I f  a policy is a solution, “actionable knowledge is the 
actual behavior required to implement the solution”
Chris Argyris, whose th irty  books include On 
Organizational Learning, is the James Bryant C onant 
Professor o f Education and  O rganizational Behavior 
in the G raduate School o f Business A dm inistration 
at H arvard  U niversity
4 55
Chris Argyris
ORGANIZATIONAL BEHAVIOR
Production of Knowledge for Action in the World of Practice
O C IA L  SC IEN T ISTS HAVE B E E N  IN T E R E S T E D  FOR MANY will use to implement that knowledge. The seamless quality is dictated 
years in producing knowledge that is relevant to everyday by the fact that in order for practitioners to implement scientific 
life. In my field it is called knowledge that has “external” va­ propositions they will have to use the concept of causality that the re­
lidity, relevance beyond the context in which the knowledge searchers used to produce the propositions.
was obtained (Campbell and Stanley, 1963). Yet a scholarly In order to illustrate this claim, I begin with a few words on the 
reviewer, in a scholarly journal, recently concluded that the meaning of actionable knowledge in the domain of practice to which 
breach between social scientists and citizens, created by the nonap­ I refer. The domain is understanding and solving complex problems 
plicability of the knowledge produced by the scientists, is growing in organizations (technical or human) that are embarrassing or 
larger. He notes that Representative George E. Brown, Jr., who is rec­ threatening. I select this domain because that is where practitioners 
ognized as one of science’s great champions, has written a report that need help the most. They learn how to solve simple, nonembarrass­
warns scholars about continuing their apparent sense of unconcern ing, nonthreatening problems as a matter of course.
about the breach (Johnson, 1993). Actionable knowledge is that knowledge that practitioners use in 
I believe that Johnson and Brown are correct. Scholars, by and everyday life to produce their intended consequences. This means that 
large, espouse relevance and actionability but then pay little attention I am focusing on human behavior in organizations where human be­
to these features when they conduct their research. For example, al­ ings are interacting, doing things to and with each other, attempting to 
though most books and articles on research in organizational behav­ influence, striving to govern the formal and informal features of orga­
ior discuss relevance (usually in a final chapter), none, that I could nizations. It m aybe helpful to think of discovering problems, inventing 
discover, give actionability the same status as validity. The advice they solutions, producing or implementing the solutions, and evaluating the 
give to researchers assumes that with decent validity, and with the ac­ effectiveness o f the action. Actionable knowledge is not a policy; a pol­
cumulation of knowledge, actionability will follow. This assumption, icy is a solution. Actionable knowledge is the actual behavior required 
in turn, contains a more questionable one, namely, that there will be a to implement the solution. The test of actionable knowledge is that, if 
relatively seamless relationship between the causality that researchers implemented correctly, it will lead to the consequences that it specifies. 
used to produce their knowledge and the causality the practitioners The test of whether individuals (or an organization) have actionable
5 6  C h r i s  A r g y r i s
knowledge is whether they can produce the consequences that they in­ very often, by easing-in or being indirect and by covering up the eas­
tend by following the dictates of that knowledge. ing-in. The cover-up is sensed by the receivers, which leads them to 
mistrust what is going on. They too cover up their doubts. The first in­
Theories-in-Use in Everyday Life dividual may then infer that the other is covering up. Of course, that 
inference is covered up. All this occurs rather naturally, spontaneously. 
The information science revolution combined with new concepts The actions are taken for granted and hence there is little awareness of 
from clinical and social psychology suggests the following way to con­ them while they are being produced. This, in turn, makes it more likely 
ceive of human nature. Human beings learn and store in their heads, that they will blame others if something appears to be going wrong. 
early in life, master programs to design their actions and to imple­ This triggers and reinforces the defensive self-fueling processes.
ment their designs.
The master design to deal with embarrassing and threatening issues Organizational Defensive Routines
has been modeled. The model (called Model I) claims to capture the 
theories of action that people use (hence their theories-in-use) as com­ Organizations populated with individuals using Model I theories-in-use 
pared with those that they espouse. According to Model I, a theory-in- create organizational cultures that reward and reinforce Model I. Organ­
use contains a set of governing values. They are to (1) unilaterally izations soon become limited learning systems in dealing with problems 
control the situation you are in, (2) strive to win and not to lose, (3) that are embarrassing or threatening (Argyris and Schon, 1978).
suppress negative feelings, and (4) be rational (Argyris and Schon, The most powerful organizational features that produce limited 
1974). This model is essentially a limited learning model. It requires the learning are organizational defensive routines. Organizational defen­
recipients to be submissive and dependent. When they respond, they sive routines are any actions, policies, or procedures designed to pre­
too will use the same governing values. Defensiveness will escalate as vent the players from experiencing embarrassment or threat and, at the 
well as misunderstanding and errors. The results will be self-fueling same time, to prevent individuals from discovering the causes of the 
processes that are counterproductive to learning (i.e., to detecting and embarrassment or threat. Organizational defenses are anti-learning 
correcting error). and overprotective.
Moreover, given the defensiveness in the relationships, the actors are Mixed messages are an example of defensive routines. “Mary, you 
not likely to receive valid feedback or to trust the feedback. They will run this department but check with Joe when you make tough deci­
therefore be unaware of their responsibility in causing these self-fuel­ sions.” “ Bill, be creative and innovative but be careful.”
ing processes. They will also hold others responsible because they can The theory-in-use behind all mixed messages contains the follow­
point to the others’ defensive behavior. ing logic and rules for implementing them: Communicate a message 
In our research, we find that most human beings are unaware of the that is mixed. Act as if the message is not mixed. Make the first two 
discrepancy between their theory-in-use and their espoused theory, rules undiscussable. Make the undiscussability also undiscussable 
especially when they are actually producing it. For example, many in­ (Argyris, 1990). For example, it is not likely that someone will say that 
dividuals espouse caring and support. They implement these ideas, he or she is about to deceive, about to manipulate, about to distort be­
cause that makes them vulnerable to the charge of acting unethically. 
Moreover, to admit and then to deceive is not to deceive; to admit and
O r g a n i z a t i o n a l  B e h a v i o r : K n o w l e d g e  f o r  A c t i o n  5 7
then to manipulate is to reduce the likely effectiveness of manipula­ There is another feature of the advice in the literature that is even 
tion; to admit and then distort is to tell the truth. more troublesome. The advice actually reinforces the very defensive 
The most powerful consequences of defensive theories-in-use and routes that cause the problems, yet the advisers are apparently un­
organizational defensive routines are that they create double binds. If aware that this is happening.
individuals tell the truth, they are likely to open up an organizational For example, a CEO of one of Americas largest corporations de­
can of worms. If they cover up and cover up that they are covering up, scribed how he sought honest feedback from his immediate subordi­
they are aware that they are living a life of dishonesty. Whistle-blowers nates. Periodically, he would assign as chair of a meeting an executive 
are often in that position. whom every one of his subordinates trusted. The executive then gave 
Most human beings whom we have studied feel helpless in such a the CEO feedback without violating any confidence. This strategy as­
context. They eventually develop a sense of cynicism about organiza­ sumes that the subordinates are not able to be candid with their CEO; 
tions ever changing. Soon they distance themselves from their “dis­ it assumes further that the first assumption should not be tested with 
honest” going-along behavior, which in turn distances them from the subordinates. Finally, it assumes that all this should be done by 
their fair share of responsibility for organizational health. They learn covering up the defenses that make all this necessary.
to live with these second-order consequences by holding organiza­ A CEO advises other executives to build trust. He tells of a colleague 
tions responsible. Hence, we now have an additional and massive set who often confided in him as to how “lousy” he thought one of his as­
of self-fueling processes. sociates was. The CEO admitted that he wondered what that colleague 
might say about him to others. The CEO mistrusted the colleague but 
Advice to Overcome Defensive Routines never told him so; indeed he covered up his feelings of mistrust.
As I reviewed the advice of scholars who described organizational de­ Such actions are sanctioned by organizational defensive routines. 
fensive routines on how to overcome them, I found little that was ac­ Once taken, they feed back to reinforce the defensive routines. We are 
tionable. For example, in the government sector the readers were back to the self-sealing, self-reinforcing processes described above.
advised to persuade effectively, to invoke national interest, to negotiate, 
to treat symptoms because they will relieve pain. If such tactics do not Model 1 and Rigorous Research Methods
work then strengthen the controls, provide monetary incentive and re­
duce red tape, decentralize, and deregulate. I submit that the advice to Why is it that scholars describe organizational defensive routines in 
reduce red tape is not news; that the issue for the actors is how to follow business, governmental, and educational organizations, decrying their 
the advice, and how to make sure it leads to positive consequences. The counterproductive consequences, yet, they end up providing advice 
literature ignores this problem of implementation. Hence the advice is that is either not actionable or consistent with the very defensive rou­
not actionable (Argyris, 1993). tines that cause the problems?
The same is true for the literature from the private sector. For ex­ One answer is that the scholars are also human beings and their 
ample, executives advise: get good people, use a disciplined approach, personal theories-in-use are also consistent with Model I. When they
generate collective responsibility, focus on the future, be polite but 
candid, question intensely. Again, this type of advice is not news, nor 
is it actionable.
58 C h r i s  A r g y r i s
are acting in everyday life they too create defensive routines that in­ are committed to the study of what is, not what might be. The latter is 
hibit reflecting on their actions as well as their community norms. If normative research and not typically a central feature of basic research.
one analyzes the actions of scholars as they compete against each There are two ways in which the emphasis upon descriptive re­
other (Watson, 1969) and as they coerce conformity upon those who search is limiting. If the universe is dominated by defensive theories- 
work with them (Mitroff, 1974), it is not difficult to see how Kuhn in-use and organizational defensive routines, then careful description 
(1962) arrived at the conclusion that scholarly communities were will produce only countless descriptions of individual and organiza­
loaded with community defensive routines that were protective of tional defensive routines (as was found in the literature review cited 
those in power and inhibited genuine confrontation of new ideas. earlier). Social science researchers become the servants of the status 
There is a second answer to the question. Methods of scientific re­ quo because they do not conduct research on changing defensive rou­
search are theories of action. If one examines the theory-in-use that is tines. They do not conduct such research because to execute it re­
the basis for rigorous (positivistic) research, be it experiments, ques­ quires a theory of a new type of organization (a normative position) 
tionnaires, or interviews, one discovers that it is consistent with Model and specification of how to get from here to there (an intervention 
I. For example, the researchers strive to be in unilateral control over position, which requires them to go beyond being descriptive).
the research in order to maximize their winning and minimize losing. In order for such research to be produced, it would have to be re­
They systematically cover up their intentions and objectives from search that studies, or more likely creates, rare events. The events are 
their subjects in order to reduce the subjects’ contamination of their rare in the sense that researchers would be producing organizational 
actions or replies. phenomena that do not presently exist. In order to do that, they would 
Scientists, while in the process of research, also cover up that they have to develop normative views of reality—views of what the world 
are covering up. This tends to lead them to produce generalizations ought to be.
that are limited to Model I action strategies. For example, the leading The second way in which the research can be self-limiting is related 
research on mass communication advises the practitioners to give to the concepts of rigor used by social scientists to produce valid 
several sides of an issue if the audience is educated and sophisticated. knowledge. These concepts do not produce user-friendly knowledge. 
It also advises giving only one view of the issue if the audience is not In order for knowledge to be user-friendly it must involve concepts of 
sophisticated. In both cases, the advice will work only if the practi­ causality, concepts of implementation, and concepts of assessment 
tioners cover up that they are following either strategy. The advice will that are usable by practitioners in their everyday life contexts. When 
not work if, for example, you tell the audience that the reason they are this requirement is not met, the breach between knowledge and ac­
getting one view is that they are judged to be unsophisticated. A re­ tion, even in the best studies, is large. For example, let us take the clas­
view of the literature indicates that this is not a rare example; indeed, sic study of frustration and regression by Barker et al. (1941).
it is an exemplar of the advice in the current literature (Argyris, 1980).
1. The researchers brought children into an experimental situation 
There is another strategy of rigorous research methodology that un- 
without briefing them about the experiment. They acted in accor­
realizingly reinforces the limited learning features of Model I. Most 
dance with good practice by keeping secret the “experimental ma­
scholars strive to describe and understand their chosen universe. They
nipulation,” namely, frustration. Nor did children make an 
informed decision to participate (although I doubt any were co­
erced if they did not wish to participate).
O r g a n i z a t i o n a l  B e h a v i o r : K n o w l e d g e  f o r  A c t i o n  59
2. They created an unambiguous situation of frustration. They got he include the subordinates? Could not the feedback exacerbate the 
the children to play with and become attached to some toys. Then feelings of frustration for some? If so, how would he find out if the 
they placed a physical barrier between the children and the toys, frustration is increasing?
thereby “causing” frustration. How would the manager strive not to create frustration that is as 
unambiguously clear as the physical barrier was for the children? Or, 
3. They had several observers behind one-way-vision glass scoring 
what behavior might lead to frustration even though the manager 
the actions of the children. They used instruments that were 
wishes to prevent this from occurring? For example, the manager will 
pretested.
have to deny some requests. But how unequivocally clear must the de­
4. They maintained the barrier long enough to collect observations nial be? If she did it with anger, it could lead the subordinates to hold 
that could be used to test their hypotheses rigorously. They con­ her responsible for being ruthless and insensitive, as well as unilateral. 
trolled the time perspective of the children and the experiment. Would not such feelings have an impact on the feelings of frustration? 
For example, what if the subordinates said privately to themselves, “ If 
5. The results led to the conclusion that mild frustration could lead to 
you’re so sure of yourself, you take responsibility for the conse­
creative behavior. Beyond that point the children regressed. 
quences,” and then psychologically withdrew, but covered it up. Such 
Regression led to more primitive behaviors including aggression.
withdrawal could ameliorate the feelings of frustration.
Now let us turn to using this knowledge in real life. It is, I believe, All these features indicate a relationship between the researcher and 
clearly applicable. For example, leaders can be taught that if they frus­ the subjects that is consistent with Model I. The moment the practition­
trate people, the latter will regress and a likely consequence is that ers attempted to create these conditions in their context, they would 
they will become angry and aggressive. have to create relationships that are consistent with Model I. At that 
How does a leader translate this applicable knowledge into actionable point they would be reinforcing the status quo. Moreover, their actions 
knowledge? For example, a leader goes into a meeting to allocate scarce may be seen as disingenuous and gimmicky, if they say they wish to re­
financial resources. He wants to do it fairly, in line with the objectives of duce frustration in order to insure openness, trust, and empowerment.
the organization as a whole and not in line with the parochial views of 
the different subordinates. He also wants to use the opportunity to gen­ What Can Be Done
erate internal commitment on the part of the subordinates to the final 
results so that the likelihood of effective implementation is increased. Producing knowledge about dealing with embarrassing and threaten- • 
He knows that he must be careful to minimize frustration. ing problems in organizations is not an easy task. It requires finding 
How does he find out what sort of impact he is having? One way is ways to create rare events that in everyday life are often viewed as ide­
to ask the individuals. But doing so maybe playing it safe. Another alistic or even dangerous. In order to do so, we need to educate 
way is to ask them to complete a written instrument. But doing so human beings in theories-in-use that discourage defensiveness and 
might frustrate the people more, especially those who believe they are encourage learning. We also need to create organizations that not only 
“winning,” and many see the filling out of forms as a diversion away reduce their defensive routines but reward change in the underlying
from progress. He might use the observational instruments the re­
searchers used by having an observer present. But how would the ob­
server feed back the data? Would he give it only to the leader? Would
6  o C h r i s  A r g y r i s
governing values and policies. We also require a theory of interven­ The concepts of Model I and of organizational defensive routines pro­
tion—how to get from here to there. Finally, we require research meth­ vided the basic framework for the diagnosis. I developed an action 
ods that do not compromise validity and generalizability yet map of the individual, group, intergroup, and organizational defenses 
overcome some of the self-limiting features around actionability de­ that caused such activities as politics at the level of the directors and 
scribed above. the negative consequences that the politics had on creating a gen­
A master program has been developed that, when it becomes a part uinely learning organization.
of the actors’ theories-in-use, will lead to a reduction of the defensive­ The second step was to feed back the map to assess the extent to 
ness of Model I and facilitate underlying learning and change. The which the directors would confirm it. The directors made a few changes 
governing values of this theory-in-use (Model II) are valid informa­ and then confirmed the map. Some expressed bewilderment about how 
tion, informed choice, and internal commitment to monitoring the ef­ they could alter their actions in order to change the map. Others main­
fectiveness of one’s actions. Model II, when used competently, reduces tained that now that they had a map of the causal factors they should be 
organizational defenses, which in turn greatly deepens the learning able to change the map. This created a dialogue that when analyzed (the 
that occurs in organizations around embarrassing and threatening sessions were recorded) showed that they were creating the dysfunc­
problems (Argyris and Schon, 1974,1978). tional features of the map.
A theory of intervention is now available to get us from here to The third step was to provide them with learning seminars to learn 
there. This theory of intervention uses a research methodology that Model II governing values and actions and make them a part of 
values generalizability and validity but gives actionability an equal Model II theory-in-use. (They could still use Model I for the routine, 
status (Argyris, 1980,1993). nonthreatening issues.) The instructional vehicle for the seminar was 
The concepts and skills around discovery, invention, production, a case method designed to diagnose each director’s theory-in-use and 
and evaluation that are used in the research interventions will be the to provide a realistic basis for changing it. The discussion was orga­
same skills that the researchers will teach the practitioners. nized around the cases. Each participant led discussions of his case. At 
Practitioners now become implementers of the concepts and, at the the same time, I helped the group to focus on the way all of them were 
same time, the implementation will become an occasion for the test of inquiring. They not only saw that their cases were Model I, but they 
these skills that are derivable from the theory. saw that their advice to each other on how to reduce the Model I fea­
For example, a group of directors of a consulting firm wanted to tures was also consistent with Model I.
make their organization truly, and persistently, a learning organization. The way out of this dilemma was practice. The seminars provided 
Yet, as they observed their own actions, they realized that they were the practice by using the cases the directors wrote. The participants 
helping to create the very organizational defenses that they abhorred. began to see how to redesign their actions. They tested their designs 
They invited me to help them change this situation (Argyris, 1993). with each other. In learning from each other, they obtained undeni­
The first step was to observe them in action, especially when they able evidence that their comrades were not as antilearning-oriented as 
were dealing with problems that involved embarrassment or threat. each director thought was the case. Otherwise they would have pro­
duced the bypasses, cover-ups, and undiscussables that they illus­
trated in their cases and during the early discussions.
The fourth step was more practice but in everyday organizational 
settings. For example, there was a hot discussion on ownership in
O r g a n i z a t i o n a l  B e h a v i o r : K n o w l e d g e  f o r  A c t i o n  6 1
which all directors participated. There were two-person discussions References
around issues of assignment to clients, the staffing of case teams, the Argyris, Chris. 1980. Inner Contradictions of Rigorous Research. New York:
defensiveness of younger consultants, and who gets promoted and why. Academic Press.
All these sessions were recorded. The respective participants found ----- . 1990. Overcoming Organizational Defenses. Needham, Mass.: Allyn & Bacon.
the transcripts very helpful in reviewing their actions. The researchers 
----- . 1993. Knowledge for Action. San Francisco: Jossey-Bass.
used the transcripts to evaluate evidence of change as well as to design 
future interventions. We could, for example, quantify the behavioral Argyris, Chris, and Donald Schon. 1974. Theory in Practice. San Francisco: 
Jossey-Bass.
changes from Model I to Model II and observe the consequences. One 
----- . 1978. Organizational Learning. Reading, Mass.: Addison-Wesley.
of the most powerfully documented results was that all the problems 
that the directors described, at the outset, as undiscussable and unalter­ Barker, R., T. Dembo, and K. Lewin. 1941. “Frustration and Regression.” University of 
Iowa Studies in Child Welfare 1:1-43.
able became discussable and alterable—for example, making changes in 
the financial ownership and about the directors’ commitment to the Campbell, D. T., and J. C. Stanley. 1963. Experimental and Quasi-experimental Design 
for Research. Skokie, 111.: Rand McNally.
firm. Another result was that most directors attempted to introduce the 
new concepts in the managing of case teams. Finally, some directors Johnson, David. 1993. “Psychology in Washington: Measurement to Improve Scientific 
Productivity: A Reflection on the Brown Report!’ Psychological Science 4(2): 67-69.
began to experiment with their clients. So far, the experiments have 
been carefully designed and successfully executed. This is leading the Kuhn, Thomas. 1962. The Structure of Scientific Revolutions. 2nd ed. Chicago: 
University of Chicago Press.
firm to conceptualize and produce new consulting services that 
Mitroff, 1.1.1974. The Subjective Side of Science. Amsterdam: Elsevier.
promise to add additional value to the advice they give clients.
Watson, James D. 1969. The Double Helix. New York: New American Library.
Concluding Comments
One of the most persistent reactions to the argument above is that it 
disregards the fact that human beings in private, public, and volunteer 
organizations are political animals.
The research suggests that human beings behave politically. But 
there is no evidence that this proclivity cannot be altered. Indeed, re­
search results are just beginning to appear indicating that political pro­
clivity is alterable, as illustrated in the case of the consulting firm. I 
hope that this essay illustrates that true revolution in the management 
of organizations is needed and that it will not occur by changing boxes 
around in organizational charts. Genuine revolution that will persist 
requires changing organizational and individual defensive routines.
In my experience the universities most likely to take this challenge 
seriously are those that take action seriously, such as professional 
schools like ILR. ■