The aim of this thesis is to challenge the tradition of theories of law to be theories of legal system. It argues that the frequently criticized individual failings of these theories are due to their common mistake in attempting to understand law in systematic forms at all. "Positivist" systems are composed of rules, norms or reasons (Hart, Harris and Raz respectively). These are linked into structured wholes by their sources. Dworkin's "Content" system links the content of "principles" into a whole which provides the "best justification" law. It is argued that from the point of view of legal actors, let al social observers, neither kind of legal system adequately encompasses describes law. Sociological theories of legal system are usually systems of individual roles and institutional functions that jointly perform the social function of law. These are rejected because legal institutions rarely fulfil their allotted function, yet do so much else besides. A non-systematic theory of society is then proposed. This helps explain why law, as a key part of that society, is unlikely to be understood in systematic terms. It also provides a model for a nonsystematic theory of law. Both law and society are seen to comprise three kinds of social relations, power (including authority), unintended effect, and value-effect (including norms). Unpatterned sets of relations found between persons in more frequent or intense interaction with each other than outsiders form institutions. These social relations are such that they generate disordered conflict within and between institutions, the former muting the latter. Law is seen as a set of institutions that are part of, and subject to, that conflict. It holds a special place within society because much of the conflict is channelled through the social relations within law.