After having spent more than a decade investigating children’s behavior in the laboratory [1
], Roger Barker set out in the late 1940s to catalog children’s activities in more common everyday settings [2
]. He was prompted to begin this laborious project after coming to the realization that in spite of his familiarity with how children behave in specific, controlled laboratory investigations, he knew little about children’s daily lives in their homes and communities. It soon became apparent to him that this state of affairs characterized the field of psychology more generally. Unlike the other sciences, psychology has neglected to develop a systematic understanding of the frequency and distribution of its most essential subject matter, behavior. He pointed out, e.g., that whereas biologists know how common or rare particular flora and fauna are, and how they are distributed in nature, psychologists have little way of knowing whether some psychological phenomenon studied in the laboratory is commonplace or atypical, and under what everyday circumstances it occurs.
Psychological tradition led Barker initially to make two assumptions at the outset of his attempt to find the causes of everyday actions. First, he assumed quite reasonably that the basis for order in everyday behavior was to be discovered solely at the level of individual action. After all, psychology’s focus has long been, and continues to be, on the individual, and in turn, on the ways in which happenings in the environment affect individual action and thought. The second assumption followed directly from the first: It was assumed that much of the time the specific individual actions that were observed could be tied to antecedent events stemming from other individuals and directed toward the child, such as directives from a parent, teacher, or peer. Under the broad influence of stimulus-response (S-R) thinking in psychology (or, in later decades, of input-output formulations of cognitive psychology), it seems logical to assume that actions can be tied to immediately antecedent conditions. However, data collected over a number of years by Barker’s research team indicate that both of these assumptions are questionable.
Their observational data indicated that antecedent individual occurrences (“social inputs”) were, at best, rather weak predictors of a child’s actions. Typically, fewer than 50% of child-directed actions or utterances were followed by behavior congruent with those “inputs”. This unexpected finding forced Barker to reexamine the considerable body of evidence that had been gathered and the prior assumptions that he had made. After all, there was order to children’s actions. In classrooms, children usually acted as students should; in the town’s stores, children generally behaved in ways that were appropriate to those establishments; in religious services children acted in socially normative ways for the setting, and so on. And yet, typically only in a minority of instances were these behaviors initiated by a social input from another individual. How then were these regular patterns to be explained? In an imaginative leap, Barker saw that focusing solely on the actions of an individual (the first assumption above) was inadequate to account for the data. He recognized that the best predictor of behavior was “where” in the community the child was at any particular time.
1.1. The Discovery of Behavior Settings and Their Components
Barker was initially led to this conclusion by recognizing that the actions of an individual child had greater variability across different locations in the community than did the actions of different children within the same location. This pattern indicated that something quite powerful psychologically was tied to place. However, this observation is strikingly out of step with psychological tradition in that it runs against the view that the best level of analysis for understanding individual action is at the scale of the individual (much less at more molecular levels). Indeed, Barker found that a child’s actions at any particular moment were most accurately predicted by knowing “where” she was (e.g., a music lesson, a game of dodge ball, a church service) rather than knowing details about individual child. From this starting point, efforts to understand what “where” means in psychological terms led Barker to the discovery of extra-individual, dynamic structures that he called “behavior settings”.
Not infrequently, the most surprising discoveries are hidden in plain sight by the stance one adopts. In spite of their rather “late” discovery in psychology, behavior settings have always been ubiquitous in human environments. After all, the actions of an individual are never free-standing, as if an individual is ever positioned in some empty Cartesian space—although that would seem to be how psychology has long operated. Instead, action is always situated. Behaviors always occur somewhere, in some context. Notably, a significant feature of the context is a dynamic, yet stable pattern of actions generated by joint participation of two or more individuals with the functional support of affordances (“milieu”). Barker called these ecological structures behavior settings. Consider, for example, an on-going class session. The setting “class session” is constituted by the patterned action of students and teachers, with the support of affordances, such as places to sit and surfaces on which to write. That is, if some number of students sit at their desks quietly, raising their hands when they wish to speak, speaking after the teacher calls on them, recording information in notebooks placed on their desks, etc., then collectively their actions in conjunction with material features of that setting generate a higher-order pattern of behavior that is conducive to teaching and learning. (Of course, there are multiple patterns of action that are suitable to teaching-learning, this being only the most traditional.)
It is revealing to note that each individual’s actions are “double-sided”. On the one hand, by operating in a manner that is socially normative with respect to e.g. classroom settings, each individual directly contributes to the constitution of that setting. On the other hand, in order to be a participant in a classroom setting, the individual must limit his or her actions in ways that are socially normative in relation to such settings. In other words, individuals’ actions simultaneously contribute to the creation and maintenance of a setting and are constrained by virtue of their participation in the setting. By acting in ways that help to constitute a particular behavior setting, individuals as a matter of course limit their actions in a manner that helps to sustain that setting, even as they may be seeking their own ends within the setting.
Such reciprocal processes, whereby structured actions create social structures that constrain subsequent actions have come to be seen as commonplace by social theorists such as Giddens [5
] and Bourdieu [6
]. Psychologists, by remaining focused solely on individual action, are prone to miss the role of these collective eco-behavioral structures. Moreover, the occurrence of higher-order patterns that are generated by their components, and concurrently constrain the operational possibilities of those components, are typical of many natural systems [7
The defining attribute of any behavior setting—that is, what distinguishes a behavior setting from a mere collection of individuals, or from a location without any individuals (e.g., an empty classroom)—is a requisite degree of interdependence
among its components, the participants and the milieu. Barker [9
] (p. 41) writes in this respect, “the phenomena of psychology and the environments in which (they occur) are interrelated; they are interdependent in the way a part of a system and a whole system are interdependent”. Establishing the degree of interdependence is an empirical matter determined by assessing the degree to which alterations in one part of a setting reverberate in other parts of the setting (for details see [3
Barker and his colleagues investigated behavior settings and their operations extensively over many decades [10
]. Still, there is an untested assumption that underlies this perspective. For individuals to participate appropriately in an on-going setting, they must first be able to perceive the type of setting that they are entering. In other words, they must recognize the identity (or meaning) of that setting. This presupposition is critical for obvious reasons: if an individual cannot recognize the type of setting she is about to enter—for example, whether it is a classroom setting, or a religious service, or a team game—how is she to know what to do, and what not to do, when joining it as a participant?
Of course, this understanding must stem from a prior history of engagement in the various behavior settings one encounters within his or her community. No doubt one important facet of early social development involves being engaged as a participant in a variety of settings. Through such experiences, the individual develops a repertoire of actions-in-context. Indeed, Barker has delineated the ways in which the range of experiences, as well as the level of responsibility in a setting’s operations, covary with age [4
] (pp. 116–118).
To be a participant in a behavior setting, and in doing so, to contribute to its operation, the individual must select the range of permissible behaviors that are appropriate for the setting; and such selection is contingent on perceiving the setting type. (Barker employed the term “behavior setting genotype” to refer to different settings that shared common functions and operations. For example, two different fire stations in a town would share a common genotype.) Moreover, to maintain the necessary operations that constitute a setting, individuals must be able to perceive when collective actions are beginning to deviate beyond a normative range and to threaten the viability of the setting’s purposes.
Barker was well aware of this fundamental premise about the perceived meaning of a setting. He writes, “The identification of a setting rests upon the direct perception of a synomorphic (i.e.
, mutually supportive) relation between a standing pattern of behavior and (its) milieu” [4
] (p. 53). He does offer some evidence for this possibility by appealing to the agreement among his research team in their efforts to identify existing behaviors settings. However, these individuals likely had prior familiarity with the concept of a behavior setting. Alternatively, does the documented existence of behavior settings not stand as prima facie evidence that their identity is perceivable? That does seem reasonable, but it remains to be demonstrated more directly that individuals in the course of their daily activities can recognize the identity of familiar behavior settings. Moreover, and more critically, assuming that individuals can perceive the identity of a familiar behavior setting, what is the nature of the stimulus information that specifies a particular behavior setting from the standpoint of a perceiver? These two questions—can the identity of a behavior setting be perceived, and if so, what is the nature of the information specifying its identity—are the primary concerns of the research to be reported here.
Generally speaking, there are two likely sources of information that specify the identity of a behavior setting. The most obvious source of information is the physical layout of the setting, such as its furnishings and their arrangement. Although this information is apt to be the most salient, it is not necessarily reliable. Physical layouts can sometimes be utilized for uncharacteristic purposes. A wedding can occur in an open area of a park, a raffle can occur in a church, and a classroom can be utilized for a play performance. These examples indicate that when patterns of action seem to be in conflict with physical layout, the former usually overrides the latter. Admittedly, these circumstances are atypical, but still such examples indicate that the physical layout plays mostly a supportive role in (and in some cases a hindrance to) the social dynamics of a behavior setting.
These considerations suggest that the most distinctive and reliable perceptual information specifying the meaning of a behavior setting is likely to be the collective action pattern of the individuals who contribute to its constitution. Barker posited that a behavior setting is principally the quasi-stable “standing pattern” of behavior among its participants. For this reason, our goal in this investigation was to assess whether individuals can perceive the identity of a behavior setting based solely on the pattern of collective actions by its participants. Admittedly, the pattern of actions is shaped by milieu possibility and constraints, but in the present work the milieu structure was not perceivable, only being at best implied through actions. We predicted that the dynamic structure that is revealed among the joint actions of setting participants over time could function as perceptual information that specifies the meaning of a familiar behavior setting.
1.2. Methodological Approach
To explore these issues, we adapted a methodology that was initially developed by Johansson [12
] in the domain of event perception. Specifically, he was interested in the visual perception of motion stemming from both inanimate and animate sources. To illustrate his approach with a comparatively simple case, consider a rigid rod rotating around an axis in the field of view at 45 degrees from the fronto-parallel plane (i.e
., an oblique rotation). He hypothesized that this object in motion would be perceived with reference to invariant geometric relations in an otherwise changing pattern of visual information. The methodology that he devised for testing this hypothesis, a point light display
, involved placing lights at both ends of a rod, and then filming the rod in darkness with only the lighted ends of the rod visible as it rotated. Even with this minimal kinematic information, namely, two moving points of light, a rotating rod oriented at 45 degrees was readily perceived, and this is because the relative distance between the lights was preserved (invariant) in relation to the perceived angle of rotation. Specifically, as the rod (unilluminated but for the ends) rotated around a fixed axis, and end-to-end away from the viewer (projectively in the picture plane), the relative distance of the two lighted end points of the lights remained invariant. Indeed, they specify the perceived angle of rotation out of the fronto-parallel plane. Research using such point light displays has generated an extensive research literature on what has come to be called the perception of structure from motion.
] applied this same methodology to the study of animate motion by attaching lights to the joints of a moving person, and filming various actions in otherwise total darkness. He found that perceivers were readily able to identify actions such as walking, dancing, bicycling, climbing a ladder, and doing pushups, based solely on the kinematic patterns of the moving lights. Remarkably, the action could often be identified within less than a second of exposure. Later research demonstrated that other qualities of the individual are also readily perceivable, such as their gender and the weight of an unseen lifted object [14
] described these visual displays as offering a “kinematic specification of dynamics” or in short, the KSD-principle. In the context of biological motion, Runeson [15
] (p. 386) describes the KSD-principle in the context of person perception studies in the following manner: “properties pertaining to a person that have a dynamic (“causal”) role in the generation in his or her movements are specified by the resulting kinematic patterns”. In other words, he is proposing that the meaning or purpose of the action, that is, the individual’s intentions, is specified in the perceived kinetic patterns.
Those familiar with Gibson’s ecological approach to perceiving will note a strong similarity between it and Johansson’s analysis. In spite of some differences between the Johansson and the Gibsonian research programs concerning how to best analyze such motion patterns [16
], there is agreement that perceiving involves detecting lawful regularities in a changing visual array. Both also agree that these regularities are detected by the perceptual system without the mediation of non-perceptual processes. In other words, the available visual information considered over time is sufficient to account for what is perceived. For this reason, from the standpoint of both approaches, perception of the event is taken to be direct. (Discussion of the theoretical and philosophical implications of this point goes beyond the bounds of this paper, but see [18
Applying this reasoning and methodology to our research question, we examined whether behavior setting type could be recognized with reference to the pattern of joint or collective actions among participants in a setting. The experimental materials developed for the present project were modeled after Johansson’s methodology. Instead of applying points of light to joints of the body, we represented the person schematically as a whole (see below). Our goal was to examine whether observers could identity a familiar behavior setting type from a kinematic display of actions among those involved in the setting. Because our focus was on the kinematic display of information, fixed, physical features of a setting, such as furnishings, were not included in the displays. Our assumption was that specific patterns of collective action that emerge from the operation of a behavior setting can serve to specify to a perceiver the type of setting that it is.
1.3. The Primary Research Materials
One member of the research team (the second author) observed and recorded the actions of participants in four public settings in a small Midwestern town (population approx. 4000). Observations were carried out over a 30 min period of time when there was a relatively high level of activity in each setting. In some cases, such times were easy to determine, as in the case of lunchtime at a small restaurant. In other cases, such as a bank lobby, several visits were necessary in order to find a time when there was considerable activity on-going.
The observer developed detailed written records of the activities of participants in each setting by noting on prepared, properly scaled floor plans the paths of locomotion of individuals through each setting and the locations where they remained in a stationary position, either standing or sitting. The estimated length of time it took for individuals to walk from one point to another in the settings was also noted on the floor plans. The public settings were the lobby of a bank, a reading room in a community library, a small restaurant that serves mostly lunches, and an ice cream shop. A fifth setting was also included, namely, a portion of an organized basketball practice that existed on videotape.
Working from the detailed transcribed records, as well as the one video recording, computer animations of the activities in these settings were generated using Autodesk®
3D animation software. The animations were created through the collaboration of the setting observer and a computer graphic designer (the third author) in order to capture the dynamics of the setting.
Each of the animations displayed only the actions of the participants in the setting against a black field, while omitting all layout or design features (e.g., tables, chairs, walls). Still screen shots from two of the displays are presented in Figure 1
Still screen shots (cropped) of two of the setting animations. Note that in the absence of movement, the identity of these settings is nearly impossible to discern.
Still screen shots (cropped) of two of the setting animations. Note that in the absence of movement, the identity of these settings is nearly impossible to discern.
Each behavior-setting participant was represented in the display by a blue cylinder, with a red dot on its surface marking the direction the individual was looking. Seated participants in the behavior setting were represented by a cylinder 2/3rds the height of a moving or standing one. If a participant sat or stood up during the simulated time interval, the cylinder shrank or become elongated, respectively. The only other action represented in the displays was movement (i.e., displacement of the cylinders) within the setting.
Initially, each animation was based on the ten-minute interval during the observation period when activity was greatest. However, because asking participants in the subsequent experiments to view five displays, each ten minutes long, was anticipated as excessively demanding, and also because in some of the settings, several minutes transpired during the ten-minute interval when relatively little activity occurred, we further edited the animations to three-minute displays. In the case of those settings in which activity was relatively continuous, specifically, the restaurant, the ice cream shop, and the basketball game, a three-minute interval that seemed representative of the initial ten-minute animation was selected. In the case of the library reading room and the bank lobby, short intervals of time when no activity was occurring were edited out. In the end, each animation consisted of three minutes of somewhat continuous action based on the original observational records.
One obvious quality of the still screen shots in Figure 1
is noteworthy to point out. As static representations, the identity of the setting each still shot captured in a “frozen” second is indeterminate. This quality generally holds for all of the investigations that utilized the Johansson point-light display technique. When the point-light displays (and in the present case, the setting patterns) are presented dynamically, the indeterminateness is greatly reduced. In the case of the point-light displays of human movement, as noted above, clarity occurs in under a second. The research presented here assesses whether a setting’s identity can also be recognized in kinetic displays of joint action.