Sport is a human activity characterized by particular organization and functioning in given performance contexts. The ecology of sport is not only distinguished by physical characteristics of locations at which player activity takes place, but also by its social significance and cultural aspects. A factor of interest for audiences as well as sport scientists and performance analysts, is to observe which players or teams succeed in competition, characterized by a complex interaction of physical (e.g., surfaces, areas, weather conditions) and social (e.g., rules) constraints specific to each sport. In team sports, the overarching aim of a team is to score more points/goals than the other team, implying an initiative to score (attacking), and to prevent the other team from scoring (defending). This distinction between attacking and defending is somewhat of an oversimplification, because there are sports where a team can be defending a lead, when in possession of the ball, and a losing team without the ball may be pressurizing space or the opposition ball carrier to re-gain possession. It may be more appropriate to talk about each team’s offensive and defensive capacity (Mateus, 2005). In each case, a specific feature of team ball sports is the flow of interactions between cooperating and competing players, with or without the ball, to achieve the competitive performance aims of each team (Mateus, 2005). The aims of the teams are thus mutually exclusive during competitive performance. Essentially, each team tries to prevent the implementation of the performance strategy of the other team, as well as dissipate their tactical actions (Davids et al., 2005). The result is the emergence of a complex web of interactive behaviors of players and teams during competition.

Team sport competitions have been conceptualized as complex dynamical systems composed of many interacting parts (e.g., Davids et al., 2005). A dynamical systems approach to sport performance describes how patterns of coordinated movement come about (‘emerge’), persist, and change. It builds on the insight that teams, as social systems, consist of a large number of interacting parts, endowing them with a capacity for spontaneous pattern formation or self-organization. The spontaneous creation of coherent macroscopic patterns (e.g., team coordination) is important scientifically, and the resulting macroscopic patterns of the dynamics of one or a few collective variables or order parameters can be studied carefully (e.g., the cluster phase; Duarte et al., 2013), without needing to record all the microscopic states of the individual parts (e.g., the movement of each player; see Kelso, 1995). Conversely, when the dynamics of macroscopic phenomena have been identified, the contributions of relevant dynamical components (e.g., the movement of certain players) to the overall dynamics may be investigated in a top–down fashion (Araújo et al., 2015).

Performers can generate behavioral patterns that are tightly coordinated with environmental events (e.g., the match configuration), in order to achieve a specific performance goal. In team sports, athletes are surrounded by physical (e.g., gravity, altitude, ambient temperature) and social (e.g., audience, rules of the game, local rivalries) constraints. Successful performance in sport is predicated on the constraints of an individual’s perceptual and action capabilities, and is grounded on information used for action selection and goal achievement (Araújo et al., 2006).

Assuming the mutuality and reciprocity between the performer and the environment, i.e., the performer-environment system as the key level of analysis, implies, not only an active agent, but also that the environment is an active part of a system, facilitating specific behavioral outcomes (Davids and Araújo, 2010). One consequence of this account is that behavior can be understood as self-organized, in contrast to organization being imposed from inside (e.g., by instructions of the team captain in sports teams) or outside (e.g., directions of the coach). Team performance does not always need to be prescribed by internal or external structures, due to inherent self-organization tendencies that exist for exploitation in sports teams (Araújo et al., 2016) yet traditional sports psychology and pedagogical practice has decreed that internal/external prescription is the default mode to face this challenge (e.g., Lidor and Henschen, 2003). Ecological dynamics suggests that, from a player’s point of view, the task is to exploit physical (e.g., pitch characteristics determined by federation rules) and informational (e.g., perception of the movements of other players) constraints to stabilize emergence of intended behaviors. Constraints have the effect of reducing the number of configurations available to a dynamical sports team (conceptualized as a social adaptive system) at any instance. In team sports competitions, coordination patterns (individual or collective) emerge under constraints as less functional states of organization are dissipated. Every team sport presents its own set of interacting constraints that helps define competitive functioning. This view contrasts with the traditional use of team statistics and notations that are often used to mechanistically operationalize team sports performance in a data-driven way (e.g., frequency counts and averages in sports: see Vilar et al., 2012; Travassos et al., 2013; Couceiro et al., 2016).

In fact, the ever-increasing successful implementation of new technologies in sports inevitably leads to sequential stages of: data capture, pre-processing, transfer, post-processing, visualization, and analysis. This relentless emphasis in current performance analysis (PA) methodologies involves a high volume, variety and complexity of data, to meet the demands of practice, evaluation and training in different sports.

As fundamental theories have failed to make any reliable analysis or predictions around football, most contemporary research has focused on probabilistic models (e.g., Owramipur et al., 2013). Nevertheless, and despite the success of some of these methods, they rely only on the overall information retrieved from previous matches, without considering on-the-fly information retrieved during a current match, such as each athlete’s positional and physiological data, nor producing any macroscopic tactical metric inherent to such data (see Couceiro et al., 2016) for a critical analysis of these methods. A critical point here is that, typically, notational analysis tends to describe performance, but omits reference to the whys and hows that underlie the structure of many recorded behaviors, which would clearly define their functional utility (McGarry, 2009). Recently, Gudmundsson and Horton (2016), in an exhaustive review of measures of team behavior in “invasion sports” (i.e., association football, hockey, rugby union), proposed how information obtained from spatio-temporal data (Araújo et al., 2004; McGarry et al., 2014) could be categorized in what they called a “coherent framework.” The categories of this framework were: (i) description of variables that can be obtained from spatio-temporal data (i.e., player trajectories, sequence of events); (ii) divisions of the playing area; (iii) players’ interactions and temporal sequences of events (captured by networks and related metrics); (iv) types of information obtained by data mining techniques (labeling events, predicting future events, identifying formations, identifying plays and group movements, segmenting a match); (v) metrics to measure player and team performance; and finally (vi) visualization techniques used to present metrics calculated from the visuo-spatial data. This framework, as well as others (Bartlett et al., 2012), is useful, but its rationale is essentially methodological, focusing on the production of an operational account of performance. This means that it has limited use for understanding team behaviors, only a capacity to measure them and operationally define them. The questions that arise are: what is the meaning of these metrics? What theoretical concepts of a theory of team behaviors can be used to explain how these behaviors emerge?

To address these, and related, questions PA should not focus on production of operational measures and performance descriptors only. Rather, PA could really benefit from the development of a detailed account providing theoretical support for interpretation of data that arise from the descriptive measures. Ecological dynamics is such a theoretical framework for studying behaviors in team games (Davids et al., 2005; Araújo et al., 2006). It has the advantage of recognizing the ‘flexibility’ of social systems (i.e., teams), and its principles can explain how the same performance outcomes can emerge from different behavioral patterns. More than a decade ago (e.g., Araújo et al., 2004; Davids et al., 2006), we began to develop a theory of interpersonal behaviors in team sports, applying concepts such as that of “superorganism” to characterizing a team (Araújo, 2006; Duarte et al., 2013; Araújo et al., 2016). This path led us to the “team synergy” hypothesis and its operationalization, presented here (see Passos et al., in press for an extensive synthesis of findings and methods applied in different team sports). The hypothesis of team behavior as a synergy is possible only if we conceive the origin of a team synergy as an emergent phenomenon, which is a consequence of players’ perceptual attunement to shared affordances, as framed by the ecological dynamics approach that we briefly overview next. But before that, we concisely present an overview of other competing hypotheses exiting in the sport sciences literature.

There are a limited number of hypotheses in the literature to explain team behaviors (see Araújo and Bourbousson, 2016, for a review). Here we outline the two more prominent proposals for the team synergy hypothesis: the hypothesis of shared knowledge grounded in the socio-cognitive approach and the participatory sense-making hypothesis grounded in the enactive approach.

The assumption of shared knowledge is predicated on the possession by team members of mental models that provide a basic shared understanding of how to achieve desired performance outcomes (Ward and Eccles, 2006). It is argued that team efficacy could increase when a sophisticated, global and comprehensive mental representation of a performance context of a collective action is somehow shared by all players and put into practice. Lack of coordination between the intentions of individual performers and those of the team imply that a shared cognitive state has yet to be achieved, with resulting difficulties in team performance (Eccles, 2010). In this view, it is the construction and updating of the individual’s mental model that explains how multiple performers may simultaneously act together. As many team members are simultaneously coordinating, the amount of similarity within individuals’ representation acting together becomes a key feature, thus indicating a state of shared understanding within the team. However, knowing “who knows what” at each moment of a match would involve an immense computational load for a representational system. Particularly, the mechanism to explain re-formulations of a team member’s schema, when changes occur in the content of another member’s schema, has proved difficult to verify (Mohammed et al., 2000). A key difficulty is to justify how mental representations exist beyond the mind of an individual organism and can be somehow shared in a collective representation (Riley et al., 2011; Silva et al., 2013).

Another alternative is the participatory sense-making hypothesis which suggests that team coordination processes should be investigated by reconstructing how individual ‘cognitions’ articulate during performance environments. With the claim that a performer possesses a unique “interiority,” this hypothesis has given a special attention to the implicit ways how each performer experiences his/her ongoing activity. Participatory sense-making processes refer to how the meanings that each performer internally builds from his/her activity corroborate with the meanings simultaneously built by co-performers, and how this participation in sense-making is experienced (De Jaegher and Di Paolo, 2007; Froese and Di Paolo, 2011). From this perspective, it is assumed that putative ‘participatory sense-making’ emerges from a cooperative effort (De Jaegher and Di Paolo, 2007). Performers working together achieve the experience of mutual engagement in real-time. Any divergence in how each member experiences a performance situation leads to varied degrees of participation in sense-making, variations in feelings of connectedness with the other, and variations in expectations for actions from others (e.g., Bourbousson et al., 2010a, 2015). This enactive approach tries to avoid representationalism, but by being grounded in the “interiority” of each individual it needs to operationally define what the internal sense-making process is and contrast it with representations of lived situations which are heavily reliant on memory. In fact, it should be clarified why there is a need to add the label “sense-making” to the process of team coordination, unless we are operating in an approach that overstates the asymmetry of organism and environment (Fultot et al., 2016). The idea proposed by enactive theorists that meaning (about the world) is internally built is not operationally defined in terms that exclude representations (prompting the question: where is meaning constructed, and how, and in what form?) and therefore, may be subject to the same criticisms as the shared knowledge hypothesis about the overreliance on mental representations, individual or collective.

In general, both the participatory sense-making and the shared knowledge hypotheses rely on data from a posteriori verbalizations of team sports performers. Therefore, the organization of behavior, from these perspectives, is mainly understood and derived from the verbalized conceptions and perceptions about behavior, not from the actual behaviors themselves. From these perspectives, overt behavior and its organization in contexts like team sports is a surrogate of verbalized shared sense-making or knowledge, without a self-organization of its own. In contrast to these perspectives, it is argued here that behavior has an organization that goes beyond what a performer possesses (verbal information expressing knowledge or its interiority). Rather it is deeply rooted in the specific circumstances of behavior, in which continuous performer-environment interactions are not determined by one component of such a system.

Understanding group coordination from the perspective of ecological dynamics requires investigating the dynamical principles that underpin the self-organizing patterns of group dynamics in performance contexts (Araújo et al., 2016; Passos et al., in press). Interpersonal patterns at the behavioral scale can be understood and modeled in terms of their own dynamics, without the need to investigate each performer’s particular movement patterns (nor their verbalizations of what these might be; Schmidt et al., 2011). Analyses of group dynamics capture the continuous interactions of system components to form stable behavioral patterns, characterized as attractor states of system dynamics (Schmidt et al., 1990). Important for the study of group dynamics is the fact that synchronization processes (i.e., temporal coordination of unfolding events in a system) have been found to occur between organisms that are connected, not only mechanically, but also informationally. In this approach, information is conceptualized in the specific patterns in surrounding energy distributions (e.g., light, sound, neural) that can specify properties of the world, meaning that the mapping from patterned energy distributions to properties of the world is direct (not reflected upon or inferred from representations of these relations). Much research has demonstrated how this dynamical synchronization process can operate predicated on information to produce coordinated timing of interpersonal interactions (e.g., Richardson et al., 2007). In these studies, the only possible way that the rhythmic units could have interacted was via the information available to the visual systems of the human participants. Importantly, distributions of energy surrounding an organism, when properly described, are rich with information that specify action-relevant properties of the world. More to the point, specificational information is the invariant structure of (low) energy distributions lawfully structured by emergent individual–environment interactions that, because of that lawfulness, specify relationships of each individual with the environment. This is precisely the type of invariant structure that supports inter-player coordination in sports teams.

Understanding of the informational basis for the emergence of group coordination, and the dynamics that permit the self-organization of coordinated group action to occur is, therefore, a priority for research from this viewpoint. This is not because it implies athletes do not have intentions, thoughts and mental states, but rather because research demonstrates how integrally intertwined are body (e.g., nervous, physiological, psychological, emotional) and contextual (e.g., social, cultural, climate, altitude) sub-systems, during interpersonal coordination (Marsh et al., 2009). Adopting an environment-individual system perspective (Richardson et al., 2008; Järvilehto, 2009; Marsh et al., 2009), an ecological approach proposes that knowledge of the world is based upon recurrent processes of perception and action through which humans perceive affordances (i.e., opportunities for action, see Gibson, 1979). The concept of affordances presupposes that the environment is directly perceived in terms of what actions a performer can achieve within a performance environment (i.e., it is not dependent on a perceiver’s expectations, Richardson et al., 2008). Moreover, to perceive an affordance is to perceive how one could act with respect to an environmental layout. This way, affordances are neither external properties of an environment nor are they representational properties of mind; they are relational properties of a performer-environment system, which capture the link between individual and environment and are perception-action system specific (Fajen et al., 2009). Affordances capture the action specific relations that exist between the skills/capacities of an individual performer and the action relevant properties of a task, including social tasks (Heft, 1989; Withagen et al., 2012). The theory of affordances is based on the dual interdependence of perception and action, where affordances are the primary objects of perception, and action is the realization of affordances (Araújo et al., 2013). Consequently, it is the role of scientists to discover information in ambient energy arrays that specifies action-relevant properties (such as team synergies) and to show how team synergies may be constrained by such information (Araújo et al., 2006).

Recently, Silva et al. (2013) challenged the hypotheses for team coordination described in the previous section. We argued that these hypotheses were predicated on ‘knowledge about’ the environment, which can be used to share knowledge and participatory sense-making and consequently organize action. Rather, during competitive performance, the organization of action by perceiving surrounding informational constraints is expressed in ‘knowledge of’ the environment. This crucial distinction emphasizes that the perception of shared affordances (for others and of others) underpins the main communication channel between team members during team coordination tasks. We grounded these explanations on Reed’s (1996) conception of affordances where affordances are resources in the environment. In this view, the relative persistence of some affordances in the animal’s environment has given rise to the evolution of several distinct action systems, enabling the animal to take advantage of these affordances (Reed, 1982, 1996). This view suggests that evolution gives rise to animals that are capable of taking advantage of the relatively persistent affordances in a performance environment. These resources in a performance environment have imposed selection pressures on some group of individuals, causing them to evolve perceptual systems to perceive these relations. Predicated on the key ideas of Reed, we went as far as arguing that affordances are collective environmental resources that have existed prior to the upskilling of the individuals that came to perceive and use them. At the timescale of team sports performance, these ideas can be taken to imply that more successful teams are composed of athletes who have learned to perceive shared affordances for and of other players.

Reed (1982, 1996) also asserted that the exertion of selection pressures by affordances is a dominant force in evolution by natural selection. However, an important criticism of this view is that the environment not only shapes individuals, but individuals also modify their environment (see Withagen and Van Wermeskerken, 2010). More precisely, individuals and their environments evolve together. In fact, Heft (2007) brought the process of niche construction to attention when arguing that the human environment is largely a product of social processes, aligned with an important discussion about the process of “niche construction” in evolutionary psychology (Odling-Smee et al., 2003). The phrase “niche construction” refers to how the modifications that the organism brings about in its environment can create new evolutionary equilibria and trajectories (Withagen and Van Wermeskerken, 2010).

Aligned with the ideas of Reed, we concur that affordances constitute the context of selection, but updating our hypothesis of shared affordances we take the dynamic view presented by Withagen and Van Wermeskerken (2010) that individuals’ dissolution and construction of affordances change this context, demonstrating the key roles affordances play in learning, development, and evolution. This clearly signifies that individuals do not passively evolve so as to fit in a preexisting environment. Rather, individuals and their environments actively co-evolve, and each individual’s modification of the environment has a constitutive role in this co-evolution process. Individuals modify their (social, physical) environment, which can change the selection pressures to which they are exposed. Lewontin (1983) emphasized that there are constraints on the co-evolution of animal and environment. Hence, contrasting with the ideas of Reed, Lewontin would not conceive the physical properties of the world as a resource that exerts selection pressures, but would conceive of them as constraints on the evolution of the individual–environment system. This ideation indicates, that affordances arise along with the evolution of action systems and provide the context of selection and adaptation after they have evolved (Withagen and Van Wermeskerken, 2010). In short, niche construction not only requires the utilization of affordances, it also consists of a change in the affordance layout. Hence, individuals often create and dissolve affordances, with other individuals (members of a species or, in humans, sports teams) being exposed to these modified environments as new members of a group (Withagen and Van Wermeskerken, 2010). Thus, the ecological inheritance from one group to the next encompasses an inheritance of affordances. It is important to emphasize, however, that changes in the affordance layout are not exclusively the result of individual activity. Indeed, as mentioned above, geo-physical (including built environments) and social processes can also alter the affordances in an individuals’ eco-niche, such as a training setting. Furthermore, it is important to reiterate that creating a niche is of course not always evolutionarily consequential – it does not have to change the context of selection. However, as we have discussed, niche construction can alter the developmental trajectory of a collective system in small or extensive ways, which could be significant or not for group performance.

Extending this idea to the level of interpersonal interactions, ecological dynamics predicts that the presence of others extends action possibilities that are realizable by individuals to action possibilities realizable by groups. Indeed, Gibson (1979) argued that ‘behavior affords behavior’ (p. 135) and that it was important to note that the ‘richest and most elaborate affordances’ (p. 135) of all are affordances of other people in social contexts. What do these rich insights imply for understanding of coordination in team sports? The suggestion is that affordances can be perceived by a group of individuals trained to become perceptually attuned to them (Silva et al., 2013). Given the mutually exclusive performance aims of both teams in a sports contest, affordances are shared and sustained by common task goals of team members who cooperate and compete to achieve group success. From this perspective, team coordination depends on the collective attunement of individual athletes in a team to shared affordances (Silva et al., 2013). Through practice, players can become perceptually attuned to affordances of others and affordances for others during competitive performance, and can refine their actions (Fajen et al., 2009) by adjusting behaviors to functionally adapt to those of other teammates and opponents. Moreover, individuals in a team can act in a way to create competitive circumstances (the affordances) that are favorable to them. For example, by pressuring opponents to play more in one zone of the playing area, this strategy can create affordances for attacking play by freeing other areas of the field/court/pitch. This active construction of affordances can be trained too. These processes enable a group of players to act synergistically with respect to specific match circumstances (Araújo et al., 2015). Importantly, shared affordances are predicated on perception, whereas synergies regard action, meaning that a synergy is a physical process that realizes a task goal under the guidance of affordance-specific information.

There are slightly different perspectives on the relevant properties of a synergy (Turvey, 2007; Latash, 2008; Araújo et al., 2016). Latash (2008) identified characteristics that should be met for a group of components to be considered a synergy, including: (i) sharing patterns, where the components should all contribute to performance of a particular task; (ii) error compensation, where some components show changes in their contributions to a task, compensating for a component that may not be making a relevant contribution; and (iii), task-dependence, the capacity of a synergy to change its functioning in a task-specific way, or to form a different synergy for a different purpose based on the same set of components. Task dependence is an expression of redundancy (Bernstein, 1967), or more generally degeneracy (Davids et al., 2006; Seifert et al., 2016), defined as the capacity of structurally different components to perform a similar, but not necessarily identical, function with respect to context (Edelman and Gally, 2001). Alternatively, Riley et al. (2011) identified two properties of a synergy: (i) dimensional compression, where the degrees of freedom that potentially are independent become coupled so that the synergy has fewer degrees of freedom (a lower dimensionality) than the set of components from which it arises (Bingham, 1988). Interestingly, the behavior that emerges from the interactions among degrees of freedom, not among the structural components, constitutes a second level of dimensional compression. Dimensional compression at both stages results from imposing environmental, task and individual constraints, which couple components so they change together, rather than independently. The second property of a synergy for Riley et al. (2011) is reciprocal compensation and it is a less biased description of what Latash (2008) described as “error compensation.” To label a behavior as “error” is to assume normativity in observers. Another possible label for an error is creativity, even though creativity goes much beyond reciprocal compensation (Hristovski et al., 2011).

Therefore, we address four properties of a synergy which are: (1) dimensional compression (Bingham, 1988; Riley et al., 2011), (2) reciprocal compensation (Latash, 2008; Riley et al., 2011); (3) Interpersonal linkages. According to Latash (2008), who termed this property “sharing patterns,” a way to quantify the amount of sharing is the matching of the sum of the individual contributions to the task, where the overall measurement of task performance may be related to the dimensional compression property. In contrast, and inspired by Ingold (2015), we address this property in relation to the individuality of each element of a synergy, and therefore, the different ways they can link together; and (4) degeneracy (Davids et al., 2006; Latash, 2008; Seifert et al., 2016). However, degeneracy may be seen as a more general property that can be expressed in different ways, compared to the previous properties which are more specific. This broader view is needed because it shows adaptability across tasks (e.g., competitive matches) and across changes in system components (e.g., players) as occurs in a sport team. Next we describe how these properties can be measured in team sport performance.

In this overview paper we have discussed the relevance of key concepts from ecological dynamics, a theoretical framework that has provided a rationale for explaining how specific constraints might impact on team synergies formed by players during competitive performance. It has been found that these ecological constraints shape the perception of shared affordances available for players, which underpin the assembly of interpersonal synergies expressed in collective actions within a team. These important group processes support the formation of synergies. Their key properties have begun to be identified including, dimensional compression, reciprocal compensation and degeneracy, guiding the current meaning of operational variables of relevance for PA, such as team center, team dispersion, team synchrony, and team communication. Theoretical and empirical developments in methods of analysis of team coordination and performance can benefit from a powerful theoretical approach that situates and traces relevant team properties as defined by synergies.

For example, actual networks are static: after defining a time interval and collecting all the observable edges during a set period, a network refers to a discrete aggregated view of the whole system. However, the dynamic behaviors of the network are what strongly affect its functionality and efficiency (Moody et al., 2005). Well-connected nodes can quickly become weakly connected (or even disconnected) over time (Hill and Braha, 2010), which is important to consider in future studies because each dynamical system is constrained by universal principles, but at the same time requires its unique suite of analytical and numerical tools to understand its behaviors (Barzel and Barabási, 2013). Moreover, in sport teams, spatial constraints captured in playing areas, rigorously limit, mark, and focus performance behaviors, e.g., a target area, such as a goal, hoop or tryline, has a strong effect on network connectivity patterns. This observation indicates where it is important to develop metrics for analysis and modeling, clarifying where performers attain different impacts, according to their relative positioning, considering key constraints as well as previous performance contributions in a competitive match.

In this paper, we have suggested how concepts like shared affordances, and synergies, framed in an ecological dynamics perspective, present key principles to substantiate the meaning of existing and future operational metrics in PA, and inform about team performance and training development.

DA, KD: discussed the structure of the paper; DA: lead the writing of the ms; KD: refined the writing of the paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.