Skip to main content

Tool Use and Causal Cognition

A three-day workshop

4-6 July 2008
University of Warwick


Friday, 4th July
10:30 Coffee
11:00 Elisabetta Visalberghi: The use of tools in wild and captive capuchin monkeys
12:30 Lunch
13:30 Laurie Santos: What monkeys know about causal powers: Evidence from categorization and play
15:00 Amy Needham: Learning about tools in infancy
16:30 Coffee
17:00 Deborah Kelemen: Intention, convention and design: The development of children's artifact concepts
Saturday, 5th July
9:00 Alex Kacelnik: Folk physics for crows (with apologies to DP)
10:30 Coffee
11:00 James Woodward: What can tool use tell us about causal cognition?
12:30 Lunch
13:30 Jeffrey Lockman: From object manipulation to tool use
15:00 John Campbell: Tool Use and Interventionist Approaches to Causation
16:30 Coffee
17:00 Alessandro Farne: Tool use affects space and body representations
Sunday, 6th July
9:00 Georg Goldenberg: Effects of local brain damage on tool use and technical problem solving
10:30 Coffee
11:00 Charles Spence & Nicholas Holmes: Assessing the multisensory consequences of tool use in humans
12:30 Concluding Discussion (chaired by Christopher Peacocke)


Speakers and Abstracts

John Campbell (University of California at Berkeley): ‘Tool Use and Interventionist Approaches to Causation’
It is difficult not to be struck by the apparent ingenuity of a capuchin monkey choosing the right size and shape of stone to crack a nut, or a crow bending a piece of wire so it can fetch a bucket of food from a well. Do we have to suppose that the animal has an understanding of the causal basis of successful tool use? One natural reaction is: (a) the animal has a general understanding of the mechanics of its environment, and (b) this understanding is put to work in the animal's construction, or selection, and use of a tool. The trouble with this diagnosis is that animals typically display significant limitations in their understanding of mechanics in general, and in the principles of tool use in particular. You might then suppose that the animal's tool use has no foundation in causal cognition. But now we have no way of explaining how it has the skills it does. The main point I want to make in this paper is that causal cognition is not monolithic. In particular, we can distinguish between: (1) a grasp of what would happen to an outcome variable under various interventions on an input variable, and (2) an understanding of causation as involving mechanistic interactions, such as the transmission of motion by impulse. Animals using tools seem to exhibit some limited grasp of intervention counterfactuals as described in (1); what seems to be missing is any mechanistic understanding of the situation, of the type described in (2). There are further contrasts between animal causal cognition and human causal cognition. Human causal inquiry, and in particular the construction and use of tools to establish causal connections, seems often to be driven by mere curiosity, rather than an interest in any immediate payoff. In contrast, the documented cases of animal tool use are all driven by immediate practical needs. This is not because animals are in general incapable of inquiry without immediate practical payoff. There are many cases in the cognitive mapping literature of animals exploring new environments simply to learn the spatial layout. This doubtless has long-term benefits, but does not seem to depend on there being any immediate payoff. In contrast, there seem to be no documented cases of animals constructing 'causal maps' of their environments for long-term use without immediate practical benefits. Tool use in animals is not driven by curiosity about causal connections.

Alessandro Farne (INSERM): ‘Tool use affects space and body representations’
Tools enable human beings and other animals to manipulate objects that would otherwise not be reachable by hands. Acting on objects by means of a physical tool requires sensory information that is mainly provided by vision and touch in the peripersonal space. Updating the weighting of inputs from these sensory modalities when using tools would render the possibility of reaching and manipulating far objects as if they were closer to the hand. The behavioural effects of tool-use on the multisensory representations of space, largely documented in the normal and pathological brain, will be critically reviewed. However, emphasis will be put one specific question: Is the body space affected by tool-use? Updating the body representation (body schema-BS) would be essential for efficient motor control during development and skilful tool-use in the adult life. The almost one-century-old hypothesis that tool-use induces plastic changes resulting in the tool being incorporated in the BS is widely accepted, and intuitive enough to become a popular notion. However, only indirect support to this hypothesis comes from previously documented effects of tool-use on multisensory coding of space. New evidence will be provided and discussed for the incorporation of a tool in the BS, by showing that free-hand movements are kinematically different after training in tool-use. These findings speak in favour of genuine tool-embodiment in the body schema.

Georg Goldenberg (Neuropsychological Department, Krankenhaus Muenchen-Bogenhausen): ‘Effects of local brain damage on tool use and technical problem solving’
There is more than one source of knowledge about use of tools and objects: Retrieval of "instruction for use" from semantic memory can support prototypical use of familiar objects. Direct inference of possible functions from structural properties enables non-prototypical use of familiar objects and use of novel objects. Multi-step actions with multiple objects put additional demands on working memory, planning ahead, and the skillfull application of trial and error. Results from neuropsychological studies probing each of these components demonstrate that retrieval of functional knowledge from semantic memory as well as direct inference of function from structure are bound to left hemisphere integrity, albeit to different structures within the left hemisphere. By contrast, multi-step actions and solution of multi-step mechanical puzzles are equally impaired in patients with right or left brain damage without predilection for distinct locations within the hemispheres.

Alex Kacelnik (Oxford University): ‘Folk physics for crows (with apologies to DP)’
Inferring from object manipulations what subjects understand about the physics underlying object interactions is very difficult, whether we examine humans, other apes, other primates, or more distant species. I will focus on the tool oriented behaviour of New Caledonian Crow, a species whose tool competence includes selectivity, manufacturing, sequential tool use and flexibility. I will characterize present knowledge of these competences and examine what –if anything- does this knowledge tell us about the birds’ level of causal understanding in object manipulation. The overall conclusion is not a binary distinction between total lack of understanding and full blown causal reasoning-controlled behaviour. What emerges instead is a patchwork of partial and sometimes conflictive evidence based both on the animals’ achievements and on their mistakes, leading slowly but progressively to a detailed picture of how tool oriented competences are inherited, develop ontogenetically, impact on the ecology of the species, have evolved and –crucially for this meeting- are controlled in the experienced adult animal.

Deborah Kelemen (Boston University): ‘Intention, convention and design: The development of children's artifact concepts’
Previous research suggests that while adults rationalize an artifact's existence, identity, properties, and function by reference to a designer's creative intent, a "design stance" does not constrain children's reasoning about artifacts until late preschool. In this talk, I will present research that explores young children's reasoning about artifacts prior to understanding design as well as their emerging sensitivity to various components of the design stance. I will also discuss the extent to which children's functional construal of artifacts might influence their understanding of natural objects to produce "promiscuous teleology"–the tendency to broadly ascribe purpose to all kinds of phenomena.

Jeffrey Lockman (Tulane University): ‘From object manipulation to tool use’
Tool use has long been considered an ability predicated on the capacity to engage in representational or symbolic thinking. In contrast, I suggest that early forms of tool use may be better conceptualized as instances in which human infants engage in coupling perception and action. In this connection, I consider the ontogenetic and phylogenetic origins of tool use by examining the ways in which exploration of objects and surfaces by infants can provide a platform from which tool use may emerge. I describe our research involving behavioral and kinematic methods to argue that the origins of human tool use can be found in the ways in which infants couple perception and action when combining objects with nearby substrates. This research indicates that well before a year of age, infants relate objects and surfaces appropriately, based on the physical composition of each. Equally important, infants also produce such appropriate combinations when using objects attached to handles. Implications of this research for understanding the phylogenetic origins of tool use will be considered as well.

Amy Needham (Duke University): ‘Learning About Tools in Infancy’
The notion that perceptual-motor activity is the basis for improvements in cognitive development is typically attributed to Jean Piaget.Although some of Piaget's claims regarding the specifics of development have been discredited, recent research has confirmed many of his broader ideas about process. In this talk I will discuss some of our recent findings regarding how infants learn about hand-held tools and what these findings mean for mechanisms of learning and development.

Laurie Santos (Yale University): ‘What monkeys know about causal powers: Evidence from categorization and play’
We present several studies aimed at exploring causal understanding in two species of closely related primate species. Our first question of interest concerned whether monkeys can distinguish between objects on the basis of whether the objects possess a novel causal power. Adapting a task developed for use with human infants, we presented capuchin monkeys (Cebus apella) with a box that lit up when certain objects were placed on top. We first allowed monkeys to place objects on the detector, learning which patterns of objects caused the detector to light up. We then presented monkeys with a forced choice between different kinds of objects. We observed that monkeys’ choice of objects depended on causal and not merely associate features of their experience. In a second set of studies, we explored whether a different monkey species, the rhesus macaque (Macaca mulatta), attends to causal information in a naturalistic exploratory task. Specifically, we examined whether monkeys play with novel objects in ways that are systematically geared towards acquiring causal knowledge. We presented monkeys with either confounded or unconfounded evidence about how a box works and subsequently allowed monkeys to approach and play with the familiar box and with a novel box that was not manipulated during the presentation. Similar to the patterns of exploratory play observed in human children, we found a significant interaction between the type of evidence monkeys received (confounded or unconfounded) and the duration of their exploratory play with the familiar and novel boxes; monkeys presented with confounded evidence showed a greater tendency to play with the familiar box than monkeys presented with unconfounded evidence. Our findings suggest that monkeys’ exploratory play is influenced by the quality of evidence they receive and raise the question of whether there is an evolutionarily ancient basis for the strategies that underlie children’s exploratory play, perhaps as part of a core causal knowledge. The proposed paper will discuss these findings in light of other results on causal cognition in primates and humans.

Charles Spence (Crossmodal Research Laboratory, Oxford University) & Nicholas Holmes (Department of Psychology, Hebrew University of Jerusalem, Israel): ‘Assessing the multisensory consequences of tool use in humans’
In this talk, we review the rapidly growing empirical literature on tool use in adult humans. We will highlight the key findings to have emerged from the recent psychophysical, and to a lesser extent neuroimaging, research in which normal participants manipulated a variety of tools in performing target-directed actions. We focus, in particular, on changes in the multisensory representation of peripersonal space that such tool use can induce, and ask the question of just how important is the particular kind of action that participants have to make with the tool to the particular pattern of results that are obtained. We will also contrast the multisensory consequences of tool use with those seen following other kinds of bodily ‘transformation’, such as, for example, induced by the rubber hand illusion.

Elisabetta Visalberghi (Istituto di Scienze e Tecnologie della Cognizione, Consiglio Nazionale delle Ricerche): ‘The use of tools in wild and captive capuchin monkeys’
The recent discoveries of tool use in wild bearded capuchin monkeys (Cebus libidinosus) in Brazil require a new way to look at this behavior. In Boa Vista (Piauí, Brazil), capuchin monkeys routinely crack open palm nuts with hammer and anvils, i.e. they exhibit a very complex type of tool use which requires the proper spatial combination of three objects (percussor, nut, and anvil). Capuchins also transport tools and nuts to the anvil and seem able to actively select functional hammer tools, in terms of material, size and weight. Selection of functional tools requires distinguishing between the relevant and the irrelevant features of objects: whereas the former are necessary for achieving success, the latter are not. Overall, it seems that the expertise reached by wild capuchins is by far superior to that reported for captive ones. It is possible that from their daily experience with nut cracking wild capuchins acquire better knowledge about tools than it is usually the case with captive subjects, that are presented with tool using tasks only for limited periods of time. Whether or not wild capuchins have a causal understanding of the effects of their actions is open to discussion and require experimental investigation aimed to rule out cognitively simpler explanations. Funded by EU FP6 NEST Programme, ANALOGY (No 029088).

James Woodward (California Institute of Technology): ‘What can tool use tell us about causal cognition?’
This talk will explore, from the point of view of a philosopher interested in causation, some issues having to do with the role of casual cognition in tool use. It will also ask what tool use can tell us about human and non-human causal cognition. Despite the impressive abilities displayed by some non-human primates and by some varieties of birds (e.g. corvids) in certain specific tool using tasks, it seems uncontroversial that these animals differ in dramatic ways from humans, including young children, in their tool using abilities. This is so despite the fact that some of the components or building blocks one would think underlie human causal learning abilities are present in many animals. For example, many animals learn readily from both classical and instrumental conditioning, and primates, at least, show, as far as evidence from looking time studies goes , many of the same “physical understanding” abilities regarding object behavior in connection with tasks involving causal perception in launching events, trajectory tracking, and appreciation of object cohesion as do human infants. Whatever the abilities involved in human causal cognition and successful tool use, they apparently go well beyond the abilities just described. So two obvious questions this raises are 1) how should we characterize or describe these differences and 2) can we say anything about what their explanation might be? This talk will discuss some possible answers to both of these questions. These include Povinelli’s suggestion that certain properties that are crucial to human causal understanding and sophisticated tool use such as “weight” and “center of gravity” are “unobservable” by non-human primates. Another possibly relevant consideration that I hope to explore concerns the apparent failure (as reported by researchers like Hauser and Santos) of even adult non-human primates to successfully integrate the physical understanding of object behavior that they apparently exhibit in passive observation tasks into action and manipulation.

Workshop Description

This workshop was sponsored by the Consciousness and Self-Consciousness Research Centre, with additional support from the Mind Association.