Surface-mode in monkeys

Keating and Keating (1993) for example found that subjects that learned to sort faces of various primates, including humans, readily generalised to so called Identi-kit faces, which are composite sketches used to depict people from eyewitness ac-counts. Thus, the testing materials used were black-and-white shaded pencil draw-ings of human faces. The subjects were then trained to recognise one standard face.

Eye-tracking revealed that the eyes, just as for photographic stimuli, attracted the most attention. However, removal of the eyes (brows, nose or lip) did not lower

rec-ogni bling tial r targe perfo revea caqu the t woul study to re woul sis.

Dittr long inste be co stimu to re perce pand In a indiv did n been

Figur (1994

The tions

60 Eye perim

tion, but m g of the fac relations be et face sket ormed as w aled that by ue could ide task as one

ld mean th y which im ecognise d ld speak ag

rich (1994) -tailed mac ead of phot ontrolled b uli has bee ecognise car

eption of m ded upon h

preferentia viduals in p not preferen n familiarise

re 10. Drawn 4; 1990).

macaques s of monke

e-tracking see mental tasks.

modification ce disrupted

etween feat tch. One fi well with a s

y isolating entify the t e of finding hat it might mplies that i rawn eyes, gainst a real

) used black caques (Ma tographs w between im en used in s

ricature fac movement here, and th

al looking photograph

ntially targ ed with a ph

n stimuli used

tested in D ey bodies. O

ems to be a m

n of the ey d recognitio tures as wel

inding is en scrambled a

an eye of target face.

g a particul t also had so

in the trans and neve lity mode b

k-and-whit acaca fascicu

as motivate ages. In ad studies of m ces. That it

to the per he cited inf paradigm hs if they h

et a novel i hotograph.

d in Dittrich

Dittrich (1 One of the

most promisin

es or brows on. It thus ll as the sp nlightening as with the the picture .⁶⁰ It is reas lar eye rath orted faces sfer from p er whole fa based perfor

te drawings ularis). Ditt ed by the e ddition Ditt movement t might be rception of fant study one month had been fam

individual i

(1994) (left),

994) had t species was

ng method fo

s did. Rota seems that ecific look g, however.

e standard f e, regardless

sonable tha her than a in this ma photograph aces. Idiosy rmance and

s to test rec trich’s reaso ease by whi trich maint perception hasty to ge f still imag (Tyrrell et h old infan miliarised w in the form

, and Dittrich

to choose b s chosen to

r pinpointing

ation, inver t the subjec

of the eyes . One of th face. The ey s of where at the subje face the w anner durin s to mug-sh yncratic re d for a featu

ognition of on for usin

ich individ tains that h n, and that eneralise fin ges do not al., 1987) nts oriented with a caric m of a carica

h (1990) (rig

between fo be the con

g individual s

sion and sc cts used the s to identif he four sub ye-tracking

it was, the ect has rega whole time.

ng the rest o hots it only esponse pat

ure mode a

f body shap ng drawn st

dual feature highly schem

infants are ndings from

need to b is inconclu d towards n cature, but ature if they

ght). From D

our slide pr nstant targe

strategies in e

cram-e spa-fy the

bjects g data e

ma-arded This of the y had tterns

analy-pes in imuli es can matic e able m the be

ex-usive.

novel t they y had

ittrich

rojec-et and

ex-the same four monkey bodies were used during ex-the whole study, but with manipula-tions (fig. 10, left). During a training phase where pictures were held constant all subjects learned to discriminate the correct body picture in about 70 trials, much faster than for discrimination of geometrical or “non-natural” stimuli. In a transfer test reduction of picture size induced more errors than did rotation of the stimuli, but performance were still at 80% correct. When different features were removed from the bodies correct discrimination deteriorated differently for different features but remained above chance for most subjects and most features.

Taken together it was found that the torso can be judged to have been the most informative area. This strongly suggests that the subjects did not view the stimuli as pictures of monkeys. In a generalisation test to a frontal view of only the faces (the pictures had previously been presented in profile) three of four subjects responded above chance. This might, according to Dittrich, suggest that face recognition was involved, but the necessary controls of rotating other body parts or replicate the test with nonsense stimuli were not done.

The experiment was rerun, but this time with another of the four species as target stimuli in all manipulations. The findings were reproduced, but with one important difference. This time the torso was not the most critical feature, but the extremi-ties.⁶¹ Dittrich concludes that this difference was due to differences in the task, maintaining that in the second experiment the pictures were known and therefore had been comparable to discrimination of individuals, while in the first experiment the task had been to separate species. The former is done with a local strategy, and the latter with a global one, i.e. looking at body shape, according to Dittrich (1994).

I would rather attribute the differences to the possibility that the subjects did not view these drawings as neither monkeys nor representations of monkeys, and simply found different parts of different pictures useful for discriminating them. That is, performance in a surface mode.

Dittrich (1990) also used line-drawings, but this time of faces (fig. 10, right). The task was for long tailed macaques to discriminate between facial emotional expres-sions. Again there was no detrimental effect by rotating the stimuli, or by changing colour, brightness or size of the stimuli. Again local features proved important, no-tably the facial outline followed by the eyes and mouth, but also configurational processing was involved since scrambled faces were less successfully discriminated.

However, claims that recognition of emotional expression took place are not sup-ported. No transfer trials to novel exemplars of the expressions were given, other than distortions of the four original stimuli pictures. Also, the local features that proved crucial for discrimination did not contribute to emotional information in the drawings. The absence of an inversion also speaks against recognition of faces (see section 6.3.4).

D’Amato and van Sant (1988) remains the pinnacle example of performance on a level that works very well in surface mode. On a task that the experimenters in-tended to work on a completely different level, capuchin monkeys (Cebus apella) used the frequent occurrence of reddish patches in human clothes etc. to, quite

61 Amusingly, the target stimuli was the patas monkey (Erythrocebus patas), the world’s fastest run-ning monkey

cessfully, form a “person versus non-person” categorisation. Red flowers and water-melons were thus included as “people.” Naturally, this categorisation broke down when colour was controlled for. This effect can occur in all forms of pictorial test-ing, depending on which discriminative cues the subjects attend to and are subse-quently reinforced for. Even when seeing that a picture depicts a person, one can judge this information to be less relevant than the fact that it wears a hat in a specific colour. But the reason the monkeys settled to attended to colour strongly suggests that the photographs were not seen as depictions of real entities. This is a more par-simonious explanation than that the life-world of monkeys is one predominantly of colour, rather than a world of objects.

Recognising species, or species typical features, in photographs seem not to be lim-ited to macaques but can be extended to New World monkeys. In Neiworth et al.

(2004) cotton top tamarins (Saguinus oedipus) demonstrated apparent categorical recognition of individual species, as well as primates in relation to other animals.

Complex colour photographs displayed on a television screen served as stimuli. The test method was the “preference for novelty” paradigm commonly used with human infants. After familiarisation with a category (specific species or primates in general) exemplars from another category (another species or nonprimate animals) were pre-sented together with novel pictures from the old category. If the subject preferred to look at exemplars from the novel category rather than at novel pictures, familiarisa-tion on a categorical level had taken place.

Complex pictures (i.e. with natural backgrounds) and a diverse mix of exemplars, as to counteract response to invariant local properties, were used. Therefore surface mode processing unlikely accounts for the results. However, the first alarm bell should go off when considering that it is assumed that these laboratory raised sub-jects recognised photographic animals that they had never experienced before. And how diverse were the visual patterns if considering combinations of local properties?

When experiencing a novel pattern of colours and shapes, it is reasonable that inter-est is evoked. The only potentially convincing indication that recognition took place, was that when familiarised with four species of monkeys, the subjects pre-ferred to look at non primate mammals rather than at a novel monkey species or photographs of apes. But species and their visual patterns can be expected to covary.

When mammal pictures were contrasted to reptile photographs, on the other hand, looking preferences did not deviate from chance. This was interpreted as an avoidance to look at reptiles, or possibly a continued interest in novel exemplars of mammals. But it could also be explained by the fact that the visual patterns as such are more varied when using pictures from such a broad category as mammals, and that they therefore never really got familiarised. A necessary control for this type of study is to include pictures of nonsense shapes and record the relative interest for nondepicting stimuli.