The contrast goes back to the founders, including Marx, Durkheim, and Weber. And authors such as Anthony Giddens, Maurice Godelier, and Pierre Bourdieu provided second-generation refinements. But new ideas have come forward in the past decade that merit reconsideration.
Some of this new thinking is contained in Peter Martin and Alex Dennis, eds., Human Agents and Social Structures. Here is a thumbnail of the approach taken by this group of philosophers and sociologists.
[The current book's] aim is to be an intervention which seeks to make the case that structural, system, or holistic approaches to the understanding of social life and the explanation of human action are fundamentally misconceived -- as, equally, are efforts which rest on individualistic assumptions. In essence, our view is that human social life is conducted in and through patterns of collaborative interaction: sociologically, our interest is thus not in the subjectivity of individuals but in the ways in which intersubjectivity is achieved and maintained. (7)This group of researchers addresses the contrast between agency and structure; but really their goal is to help to dissolve the distinction. They want to show that "structures" do not exist in any strong sense (including the senses associated with critical realism), and that a proper understanding of "agency" involves both subjective and objective features of the individual's actions, thoughts, and situation. Social relationships are densely intertwined with reasons, emotion, commitments, beliefs, and attitudes -- the aspects of consciousness that make up agency and action.
Here is a representative statement about social structures:
The collective concepts (such as family, state, organisation, class and so on) -- which have often been seen as fundamental to sociological analysis -- have often encouraged 'the temptation to reify collective aspects of human life' (Jenkins 2002a:4); that is, to treat them as if they were real entities, independent of the human beings who constiTute them. (7)Their affirmative theory of agency -- now stripped of the notion that it is a polar opposite to structure -- has much in common with the traditions of micro-sociology -- Goffman, ethnomethodology, symbolic interactionism, and phenomenological sociology. The idea here is to emphasize the very concrete ways in which each of these traditions succeeds in identifying the agent, the social actor, as both subjective and objective. He/she is a subject, in the sense that the agent possesses thoughts, emotions, desires, aversions, allegiances, and the like, which in turn contribute to the actions and lives they live. But the agent is objective, in the sense that he/she is embedded and developed within a concrete set of social relationships and institutions.
Thus each of these approaches develops in its own way the idea that human social life is carried out through processes of interaction among real people in specific situations, and each seeks to avoid the reification of collective concepts -- there are no such 'things' as social 'structures,' 'classes', or indeed 'societies', yet terms such as these are indispensable, not only for sociologists but for the purposes of everyday communication. (14)Martin and Dennis quote Anthony King with approval: "human agency is a collective product, germinated with others and dependent upon the social networks in which we al exist. Human agency is better understood as the collective product of social relations ... than as an autonomous individual power" 14).
So is there space within this view of sociology for investigation of "macro" features of society? They argue that there is, but not as an autonomous domain:
There is nothing in the perspective developed here that would prevent the investigation of social phenomena conventionally described in terms of 'macro' structures. But what we are suggesting is that, for example, the 'class structure' must be conceived as the outcome of stratifying processes and practices, many of which -- like the grading of students' work or the awarding of educational credentials -- may appear to be routine and mundane. This interactionist sociologists have investigated the processes through which social class differences in educational attainment are produced. (15)One reason I'm intrigued by the approaches taken to this aspect of social ontology by the contributors to Human Agents and Social Structures is that these approaches seem to converge with the ideas about "methodological localism" that I've been drawn to as a bottom-level description of social ontology (link). On this approach, neither "structure" nor "agent" can be specified in its own terms alone; rather, we need to base our social concepts on the socially situated and socially constituted individual, located within a set of locally manifest social relations (link, link). So we cannot separate agency and social location (structure); rather, the fundamental unit of social activity involves both aspects.