Does the Viewpoint Deviation Effect Diminish if Canonical Viewpoints are used for the Presentation of Dynamic Sequences?

Two studies examine the visual presentation of dynamic sequences. Experiment 1 tests if there are canonical viewpoints, that are especially appropriate for presentation. Participants agreed that viewpoints with 90 degree deviation between axis of sight and axis of main movement in the sequence are better than other viewpoints. Experiment 2 examines if these canonical viewpoints weaken the perspective deviation effect in a recognition task according to their postulated information richness. A perspective deviation effect was found both for canonical and less canonical views, even if it was weaker for the canonical views. Viewpoint Deviation and Canonicality This paper deals with questions concerning the cognitive representation of visually presented dynamic sequences, specially the role of viewpoint. A first experimental series (Garsoffky, Schwan & Hesse, 2002) showed that the viewpoint from which one sees a dynamic sequence becomes part of the cognitive representation of that sequence and therefore influences later memory retrieval processes. This viewpoint deviation effect appeared in three experiments examining recognition memory for visually presented dynamic sequences (Garsoffky et al., 2002) and comprises the stable result, that cuttings from sequences are best recognized if they are presented in a viewpoint most similar to the viewpoint from which participants before saw the whole sequence. This means the cognitive representation of dynamic sequences is not uncoupled from the viewpoint from which one primarily saw the specific sequence and therefore influences later memory retrieval processes. The question now is, if this viewpoint deviation effect holds for all kinds of viewpoints or if the use of special viewpoints may reduce this effect. The following studies therefore ask, if various viewpoints differ in their qualification to present a sequence – i.e. if there exist so called canonical viewpoints that could by now only be shown for static objects (e.g. Palmer, Rosch & Chase, 1981) (Experiment 1), and further it will be investigated, if these canonical viewpoints have an influence on the viewpoint deviation effect found by Garsoffky et al. (2002) (Experiment 2). The concept of canonicality in connection with visual viewpoints was firstly empirically investigated and defined by Palmer et al. (1981). They discuss the idea of canonical viewpoints from an information-processing approach, a categorization perspective, in terms of phenomenology, and with regard to the concept of affordances (Gibson, 1982) and they conclude, that canonical viewpoints compared to other viewpoints contain more information as well as information of high salience, are the most typical viewpoints of an object, are those viewpoints from which an object is most perceivable, and are especially qualified to present the affordance structure of an object. Empirically canonical viewpoints are defined e.g. by asking participants to imagine an object and then to describe the viewpoint from which the imagination took place, or participants were asked from which viewpoint they would make a photo of an object, or participants had to choose between photos with varying viewpoints which photo in their opinion presented the object best (Blanz, Tarr & Bülthoff, 1999; Palmer et al., 1981). Evidence for canonical viewpoints is stated if there is high interand intraindividual agreement. At least for static objects some conclusions about the nature of canonicality can be made that do not mutual exclude each other. (i) Functionality and familiarity: Especially objects of everyday life we often see from a specific viewpoint that corresponds with the functionality of that object, i.e. when interacting with that object we see the object from a specific, i.e. canonical viewpoint that allows optimal interaction (Blanz et al., 1999). (ii) Information richness: In some studies canonical viewpoints were discovered even for abstract or nonsense objects – a fact that can not be explained by familiarity or functionality (Cutzu & Edelman, 1994; Edelman & Bülthoff, 1992; Perrett & Harries, 1988). It was concluded that canonical viewpoints present more information and especially more salient information of an object than other, less canonical viewpoints. They present a high number of visible surfaces of an object, important parts of an object are not covered, and they are stable against small variations of the viewpoint, i.e. the informational advances of that viewpoint remain the same even if the viewpoint is changed slightly (for a comprehensive list see Blanz et al., 1999). (iii) Discriminability: Cutzu and Edelman (1994) concluded from their findings using abstract objects, that because of limited cognitive capacity for every specific object only the diagnostically valuable attributes are stored, that help to distinguish this object from other objects. This means that it varies with changing contexts or tasks, which viewpoint is more canonical than other viewpoints. This paper investigates if the usage of such high informative viewpoints, that allow optimal discrimination, leads to a more viewpoint independent cognitive representation of visually presented dynamic sequences. I.e. the question is if so called canonical viewpoints help to recognize sequences better even if they are presented from new viewpoints.