6.3 Monkeys
6.3.4 Inversion effects
The analysis above of the effects of reinforcement can be applied also to the mixed results regarding inversion effects in monkeys. Does a subject recognise a stimulus turned upside down as readily as one facing the typical end up? It is generally be-lieved that inversion effects can be attributed to the fact that certain expert stimuli, e.g. faces, are recognised by configural processing, and that the relations between features of a face get different from those stored in experience when the whole image is rotated (Vermeire and Hamilton, 1998).
However, this can be argued to be true for many local features as well, it is just a matter of where the line for holistic processing is drawn. Local features can be proc-essed configurally within its own boundaries. Responding to eyes instead of faces, for example, might very well entail configural processing of the iris and pupil’s
63 See section 11.3 for discussion of this paradigm.
tion to the sclera, while ignoring the eye’s relation to the face at large. Nevertheless, the result when encountering a view which is incongruent with the perceptual strat-egy you normally use is of course that recognition is slowed down.
Humans seem to have a strong inversion effect for faces. Presence or absence of inversion effects for facial photographs has therefore been used to argue for the pres-ence or abspres-ence of face specific visual processing in animals. Results have been mixed, with neurological (e.g. Bruce, 1982) or stimuli centred (e.g. Parr et al., 1999) explanations offered. As argued above, local or global processing strategies as a result of procedural variables is perhaps also worth considering, and notably the effect of reinforced drilling. Vermeire and Hamilton (1998) suggest that the number of ex-emplars during discrimination training might affect the processing mode, with local-ised strategies being the result of small stimulus sets. This might certainly be the case for some studies, but the stimulus quantity alone is unlikely to explain the effect.
To consider a few examples of studies that have drawn negative conclusions re-garding inversion effects in monkeys. Martin-Malivel and Fagot (2001), where ba-boons were found to respond to cut-out shapes rather than faces, found no inversion effect. Neither could Bruce (1982), Rosenfeld and Van Hoesen (1979) and Dittrich (1990). The latter with line-drawn facial stimuli (see fig. 10, right, p. 94). What these studies had in common was a testing paradigm based on heavily reinforced discrimination.
Studies of monkeys that instead have found evidence in the affirmative have relied on other types of paradigms, such as preferential looking (e.g. Tomonaga, 1994;
Swartz, 1982) or MTS (e.g. Parr et al.,1999; Phelps & Roberts, 1994; Overman &
Doty, 1982). A notable exception is Vermeire and Hamilton (1998) who found in-version effects in rhesus macaques using a go/no-go paradigm, but their monkeys were split-brain subjects and comparison to intact subjects is complicated. Further-more, inversion effects could only be shown for about half of the subjects. They found that the right hemisphere was more sensitive to inversion of macaque facial photographic stimuli than the left one, but also that macaques can rely on process-ing of local features regardless of hemisphere when needed, i.e. when recognition (as in reality mode processing) is not possible. “[...] the fact that the same stimuli made of exactly the same features are subjected to different modes of processing depending on their orientation suggests that it is not merely sufficient that the stimuli have both component and configural properties available to be processed, but rather that the perceiving organism finds itself unable to extract configural information from an inverted face and resorts to an analytical, component mode of processing” (Vermeire
& Hamilton, 1998, p. 1012). This flexibility is an important reason for why find-ings from pictorial experiments have to be analysed as pictorial problems before gen-eralisations to real life perception can be drawn. True recognition must be teased apart from apparent recognition because mode of picture processing will affect the processes underlying performance on the task and the conclusions that can be drawn from it all.
There are reasons to believe that the inversion effect is dependent on recognising the depicted stimuli as real objects.
Overman and Doty (1982) not only found that recognition of inverted faces were more difficult than upright images for pig-tailed macaques, but also that the subjects displayed social responses towards monkey and human faces. This response seemed to have been strongest at initial exposure. Naturalistic images, like scenery, flowers, birds and insects, did neither yield emotional responses nor inversion effects.
Using a MTS procedure to study memory Phelps & Roberts (1994) demon-strated inversion effects for primate facial black-and-white photographs in humans and squirrel monkeys (Saimiri sciureus), but not in pigeons. Humans but not squir-rel monkeys also showed inversion sensitivity for photographs of outdoor scenery.
Thus it is likely that the squirrel monkeys did not recognise scenery in photographs.
Parr et al. (1999) showed that the inversion effect is not face specific in rhesus macaques. The effect was not elicited by human faces and abstract shapes but by cars and faces of an unknown monkey species.64 So why did the inversion effect manifest itself for monkey faces and cars, but not for human faces and abstract shapes? Can it be that human faces were not recognised as faces in the photographs, and/or that attention to local features was selected for in the food reinforced MTS procedure used? According to Parr et al. (1999) discrimination of human faces was acquired quickly compared to other stimulus classes, which is supposed to be consistent with a configural, or global, processing strategy. However, fast acquisition could likewise occur for the typical human eye or another more local feature which is highly invari-ant between photographs of human faces. After all, the food reinforced training pro-cedure gave no incentive to actually recognise the content of photographs.
Chimpanzees in Parr et al. (1998) showed a more expected performance with the very same stimuli and method (MTS), with inversion effects occurring for familiar face categories, like chimpanzees and humans, but failing for capuchin monkeys, cars and abstract shapes.65 However, responses became notably less accurate across all categories during inversion trials, suggesting configural processing.
The real question is why cars did evoke an inversion effect for the macaques above.
Further studies of this phenomenon must take into account whether the subjects recognise the stimuli as objects or not, and whether they retain this recognition across the whole experiment. Perhaps it will be found that faces happen to be extra salient motifs in pictures (several studies to this effect has been presented in this chapter) and are sensitive to inversion for this reason, as are other recognisable ob-jects, although they might be novel, like cars. Studying the inversion effect with real objects ought to be a given control.
64 That known objects, or “expert categories”, are more sensitive to inversion has been shown for humans (Diamond & Carey, 1986) and chimpanzees (Parr et al., 1998; Parr & Heintz, 2006) but was not supported in Parr et al. (1999).
65 The study fails to tease apart the face specificity of inversion from the familiarity effect since no pictures of well known non-face objects were included.