Of Minds and Language (13 page)

There is a rich literature on navigation in foraging ants and bees, which make ideal subjects, because they are social foragers: they bring the food they find back to the communal nest, then depart again in search of more. In this literature, one finds many demonstrations of the subtlety and sophistication of the spatial reasoning that goes on in these miniature brains, which contain only on the order of 1 million neurons. For some recent examples, see Collett and
Collett (2000); Collett et al. (2002); Collett and Collett (2002); Harris et al. (2005); Narendra et al. (2007); Wehner and Srinivasan (2003); Wittlinger et al. (2007); Wohlgemuth et al. (2001). For a review of the older literature, see Gallistel (1990a:
Chapters 3
–
6
). Here, I have time to recount only two of the most important recent findings.

For many years, researchers in the insect navigation field have questioned whether ants and bees make an integrated map of their environment (e.g., Collett and Collett 2004; Dyer 1991; Wehner and Menzel 1990; but see Gould 1990). The alternative generally proposed is that they have memorized range-bearing pairs that enable them to follow by dead reckoning routes back and forth between familiar locations. They have also memorized snapshots of the landmarks surrounding those locations (Collett et al. 1998; Collett et al. 2002; Collett 1992; Collett and Baron 1994) together with the compass directions of those landmarks, and they have memorized snapshots of landmarks passed en route between these locations (Fukushi and Wehner 2004). But, it is argued, all of this information is integrated only with regard to a particular route and summoned up only when the ant or bee is pursuing that route (Collett and Collett 2004).

Part of what has motivated skepticism about whether the information from different routes is integrated into an overall map of the environment is that bees often appear to fail a key test of the integrated-map hypothesis. The question is, can a bee or ant set a course from an arbitrary (but recognizable!) location on its map to an arbitrary goal on its map? One way to pose this question experimentally is to capture foraging bees when they are leaving the hive en route to a known goal and displace them to an arbitrary point within their foraging territory. When released at this arbitrary new location, do they reset their course, or do they continue to fly the course they were on when captured? Under some conditions, they do reset their course (Gould 1986; Gould and Gould 1988; Gould 1990), but in most experiments, most of the bees continue to fly the course they were on (Dyer 1991; Wehner and Menzel 1990). This suggests that they cannot recompute the course to their old goal from their new location.

Against this conclusion, however, is the fact, often reported in footnotes if at all, that the bees who take off for the wild blue yonder on a course inappropriate to their goal (given their release location) are nonetheless soon found either at the goal they had when captured or, more often, back at the hive. They do not go missing, whereas bees released in unfamiliar territory do generally go missing, even if that territory is quite close to the hive.

The problem has been that we had no idea what happened between the time the bees disappeared from the release site flying on the wrong course to the
time they reappeared, either at their intended goal or back at the hive. Menzel and his collaborators (2005) have taken advantage of the latest developments in radar technology to answer the question, what do misdirected bees do when they discover that they have not arrived at their intended goal? Radar technology has reached the point where it is possible to mount a tiny reflector on the back of a bee and track that bee at distances up to a kilometer. Thus, for the first time, Menzel and his collaborators could watch what misdirected bees did. What they did was fly the course they had been on when captured more or less to its end. This brought them to an equally arbitrary location within their foraging terrain. They then flew back and forth in a pattern that a sailor, aviator, or hiker would recognize as the sort of path you follow when you are trying to “get your bearings,” that is, to recognize some landmarks that will enable you to determine where you are on your map. At some point this flying back and forth hither and yon abruptly ended, and the bee set off on a more or less straight course either for the goal they had been bound for when captured or back to the hive. In short, they can set a course from an arbitrary location (the location where they find themselves when they realize that they are not getting where they were going) to another, essentially arbitrary location (the location of the feeding table they were bound for). This result argues in favor of the integrated map hypothesis.

The final result I have time to report (Gould and Gould 1988; Tautz et al. 2004) moves the level of abstraction at which we should interpret the information communicated by the waggle dance of the returned bee forager up another level. These little-known results strongly suggest that what the dance communicates is best described as the map coordinates of the food source. Moreover, it appears that before acting on the information, potential recruits consult their map for the additional information that it contains.

In these experiments, a troop of foragers was recruited to a feeding table near the hive, which was then moved in steps of a few meters each to the edge of a pond and then put on a boat and moved out onto the pond. At each step, the table remained where it was long enough for the troop foraging on it to discover its new location and to modify appropriately the dance they did on returning to the hive. So long as the table remained on land, these dances garnered new recruits. But when the table was moved well out onto the water, the returning foragers danced as vigorously as ever, but their dances did not recruit any further foragers – until, in one experiment, the table approached a flower-rich island in the middle of the pond, in which case the new recruits came not to the boat but to the shore of the island, that is, to the nearest plausible location. In short, bees' past experience is spatially organized: like the birds, they remember
where they found what, and they can integrate this spatially indexed information with the information they get from the dance of a returning forager.

The findings I have briefly reviewed imply that the abstractions of time, space, number, and intentionality are both primitive and foundational aspects of mentation. Birds and bees organize their remembered experience in time and space. The spatio-temporal coordinates of remembered experience are accessible to computation. The birds can compute the intervals elapsed since they made various caches at various locations at various times in the past. And they can compare those intervals to other intervals they have experienced, for example, to the time it takes a given kind of food to rot. The bees can use the dance of a returning forager to access a particular location on their cognitive map, and they can use that index location to search for records of food in nearby locations. Birds can subtract one approximate number from another approximate number and compare the result to a third approximate number. And birds making a cache take note of who is watching and modify their present and future behavior in accord with plausible inferences about the intentions of the observer.

To say that these abstractions are primitive is to say that they emerged as features of mentation early in evolutionary history. They are now found in animals that have not shared a common ancestor since soon after the Cambrian explosion, the period when most of the animal forms now seen first emerged.

To say that they are foundational is to say that they are the basis on which mentation is constructed. It is debatable whether Kant thought he was propounding a psychology, when he argued that the concepts of space and time were a precondition for experience of any kind. Whether he was or not, these findings suggest that this is a plausible psychology. In particular, these findings make it difficult to argue that these abstractions arose either from the language faculty itself or from whatever the evolutionary development was that made language possible in humans. These abstractions appear to have been central features of mentation long, long before primates, let alone anatomical modern humans, made their appearance.

RIZZI: I was wondering how far we can go in analogy between the foraging strategy that you described and certain aspects of language. I wondered whether there is experimental evidence about strategies of rational search of this kind:
first you go to the closer spots and later to more distant spots. A particular case that would be quite interesting to draw an analogy with language would be the case of intervention, presenting intervention effects in these strategies. For instance, just imagine a strategy description of this kind, that there is a direct trajectory for a more distant cache; there is one intervening spot with a less desirable kind of food (let's say nuts rather than peanuts, or rather than worms). Would there be anything like experimental evidence that this kind of situation would slow down somehow the search for the more distant spots – or anything that would bear on the question of whether there are distance and/or intervention effects in search strategies? Because that is very typical of certain things that happen in language – in long-distance dependencies.

G
ALLISTEL
: As regards the second part of your question, on the interfering effect of an intervening, less desirable cache, I don't know of anything that we currently have that would be relevant, although it might very well be possible to do this. The setup that Clayton and Dickinson used, as I just said, doesn't lend itself at all to that because it's not like a natural setup where this situation would arise all the time. The birds are just foraging in ice-cube trays. However, some years ago we did a traveling salesman problem with monkeys, where they very much have to take distance into account, and where they have to take into account what they are going to do three choices beyond the choice that they are currently making. That is, the monkeys had to harvest a sequence, going to a number of cache sites. This was done by first carrying a monkey around and letting it watch while we hid food, before releasing it to harvest what it had seen hidden. The question was, would it solve the traveling salesman problem by choosing the most efficient route, particularly in the interesting cases where to choose the most efficient route, the least-distance route, you would have to, in your current choice, foresee or anticipate what you were going to do in a subsequent task. And they very clearly did do that. They clearly did show that kind of behavior, so I think that's relevant.

H
AUSER
: One of the puzzles of some of the cases that you brought up is that lots of the intimate knowledge that the animals have been credited with seems to be very specialized for certain contexts, which is completely untrue of so much of human knowledge. So in the case of the jays, it seems to be very, very located to the context of cache recovery. Now, maybe it will eventually show itself in another domain. We're taking advantage of natural behavior so maybe it will not. But in the same way that the bees seem to be one of the only species that externalize this knowledge in the communicative signal in a richness that is totally unparalleled in any other species but humans, so you get this kind of odd thing where the bees are only really sort of talking about one specific context.
You have rich social relationships, but there is no communicative signal outwards at all. So the question is – the way I've put it in the past is – animals have this kind of laser-beam intelligence and we have this kind of floodlight, and what happens? How do you get from this very, very selective specialization to probably a promiscuous system in humans?

G
ALLISTEL
: Well, of course the competence–performance distinction is just as important in interpreting the behavior of animals as it is in interpreting the language of humans. They have a lot of competences that they don't always choose to show us. But I agree with your basic point, and in fact it is something I have often emphasized myself. Animals show a lot of competence in a very sharply focused way. If I were to venture into perilous terrain and ask what language does for thought, one suggestion that one might offer is that, because it allows you to take these representations that arise in different contexts with, on the surface, different formal structure, and map them onto a common representational system, it may enable you to bring to bear the representational capacity of this module on a problem originally only dealt with by that module, and so this module can contribute something that the original module wouldn't have been able to do on its own. And that would be where the floodlight quality of human reasoning came in perhaps. The idea that language didn't really introduce new representational capacity, except perhaps insofar as it created a representational medium in which anything could be, to some extent at least, represented.

U
RIAGEREKA
: At some point I would like to hear your opinion, Randy, on this
Science
report on the bees doing their dance also for the purpose of finding a new nest, so the behavior is apparently not fully encapsulated for the purposes of foraging. I had no idea that they also did that, find a viable nest with procedures akin to those involved in foraging. I don't know how plastic that is. The point I'm trying to emphasize is this: would we find more of those apparently plastic behaviors if we knew where to look? That said, in the case of plasticity that we have seen in our system, my own feeling (and this is sheer speculation) is that generalized quantification – that is, the type of quantification that involves a restriction and a scope – is certainly central to much of human expression, but may be hard to find in other species. In fact, if Elena Herburger is right in her monograph on focus, this sort of full-fledged, crucially binary quantification may even be central to human judgment, especially the way Wolfram Hinzen is pushing that idea. It may be that the type of syntax you require for that type of quantification (which is one of the best understood systems in linguistics), however it is that we evolved it, might as well liberate, if you will, a kind of richly quantificational thought that I would actually be very
interested to see if animals exhibit. I mean, you know much more than I do about these things, Randy, but the experiments I have read do not get to generalized quantification. For example, in dolphin cases in the literature, it is reported that these animals get, say,
bring red ball, bring blue ball
, and so on; let's grant that much. But apparently they do not get
bring most ball
or even
bring no ball
. So maybe that would be another way to push these observations, another thing to look for, constructing experiments to test for behaviors of that truly quantificational sort.