Abstract: We present the results of two experiments that probe the ability of rhesus macaques to match visual arrays based on number. Three monkeys were first trained on a delayed match-to-sample paradigm (DMTS) to match stimuli on the basis of number and ignore continuous dimensions such as element size, cumulative surface area, and density. Monkeys were then tested in a numerical bisection experiment that required them to indicate whether a sample numerosity was closer to a small or large anchor value. Results indicated that, for two sets of anchor values with the same ratio, the probability of choosing the larger anchor value systematically increased with the sample number and the psychometric functions superimposed. A second experiment employed a numerical DMTS task in which the choice values contained an exact numerical match to the sample and a distracter that varied in number. Both accuracy and reaction time were modulated by the ratio between the correct numerical match and the distracter, as predicted by Weber's Law.

Abstract: Patterned string tasks are a test of perceptual capacity and the understanding of means-end connections. Primates can solve complex forms of this task in laboratories. However, this may not indicate the level of such cognition that is commonly employed in the wild, where decision-making time is often short and distractions such as predator avoidance and competition between conspecifics are often prevalent. The current study tests whether wild common marmosets (Callithrix jacchus) can successfully complete the simplest form of the patterned string task, parallel strings, while in their natural environment. Although 12 out of 13 marmosets could successfully complete the task, in previous laboratory-based studies on primates, the errors at this task by all primate species tested were consistently lower than in the present study. This is probably explained by the added difficulties imposed by the natural setting of the task in the present study, exemplified by a significant increase in observed vigilance behaviour by subject animals prior to attempts at the task that were unsuccessful. The undertaking of such tasks by common marmosets in situ probably provides a more reasonable representation of the levels of cognitive capacity expressed by this species in the wild than do laboratory-based studies of the task.

Abstract: The sensitivity of eleven pet dogs and eleven 2.5-year-old children to others' past perceptual access was tested for object-specificity in a playful, nonverbal task in which a human Helper's knowledge state regarding the whereabouts of a hidden toy and a stick (a tool necessary for getting the out-of-reach toy) was systematically manipulated. In the four experimental conditions the Helper either participated or was absent during hiding of the toy and the stick and therefore she knew the place(s) of (1) both the toy and the stick, (2) only the toy, (3) only the stick or (4) neither of them. The subjects observed the hiding processes, but they could not reach the objects, so they had to involve the Helper to retrieve the toy. The dogs were more inclined to signal the place of the toy in each condition and indicated the location of the stick only sporadically. However the children signalled both the location of the toy and that of the stick in those situations when the Helper had similar knowledge regarding the whereabouts of them (i.e. knew or ignored both of them), and in those conditions in which the Helper was ignorant of the whereabouts of only one object the children indicated the place of this object more often than that of the known one. At the same time however, both dogs and children signalled the place of the toy more frequently if the Helper had been absent during toy-hiding compared to those conditions when she had participated in the hiding. Although this behaviour appears to correspond with the Helper's knowledge state, even the subtle distinction made by the children can be interpreted without a casual understanding of knowledge-formation in others.

Abstract: This study explored whether the tendency of chimpanzees and children to use emulation or imitation to solve a tool-using task was a response to the availability of causal information. Young wild-born chimpanzees from an African sanctuary and 3- to 4-year-old children observed a human demonstrator use a tool to retrieve a reward from a puzzle-box. The demonstration involved both causally relevant and irrelevant actions, and the box was presented in each of two conditions: opaque and clear. In the opaque condition, causal information about the effect of the tool inside the box was not available, and hence it was impossible to differentiate between the relevant and irrelevant parts of the demonstration. However, in the clear condition causal information was available, and subjects could potentially determine which actions were necessary. When chimpanzees were presented with the opaque box, they reproduced both the relevant and irrelevant actions, thus imitating the overall structure of the task. When the box was presented in the clear condition they instead ignored the irrelevant actions in favour of a more efficient, emulative technique. These results suggest that emulation is the favoured strategy of chimpanzees when sufficient causal information is available. However, if such information is not available, chimpanzees are prone to employ a more comprehensive copy of an observed action. In contrast to the chimpanzees, children employed imitation to solve the task in both conditions, at the expense of efficiency. We suggest that the difference in performance of chimpanzees and children may be due to a greater susceptibility of children to cultural conventions, perhaps combined with a differential focus on the results, actions and goals of the demonstrator.

Abstract: Most studies of animal tool use require subjects to use one object to gain access to a food reward. In many real world situations, however, animals perform more than one action in sequence to achieve their goals. Of theoretical interest is whether animals have the cognitive capacity to recognize the relationship between consecutive action sequences in which there may be one overall goal and several subgoals. Here we ask if cotton-top tamarins, a species that in captivity uses tools to solve means-end problems, can go one step further and use a sequence of tools (means) to obtain food (end). We first trained subjects to use a pulling tool to obtain a food reward. After this initial training, subjects were presented with problems in which one tool had to be used in combination with a second in order to obtain food. Subjects showed great difficulty when two tools were required to obtain the food reward. Although subjects attended to the connection between the tool and food reward, they ignored the physical connection between the two tools. After training on a two-tool problem, we presented subjects with a series of transfer tests to explore if they would generalize to new types of connections between the tools. Subjects readily transferred to new connections. Our results therefore provide the first evidence to date that tamarins can learn to solve problems involving two tools, but that they do so only with sufficient training.