The theory (or rather the notion) that the international economy functioned more or less effectively for roughly a century down to 1914 because Great Britain provided the ‘public goods’ required for the smooth operation of the ‘liberal international order’ has become a textbook generalization. That notion emerged quite recently and can be traced to Kindleberger's attempt to explain the pronounced cyclical fluctuations experienced by the world economy during the interwar years 1919–39, as well as the severity and duration of the Great Depression from 1929–33 in terms of the American failure to sustain conditions necessary for the financial stability of an interdependent global economy. In Kindleberger's view, Britain, which had acted as a hegemonic power before 1914, lacked the resources to continue with its historic role after the Great War, while the United States (which by 1918 enjoyed a position in the world economy of arguably greater weight and significance than the United Kingdom had ever possessed during the long nineteenth century) commanded neither the knowledge nor the political will to replace Britain as the responsible hegemonic power until after the Second World War.

Arnold Toynbee was Director of Studies at Chatham House from 1924 until his retirement in 1954. His fame rests on two monumental projects to establish a global perspective. The Annual Survey each year reviewed the world as a geographical whole, selecting four or five themes for special treatment. It was not an edited volume. The Study of History put Western civilization in the context of a historical survey of all previous civilizations. Quite apart from his considerable journalistic output, the thirty-four volumes of the Survey and the ten volumes of the Study show how seriously Toynbee had taken a schoolmaster's injunction ‘first to see your subject or your problem as a whole’. Toynbee believed that a global perspective in terms of both time and space was of practical use.

The Italian city-state system occupies a special place in the canon of orthodox international relations. For, as Martin Wight says, ‘it was among the Italian powers that feudal relationships first disappeared and the efficient, self-sufficient secular state was evolved, and the Italian powers invented the diplomatic system’. And of course this was not all they invented. In addition to the earliest modern discourse of Realpolitik (‘Machiavelli’, Carr tells us, ‘is the first important political realist’), it is in the Italian city-states that we find the first routine use of double-entry book-keeping, of publicly traded state debt, of marine insurance, of sophisticated instruments of credit (such as the bill of exchange), of commercial and banking firms coordinating branch activity across the continent, and so on. Here, too, the citizen militias gave way earliest to the mercenary armies that would later characterize European Absolutism; and within the town walls, a population given over increasingly to commerce and manufacture elaborated new forms of urban class conflict.

‘World order’ has become a current catchphrase of political discourse and journalism. ‘Multilateralism’ has become something of a growth sector in academic studies. What current events have brought into prominence, scholarship has an obligation to subject to critical analysis. This article raises some of the questions that should be probed in this analysis.

I welcome this opportunity to respond to Martin Hollis and Steve Smith's ‘Beware of Gurus: Structure and Action in International Relations’, their reply to my review2 of their book, Explaining and Understanding International Relations. Their constructive comments have helped me clarify my own thinking, and I hope by extending my previous remarks in the same constructive spirit I can return the favour. In ‘Beware of Gurus’ they took up both issues I raised about their book: the relationship between the levels-of-analysis and agent-structure problems, and that between causal and interpretive explanations. In part for reasons of economy and interest, and in part being more persuaded by their comments regarding to the latter, I shall limit myself here to the former, taking issue in particular with what I see as their reduction of the agent-structure problem to one of levels-of-analysis.

We were at first puzzled by Alexander Wendt's latest rejoinder, since systemic theorists need concede nothing in recognizing states as self-interested actors with identities and interests. They have only to add that these identities and interests are shaped by the system in general and are given specific direction at any one time by the systemic pressures operating in the previous period. After all, ‘self-interested’ is a dummy term, until supplied with content, and systemic theorists can regard the system as the source of what matters. Explanations which proceed ‘top-down’ by explaining the behaviour of the units in terms of the system need some kind of feed-back mechanism involving the units. It is no objection to systemic theory that the units contribute to the process, provided that they do so in ways shaped by the demands of the system. In short, why does Wendt believe that the identities and interests of actors cannot be formed by the system? Waltz certainly offers a powerful explanation cast in precisely these terms.