[From Bill Powers (970306.0400 MST)]
Bruce Gregory --
My mind apparently wants me to get up to see the comet, so it woke me with a
reminder of a nagging loose end.
I believe it was you who showed the evolution stuff to a friend named Roy
who knows Cairns; Roy responded with this:
My only comment would be that the Cairns directed mutation is likely >NOT
random -- i.e., in addition to other non-directed mechanisms, the >bacterium
can actually alter its DNA so as to increase the likelihood >that DNA will
produce a gene product with the desired function. The >organism doesn't
need complete control over this process -- only >control over probabilities.
Of course direct, specific control over mutation would greatly strengthen my
hypothesis about the origins of life and of control being the same thing.
But I have a hard time believing that E. coli could know that the cause of
its starvation is not simply that galactose is no longer available, but that
fructose is available instead (I don't know that those are actually the
sugars involved). There might be many other causes of the lack of galactose
-- for example, the supply might have become more scattered, so that what is
needed is a mutation that produces faster locomotion. If that were the case,
the same starvation would have to lead to increasing the probability of a
What's not obvious about what we've come to call the "E. coli mode of
reorganization" is that it will give the _appearance_ of biasing a random
process toward producing a specific outcome. If a change in the wrong
direction (which would include a mutation that makes matters worse in any
respect) leads to a shortening of the interval to the next mutation, then
the irrelevant or deleterious mutation will decrease in frequency relative
to those that are beneficial. That is, the beneficial mutation will appear
to become more probable, because its effects persist longer before the next
mutation. But no change in the actual relative frequency of this mutation is
In our demo of the E. coli effect, the probability of travelling in any
direction after a tumble is the same. Yet if there were a large population
of E. colis on the screen, it would be found that many more of the E. colis
are moving in the right direction t a given time than in any other
direction. Furthermore, if you suddenly changed the direction of the
gradient, the great majority of E. colis would quickly be observed to be
moving in the new direction. The probability of moving in a specific
direction would appear to have shifted, but it has not. All that has changed
is the time spent swimming in a given direction before the next random
tumble. The outcome of the tumbles remains truly random.
The power of the E. coli mode of reorganization is that it requires NO
KNOWLEDGE of the environment. If some critical variable is departing from
its reference level, make a random change right away; if the error is
decreasing, postpone the next random change. That is ALL that is needed to
(apparently) increase the probability of a favorable change.
I'm beginning to think that this principle is unknown to evolutionary
biologists, and that it's important enough that they should be told about
it. Maybe "Roy" should show the E. coli demo to Cairns.