Visual data have been of concern to the social sciences in two ways: visual records produced by the investigator, and visual documents produced by those under study. In recent years, however, this dichotomy between the observer and the observed has begun to collapse (as it has across the qualitative social sciences more generally) and a third kind of visual record or more accurately representation has emerged: the collaborative representation.

Thus visual anthropology and visual sociology proceed methodologically by making visual representations (studying society by producing images), by examining pre-existing visual representations (studying images for information about society), and by collaborating with social actors in the production of visual representations. Methodologically, the use of photography, film and video to document areas of social and cultural life would appear to be straightforward and unproblematic.

Photography was also employed as a 'visual notebook' by anthropologists to document aspects of material culture produced by a particular society. After the invention in 1895 of the portable motion picture camera, film was employed to the same ends. In recent years anthropologists and others have begun to re-examine the products of colonial photography, being as interested as much in the ideas that led to the production of such photographs as in the societies and cultural forms they supposedly document (see the essays in Edwards 1992, and Scherer 1990).

The study of early ethnographic film is less well advanced, largely because the sources are less well-known and less accessible, but an ESRC-funded project is currently underway at the University of Oxford to catalogue much of this earlier material (see under 'Electronic Resources' below), and it is hoped that this will stimulate further research. Following on from the Victorian taxonomic and classificatory uses of visual media, photography, film and video have been used more recently to gather data for various other kinds of formalist analysis: proxemics, choreometrics and kinesics and conversation analysis.

What many of these recent projects have in common with their Victorian and Edwardian antecedents is an approach to mechanical visual recording media, which tend to treat them as neutral technologies capable of objectively recording social behavior or visible 'givens'. Images are no more 'transparent' than written accounts and while film; video and photography do stand in an indexical relationship to that which they represent they are still representations of reality, not a direct encoding of it.

As representations they are therefore subject to the influences of their social, cultural and historical contexts of production and consumption. Thus the visual sociologist or anthropologist adopts a dual perspective on visual media. On the one hand they are concerned with the content of any visual representation what is the 'meaning' of this particular design motif on an art object? who is the person in the photograph?

On the other hand, they are concerned with the context of any visual representation who produced the art object, and for whom? why was this photograph taken of this particular person, and then kept by that particular person? When studying visual representations that have been created by others the dual strands of content and context are fairly easy to investigate in tandem. When, however, the investigator produces the visual representations there is a danger of the content taking priority over the context.

Within documentary film, the 'direct cinema' movement in the 1960s sought to correct this imbalance by ensuring that the conditions of filmmaking were revealed to the viewer (see Barnouw 1974, Rosenthal 1988, and Loizos 1993). Typically this involved the deliberate inclusion of the filmmakers' kit in the image or even the filmmakers themselves. Such ideas were absorbed into ethnographic film practice, simultaneously with techniques that were thought to bring the human subjects of the film closer to the viewer. With still photography, more sensitive or reflexive representations are perhaps slightly harder to accomplish.

In many cases, social investigators choose to create some marriage of text and image, where each provides a commentary on the other. Doug Harper, a visual sociologist, has accomplished this to particularly good effect in his work (Harper 1987; see also Berger and Mohr 1975). It is important to remember, however, that all visual representations are not only produced but are consumed in a social context, one that invokes a family resemblance to similar representations television and cinema in the case of film and video.

Members of an audience will bring to the screening certain expectations of narrative form, 'plot' development, 'good' and 'bad' composition, and so forth, however unconscious or inchoate their understandings. Nor can a single 'reading' of a film necessarily be presumed. Sociologists such as Stuart Hall have advocated the notion of 'preferred readings' (Hall 1977), while an anthropological study of ethnographic films shown to students refutes the liberal assumption that such films encourage the viewers empathetically to narrow the gap between self and a radically different other (Martinez 1990).