deliver consequences differentially based on what configurations occurred (e.g., Tadaiesky & Tourinho, 2012), or they produce particular automated outcomes when responses are measured by experimental software (e.g., Hunter, 2012). Almost all formal data analyses involve demonstrating the consistency by which aggregate products were produced (e.g., graphs of cumulative production of aggregate products across trials), allowing readers to readily analyze relations between occurrences of specific aggregate products and consequences contingent on their production.

When aggregate products are shown to be consistently produced across trials, culturants can only be assumed if the consistent production of aggregate products necessitates stimulus control of one participant’s operant behavior over another’s. As such, experiments like those by Tadaiesky and Tourinho (2012) and Sampaio et al. (2013) provide the clearest evidence of culturants because potential selecting events were made contingent on participants behaving with respect to how other participants respond. Almost all metacontingency researchers have provided reports of verbal interactions between participants or incorporated features into their experiments that allow for participant’s behavior to come under discriminative control of the behavior of other participants, but these analyses are almost always incomplete considering the myriad of possible interactions that could have occurred. Even when possible interactions that could constitute IBCs are reported, it is not always clear that such interactions were functionally related to the production—or continued production—of aggregate products. For example, many interactions observed in metacontingency experiments reportedly involve one participant instructing others how to respond (e.g., Sampaio et al., 2013; Smith et al., 2011). On one hand, if a participant is told how to respond on one trial and circumstances are kept constant for subsequent trials, additional delivery of instructions may be irrelevant to responding. On the other hand, if a specific instruction applies across trials, a participant may only need to provide instructions once and not repeatedly, raising doubt to whether IBCs reoccur across trials or if behavior is simply rule-governed. Even in the case of Tadaiesky and Tourinho’s (2012) study, IBCs may not have been replicated across some trials given that certain patterns of behavior extending across trials could have been reinforced (e.g., participants may have maintained consistent production of aggregate products by behaving in the same way on a trial as they responded three trials ago).

Cultural Selection

Special attention has been given to differentiating operant behavior controlled by reinforcing events and culturants controlled by selecting events across metacontingency experiments. This is typically sought by either arranging the delivery of programmed individual consequences (that should only maintain the behavior of one individual) and cultural consequences (that should maintain culturants) to (1) be contingent on responses that are incompatible with one another in order to demonstrate that culturants may be maintained by selecting events despite the availability of more reinforcing alternatives that would otherwise maintain different responses or (2) be contingent on multiple participants behaving according to different—but not mutually exclusive—requirements that either do or do not involve stimulus control of one participant’s behavior over another’s. In the first case, participants should make different, incompatible responses (e.g., selecting A rather than B) that can only be maintained by either individual or cultural consequences. In the second case, participants should be able to satisfy a response requirement along one dimension to produce individual consequences (e.g., selecting a variant of A rather than a variant of B) and another dimension to produce cultural consequences (e.g., selecting a variant of A that has not been selected by another participant). Although most studies only utilize one approach, both of these controls were integrated across different groups in an IWMP study conducted by Soares et al. (2018). In their study, all participants could earn more points by selecting odd-numbered rather than even-numbered rows. However, in some groups, participants could also produce programmed cultural consequences if all participants selected even-numbered rows of different colors, whereas in other groups participants could produce cultural consequences if all participants simply selected different colored rows. They found that participants only consistently produced cultural consequences when they could maximize all rewards with the same response (i.e., selecting odd-numbered rows of different colors), suggesting that metacontingent control is not likely to be demonstrated when behavior required to produce selecting events also requires forgoing potent reinforcing events.

Studies utilizing IPDGs have found results that may contradict those found by Soares et al. (2018). In IPDGs, each participant in a group can choose to either cooperate or defect on a given trial to earn points. Points are distributed depending on (1) what a given participant selects and (2) what all other participants select. If behavior is primarily controlled by the most local programmed point contingency, participants should always choose to defect because defecting always produces more points relative to those that could be gained by cooperating on any given trial regardless of what other participants select. In other words, gaining more rather than fewer points on a trial is a consequence that is only dependent on the behavior of one individual. However, participants earn more points for either defecting or cooperating for each other participant that cooperates. If all participants cooperate (i.e., exclusive cooperating), such behavior cannot be said to be controlled by the most local point contingency given that participants are not behaving to maximize points on that trial. Metacontingency experiments typically aim to achieve and maintain the production of exclusive cooperating by all participants—as well as other aggregate products (e.g., exclusive defecting by all participants after having maintained exclusive cooperating)—by programming cultural consequences (e.g., gains or losses of points) to be contingent on aggregate products. Several studies have demonstrated maintenance of exclusive cooperating by all participants despite being able to earn greater individual consequences on any given trial by defecting instead (e.g., Morford & Cihon, 2013; Ortu et al., 2012).

Metacontingency IPDGs differ in how and when selecting events are programmed to occur. In a study by Ortu et al. (2012), market feedback (i.e., additional gains or losses of points) contingent on specific combinations of group member responses adjusted points kept in a separate bank that were dispersed evenly among participants every other trial on average. Point gains or losses for each aggregate product differed across conditions and experiments, but in all cases only one aggregate product produced point gains whereas all others either produced losses or simply no points. Costa et al.’s (2012) preparation was similar, except that (1) market feedback was programmed to only be delivered intermittently based on the aggregate product produced on a given trial rather than accumulate across trials and (2) multiple aggregate products produced gains of points rather than just one. Morford and Cihon’s (2013) market feedback contingency was more similar to Ortu et al.’s (2012) except that market feedback was delivered on every trial in conditions where market feedback occurred. Across experiments in most of these studies, participants were shown to consistently produce aggregate products that maximized market feedback, even if greater programmed individual consequences could be earned by making the alternative response.

Although all metacontingency IPDGs program distinct consequences to function only as reinforcing events and not selecting events, the extent to which selecting events can be identified without reinforcing properties is unclear and seemingly dependent on preanalytic assumptions. Considering that all participants contact rewards contingent on both individual responses and aggregate products, potential selecting events like market feedback may be characterized as reinforcing or punishing events depending on circumstance. If an organism does not respond in a way that produces potential reinforcing events, those events cannot be said to be reinforcing. Several experiments have shown a relatively higher probability of participants defecting in the absence of market feedback contingent on exclusive cooperating than when such market feedback occurred (e.g., Morford & Cihon, 2013; Ortu et al., 2012), but it does not necessarily follow that market feedback did not constitute reinforcing events because, in their absence, participants responded differently. On one hand, if defecting by any single participant increased the probability of subsequent aversive events, such as getting “kicked,” participants’ behavior may have been reinforced by avoiding aversive events. On the other hand, if such punitive events occur, participants may only be able to maximize points across trials by exclusively cooperating. However, during several IPDG experiments (Costa et al., 2012; Morford & Cihon, 2013; Ortu et al., 2012), exclusive cooperating actually produced the most amount of points that could be earned on a given trial in conditions where the highest market feedback gains were contingent on all participants cooperating, although such responding did not maximize rewards on each trial when market feedback was scheduled to occur intermittently (Costa et al., 2012; Ortu et al., 2012). In any case, behavior observed in IPDGs may always be considered to be controlled by temporally extended reinforcing events in which participants behave with respect to one another to maximize points for the group across trials rather than only maximize points for oneself on current trials. This type of molar interpretation has been repeatedly promoted in the behavior-analytic enterprise (Locey & Rachlin, 2012; Rachlin et al., 2000) and suggests that selecting events may only be divorced from their reinforcing properties if one considers reinforcement to be based on contingent relations between responses and immediate outcomes rather than correlations between patterns of events (Baum, 1973; Rachlin, 2013). If this is not the case, it is unclear why (1) market feedback would need to be programmed as selecting events given that maximization of points across trials can be considered selecting events dependent on the behavior of multiple individuals whereas maximizing points on any one trial is only dependent on the behavior of one individual and (2) additional point contingencies that negate the critical difference in IPDG contingencies—when considered from a molar perspective—would be integrated. Given that exclusive cooperating has been shown to be maintained in the absence of market feedback under contrived circumstances (e.g., Locey & Rachlin, 2012; Yi & Rachlin, 2004), there is difficulty in asserting that selecting events in IPDGs are more than just additional, temporally extended reinforcing events contingent on the interrelated behavior of multiple individuals.

There is also difficulty asserting that IBCs are evident in IPDGs. Metacontingency IPDGs are typically arranged so that participants can contact stimuli discriminative of other participants’ behavior, which may allow for participants to exert stimulus control over the behavior of others. However, point delivery may maintain the behavior of all participants regardless of whether they contact such potential discriminative stimuli, especially in experiments where larger losses of points are correlated with a larger proportion of participants choosing to defect (e.g., Morford & Cihon, 2013; Ortu et al., 2012). If behavior in IPDGs can be maintained by events that are discriminative of access to more rather than less points on subsequent trials, other events that occur in IPDGs may have little relevance as to why certain responses are observed. Furthermore, if participants are aware of responses of other participants and how to produce different trial outcomes, they may respond to effectively alter other participants’ behavior through the delivery of points (although this could still happen even if they are not aware). Such behavior may be indicative of IBCs, but if market feedback or other types of selecting events can effectively serve as a means of reinforcing or punishing behavior of participants, those events may be better described as reinforcing or punishing events involved in interactions that extend across trials rather than selecting events that only exert control over events occurring on one trial. Although data on supplementary responses—like those reported by Morford and Cihon (2013) on the probability of switching responses after fining—may provide greater evidence of IBCs, IBCs should be unequivocally demonstrated across the production of all aggregate products. Without doing so, evidence of cultural selection is depreciated.

Conclusion

The purpose of this analysis was to describe how metacontingency constructs and other descriptions participate in experimentation. Toward this end, specific participations of description in experimentation—specificity, levels of analysis, types of analysis, and procedures—were discussed to highlight where inconsistencies between metacontingency constructs and experimental arrangements, findings, analyses, and aims could be identified. Experiments were described with respect to these participations to demonstrate the extent to which relations of interest to the metacontingency enterprise can be used to describe events observed and analyzed during metacontingency experiments. Many other studies could have been described (e.g., Borba et al., 2017; de Carvalho et al., 2017; Smith et al., 2011, Toledo et al., 2015; Vasconcelos & Todorov, 2015; Velasco et al., 2017), and many that have been described here could have been described in more detail. Despite this, additional descriptions do not seem to be necessary to detail discontinuities between descriptions and relations among reported experimental events within the metacontingency enterprise.

Metacontingency experiments typically allow for contact of the following events and relations by both experimenters directly and readers of descriptions of such experiments indirectly:

1. The production of aggregate products

2. Dependence of programmed consequential events on the production of aggregate products

3. Changes in probability or rate of aggregate products or variants of aggregate products correlated with contingent programmed consequential events

The reliability of these events and relations observed suggest that the macrocontingency construct does not describe these experiments well. Macrobehavior—the behavior of multiple individuals controlled by independent sources—should not produce cumulative effects with selective functions. Across most studies considered here, aggregate products were shown to be sensitive to contingent consequential events. Responding in certain groups within Soares et al.’s (2018) study may be described in terms of macrocontingencies considering that, when producing selecting events conflicted with producing reinforcing events, participants simply responded in similar ways to produce reinforcing outcomes. Regardless, their study is still not well described in terms of macrocontingencies because of the absence of any observed or reported cumulative effect of social significance, other than that to the current analysis. Other studies that have specifically examined macrocontingencies (e.g., Borba et al., 2014) utilize procedures markedly distinct from those examined here.

The events and relations readily apparent in these experiments, although important for demonstrating cultural selection, do not encapsulate all events and relations required to distinguish cultural selection occurring through metacontingencies from reinforcement occurring through coordinated operant contingencies. In particular, there is reasonable doubt—based on analysis of the procedures and analyses used in metacontingency experiments—considering the extent to which:

1. IBCs are observable;

2. Selecting events without reinforcing functions can be identified;

3. Selecting events are external to IBCs.

In light of these ambiguities, the extent to which events and relations observed in metacontingency experiments are described more succinctly and accurately by metacontingency constructs than by coordinated operant contingency constructs is questionable. Descriptions are not necessarily mutually exclusive, but more assumptions and allowances must typically be made to describe experimental events observed in terms of metacontingency constructs than coordinated operant contingency constructs.

In most experiments, IBCs are either completely assumed or only partially described due to either insufficient analyses or lack of measurements. Although some experimenters (e.g., Hunter, 2012; Sampaio et al., 2013; Tadaiesky & Tourinho, 2012) have devised experiments to address observation of IBCs, metacontingency theoreticians have commonly suggested that IBCs can be assumed by observation of aggregate products like how lever presses can be assumed by contact with “switch closure[s]” (Glenn et al., 2016, p. 13). This should not be standard practice in experimental settings in which the reproduction of IBCs and their aggregate products is a primary goal. Assuming IBCs by observation of an aggregate product seems to be indicative of conflation between dependency and functional relations. As Figure 4 illustrates, reinforcing events can be contingent on the behavior of multiple individuals without environmental alterations produced by one individual participating as discriminative events for the behavior of others. When a lever depression closes a switch in free operant settings, there is no doubt that an organism pressed the lever because contingencies are arranged so that switch closures are necessarily dependent on lever pressing. Aggregate products may require responses by multiple individuals, but such requirements do not mean that their behavior is functionally related. Aggregate products may not be indicative of IBCs even if necessary environmental alterations occur and, thus, may not actually be aggregate products. Demonstrating functional relations between responses of multiple individuals is an empirical matter that cannot be observed in a single instance of generating an aggregate product. The consistent production of aggregate products may be indicative of IBCs when selecting events are contingent on responding that must adjust across trials dependent on what other participants do, but IBCs may be observed in situations where aggregate products do not necessarily require contact with anything but coordinated reinforcing events. In such cases, IBCs may be better characterized as contingencies that persist through coordinated behavior that produces outcomes discriminative of such contingencies than selected units. Many studies in the behavior-analytic enterprise suggest that functional relations would form between responses across individuals in metacontingency settings given that stimuli that are present when reinforcing events occur typically acquire discriminative functions, but such a process should be demonstrated—not assumed—especially considering that such a process is indicative of reinforcement. Even when IBCs are observed in cases where one or more participants instruct other participants on how to respond or participants respond with respect to stimuli discriminative of how other participants have responded, the extent to which IBCs reoccur is questionable considering that such responding may not persist.

The issue of demonstrating IBCs should be addressed by adjusting experimental procedures. Scientific workers should utilize procedures like those promoted by Sampaio et al. (2013) that necessitate discriminative control of participant’s behavior over that of others in order to consistently produce aggregate products, although some alterations may be constructive in terms of ensuring consistent demonstration of IBCs across trials with greater precision. In their study, participants could not select the same figure every trial to maximize points because they were presented with random sample stimuli across trials that differed in numbers of arrows, forcing them to alter their behavior across trials to maximize points. However, participants could have always selected arrows pointing in the same direction across trials to maximize bonuses. Even if such behavior occurred, participants would still have needed to respond in a certain order to earn bonuses. Their arrangement necessitated observation of IBCs when target aggregate products were shown to be consistently produced, but this was not observed in the case of the triad that was not allowed to talk. As such, a precise description of IBCs in their study required analyses of verbal interactions that, because talking was not restricted in any way when it was allowed to occur, were seemingly cumbersome to conduct and report.

The reliance of consistent aggregate product production on verbal behavior found by Sampaio et al. (2013) and others (e.g., Soares et al., 2018) suggests that IBCs can more precisely be described in circumstances in which participants are not allowed to talk to one another yet promote the acquisition of discriminative functions for discrete, observable, and measurable environmental events produced by behavior. For example, with minimal changes to Sampaio et al.’s procedures, this may be accomplished by (1) reducing the number of figures participants can select from; (2) not allowing participants to see or talk to one another; (3) having participants respond separately at different computers; (4) controlling the order by which participants respond each trial; and (5) presenting every other participant with figures selected by other participants as they respond. If target aggregate products could be produced under such circumstances consistently, an experimenter could isolate stimuli with discriminative functions controlling the behavior of participants (i.e., those presented to participants dependent on what other participants select) and allow for more precise descriptions of IBCs that are more accurately summarized by specific aggregate products. In the absence of such alterations, experimenters should perform more systematic analyses to account for functional relations between participant’s vocal and nonvocal responses if participants are allowed to talk to one another (e.g., analyses in which the functions of vocal responses can be classified). As an alternative, experimenters may also systematically restrict which communicative responses participants are allowed to perform to make analyses of verbal activity more manageable. Whether or not these recommendations are upheld, procedures and analyses developed by Sampaio et al. (2013) and Tadaiesky and Tourinho (2012) should be adopted by other researchers, as they provide clearer evidence of IBCs than are reported elsewhere. Experiments will always be constructed differently in accordance with differences in events and relations of interest, but commonalities among experimental procedures are important for describing meaningful relations among events occurring across experiments (i.e., increasing breath with descriptions of events and relations).

Experiments may more widely utilize procedures that allow for more clear and precise observation of IBCs if relations between IBCs and their aggregate products are described in different ways. It is implicit that if IBCs are defined in terms of individual’s behavior producing environmental alterations that function as environmental events for the behavior of others (Glenn, 2004) and aggregate products are defined as an effect produced by and dependent on multiple individuals behaving within IBCs (Hunter, 2012), an aggregate product should be indicative of (1) multiple changes in environmental circumstances and (2) individuals behaving in different environmental circumstances from one another. An experimental procedure like that used by Sampaio et al. (2013) forces such situations because how one participant should behave to maximize points is dependent on how other participants have behaved. In other situations, such as IPDGs where consequence requirements do not necessitate attending to the behavior of others or stimuli correlated with them, participants may effectively behave in environmental circumstances irrespective of what other participants are doing or only respective by virtue of contact with consequences contingent on all of their behavior. If aggregate products are considered to be terminal environmental alterations dependent not solely on the behavior of multiple individuals but multiple individuals behaving in different, discriminable environmental circumstances produced by the behavior of other individuals, IBCs may be more readily observed in experimental settings if circumstances are arranged to promote such patterns of environmental alterations and analyses are conducted to demonstrate discriminative stimulus control. Likewise, definitions of IBCs may be more precise if they emphasize necessary discriminative functions of environmental alterations produced by behavior within IBCs. Glenn et al. (2016) discuss aggregate products in these ways, but experimenters do not always arrange contingencies or perform analyses accordingly, which may be in part due to the assertion that IBCs can be assumed by an aggregate product.

As an alternative, IBCs may be easier to examine in experimental circumstances if the concept of IBCs is simplified. If IBCs are thought to refer to the behavior of multiple individuals that produces a particular environmental alteration, all metacontingency experiments demonstrate IBCs. However, this characterization would not be productive for the metacontingency enterprise because replication of IBCs would not be sufficiently differentiated from simply reinforcing behavior by events contingent on environmental alterations produced by multiple individuals. Selecting events would simply be selective of certain responses that would not necessarily be related in any other way.

Although how to best observe or define IBCs is a serious issue within the metacontingency enterprise, differentiating between reinforcing and selecting events is arguably a more pressing concern. Considering that IBCs should necessarily involve discriminative stimuli produced by behavior and that stimuli acquire discriminative properties through reinforcement, selecting events may always be described as reinforcing events for all individuals participating in IBCs who contact them. As such, IBCs may simply be a description of discriminated operant behavior reinforced by the same events across multiple individuals. Whenever selecting events cannot be distinguished from reinforcing events, the metacontingency construct is potentially compromised as a description of events observed because events may be more accurately described as single, ongoing IBCs comprising responses reinforced by events contingent on multiple responses. Considering that, “IBCs themselves are made up of interlocking contingencies of reinforcement in which the local behavior of participants is directly reinforced” (Glenn et al., 2016, p. 13), selecting events may be external to IBCs in the sense that they occur contingent on behavior but not in the sense that they occur contingent on IBCs in which behavior is reinforced by other events. Because IBCs should constitute part of the cultural unit selected by selecting events, coordinated operant contingencies may offer a more precise description of more complex and less ambiguous events.

As discussed above, a distinction between selecting and reinforcing events seems to require the preanalytic assumption that reinforcing events can only consist of those immediately occurring after responses on which they are contingent. This excludes a description of metacontingencies when behavior is considered from molar perspectives (e.g., Baum, 1973; Rachlin, 1992), because selecting events—being correlated with responses occurring in IBCs—can always constitute reinforcing events for the behavior within IBCs for individuals that contact them. Although no empirical study can validate a molecular or molar perspective of behavior or invalidate the other, a decidedly molecular view of behavior in the metacontingency enterprise would not only be inconsistent with selecting events having reinforcing functions for patterns of behavior but would also needlessly complicate experimental findings. For example, in Tadaiesky and Tourinho’s (2012) study, IBCs and their aggregate products were shown to reliably reoccur even though only consequences contingent on responses of multiple individuals responding in a linear sequence were delivered. If one was to describe such events as selecting events without reinforcing functions, then it would be unclear if reinforcement was occurring at all and would have to be assumed. Although Glenn et al. (2016) have suggested that the replication of IBCs may be maintained by selecting events in the absence of local reinforcing events, as they claimed was shown in Tadaiesky and Tourinho’s (2012) study, behavioral processes are considered necessary to maintain operant behavior from both molar and molecular perspectives. The metacontingency enterprise—as its own enterprise with its own standards and conventions—does not need to be entirely congruent with the behavior-analytic enterprise at large, but it cannot be congruent at all if behavioral processes are considered unnecessary to maintain operant behavior. If they are considered necessary, selecting events may simply constitute temporally extended reinforcing events that reinforce interrelated patterns of behavior of multiple individuals.

When viewed from a molar perspective, metacontingencies are only distinguished from coordinated operant contingencies based on features that are largely arbitrary when not related to other factors. Culturants require multiple individuals behaving together to produce common outcomes, but coordinated operant contingencies only stipulate that reinforcing events are contingent on multiple environmental alterations produced by patterns of behavior. The arbitrariness of this distinction is clear when considering tasks in metacontingency experiments. For example, in Sampaio et al.’s (2013) study, if the authors had not required that each individual needed to make a response to produce points on a given trial, one individual could have completed the task without help from other participants (i.e., she could have selected certain stimuli in a certain order to produce bonuses). Removing that requirement in their study—as well as most other metacontingency studies—may have produced interactions that would have been poorly described in terms of metacontingencies but still could have been described in terms of coordinated operant contingencies whether or not multiple individuals participated in performing all necessary responses to produce reinforcing events. Integrating certain response requirements may have made it more likely for multiple individuals to participate even if they were not required to do so, such as imposing time requirements on responding, exposing some participants to stimuli discriminative of target responses that other participants were not shown, or requiring certain types of responses that some participants were more likely to emit than others (e.g., responses only probable for participants with particular histories of reinforcement). Systematically investigating the participation of these types of factors may not only make metacontingency experiments more socially valid, but they may also be useful for differentiating interactions characterized by metacontingencies from other interactions. In other words, constructing circumstances in which participants work together to produce specific outcomes even though they are not required to do so may be more important to understanding cultural events than demonstrating differential control of consequences that are at best always correlated with one another and at worst simply different names for events contingent on different numbers of environmental alterations produced by different numbers of organisms. To the extent IBCs have been demonstrated, this has been somewhat observed in IPDGs where participants behave differentially before and after programmed selecting events are introduced (e.g., Ortu et al., 2012), and in particular in those in which the participation of communication was investigated (e.g., Costa et al., 2012) and related studies (e.g., Ribes-Iñesta et al., 2006).

As such, a metacontingency may only constitute one type of coordinated operant contingency that describes certain events and relations involving particular factors (i.e., environmental alterations that cannot be produced by the behavior of only one individual) based on experiments performed thus far. Differentiating classes of contingencies is useful in any behavioral enterprise, as different descriptions orient research toward investigating different relations among events. When contingencies are considered to operate on different levels of selection, though, constructing relations between events may become needlessly complex. For example, in metacontingency experiments in which events constituting leader–follower relations were observed and described (e.g., Hunter, 2012), an important factor for determining the “leader” in each interaction may be sensitivity to temporally extended events. In the behavior-analytic enterprise, this type of sensitivity has been characterized in terms of time horizons (i.e., the temporal window in which events can constitute reinforcing events; Bickel et al., 2006) and quantified using delay discounting measures (e.g., Jones et al., 2009). If selecting events are considered distinct from reinforcing events, a different construct would have to be constructed to measure sensitivity to selecting events, and even more constructs would be needed for describing relations between each type of sensitivity. Although researchers have promoted such approaches based on examining relations between behavioral and cultural events (e.g., Houmanfar et al., 2010), such intermediary analyses are unlikely if interactions do not involve different processes occurring at different levels of analysis.

The metacontingency enterprise may benefit from constructing its own orientation toward behavior and behavioral processes. Given that consensus on molar (e.g., Baum, 1973) or molecular (e.g., Michael, 2004a) views of behavior has not been reached in the behavior-analytic enterprise itself, the metacontingency enterprise cannot rely on constructs from that enterprise to account for descriptions of events and relations pertinent to its own subject matter, which itself is directly informed by how behavior and reinforcement are conceptualized. Although Glenn (1986, 1988) and Houmanfar et al. (2010) promote molecular behavioral orientations, Glenn et al. (2016) seem to describe molar relations among behavioral events, such as in suggesting that selecting events can have reinforcing properties for complex patterns of behavior like playing basketball. Describing these types of assumptions may not only alter descriptions of events within the enterprise (e.g., boundaries for reinforcing and selecting events) but may also orient specific experimental activities.

If a more molar orientation to behavioral events is taken, metacontingency experiments would likely differ significantly from those already conducted. For example, several experiments program selecting and reinforcing events to be contingent on the same responses (Sampaio et al., 2013; Soares et al., 2018; Soares et al., 2019; Tadaiesky & Tourinho, 2012). This poses a challenge for differentiating activities functionally related to different consequences in terms of time allocation (Baum & Rachlin, 1969) considering that the same responses may be abstracted from overlapping activities. Similarly, supposedly different activities in metacontingency experiments (i.e., those controlled by selecting and reinforcing events) may actually be part of the same pattern (e.g., earning points). Choice between different cultural activities (i.e., activities that produce different aggregate products upon which different selecting events are contingent) would likely take a prominent role in the enterprise, which has only been examined in IPDG studies where rewards of different magnitudes were made contingent on different aggregate products (e.g., Ortu et al., 2012). A molar orientation may also change how events and relations are conceptualized and constructed, including metacontingency and coordinated operant contingency constructs themselves. If culturants are considered in terms of activities extended across time, cultural analyses may pivot toward observation of patterns in which multiple individuals take part rather than the replication of selected units abstracted from such patterns.

Regardless of how behavior is conceptualized by the metacontingency enterprise, differentiating between selecting and reinforcing events may be complicated by verbal behavior and processes that are largely uncontrolled in metacontingency experiments. If participants are able to contact selecting events, such events—and events upon which they are dependent—may participate in equivalence relations (Sidman, 2000), relational frames (Hayes et al., 2001), contingency substitution (Ribes-Iñesta, 1991), or some other verbal unit or activity by which properties of events acquire both reinforcing and discriminative functions that extend the boundaries of IBCs. Although verbal relations may be a defining factor of cultural events, Glenn et al. (2016) contend that selecting events do not always have reinforcing properties. This may be shown if experiments are designed so that only specific group members contact selecting events. For example, consider a situation in which four participants sit at separate computers and are not allowed to see or talk to one another as they complete a task. On any given trial, three participants are asked to select either a red, blue, or green button. After all three participants have selected, the fourth participant is shown what the three participants have selected. That participant is then asked if their selections match a selection target (e.g., one red, one blue, and one green selection) or not. Point delivery contingencies for the fourth participants are arranged so that the most points are delivered when they state that the selections match the target when they actually do. After receiving points, the fourth participant is then able to distribute them to the other participants, keep them, and/or put them in a bank to use on later trials. In such a circumstance, that participant could deliver more points to participants who should select the same button again and less or no points to participants who should select different buttons. Although delivering points to the fourth participant could still be considered a reinforcing event for the behavior of the fourth participant, those events could not be described as reinforcing events for the behavior of other participants because they never contact those events. Maintenance of the behavior of any one of the other participants could be described in terms of reinforcement arranged by the fourth participant, but contingencies imposed by the fourth participant to maintain specific behavior of the other participants could not be described without accounting for points delivered to the fourth participant contingent on the behavior of all participants. Even if the three participants selecting colored buttons were interacting verbally with experimental stimuli or each other, they could not directly contact the events or stimuli that would determine which stimuli they should select to maximize points delivered to the fourth participant. In the absence of such controls, selecting events may simply constitute reinforcing and discriminative events that have acquired properties through verbal processes. From a more molar perspective, by not restricting contact of selecting events, selecting events may be characterized as reinforcing events participating in abstract, temporally extended patterns of behavior.

Given that consistent aggregate product production is generally not observed when participants are not allowed to talk to one another, selecting events may in part be distinguished from reinforcing events on the basis of verbal properties. This is not only suggested by groups of nonhuman organisms failing to consistently produce aggregate products that verbally competent human participants seem to produce reliably, such as exclusive cooperating in IPDGs (Green et al., 1995), but by features of IBCs that experimenters have been able to demonstrate clearly. IBCs observed in both Sampaio et al.’s (2013) study and Tadaiesky and Tourinho’s (2012) study can both be characterized in terms of higher-order discriminations in which participants are not only behaving with respect to relations between formal properties of stimuli but arbitrary properties based on the order of responses. For example, behavior in Sampaio et al.’s (2013) study may, if interpreted from a relational frame theory perspective (Hayes et al., 2001), involve frames of coordination (i.e., when participants describe the figures they select), frames of comparison (i.e., when participants select a figure with arrows pointing in a direction relative to those selected by another participant), and deictic frames (i.e., when participants respond according to whether or not another participant has already responded) among others. The behavior of nonhuman organisms has certainly been shown to be organized when potential reinforcing events require responses of multiple individuals (e.g., Velasco et al., 2017), but there is little evidence to suggest that their behavior can even be reinforced by consequences contingent on responses of multiple individuals without directly reinforcing similar responses prior to imposing such contingencies (e.g., Lanza et al., 1982). Although differentiating selecting events from reinforcing events on the basis of their participation in verbal relations would suggest that cultural selection may amount to reinforcing complex discriminations simultaneously across organisms by events discriminative of such interactions, it may also prove to be constructive for understanding determinants of both establishing and maintaining cultural events. Whether or not selecting events are distinguished on this basis, scientific workers in the metacontingency enterprise should agree on a theoretical orientation to verbal processes to effectively understand functional relations involved in IBCs and the proliferation of group activity, especially considering the prevalence of rule-governed behavior reported in metacontingency studies (e.g., Sampaio et al., 2013; Smith et al., 2011).

It seems unnecessarily artificial to suggest that selecting events may be differentiated from reinforcing events based solely on the nature of dependency relations between events. If aggregate products only refer to outcomes of the behavior of multiple individuals “that could not be produced otherwise” (Vichi et al., 2009, p. 42), then selecting events can be differentiated from reinforcing events when they are contingent on different types of outcomes (i.e., those that can only be produced by multiple individuals and those that cannot). However, if aggregate products are described in accordance with Hunter’s (2012) more general definition where they are simply, “the effect on the environment produced by an IBC” (p. 44), then such a distinction between selecting and reinforcing events would not be applicable with respect to all culturants. Glenn et al. (2016) do not offer a definition for aggregate products, but specifying that aggregate products are only outcomes that can be produced by groups seems needlessly restrictive. There may be outcomes that individuals working in groups can produce more effectively, efficiently, or consistently than they could on their own (Glenn, 2004), but individuals may still be able to produce those outcomes. In each of these cases, individuals may work together rather than individually in part because doing so is more reinforcing. Describing how reinforcing events are contingent on multiple environmental alterations is important, but—as the coordinated contingency construct shows—does not require any reclassification of reinforcing events. This is echoed by language used to describe metacontingency procedures, as experimenters typically describe programmed consequences as individual and cultural consequences instead of events that select for operant behavior or culturants, respectively.

Several of these issues may be irrelevant if metacontingencies are considered a procedure, as proposed by Baia and Sampaio (2019), rather than conditions under which a specific, distinct process occurs like Glenn et al. (2016) suggest. If reinforcing and selecting events are considered different types of consequences contingent on responses of either one or more individuals, respectively, most metacontingency experiments can be described in terms of metacontingencies. However, it is not clear how such a conceptualization of metacontingencies is productive toward orienting to relations among events that participate in the reoccurrence of culturants. Given that selecting events will always be correlated with behavior occurring in IBCs, cultural selection may always be considered to summarize reinforcement that maintains behavior across multiple individuals and—as a product of such—establishes discriminative properties among their behavior that may or may not be facilitated by verbal processes (which are also typically based on reinforcement in the behavior-analytic enterprise). When properties of selecting events, such as relative magnitude (e.g., Morford & Cihon, 2013) or intermittency of rewards (e.g., Soares et al., 2019), or requirements for their occurrence like the complexity of discriminations (Tadaiesky & Tourinho, 2012), are shown to be functionally related to the production of aggregate products, such relations always reflect previously established relations observed in experiments on operant behavior. Like the metacontingency construct itself, metacontingency experiments seem not only to be analogous to experiments on operant behavior with respect to conditions manipulated but also produce parallel findings. In some cases, this may simply be the case because selecting events are reinforcing events that can endow correlated stimuli—including responses of people—with discriminative functions. In other cases, relations are likely to be much more complex. That complexity may be well-organized and simplified by metacontingency constructs, but doing so may hinder recognition of important relations that could be targeted to produce cultural change. If metacontingencies are considered a specific class of coordinated operant contingencies involving particular factors, the participation of several relevant factors that would otherwise be obscured by not analyzing relations between individuals’ behavior and group consequences would be emphasized without having to develop terms to describe relations among events that can be described as outcomes and properties of behavioral processes. Considering that producing cultural change—not just understanding it—is a goal of the metacontingency enterprise (Mattaini, 2019), a coordinated operant contingency analysis of cultural events is likely to be just as advantageous as the metacontingency construct, if not more, for the enterprise.

Discussion

Although there are several other implications as described above, describing cultural events in terms of coordinated operant contingencies would not necessarily change the unit of analysis of metacontingency experiments, just how it is described. In the metacontingency enterprise, the unit of analysis has been designated as the culturant (i.e., IBCs that produce a specific aggregate product; Baia & Sampaio, 2019). In operant terms, this unit amounts to functionally related behavior of multiple individuals on which specific reinforcing events are contingent. This unit is not corrupted when described in terms of coordinated operant contingencies; such descriptions only emphasize the complexity of events described and the participation of behavioral processes. As evident by IPDGs and related studies (Hunter, 2012), much of this complexity involves social dynamics established by reinforcing events contingent on the behavior of multiple individuals, but understanding these dynamics is a matter of describing interactions observed on a psychological level that involve individuals contacting outcomes contingent on their behavior. No experiment in the metacontingency enterprise has demonstrated control by selecting events that could not be described as reinforcing events for the interrelated behavior of multiple individuals unless viewed with a particular molecular orientation that stipulates that temporally extended outcomes cannot function as reinforcing events. If the metacontingency enterprise does not maintain that orientation, it may promote rapid development given the incredible amount of activity in behavior-analytic subenterprises pertaining to understanding relations between patterns of behavior and delayed outcomes. Otherwise, scientific workers may continue investigating properties of processes well-established in the behavior-analytic enterprise and developing parallel terms instead of focusing on the unique contribution their unit of analysis affords: complex, evolving social dynamics maintained by reinforcing events contingent on the behavior of multiple individuals. This unit is, of course, a special case of a more general unit (i.e., functionally related behavior that produces specific outcomes), that the enterprise may find useful to study in order to identify factors that contribute to multiple individuals behaving to produce outcomes together when they may not necessarily have to. In other words, shifting focus away from demonstrating cultural selection to conditions under which individuals produce outcomes more effectively, efficiently, or consistently as a group than they can on their own—when they conceivably could actually generate those outcomes—may be progressive for the enterprise.

The current analysis does not discount the metacontingency construct. Metacontingencies, like any other construct, are descriptions of relations among events and not events observed (Fryling & Hayes, 2017). No construct can ever be “observed” through experimentation; events are observed, measured, and related to other events and relations as constructs. Constructs are adopted, transformed, and abandoned by workers within scientific enterprises in accordance with truth criteria of that enterprise. In the behavior-analytic enterprise, constructs are weighed in part by the effective action they afford (Barnes & Roche, 1997; Schoneberger, 2016; Skinner, 1961). Reinforcement, as a construct, has not persisted in the behavior-analytic enterprise because it is more “true” in an absolute, objective sense but because it reliably orients effective behavioral adjustments. As a construct in an enterprise with similar philosophical underpinnings, the scientific utility of the metacontingency construct should also be weighed on a similar basis. As Glenn et al. (2016) state, metacontingency constructs warrant discussion because they are, “useful conceptual tools both in guiding experimental analysis . . . and in understanding and changing everyday cultural-level phenomena” (p. 12).

Given the scale of events metacontingency theorists are generally concerned with, the metacontingency construct has been especially effective in interpreting large-scale organizational and sociological events. Ardila Sánchez et al. (2019) used an elaborated metacontingency model (Houmanfar et al., 2010) to describe complex relations between participations of governmental agencies, private corporations, ecological conditions, and cultural practices in Puerto Rico after Hurricane Maria. Foxall (2015) outlined symmetrical and asymmetrical relations between marketing firms functioning as metacontingencies themselves. Al-Nasser et al. (2019) theorized several loci of influence at the cultural level that could be targeted to prevent honor killings in Jordan. These interpretative accounts exemplify how metacontingency models have been applied to complex relations between groups, cultural practices, and sociological circumstances that are difficult, cumbersome, or impossible to articulate in terms of operant processes. They also, more so than operant-based models that only specify contingencies, bring attention to nonbehavioral factors that are relevant to events of study. In all of these cases, though, dependency and functional relations that are related to behavior on the operant level and may be of interest are obscured by performing analyses at the cultural level of analysis.

There is a substantial jump to be made from events and relations investigated in metacontingency experiments to these interpretative accounts in terms of  (Zilio, 2019). The metacontingency construct seems to engender experimental analyses that are sometimes difficult to describe in terms of operant contingencies because certain procedures and analyses are not incorporated. However, not demonstrating certain events and relations—in particular those composing IBCs—necessarily restricts metacontingency-based interpretations because not all relevant events and relations are demonstrated. This introduces doubt with respect to the functions of programmed events, especially whether selecting events can maintain culturants without reinforcing functions. When experimental arrangements are devised to account for observation of events composing IBCs, experimental events can always be described more precisely and less ambiguously in terms of coordinated operant contingencies because (1) events and relations that need to be demonstrated do not need to be assumed; (2) specific dependency and functional relations are described; (3) the importance of accounting for other factors is recognized; and (4) functional properties of events at one level of analysis are not dependent on having properties on another level. This suggests that similarities observed between cultural selection and reinforcement may be, in part, an artifact of relinquishing precision by focusing on how reinforcement alters the production of aggregate products without looking at behavior.

If interest in the metacontingency enterprise shifts away from cultural selection toward alterations in IBCs, coordinated operant contingency models may be more useful than metacontingency models. Describing events in terms of coordinated operant contingencies does not require differentiating events that select culturants from those that reinforce operant behavior, a difference in function that cannot be wholly demonstrated if current experimental procedures are maintained. Coordinated operant contingency descriptions are highly flexible and can be altered to highlight progressive changes in functional and dependency relations that occur with respect to reinforcing events. A coordinated operant contingency account retains continuity of behavioral processes when analyses are scaled from one to multiple individuals whereas metacontingency models stipulate that different processes emerge for environmental alterations produced by groups. Additionally, coordinated operant contingency descriptions are applicable to a wide range of circumstances, even those in which the influence of certain events cannot be attributed to reinforcing properties because not all group members contact those events. If the probability or rate of occurrences of culturants is shown to be sensitive to changes in dependency and functional relations within IBCs, coordinated operant contingency models may be preferred to account for deviations in functional relations obscured by more general metacontingency models.

Whether the metacontingency enterprise utilizes coordinated operant contingency constructs or not, specific recommendations for metacontingency experimentation can be made. Metacontingency experiments should begin manipulating how many individuals can contact selecting events to demonstrate control of behavior by contingent consequences that cannot function as reinforcing events for all group members. The relevance of factors that promote cooperation to produce outcomes when participants could produce such outcomes individually should be investigated. Communication should be restricted to various degrees to demonstrate how stimuli acquire discriminative functions in metacontingency contexts. Procedures and analyses like those of Sampaio et al. (2013) and Tadaiesky and Tourinho (2012) should be integrated into experimental designs to isolate all events and relations described by the metacontingency construct. IPDGs and IWMPs have been shown to be useful for the enterprise, but the metacontingency construct may be more descriptive of observed experimental events if procedures specific to the enterprise are utilized, like those constructed by Sampaio et al. (2013) and Smith et al. (2011).

Other recommendations can be made in terms of elaborating, refining, and transforming descriptions. Preanalytic assumptions should be more clearly described to clarify the enterprise’s orientation to behavioral events. Individual aggregate products should not be described as indicative of IBCs to demonstrate more assuredly metacontingency events and relations in experimental settings. Unless additional experiments demonstrate otherwise or a decidedly molecular view of behavior is agreed upon, selecting events should be described as always having reinforcing properties. An account of the participation of verbal activity in cultural events should be formalized, and cultural selection should be described in terms of behavioral processes unless experiments demonstrate events that cannot be explained by them. More elaborate metacontingency models (e.g., Houmanfar et al., 2010) address some of these concerns, but additional system building is necessary to formulate a comprehensive and conventional account of cultural events.

Although it is not fair to critique the metacontingency enterprise by standards of another, an interbehavioral orientation suggests that metacontingency experiments have yet to examine significant cultural events of interest. According to Kantor (1982), cultural interbehavior refers to behavior shared across individuals characterized by arbitrary, institutional stimulus functions acquired under the auspice of a group. As such, cultural behavior is delimited from other behavior because it is conventional. Individual behavior is not the focus of the metacontingency enterprise, but neither is the establishment of conventionality. Most metacontingency experiments that have examined the participation of verbal behavior have found it to be crucial for the consistent production of aggregate products, but no experiments have demonstrated the establishment of communicative interactions—specific to groups examined—in which linguistic responding between participants is maintained despite reinforcing events not being dependent on such behavior. Language is a defining feature of cultural groups, and a model of cultural selection should not only be able to account for the functionality of verbal behavior in groups but facilitate experimentation in which it is established. Kantor (1970) once warned that fixating on reinforcement would debilitate the experimental analysis of behavior. The metacontingency enterprise should be similarly cautious with respect to cultural selection as it is currently conceptualized within the enterprise.

The extent to which this analysis has any bearing on the metacontingency enterprise is for workers within that enterprise to. While some findings were similar to Zilio’s (2019), a focus on functions pertaining to description and experimentation allowed for both an analysis with criteria specific to the metacontingency enterprise’s own logic and standards and constructive orientation toward system building activities. This analysis was performed specifically with respect to the metacontingency enterprise, but it also serves as model for understanding science and other enterprises through an interbehavioral lens. It is different from interbehavioral analyses by Kantor (e.g., 1958), which give more attention to the participations of presuppositions and assumptions within an enterprise. Similar analyses may also be useful for the metacontingency enterprise, and perhaps others.

Funding

This research was not supported by any federal or non-federal funding.

Data availability

Not applicable.

Declarations

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Code availability

Not applicable.

Conflicts of interest/competing interests

The authors declare that they have no conflict of interest.