The Older Paper: The Epistemology of the Dead Parrot Sketch

Anthony Greenwald's 2004 article "The Resting Parrot, the Dessert Stomach, and Other Perfectly Defensible Theories" (low-quality pdf) is a rare case of a readworthy contribution to a Festschrift. Greenwald laments the existence of endless theoretical debates in psychology which aren't as easily resolved as the methods textbook would have you believe. I'll try to give you the gist of it by means of quotation:
A common property of many long-unresolved theoretical debates in psychology is the flexibility of the competing theories: They are readily modified to accommodate unanticipated findings. Although theory modification is essential to scientific progress, repeated modifications can make contending theories effectively interchangeable, in turn making their competition illusory. Because theory competitions can be either sustained or resolved by voluntary actions of researcherltheorists, there is no way, beyond generalizing from the past, to predict that a specific theory competition will be sustained in illusory fashion.


If competition among theories is a good way of doing science, we should expect theoretical controversies to have a short life expectancy. When a controversy occurs, we should expect that experiments designed to choose among the theories will resolve the controversy within perhaps a few years. It is therefore informative to examine the life expectancy of theoretical controversies.
Philosophy of science doesn't help, Greenwald says:
[C]onsider a theoretical controversy that has recently occupied the time and attention of astronomers: Is the astronomical object Pluto a planet or is it a large comet? Regardless of any conclusion that philosophy of science might reach about the possibility of resolving this debate, astronomers have it in their power to prolong the debate endlessly, just as it is possible for them to achieve a speedy resolution. More generally, it is always an option for researchers either to prolong or to resolve any competition between theories. As a result, working scientists effectively make the philosophers' analyses irre1evant.
In a footnote he adds:
At the Festschrift conference, I showed a video recording of Monty Python's "Dead Parrot Sketch"(Chapman, Monty Python, 1969/1989) to illustrate the opportunities that exist for optionally prolonging theoretical debates. In that episode, a customer returns a parrot to the pet shop from which he recently purchased it and presents the pet shop owner with the complaint (theory) that the parrot is now dead and was indeed dead at the time of purchase. The pet shop owner presents and defends several alternative views, especially (and repeatedly) the theory that the parrot is resting (providing part of the title of this chapter). In this inspired and hilarious piece of comedy, the shop owner's persistent and imaginative refutations of the dead parrot theory illustrate the possibility of prolonging a theoretical debate indefinitely by defending alternative interpretations even in the presence of compelling data.
You may have seen that sketch before, but here it is anyway (partial transcript):

It's a nice sketch, but Greenwald's interpretation strikes me as exactly wrong. Let's have a closer look:

1. The customer advances the hypothesis that the parrot is dead, which is in line with the observation that the parrot is lying on the cage's base and isn't moving.

2. The shopkeeper advances the alternative hypothesis that the bird is "resting". (I'm fairly sure that parrots actually don't lie down to rest, but let's overlook this.)

3. The customer proposes to test the hypothesis: "If it's resting, I'll wake him up."

4. He proceeds to do so. The result of the test is in line with the hypothesis that the parrot is dead and not with the alternative that it's resting.

5. The same is true of alternative methods, such as shaking the parrot and throwing it onto the floor.

6. In quick succession, the shopkeeper then proposes alternative hypotheses (e.g., "it's stunned"), which are not in line with the data already gathered. To call this a degenerating research programme would be kind.

7. Moreover, the customer calls attention to the fact that data in line with the view that the "dead" hypothesis is wrong were fabricated: "I discovered the only reason that it had been sitting on its perch in the first place was that it had been NAILED there."

8. In the end, even the shopkeeper admits the customer's hypothesis was correct ("I'd better replace it").

Yes, the shopkeeper does propose "alternative interpretations even in the presence of compelling data", but the whole premise of the sketch is that those hypotheses are wrong - and we assume the shopkeeper himself knows this full well. Much of the comedy derives from the stupidity of the shopkeeper's excuses. Analogous hypotheses wouldn't fly in any scientific community.

If you teach introductory methods or whatever it's called at your place, you may want to use the "Dead Parrot Sketch" as an illustration of how hypothesis testing works, instead of how it doesn't.

No comments: