Previously, the construct of situated conceptualization developed as an account of how simulations of conceptual knowledge become situated (Barsalou, 2003a–b, 2005a–b, 2008a–b, 2009; Barsalou et al., 2003; Yeh & Barsalou, 2006; also see Barsalou et al., 1993). Simulating conceptual knowledge about a bicycle, for example, does not simply represent a bicycle alone against an empty background. Instead, simulating a bicycle typically occurs in a background situation, such as riding cautiously along a busy street on the way to work (one of infinitely many situated conceptualizations associated with the category of bicycles). By simulating background situations this way, agents prepare themselves for situated action with the focal object or event. Simulating the ride to work, for example, generates useful inferences about the setting, relevant agents and objects likely to be encountered, relevant actions to perform, and mental states likely to result.
This chapter develops the construct of situated conceptualization beyond earlier treatments. After the first section establishes theoretical properties of situated conceptualization, the second demonstrates its applications to a variety of cognitive, affective, and behavioral abilities.
Situated conceptualization: Theory
Concepts
Because situated conceptualization is a construct associated with conceptual processing, it is essential to define what is meant by a “concept” (also see Barsalou, 2012; Murphy, 2002). Following the account developed here, a concept is a dynamical distributed system in the brain that represents a category in the environment or experience and that controls interactions with the category’s instances (e.g., the concept of bicycle represents and controls interactions with bicycles). Within the human conceptual system, thousands of concepts represent diverse categories of settings, agents, objects, actions, mental states, properties, relations, and so forth.
Although many accounts of concepts exist, they generally assume that a given concept aggregates information across interactions with a category’s members. The concept of bicycle, for example, aggregates information accumulated across interactions with bicycles. Using selective attention to isolate information relevant to the concept (e.g., perceived bicycles) and then using integration mechanisms to integrate it with other bicycle information in memory, aggregate information for the category develops continually (Barsalou, 1999, 2003a; Schyns, Goldstone, & Thibaut, 1998). Although learning plays a central role in establishing concepts, strong genetic constraints limit the features that can be represented for a concept and also their integration in the brain’s association areas (Simmons & Barsalou, 2003).
Once the conceptual system is in place, it supports virtually all other forms of cognitive activity. During online interaction with the environment, concepts contribute to perception via inferences that support perceptual constancy, pattern completion, anticipatory movement, and so forth. Concepts enable categorization, making it possible to identify the objects, agents, actions, and so on currently present in a situation. Concepts support action via inferences that establish the affordances of objects, actions likely to be effective, and probable outcomes (e.g., affect, reward). In general, concepts make it possible to go beyond the information given, providing an agent with diverse forms of expertise about perceived category instances (Bruner, 1973).
Concepts also play central roles in offline processing when people represent nonpresent situations during memory, language, and thought. As Donald (1993) reviews, humans, unlike other species, are prolific in representing and analyzing past situations, planning and coordinating future situations, and developing counterfactuals to current situations. Concepts provide the building blocks for representing and processing nonpresent situations. Without concepts, representing nonpresent situations would not be possible.
Grounded cognition
Because the construct of situated conceptualization draws heavily on the framework of grounded cognition, it is useful to place the construct within this framework. A natural way of doing so is to begin with the historical perspective. Since the cognitive revolution, the so-called sandwich model has dominated theories of cognition, viewing cognition as processes “sandwiched” between perception and action (Hurley, 2001). As a consequence, cognitive processes are often viewed as relatively modular, making it possible to study them without taking perception and action into account. By simply focusing on mechanisms associated with attention, working memory, long-term memory, language, and thought, it is possible to develop satisfactory accounts of cognition. Based on this assumption, paradigms for studying cognition – together with theories that explain results from these paradigms – typically ignore perceptual and motor processes.
From the perspective of grounded cognition, the sandwich model will never explain cognition successfully. Instead, proponents of grounded cognition argue that cognition will only be understood once the relevant domains of study are expanded significantly beyond classic cognitive mechanisms (Aydede & Robbins, 2009; Barsalou, 2008a, 2010; Clark, 2008). Only when these additional domains are included will accounts of cognition be successful.
Across the literature on grounded cognition, researchers often argue that four additional domains beyond classic cognitive mechanisms must be included. First, researchers increasingly propose that cognition relies heavily on the modalities that constitute perception, action, and interoception. As described in the next section, the basic cognitive process of simulation utilizes mechanisms in the modalities. When conceptually representing the color of a nonpresent object, for example, the cognitive system utilizes simulations of color in the visual system (e.g., Hsu, Frankland, & Thompson-Schill, 2012; Simmons et al., 2007). Analogously, when conceptually representing how an object sounds, people do so with simulations of sounds in the auditory system (Kiefer et al., 2008; Trumpp et al., 2013).
Second, researchers increasingly propose that cognition often (but not necessarily) relies on bodily states and physical action (for reviews, see Barsalou et al., 2003; Niedenthal et al., 2005). On the one hand, cognitive states often produce related bodily states. When people perceive tools, for example, their motor systems anticipate the actions associated with object affordances (Caligiore et al., 2010; Tucker & Ellis, 1998). When people perceive the facial expressions of others, they sometimes mimic and embody them (e.g., Niedenthal et al., 2010). On the other hand, bodily states can influence cognitive states. When people experience physical warmth and cleanliness, for example, they may feel socially connected and psychologically cleansed, respectively (e.g., IJzerman & Semin, 2009; Lee & Schwarz, 2010).
Third, researchers propose that cognition depends on the physical environment. Since Gibson (1966, 1979), many researchers have argued that it is impossible to understand and study perception by only considering sensory systems. Because perception results from the coupling of sensory systems with the physical environment (together with the body), it is essential to include the physical environment in accounts of perception. More recently, researchers working from the perspectives of situated action and situated cognition have similarly argued that cognition cannot be explained without incorporating its coupling with physical environments (e.g., Aydede & Robbins, 2009; Clark, 1998, 2008). Because the brain establishes distributed patterns for processing familiar situations, taking the physical situations that produce and support these patterns into account is essential for satisfactory theories of cognition.
Fourth, researchers propose that cognition depends on the social environment. As evolutionary theories often argue, increasingly powerful social cognition constituted the primary adaptions of cognition in humans (e.g., Donald, 1993; Tomasello, 2009). Related to action, humans developed increasingly sophisticated representations of agency and self, together with increasingly powerful abilities for social mirroring, imitation, and cooperative action. Related to theory of mind, humans developed the abilities to establish joint attention and represent the minds of others. Related to communication, humans developed remarkable new abilities to use language, establish social groups, create culture, and archive cultural bodies of knowledge. For all these reasons, understanding human cognition successfully requires understanding its coupling to the social environment. Analogous to understanding how the physical environment shapes and supports cognition, it is essential to understand how the social environment shapes it as well.
Thus, from the grounded perspective, cognition does not simply reside in a set of cognitive mechanisms. Instead, cognition emerges from these mechanisms as they interact with sensory-motor systems, the body, the physical environment, and the social environment. Rather than being a module in the brain, cognition is an emergent set of phenomena that depend critically on all these domains, being distributed across them (e.g., Barsalou, Breazeal, & Smith, 2007; Clark, 1998, 2008).
Finally, referring to this perspective as “embodied cognition” is relatively narrow (Barsalou, 2008a, 2010). Certainly, cognition depends on the body in critical ways. Nevertheless, it also depends on sensory-motor systems, the physical environment, and the social environment. The classic way of describing this perspective as “grounded cognition” acknowledges all the domains in which cognition is grounded and from which it emerges (e.g., Pecher & Zwaan, 2005; Searle, 1980). As we will see shortly, the construct of situated conceptualization integrates cognition across these domains.
Simulation
As we will also see shortly, the construct of simulation plays central roles in situated conceptualizations (Barsalou, 1999, 2008a, 2009). Most basically, a simulation re-enacts the kind of brain state that occurs while interacting with a category’s members. When simulating a bicycle, for example, the brain re-enacts the kind of brain state that occurs while experiencing bicycles. As we will see, simulations play diverse roles in representing a category, producing a variety of situated predictions and controlling action.
For simulation to occur, experiences of actual category members must become established in long-term memory. Consider experiencing instances of the category hammers. As people experience hammers, brain areas that process their properties become active and associated together (Martin, 2007). Specifically, distributed associative patterns are likely to become established across the fusiform gyrus (shape), premotor cortex (action), inferior parietal cortex (spatial trajectory), and posterior temporal gyrus (visual motion). Following many learning episodes, an increasingly entrenched associative network reflects the aggregate effects of neural processing distributed across these areas. From the perspective developed here, this entrenched network represents the concept of hammer, given that it contains aggregate information about its respective category (elsewhere these distributed networks have been referred to as “simulators”; e.g., Barsalou, 1999, 2009). For a similar perspective, see the chapter in volume 2 by Brunel, Vallet, Riou, Rey, and Versace (also see Versace et al., 2009; Versace et al., 2014).
Once a concept has become established in memory, it produces specific simulations of the category dynamically. On experiencing a hammer (or hearing the word “hammer”), a subset of the hammer network becomes active to simulate the processing of a hammer in one of infinitely many ways. Typically, these simulations remain unconscious, at least to a large extent, while causally influencing cognition and action. To the extent that part of a simulation becomes conscious, mental imagery is experienced. Such simulations need not provide complete or accurate representations but are likely to be incomplete and distorted in many ways, representing abstractions, caricatures, and ideals, as well as specific learning episodes.
In a Bayesian manner, the hammer simulated on a given occasion reflects aspects of hammers experienced frequently in the past, together with aspects that are contextually relevant (Barsalou, 2011). In other words, the underlying network generates one of infinitely many hammer simulations dynamically, each adapted to the current situation. Once this simulation exists, it represents a hammer temporarily in working memory, producing, for example, anticipatory inferences about the object’s affordances.
As Barsalou (2008a) reviews, simulation appears to be a basic computational mechanism in the brain. Not only is it central for conceptual processing, it also plays important roles across the spectrum of cognitive processes, from perception to social cognition. By no means, however, is simulation the only representational process in the brain. Instead, other important representational mechanisms work together with it to produce cognition, especially linguistic forms and perhaps various abstract representations, including conjunctive neurons in association areas (e.g., Barsalou et al., 2008; Simmons & Barsalou, 2003).
Situatedness
When a simulation is constructed to represent a category, it is not constructed in a vacuum. Instead, much evidence suggests that simulations are situated (e.g., Barsalou & Wiemer-Hastings, 2005; Wu & Barsalou, 2009; for a review, see Yeh & Barsalou, 2006). When representing the category of chairs, for example, a simulated chair is likely to be embedded in a background setting, together with agents and objects likely to be present, and also with actions, events, and mental states likely to occur. By representing a cate...
