Action Science

We focus on the book Action Science, by Chris Argyris, Robert Putnam, and Diana McClain Smith because it contains the major ingredients of the action science approach; we have been able to use the book in the classroom, where it succeeds in making the arguments clear to first-time readers. We believe that the approach deserves an extended presentation because action science is a major strand of development in action research, combining elements from systems theory, psychoanalysis, and organizational behavior perspectives in an overarching approach. It also explicitly takes on the issues of scientific knowledge in the practice of social change. To date, it is one of the best efforts to deal with the relation between AR and scientific method.

Early on in Action Science, the authors state their main objective clearly: “Our focus is on knowledge that can be used to produce action, while at the same time contributing to a theory of action” (p. ix). In so doing, they argue for a link between theory building and theory testing in action as a single repertoire of actions.

Argyris, Putnam, and McClain Smith (1985) recognize that this has rarely been attempted; they work to frame an explanation for this failure by dis­cussing what they call the false conflict between “rigor” and “relevance.” The authors point to the long-standing institutional habit in the social sciences to assume that what is relevant, what touches the real world in known locations, cannot be by definition the source of rigorous knowledge. They argue that rigor and relevance is a false dichotomy. Referring to Lewin’s view that the best way to understand something is to try to change it, they argue that the road to rigor lies in the attempt to apply social theory to social action, a view consis­tent with the philosophy of AR we have laid out in previous chapters.


A key concept and method in AS is confronting. Confronting is a process by which social actors are forced to come to terms explicitly with their own defensive reactions to changes and perceived threats by inquiring into the causes of those reactions and analyzing the consequences of giving into them. Though Argyris et al. (1985) point out that not all defensive reactions have negative consequences, they strongly believe that defensive behaviors are the key causes for the widespread observation that groups often cycle endlessly between conflicting demands, when the only way forward is to confront and resolve the conflicts.

Theory of Change and Stasis

In Argyris et al.’s view, the aim of AS is to create “an inquiry into how human beings design and implement action in relation to one another. Hence it is a science of practice” (1985, p. 4).

But their goals are even more ambitious because they intend to inquire into ( 1) the variables embedded in the status quo that keep it the status quo; (2) the variables involved in changing the status quo and moving toward liberating alternatives; (3) the variables in a science of intervention that will be required if the previous propositions are ever to be tested; and finally (4) the research methodology that will make change possible and simultaneously produce knowledge that meets rigorous tests of disconfirmability ( 1985, p. xii).

By means of this effort, Argyris et al. (1985) seek to develop a “science of practice” (p. 4) through which individuals and groups can be assisted in “creating and maintaining behavioral worlds conducive to generating valid information [under] conditions in which agents can make free and informed choices and feel internally committed to their choices” (p. 77).


Action science takes on the social science bastion of objectivity directly because Argyris et al. (1985) correctly anticipate that the core objections to their formulations will center on this standard positivistic defensive routine. Their response to the objectivity argument is that it is not possible to achieve even the minimal “valid description” (p. xii) until at least some of the defensive routines of the participants have been directly engaged. Because in their view these patterns of defensive behavior can be both functional and dysfunctional, it is not possible to understand behavior until some sorting out of these ele­ments has been undertaken. In other words, empirical description itself is impossible without intervention, a direct attack on the conventional social science position.


Argyris et al. ( 1985) argue for intervention as the principal source of meaningful descriptions on the basis of which a science of action can be built. They invert the conventional social science approach to rigor and relevance by arguing that the standard approach to rigor produces irrelevant, untested, and untestable propositions. “We will argue … that theory that intends to con­tribute to practice should have features that differ from those of theory respon­sible only to the criteria of pure science” (pp. 18-19).

In their view, what makes the human sciences unique is that they study a group of people in practice, and the action scientist is a practitioner engaged in parallel processes of practice, reflection, defensiveness, and objectification with these cosubjects (p. 22). In the end, their goal is no less than the foUowing:

Action science is centrally concerned with the practice of intervention. It is by reflecting on this practice that we hope to contribute to an understanding of how knowledge claims can be tested and justified in practice and of how such inquiry is similar to and different from that of mainstream science. (p. 35)

Espoused Theory and Theory-in-Use

In developing the argument for AS, Argyris uses concepts and theories devel­oped in a host of previous works, including some written with Donald Schon (Argyris & Schon, 1978, 1996). Among them are espoused theory, theory-in­use, single-loop learning, double-loop learning, and Model I and Model II theories of action. We develop these notions briefly here because they are fundamental elements in the infrastructure of AS as a form of practice.

The espoused theory/theory-in-use terminology does not refer to newly discovered concepts, but rather names well-known ideas that are important in any kind of competent social research. Espoused theory refers to the account actors give of the reasons for their actions. Theory-in-use refers to the observer- analyst’s inferences about the theory that must underlie the observed actions of the same people if their actions are to be made sense of. Often, espoused theory and theory-in-use do not coincide; occasionally, they are directly at odds with each other.

These are not new distinctions. In anthropology, they have been rendered as the distinction between emic and etic approaches. Historical materialism poses the same issue in terms of ideology and infrastructure as does Gramsci’s (1975) use of the concept of hegemony. What is new about AS is that the dis­tance between the espoused theory and theory-in-use becomes the focus of attention in a group’s inquiry into its own actions as a means to try to move the group to a more liberating dynamic.

Single-loop learning refers to a situation in which people or organizations alter their behavior but do nothing to change the behavioral strategies that gave rise to the problematic situation initially. The problem situation is taken as given, and the participants improve their ability to solve specific challenges. The effect is to achieve, possibly, a brief amelioration of a problem, but because the underlying causes are not confronted, the problems return. They persist and regain strength as soon as another dilemma is encountered.

By contrast, double-loop learning results from responding to a problem by stepping back and examining alternative larger frames into which the problem can be put. The immediate problem is understood to be the product of a con­text that itself must be altered. By altering this context, a group can move to a new plane of OL and change. Action science generally views the persistence ot single-loop learning as the product of defensive reactions and improper infer­ences about the motives of others.

The import of the single-loop/double-loop distinction is that it identifies cer­tain kinds of problems as those toward which AS interventions should be aimed.

These are “problems that persist despite efforts to solve them      [They] are likely to have double-loop issues embedded in them” (Argyris et al., 1985, p. 87).

Linked to these two kinds of learning are theories of action. In what Argyris et al. ( 1985) call “Model I,” the underlying model is based on having unilateral control over others. Few people espouse Model I, but many people practice it. Another theory of action is “Model O(rganizational)-I.” This kind of theory of action gives rise to a limited learning system that corrects errors that cannot be hidden and do not threaten the group’s underlying norms. Here the center is broad participation, a focus on win-win approaches, and a strong emphasis on expressing feelings while suppressing intellectual analysis.

Counterposed to these are “Model II” theories of action. In Model II, there are “minimally defensive interpersonal and group relationships, high freedom of choice, and high risk taking. The likelihood of double-loop learning is enhanced, and effectiveness should increase over time” (Argyris et al., 1985, p. 102).

Model O(rganizational)-II is the same, but the individuals making up a collectivity are acting out Model II theories-in-use. The result is the creation of a community of inquiry in which issues and conflicts can be opened up and in which both single- and double-loop learning occurs.


One of the most interesting features of AS is its strong attention to meth­ods for developing tests of interpretations. The perspective is based on an extensive development of the kinds of concepts used in attributing reasons for people’s behaviors, assigning causal responsibility, and achieving intersubjec­tive agreement about the data.

A technique called the “ladder of inference” is used to link subject dia­logues, interpretations, and actions into an analyzed interpretation of interac­tions: “In AS we deal with this issue with the help of a conceptual device, the ladder of inference. This is a schematic representation of the steps by which human beings select from and read into interaction as they make sense of everyday life” (Argyris et al., 1985, p. 57).

In constructing this analysis, the first round is utterances from ordi­nary speech in a specific situation. Then the observers and participants assign meanings of the utterances (both their own and those of the people they are dealing with). These meanings are then examined and compared, and the inferences used to arrive at the meanings are analyzed. The ladder of inference refers to the connecting links of analysis between the utterance and the inter­pretation arrived at. Typically, most people make very powerful inferences about the aims of others on the basis of shaky data. The point is to move up and down the ladder together with the actors in the situation being examined, checking how conclusions are drawn, what is paid attention to, and what is ignored. From this, the patterns of behavior leading to a persistence of single­loop learning surface and can be examined.

Note that the ladder of inference is a technique of organizational inter­vention, not a mere research tool. It is applied because a group has a problem that it has commissioned an action scientist to help it try to solve. Intervention is not at the opposite end of some continuum leading from action to research.

Rather, as Argyris et al. (1985) put it, “Intervention is the AS analogue of experimentation” (p. 64). Without intervention, there is no AS! The Lewin and Dewey inheritance is clear.


One of the most outstanding features of Argyris et al. ( 1985) is a fascinat­ing critique and reformulation of the famous experiments by the U.S. psycho­logist Stanley Milgram (1974) on the willingness of ordinary people to inflict harm on fellow human beings. By commenting on the Milgram experiments and distinguishing their strategy from Milgram’s, Argyris et al. succeed in showing how different a science of action would be from conventional social science, even on socially relevant subjects.

Milgram used people recruited by an advertisement putatively to teach an experimental subject some word associations. Whenever the subject failed, the teacher was to administer electric shocks, and the shocks increased over the course of the experiment to dangerous levels. In fact, Milgram and the experi­mental subject were secretly collaborating, and no electric shocks were admin­istered. The “teachers” did not know this, however.

Milgram interviewed the teachers beforehand, and all asserted that they would not knowingly harm a fellow human being. Though Milgram found lots of variation in people’s reactions, manypeople were in fact willing to shock the experimental subject. From this, Milgram concludes, without a dear line to the data, that humans are not intrinsically hostile or aggressive, but rather are weak willed and prone to follow orders by those in authority. As a result of this find­ing, the work was dubbed the “Eichmann experiments.” Milgram reported his findings almost 10 years after completing the work, but made no social inter­vention other than writing his book.

In commenting on Milgram’s (1974) work, Argyris et al. (1985) are- respectful of his accomplishments and yet distinguish their approach from his. “In order to reliably describe some phenomenon, one ought to retain its essen­tial features and construct a situation that captures its essence” (p. 111).

In Argyris et al.’s view, Milgram did not try to alter the situation and its outcomes. This lost him the possibility of understanding the genesis of the observed behavior. As a result, his experiments could not yield knowledge that might help individuals break out of this dilemma. We never learn of alterna­tives that might better manage it, and we do not discover the deep structures that maintain it.

Argyris et al.’s Model II approach would have been to alter the parameters of the experiment to change the outcomes. Because what is socially desired is a population in which no one is willing to follow immoral orders, Argyris et al. argue that AR should focus on the disobedient teachers and inquire into the causes of their unwillingness to follow such orders. From this, theories can be developed about the causes of disobedience, and the experiment could be var­ied to increase the causes of disobedience until the maximum disobedience is achieved. In this way, a Model II inquiry both inquires into the causes ofbehav- ior and intervenes directly to promote morally desirable behavior in the research subjects. The distance between this and conventional social science is dear, and the basis of conventional social science itself is revealed to be single­loop, Model I behavior.

Practicing Action Science

Not content simply to lay out abstract theories of AS, Argyris et al. ( 1985) also formulate a number of rules of practice that guide their AR. Paralleling these are a series of rules for testing hypotheses, a subject almost never broached in the AR literature. Whatever one thinks of this version of AS, Argyris et al. are correct in making attention to scientific reasoning a higher priority.


There is much that is useful in this framework. Rather than justifying action by using some kind of ethical argument about dealing with social prob­lems, Argyris et al. ( 1985) argue for AS as a better form of scientific inquiry than conventional science. They also pick up the core of Dewey’s and Lewin’s arguments that it is through action that learning can occur. Thus, in these authors’ view, to be scientific, social research must be socially engaged. To put it in the frame that Argyris has long used, the aim of AS is to increase the pos­sibility of unlikely but socialiy beneficial (liberating) outcomes. They want to achieve this through the deployment of the scientific method. The core logic of their argument is that any social research that is not interventionist cannot be scientific, an argument they make quite effectively in their analysis of the Milgram experiments.

The reverse side of this logic is a less dearly expressed but equally severe judgment of many other action researchers. In Argyris et al.’s view, too many action researchers routinely accept the separation of thought and action that characterizes conventional social science, choosing to justify their work by the urgency of the problems they study or the goodness of the goals they have. More bluntly, the greater part of AR is characterized by foggy epistemologies and inco­herent or careless methodologies. By the logic of AS, this is itself a Model I single-loop behavior and will not produce a successful community of inquiry.


Psychologism, Defensive Routines, and Intervener Paternalism

For us, AS takes a very narrow cut of the complexities of human psychol­ogy, even though we welcome its analysis of motivation and behavior at this level. In the main, human psychology, as relevant to action, is reduced to the production of defensive routines leading to single-loop outcomes. It does not seem plausible to us that defensiveness is the only major psychological process relevant to these group phenomena. The richness of human motivations, the complex interactions between cultural ideas and the economics and politics of particular situations, and the complex differences among all the partici­pants in a particular situation are not explored if the analysis focuses only on defensiveness.

It is unclear how the action scientists themselves overcome this defensive­ness. According to their own view, defensiveness is the “default” form of human action (they describe the Model I responses of individuals as “natural” and “automatic”; Argyris et al., 1985, p. 151). This assumption is quite important because it creates an unexplained gulf between the facilitator and the subject.

Left to their own devices, participants would be unable to redesign the Model I predispositions that lead to repetitive failures. Rather than continue to feel frustrated and hopeless, they might decide that it is impossible to produce Model II action and thereby justify their withdrawal; or they might decide that some Model I strategies are as good as could be expected, and not focus on their counterproductive features. In other words, the defenses that enable people to remain unaware of their theories-in-use in the Model I world would reassert themselves. The task of the interventionist is to help participants begin to redesign their theories-in-use genuinely (Argyris et al., 1985, p. 338).

No justification is ever given for this state of affairs, nor is any explanation provided about the sources of the interventionists’ “unnatural” human capac­ities to overcome these limitations. Argyris et al. appear to be natural-born action researchers in this account.

In speaking of choosing to change, they state that although individuals have no choice in their theory-in-use and the 0-I learning system, they can choose to alter their theory-in-use and, hence, the OL system and culture. But such changes will not occur unless the players are committed (p. 152). But why individuals have no choice, what commitment is, and how commitment devel­ops are not discussed.

This view is highly charged politically because it forces us to conclude that action scientists are different kinds of human beings from natural ones. Although this view might be justified through a discussion of the process of people becoming trained to be action scientists, AS does not contain such a dis­cussion. We are left with the action scientist as an unchallenged, self-conscious, and self-contained individual capable of acting on others.

Another way of looking at this problem in AS is to note that the analysis has a strongly dyadic bias. That is to say, although many of the examples occur in group contexts and Model I and Model II refer to group behavior, the pre­dominant image that emerges from a reading of AS is that of a skilled practi­tioner or teacher confronting a group member and getting that group member to inquire into and change his or her behavior.

We believe that this image of the teacher and student, the therapist and the patient, though having the merit of focusing attention on some of the psy­chological dimensions of group processes, also incurs significant costs. Until recently in AS, we did not see an analysis of groups as groups. Rather, groups were assumed to function well when all the dyads in them are functioning well. Notions of group structure, political economy, gender differences, ethnicity, and the like have been left out. Groups are portrayed as being constituted of individuals engaged in exchange and as ideally moving toward some kind of rational choice model. In particular, this makes the analysis insensitive to power relationships, including the power this approach bestows on the expert intervener.

This matters, not because there are no good action scientists; indeed we have both seen marvelous AS practice. Rather, it matters because we think the good practice we see arises, in part, because these practitioners have a more sophisticated social theory than they articulate in their writings and that they practice in a less hierarchical way than the model suggests. Developing a better analysis of these elements is a pending assignment for AS.

These issues of hierarchy matter a great deal because AS proceeds by iden­tifying problems and agreeing when solutions have been found. Thus, the authority to make these decisions is central. Deciding what kind of behavior is appropriate, for instance, is crucial. Yet, for example, in referring to “brittle­ness” in social relationships, action scientists define it as a “predisposition to express an inappropriately high sense of despair or failure when producing error” (Argyris et al., 1985, p. 156). What is inappropriate and who decides is not discussed; this apparently is a decision to be made by the intervener. Argyris et al. also speak of “genuine organizational change” without defining what is genuine, again permitting the intervener the authoritative position of deciding what constitutes change and what does not.

The way in which the ladder of inference is used and the way segments of dialogue are separated for analysis also reveals a highly rule-based vision of culture, a view supported by Argyris et al.’s repeated use of the term routines to describe behaviors. Although many schools of thought do this, including com- ponential analysis in anthropology, there are significant limitations to such a view. Behavior is more than rules, just as language is more than grammar. This is particularly important because AS’s effort to be scientific requires some sort of notion of “disconfirmation” as part of the approach. But the discussions these authors give of disconfirmation rest heavily on this rule-based view of behavior, giving us little sense of the experiential difficulties of a disconfirma- tion strategy when the flow of human behavior is viewed in its ethnographic complexity.


M ofus who are posing critiques of conventional social science approaches must have an explanation about why such conventional social science prevails. If we are right and conventional social science is wrong, then we must explain why it is dominant and we are not. We have already devoted attention to this issue in the present book (see Part 2).We only point out that AS lacks an explanation why the social sciences chose to mimic the natural sciences rather than Dewey and Lewin (Argyris et al., 1985, p. 5) or why there exists what Argyris et al. call “pernicious separation” of theory and practice (1985, p. 7).

In our view, this inattention to larger issues of social structure and politi­cal economy stems from the same dyadic and therapeutic view we commented on above. Action scientists assume that people are misguided and can be brought back to a better view of the matter through high-quality intervention. This ignores the existence of the whole political economy of social research that always moves in the direction of blunting the reformist and democratiz­ing elements in social science for reasons that seem better explained by matters of power than by “defensive routines” of the members of particular groups.

None of these criticisms is unanswerable, and some of them have been addressed by action scientists in recent publications. This framework has gone farther than any other in trying to address some of the methodological and epistemological issues raised by the notion of a science of AR, and deserves close attention for this reason.

Source: Greenwood Davydd J., Levin Morten (2006), Introduction to Action Research: Social Research for Social Change, SAGE Publications, Inc; 2nd edition.

2 thoughts on “Action Science

  1. Carlo says:

    Great beat ! I wish to apprentice while you amend your site, how can i subscribe for a blog website?

    The account helped me a acceptable deal. I had been a little bit acquainted of this your
    broadcast offered bright clear idea

  2. Dominick says:

    Hi i am kavin, its my first time to commenting anyplace,
    when i read this post i thought i could also create comment due to this good piece of writing.

Leave a Reply

Your email address will not be published. Required fields are marked *