Given that the methodological issues surrounding the agent–structure relationship are of most interest to the majority of practising IR researchers, it is surprising that they have been accorded so little attention. Gil Friedman and Harvey Starr attempt to illuminate the methodological consequences of the agent–structure problem, but ultimately their account is very one-dimensional owing to the lack of attention paid to metatheoretical matters. This means that the ontological issues at the heart of the agent–structure problem are not adequately addressed, and as a result the methodology they propose, methodological individualism, is limited and has been the subject of a sustained critique. The two theorists who have addressed the methodological difficulties – Martin Hollis and Steve Smith – have consistently argued that ‘there are always two stories to tell, one explanatory and the other interpretative, and that they cannot finally be combined’. In chapter 6 I presented their arguments, correctly in my opinion, since this is how Hollis and Smith themselves depict them, as epistemological in nature. As they put it, ‘[e]pistemologically, however, compromises seem to us far less problematic on the vertical axis than on the horizontal.’

As it transpired, however, (1) these epistemological arguments, were more accurately described as methodological problems embedded within ontological assumptions; and (2), where genuinely epistemological, the conclusions drawn by Hollis and Smith did not follow from the premises.