Grimm's Law and Predictability (ex Re: The Neolithic Hypothesis)

[Robert Whiting]

On Tue, 20 Apr 1999 X99Lynx at aol.com wrote:

> In a message dated 4/20/99 1:44:51 AM, whiting at cc.helsinki.fi wrote:

> << "Predictable" does not mean "capable of being
> hypothesized about >>

> Lest any readers be confused about this:

> 'Predictability' and 'reproducible results' are the two 'traditional'
> requisites of scientific methodology.  This is from Dewey and those
> fellows.

And this is absolutely correct so long as "predictability" means
knowing what will happen in advance and "reproducible results"
means that the same thing (within the limits of experimental error)
will happen every time.

> The basic idea is that a hypothesis or premise ought to predict
> observable  results.

No, this is not the basic idea of hypothesis.  It is just your idea
of what a hypothesis should be; it is "fuzzy-think" based on
inaccurate definitions and incomplete reasoning.  The following
is what a hypothesis is according to the definition at
     http://www.writedesignonline.com/organizers/hypothesize.html

     hypothesis - 1. a proposition, or set of propositions, set
     forth as an explanation for the occurrence of some specified
     group of phenomena, either asserted merely as a provisional
     conjecture to guide investigating (working hypothesis) or
     accepted as highly probable in the light of established
     facts.  2. a proposition assumed as a premise in an argument.
     3. the antecedent of a conditional proposition.  4. a mere
     assumption or guess.

You will notice that none of this says anything about predictions.
Prediction is not part of the definition of hypothesis.  A
hypothesis may make predictions, but it does not have to.  A
hypothesis (in the scientific sense, i.e., definition 1 above) is
simply an explanation put forth to account for observed facts.
But the hypothesis cannot make predictions about these observed
facts without being circular.  If the hypothesis is created to
explain the observed facts one cannot say that the observed facts
are "predicted" by the hypothesis.  The only prediction that is
implied by the hypothesis is that the hypothesis is true, which is
trivial ("the universe works the way it does because that's the
way the universe works"; "this hypothesis is true because it
explains the observed data").

Now for any set of observed data there are an infinite number of
explanations of how that data came to be.  But while anything is
possible, not everything is probable.  Some explanations will
just be more plausible than others.  Predictions often come from
an attempt to generalize the hypothesis so that it accounts not
only for the observed facts, but also for other facts not yet
observed.  This reasoning from the specific to the general is
called induction.  But the problem with induction is deciding
which of the competing hypotheses is the one that should be
generalized, because it is often not at all obvious which
hypothesis is more plausible than the others.  This is summed up
neatly by Conan Doyle when he has Holmes say:

  "Ah! my dear Watson, there we come into those realms of
  conjecture, where the most logical mind may be at fault.  Each
  may form his own hypothesis upon the present evidence, and
  yours is as likely to be correct as mine."
Another way that predictions arise from hypotheses is by assuming
the truth of the hypothesis and using it as a major premise to
deduce some consequence that would follow from the truth of the
hypothesis.  This logical consequent seems to be what you are
considering to be the "prediction" that the hypothesis makes.
This process is called deductive reasoning.  One of the problems
with this is that it is full of pitfalls, especially for those not
trained in it.  Although a true major premise and a true minor
premise connected by a valid argument must produce a true
conclusion, any other combination of true and/or false premises
and/or valid and/or invalid arguments can produce either a true
or false conclusion.  Furthermore, using a different hypothesis
as the major premise in conjunction with a different minor
premise and/or a different argument might very well result in the
same conclusion (logical consequent).

Some people reserve "theory" for a hypothesis that satisfactorily
explains new observations, but I don't know that this is
standardized.  My Webster's says that the shared meaning element
of 'hypothesis', 'theory' and 'law' is "a formulation of a
natural principle based on inference from observed data."  Some
people have a strict hierarchical ranking of these, with
hypothesis referring only to explanations of observed data,
theory referring to satisfactory explanation of new observations,
and law referring to a time-tested theory that has never failed
to explain any observation (thus the law of gravity, the theory
of relativity, and the Neogrammarian hypothesis of the regularity
of sound change).  On the other hand, many people use some (or
all) of these terms interchangeably.

> Otherwise it cannot be tested.

Many hypotheses make predictions but still can't be tested.
Many hypotheses do not make predictions but still can be rejected.

The scientific method is extremely simple:  1) a problem is
identified, 2) relevant data is collected, 3) a hypothesis is
formulated to account for the data, 4) the hypothesis is
empirically tested.  Steps 1 and 2 do not have to occur in this
order.  The problem may emerge during the collecting of data or
may become apparent only after data has been collected for some
other reason.  But step 3 should be preceded by 1 and 2 in
whatever order they may come.  Formulating a hypothesis and then
collecting data to support it is not part of the scientific
method.  It is known as "speculating in advance of the evidence"
or "counting chickens before they hatch"; or as the famous
recipe for hassenpfeffer begins:  "First, catch your hare..."

But your claim that only a hypothesis that makes predictions
can be tested is false for a number of reasons.  First let me
quote to you from S. F. Barker, _Induction and Hypothesis: A
Study in the Logic of Confirmation_ (Ithaca, 1957), 157:

     ... the uncritical proponent of the method of hypothesis is
     claiming that any hypothesis is confirmed if and only if
     consequents deducible ["predictions"] from it are verified;
     but this would entail that many hypotheses which we set
     store by cannot be confirmed at all in a direct and natural
     way.  Hypotheses universal in form would fall under this ban
     for instance.  Only observational statements are verifiable
     and an observational statement needs to be existential in
     form; but from a universal statement no existential
     statement can be deduced.  Thus no universal statement
     (considered in isolation) can have any verifiable
     consequent, and thus no universal statement can be directly
     confirmed.  This would be a serious defect in the method of
     hypothesis.

Second, verifying "predictions" (logical consequents) of a
hypothesis does not necessarily confirm a hypothesis, because
competing hypotheses might have the same or similar "predictions"
and confirming the one could just as easily be considered as
confirming the others.  It is a test that is a non-test.

The testing of hypotheses, then, is not based on confirming the
"predictions" of the hypothesis, but rather on confirming some
rival hypothesis, or, as it is often stated, "falsifying" the
original hypothesis.  If a hypothesis that conflicts with the
original hypothesis can be confirmed, then the original
hypothesis is falsified and must be either rejected or
reformulated.  This concept of "falsification" was put forth by
Karl Popper in the mid-1930's and has since become the standard
on which modern theories of hypothesis testing are based.  If a
hypothesis has no test for falsification (i.e., there exists no
hypothesis that would contradict the original hypothesis that is
capable of being confirmed) then the hypothesis is labelled
"non-falsifiable" and considered to lie outside the scientific
method.

Thus it is not "making predictions," but rather "falsifiability"
that is the principal criterion of a well-formulated hypothesis.
Many quite valuable hypotheses do not make predictions that have
any observable results, but this does not mean that they are not
scientific or must be considered poor hypotheses.  Rather it is
the "non-falsifiable" hypothesis that should be considered
unscientific, regardless of whether it makes predictions or not.

So either a hypothesis will be falsified and rejected (a
hypothesis that has a test for falsification but is not
falsified is not "proved", it simply gains credibility), or no
new observations will be possible and it will simply remain one
hypothesis among many, or it will be confirmed by more and more
observations and confidence in it will increase.  But even when
it has been confirmed repeatedly, it still can't be proved.  It
always remains provisional even though as confidence in it
increases it may be upgraded from hypothesis to theory to law.
But no matter how many times the results of experiments agree
with a theory, one can never be sure that the next time the
result will not contradict it.

> "Reproducible results" means that the premise can also be
> tested by others.

No, this is not what it means.  It implies this, but what it
means is that every time an event that is accounted for by a
theory with reproducible results is repeated (by whomever), the
result will be the same.  Predictability and repeatability are
indeed the hallmarks of a sound scientific theory.  But they are
the results of hypothesis testing, not the basis for hypothesis
formulation and testing.  What they imply is that it is possible
to use Newton's laws of motion and gravity (his "Principia
Mathematica Philosophiae Naturalis") to design a lunar lander
that, despite severe limitations on size and weight, will be able
to generate the required amount of thrust to lift it and its
specified payload off the moon's surface and put it in orbit
around the moon *without ever having been to the moon*, and that
it will work every time.  Now *that's* predictability and
repeatability.

<snip>

Bob Whiting
whiting at cc.helsinki.fi