Mental representation (Imagery)

Mental representation is one of the central paradigms in cognitive psychology. The ability for mental representation implies generalisation of a familiar image to its visual transformations. In terms of an experimental procedure, subjects prefer transformed (say, the mirrored image or the left-right transformation) of the rewarded stimulus or group of stimuli to a neutral stimulus when the learned stimuli are absent. Being confronted with these transformations, some animals respond to them as if they were the original learned stimuli, thus showing a capacity for flexible stimulus use. This kind of ambiguity is associated with cognitive processes such as mental rotation, the capacity of mentally changing the orientation of an image in order to reassess it from a new imaginary perspective. It is assumed that the underlying mechanisms allowing such generalisations have an important cognitive basis. For instance, in generalising to a rotated image, mental rotations of the learned image are assumed to occur in the brain. A combination of several techniques devised by psychologists for studying mental manipulations with images has been extended to develop an imagery paradigm for animals. Processes of imagery in animals could be based on what is called “detached representation”. A chimpanzee that walks away from a termite hill to break a twig in order to make a stick for fishing termites has a detached representation of a stick and its use.

Operationally, imagery refers to processing visual information not currently before the subject, that is, the processing of representation of visual stimuli. This capacity of organisms is closely related to extrapolation and stimulus generalisation; hence the resemblance in experimental procedures. An integral part of the method for testing imagery in animals is to induce transformation of the remembered stimulus.

Neiworth and Rilling (1987) have suggested an experimental procedure based on both elements mentioned above: stimulus rotation and extrapolated movement. The first element was derived from the mental rotation task that was initially developed for humans by Shepard and his colleagues (Shepard and Metzler, 1971; Shepard and Cooper 1982). In a series of their psychological experiments done in the 1970s, human subjects were asked to report if an object that has been rotated from its standard form was the same as the standard form or was a mirror image of the standard form. Subjects' response time in tasks involving mental manipulation and examination of presented figures was found to vary in proportion to the spatial properties (size, orientation, etc.) of the figures presented. Reaction times have been found to be a linear function of the angular rotation of the stimulus. In other words, reaction time indicated that humans took more time to respond when there was greater angular disparity between the standard and rotated forms. This isomorphism between reaction times and angular rotation suggested that subjects were rotating mental images to solve this problem.

Hollard and Delius (1982) were the first to conduct a mental rotation study simultaneously with humans and animals. Pigeons and humans chose which one of two alternative visual forms was identical to, or a mirror image of, a previously presented sample form. The two comparison forms were presented in various orientations with respect to the sample. The two species yielded similar accuracies, but although human reaction times depended linearly on the angular disparities, those of the pigeon did not. Humans appeared to apply a well-known, thought-like, mental rotation procedure to the problem, whereas pigeons seemed to rely on different recognition processes. Mirror-image forms may be better discriminated by the pigeon's visual system than by the human one. In general, it was suggested that pigeons may even be superior to humans in their capacity to mentally rotate visual information, as they showed no increase in reaction time when judging the identity of stimuli presented at increasingly different angular orientations, whereas humans did (Cook, 2001).

Neiworth and Rilling (1987) elaborated tasks for pigeons and humans based on stimulus rotation on a video monitor. The important element of their procedure was based on extrapolation tests when subjects were required to extrapolate the location of a moving target that had disappeared. This approach was applied to testing the hypothesis that perceptual information is maintained by the subject and that mental transformations occur while the moving stimulus is absent. The experimental procedure thus includes two ingredients such as a perceptual trial, in which the stimulus and its movement are fully presented, and an “imagery” trial, in which a portion of the stimulus transformation is invisible. By comparing the behaviour of subjects on perceptual and imagery trials, it is possible to test whether or not subjects can cognitively change a representation in the same way they have learned to discriminate the perceptual change.

The task for pigeons involved observation of a clock hand stimulus that rotated from an initial location of 0 ⁰ (12.00 position). On perceptual trials, the clock hand moved at constant speed and was always visible to the subject. Imagery trials were identical except the clock hand disappeared at the 90⁰ position (3.00) for a specific delay and then reappeared at an appropriate location as if it had rotated with constant speed during the delay. On violation trials, the clock hand also disappeared at 90⁰ but reappeared after a delay at a position inconsistent with constant velocity whether or not the hand was always visible. After discrimination training involving two locations (135⁰ and 180⁰) pigeons showed immediate and positive transfer to a novel intermediate location (158 ⁰) and to a novel location outside the boundaries trained (202⁰).

The experimenters found an essential parallel between imagery in 3pigeons and humans basing on evidence of an interaction between the perceptual illusion of momentum and visual memory. The fact is that humans do not accurately represent the final location of the moving object; instead, visual memory of the final location shifts forward by some small amount, a distortion similar to the real momentum that occurs while moving objects stop. This phenomenon, called representational momentum (Freyd, 1983), is assumed to be fundamental and automatic, and it turned out that pigeons made the same kinds of errors as humans when representing movement. In sum, the results of experiments supported structural, functional, and interactive descriptions of an imagery process in pigeons.

Wasserman et al. (1996) examined pigeon’s capacity to recognise depth-rotated stimuli. They used a discrimination learning paradigm to explore pigeon’s recognition on line-drawings of four objects (an aeroplane, a chair, a desk lamp, and a flashlight) that were rotated in the depth dimension. The pigeons readily generalised discriminative responding to pictorial stimuli over all untrained depth rotation.

Monkeys also demonstrated a clear evidence of stimulus generalisation with drawings of depth-rotated objects (Logothetis et al., 1994). After training monkeys for 8 months on successive same-different discriminations, the researchers tested their subjects with drawings of either the target shape at its given orientation or the target shape at different depth-rotation orientation. “Same” responses were found to be highest when the test drawing depicted the target shape at its given orientation, these reports systematically falling as the test drawings depicted greater rotation away from this orientation. It is interesting to note that stronger stimulus generalisation was obtained with rendered drawings of realistic objects than with computer-generalised wire-frame and blob-like shapes and with exposure to multiple views of the target shape than with single views of it.

Cook and Katz (1999) have found some similarities between humans and pigeons in mental representation of three-dimensional objects. In their experiment pigeons were taught to discriminate between computer- generated three-dimensional projections of cubes and pyramids. These object stimuli were then presented on each trial either dynamically rotating around one or more of their axes or in a static position at a randomised viewing angle. Pigeons were rewarded for pecking in the presence of only one of the objects. Tests with different rotational transformations of the stimuli suggested that the pigeons may have been using a three-dimensional perception of these objects as the basis for their discrimination. The pigeon’s performance was consistently better with the dynamic presentations than with randomly oriented static views. Further, their performance was relatively unaffected by transformations in object size; the rate, direction and combination of motions; and changes in surface colour of the stimuli. Three of the four pigeons showed some evidence of recovering the structure or shape of these objects from just the pattern of their motion on the display. When all contour and surface information was removed in test conditions leaving only the rigid projective geometry of the moving objects to guide performance, these birds were again better at discriminating the dynamic stimuli. Cook (2001) suggested that pigeons may have some higher level similarities with humans, as the experiments showed some capacity for recognising objects across different transformations and deriving structural information from the pattern of an object’s motion.

There are some clear experimental evidences that complex visual information processing can be implemented on even smaller but also highly mobile animals like bees, although it is still discussible whether the obtained results can be explained with the use of mental representation paradigm. The ability to generalise a familiar image to visual transformations like a mirror image or a left-right transformation allows recognition of familiar images from a different viewpoints for these small creatures. The fact is, bees learn and memorise not only the location of the food sources but also specific floral features that allow them to discriminate between rewarded and non-rewarded flowers, and visual cues are of fundamental importance in this context.

A question whether bees are able to transfer acquired information about a previously rewarded pattern to its mirror image and its left-right transformation was first raised by Gould (1988) on honey bees. Later this problem was studied on bumblebees (Plowright 1997). In Gould’s study each stimulus consisted of four circles of the same diameters and different colours. The transformations were obtained by changing the positions of the circles. The same technique was used in the experiments with bumblebees. Nevertheless in these experiments the possibility of the use of preference of colours by the insects was not excluded.

Guirfa and co-authors (Giurfa et al., 1999; Giurfa, Lehrer, 2001) in their experiments with honey bees used achromatic (black and white) stimuli whose transformations resulted in different alternatives for mirror-image and left-right permutation. Individually marked honey bees were trained to collect syrup on vertically presented stimuli lying flat on the back walls of a wooden Y-maze that was placed close to a large open window in the laboratory, through which bees could enter the maze. Bees were trained either with a single pair of patterns or with six different pairs of patterns presented in a random succession. Within each pair one pattern was rewarded and the other was not. All patterns had four quadrants, each displaying a different stripe orientation. In multiple-pattern training the six rewarded patterns shared a common configuration different from that of the six non-rewarded ones. After both kinds of training, the bees preferred the mirror image and the left-right transformation of the rewarded pattern (or rewarded configuration) to a novel pattern. They also preferred the left-right transformation to the mirror image. The researchers explain this performance by: (1) matching with a retinotopic template of the trained patterns after training with a single pair of patterns; and (2) matching with a generalised pattern configuration after training with randomised sets of patterns. Although both strategies are based on comparison of an image currently perceived with one that has to be accessed from memory, they constitute different options as the former is less flexible while the latter allows categorisation of novel patterns.

In experiments of Plowright and co-authors (2001), bumble bees were trained to discriminate between a reinforcing pattern (S+) and a non-reinforcing one (S-) which differed only in the configuration of four artificial petals. They were subsequently tested for recognition of the S+ rotated by 90° (S + 90). Experiment 1 used petals of four colours, and the other experiments used four symbols. The symbols either remained unchanged when the whole pattern was rotated (e.g., “+” in Experiment 2) or changed appearance (e.g., “< ” in Experiment 3). The bumble bees failed to recognize the S + 90 in the first two experiments, but in Experiment 3, the choice proportion for S + 90 in the presence of a new pattern was significantly higher than chance. The researchers concluded that bumble bees can recognize a rotated pattern, possibly by using mental rotation, provided that a cue as to the extent of the pattern transformation is given.

The explanation of the results obtained on bees possibly does not invoke mental rotation processes. As Giurfa (2003) gives this, the processes involved in left-right transformation and mirror image generalisation seem to be based on “lower” processes such as integration of pattern features into a simplified configuration and comparison of the resulting representation with the currently perceived image. In any case, bees can generalise the excitatory strength of a trained pattern to the mirror-image and the left-right transformations. These insects are capable of the flexible use of the resulting representation to categorise and thus respond appropriately to novel patterns.

16. CONCEPTUAL BEHAVIOUR BASED ON RELATIONS

Here we will consider experimental studies aimed to a problem of whether animals posses a type of reasoning which allows them to combine knowledge about specific relationships in order to infer another relationship. As we have seen in the previous issue, some animals possibly can operate with perceptual concepts basing on categories such as “people”, or “trees” in which application of the concept leads to the inclusion that novel exemplars belong to a given category. Relation concepts are based on common abstract relationships shared by sets of stimuli. For example, a generalised identity concept is shown by a subject that discriminates among novel sets of stimuli based on the perceptual “ sameness ” of identical stimuli, whereas a generalised “ oddity ” concept is shown by a subject that discriminates among novel sets of stimuli based on their perceptual difference. Finally, associative concepts are those in which stimuli are categorised on the basis of common associations with other stimuli, responses, or outcomes. For example, disparate arbitrary stimuli, such as A and C, may become related to one another through their common relationships with a given mediated stimulus, B. Although the stimuli in such associative categories are not perceptually similar, they are treated in a similar fashion, as demonstrated by a transfer of function among the stimuli in a given category.