Repost: Insightful Animal Behavior: A "Sufficiently Advanced Technology"

October 11, 2013

The news is chattering over a new paper by Smet and Byrne entitled “African Elephants Can Use Human Pointing Cues to Find Hidden Food” [link]. The lede is frequently the typical one for comparative cognition studies. Take this example from VOA:

Elephants are able to recognize human gestures without any sort of training, new research shows. Scientists believe the finding indicates that elephants are able to understand humans in a way most other animals do not.

They might be excused for this since the authors themselves write, in the Results and Discussion section “Here we found elephants capable of responding spontaneously to pointing gestures that require attention to subtle differences in the position of the forearm and hand.“. This is, however, a tired and old problem with this type of study.

Even Carl Zimmer, writing in the NYT, makes most of his post about this wonderous “spontaneous” property of all elephants. Still, to his credit he does include the critical caveat.

Diana Reiss, an expert on elephant cognition at Hunter College, wondered if the elephants had already learned about pointing by observing their handlers pointing to each other.
“In these elephant camps such opportunities can easily go unnoticed by their human caretakers,” said Dr. Reiss.

The authors themselves point this out, although they try to handwave it away:

All of these elephants have lived in captivity since infancy: they have had the opportunity to witness pointing used between humans. However, observation of human interactions does not automatically translate into aptitude at interpretation of these interactions.

Whoa dude. Whoa. Hold up. Extraordinary claims require extraordinary evidence. Except, apparently, in comparative cognition when we are just sooooo keen to believe findings that show species X is “just like humans” in some cognitive or behavioral property.

I have at least one observation in the archive that points out where not thinking hard about the study design can lead to unsupported conclusions being widely disseminated. This post was originally published Feb 25, 2008.


In the midst of World War I, Wolfgang Köhler conducted a famous series of experiments to investigate problem solving ability in chimpanzees. The lasting impression of these experiments, reinforced by just about every introductory Psychology text, was Köhler’s assertion that the chimps demonstrated “insightful” learning.
Did they now?


From this site:

Köhler’s most well-known work on chimp cognition was in the use of tools to gain access to food. A chimp would have to stack boxes to reach a banana that was suspended out of reach, or insert a narrow stick into a thicker one to produce a tool long enough to reach food. While Köhler’s star chimp, Sultan, did not immediately put two shorter sticks together to make one long one, he worked on the sticks for over an hour. When they had fitted together, Sultan immediately used the new tool to retrieve the bananas. This solution demonstrates insight – recognizing the “problem space” – rather than foresight.

Another set of descriptions from Köhler gushing over how the chimpanzee “spontaneously” jumps to the correct solution (“Eureka!”) without evidence of incremental training. Amazing!
Or, is it…

OMG! Pigeons have insight! We must reorganize our whole understanding of the supremacy of the apes. Right?
The scholarly description of this little experiment, while less intuitive than the video, resulted in a peer-reviewed publication. From some OldeSkool behaviorists making the point that when the behavioral and environmental antecedents are unknown, we might be led astray by our assumptions.

Epstein R, Kirshnit CE, Lanza RP, Rubin LC. ‘Insight’ in the pigeon: antecedents and determinants of an intelligent performance. Nature. 1984 Mar 1-7;308(5954):61-2.

Somehow I am reminded of Arthur C. Clarke’s Third Law: “Any sufficiently advanced technology is indistinguishable from magic“.

Now it may be the case that all brain/behavior relationships in animals more complicated than a cockroach approach Clarke’s “sufficiently advanced technology” as far as we are concerned at present. Nevertheless, at least the behaviorist tradition results in demonstrations that seem understandable to the average college freshman. You have stimuli, you shape incremental responses with the provision of consequences for particular responses and you observe learning or acquisition curves. It makes sense.

However sometimes we visit phenomena for which we do not have a great understanding of, or appreciation for, the “technology”. We understand much less about what generates the higher cognitive skills out of the (admittedly complicated) anatomy and signaling (chemical and electrical) properties of the brain. We may not have much appreciation for the behavioral history of animals in the wild or semi-feral captivity. And so. And so. When faced with phenotypes which seem beyond our ability to explain with our comprehensible level of technology, we grasp for magical explanations. “Insight”? “Consciousness”? “Self-awareness”? We struggle with understanding and classifying these properties of the brain.

I don’t want to belabor the analogy too much. I am not suggesting that believing in “insightful learning” or “consciousness” is the same thing as believing in “magic”. My point is that we should be cautious in assigning inferences of qualitatively different processes just because it is not immediately apparent to us, e.g. how a particular skill was acquired. Cautious in assuming that just because a specific species exemplar sitting before us can or cannot demonstrate a particular behavior this means that all members of the species can / cannot perform similarly.

And most importantly, we should be cautious in assigning order-of-magnitude or quality-based rankings. For example if we have the “understandable” technologies and the “sufficiently advanced” technologies lined up, the lowliest of the “magic” technologies may be, in objective terms, much closer to our higher “understandable” technologies than they are to the highest of the “magic/incomprehensible” technologies. In other words, the “insight” expressed by Köhler’s chimpanzee subjects may be in objective terms much closer to the “insight” expressed by Epstein’s pigeon than it is to some version of human “insight” which leads, for example to an advance in understanding brain function in a neuroscience laboratory.

Returning to a couple of additional “technologies” of the nonhuman ape brain, it turns out that chimpanzee’s highly touted ability to fish for termites with a bit of stick is a meticulously learned behavior taught by the mother. But..but.. tool use was supposed to be some qualitative watershed. A hugely significant difference in the animal kingdom between the “haves” and “have nots”. And believe you me it is a very large part of the comparative cognition theme to insist on examining “spontaneous” behaviors that are not trained and shaped though incremental learning. (It will not surprise you that the above Epstein video and similar behaviorist demonstration are precisely the reason for this distinction.) Now we have come to appreciate that this supposedly species-typical behavior, termite fishing, is incrementally learned. Now what would happen if some members of a species received incremental training and some did not?

It has been shown that other tool use in chimpanzees is culturally limited, meaning that one group of the species may be found to use rocks to crack hard-shelled coula nuts whereas another group many miles away may never do so, despite plenty of nuts and rocks lying around. So at one point, a theoretical pair of field primatologists working with distant groups of subjects might reach opposite conclusions about the “innate capability” of a single species for use of a particular tool. Should we conclude from this that chimpanzee group A is more “insightful” than chimpanzee group B in a constitutive, unalterable way? That they exhibit qualitatively and significantly different cognitive “ability”? Of course not…once we put all the data together. Then we realize we need to go looking for additional environmental contributions to the observed behavioral phenotype.
And tone down our rhetoric on nice neat species differences in “innate capability”.

A nice side effect, is that by considering the different environmental contributions to brain/behavior phenotypes which may on first blush seem fantastical or nearly miraculous, we can gain a more realistic perspective on the “technological difference” scale from simpler organisms all the way up to what is the most complicated brain technology of our acquaintance, the human.


Returning to the elephant study, the lesson from Epstein’s pigeon demonstration is clear. Without knowing the behavioral history of the subjects we cannot ascribe any behavior to “innate” or “spontaneous” properties of that animal. The chimpanzee “tool use” examples show that ascribing the behavior of one natural subpopulation of individuals to the entire species is an uncertain proposition, at best. This elephant study is a classic case in point and I recommend you read the coverage of it with a close eye. How many reports brush away, as do the authors, the devastating flaw that they do not really know the antecedent learning of the elephants? How many generalize this to all elephants and draw strong species-conclusions with other species for whom this supposed phenomenon has not been ever demonstrated?

h/t: @feathered_ape

Advertisements

11 Responses to “Repost: Insightful Animal Behavior: A "Sufficiently Advanced Technology"”

  1. 1981 Says:

    Maybe “The only way to survive is to fake data” is not quite accurate. But it certainly does seem that a lot of scientists believe that “The only way to survive is to **pimp** data,” and that’s perfectly well illustrated in this study when, as DM points out, the authors “try handwave [a critical caveat] away.”

    I think I am guilty as charged here: I pimp my data. And I do think that doing so has become critical for surviving in these lean times. I feel the need to show what I do in the best possible light, if I want to keep the gravy train running. Am I being scientifically inaccurate? Sure, but at least I get to have a job.

    Like

  2. SnowBoardDude Says:

    @1981

    Dude, everyone pimps their data. If there is something I learned while getting my PhD is to show what supports the hypothesis and hide what does not, which is the path of least resistance for publishing purposes. Publishing in the best possible journal is what really counts these days. That’s the new and revised scientific method.

    Like

  3. Grumble Says:

    “How many generalize this to all elephants and draw strong species-conclusions with other species for whom this supposed phenomenon has not been ever demonstrated? ”

    No careful scientist would do either of those things. So what’s the big deal?

    Like

  4. drugmonkey Says:

    This is the entire operational backbone of the field of comparative cognition. Being from the behaviorist tradition, this means war. There can be only one.

    Like

  5. drugmonkey Says:

    There’s also a difference between fluffing up your results to be better than they really are (the hit rate was about 67% for a two-alternative problem, i.e., chance at 50%), and waving off THE most likely alternative explanation without a single shred of evidence.

    A.F.S. never observed any of the handlers using pointing to direct any elephant.

    Right? The whole point of the paper is that the elephant “spontaneously” attends to what the human is doing. Why wouldn’t the elephants be attending to what the humans that they are around are doing whether the person intends to “train” the elephant or not? Could be as simple as one person pointing to another to drop the hay or water bucket right over there and the elephant coming to associate the pointing with the location where good stuff is located.

    Moreover, elephants’ success on this task was not related to age, indicating that different amounts of experience with human handlers did not influence these elephants’ abilities to solve the task

    how on earth do they know how long it takes to get to asymptote? Did they estimate how many times an elephant may have witnessed a human intentionally pointing in its lifetime? humans point all the frigging time!

    elephants spontaneously attend to and correctly interpret human deictic gestures without extensive prior learning opportunities

    A clearly unbased statement. And the whole paper depends on it.

    Like

  6. Steve Shea (@sheacshl) Says:

    @snowboarddude

    Are you serious? Where did you get your PhD?

    Like

  7. Joanne Says:

    “Nevertheless, at least the behaviorist tradition results in demonstrations that seem understandable to the average college freshman. You have stimuli, you shape incremental responses with the provision of consequences for particular responses and you observe learning or acquisition curves. It makes sense.”

    This has nothing to do with the elephant study, but I thought it worth pointing out that Gallistel has shown that the learning curve is an artifact of group averaging. Most individual learning curves, even for simple conditioning behaviors, are characterized by sudden jumps, so “insight” may be a relatively common feature of learning.

    Like

  8. drugmonkey Says:

    the learning curve is an artifact of group averaging

    yeah, anyone who has done a learning study knows this. It is an interesting question in substance abuse research for studies where we are talking about the acquisition of intravenous self-administration in rodents – ties into both the failure of preclinical research to model the conditional probability of dependence very well (a topic I discuss occasionally wrt human drug use) and the resurgence of interest in the “rat park” studies.

    Like

  9. bashir Says:

    When faced with phenotypes which seem beyond our ability to explain with our comprehensible level of technology, we grasp for magical explanations. “Insight”? “Consciousness”? “Self-awareness”? We struggle with understanding and classifying these properties of the brain.

    this is like, half of experimental/cognitive psychology.

    Like

  10. DrugMonkey Says:

    That isn’t a coincidence Bashir.

    Like


  11. […] here. Veronique Greenwood, and especially drugmonkey, took a decidedly more sceptical view here and here. All excellent […]

    Like


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: