Levels of processing and memory for pictures
Picture recognition has long been believed to transcend the limitations of capacity and information loss found in studies of recognition in other modalities. Some researchers are critical of this beliefs however; they point out that, in conventional tests of recognition, target pictures and distractor items form a heterogeneous set. Thus one may not infer that because recognition rates are unusually high the targets have been represented in inordinate detail subjects could be using fragmentary representations which due to the nature of the recognition test still allow near-perfect performance. It was proposed that the representation of pictures occurs at two levels within the memory system. One "encoding domain" contains the specifically visual aspects of a picture the other contains a semantic representation. The visual "surface structure" permits recognition of the appearances of pictures the "deep structure" allows the picture to be the object of cognitive operations. It was suggested that the two domains might have different temporal characteristics visual detail being lost more quickly than thematic information. Two experiments were designed to test these hypotheses. A recognition task was used in which the distractor items were derived from the targets by transformations. Two types were used: one which altered spatial information but preserved meaning and a second which altered both visual and meaning information. Targets were either real-word scenes or random arrays of objects. Testing was conducted immediately or after a delay of one week. Controlling the similarity between targets and distractors produced recognition rates significantly lower those of traditional studies. In addition forgetting rates were much higher. There was some evidence for the differential loss of visual and semantic information. An unexpected finding was that meaningfulness of targets enhanced recognition more strongly at immediate testing when one might expect little difference between recognition real-world scenes and random arrays, due to the presence of well-defined surface structures. After a delay of one week, the effects of meaningfulness had not increased, in contrast with the predictions on the basis of the levels of processing framework. Thus schemata seem to guide the pick-up of visual information but have less effect on the memory process itself. A second series of experiments addressed the question of what effect attention to different levels of representation would have on recognition performance. The consistent finding of these experiments which employed a variety of orienting tasks designed to make salient either structural or thematic attributes in conjunction with distractor transformations intended to test for the encoding of different: types of information was that the basic pattern of recognition errors was similar across orienting tasks. Subjects appeared to be encoding the same information in all cases. The orienting tasks modified only elaborative activity. This result was interpreted in terms of an interactive model of perceptual processing in which information derived from different levels of analysis is communicated between different subsystems modifying their operations. The final series of experiments addressed the role of the criteria I task in picture memory. A reconstruction procedure was employed to assess pictorial recall uncontaminated by individual differences in artistic ability. The effects of meaningfulness were relatively stronger in the case of reconstruction compared with recognition. Spatial information unless it contributed to schemata was not well remembered. When reconstruction was used in conjunction with orienting tasks it was again found that the basic pattern of responding was similar between encoding conditions but once more elaborative activity appeared to have been influenced by the particular orientation. This time however semantic orientation resulted in superior performance over time. In addition differential accuracy of reconstruction for the horizontal and vertical dimensions of pictures suggested that pictures were construed and represented in depth. Finally significant correlations were discovered between subjects' rated mental imagery for scenes and their subsequent reconstruction, performance suggesting that imagery might well be an internalized analogue of the process of reconstruction. The differences observed between recognition and recall performance were attributed to the fact that a test of recognition provides information which may be reperceived guided by memory of the original experience. Recall however requires a reconstruction of the original experience and is thus subject to a greater number of intervening variables. The differences between the results of these series of experiments and those of similar verbal experiments, were interpreted in terms of the different roles that verbal and pictorial information play in communication. The former modality is used, primarily, as a vehicle for semantic information the latter for conveying information about the form and appearances of events, in addition, to their meaning. It was suggested that comparison of reading and picture perception might be a more fruitful approach to the investigation of the symbolic processes.