Here I show why no one-to-one correspondence between language and the world is possible and how this is connected to the social nature of language-using and the normal conditions of empirical description. In particular, the role of the ‘open-texture’ of empirical terms in accommodating the epistemic possibilities opened up by the incompleteness of empirical description and verification is explored. The analysis of descriptive language-using also provides us with a paradigm of social activity and ontology. This allows me, throughout the work, to explore certain homologies along two dimensions of connection – the first defined by the consideration that science uses language and, being social, it must also be a possible object of social science; the second by the ontological sequence: meanings, beliefs, things. For example, a certain necessary ‘relativity’ in speech action is reflected in a corresponding epistemic relativity (central to Chapter 2) and in turn in a wider social relativity (discussed in Chapter 3), which is roughly the condition that we can act only relative to situations that we never individually create and are always effectively given to us. This might reasonably be said to be a transcendental property of social life – one of a number that we attempt to develop in this thesis.