Abstract: Categorization refers to the classification of perceptual input into defined functional groups. We present and discuss evidence suggesting that stimulus categorization can also be found in an invertebrate, the honeybee Apis mellifera, thus underlining the generality across species of this cognitive process. Honeybees show positive transfer of appropriate responding from a trained to a novel set of visual stimuli. Such a transfer was demonstrated for specific isolated features such as symmetry or orientation, but also for assemblies (layouts) of features. Although transfer from training to novel stimuli can be achieved by stimulus generalization of the training stimuli, most of these transfer tests involved clearly distinguishable stimuli for which generalization would be reduced. Though in most cases specific experimental controls such as stimulus balance and discriminability are still required, it seems appropriate to characterize the performance of honeybees as reflecting categorization. Further experiments should address the issue of which categorization theory accounts better for the visual performances of honeybees.

Abstract: Recent water maze experiments suggest that rats performing place navigation primarily use the geometric information provided by a set of landmarks, and neglect the featural information provided by the identities of the landmarks. Here, I develop a model that explains how an animal may perform place navigation by relying only on geometric information. The core of the model is the representation of places as panoramas defined by circular bar-codes embodying the relative bearings and apparent sizes of the landmarks, irrespective of their identities. There are two stages in the model. During the first stage, the animal freely explores its environment in order to acquire spatial information at the local level. During the second stage, the animal uses the information previously memorized to perform place navigation towards the goal it intends to reach. The possible role of two brain areas in place navigation is discussed within this framework. Beyond their primary role in landmark-based representations of places, hippocampal place cells may be involved in computing the current distances to the landmarks. Beyond their primary role in landmark-based representations of headings, post-subicular head-direction cells may be involved in computing the “compass bearings” of the landmarks.

Abstract: Nonhuman animals demonstrate a number of impressive quantitative skills such as counting sets of items, comparing sets on the basis of the number of items or amount of material, and even responding to simple arithmetic manipulations. In this experiment, capuchin monkeys were presented with a computerized task designed to assess conservation of discrete quantity. Monkeys first were trained to select from two horizontal arrays of stimuli the one with the larger number of items. On some trials, after a correct selection there was no feedback but instead an additional manipulation of one of those arrays. In some cases, this manipulation involved moving items closer together or farther apart to change the physical arrangement of the array but not the quantity of items in the array. In other cases, additional items were added to the initially smaller array so that it became quantitatively larger. Monkeys then made a second selection from the two arrays of items. Previous research had shown that rhesus monkeys (Macaca mulatta) succeeded with this task. However, there was no condition in that study in which items were added to the smaller array without increasing its quantity to a point where it became the new larger array. This new condition was added in the present experiment. Capuchin monkeys were sensitive to all of these manipulations, changing their selections when the manipulations changed which array contained the larger number of items but not when the manipulations changed the physical arrangement of items or increased the quantity in one array without also reversing which of the two arrays had more items. Therefore, capuchin monkeys responded on the basis of the quantity of items, and they were not distracted by non-quantitative manipulations of the arrays. The data indicate that capuchins are sensitive to simply arithmetic manipulations that involve addition of items to arrays and also that they can conserve quantity.

Abstract: As previously reported (Beran and Rumbaugh, 2001), two chimpanzees used a joystick to collect dots, one-at-a-time, on a computer monitor, and then ended a trial when the number of dots collected was equal to the Arabic numeral presented for the trial. Here, the chimpanzees were presented with the task again after an interval of 6 months and then again after an additional interval of 3.25 years. During each interval, the chimpanzees were not presented with the task, and this allowed an assessment of the extent to which both animals retained the values of each Arabic numeral. Despite lower performance at each retention interval compared to the original study, both chimpanzees performed above chance levels in collecting a quantity of dots equal to the target numeral, one chimpanzee for the numerals 1-7, and the second chimpanzee for the numerals 1-6. For the 3.25-year retention, errors were more dispersed around each target numeral than in the original study, but the chimpanzees' performances again appeared to be based on a continuous representation of magnitude rather than a discrete representation of number. These data provide an experimental demonstration of long-term retention of the differential values of Arabic numerals by chimpanzees.

Abstract: A previous study demonstrated that horses can learn socially from observing humans, but could not draw any conclusions about the social learning mechanisms. Here we develop this by showing horses four different human action sequences as demonstrations of how to press a button to open a feed box. We tested 68 horses aged between 3 and 12 years. 63 horses passed the habituation phase and were assigned either to the group Hand Demo (N = 13) for which a kneeling person used a hand to press the button, Head Demo (N = 13) for which a kneeling person used the head, Mixed Demo (N = 12) for which a squatting person used both head and hand, Foot Demo (N = 12) in which a standing person used a foot, or No Demo (N = 13) in which horses did not receive a demonstration. 44 horses reached the learning criterion of opening the feeder twenty times consecutively, 40 of these were 75% of the Demo group horses and four horses were 31% of the No Demo group horses. Horses not reaching the learning criterion approached the human experimenters more often than those who did. Significantly more horses used their head to press the button no matter which demonstration they received. However, in the Foot Demo group four horses consistently preferred to use a hoof and two switched between hoof and head use. After the Mixed Demo the horses' actions were more diverse. The results indicate that only a few horses copy behaviours when learning socially from humans. A few may learn through observational conditioning, as some appeared to adapt to demonstrated actions in the course of reaching the learning criterion. Most horses learn socially through enhancement, using humans to learn where, and which aspect of a mechanism has to be manipulated, and by applying individual trial and error learning to reach their goal.