> How about studing the progression of a joke? Start with a brand new
>joke in a school. Let the joke float around and at a certain point query
>everyone to repeat the joke. Find out just how many variations there are
>and what was the most common variation. Then once you have figured out
>the most common variation tell the joke again to a far away school that
>is unlikely to have heard the joke. To another far away school tell the
>variation. Ideally both schools will have the same amount of students
>and be similiar in as many repects as possible. After a set period of
>time query the students of both schools. If more of them know the
>variation in one school then the original in the other then you have
>shown that the mutation of the joke increased its chances of survival!
>For extra measure you might want to start with another joke again but
>this time the variation and the original are started visa versa in the
>two far away schools.
My only problem is that this presents an enormous tracking problem. Not
that it isn't a good idea, but it simply seems beyond anyones means to
accomplish it.
Just ensuring that there is no crossover between the test subjects and those
of the control group seems nearly impossible. The very media we're talking
in makes a mockery of the concept of two schools in some way "isolated"
enough from one another for these experimental purposes.
My approach to forming experiments to prove/disprove this hypothesis is a
little different, but my ideas still need to percolate a bit more before
they're ready for the light of day. So let me run with your ball for a
minute since it might help me get my own juices get flowing as well...
How about using this media (e-mail/Net) to our advantage? Say we just want
to show that criteria #4 is met:
>4) At least some of those errors in copying can be proven to have
>an effect the replication rate of the copy.
We'll need to assume that we can reproduce those copying errors in the
lab--and we may need another experiment afterward to prove that that's
truly the case. (But that's why the criteria was broken down in the way it
was in the first place--so that we can whittle away at it one piece at a
time.) Let's assume we can reproduce the changes that take place during
transmission.
We come up with 5 or 6 variations on the same joke. It's important here
that we arrive at each variation using a specific alteration of the "parent"
joke, because we'll need to repeat the entire process with several different
parent jokes in order to prove any statistical tendency in the data.
Now, the controls: All we need to do for this experiment is to prove that
the reproduction rates are _different_ for the _different versions_ of the
same test joke. So, if our distribution of versions is completely random
and we repeat the experiment enough times, each time with different
distributions, we should be able to draw conclusions about differences in
replication rates (if any) between the versions.
So next we've got to find every Joke-of-the-Day mailing list, rec.humor,
alt.funny.haha, and the like that we can and subscribe to all of them. Put
the ones which we can post onto in a pool--it's from this pool, and
completely at random, that you will be selecting distribution points for
each version of the joke.
Post all versions of the joke to each of their appointed destinations at the
same time, and then simply record the data on how long it takes each version
to re-appear on another list. Gather all the data for several trials, plug
it all into a computer, and number-crunch the living bejezus out of it.
Not only do you test the "Errors in copying can be proven to have an effect
the replication rate of the copy" sub-hypothesis, but you're more than likey
to come up with some interesting data showing a corespondence between the
alterations you made in each version and its replication rate.
-Prof. Tim