%20-%20https://sketch.com%20--%3e%3ctitle%3e3ECBBF50-FD08-412A-943D-2CFC5E025A16%3c/title%3e%3cdesc%3eCreated%20with%20sketchtool.%3c/desc%3e%3cg%20id='Stylesheet'%20stroke='none'%20stroke-width='1'%20fill='none'%20fill-rule='evenodd'%3e%3cg%20transform='translate(-926.000000,%20-668.000000)'%20id='BUTTONS'%3e%3cg%20transform='translate(100.000000,%20597.000000)'%3e%3cg%20id='BTN-/-RIGHTS'%20transform='translate(816.000000,%2065.000000)'%3e%3cg%20id='ICON-/-Rights'%20transform='translate(10.000000,%206.000000)'%3e%3cpath%20d='M13.5752841,0%20L1.34943182,0%20C0.59375,0%200,0.586666667%200,1.33333333%20L0,22.6666667%20C0,23.4133333%200.620738636,24%201.34943182,24%20L17.6505682,24%20C18.40625,24%2019,23.3866667%2019,22.6666667%20L19,5.36%20L13.5752841,0%20Z%20M13.5752841,1.89333333%20L17.0838068,5.36%20L13.5752841,5.36%20L13.5752841,1.89333333%20Z%20M1.34943182,22.6666667%20L1.34943182,1.33333333%20L12.1988636,1.33333333%20L12.1988636,5.36%20C12.1988636,6.10666667%2012.7926136,6.69333333%2013.5482955,6.69333333%20L17.6235795,6.69333333%20L17.6505682,22.6666667%20L1.34943182,22.6666667%20Z'%20id='Shape'%20fill='%23595959'%20fill-rule='nonzero'%3e%3c/path%3e%3ccircle%20id='Oval'%20stroke='%23595959'%20stroke-width='1.5'%20fill='%23595959'%20cx='9.5'%20cy='13.5355339'%20r='5.5'%3e%3c/circle%3e%3cpolyline%20id='Path-2'%20stroke='%23FFFFFF'%20stroke-width='1.5'%20transform='translate(9.500000,%2013.035534)%20rotate(45.000000)%20translate(-9.500000,%20-13.035534)%20'%20points='8%2016.0355339%2011%2016.0355339%2011%2010.0355339'%3e%3c/polyline%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/svg%3e)
1. Introduction
Teppo Felin and Nicolai Foss provide a provocative critique of past research on routines and capabilities, and seek to point us in a different direction (Felin and Foss, Reference Felin and Foss2011). While I believe I understand most of their specific points, and agree with quite a few of them, I find myself quite puzzled about the underlying presumptions and ambitions of their discussion.
They give an extended account of the intellectual antecedents of the key ideas, an account that receives space commensurate with a considerable and wide-ranging intellectual effort. This account, however, is barely recognizable to me; that is, it corresponds very little to the intellectual history of the subject as I, at least, experienced it. Further, the weakness of the correspondence is something that would be readily ascertained with only a modest amount of checking. This suggests that perhaps the account is not supposed to be read as intellectual history at all; it is, rather, the authors’ theoretical exploration of the question, ‘Where did ideas about routines and capabilities come from?’ In that case, we should ask what grounds there are for accepting this theoretical proposal.
Complementing the analysis and critique, they offer a constructive program for the field. Again, there is a good deal here with which I agree, but much of it also seems to correspond to the actual path of the field in recent years. For some time now, there has been widespread recognition of the need to move beyond the simplest models of experiential learning. As for the featured idea of ‘the poverty of stimulus’, it is harder to interpret concretely in the relevant context, perhaps because the authors do not identify extant examples of this approach that might deserve emulation. The approach may be too fresh for that, or – more likely in my view – the idea is actually more cautionary than constructive. It offers further guidance on what not to do.
My complaint here is reminiscent of the one that we heard from neoclassical economists years ago: ‘You criticize our theories and you propose very different ways of thinking – but what do you have to show that might convince us that these different ways are actually helpful in specific research contexts that are of interest to us?’ But that was many years ago, and now the research program that seeks to respond to that complaint is the one that Felin and Foss criticize, while I am the one to ask the skeptical question about what is really on offer. For me, this is a welcome change relative to the earlier era.
At a deeper level, Felin and Foss seek to ground their discussion primarily in philosophical considerations, as distinguished from considerations of theory and evidence. They do allude to some lines of empirical inquiry that they seem to approve (as distinguished from the ‘empiricism’ they deplore), but for the most part these are remote from the routines and capabilities discussion itself. Thus, for example, they refer to an important line of inquiry about the mechanisms of language acquisition in humans, in which the contributions of Noam Chomsky are of historic importance. Since language is a complex competence, it is easy to accept the idea of a high-level analogy between the challenges of understanding language acquisition and those of understanding capability learning. Certainly it is more than legitimate to explore this. Reflecting that pursuit, the Felin–Foss bibliography contains five references to Chomsky – but on the other hand it is thin and very selective in its treatment of the literature on capability learning, tending to focus on examples close to the learning-curve tradition that are said to be tainted with ‘empiricism’.Footnote 1 As I shall illustrate further, this example captures a pervasive tendency in the paper: it often seems that the topic referenced in the paper's title is only in the peripheral vision of its authors. Focal attention is occupied by more abstract concerns, illuminated by that diverse array of philosophers, psychologists, etc who carry such weight in the bibliography.
Since it appears that we may not really agree on what is or should be under discussion, it could well be that there are some fundamental communication problems here. While difficulties of that kind are notoriously difficult to straighten out, I welcome the occasion to make the attempt. Certainly a portion of the critical view that Felin and Foss present is shared by others, including some others who are themselves active contributors to the routines literature. There is also some detectable convergence regarding desirable new directions. The effort to clear the channels is certainly worthwhile.
In the remainder of this comment, I first examine some of the sources of communication blockage, and then pursue the five ‘problems’ identified in the Felin and Foss critique. Finally, of course, I discuss the new directions.
2. Some possible misunderstandings
The first misunderstanding worth noting is possibly mine, and it is one to which I am inevitably prone: I tend to read the discussion as if the authors were speaking to me personally. While there is a certain amount of warrant for that in the mentions of my work with Richard Nelson, and later with Maurizio Zollo, I am not sure whether I am entitled to object to the proffered characterizations of the ‘literature’ on routines and capabilities merely on the grounds that it does not apply to me. Neither, however, does it clearly apply to Nelson, or to many other friends and colleagues who have worked in this area. I would be hard pressed, in fact, to think of anyone to whom the broad-brush characterization clearly applies – although some of the narrower points are at least backed by relevant quotations.Footnote 2 This raises the suspicion that the point of view criticized may be in large part the creation of the authors, or at least that the specific targets are poorly identified. In any case, while I seek to respond to the critique on behalf of the wider community whose contributions are referenced and/or neglected by Felin and Foss, I offer apologies in advance for any overly-personalized interpretations, as well as for my possible failure to appreciate fully the larger agenda that seems to motivate the authors.
A second possible misunderstanding relates to the purposes for which we study organizational routines. At least in evolutionary economics, the concepts of routines and capabilities were put forward as components for a theory of organizational behavior, and that theory is itself a descriptive theory, i.e. a predictive/explanatory theory, not a normative or prescriptive one.Footnote 3 It is not claimed that learned routines are the only things driving behavior, or the only thing worth studying. It is certainly not claimed that organizations should avoid intentionality, or creativity or the exercise of whatever free will they or their leaders may have. On the contrary, the concept was developed partly to provide a realistic backdrop against which these other sorts of activity could be examined.
What is true, however, is that the theory is intended to be an operational one that can be implemented on the basis of empirical observations that might actually be feasible in at least some cases. This accounts for the emphasis on observables, to which Felin and Foss object, including the emphasis in the empirical literature on the very simple device of counting the number of times the actor has dealt with that sort of situation. I am not aware of any empirical researcher who has proclaimed that ‘counts are plenty good enough for me’; what they affirm is that counts are available, theoretically interpretable and prove to be capable of explaining variance in performance. The admitted conceptual deficiencies of the counts measure actually offer an empirical researcher some indirect long-run encouragement for the larger research program – you might not expect anything so simple to work at all; who knows what better data might unveil? These researchers would probably be inclined to ask Felin and Foss what variance they seek to explain and what observables they think might explain it; I doubt that they would put free will, or even creativity, forward in that context.
While routines are certainly not the only things going on inside business firms, they do have distinctive appeal as an element of a theory that aspires to prediction. Unlike the inner workings of a creative mind, or the subtleties of organizational politics at the highest level, they are relatively accessible to observation. Further, there is reason to think that close observation of the routines in use at period t might actually have a payoff in terms of successful prediction of firm behavior in period t+1.Footnote 4 In our 1982 book, Nelson and I pushed this point further in a concluding statement at the end of chapter 5. We offered a broad, routine-centered formula for approaching the problem of predicting firm behavior, a formula that has a number of important nuances (Nelson and Winter, Reference Nelson and Winter1982). To our critics I would say, what specifically is it that you propose to substitute for that recipe, and what is the basis for thinking that it is both feasible and an improvement? (Needless to say, actual improvements could only rise from an operationalized version of a rival approach, which could be expected to present its own difficulties.) It appears to me here, as it has on other occasions over the years, that this question is not actually being addressed. If, hypothetically, we are aspiring to do empirical science in this area, is that not a key question to address?
I do think of successful empirical science as being the goal, though such success can be pursued in many ways. The observations of Felin and Foss leave me uncertain as to where their favored ‘internalist or rationalist approach’ lies in relation to modern ideas of scientific inquiry. What are we to make of a passage like the following?
Unfortunately, behaviorist and empiricist models in organization theory have been closely linked to natural sciences with the emphasis placed on such things as incremental and experiential evolution, randomness, and environmental selection (e.g., Nelson and Winter, Reference Nelson and Winter1982; also see Holland, Reference Holland1995), and thus the evidence of ‘man's artifice’ essentially gets lost in ascribing causal roles to the artifice rather than the man (Felin and Foss, Reference Felin and Foss2011: 14).
I take some pride in this complaint. ‘Unfortunately, . . . closely linked to natural sciences’? Such linkage is a very good thing, in my view, and the real question is who does it best. The ‘emphasis placed’ seems in the main to be appropriately placed, although there is an implicit issue here, which is discussed below, about the emphasis that should be given to prior biological evolution.
There may be a related misunderstanding involving the meaning of ‘cause’ and ‘causal explanation’. Felin and Foss argue that experience and repetition are not a fundamental cause of capability:
We argue that experience and repetition are only an epiphenomenon of more ultimate theoretical causes of behavior; experience and capability have endogenous origins. Citing experience and repetition as causes of behavior and capability only restates rather than resolves the theoretical problem of understanding and explaining organizational behavior or heterogeneity (2011: 7).
Here we have a philosophical issue of the kind that presents a nice target for the sort of analysis the authors would like to promote. The object of this analysis should be, ‘Experience and repetition are not a true cause.’ What does that mean, exactly? Consider the case of a concert pianist who spent years at the keyboard practicing, but not, thereby, promoting her chances of professional success – according to Felin and Foss. This history of experience and (tedious, let me tell you) repetition was not a ‘final’ theoretical antecedent of her musicianship, according to the authors. In their favor, we should acknowledge that there are other things to be said about the situation – mentioning innate aptitudes, musically-inclined parents, hormones and upbringing supportive of high diligence, etc – but reference to these only ‘restates rather than resolves the problem’: whence came these? Most importantly, reference to these important considerations hardly seems to devalue the significance of ‘Practice, practice, practice!’ as a causal explanation for how this person got to be on stage at Carnegie Hall.Footnote 5 The strangeness of ‘only an epiphenomenon’ in this case derives in part from the fact that much of the required antecedent behavior was not observably adjacent to the effect at all – and a corresponding point extends also to the authors’ dismissal of experience in relation to organizational learning.
The authors say ‘By an epiphenomenon we refer to the question of whether the theoretical explanation given provides causal finality, or whether the explanation induces immediate questions about the nested, actual antecedents.’ (2011: 7) I take the view that causes operate in time; everything has causal antecedents, and causal antecedents are, in a broad sense, evolutionary antecedents. In this perspective, ‘causal finality’ may be hard to reach, but the explanatory regress reaches a terminus at the Big Bang, somewhere around 14 billion years ago. This regress embraces all of the considerations that the authors offer as ‘final’ causes, including the levels of intentionality and creativity displayed in Homo sapiens. Why they believe that human intentionality is a ‘final’ cause, which does not itself require explanation, or cannot have a scientific one, is unclear to me.
3. Bibliography gaps
I have referred above to the heavy bibliographic weight carried by references that seem fairly remote from the routines and capabilities topic, and the correspondingly light weight of those that are plainly on point. A broad range of literature is referenced. I confess to being unfamiliar with perhaps the larger proportion of it, outside of the domain of obvious relevance, though familiar with some. A portion of the unfamiliar part seems, at the title and author level, to be highly relevant; in other cases I am more doubtful that this is so. On the other hand, I have a reasonable sense of the routines and capabilities literature, including contributions that were important in the background of my work with Nelson, and a range of subsequent contributions up to the present that have supplemented, tested and in some degree modified the case that Nelson and I presented.Footnote 6 As noted previously, very little of the topic-specific bibliography is acknowledged in the Felin–Foss argument. Some of the items that are missing are quite central, in my view, to the concerns raised. Thus, it seems to me that part of the difficulty I have in recognizing this map of the intellectual territory is attributable to the fact that not much familiarity with that territory is displayed.
The leading example of these difficulties is the remarkably strong emphasis given to the idea that the concepts of organizational learning that appear in the routines and capabilities literature are close cousins to the ideas advanced in behavioral psychology by B. F. Skinner, Edward Thorndike, John Watson and others.
Overall, what is quite striking about the routines and capabilities-based literature is that it utilizes the exact same inputs, mechanisms and theoretical arguments as the psychological literature that it builds on. For example, Schwartz's (Reference Schwartz1978) book on psychological learning illustrates that the same exact input–output relationships and theoretical mechanisms are featured both in behaviorist psychology and much of the organizational literature on routines and capabilities (Felin and Foss, Reference Felin and Foss2011: 6, emphasis in the original).
Here it is important to note that the ‘illustrates’ judgment derives entirely from Felin and Foss, not Barry Schwartz – whose Reference Schwartz1978 textbook (1st edn) obviously predates most of the relevant literature on capabilities. While the bibliography contains multiple works by behavioral psychologists (three by Skinner), the bibliography of the Nelson and Winter book contains none of these names. However, the latter work does cite some psychologists in chapter 4 (Skills) and chapter 5 (Organizational Capabilities and Behavior), particularly Roger Schank and Robert Abelson, and George Miller, Eugene Galanter and Karl Pribam. More important, however, is the debt of chapter 4, and indirectly chapter 5, to Michael Polanyi – not a psychologist, but a writer with vast insight into many things, including the phenomenon of procedural memory that psychologists later defined and studied. There is also a psychologist whom we did not cite in our 1982 book (Nelson and Winter, Reference Nelson and Winter1982) but is now understood as a major figure in the background of the routines and capabilities work, or at least as an earlier author with a helpful and congenial perspective on these issues – John Dewey.Footnote 7 None of these authors, who actually influenced the routines and capabilities ideas at that particular stage of their long history, are mentioned. All are ignored, in favor of authors imagined to have been particularly influential, simply to support the aggressive claim of strong parallels between the capabilities theory and behavioral psychology.
Like Polanyi and Dewey, Nelson and I devoted considerable attention to the many interactions between the relatively automatic and often tacit parts of learned behavior and the more deliberate and truly choice-laden parts that guide the directions of learning and modify learned patterns. The critique that Felin and Foss offer seems to reflect a selective reading of chapter 5 in Nelson and Winter (ibid.). For example, the authors do not engage with section 5.5, ‘Optimal routines and optimization routines’, or 5.6, ‘Routines, heuristics and innovation’. But that is only the beginning of the shortcomings of the Felin–Foss characterization. What happened to chapters 1, 4, 11, 12–15 and much in the chapters in between? On what basis, aside from unsupported reference to the alleged influence of Skinner and others is it possible to read these pages and then return a verdict of neglect of the importance of intentionality, as if the book closely paralleled a study of learning by rats and pigeons?
Perhaps Nelson and Winter (ibid.) is not in the bulls-eye of the intended target for the Felin–Foss critique. But if it is not, what is? The complaints that have little validity for Nelson and Winter (ibid.) have even less validity for our writings since then, or for those of our many collaborators and others who have followed a similar line. On the behalf of the latter, in particular, we do feel some real sense of offense.
It is particularly unfortunate that the broad-brush characterization tars scholars such as Nelson and Zollo with association with ‘rat psychology’ as practiced by Skinner and others. This in fact is the opposite of what these two scholars are about. Indeed, even the titles of their works provide a sufficient clue to the error of the characterization; I will not do here what anyone can do with a quick search on Google Scholar. Similarly, Felin and Foss identify March and Simon's Organizations (March and Simon, Reference March and Simon1958) as an important channel by which the errors of behaviorist psychology infiltrated subsequent literature.
Theoretically there is a strong link between psychological behaviorism (as pioneered by the likes of Thorndike, Reference Thorndike1932; Watson, Reference Watson1934; and Skinner, Reference Skinner1989) and subsequent models of organization (specifically see, e.g., March and Simon, Reference March and Simon1958: 9–10; 139–142) (Felin and Foss, Reference Felin and Foss2011: 3).
While March and Simon (Reference March and Simon1958) do reference Thorndike and other champions of behaviorism, a reader will also find emphasis on the internal states of actors, including memory conditions, and the suggestion that the relation between actor and environment is one of mutual interaction. Thus, the claim of a ‘foundational’ role for behaviorism in the work of March and Simon is not significantly more credible than in the case of Nelson and Winter, although the superficial evidence from citations is less clear-cut. The authors go so far as to cite Zollo and Winter (Reference Zollo and Winter2002) to support the charge of an overemphasis on experience accumulation, saying that Zollo and Winter ‘give primacy to the role of “accumulated experience” as the origin of organizational capabilities’. The claim of ‘primacy’ is quite unjustified, since the paper is actually dedicated to the cause of going beyond simple accounts of experiential learning – as might be suggested by the fact that the first two words of the cited article's title are ‘Deliberate Learning. . . .’!Footnote 8 Zollo and Winter say:
In the spirit of bridging the behavioral and cognitive approaches to the organizational learning phenomenon. . ., we attend both to the experience accumulation process and to more deliberate cognitive processes involving the articulation and codification of knowledge derived from reflection upon past experiences (ibid.: 340).
Also:
One of the recognized limitations of the behavioral tradition in the study of organizational learning consists of the lack of appreciation of the deliberative process through which individuals and groups figure out what works and what doesn't in the execution of a certain organizational task (ibid.: 341).
So much for the primacy of accumulated experience and the alleged neglect of intentionality.Footnote 9
As I briefly discuss below, the discussion of routines has broadened considerably in recent years, and many scholars have concerned themselves with the interplay of routinized behavior and agency. With respect to that development, the Felin–Foss critique is even farther from the mark than in the cases just cited. In short, the Felin–Foss interpretations of texts and of the broader intellectual history are far from trustworthy. There seems to be an underlying commitment to an idée fixe that blinds them to an abundance of obvious evidence that does not fit.
4. The problems
Origins and causation
Providing a fully compelling theoretical account of the origins of routines and capabilities is indeed a challenging goal, and in my view it is correct to say that no such account exists at present. In 1982, Nelson and I largely sidestepped the problem by focusing on the phenomena associated with established skills and routines, while making occasional contrasts with novice performance and also allowing for the contribution of routines to innovation. In this we followed Schumpeter (Reference Schumpeter1934 [1911]), whose discussion of the circular flow precedes his account of entrepreneurship and its correlates – and the latter does not so much explain as simply remark upon the phenomena that entrepreneurial history presents. If, however, we were addressing these matters today, we certainly would be able to do much better on the problem of origins, because so much has been learned in the intervening years – not nearly enough in some absolute sense, but relatively speaking a great deal.
The concerns that Felin and Foss express in this area illustrate some of the issues noted above. For example, they say ‘. . . how one (an individual or organization, any organism) sets out to have a “better” or “quality” (heterogeneous) experience(s) remains unanswered’ (2011: 7–8). This (along with the question about the ‘better’ and the ‘right’ experiences) implicitly invokes the prescriptive perspective on the phenomena, an important area of concern but one that (as previously noted) should be distinguished from the criteria of scientific success. It seems clear, however, that one does not have to ‘set out’ to have heterogeneous experiences; often it is hard to avoid that. As to quality, it is true that ex ante choice often plays a limited role, but so what? I can certainly provide serious causal explanations for observed effectiveness without presuming that the actors involved ‘set out’ to achieve it – as the theory of biological evolution powerfully illustrates.Footnote 10 Similarly, the idea of an ‘infinite regress’ is invoked by way of critique of the idea that capabilities arise from experience. This distorts a question of history or evolution into a question of logical entailment. The origin of today's organizational routines and capabilities lies in the past, along with the origins of the Constitution of the United States, the Earth, and the element carbon. New entities such as these (recognized in the various ontological domains created by H. sapiens) do appear from time to time. Broadly speaking, they all have identifiable evolutionary origins. The challenge is not to ‘unpack’ an alleged regress, but to understand how those specific novelties emerged by the operation of dynamic processes of a more general character, which transcend the narrow ones invoked to explain the relevant phenomena within the established ontologies (e.g., orbital calculations for Earth, organic chemistry).
Next we encounter the rhetorical quest for ‘the first experience’ and the claim that ‘. . ., if experience and repetition is the origin of learning, then logically the experiences themselves, and input and environmental stimuli more generally, should in essence carry the contents of experience or “lessons learned” within them’ (Felin and Foss, Reference Felin and Foss2011: 8). This seems like an attempt to ‘problematize’ the obvious. Presumably the authors have had a lot of recognizable ‘first experiences’ in various domains, at the personal level. It is obvious that these experiences did not impinge on a logical blank slate but actually impinged on a complex, already experience-laden human being. It should be quite clear also that ‘capability’ is not antecedent to ‘repetition’ (as implied; ibid.: 10) – not, at least, if first experiences with pronouncing a foreign language are indicative. To understand the nature and import of such ‘origin’ events is a worthy but challenging objective. It requires understanding how the system (e.g. a Felin or a Foss) state at t, which emerged from the full history (back to the Big Bang) prior to that time, interacts with the time t input to the system (e.g., instruction in some skill) to produce the system outputs and its state at t+1. It is difficult to do that for a complex system because a large fraction of the relevant history lies buried in the system state at t, and much of it is unobservable. These big scientific challenges are not, however, logical puzzles to which the idea of ‘infinite regress’ applies.
Aside from reflecting on the processes that produced their own skills, the authors might examine some of the accounts in the literature of the development of routines and capabilities at the organizational level – e.g., the Cohen and Bacdayan experimental study previously cited (1994), or studies such as Usselman (Reference Usselman1993) on IBM's capabilities, or Szulanski (Reference Szulanski, Dosi, Nelson and Winter2000) on the origins of Banc One's replication capability (Narduzzo et al. Reference Narduzzo, Rocco, Warglien, Dosi, Nelson and Winter2000) on the creation of a cellular phone network, or Raff (Reference Raff2000) on the book superstores, or the great book on the German synthetic dye industry (Murmann, Reference Murmann2003). In their different ways, these studies make clear that the development processes certainly cannot be fully accounted for by individual-level psychology, and that, yes, shared experience does play a crucial role in the origins of effective group performance. There are also, however, papers that anticipate the Felin–Foss warnings against an overly mechanical conceptualization of learning from experience (e.g., Mishima, Reference Mishima, Lamoreaux, Raff and Temin1999; Sinclair et al. Reference Sinclair, Klepper and Cohen2000). Those warnings, however, have much greater force with respect to some limited, and older, parts of the learning curve literature than to the routines and capabilities literature – which is perhaps why Felin and Foss are focused there. The learning-curve literature itself has progressed a long way from its origins as a planning device for aeronautical engineers (Argote, Reference Argote1999). Contemporary enthusiasts of the capabilities approach cannot reasonably be accused of neglecting the role of deliberate, intelligent problem-solving efforts; consider Dosi and Marengo (Reference Dosi, Marengo and England1994), Winter (Reference Winter2000), Zollo and Winter (Reference Zollo and Winter2002), Becker and Lazaric (Reference Becker and Lazaric2003), and see also the generous assessment of evolutionary theory provided by the late Alfred Chandler (Chandler, Reference Chandler1992), hardly a man to minimize the role of intentionality and intelligence in the creation of capabilities. Most important, consider the full body of Nelson's work, past and present.
Today, inquiry into the origins of capabilities promises to be greatly invigorated by the recent burgeoning of research on the ‘spinoff’ phenomenon. Surprising things have been discovered about the typical origins of new firms, and questions raised about the specific content of the ‘genetic’ information with which new firms are apparently endowed. The answers for those questions are being vigorously sought on this exciting new frontier. Steven Klepper has played the leading role in this movement; see, e.g., Klepper (Reference Klepper2002), Klepper and Sleeper (Reference Klepper and Sleeper2005) and several subsequent papers.
Extremes
Under this heading, some of the observations have a welcome grounding in familiar themes of the strategic management literature, particularly the ‘entrants vs. incumbents’ theme. This issue is linked to the critique of experience accumulation as the basis of capability. For example:
To take one glaring example from the literature, how do experiential models account for organizational ‘diseconomies’. . ., that is, the oft-noted discrepancy between the radically skewed innovation outcomes in small and young versus large and old organizations (the former being much more productive and innovative in relation to their size)? That is, experience or repetition-based learning models would of course predict the primacy of older, larger (thus more experienced) organizations, given that the theoretical mechanisms (e.g., repetition, accumulation . . .) have ‘operated’ both longer and more frequently, thus leading to increased organizational capability (Felin and Foss, Reference Felin and Foss2011: 9–10, emphasis in the original).
There is a misapprehension here, which has become quite common in the strategic management literature, and I welcome the opportunity to comment on it.
To begin with, it is worth recalling here that the second paragraph of the capabilities chapter in the Nelson and Winter (Reference Nelson and Winter1982) book concludes with the following statement about the scope of the discussion being introduced: ‘. . . organizations that are involved in the production or management of economic change as their principal function – organizations such as R&D laboratories and consulting firms – do not fit neatly into the routine operation mold’ (ibid.: 97). The caveat I would offer today would actually be significantly weaker than that (Winter, Reference Winter, Beck and Ebner2008), but it remains true and important that high routinization is much more prominent in some situations than in others. Operational capabilities, such as production and distribution, are expected to display more routinization than first-order dynamic capabilities such as research and development (R&D) (Winter, Reference Winter2003). This is not a limitation of capabilities theory or a denial of the role of experience, but an illustration of it. The newer or less repetitive the activity, the less weight we should give to experience in predicting success.
In the domain of production, the predicted ‘primacy of older, larger (thus more experienced) organizations’ does correspond to the empirical reality. It is overly generous of Felin and Foss to attribute this accurate prediction specifically to capabilities theory. There are alternative explanations, but we are happy to accept the scoring of this point. If the question is restricted to innovative activity, then the balance does shift toward newer firms. It is true that a good deal of evidence points to the higher efficiency of small firms in R&D projects of a given type. Whether the efficiency difference implies primacy in an important sense is not so clear. R&D projects are heterogeneous, and some important ones are not realistically within the capabilities of small firms (Schumpeter, Reference Schumpeter1950). This heterogeneity is only one of several measurement problems that may detract from the meaningfulness of the comparison.
In the strategic management literature, claims for the innovative superiority of small firms and entrants often go beyond the narrow efficiency question, and are often exaggerated. The observation ‘Isn't it interesting that David sometimes defeats Goliath in business competition’ became so familiar and well illustrated with selected examples that it was transmuted into ‘What is the explanation for the fact that David generally beats Goliath?’ If this generalization were true, it would indeed count against the most straightforward sort of prediction from the capabilities viewpoint. But it is not remotely true, as any inspection of readily available descriptive statistics will amply demonstrate. The great bulk of economic activity in the economy is performed by the larger, older firms. As far as I know, nobody has ever sought to demonstrate in a statistically credible way that ‘David generally beats Goliath’, which is not surprising given that the aggregate figures are so much to the contrary. Such a demonstration would require (as usual) some care with the operational definitions, particularly on the question of when a Goliath is declared ‘beaten’. It is quite common that dominant incumbents get wounded by new competition, particularly new, innovative firms, but it is also common that they rally and recover, if not all the way to their former dominance.Footnote 11 Sometimes the wounding is severe enough to carry them at least temporarily into bankruptcy, but if they retain a major market share, is that not a retention of ‘primacy’? Such questions raise basic conceptual issues for evolutionary theory, which I addressed some time ago (Winter, Reference Winter and Singh1990). Equally fundamental to an adequate statistical demonstration is a definition of the universe of David–Goliath contests, from which we need a representative or random sample. (What we mostly have on record reflects sampling on the dependent variable value ‘David wins!’.) If every aspiring start-up in an established industry counts as a new David, then the proposition that David generally wins will be overwhelmingly rejected, so we probably need a more stringent criterion to locate a question that might conceivably be open.
In the perspective of evolutionary theory, innovative small firms are seen as an important contributor of variety to the system. They are important not because they do given tasks better than more experienced large firms, but because they often attempt different tasks. Neither efficiency nor financial success is necessary for the role performance to be a valuable one, since the fruits of failed efforts are often not lost to the economy as a whole (Dosi and Lovallo, Reference Dosi, Lovallo, Garud, Nayyar and Shapira1997; Knott and Posen, Reference Knott and Posen2005; Hoetker and Agarwal, Reference Hoetker and Agarwal2007). Meanwhile, in established domains, firms like Intel exercise their dynamic capabilities to pursue trajectories of world-transforming innovation on a world-transforming scale – though some might disparage such experience-based innovation as ‘incremental’, relying mainly on the fact that the innovators make it look easy (Winter, Reference Winter, Beck and Ebner2008).
Intentionality
There is a germ of truth in the observations that Felin and Foss (Reference Felin and Foss2011) offer under this heading. In a part of the routines and capabilities domain, the themes of experiential learning and relatively automatic response arguably became over-developed relative to rival themes of cognition, problem-solving and intentionality. Similarly, the contribution of routines to organizational inertia may have been overstated relative to the role of routines in producing change. As already suggested in the ‘origins and causation’ discussion, the extent of this misallocation should not be overstated, for these dualities are certainly recognized in Nelson and Winter (Reference Nelson and Winter1982) and in the deeper antecedents in the routines literature. Nevertheless, an assessment of the situation similar to that of Felin and Foss was part of the motivation for the convening of a workshop on Cognition and Capabilities, held at the Harvard Business School in 2002. Several of the recent works cited by Felin and Foss as exceptions to their general point had some connection to that workshop. There are others, some published (Gavetti, Reference Gavetti2005; Gavetti et al. Reference Gavetti, Levinthal and Rivkin2005) and some still in progress. Meanwhile, another line of work had been developing independently, in which agency in seen as operating through the ongoing practice of a routine to modify the routine itself. An important example of this body of theoretical and empirical work is Feldman and Pentland (Reference Feldman and Pentland2003) – a paper notable for its award-winning contribution (2009 ASQ [Administrative Science Quarterly] Scholarly Contribution Award), the suggestive juxtaposition of the words ‘routines’, ‘flexibility’ and ‘change’ in its title, its high citation rate, and its absence from the Felin–Foss bibliography. For related work, see Pentland and Rueter (Reference Pentland and Rueter1994), Feldman (Reference Feldman2000), and Pentland and Feldman (Reference Pentland and Feldman2008), among others. Thus, the recommended adjustment of research direction actually has been developing for several years now (Becker et al. Reference Becker, Lazaric, Nelson and Winter2005; Gavetti et al. Reference Gavetti, Levinthal and Ocasio2007).
Beyond this broadly agreed territory, most of what the authors say under this heading derives simply from their dedication to the perception that routines and capabilities theory is grounded primarily in behaviorist psychology. To oversimplify the points already made in this connection, the reader is advised to note that the behaviorist psychologists multiply referenced by Felin and Foss (Reference Felin and Foss2011) are not referenced at all in Nelson and Winter (Reference Nelson and Winter1982); meanwhile, many of the authors cited by Nelson and Winter as among their antecedents are not cited at all by Felin and Foss. Perhaps the most crucial and extreme of these problems is the neglect of the influence of Polanyi; nothing would do more to lay to rest the ‘rat psychology’ charge than to read Polanyi (Reference Polanyi1964), especially chapter 4 on Skills.
One issue possibly deserving of further comment is the relationship of intentionality to evolutionary logic. The opening sentence of Dosi et al. (Reference Dosi, Nelson and Winter2000: 1) notes that ‘It is familiar enough that business firms and other organizations “know how to do things” – things like building automobiles or computers, or flying us from one continent to another’. It would be remarkable if anyone were to believe that organizations learned to do these things in the manner that a rat learns to run a maze, running strictly on short-run feedback and without intentionality (of a human kind). It is correspondingly remarkable that such a belief is vigorously attributed by Felin and Foss to persons who consider themselves students of the phenomena of organizational capability.
A more penetrating question could be raised about the compatibility of high intentionality with an evolutionary viewpoint. In biology, it is the blind watchmaker of evolution who offers an explanatory alternative to ‘intelligent design’. But in the development of capabilities, there really is a lot of intelligent design – though not to the exclusion of experiential learning. Is the evolutionary approach therefore unsuited to the explanatory task in the domain of capabilities? No, the point is that while the ‘watchmakers’ in human organizations are far from blind, they are quite near-sighted. They can do complicated instrumental things in the interest of making an attempt at a complicated performance – but that does not mean that they can accurately foresee the long-run consequences of the attempt. From the system-level viewpoint, this means that genuine experiments are being run; new ways of doing things are being put to the reality test, and outcomes are not highly predictable. It is the weakness of foresight that brings the evolutionary logic into play (Hodgson and Knudsen, Reference Hodgson and Knudsen2006; Jantunen and Sainio, Reference Jantunen and Sainio2008).
Thus, in making appropriate room for intentionality in our theories, it is important to avoid the distortions that crippled rational choice theory, where ideas of rational intentionality became stretched to the point where they effectively impute a kind of perfect foresight to the actor.Footnote 12 Intentionality can be very effective locally, in time and space, and is a crucial factor in the appearance of sophisticated artifacts – but this does not imply that the actor has reliable insight into, much less control over, the future. Absent such insight and control, and given the reality of continuities in behavior, an evolutionary approach to the task of explanation is a forced move.
Creativity and new knowledge
Much of the above discussion of the ‘origins’ problem is applicable to the obviously closely related problem of the origins of new knowledge. In particular, as just noted, there is simply no substance in the suggestion that the capabilities approach is at odds with the existence of intelligence or its occasional application in episodes of creativity and problem-solving. The question is one of its empirical prevalence in various contexts, especially in organizational behaviors of a complex but quite repetitive kind – a kind represented in a great deal of productive activity.
The authors have very high standards for what they would concede to be new knowledge. Computation evidently cannot produce new knowledge, per such long-familiar claims as ‘computers and artificial agents can only produce what they are programmed to do, originating no new knowledge and showing no creativity’ (Felin and Fosse, Reference Felin and Foss2011: 13). By the time such statements are trimmed and qualified to the point where they reduce to a core truth, the truth thus disclosed has become so small that it is uninteresting. There is a basic distinction to be made between the role of the programmers in creating algorithms and the results of those algorithms when run. If a computer extends the known digits of pi, or the known set of prime numbers, is it reasonable to say that no new knowledge is thus created? Relatedly, there is considerable overstatement in the claim that computers are ‘deterministic’ in a sense that somehow precludes the appearance of interesting novelty. From today's perspective, these conceptions all seem rather quaint (see McCorduck, Reference McCorduck2004, an update of a book on artificial intelligence published 25 years earlier). For decades now, the quest for computational models of ‘emergent’ phenomena has been pursued to impressive effect. The ‘species chauvinism’ of H. sapiens continues to fuel the quest for a definition of the domain of creative accomplishment that computers assuredly cannot get into – but the territory claimed keeps shrinking, and it has long been apparent that most humans do not get into it either.
Perhaps the important question in the context of capabilities research is whether computers are so limited in their capacities as to make them necessarily ineffective in modeling the creation of novelty. Practitioners in the field generally perceive that it is not the limitations of the computers that principally stand in the way of progress, but the cognitive limitations of the human beings who need to understand what the computer has done.
The environment, the ‘internal approach’– and the ‘poverty of stimulus’
Under these headings, the Felin and Fosse (Reference Felin and Foss2011) press their assault on a hypothetical theory in which experience alone accounts for the learning underlying routines and capabilities, with no weight at all given to the nature of the system exhibiting the learning. This notion is evoked repeatedly in the paper, for example in the attribution to Karl Popper (Reference Popper1972) of the claim that behaviorists view the mind as a tabula rasa (Felin and Foss, Reference Felin and Foss2011: 5). The argumentation in this vein reaches some kind of extremum with the rhetorical invocation of the ‘pet bee’, whose experience hypothetically matches that of a learning child:
The child would not develop the navigational ability of the bee, and the bee would not develop language capabilities, despite uniform environmental stimuli (ibid.: 16).
I ask, very seriously, whether anybody ever actually believed to the contrary. True, there has been plenty of back and forth on the matter of allocation of emphasis; Gestalt psychologists such as Wolfgang Köhler were a lot different from the Skinnerians. But the ‘tabula rasa’ and ‘pet bee’ language is extreme; it is unfortunately the kind of thing that often happens when somebody is tendentiously characterizing somebody else's opinion. It is not the kind of thing that anybody ever seriously advanced as his own opinion. Felin and Foss devote a great many words to attacking a position that probably nobody ever held – not just nobody in the capabilities camp, nobody at all.Footnote 13 They come out, constructively, for the program of giving more weight to the ‘internal’ attributes of the learner. If I understand them correctly, it is those attributes that make up the explanatory deficit called ‘the poverty of stimulus’. In that stance they have a lot of distinguished company going back a long way. It was very much what the Gestalt psychologists said by way of criticism of behavioral psychology. (Personally, I learned my (undergraduate) psychology from the Gestalt viewpoint long ago, and was never tempted by the other view.)
It should be conceded that there is a glaring ‘poverty of stimulus’ issue in the logic of the simplest models of evolutionary economics, though it is not strictly a matter of capability learning. Many such models presume that a very simple feedback mechanism involving profitability and growth is sufficient to shape the detailed destiny of a large, complex firm. That is hardly credible on its face, and (consistent with the general Felin–Foss point), there is an implied need to inquire much more closely into what is going inside those complex firms.
Perhaps the capability theorists should have given explicit emphasis to the premise that the organizations of which we spoke were indeed inhabited by representatives of H. sapiens, and thus came to the situation endowed with the many remarkable capacities of that species (including language), and ready to learn. Probably we should have gone on to discuss the rich heritage of social, cultural and personal factors that also shaped these actors, and shrunk explicitly from any claim that species membership by itself would suffice to define the class of actors to which the ideas might relate. If that failure of exposition is the ultimate source of the misunderstandings that Felin and Foss pursue, it was an expensive failure indeed. Unfortunately, we probably took that all for granted.
A curious thing about the Felin and Foss account is that evolution (in a general sense) is not prominent as a possible explanation for the ‘human nature’ that needs to be understood if we are to complete the behavioral picture. Yet that seems like a plausible direction for an inquiry into origins. As a practical matter, however, we always identify punctuation points in the causal chain and pursue the argument forward from there. We do not pursue the question of the origin of life back into the question of the origin of the chemical elements, or the origin of primates back to the question of the origin of life. We tend to take the typical attributes of H. sapiens as given when discussing human learning, although the diversity within the category is enormous. Similarly, the question of the origins of routines and capabilities can be, and to some extent has already been, pursued back into the origins of the particular organization and beyond – but perhaps not far enough into historical and evolutionary antecedents, and certainly not all the way back to the origins of the human species. Remarkably, there is a sense in which the possibility of adjusting that balance is perhaps better than ever before – because modern instrumentation gives us unprecedented access to the internal workings of human bodies and minds, shaped as they are by biological and cultural evolution. A dramatically inviting future is beckoning, not just for scholars interested in routines and capabilities, but for social science generally. Stronger ‘foundations’ for our future work can be located by looking in that direction, where there is an opportunity to craft a strong explanatory connection to the present attributes of the human species, and beyond that to their evolutionary sources. But . . . it's not going to be simple.
Perhaps Felin and Foss are simply urging us to look back beyond the punctuation points presumed in some discussions of routines and capabilities. They may favor more careful case studies of organizations, perhaps with a developmental focus, on the model provided by Gavetti and Rivkin (Reference Gavetti and Rivkin2007). If this is the proposal, I agree with it. In that case, however, the considerable volume of existing research that also supports this proposal should not be ignored.
