2. Lack of construct clarity: why it may not be helpful to ‘focus on real routines’

It is preferable in a debate if the debating parties can agree about the nature of the subject being discussed, or, if this is not possible, define some shared conceptual ground that may help to resolve disagreements. A basic problem with much routines-based reasoning and associated work on capabilities is that it suffers from problems of construct or conceptual clarity, which makes debate, such as the present one, inherently difficult. Moreover, it also hampers scientific progress (Williamson, Reference Williamson1999). An example is the difficulty of pinning down what exactly we may mean by the ‘foundations’ of a concept that is still so elusive as that of ‘routine’ (see Becker, Reference Becker2004). A further problem is the fair amount of ad hoc theorizing that is thriving in this literature – and which we discern in the responses of our opponents.

Specifically, proponents of routines are characteristically inclusive in their understanding of routines – at the expense of construct clarity.Footnote ¹ New constructs do not come ready-made with clear definitions (Leijonhufvud, Reference Leijonhufvud and Lakatos1976), and, in recognition of this, sorting out definitional issues, scope conditions, semantics, etc. constitutes a quite significant and important part of scientific activity, not the least in the social sciences.Footnote ² However, the routines literature has been significantly underperforming with respect to promoting construct clarity. Consider, for example, Pentland's definition of a routine as ‘repetitive, recognizable patterns of interdependent actions, carried out by multiple actors’ (Pentland, Reference Pentland2011: 280). According to this definition, the patterns of pricing decisions we observe in oligopolistic markets are routines. So is the fact that drivers tend to stop when the traffic lights are red (they are interdependent because other drivers adjust their actions). Voting behaviors in parliament are routines according to the definition. In fact, Pentland's definition is so inclusive that much of what is meant by an ‘institution’ is covered by it.

Pentland tells us to ‘focus on real routines’ and this, he submits, will help us ‘avoid confusion’ (ibid.). Accordingly, he offers an example of invoice processing. But, as he describes it, it sounds exactly like what may be called a ‘standard operating procedure’. These are typically designed. Fine. But a planned/designed sequence of interdependent actions is just one interpretation of routines. Others stress the emergent, unplanned character of routines (Nelson & Winter, Reference Nelson and Winter1982). Unfortunately, the literature seldom makes it explicit whether a given discussion concerns intentionally designed standard operating procedures à la Richard Cyert and James March (Reference Cyert and March1963) or emergent routines à la Richard Nelson and Sidney Winter (Reference Nelson and Winter1982).

One may question whether it is a problem that the routines literature works with such an inclusive key construct: if a host of heterogeneous phenomena in the social world share the characteristic of ‘routine-ness’, then why not analyze them all as manifestations of the same thing? While this research strategy sounds compelling, we are worried that it may suppress too much variation. Thus, are routines in a hospital emergency room really the same ‘thing’ as routines in a research and development laboratory in a major industrial corporation? Do they emerge in the same manner? Are they maintained and changed in the same manner? Are the underlying microfoundations the same? Thus, we are simply not too charmed by the proposal to ‘focus on real routines’. First, there is so much heterogeneity with respect to what may conceivably qualify as ‘real routines’. Second, pace Pentland, ‘focusing on real routines’ does not resolve any of the deep-seated philosophical, theoretical, explanatory, etc. problems that we focused on in Felin and Foss (Reference Felin and Foss2011). Third, Pentland's admonition fails to recognize that observation and reality are preceded by theory. We discuss these issues further below.

3. Universal mechanisms or comparative chauvinism?

While the arguments of our critics are sometimes hard to pin down, nonetheless there is strong evidence in their responses that very directly supports the claims in our original paper. To illustrate, Winter says that he ‘take[s] some pride’ in the direct links between routines-based reasoning and the sciences that place an emphasis on ‘incremental and experiential evolution, randomness, and environmental selection’ (Winter, Reference Winter2011: 261). He even argues that we engage in ‘species chauvinism’ (ibid.: 272) (when comparing human capability with machine capability). We are proud of this type of chauvinism. Namely, the chauvinism comment is highly instructive as it precisely illustrates our point about the need to focus on the endogenous characteristics of actors themselves, whether individual or collective, rather than – as Winter would have it – focus on universal, species-independent, environmental mechanisms that (somehow) operate outside the organism under study, as if the nature of the organism itself did not matter. One way to summarize the argument of our original paper is to indeed say that we think that comparative chauvinism is warranted.

Evolutionary economics embraces a type of scientific universalism where environmental selection is seen as the key general mechanism (Winter, Reference Winter1964; Aldrich et al., Reference Aldrich, Hodgson, Hull, Knudsen, Mokyr and Vanberg2008; Hodgson and Knudsen, Reference Hodgson and Knudsen2010). Winter has in fact argued that:

However, while there might be some similarities across species and mechanisms – thus suggesting the possibility for a grand unified theory – we remain skeptical of this and would rather focus elsewhere. Specifically, there are important, comparative differences between species, in the underlying nature of the things under study. Part of the impetus for our original article was to make a point about the unique nature and capability of humans, something that universal, species-independent and grand theorizing tends to ignore.

The eagerness to apply theories and methodologies from the natural sciences directly to the social sciences of course is not new and is the impetus for the ‘empiricism’ (and associated ‘behaviorism’) that we criticize. From the British empiricists to Auguste Comte and Herbert Spencer onwards there has been a desire to ‘unite’ the sciences under a common, empiricist framework. This grand program, and its direct links to evolutionary mechanisms, led Ernst Mach to the ‘conviction that the foundations of science as a whole, and of physics in particular, await their next greatest elucidations from the side of biology, and especially, from the analysis of the sensations’ (Mach, Reference Mach1897: preface). Thus, while Winter says that he has ‘difficulty . . . in recognizing . . . the intellectual territory’ (Winter, Reference Winter2011: 262) that we place him in, nonetheless this brand of empiricism, as well as the associated universal Darwinism that evolutionary economics subscribes to, falls directly into this intellectual tradition and larger program of research.

Granted, there are many types of empiricisms – for example, ‘sensationalist’ or ‘logical’ (Ernst Mach and the Vienna Circle), ‘instrumentalist’ (Dewey) – with differing foci, whether perceptions, immediate actions or doing, various types of experiences and observations (associational mechanisms) and so forth. But what all of these empiricisms share is an underlying unity in black boxing the characteristics of the organism itself – its nature – and focusing largely on the environment. All this is done in the name of a seemingly more objective science where a priori knowledge and/or native characteristics and capabilities of the organism itself are rejected (or, certainly these are not the primary drivers of any empiricist model). Evolutionary economics is not alone in linking with this type of empiricism – various aspects of empiricism have also been central for other streams of economic thought. For example, the German Historical School placed an emphasis on real, observable history and experience and shunned ‘abstract’ theory. American institutionalism, including John Dewey's student Wesley Clair Mitchell, also explicitly sought to lay behaviorist and empiricist foundations for economics (for some historical discussion of this point, see Caldwell, Reference Caldwell2004).

In his response, Pentland hopes to circumvent the above philosophical and methodological issues, but he inadvertently places them center stage. Pentland (Reference Pentland2011: 285–286) argues that we fail to ‘incorporate current thinking about the ontology of organizational routines’ (ibid.: 285). Specifically, we allegedly neglect technology, managerial design, institutional rules and norms, and experience, and we do not realize that in ‘. . . any given context, when people need to get their work done, they adopt whatever strategies (patterns of actions) that seem best, and adapt them as necessary. For the people involved, there are no deep philosophical issues, and no shaky foundations’ (ibid.). Apparently, Pentland thinks that because people seem perfectly capable of adopting routines, there is no need to ask microfoundational questions. Try telling an economist or a sociologist that because we can observe people interacting fairly successfully in the market place, there is no need for further inquiry into their motives, preferences, actions, and so on. This is obviously a glaring non sequitur.

But methodological stands are inevitable. In fact, Pentland's focus on observation ironically parallels some of the basic tenets of the ‘wissenschaftliche auffassung’ (‘scientific view’) of the Vienna Circle empiricists (for an overview, see Carnap and Morris, Reference Carnap and Morris1969). Specifically, throughout his response Pentland calls for us to focus on ‘real problems’, with an emphasis on the observation of ‘actual routines’ and ‘work practices’. But our precise point is that these observations – whether in the case of the subjects we study or in the case of scientists themselves – are highly problematic as sources for understanding actual behavior and reality.Footnote ³ To illustrate, the observation of an apple falling in no way embodies the law of gravitation within it for the observer. Or to provide another example, while it still appears that the sun orbits the earth, the truth is the opposite. Whether we are dealing with scientists or the ‘lay’ people that we study, observation and reality are preceded by theory.

We are scarcely the first to highlight serious problems associated with naive empiricism in economic and social theory. Freidrich Hayek (Reference Hayek1952) criticized social scientists of his time for their ‘scientism’, the effort to mimic the natural sciences even though there are comparative and obvious differences between humans and the subjects that the other sciences study. Karl Popper (Reference Popper1972) also made strong claims about the highly problematic and simplistic input–output, ‘bucket’ nature of empiricist theorizing. Of course, such critiques do not mean that observation or experience (along with other perceptual mechanisms) do not matter, of course they do – as we discuss, they play a ‘triggering’ role (see our ‘Poverty of stimulus’ discussion; Felin and Foss, Reference Felin and Foss2011: 245–250) – but our point is that these ‘externalist’ factors should not be given primacy as they may (and indeed often do) lead theorizing astray.

4. Models of mind and man

The center stage of the present debate is occupied by questions about the appropriate model of mind and man. As discussed above, the problem with universal and empiricist models is that they do not require theories about the nature of the organism itself, as the heavy lifting of these theories is largely done by the environment. It is at this level where we see significant overlap between routines and capabilities-based theorizing of evolutionary economics and psychological behaviorism and empiricism. Thus, while Winter (Reference Winter2011) takes issue with our ascription of ‘rat-psychology’ to leading scholars in the routines and capabilities space, we nonetheless see crucial overlaps in the key constructs and mechanisms, specifically between the behaviorism of psychology and ‘behavioralism’ of organization theory.

Behavioralism partly emerged as a critical reaction to a number of extreme assumptions in the modeling of man in economics (see Simon, Reference Simon1955) and was an attempt to add greater descriptive realism to the analysis of individual and organizational behavior (March and Simon, Reference March and Simon1958; Cyert and March, Reference Cyert and March1963).Footnote ⁴ Herbert Simon's boundedly rational actor has been the underlying model of many extant theories of organization (March and Simon, Reference March and Simon1958; Mahoney, Reference Mahoney2005), notably providing important foundations – and the point of departure vis-à-vis the neoclassical theory of the firm – for Cyert and March (Reference Cyert and March1963), a key source publication for much strategic management thinking as well as institutional economics (Williamson, Reference Williamson1975; Pierce et al., Reference Pierce, Boerner, Teece and Teece2008). Simon's conception of rationality is, of course, an explicit reaction to the modeling of man in economics. It models agents in terms of information processing, and adds situational determinants of decision-making, such as information access or availability. Thus, rather than assume that somehow information disseminates perfectly, bounded rationality points to the localness of information and its diffusion and bounds. This intuition is important in models of organizational search (Gavetti and Levinthal, Reference Gavetti and Levinthal2000).

Winter suggests that our reading of the routines and capabilities literature and the way we link it to behaviorist psychology may be a rational reconstruction rather than a literal depiction of ‘how it really happened (Winter, Reference Winter2011: 264). He specifically takes issue with our attribution of ‘rat-psychology’ to leading scholars who work on routines and capabilities. We agree with Winter that behaviorism and behavioralism are not fully congruent; nevertheless, there are crucial overlaps.Footnote ⁵ At first glance behavioralism mainly seems to be about adding realism to the understanding of behavior (vis-à-vis the rational economic man bogey). However, behavioral approaches start with environmental and experiential factors as the causal starting-point for explaining human activity and behavior, and add behavioral programs (i.e. routines) that are triggered by environmental stimuli to understand decision-making. To understand behavior, so the argument goes, we need to understand these programs as well as the environment that supplies the stimuli.Footnote ⁶ In marked contrast, rational choice-type (and more generally ‘rationalist’) models are strongly future-oriented and focused on the a priori capabilities and choices of the subject itself. Thus, in behavioral approaches the temporal orientation on the past manifests itself with a central focus given to experience in its various forms – which we find very problematic – where experience provides the key driver of organizational decision-making and behavior.

While we can agree that the behavioral program in organizational research has not gone to the extremes of Skinnerian behaviorism, we maintain that behavioralism, like behaviorism, puts a premium on treating individual cognition and decision-making in the very simplest possible terms. Individuals are represented as rather rigid followers of rules, routines, heuristics, or just plain dumb (an illuminating panel discussion by leading scholars confirms this: see Murmann et al., Reference Murmann, Aldrich, Levinthal and Winter2003).Footnote ⁷ In fact, there is an explicit attempt to cram virtually all behavior into the routines box, so that even highly creative behaviors are placed under ‘dynamic routines’ or ‘dynamic capabilities’ or ‘second/nth order’ headings (Nelson and Winter, Reference Nelson and Winter1982; Teece et al., Reference Teece, Pisano and Shuen1997; Eisenhardt and Martin, Reference Eisenhardt and Martin2000; Winter, Reference Winter2003). Behavior that cannot be treated in this manner is relegated to a residual category of ‘ad hoc problem solving’ (Winter, Reference Winter2003), and by implication, we fear, seen as deserving of less interest on the part of researchers.

The attraction of behavioralism was that it promised an ostensibly more realistic model of man that could take behavioral variety, learning behavior, adaptive preferences, and so on more fully into account. In contrast, economics models of behavior were seen as static, excluding learning (Cyert and March, Reference Cyert and March1963). However, as Mark Casson observes, the exact opposite is closer to the truth:

More broadly, Michael Jensen and William Meckling (Reference Jensen and Meckling1994) argue that the view of man in the ‘rational action’ or situational analysis approach is that of a ‘resourceful, evaluative, maximizing’ man (i.e., the REMM model) – a view that they argue is entirely consistent with differences in observed behavior and with learning: ‘Human beings are not only capable of learning about new opportunities, they also engage in resourceful, creative activities that expand their opportunities in various ways’ (Jensen and Mecking, Reference Jensen and Meckling1994: 5). If Casson and Jensen and Meckling are right, the behavioralist criticism of the rational action model may have been based on a strawman right from the beginning.

What is more, the behavioralist critique of the neoclassical view of man can justifiably be leveled at behavioralism itself. As noted by Richard Langlois and László Csontos: ‘. . . behavioralists tend to assume that agents are (1) hard-headed rule followers or (2) pre-programmed satisficers ab ovo’ (Langlois and Csontos, Reference Langlois, Csontos, Mäki, Gustafsson and Knudsen1993: 118). The human rats in behavioralism are surely hardwired with more sophisticated wirings than real rats, but their psychological makeup still makes them behave in a rat-like manner. Furthermore, an additional problem with behaviorism of course is that it not only is an inappropriate scientific model for humans but for rats as well (as highlighted by a large body of work in ethology) – a pseudo-scientific program of research (Chomsky, Reference Chomsky1959).

One way to then restate a central point of our original article is to argue that models that feature bounded rationality should focus more on the nature of rationality itself and not just its historical, structural, contextual, perceptual, experiential and environmental boundedness. The latter of course are the key mechanisms of work on routines and capabilities. While experience and history provide external data that appear scientific (as we discussed previously), nonetheless it does not help us understand the intentional and purposive factors associated with individual and organizational behavior. As Lionel Robbins argued:

In short, we are dealing with human actors and thus a theory of these actors is needed, a theory of their nature, creative capacities, rationality and so forth.

5. Levels of analysis

Our effort in the target article (Felin and Foss, Reference Felin and Foss2011) was to raise a levels-independent argument about the problematic nature of experience as a key mechanism (and associated empiricist foundations) for explaining behavior, whether individual or organizational, but both Hodgson and Knudsen (Reference Hodgson and Knudsen2011) and Pentland (Reference Pentland2011) extend their critiques and also raise concerns about some levels-related issues from our previous work on microfoundations (e.g., Felin and Foss, Reference Felin and Foss2005; Abell et al., Reference Abell, Felin and Foss2008). We thus address each in turn.

Hodgson and Knudsen applaud us for emphasizing ‘the role of human agency and individual psychology in the analysis of routines’ (Hodgson and Knudsen, Reference Hodgson and Knudsen2011: 295), but take issue with our endorsement of methodological individualism, because they think we pose ‘individuals versus social structure as alternative explanantia’ and doing this is a ‘false dichotomy’. Pentland also seems to believe that we think that routines are ‘individual-level phenomena’ and that ‘routines are . . . exclusively a function of human psychology’ (Pentland, Reference Pentland2011: 287). This rests on a misreading of our papers: nowhere do we deny that social structure influences individual action and interaction. In fact, we have often used the simple expository device known as the ‘Coleman bathtub’ (Coleman, Reference Coleman1990) – which brings out the macro–micro link quite explicitly (see Abell et al., Reference Abell, Felin and Foss2008).Footnote ⁸ We do think that there are significant benefits for bottom-up-type modeling which explains the origins of collective structures, such as norms – rather than just assuming their existence – but naturally these collective structures in turn also play a role in enabling and constraining subsequent interaction and outcomes. However, we maintain the basic point that there is no meaningful role for macro-level (macro-to-macro) causation. In the social domain, at least, macro-variables are necessarily linked to other macro-variables through mediating micro-variables.

Pentland argues that ‘the most serious flaw of the microfoundations project’ and its ‘individualism’ is that it ‘fails to incorporate the ontology of organizational routines’ and ignores ‘sociomateriality’. In turn, Pentland argues that that the proper foundations, rather than individualism, are in the work of Anthony Giddens and Pierre Bourdieu, and levels-independent concepts such as ‘structuration’ and ‘practices’. The problem with Pentland's solution of essentially ignoring levels (individual and collective) and supposedly unproductive dichotomies (agency and structure) is that theoretical explanation, and the uncovering of mechanisms, gets subverted by the invention of vague concepts such as ‘structuration’ that only seem to mystify rather than explain social outcomes. We are not the only ones to make this point. Margaret Archer notes that Giddens's notion of ‘structuration . . . throws a blanket over the two constituents, structure and agency, which only prevents us from examining what is going on beneath it’ (Archer, Reference Archer1995: 102). Precisely. In other words, concepts such as structuration and practice ‘lose’ their problem at the very outset by simply giving them a name (structuration or practice), a collective equilibrium that should be explained rather than named (see Coleman, Reference Coleman1990). While we appreciate how these approaches might attune us to process, nonetheless there is little modeling or theoretical development about how, for example, heterogeneous interests or beliefs aggregate to collective-level consensus and emergent outcomes (or not). The precise point of the microfoundations program is to unpack, where possible, the central constituents, processes and interactions among individuals in order to explain the origins, maintenance and reproduction of collective outcomes such as norms, routines, capabilities and so forth. To simply tightly couple these constructs, then name them, does not meaningfully advance our theoretical understanding. Our preference is for sharper, analytic models.

Related to this levels issue, we might briefly note here that we indeed see the target paper of the present debate (Felin and Foss, Reference Felin and Foss2011) as part of a wider microfoundations discussion and thus welcome the opportunity to clarify what we think the proper foundations are. Microfoundations, after all, could also be based in behaviorism. George Homans, for example, was an important advocate of microfoundations in social theory, but, as noted by Coleman (Reference Coleman1986: 1131), his social theories ended up being highly reductionist and essentially grounded in precisely the type of behaviorism that we criticize. Thus, there are various forms of reduction, of how we might study and specify the constituent elements that make up organizations, and we argue that subjective, forward-looking and choice-based – rather than observation-, experience- and environment-based – models more fruitfully account for human and organizational behavior.

6. Agency, choice and uncaused causes

Hodgson and Knudsen (Reference Hodgson and Knudsen2011) raise the issue of the nature and explanation of human agency, which indeed is a difficult and fundamental concern. Essentially, they argue that we want to treat choice as an uncaused cause (à la George Shackle), and that this is fundamentally unscientific as the hallmark of scientific explanation is a causal approach. They argue that ‘asserting the importance of human agency does not give an excuse to terminate the search for causal explanation, which is the essence of science’ (Hodgson and Knudsen, Reference Hodgson and Knudsen2011: 296).

We agree with Hodgson and Knudsen here, and suggest that they put their finger on one of the most difficult issues in the understanding of agency: how much determinacy do we want to introduce to our explanations of human action and how do we introduce such determinacy? The issues here, as Hodgson and Knudsen note, are ultimately deeply philosophical. They go back to David Hume who argued that human behavior must either be determined and (therefore) predictable, or it is random and (therefore) unpredictable. Both options eliminate notions of ‘free will’, ‘creativity’, ‘innovation’ and indeed choice itself.

In a classic discussion of this dilemma, Popper offers the notions of ‘clocks’ and ‘clouds’ as illuminating metaphors. Specifically, he uses clocks to represent systems ‘which are regular, orderly, and highly predictable in their behavior’ (Popper, Reference Popper1972: 207; emphasis in original). Clouds, on the other hand, represent systems that are disorderly and unpredictable. Real-world behavior, according to Popper, resides somewhere between clouds and clocks. A key notion in this ontology is that of ‘plastic control’. Popper argues that biological evolution has added a number of higher-level functions in organisms that ‘. . . do not replace the lower ones . . . but they establish a kind of plastic control in them – a control with feedback’ (ibid.: 239). Thus, plastic control is a system property, whether that system is the mind, an organism, or, perhaps, an organization, and it is key to Popper that plastic control differs from iron-clad control, i.e., determinism.

In contemporaneous work, Popper (Reference Popper1972) offers the rationality principle as the middle ground between clouds and clocks. Popper argues that the social sciences should explain behavior in terms of the conjunction of a ‘situational analysis’ and the principle of rationality. The situational analysis includes both external constraints (e.g., prices in microeconomics) and internal features of the agent himself, such as his goals and knowledge. In Popper's account these features should be objectified in the sense of building a model of the agent's situation which may be more or less detailed (or anonymous) depending on what one seeks to explain. Popper relates his approach to the traditional covering law model in the philosophy of the natural science (according to which one explains by subsuming particulars under universals). The features of an agent's situation then correspond to initial conditions in the explanation of a physical phenomenon, and Popper argues that what stands in for the role of a ‘law’ in social models is the rationality principle (Latsis, Reference Latsis, Cohen and Wartofsky1983: 133).

The key link to Popper's (Reference Popper1972) analysis of clouds and clocks is that situational analysis implies that decision-making behavior is not deterministically determined, for example, by inbuilt heuristics or decision rules, but is plastically controlled. What drives such control is in the first instance mental states, such as purposes, expectations, preferences, etc., but also theories and critical discussion (ibid: 247). Thus, situational analysis is in essence a description of plastic control, that is, an account of how action emerges from situational features that, however, do not force particular behaviors (Latsis, Reference Latsis, Cohen and Wartofsky1983: 140). It reconciles the ontology of plastic control, and therefore such notions as free will, real choice, creativity, etc., with the notion that we can in fact explain in a deductive manner in the social domain. To us this provides a strong argument why situational analysis is preferable to behavioralism: the distinction, so basic to behavioralism, between routinized/programmed and non-routinized/non-programmed behavior suggests that creativity, new solutions, innovation, and indeed choice ‘are activities present only in extraordinary, exceptional circumstances’ (Bianchi, Reference Bianchi1990: 161).

Note that this actually highlights some key behavioralist ideas: The Popperian emphasis on the fallibility of knowledge and the strong stress on the importance of procedures in testing theories (Popper, Reference Popper1972) explicitly take bounded rationality and heuristics into account. However, what is at stake is that behavioralist agents have been programmed to follow experience-based heuristics, rules, routines, etc. in a non-critical, unreflective manner. This is particularly problematic given that the behavioral tradition has had a penchant for dealing with those biases to decision-making that, in the light of classical decision theory, stand out as decision-making errors. Grandori relatedly points out that the behavioral tradition:

Different attempts in social science at explaining human action/behavior reach determinacy in quite different ways, one by means of situational constraints (in a broad sense) and one by means of programming the agent. Note, however, that in the situational analysis, while the agent's behavior is not forced, it is still influenced by factors at different levels, some of them internal and some external to the agent (Jensen and Meckling, Reference Jensen and Meckling1994). Such factors may be more or less ‘influencing’. Thus, the situational constraints may be such that it seems that they actually fully control (rather than plastically control) the decisions of the agent. If a fire breaks out in a movie theater, an agent does have the possibility of not trying to escape through the closest exit. Or, a producer in a competitive market may decide to sell at a price above the competitive price. But those choices probably mean death (physically and economically). The point is that certain features of the situation may overwhelm other ones; for example, environmental features (a fire, competitive pressures) may overwhelm the features of beliefs and expectations. In such cases, situational analysis may be observationally indistinguishable from behavioralism: the agent is forced to adopt a certain action (a single-exit solution). However, in the case of situational analysis, the environment forces him to do so, while in the case of behavioralism he is programmed to adopt the action based on a given environmental stimulus. While both Popperian situational analysis and behavioralism seek to reach determinate implications concerning human action/behavior, determinacy is reached in very different ways. Overall, in behavioral models, agents are hard-wired to choose certain courses of actions, whereas situational analysis tries to reach determinate outcomes without compromising the free will of the agent. Langlois and Csontos neatly sum up the basic differences between the two models:

All that said, we think that starting with individual-level preferences, beliefs and expectations provides a natural starting point for economic and social theory. While the reductionist logic might be pushed to look at issues of free will and compatibilism versus incompatibilism (as discussed by Hodgson and Knudsen, Reference Hodgson and Knudsen2011: 297), or to even look at genes (or as Winter argues: the Big Bang; Winter, Reference Winter2011: 266), we think that choices are a very basic nexus where alternative trajectories and possible histories can and do occur. In other words, the free will discussion need not necessarily be solved to look at choices in economic and social situations. Furthermore, the type of evolutionary psychology pushed by some is nigh irrelevant, due to the excessively long timeframes, to the very basic, day-to-day decisions actors make.

Thus, we essentially agree with Popper that ‘what social science wants to understand is how such non-physical things as purposes, plans, decisions, theories, intentions and values, can play a part in bringing about physical changes in the physical world’ (Popper, Reference Popper1972: 229). To arrive at an understanding of this, we need appropriate models of human nature and behavior. While these models can hardly be fully predictive and still be models of decision makers with free will, they can certainly still further our understanding. As Gerald O'Driscoll and Mario Rizzo argue:

7. Toward a rationalist alternative: key points

Winter insists that we specify an alternative to the literature we criticize:

Now, the for-your-critique-to-be-taken-seriously-you-must-specify-the-alternative objection is a classic conversation stopper. The history of science provides many examples of important refuting instances, decisive critique, etc. that were made without any clearly stated alternative.Footnote ⁹ Nevertheless, it may well be that old theories are only abandoned when new theories emerge. So, Winter's objection has some force.

As we argue in our original article (Felin and Foss, Reference Felin and Foss2011), central to our rationalist approach is the specification of the proper psychological foundations and understanding of the nature of man. We think individuals are rational decision-makers who engage in best-efforts satisficing, but also creative problem-solving and solution-generation. Our effort to build up a model of man scarcely needs to be done in a vacuum, as our conception shares many similarities with a ‘rationalism’ that has an intellectual history going back to Plato (e.g., Chomsky, Reference Chomsky2003 [1966], provides an overview). Central to this rationalism is the view that humans theorize, reason and imagine possibilities. As noted by Samuel Taylor Coleridge, ‘the meanest of men has his Theory: and to think at all is to theorize’ (Coleridge, Reference Coleridge1969[1840]: 430). In economics this sentiment links with Adam Smith, who also saw the forward-looking, anticipatory aspects of individuals as central for his theories of markets and human interaction (see Rothschild, Reference Rothschild2001). So have most economists since Smith. Individuals have subjective plans, beliefs, theories and expectations about the environment and these inform their choices and behavior. This conception of humans is diametrically opposed to empiricist and behaviorist views of decision-making.

This form of rationalism differs markedly from the caricatured versions of objective ‘rational choice’ in economics. Actors need not be super-human in their abilities to process information, but more realistically they both utilize local information and make conjectures and have theories about the future: about what actions to take, the value of alternatives, possible forms of organization, etc. In other words, we allow economic actors the same prerogative as we give to scientists within their domain, rather than somehow treating them altogether differently. A rationalist approach, then, can be seen as a response to the somewhat glib portrayals of dumb and bounded (in various ways) views of human subjects. After all, human societies have made stunning progress in various technological and other domains, and this progress and the associated organization behind it requires better theoretical thinking and explanation.

The poverty of stimulus notion, we feel, shows some promise as a way of meta-theoretically thinking about what needs to be explained vis-à-vis human nature and behavior in economic contexts. That is, the capabilities that individuals and organizations have are completely under-determined by the experiences they have (see Felin and Zenger, Reference Felin and Zenger2009). While the senses receive significant weight in extant theories, we think that observation and perception are preceded by important mechanisms related to endogenous factors such as human nature, a priori knowledge, theoretical conjecture and so forth. Thus while the responses (particularly by Winter, Reference Winter2011) call for a focus on universal mechanisms primarily associated with the environment, we think that a more fruitful approach to theorizing is to study the endogenous factors and capabilities of individual and collective actors.

8. Coda

We might note that our call to understand the endogenous factors associated with human action and behavior in economic settings coincides with a broader discussion and debate in biology and philosophy about the appropriate models for understanding behavior and evolution. While we have argued for ‘comparative chauvinism’ vis-à-vis understanding and specifying human nature, the original critiques of behaviorism applied to rats just as well as humans: neither can meaningfully be said to behave according to stimulus–response-type models. Some have indeed recently linked the problems associated with behaviorism and natural selection to the root problem of giving excessive primacy to the environment (Fodor and Piattelli-Palmarini, Reference Fodor and Piattelli-Palmarini2010). We have no need to take sides in this heated debate as it relates to biology, though we whole-heartedly agree that understanding the endogenous nature of organisms themselves – which in the case of humans includes thought, beliefs, expectations, preferences, theorizing and so forth – should be center-stage rather than relegating the explanatory apparatus to the environment, as evolutionary economics and empiricist models do.