In a statement that is relatively famous, considering its position at the back of an old book, Alfred Marshall remarked on the ‘the manifold influences of the element of time’. He noted the obstacles those influences pose to mathematical analysis (or, he said, any analysis) of a complex, ‘real life’ problem – and the tendencies to over-simplification that often result (Marshall, 1920, p. 850). Marshall's statement remains broadly relevant today, notwithstanding the very major advances in the analysis of dynamical systems that have occurred in the interim. It is relevant in particular to the complex problem featured in this special issue, understanding the origins of organizational routines and capabilities. In this essay, I argue that an adequate answer to the origins question must fully respect the element of time. That this is the case might seem obvious, considering the connotations of the word ‘origins’. Yet, if the call for this special issue is indicative of the general state of the discussion on origins (and the related ‘microfoundations’ discussion), the centrality of time is not widely acknowledged. The word ‘time’ appears just once, and important related words appear not at all. Particularly striking is the absence of ‘history’ and ‘historical’, considering the expressed intention to emphasize empirical studies of the origins question. The explanation for this may be the one suggested by Marshall, i.e. an understandable drive towards analytical simplification. Regardless, I argue that it is inherent in the nature of the topic that the obstacles must be confronted, and that evolutionary theory is the effective way to organize the confrontation. The invitation extended by the editors asked specifically for a response to the question, ‘What is the one central issue related to the micro-level origins of routines/capabilities that deserves attention in future work?’ This is an excellent question, and I will offer a fairly detailed proposal. First, however, it is necessary to inquire into the meaning(s) of the ‘micro-level origins’ question itself, and how the element of time is or should be handled in it. Consider question 2 of the call for papers, ‘How do individual-level factors (such as traits, abilities or emotions) aggregate to create collective capabilities?’ (emphasis supplied). Should we understand this ‘aggregation’ to be occurring in a limited period of capability creation, and the referenced individuals to comprise only the members of the collective at that time? Should we presume that all the abilities exercised in the collective capability are among those initially brought by the individuals, or are the latter simply the basis of contributions to a protracted organizational process that creates the capability? Or, is the intended scope much more open in these respects and others; for example, are long-dead inventors, entrepreneurs, and scientists entitled to a place in the aggregation? I would argue that the broad-scope version probes much closer to the truth about the origin of a capability and its relation to individual behaviours – but in that case the ‘aggregation’ language seems out of place. Perhaps the language was intended to invite creation of simple models – an activity that is legitimate enough in general, but for the origins problem it seems particularly subject to the hazards that Marshall pointed out.11 The tools of computational modelling (as distinct from Marshall's ‘mathematics’) make formal theorizing feasible for complex problems – such as explaining the origin of capability by reference to both initial individual abilities and shared experience within the organization. This is well illustrated by the model of Miller et al. (2012). A key question, then, is whether we understand ‘origins’ to refer to events that are located in time. The connotations of ‘origin’ can be spatial rather than temporal, or an account of origins may feature causal mechanisms, together with some temporal or spatial reference. For example, Darwin's account of the origin of species offers a time-embedded causal mechanism: for any particular species there was a time when it first appeared, and this happened ‘by means of natural selection’. According to the evolutionary interpretation of ‘origins’, the way to understand the origins of any particular phenomenon is first to ask what preceded it in time – and then ask what mechanisms account for the transition between what was observed earlier and what was observed later, the phenomenon in question. The most likely first-order answer to a question about the origin of a particular capability is, ‘it evolved from another capability, one that was similar in many respects’ – exactly paralleling Darwin's answer for species.22 I would not make the same statement regarding routines in general, since a richer set of possibilities appears at more micro levels of organizational behaviour. In the case of capabilities, it can happen that the ‘ancestor’ is a more chaotic, ad hoc version of a similarly motivated undertaking – one that might be insufficiently routinized and reliable to count as a ‘capability’. The study of origins of capabilities is primarily a study of transition (and transmission) mechanisms between ancestors and descendants, and it merges continuously into the study of incremental change of existing capabilities (much as it does for species). This observation raises the question of how we demarcate discrete, stationary categories within a process of relatively continuous change; see Birnholtz et al. (2007). Is the capability to provide airline passenger service with jet aircraft different enough from the corresponding case with piston engines to be a different capability, with a different origin? A general-purpose answer to such questions is neither available nor necessary, but it is important to remember that they are there. If the story of a particular organizational capability is pursued back into history, the quest virtually always turns up a situation where the organization was much smaller and the core performance was primitive and unreliable by current standards. The specific individuals involved, and their abilities, were then much more prominent in the picture. Also, historically significant inventions loomed larger in the technology picture than any particular technology does in the descendant capability today. Seek the origins of the capabilities of today's Ford Motor Company and find Henry Ford, among other individuals; but find also the invention of the internal combustion engine. These are significant partial responses to the origins question regarding Ford capabilities today – but getting these puzzle pieces in hand does not provide much of a picture of where Ford's current capabilities came from. It is the evolutionary process of ‘descent with modification’, operating over decades, that accounts for that picture. Among the phenomena that need to be addressed are the obvious facts that some organizational capabilities persist in recognizable form beyond individual lifetimes and (more plainly) beyond the typical tenure of individuals in organizational roles. Also, many individuals enter organizations through the routinized procedures enacted by HR departments, which surely are among the most powerful shapers of connections between the individual and organizational levels, and thereby are a major source of continuity/inertia in organizational character. These facts involve, obviously, the temporal setting of the relationships between individuals and the organization. We will not benefit much from an ‘individuals first’ account of origins that has no convenient place for these obvious facts, with their intrinsic connection to the element of time. There are already many studies in existence that can be thought of as illuminating the origins question via a focus on the historical evolution of technologies and organizations, though these contributions have been little noted in the recent discussion of the microfoundations and origins issues in management and organization theory; see David (1992); Usselman (1993); Raff (2000); Rosenbloom (2000); and Murmann (2003), among many others. Historians naturally tend to keep the time axis in view, and by doing so they produce valuable insight into the questions posed for this special issue – although not one of those questions was plainly posed as a historical question. The question that I now propose, in response to the specific editorial request, is akin to many that the historians have illuminated, but requires contemporary ethnographic observation. When a specific capability first appears at a specific site, where does the requisite knowledge come from? This question embraces a broad spectrum of situations.33 A compelling study of a situation that is within that spectrum, and also highly relevant as to method, is Bechky (2003). The capability in question may be new to the world, or only to the site. Abundant examples at the latter end are provided by the activities of replicator organizations, which go about the world creating new ‘copies’ of productive facilities they have already established elsewhere (Winter, 2010). The perspective offered here is that this is a continuum in which common mechanisms operate, in varying proportions, at all points. Thus, there is creativity, and problem-solving, even in the most straightforward replication cases. There is re-use of existing knowledge in the mix that yields the most innovative results. For various reasons, the organizational aspects of the problem are more prominent where novelty is less extreme. The challenge is to understand the mechanisms, and how the mix responds to the specifics of the situation. To explain the interest in the question, it is important first to emphasize how fundamentally the idea of productive activity summarized by ‘organizational capability’ departs from a prominent alternative, the ‘recipe theory’. The recipe theory goes unquestioned in mainstream economics (where it is called ‘production theory’), more importantly, it goes largely unquestioned in an enormous amount of less formalized discourse. According to the recipe theory, having the recipe and the means to assemble the named ingredients suffices to confer the ability to produce the output described.44 As I hope is obvious, I use the term ‘recipe’ here to reference the full range of symbolically rendered accounts of how a thing is done, including, for example, manuals, blueprints and instructions delivered orally. Problem: How does the symbolic account in the recipe actually shape its enactment – particularly, via the muscles of the enacting humans, including the muscles employed in the wrench turns, mouse clicks, and sceptical looks, and in the generation of other signals received by other personnel as well as by the diverse physical artefacts in the organization? It is the indubitable existence of this gap that underpins confidence in this proposition: even the most straightforward re-enactment of existing knowledge necessarily involves new learning if it involves a new performer – and the more so in a new context. Further, this new learning frequently gives rise to significant novelty. By studying the arrival of a specific capability at a particular site, we could assess the force of the recipe theory against the capability theory. The capability theory, of course, does not deny that recipes matter. They matter a lot more in some situations than in others, and it is important to understand the distinctions among the situations.55 The accumulating evidence on heterogeneity speaks very directly to the contrasting expectations of the recipe theory and the capabilities view of knowledge, with the evidence now strongly tilted in the latter direction (see Jacobides and Winter, 2011). A researcher stationed at an incipient site of capability ‘origin’, and with good access to the participants, could learn a lot about where the knowledge was coming from. The immediate origins and important content of the key recipes could be identified. The enactment of the recipes could be observed; the reasons why it was sometimes straightforward and sometimes problematic could be pursued. The processes by which the organization arrived at the answers that were not already there, ‘in the book’, could be identified, as well as the circumstances where adherence to the book produced more problems than progress. The role of diverse local contingencies – which capabilities theory deems important collectively – could be assessed. A programme of research of this kind would provide fundamental insight into the micro-level origins of the capabilities in use today, by showing in detail how the process of descent with modification actually works in the domain of capabilities. I am indebted to Michael Cohen for constructive comments on an earlier draft; the usual exoneration is supplied. Research support from The Mack Center, The Wharton School, is gratefully acknowledged.