Spontaneous categorization of tools based on observation in children and chimpanzees

Abstract

The acquisition of the concept of 'tool' remains intriguing from both developmental and comparative perspectives. Our current model of tool use development in children is based on humans' supposedly unique ability to adopt a teleological stance: the understanding of a demonstrator's goal-based intentions when using a tool. It is however unclear how children and chimpanzees, our closest relatives, combine their knowledge of different objects whose function is to act on other parts of the environment, and assign them to a single category of 'tools'. Here, we used a function-based approach to address this question. We exposed 7 to 11-year-old children and adult chimpanzees to a Matching-to-Function (MTF) task to explore whether they would sort tools and non-tools separately after demonstration of their function by an experimenter. MTF is a variant of Matching-to-Sample where the sample and the target are from the same category/kind rather than identical. Around 40% of children paired objects according to their function in the MTF task, with only one child younger than 8 years doing so. Moreover, when verbally questioned, these children offered a function-based answer to explain their choices. One of six chimpanzees also successfully paired objects according to function. Children and at least one chimpanzee can thus spontaneously sort tools into functional categories based on observing a demonstrator. The success of a single chimpanzee in our task suggests that teleological reasoning might already have been present in our last common ancestor but also shows that human children more readily conceptualize tools in a spontaneous fashion.

Figure 1

The box and objects used in the experiment with children and chimpanzees. Left: non-tools, the wedge (W) and the stick (S). Right: tools, the hammer (H) and the pizza slicer (P). Center: box closed (left) and opened through two mechanisms (top: window down; bottom: hinge open).

Figure 2

Chimpanzee subject (Ai) and a child participant in the identity Matching-to-Sample phase of the experiment. Samples are presented in compartment 3 of the apparatus; the two alternatives (one of which is identical to the sample) are located in compartment 2. Subjects can be seen making their choice between the alternatives presented, through pointing.

Figure 3

Pairing results in probe trials for the three chimpanzee subjects that progressed to the MTF phase (A,B): Ai, (C,D): Chloe, (E,F): Pendesa). Panels B, D, F show the distribution of results obtained after subtracting the random distribution for 128 or 120 trials (depending on subject) from the observed results obtained by each chimpanzee. G and H show expected distributions after subtraction for fully function based (G) and fully random (H) pairing of the objects (128 trials, W = Wedge, H = Hammer, P = Pizza cutter, S = Stick) respectively. In particular, for panel H, because the forced-choice always involved at least the other tool or non-tool (i.e. W with S; and H with P), paired with one of two distractors, there are twice as many chances to hit the correct pairing by chance, hence double the distribution for the pairs W-S and H-P.

Figure 4

Pairing results in MTF probe trials in the four age classes of children tested (7–8 y: (A,B); 8–9 y: (C,D), 9–10 y: (E,F), 10–11 y: (G,H); W = Wedge, H = Hammer, P = Pizza cutter, S = Stick). This figure shows cumulative data for each age group, i.e., all trials across all subjects in the given group pooled together. Panels B, D, F and G show the distribution of results obtained after subtracting the random distribution for the respective number of trials corresponding to the number of tested children per age class (see legend of Fig. 3 for explanation).