Abstract: The theory of situated learning claims that every idea and human action is a generalization, adapted to the ongoing environment, because what people see and what they do arise together. From this perspective, thinking is a physical skill. As we create names for things, shuffle around sentences in a paragraph, and interpret what our statements mean, every step is controlled not by reinstantiated grammars and previously constructed plans, but adaptively recoordinated from previous ways of seeing, talking, and moving. Situated learning is the study of how human knowledge develops in the course of activity, and especially how people create and interpret descriptions (representations) of what they are doing. This introduction provides a historical perspective of situated learning, including the work of Dewey, Bartlett, Vygotsky, and Ryle. I provide examples of how situated learning is being applied today in business process redesign.
The theory of situated learning claims that knowledge is not a thing or set of descriptions or collection of facts and rules. We model knowledge by such descriptions. But the map is not the territory: Human knowledge is not like procedures and semantic networks in a computer program. Human knowledge should be viewed as a capacity to coordinate and sequence behavior, to adapt dynamically to changing circumstances.
If the above description of situated learning still seems confusing, it is probably because of how the AI literature has equated the following terms:
Identifying knowledge with descriptions often leads to interpreting situated cognition to mean that there are "no internal representations" or "no concepts in the mind." Rather, the claim is that "knowledge" is an analytic abstraction, like energy, not a substance that can be in hand. You cannot inventory what someone knows. Knowledge representations (e.g., textbooks, expert systems) are descriptions, which are tools, not knowledge itself. "The map is not the territory."
Because knowledge is not a thing or set of descriptions, we do not learn by transferring facts and rules from one head to another. Understanding how learning is a process of conceiving an activity, and activities are inherently social, puts emphasis on improving learning addressing issues of membership, participation in a community, and identity.
Notice that we are referring to the conception of activity, not tasks. Activities embrace the norms of social groups. Examples are: working home alone, phone conversations, attending a lecture, gardening, weekend relaxing, being a researcher, being a business trip, etc. Saying that activities are social means that norms constrain how we dress and talk, what constitutes being "on task," what constitutes an interruption, etc. This knowledge of social choreography orients how we communicate information, what constitutes an interesting theoretical inquiry, how we justify design judgments, and how we interpret policies. This is the social dimension of all knowledge.
Saying that activities are social has nothing to do per se with whether the activity is done alone or with other people present. Again, the superficial view that social means "in the presence of other people" (compare to the superficial view that situated means "in some location"), fully distorts the psychological claim. Action is situated because it is constrained by a person's understanding of his or her "place" in a social process.
Because knowledge is not merely indexed, retrieved, and applied, but problem solving involves reconceptualization, current models of problem solving are impoverished. Controlled experiments may reveal interesting patterns in how concepts are related, but they tell us little of how people come to conceive of a situation as being a problem. Because knowledge has been wrongly viewed as descriptions, problems have been viewed as descriptions--reducing the everyday sense of trouble, discomfort, and difficulty to mathematical puzzles, neatly expressed on paper and neatly solved by logic.
Because knowledge has been equated with scientific theories, the expertise of applying theories to develop a good design and the everyday problems of interpreting policies and rules have been ignored. The process of articulating theories about the world, expressing values, and arguing about conflicts has been relegated to simply knowing what the facts are and making logical inferences. In practice, the problem is that people disagree about the facts and how to make tradeoffs. Expertise is not merely knowing the rules, but knowing how to make good interpretations.
Practice, what people say and do is not the same as theory, descriptions of what people say and do. Therefore we cannot write down once and for all what people know. Learning to become a member of a community of professionals is not accomplished by transferring the rules and handing over the tools. Knowledge of the professional is conceptual, embodied in ways of seeing, roles, ways of interacting. And because concepts are not words, learning cannot be accomplished by describing or telling alone.
Again, saying that learning is situated is explaining the nature of human concepts. We are not saying that people learn best by "trying something out" (even that is a common oversimplification of the idea). We are saying that learning occurs in all human activity, all the time. The recommendations are to examine the nature of how problems arise, how people use theories in practice, how interpretations are biased by a person's conceptualization of his or her role. In effect, situated learning calls us to see how rules make ongoing learning necessary in the workplace, and how the development of knowledge is constrained by every individual's conception of what he or she is supposed to be doing.
Scaffolding2 is a dynamic process by which an expert deliberately expands a novice's skills, based on feedback from the novice in practice. Expert and novice progress together through the ZPD, reciprocally building shared experience and representations.
Both expert and novice are relating their own task actions (self-regulation) to the speech and actions of the other (other-regulation), by which they adjust their perception and make new choices of language and activity.
Put another way, the ZPD is mutually constructed to maintain a correspondence between other- and self-regulated behavior achieved through scaffolding.
Signs (language and other representations) are representations of external activity that become reconstructed and internalized. In this way, speech, which organizes meaning encountered in the social world, is internalized to become thought, allows for speech production (necessary to take part in the social community), and becomes the basis for activity.
Tools and signs are acquired by participation in the social environment.
This short tour through activity theory helps explain why and in what sense learning is situated. It helps to remember that the learner's conception of the situation is what frames and orients his or her experience, that is, what is learned. Activity theory partially explains how individual conceptions are brought into alignment with a community of practice through tools, signs, and activities.
It is the community rather than the individual, that defines what a given domain of work is and what it means to accomplish it successfully. (Suchman and Trigg, "Understanding Practice," Design at Work, p. 73)
Learning is a process that takes place in a participation framework, not in an individual mind. (Hanks, Situated Learning, p. 15)
The point is not so much that arrangements of knowledge in the head correspond in a complicated way to the social world outside the head, but that they are socially organized in such a way as to be indivisible. (Lave, Cognition in Practice, p. 1)
--constituting an evolving membership and capability to participate in different forms.
--the means of reproduction and development of communities of practice.
To understand and change the result of cognitive-social activity--what people believe and do--we must take into account their social-material interactions. Viewing context as merely the data for information processing leads to an over-malleable view of behavioral change, as if it is merely shaped by internal contents. Rather, we need to understand the duality of structuring inside and out, and how these transform each other.
The strongest effect is not in "how to teach," but in "how to change" a social system. Formal education alone--teaching theory more effectively--is not sufficient. Rather, we must understand change within the trends and resources available for activity. For example, we can't simply teach ITS researchers a new view of cognition and expect a change to occur in the practice of instructional design. We must work within the constraints of funding, computational methods, and forms of practical involvement already underway. Researchers must "get on board" the trends of the instructional design community and reshape practice as a participant (as opposed to lecturing or just delivering tools).
According to Lave, ethnographic studies emphasize apprenticeship in order to
reveal the indivisible character of learning and work practice. This in turn
helps to make obvious the social nature of learning and knowing:
'Intelligent' cannot be defined in terms of 'intellectual' or 'knowing how' in terms of 'knowing that'; 'Thinking what I am doing' does not connote 'both thinking what to do and doing it.' My performance has a special procedure or manner, not special antecedents.
Efficient practice precedes the theory of it; methodologies presuppose the application of the methods, of the critical investigation of which they are the products.
It was because Aristotle found himself and others reasoning now intelligently and now stupidly and it was because Izaak Walton found himself and others angling sometimes effectively and sometimes ineffectively that both were able to give their pupils the maxims and prescriptions of their acts.
It is therefore possible for people intelligently to perform some sorts of operations when they are not yet able to consider any propositions enjoining how they should be performed. (Ryle, 1949)
Instead of deriving from stored structures, interaction creates structure,
which is always adapted in constructing future behavior. W.F. Hanks explains
this in the preface to Situated Learning (Lave and Wenger, 1991):
[In the work of Lave and Wenger] it is not merely that the structure issue is transposed from the level of mental representations to that of participation frames. Rather this transposition is compounded by a more subtle and potentially radical shift from invariant structures to ones that are less rigid and more deeply adaptive. One way of phrasing this is to say that structure is more the variable outcome of action than its invariant precondition.... (p. 16)
It involves a prereflective grasp of complex situations, which might be reported as a propositional disposition, but is not one itself. (p. 20)
Contrasting different activities (not two puzzles), Lave found that transfer didn't occur between "best-buy calculations in supermarkets where the problem solver is agent and on paper where the problem solver is the object of the exercise." That is, experience in real-life didn't transfer to school work.
On the one hand, we could criticize Lave's work because the essence of the transfer claim is that theoretical generalizations, that is, descriptions, would be useful in multiple settings. Lave has implicitly identified knowledge with descriptions once again. In saying that knowledge doesn't transfer, Lave hasn't addressed the issue of whether descriptions are useful guides in multiple settings.
Knowledge engineering is a prime example of the power of descriptive generalization (Clancey, 1988). Domain-general representations are applied to build new expert systems. The idea that descriptive theory is "decontextualized" is a superficial rendering of what a theory is and how it is made useful in a new setting. The point is that a theory, like any map, must be interpreted for the problems at hand.
On the other hand, Lave is correct that in AI studies context is treated
equivalently within a problem, across problems, across domains, and across ways
of life (home, school, and experimental lab). Claims about transfer must be
relative to the grainsize of the change in situation. In claiming that a
descriptive theory "transfers" we must always say "more or less, depending on
the perceptual similarity and social-interactional activity of the new setting."
Lave summarizes the "decontextualized" position she is criticizing:
We will examine Lave's weight-watcher story in more detail later. The example, appropriately interpreted, suggests that a mathematical conceptualization more general than the situation at hand is orienting the person's behavior. Greeno has been aware of this idea, in his consideration of the conceptual invariants of mathematical reasoning across situations (Greeno, 1988), and not just their manifestation in situated activity. Because Lave and others have avoiding talking about the brain, they have poorly articulated the psychological implications of their claim: Conceptual organizers do get reused, but they are always adapted.
To be clear, Lave's analysis makes major, important claims for learning
research. She shows in her studies that cognitive studies of learning based on
descriptive mechanisms alone have several deficiencies:
The statements in Figure 1 indicate how the idea that knowledge consists of stored facts and laws derives from the philosophical assumption that the world consists of segmented objects with defined properties, which are selected and filtered by perception. Furthermore, the rationalist view holds that the laws we observe are somehow embedded in the world, rather than being descriptions of patterns we observe. An alternative view is that there are many ways to describe the world (Lakoff, 1987; Edelman, 1992; Gregory, 1988). The direction of science is culturally influenced by language and tools. But the important impact of this observation is in everyday design and decision making when tradeoffs are required.
The relevant facts when designing something or interpreting a policy are culturally determined. The force of the social construction of knowledge is not to say that science is impossible and anything can be believed. Rather, the force is on the production of designs and the learning about how to act within a social matrix of policies and values. By reducing knowledge to scientific facts and laws, the cultural expertise of a practitioner in creating and evaluating good designs is inadequately understood (Schön, 1987). In effect, knowledge has been equated with theories, designs, and policies; such descriptions must be properly understood in relation to each other, and not equated with concepts.
Figure 2 presents the common view that equates descriptions in the world with concepts in the mind. The term "representation" is used equivalently to refer to external representations such as programs, text files, and diagrams, and internal acts of representing such as visualizing something, talking to oneself; and these experiences are equated with conceiving, a subconscious process. Instead of distinguishing between descriptions, experiences, and concepts, the exclusively symbolic view insists that these are all isomorphic, all the same thing, just symbols (Vera and Simon, 1993).
The field of ITS was founded on this idea:
The epistemology of knowledge driving the early design of ITS programs was of course the epistemology of expert systems and symbolic learning programs. Figure 3 illustrates how learning was conceived as a separate module and hence a process independent from acting. Learning was viewed as something that only occurs on reflection, in which the problem solver generates descriptions criticizing the performance and finding ways to improve the models and inference procedures: "A learning systems responds acceptably with respect to some performance element within some time interval following a change in its environment" (Smith, et al. 1977).
This is a reasonable description of how people modify descriptive models, but it doesn't allow for recoordination to occur during problem solving itself. It suggests that all learning involves manipulating descriptions and fundamentally ignores perceptual recoordination. This model fundamentally distinguishes between physical skills and intellectual skills, suggesting that there is no important physical aspect to conceptualization. Newell made these assumptions explicit in describing "the total cognitive system," a generalization of Soar (Figure 4).
In Figure 4 see that perceptual and motor systems are independent from the deliberation process (P->E, C, and D->M are disjoint during performance). Features for segmenting the world are learned independently from productions (P, E->C-D, and M are disjoint). And once again, performance is distinguished from learning: Doing and learning are viewed as separate kinds of action.
Finally, because the assumptions of the symbolic approach identifying
situations with descriptions are not called into question in the design of
laboratory experiments or the creation of symbolic models, this approach cannot
be falsified on its own terms. On the one hand, researchers claim that
philosophy is irrelevant (Vera and Simon, 1993). The philosophy of science might
appear irrelevant to someone (following Figure 1) who believes that science is
objective, without cultural underpinning. Lave argues that the manner of
carrying out an experimental investigation of cognition is cultural because
it makes assumptions about the nature of knowledge, situations, and problems
(e.g., that problems in everyday life are experienced as descriptions like
puzzles). Unfortunately, this suggests a symbolic theorist that situated
learning is unscientific, because it is attempting to eliminate experimentation.
But the issue is understanding the starting point for the experimental subject
and how experiments are related to everyday life:
If we criticize a symbolic model, such as a model of medical diagnostic problem solving, the symbolic theorist can always make post hoc repairs. Examining protocols, the symbolic theorist will claim that the subject's vocabulary and beliefs that don't fit the initial cognitive model were in their heads before the experiment began. Of course, this is the very assumption which must be experimentally demonstrated. Schön (1979) calls this "historical revisionism."
Situated learning research amply demonstrates how "normative" models don't fit human behavior when we give people a chance to show us what kinds of descriptions they prefer for modeling their own experience. Bamberger shows this clearly in her studies of children creating their own music notation (Bamberger, 1991). In related work (Clancey, 1994), I showed how two children learning basic geometry are unable to identify the straight lines on a screen because they are unsure what grainsize is appropriate for segmenting the markings on the screen. Confused by the pixels which are visible, and the artifacts of a low-resolution screen, they describe the lines in terms of "the little dots," "these little lines," "bigger" and "thicker." Claims that the world consists of symbols that are objectively available and matched against production rules in long-term memory ignore how perception is a process of segmenting and grouping, not selecting or filtering from given objects. Furthermore, a close study of such problematic situations shows that what learners see or hear is coupled to the kinds of interpretations they are capable of making. In a complicated, conceptual process, perceiving and activities are learned together as a unit.
As Dewey pointed out, the terms we use for describing a situation are
grounded in our physical method of looking: orienting the paper,
gesturing, and aligning reference objects such as rulers. Dewey emphasized that
the context, the world, is not a given entity, but consists of conceptual
situations. Again, a situation is not the observer's description, but
the person's experience:
Instead of interpreting the character of sensation, idea, and action from their place and function in the sensorimotor circuit, we still incline to interpret the latter from our preconceived and preformulated ideas of rigid distinctions between sensations, thoughts, and acts. The sensory stimulus is one thing, the central activity, standing for the idea, is another thing, and the motor discharge, standing for the act proper, is a third.
As a result, the reflex arc is not a comprehensive, or organic, unity, but a patchwork of disjointed parts, a mechanical conjunction of unallied processes...
What is wanted is that sensory stimulus, central connections and motor responses shall be viewed, not as separate and complete entities in themselves, but as divisions of labor, functioning factors, within the single concrete whole, now designated the reflex arc.... What shall we term that which is not sensation-followed-by-idea-followed-by-movement...? Stated on the physiological side, this reality may most conveniently be termed co-ordination. (Dewey, 1896, p. 137)
The "cottage cheese story" demonstrates that inquiry is a complex coupling of physical materials, sensorimotor coordinations (including non-descriptive ways of seeing), plus the articulation and manipulation of constraints, in the manner described by Dewey. Pedagogically, the cottage cheese story celebrates inventiveness and the importance of teaching representational conventions without destroying creativity, without leading children to believe there is only one way to think, only one "correct" way of describing facts and working problems (Brown, et al., 1988).
Lave's analysis of the cottage cheese story is inherently focused on
understanding the "multiple and mutually constitutive character of
problem-solving processes" (p. 165). Her analysis is concerned not only with
social processes, but also perception, hand-eye coordination, conceptualization,
and talk. In studying how a shopper judges which items to buy, she observes the
precise ordering of attention and motion:
Lave explains that the aspects of problem-solving activity, such as
calculating which item in a store is a better buy, are often assigned in our
analysis to different times and locations. But the process of knowing
similarities and differences and articulating them is not easily located within
With this introduction, we consider now Lave's presentation of the cottage
But besides contrasting this enactment with calculation on paper, we must also draw out the psychological aspects of perception and conceptualization, which themselves contrast with symbol manipulation on paper. That is, our analysis of the cottage cheese story needs to relate three phenomenon:
Unfortunately, the attempt to bridge social and cognitive views in examples like this has caused an number of misunderstandings. For example, Vera and Simon believe that the cottage cheese anecdote is meant to illustrate that "knowledge about interaction with real-world objects is not symbolically represented."(p. 18) We can't be sure what the subject muttered to himself, but he did say "got it!" and talked about calculus. This suggests that the subject was symbolically representing the situation in his imagination. That is, the subject was interpreting the meaning of his thoughts, moves, and experiences. "Symbolic" here refers to anything interpreted as meaningful, not to conventional marks or symbols in a symbol structure (Newell's (1990) definition).
Of course, the issue raised by Lave is whether comprehending the specification ("take three-quarters of two-thirds of a cup of cottage cheese") involved a description, which was then coded into movements (Figure 4). Or was the subject visualizing and projecting an ordering of movements that arose as an embodied physical coordination?
Lave argues for a dialectic process, but without a theory of perception and coordinated movement, her explanation fails to account for individual differences. This is ironic because the theory of situated learning "promotes a view of knowing as activity by specific people in specific circumstances."(Lave & Wenger, p. 52) We understand the circumstances better, but without a psychological foundation for knowledge-level descriptions, there can be no specific accounting for where ideas come from, how situations are recognized (even though they are always different in particulars), and how habits are developed and sustained over time. In this respect, situated learning theory requires psychology and perhaps neurobiology to be comprehensible to educators and cognitive scientists.
Reify individual employee, a constant player moving in the corporation
|Technical details of products and services (internal
(stored in individual heads)
|Training given to individual |
Reify company-customer relations as stable & responsive
|Customer relations (interactive capacity)||Cross-functional team
(manifest in activity)
|Project activity of functional workgroup and teams|
As a second example, consider the problem of exploiting computer technology in the workplace to facilitate learning. Typically, applying the symbolic, individual view, we would say that the problem to be solved is delivering the proper facts and policies to the desktop. Again, we would adopt a normative view and say that people fail because they don't know the rules.
But when we study what happens everyday in the workplace, we find that
problematic situations arise not only because people don't know the rules, they
don't know what to do about the rules when they conflict with the
company's goals and values:
There are issues out there that we have no control over. This is one of them. And I have begged for a year and a half to get this. And I haven't gotten it and I'm not going to. So then, we're going to have to be more flexible as a team and use our creativity to support what we have to do to get our jobs done....
I'd just like to make you aware, this is not a new problem. It's a very,
very old problem. I've worked with it. And, when you're told no, you just have
to deal with the 'no.' And, get smart about how to work around that 'no.'
In designing computer tools for learning on the job for situations like this, we will not help if we simply deliver the company's policies to the desktop. We must help people invent creative interpretations and develop good designs which resolve conflicts. We could start by understanding the context in which people are operating and focus down to develop a genuinely useful tool.
This team is not trying to rotely learn the company's methods and procedures, but to find creative interpretations of policies, creative adaptations of information systems, and new forms of collaboration that will enable them to get their job done, despite other people's obstructions. Learning is again occurring on several levels: The team leader is teaching her cohorts how to deal with difficult situations. She is conveying a constructive attitude and orientation for "dealing with the 'no.'" To understand what kinds of tools would be useful, we would want to follow such interactions in the workplace to see how the workers represent their situation and learn how to deal with it.
Broadly speaking, we find that learning on the job often arises when trying
to manage activities. They are not solving puzzles but trying to describe
the situation that appears problematic. Such on the job learning is
fundamentally not rotely digesting somebody else's representations, but
improving capability to create vocabularies and notations, in order to explain,
document, and manage activities.
People are not just digesting and "implementing" pre-existing, written methods and procedures that formalize work processes. "Learning on demand," a catch-phrase we hear today, must be more than classroom learning at a different time and location, as if someone sitting at his or her desk finally decides to memorize someone else's theory. In the workplace, learning occurs within socially-coordinated inquiry, as a process of creating new representations to document, understand, or gain control of everyday work. Talk about a "curriculum" obscures the learner's perspective: 1) how to use existing tools and languages, 2) how to represent one's own activities within a new language or visualization, and 3) how to participate effectively in multiple communities of practice.
Learning on the job often is a process of finding words and frameworks for describing one's conceptualization. These descriptions, diagrams, and partial plans then become guides, as they are later interpreted for articulating the pros and cons of alternative designs and ways to work around policies.
That is, workers need tools for representing what is on their mind, as ways
Eleanor Wynn (1991) summarizes the limitations of the normative, business
functional view of work:
The alternative view, which Wynn advocates, is to understanding what is problematic in everyday work. Rather than treating knowledge as static, we view categorizing, sorting, evaluating, and judging as dynamic processes. So how to interpret a policy means adapting a prior interpretation, not merely applying a rule. By definition, what constitutes information and appropriate criteria must be adjusted with every problematic situation--otherwise the situation would not be a problem, indeed, there would be no "situation" to speak of. In effect, situated learning calls for the kind of "metacognitive" approach that ITS has advocated. But the emphasis is again not on learning a fixed strategy or way of organizing the world, but learning how to learn. It's a cliché, but situated learning reveals the idea in a new way. The focus is not on "learning how to digest" or "learning how to access facts" but learning how to invent, to work around and through difficulties, and to interact productively.
Situated learning theory reveals the limitations of computer-human interaction analysis based on a descriptive, stored view of knowledge. Such assumptions led us to use computers for automation, to replace people and teachers by machines. We sought to store knowledge in computers and deliver expert systems to workers and intelligent tutoring systems to students. We didn't ask, "Why are representations (diagrams and descriptions) created? How are categories, priorities, and policies interpreted within the constraints of an activity?" We viewed people as information processors working alone at a workstation. We didn't understand the origin of problems in conflicting values and identities, different paradigms influencing decisions about roles, rights, and access to information (Zuboff, 1988).
In the idealistic, objectivist world of the symbolic approach, we ignored that the problems arise not only because people don't know scientific facts and laws, but they disagree conceptually about how to create and value alternative designs. For all the concern about knowledge, the symbolic view fundamentally misconstrued the psychological nature of how decisions get made. As we saw in the cottage cheese example, the same criticism can be made about the social-interactional view, which appropriately encourages us to look outside the web of information-processing computers and telecommunications links, but never explains why changing human behavior is so difficult.
Some of the implications for ITS research are:
"They have knowledge but cannot act."
"Knowledge should be stored before it gets lost."
"Mental models are in the head of the user."
"Perception is controlled by theory."
"Learning is an individual, knowledge-based process and a social communication process."
How would we criticize or reinterpret these statements from the perspective of situated learning theory?
Alexander, C., et al. (1977). A Pattern Language. New York: Oxford University Press.
Bamberger, J. (1991). The mind behind the musical ear. Cambridge, MA: Harvard University Press.
Bamberger, J. and Schön, D.A. (1983). Learning as reflective conversation with materials: Notes from work in progress. Art Education, March, pp. 68-73.
Bartlett, F. C.  (1977). Remembering-A Study in Experimental and Social Psychology. Cambridge: Cambridge University Press. Reprint.
Bateson, G. (1972). Steps to an Ecology of Mind. New York: Ballentine Books.
Bateson, G.(1979). Mind and Nature: A necessary unity. New York: Bantam.
Bickhard, M. H. and Terveen, L. (in preparation). The Impasse of Artificial Intelligence and Cognitive Science.
Brooks, R.A. (1991). Intelligence without reason. In Proceedings of the 12th International Conference on Artificial Intelligence (pp. 569-595). San Mateo, CA: Morgan-Kaufmann Publishers.
Brown, J. S., Collins, A., and Duguid, P. (1988). Situated cognition and the culture of learning. IRL Report No. 88-0008. Shorter version appears in Educational Researcher, 18(1), February, 1989.
Clancey, W. J. (1988). The knowledge engineer as student: Metacognitive bases for asking good questions. In H. Mandl, & A. Lesgold (editors), Learning Issues in Intelligent Tutoring Systems Springer-Verlag. pp. 80-113.
Clancey, W.J. (1992). Representations of knowing: In defense of cognitive apprenticeship. Journal of Artificial Intelligence in Education, 3(2),139-168.
Clancey, W.J. (1993). Situated Action: A neuropsychological interpretation (Response to Vera and Simon), Cognitive Science, 17(1), Jan-Mar.
Clancey, W.J. (1994). Situated cognition: How representations are created and given meaning. In R. Lewis and P. Mendelsohn (editors), Lessons from Learning, IFIP Transactions A-46. Amsterdam: North-Holland, pp. 231-242.
Clancey, W.J., Smoliar, S.W., and Stefik, M.J. (1994). Contemplating Minds: A Forum for Artificial Intelligence. Cambridge, MA: The MIT Press.
Collingwood, R. G. (1938). The Principles of Art, London: Oxford University Press.
Dewey, J.  (1981). The reflex arc concept in psychology. Psychological Review, III:357-70, July. Reprinted in J.J. McDermott (editor), The Philosophy of John Dewey, Chicago: University of Chicago Press, pp. 136-148.
Dewey, J.  (1981). The Child and the Curriculum. Chicago: University of Chicago Press. Reprinted in J.J. McDermott (editor), The Philosophy of John Dewey, Chicago: University of Chicago Press, pp. 511-523.
Edelman, G.M. (1992). Bright Air, Brilliant Fire: On the Matter of the Mind. New York: Basic Books.
Gardner, H. (1985). The Mind's New Science: A History of the Cognitive Revolution. New York: Basic Books.
Greenbaum J. and Kyng, M. (1991). Design at Work: Cooperative design of computer systems. Hillsdale, NJ: Lawrence Erlbaum Associates.
Greeno, J.G. (1988). Situations, mental models, and generative knowledge. In D. Klahr and K. Kotovsky (editors), Complex Information Processing: The impact of H. A. Simon. Hillsdale, NJ: Lawrence Erlbaum Associates.
Gregory, B. (1988). Inventing Reality: Physics as Language . New York: John Wiley & Sons, Inc.
Iran-Nejad, A. (1987). The schema: A long-term memory structure or a transient functional pattern. In R. J. Tierney, Anders, P.L., and J.N. Mitchell (editors), Understanding Readers' Understanding: Theory and Practice, (Hillsdale, Lawrence Erlbaum Associates)
Lakoff, G. (1987). Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago: University of Chicago Press.
Lave, J. (1988). Cognition in Practice. Cambridge: Cambridge University Press.
Lave, J. and Wenger, E. (1991). Situated Learning: Legitimate Peripheral Participation. Cambridge: Cambridge University Press.
Mead, G. H. (1934). Mind, Self, and Society. Chicago: University of Chicago Press.
Newell, A. (1990). Unified Theories of Cognition. Cambridge, MA: Harvard University Press.
Ryle, G. (1949). The Concept of Mind. New York: Barnes & Noble, Inc.
Sandberg, J.A.C. and Wielinga, B.J. (1991). How situated is cognition? In Proceedings of the 12th International Conference on Artificial Intelligence (pp. 341-346). San Mateo, CA: Morgan-Kaufmann Publishers.
Schön, D.A. (1979). Generative metaphor: A perspective on problem-setting in social policy. In A. Ortony (editor), Metaphor and Thought. Cambridge: Cambridge University Press. 254-283.
Schön, D.A. (1987). Educating the Reflective Practitioner. San Francisco: Jossey-Bass Publishers.
Smith, R. Mitchell, T., Chestek, R., and Buchanan, B.G. (1977). A model for learning systems. Proceedings of the International Joint Conference on Artificial Intelligence. Boston, MA, pps. 338-343.
Suchman, L.A. (1987). Plans and Situated Actions: The Problem of Human-Machine Communication. Cambridge: Cambridge Press.
vanLehn, K. (1991). Architectures for Intelligence: The Twenty-Second Carnegie Symposium on Cognition, Hillsdale: Lawrence Erlbaum Associates.
Vera, A.H. and Simon, H.A. (1993). Situated action: A symbolic interpretation. Cognitive Science. 17(1) 7-48.
Vygotsky, L.S. (1978). Mind in Society: The Development of Higher Psychological Processes. Edited by M. Cole, V. John-Steiner, S. Scribner, and E. Souberman. Cambridge, MA: Harvard University Press.
Wenger, E. (1990). Toward a theory of cultural transparency: Elements of a social discourse of the visible and the invisible. PhD Dissertation in Information and Computer Science, University of California, Irvine.
Winograd, T. and Flores, F. (1986). Understanding Computers and Cognition: A New Foundation for Design. Norwood: Ablex.
Wynn, E. (1991). Taking Practice Seriously. In J. Greenbaum and M. Kyng (editors), Design at Work: Cooperative design of computer systems. p. 45-64.
Zuboff, S. (1988). In the Age of the Smart Machine: The Future of Work and Power. New York: Basic Books.