A blog about problems in the field of psychology and attempts to fix them.

Saturday, October 1, 2011

Tool Use in Fish = Good, Future Planning in Fish = Bad

As I was working on the next embodiment post, my attention was drawn to the recent report of tool use in fish (for example, this). As this report was exactly on topic with the issues Beyond the Brain is dealing with, it didn't seem a bad idea to take a minute to think about.

Basically, researchers found a fish that digs small clams out of sand, then swim around with the clam in their mouths until they find an 'appropriate' rock sticking up out of the sand. Then the fish hurl the clam at the rock to break it open and get to the yummy meat inside. My background is in animal behavior, so this all seemed very plausible and very cool. I had never heard of fish doing something exactly like this, but there are species fish do related things, like gathering stones to make 'nests'. So, the argument that fish 'use tools' seems strong, especially if we are willing to accept a graded notion of tool use, and are merely accepted that fish do something on the low end of that spectrum. However, the authors of the report also seem to believe that they have found evidence that the fish plan ahead. That last part seems problematic.


This point seems so obvious to me, with the way I was trained to approach behavioral studies, that I am struggling to explain it. That fish do a long series of events that end up with a given outcome simply does not mean that the fish had that outcome 'in mind' when the behavior-sequence started. A huge chunk of the findings from classical Ethology are relevant. Over and over again, researchers have identified specific aspects of the world ("sign stimuli" or "releasers" in the old dialect), in response to which organisms initiate behavior. Some times, a complex behavior is supported by the encounter of a long series of crucial variables, other times a complex behavior continues unsupported after initiation.

The classic example of the animal on 'autopilot' is the greylag goose rolling its egg back into its nest (this was a 'fixed action pattern' in the old dialect). This is classic Niko Tinbergen and Konrad Lorenz stuff. When an egg is found near the nest, the goose goes through an elaborate process of rolling the egg back into the nest. The researchers, through careful experimentation, determined what a goose experienced as an egg, by putting objects of various shapes and sizes near the nest, and seeing which the goose treated as an egg. (Yes, ethology is experimental phenomenology, and is entirely compatible with New Realism style behaviorism.... but that it a topic for another time.) More importantly, here, the researchers also found that if you remove the egg, or egg-like object, in the middle of the goose's behavioral sequence, the sequence continued through to the end. It is pretty hard to argue that the goose is forward looking, and basing its behavior on the goal of getting the egg into the nest, if taking away the egg does not change the sequence of behavior.

Even in cases that are not so clear cut, researchers often find that a series of stimuli are needed to support successful action. As Barrett puts it, no behavior happens in a vacuum.Those in applied behavior analysis (the useful application of operant conditioning) talk about this most obviously in the case of 'chaining'. Chaining occurs when one behavior creates a stimuli that triggers the next behavior. For example, for a well trained fencer, acting on 'autopilot', a given sword position might lead to a specific adjustment, that adjustment leads to a different sword position that in tern leads to the next adjustment, etc. Soon enough the fencer might have initiated a very technical series of moves and be honestly surprised at the position they ended up in. It is pretty hard to argue that the fencer is forward looking, and basing her behavior on the goal of getting to the final position, when the fencer's behavior is intimately intertwined with the opponents response (or lack thereof) and the fencer claims no intention of ending up where they did.

On the other hand, there are plenty of examples of animal behavior that are clearly forward looking, at least in the sense that the behavior is motivated towards accomplishing a goal. Studies of wasps or bees flying back to their nests is a classic example. That the wasp is trying to accomplish a goal (getting home) is seen when the wasp continuously adapts its behavior to environmental perturbations until it finds its nest. It will fly back various distances and try to reorient, use landmarks various distances away, and even dig in many different locations if necessary to uncover its nest.

So, I'm not saying that these tool using fish do not anticipate. Personally, I don't think fish are forward thinking, but I recognize that such is a bias, and open to empirical investigation. What I am saying is that a fish using a 'tool' does not, in any way, evidence that the fish are forward thinking. The fish digging for the clam need not have in mind its eventual need to find a rock, nor its eventual consumption of the meat. The researchers who are claiming otherwise are either unfamiliar with the decades worth of past research, don't understand the past research, or are just playing sensational angles to get undeserved media attention. (Personally, I hope it is the latter, because a little exaggeration, accidental or intentional, during an interview is understandable. Especially as the interviewer interjects their own language and spin.)

Other possible explanations for the fish's behavior abound. It could be the case that there is some sort of unalterable action sequence initiated by seeing the clam. Maybe the fish would continue seeking out its rock even if the clam was removed from its mouth. It could be the case that there are a series of behaviors each triggered by the very local stimuli. Maybe the fish would dig until something small and round filled its mouth, even a pebble, and then it would throw the pebble against the rock with equal enthusiasm. It also could be the case that fish could be forward looking. Maybe the fish varies its initial behaviors in some ways that indicate it is preparing for the final behavior, for example it may go after clams closer to ideal rocks, or avoid throwing the clam on the rocks when other fish are around that might take its prize away. There are studies that could distinguish these possibilities.

If we really, really, really, believed in embodied cognition, we would not accept speculative testimony as to what a fish was thinking, even testimony of an expert who had observed a small number of unusual actions. We would not accept the speculation, because doing so would be to fall into the dualistic premise that claims about the mind are necessarily speculative. Instead, we would attack any claims about human and animal minds as straightforward empirical questions about what the animal is doing in a given situation or set of situations.

No comments:

Post a Comment