Of Minds and Language (12 page)

The history of thought abounds in ironies. One of them is that Sir Charles Sherrington's enormously influential book
The Integrative Action of the Nervous System
(Sherrington 1906) did as much as any work to persuade many scientists that a purely material account of mental activity – an account couched in neuroanatomical and electrophysiological language – was possible. The irony is that Sherrington, who died in 1952, was himself strongly committed to a Cartesian dualism. He believed that when he severed the spinal cord he isolated the purely physical neural machinery of the lower nervous system from the influence of an immaterial soul that acted on levels of the nervous system above his cut.

Sherrington placed the concept of the synapse at the center of thinking about the neurobiological mechanisms of behavior. His student, Sir John Eccles (1903–1997), further enhanced the centrality of the synapse in neuroscientific thinking by confirming through intracellular recordings of postsynaptic electrical processes Sherrington's basic ideas about synaptic transmission and its integrative (combinatorial) role. Eccles, too, was a Cartesian dualist, even though he secured the empirical foundations on which contemporary connectionist theories of mind rest. The irony is that a major motivation for connectionism is to found our theories of mind not only on physically realizable
processes but more narrowly on the understanding of neuroanatomy and neurophysiology that Sherrington and Eccles established. Indeed, the neurobiology commonly mentioned as a justification for connectionist theorizing about the mind is exactly that elaborated by Sherrington a century ago. Discoveries since then have made no contribution to the thinking of contemporary modelers.

A similar irony is that the empirical foundations for the now flourishing field of animal cognition were laid by behaviorist psychologists, who pioneered the experimental study of learning in non-human animals, and by zoologists, who pioneered the experimental study of instinctive behavior in birds and insects. Both schools were to varying degrees uncomfortable with representational theories of mind. And/or, they did not believe they were studying phenomena in which mind played any role. Nonetheless, what we have learned from the many elegant experiments in these two traditions is that the foundational abstractions of time, space, number, and intentionality inform the behavior of the birds and the bees – species that last shared an ancestor with humans several hundred million years ago, more than halfway back in the evolution of multi-cellular animals.

Some years ago (Gallistel 1990a), I reviewed the literature in experimental psychology and experimental zoology demonstrating that non-human animals, including birds and insects, learn the time of day (that is, the phase of a neurobiological circadian clock) at which events such as daily feedings happen, that they learn the approximate durations of events and of the intervals between events, that they assess number and rate (number divided by time), and that they make a cognitive map of their surroundings and continuously compute their current location on their map by integrating their velocity with respect to time. Here, in this paper, I give an update on some further discoveries along these lines that have been made in recent years.

The most interesting recent work on the representation of temporal intervals by birds comes from a series of brilliant experiments by Nichola Clayton, Anthony Dickinson, and their collaborators demonstrating a sophisticated episodic memory in food-caching jays (Clayton et al. 2006; Clayton et al. 2003, and citations therein; see also Raby et al. 2007). In times of plenty, many birds, particularly many species of jays, gather food and store it in more than ten thousand different caches, each cache in a different location, spread over square miles of the landscape (Vander Wall 1990). Weeks and months later, when food is scarce, they retrieve food from these caches. Clayton and Dickinson and their
collaborators took this phenomenon into the laboratory and used it to show that jays remember what they hid where and how long ago and that they integrate this information with what they have learned about how long it takes various kinds of food to rot.

The experiments make ingenious use of the fact that jays are omnivores like us; they'll eat almost anything. And, like us, they have pronounced preferences. In these experiments, the jays cached meal worms, crickets, and peanuts. Other things being equal, that is the order of the preference: they like meal worms more than crickets, and crickets more than peanuts. In one experiment, hand-reared jays, with no experience of decaying food, were given repeated trials of caching and recovery. They cached two different foods in two different caching episodes before being allowed to recover their caches. In the first of each pair of caching episodes, they were allowed to cache peanuts on one side of an ice-cube tray whose depressions were filled with sand. In the second episode of each pair, they were allowed to cache either mealworms or crickets on the other side of the same tray. Thus, on some caching trials, they hid peanuts in one half of the trays and mealworms in the other, while on other trials, they hid peanuts in one half and crickets in the other.

Either 4 hours, 28 hours, or 100 hours (4 days) after each pair-of-caching episode, they were allowed to recover food from both sides of the trays. On trials with only a 4-hour delay, both the mealworms and the crickets were still fresh and tasty when retrieved. At that delay, the jays preferred to retrieve from the caches where they had hidden either mealworms or crickets (depending on whether they had cached peanuts-and-mealworms or peanuts-and-crickets). On trials where a 28-hour delay was imposed between caching and recovery, the experimenters replaced the cached mealworms with mealworms that had been artificially rotted. Thus, on the first few peanuts-and-mealworms trials with a 28-hour delay before retrieval, the jays found inedible “rotten” mealworms where they had cached tasty fresh mealworms. By contrast, on peanuts-and-crickets trials, they found crickets that were still fresh after 28 hours in their caches. On trials with a 4-day delay before recovery, both the mealworms and the crickets had rotted; the peanuts alone remained fresh.

Control birds that never encountered rotted caches preferred the caches where mealworms and crickets had been hidden no matter how long the delay between caching and recovery. The experimental birds preferred those caches when only four hours had elapsed. When twenty-eight hours had elapsed, their preference after a few trials of each type depended on whether it was mealworms or crickets that they had hidden on the “better” side of the tray. If it was mealworms, they preferred the peanut caches, but if it was crickets, they preferred the cricket caches. When four days had passed, their preference after
a few trials (during which they learned about rotting) was for the peanut caches, whether it was mealworms or crickets that they had hidden on the “better” side of the tray.

In an ingenious extension of these experiments, Clayton, Yu, and Dickinson (2001) showed that the birds would adjust their retrieval preferences on the basis of information about rotting time acquired after they had made their caches. At the time the caches were made, they did not yet know exactly how long it took the meal worms to rot.

It appears from these experiments that the remembered past of the bird is temporally organized just as is our own. The birds compute elapsed intervals and compare them to other intervals in memory. They compare the time elapsed since they cached a cricket to what they have since learned about the time it takes a cricket to rot. Like us, birds reason about time.

There is an extensive literature showing that pigeons and rats can base behaviorally consequential decisions on estimates of the approximate number of events (Brannon and Roitman 2003; Dehaene 1997; Gallistel 1990a). In many of the experiments, the animal subjects make a decision based on whether the current number is greater or less than a target number in memory. Thus, these experiments give evidence that animal minds reason about number as well as about time. Brannon and her collaborators (Brannon et al. 2001) extended this evidence using a task that required pigeons to first subtract the current number from a target number in memory and then compare the result to another target number in memory.

In their experiment, the birds pecked first at the illuminated center key in a linear array of three keys on a wall of the test chamber. Their pecking produced intermittent flashes (blinks) of the light that illuminated the key. The ratio of the number of pecks made to the number of flashes produced varied unpredictably, for reasons to be explained shortly. After a number of flashes that itself varied unpredictably from trial to trial, the two flanking keys were illuminated, offering the bird a choice.

Pecking either of the newly illuminated side keys generated further intermittent flashes. Eventually, when the requisite number of further flashes on the side key they first chose had been produced, the bird gained brief access to a feeding hopper. For one of the side keys the requisite number was fixed. This number was one of the target numbers that the birds had to maintain in memory. For the other side key, the number of flashes to be produced was the number left after the flashes already produced on the center key were subtracted from a
large initial number. This large initial number was the other number that had to be maintained in memory. The greater the number of flashes already produced on the center key, the smaller the difference remaining when it was subtracted from this large initial number; hence, the more attractive the choice of the “number-left” key relative to the “fixed-number” key. The pigeons' probability of their choosing the number-left key in preference to the fixed-number key depended strongly and appropriately on the magnitude of the number left relative to the fixed number.

The random intermittency of the flashes partially deconfounded the duration of pecking on the center key from the number of flashes produced by that pecking, allowing the authors to demonstrate that the pigeons' choices depended on number, not duration.

Jays are not above stealing the caches of others (Bednekoff and Balda 1996). Experienced jays are therefore reluctant to cache when another jay is watching. They remember which caches they made while being watched and which jays were watching them (Dally et al. 2006). When no longer watched, they selectively re-cache the food that others observed them cache (Emery and Clayton 2001). “Experienced” jays are those who have themselves pilfered the caches of other jays; those innocents who have not succumbed to this temptation are not yet wary of being observed by potential thieves while caching (Emery and Clayton 2001). Thus, nonverbal animals represent the likely intentions of others and reason from their own actions to the likely future actions of others (see also Raby et al. 2007).

The zoologist Karl von Frisch and his collaborators discovered that when a foraging bee returns to the hive from a rich food source, it does a waggle dance in the hive out of sight of the sun, which indicates to the other foragers the direction (bearing) and distance (range) of the source from the hive (von Frisch 1967). The dancer repeatedly runs a figure-8 pattern. Each time it comes to the central bar, where the two circles join, it waggles as it runs. The angle of this waggle run with respect to vertical is the solar bearing of the source, the angle that a bee must fly relative to the sun. The number of waggles in a run is a monotonic function of the range, that is, the distance to the source.

It is somewhat misleading to say that the dance communicates the solar bearing, because what it really communicates is a more abstract quantity,
namely, the compass bearing of the source, its direction relative to the north-south (polar) axis of the earth's rotation. We know this because if the foragers that follow the dance and use the information thus obtained to fly to the source are not allowed to leave the nest until some hours later, when the sun has moved to a different position in the sky, they fly the correct compass bearing, not the solar bearing given by the dance. In other words, the solar bearing given by the dance is time-compensated; the users of the information correct for the change in the compass direction of the sun that has occurred between the time when they observed the dance and the time when they use the directional information they extracted from it. They are able to do this, because they have learned the solar ephemeris, the compass direction of the sun as a function of the time of day (Dyer and Dickinson 1996). Man is by no means the only animal that notes where the sun rises, where it sets, and how it moves above the horizon as the day goes on.

Knowledge of the solar ephemeris helps make dead reckoning possible. Dead reckoning is the integration of velocity with respect to time so as to obtain one's position as a function of time. Successful dead reckoning requires a directional referent that does not change as one moves about. That is, lines of sight from the observer to the directional referent must be parallel regardless of the observer's location. The farther away the point of directional reference is and the more widely perceptible from different locations on the earth, the better it serves its function. In both of these respects, the sun is ideal. It is visible from almost anywhere, and it is so far away that there is negligible change in its compass direction as the animal moves about. The problem is that its compass direction changes as the earth rotates. Learning the solar ephemeris solves that problem.

Dead reckoning makes it possible to construct a cognitive map (Gallistel 1990a:
Chapter 5
) and to keep track of one's position on it. Knowledge of where one is on the map makes possible the setting of a course from wherever one currently is to wherever one may suddenly wish to go. The computation involved is simple vector algebra: the vector that represents the displacement between one's current location and the goal location is the vector that represents the goal location minus the vector that represents one's current location. The range and bearing of the goal from one's current location is the polar form of that displacement vector.