¹ See Williamson (1993: pp. 139ff.), who, along with Douglass North (1991) and Joel Mokyr (1992) embraced the term 'path dependency.' The latter is a variant (or, for those inclined to evolutionary modes of discourse, a 'micro-mutant form') that systematically is 'selected against' throughout these pages -- wholly on grammatical grounds. Consulting Brown (1993) for the definition of 'dependency,' one is reminded that the noun describes "a dependent or subordinate thing, esp. a country or province controlled by another"; whereas, the same authority holds "dependence" to be "the relation of having existence hanging upon, or conditioned by the existence of something else; the fact of dependency on another thing or person." My terminological preference on this minor point, however, ought not be construed as reflecting an adverse judgement upon the substantive merits of what the abovementioned authors have had to say about path dependence.

² The Theory of History (New Haven, Yale University Press, 1925) was the third work of a trilogy by John Frederick Teggart. It has been belatedly recognized as a precocious and penetrating analysis of the intellectual roots of the separation between studies of history and the nineteenth century formulation of Darwinian evolution as a continuous, event-less dynamical process -- a conceptualization that, in addition to its dubious influence upon the field of evolutionary biology, was seen by Teggart to have led social science theorizing in a paradoxically ahistorical direction. See the reprinting of Teggart (1977); Eldridge (1985: esp., pp. 22-28) on its significance for modern evolutionary theory and the concept of 'punctuated equilibria. Further discussion of its bearing on the disciplines of economics and economic history may be found in David (1993:"Historical Economics").

³ I refer in this regard to Liebowitz and Margolis's (1990) dismissal of the version of the story of QWERTY presented by David (1985, 1986) as "a specious example of market failure borne in part from insufficiently rigorous examination of the historical record." The historical arguments and evidence offered to support that critique are re-examined in my forthcoming paper:'Neoclassical economists at the keyboard: Is there a cure for 'repetitive thought injuries"?'

⁴ This will provide an opportunity to indicate some of the dispersed locations in which my early publications on conceptual and analytical issues relating to path dependence can be found -- in aid of which a comprehensive listing my papers on those topics is given by the bibliographic references at the end of this essay. The latter citations -- a selective inventory of debris scattered alongside the trail that led me to "Clio and the Economics of QWERTY"(1985), and from thence to the present -- well may be the most generally useful contribution to emerge from the present undertaking.

⁵ Although these eventually had to be excised from the brief paper printed in the American Economic Review (1985), they survive in the full text published subsequently, see David (1986).

⁶ Another minor but possibly significant novelty consists in its being the shortest paper that I had hitherto put into print.

⁷ At the time it was not possible to draw upon the detailed research subsequently undertaken on the history of the standards rivalry in the VCR market, by Rosenbloom and Cusamano (1987), Baba and Imai (1990), Cusamano, Myolandis and Rosenbloom (1992), and Grindley (1995). Otherwise that might have served as still more 'topical' illustration of the point I wished to make for economists at large.

⁸ Compare the expression of those discontents in David (1969).

⁹ According to Liebowitz and Margolis (1995: 205), "The path dependence literature comes to us accompanied and motivated by a mathematical literature of nonlinear dynamic models, known as chaos or complexity models...." It will be pointed out below why this is incorrect from a formal standpoint, but it is also misleading as to the intellectual history of the subject, as would be evident from the absence of models of deterministic chaos in the papers on path dependence up to and including the American Economic Association Proceedings (May 1990).

¹⁰ See Koot (1987), Ch.6, and further elaboration in David (1992):"Invisible Hand".

¹¹ Liebowitz and Margolis (1995b: pp. 209-210) fall into just this confusion on the one occasion on which they offer a formal definition of the meaning of 'path dependence'. They say, correctly: "The meaning closest to current use in economics is that of stochastic processes that incorporate some concept of memory." But thereupon draw from the Encyclopedic Dictionary of Mathematics (Cambridge, MA: MIT Press, 1987) the following definition of 'path dependence': Letting P(n) be the probability of event E(n) =A(1) on the n-th trial, and (1-P(n)) be the probability of E(n) = A(2) A(1), at that trial, then the general 'response probability' for the sequential process is: P(n+1) = f {P(n), E(n), E(n-1),...,E(1)}. When the function f = f {P(n), E(n), E(n-1),..., E(n-d)}, the response probability is said to be 'd-trial path dependent. "In the special case where d=0 it is 'path independent.' The text in Liebowitz and Margolis (1995b: p.210) then goes on to claim, quite erroneously:"The use of path dependence in economics is, for the most part, loosely analogous to this mathematical construction: Allocations chosen today exhibit memory; they are conditioned on past decisions." If 'allocations' are associated with 'events', E(i), and (probabilistic) decisions at moment n are characterized by the pairs [P(n); 1- P(n)], then the foregoing statement does not correspond to the mathematical construction of d-trial path dependence and, more than the latter, corresponds to the generic usage of path dependence in economics.

¹² The reference in the passage quoted to "contingency" as the conceptual counterpart in biology of the idea of path dependence is followed by Liebowitz and Margolis's (1995b: 33) statement that "In Wonderful Life, Stephen J. Gould applies this intellectual revolution to paleontology." But, it should be shiningly clear from that work by Gould (1989: pp. 282ff, esp.), and really no less from his earlier writings, that he is not drawing upon a recent intellectual revolution: "I regard Charles Darwin as the greatest of all historical scientists. Not only did he develop convincing evidence for evolution as the coordinating principle of life's history, but he also chose as a conscious theme for all his writings...the development of a different but equally rigorous methodology for historical science....Historical explanations take the form of narrative: E, the phenomenon to be explained, arose because D came before, preceded by C, B, and A. If any of these earlier stages had not occurred, or had transpired in a different way, then E would not exist (or would be present in a substantially altered form, E', requiring a different explanation)....I am not speaking of randomness (for E had to arise, as a consequence of A through D), but of the central principle of all history--contingency." [Gould (1989: 282-283)] Further on, Gould (1989:283-284) writes of the universal psychological appeal of the notion of historical contingency, in terms that leave no doubt that this is not a concept specific to evolutionary biology: "Historical explanations are endlessly fascinating in themselves, in many ways more intriguing to the human psyche than the inexorable consequences of nature's laws....Contingency is the affirmation of control by immediate events over destiny....Contingency is a license to participate in history, and our psyche responds. The theme of contingency, so poorly understood and explored by science, has long been a mainstay of literature....Tolstoy's theme in all his great novels." What Gould provides in Wonderful Life is a new interpretation of the record of life left in the Burgess Shale, but, as he takes pains to acknowledge, this interpretation "is rooted in contingency" -- a very old and far from revolutionary idea.

¹³ The latter may then be subjected to criticisms from which the original analysis would be immune. A striking instance of such a switch is to be found in Liebowitz and Margolis's (1995b: 214-215) reproduction and critique of a deterministic payoff tableau, used by Arthur (1989) purely as a heuristic device -- to convey the possibility that a sequence of myopic adoption decisions under increasing returns to adoption could result in the commitment of the ensemble of adopters to a dominated outcome. In the course of pointing out that the payoff tableau may be read in a way that is inconsistent with the results reported for Arthur's stochastic model, there appears the following commendably candid footnote (pp. 214-215, n. 15): "Actually, Arthur states that this example does not exhibit any 'non-ergodicity,' meaning that it is not path dependent in the sense that small differences in historical sequences play a role in the final equilibrium. In this example the end result is the same no matter the order of initial participants. But it illustrates lock-in very well." I might note that this footnote is the only place I have found in Professor Liebowitz and Margolis's publications on path dependence where the concept is explicitly defined with reference to non-ergodicity, and even it omits reference to probability.

¹⁴See endnote 1 on the substitution of 'dependency' for 'dependence'.

¹⁵ It is a general observation that straw men are invariably given labels associated with the work of others than the individuals who construct them. Mokyr (1990:163) begins the paragraph immediately following the passage quoted in the text by observing: "This explanation of technological progress may appear trivial to some and false to others." There follows a review of some "mechanisms for this autocorrelation" in the level of technological creativity, whose inadequacies as guarantors of unbroken chains of innovative success are deftly noted. Says Mokyr (p.164): "...these sequences do not provide a wholly convincing account of technological advance, [as] it is misleading to think that nothing leads to technological progress like technological progress....Neither nature nor history can lock a society forever into a dead-end technology." Indeed. Of course, it is only Mokyr himself who at this point in his book has mis-labeled what is little more than 'the general theory of the virtuous spiral' with the tag 'path dependence.' A bit further on, Mokyr (pp. 164-165) appears to take a different tack when, without explicit notice, he shifts the explains of the supposed "path dependent theory of technological change" from the rate of inventive activity to its direction: "The path-dependent nature of technological change, in which its course is explained mainly by its past, can be extended, though caution must be used in applying these models. The links with the past must be specified rather than assumed." [Emphasis added.] From the whole presentation a reader would have to be forgiven were he or she to form the utterly mistaken impression that path dependence is: (a) a theory about the sources of technological creativity, (b) the claim that technical progress is temporally auto-correlated, (c) a trivial or false theory, because it says merely that success breeds success, and failure is followed inexorably by more failure -- both of which are manifestly not true.

¹⁶ Closely following this, Mokyr's (1990: p.286) comments on the competitive diffusion of technological compatibility standards and conventions, and makes it clear that his evolutionary perspective on technological change does associate the concept of path dependence with 'weak' Darwinian selectionism, and the consequent possibilities of outcomes that may be locally or situationally efficient, yet not globally optimal: "There is no optimality per se in these standards [33 1/3 rpm for playing phonograph records, or driving on the right side of the road] but given that they exist, they have to be accepted, and thus impose a constraint on the techniques that can be used....Yet not all specific behavior has definite adaptive meaning, just as not all technical conventions are necessarily efficient."

¹⁷ See the bibliographic references for David, and co-authors, especially the publications from 1985 to 1997.

¹⁸ David (1985, 1986) tried out the neologism "QWERTY-nomics" when introducing the analytical elements of the story of the emergence of the QWERTY keyboard format as the de facto standard in the U.S. typewriter industry, but the more transparent phrasing of the titles ("the economics of QWERTY") took increasingly firm hold, as in Krugman (1994: Ch.9). Notice that the discussion there, and in Krugman (1991:100), is careful to associate the potential for lock-in to Pareto inferior equilibria with the economics of QWERTY, rather than with the more general phenomena of path dependence. This logical separation has been blurred in part by the rhetorical success of QWERTY, for which David (1985) may be blamed. Some of the credit for the resulting confusion also must be shared, however: the distinction also has been obfuscated by Liebowitz and Margolis's (1990) attempt first to refute the view that the early de facto standardization on QWERTY represented a sub-optimal selection by the market, and subsequently (1995b) to impose a taxonomy in which a particular form of persisting sub-optimality was labeled "third-degree path dependence." The second of these maneuvers is examined in section 5, below.

¹⁹ See, e.g., the discussion in David (1971), or David (1975: Ch.5).

²⁰ Closer analysis of both the 'naive' and 'sophisticated' economic interpretations of Veblen's (1915) argument, and their treatment in subsequent studies by economic historians may be found in David (1975), Ch. 5.

²¹ This is one of those instances in which a casual, heuristic analogy becomes confused with, and actually supplants the thing it had been invoked to explain. From the formal standpoint it is quite inappropriate to bracket the theory of chaos with that of stochastic processes which exhibit path-dependence. Although strong positive-feedback (or, alternatively, additive interactions) will result in the system's inability to shake off its past (non-ergodicity), mathematically the latter is something different from 'sensitive dependence on initial conditions.' See, e.g., Ruelle (1991: Ch.14), Steward (1990).

²² See, e.g., David (1988), David (1994:'Positive Feedbacks) and Durlauf (1996).

²³ Some will have been introduced to the existence of these by more recent economic applications of interacting particle models whose dynamics are non-ergodic, such as have appeared in Arthur (1988a), David (1988, 1993), Durlauf (1990), and Weidlich and Haag (1983). On spin systems, particularly, see Ligget (1985), and discussion in David, Foray and Dalle (1997). On percolation models, see Grimmet (1989), and other references in David (1988, 1993), with economic applications in David and Foray (1993, 1994, 1995), and David (1997).

²⁴ The properties of one such spin system, the stochastic formulation of the Ising model of ferromagnetism, were studied thoroughly by Russian probability theorists during the 1960s. Results from that analysis were taken up quite quickly by mathematical general equilibrium theorists working on what at the time were called random economies. The path-breaking paper by Hans Föllmer on general equilibrium in random economies with interacting agents built directly upon those results from Markov random field theory. David, Foray and Dalle (1997) give references and a discussion of the early contributors to this literature.

²⁵ See e.g., David (1993a, 1997), David and Foray (1993, 1994).

²⁶ For discussion of this in the context of technical compatibility standards, see, e.g., David and Greenstein (1990); on social conventions, organizational routines and formal institutions, David (1994, Les Standards), and David (1994, Institutions).

²⁷ A career engineer officer,Rickover intially had been seconded to Oak Ridge to do a paper study of a liquid-metal-cooled reactor for a destroyer, and in 1948 work on sodium-cooled reactors was underway at the Argonne Laboratories, and in the General Electric facility at Knolls, near Schenectady; but once Rickover's attention shifted to the application of nuclear propulsion in submarines, the liquid sodium designs were rejected in favor of developing a light-water pressurized water reactor. For other details, see the history related by Cowan (1991).

²⁸ On the latter, and, in the instance of the accident at Three Mile Island, another illustration of hysteresis effects in the industry, see, e.g., David, Maude-Griffin and Rothwell (1996).

²⁹ See, e.g., David (1987, 1990); David and Bunn (1988), Cowan (1991).

³⁰ Compare the detailed analyses of the VHS market in Baba and Imai (1990), Cusamano, Myolandis and Rosenbloom (1992) and Gridley (1992), none of which are noticed in Liebowitz and Margolis (1994), or the latter authors' subsequent references to this case.

³¹ Compare, e.g., the critical discussion of the early history of VCRs in the discussion of network externalities by Liebowitz and Margolis (1994, 1995a, 1995b, 1995c).

³² Especially in David (1987), David and Bunn (1988), David and Greenstein (1990) and, most forthrightly in David (1992).

³³ See David (1993, Historical Economics) for more on the teleological mode of analysis in economics.

³⁴ For further discussion of the latter topics, see, e.g., David (1994), David and Foray (1993).

³⁵ On these epistemological topics, see, e.g., the stochastic models developed in David (1997), and David and Sanderson (1997).