An important philosopher and anthropologist of science, Bruno Latour has recently outlined an ambitious programme for a new sociological empiricism, in continuation of his actor-network-theory (ANT). Interrogating issues of description, explanation and theoretical interpretation in this ‘sociology of associations’, we argue that certain internal tensions are manifest. While Latour's philosophy of social science demands an absolute abandonment of theory in all its forms, proposing instead to simply ‘go on describing’, he is in practice employing versions of common sense explanation and pragmatic-constructivist theory to make ends meet. The core of this tension, we claim, can be located in Latour's meta-theoretical commitments, in effect obscuring important ways in which human subjects employ things, effects and symbols beyond their simple, ‘empirical’ existence. To illustrate these claims, we deploy the example of how morality works in social life, and coin the term quasi-actant, in allusion to the Latourian actant, to better understand such processes. Our overall criticism of ANT is immanent, aiming at the re-introduction of what we dub ‘virtual theory’ into Latourian empiricism, thus further strengthening what remains one of the most promising contemporary attempts to reinvigorate the sociological enterprise.