Mike Frank quite helpfully recalls last year's (interminable?) discussion about objectivity and cinema. In the cases Frank cites, the image seems *more* reliable than other elements, but one could probably think of numerous examples in which the image is *less* reliable than we're inclined to remember--even without going to the extreme of 'lying' flashbacks of the _Stage Fright_ and _Mortal Thoughts_ variety. One small bit of evidence against the visual image always being highly reliable (tantamount to 'objective reality') would be those films in which, for instance, a shift in time takes place not during a cut between shots but actually during a shot. This can be aided and abetted with sound bridges. E.g., there was an _X Files_ episode a couple of years ago in which the flashbacks were handled in such a way that a character would relay events in the past, and the shot-reverse shot of the narrating character in the past would suddenly and without warning become the shot of the character relating the events in the present. Thus a single shot would change temporal status--retrospectively. I know there are theatrical films in which this happens, too--I was just reading about an example--but I can't think of any right off, and I can't recall the example I just read (!). In such cases the shot itself does not yeild up all the characteristics of the denoted situation, and thus the image becomes a poor or misleading indicator. I watched Tarkovsky's _Nostalgia_ recently, and he does something similarly misleading. The film is shot in color, but it is punctuated with sequences in black-and-white or perhaps merely a more reduced color palette. Some of these can be interpreted as flashbacks; others may be events taking place far away (in Russia, as opposed to Italy, where most of the film is set); still others are dreams. By the end of the film one can no longer tell. Thus in the case of _Nostalgia_ the image seems to be visually differentiated in such a way as to allow us to understand its epistemological status, but then the marking system breaks down or becomes too vague to be reliable. Perhaps these examples would allow us to imagine a continuum between strong and weak semantic saturation of the image. I.e., some images seem to register quite fully all of the detail which is relevant to our decoding the fictional denotation: we can see in the image the time period during which it takes place, the social status of the character, the time of day, etc. In other films individual shots are less densely packed with information which allows us to decode their narrative status and relationships to adjoining shots. Again Tarkovsky comes to mind: his characters often address unseen, offscreen characters, and we are often in doubt about whether there's anyone there or not. (Woody Allen steals this trick for the ending of _Interiors_: is the daughter speaking to the mother or to herself?) With Tarkovsky we often cannot even tell if two subsequent shots are in the same place and time. Mike Frank's discussion of various kinds and degrees of contradiction between the image and the voiceover also suggests an interesting perspective on film, one in which we examine the degree to which contradiction amongst cinematic elements becomes meaningful or disruptive. (Sometimes in a poorly made film, one thinks: is this a clue? or a foul-up?) Sincerely, Edward R. O'Neill UCLA General Education Program ---- Screen-L is sponsored by the Telecommunication & Film Dept., the University of Alabama.