Let speculation run riot
BMJ 2000; 321 doi: https://doi.org/10.1136/bmj.321.7260.0 (Published 02 September 2000) Cite this as: BMJ 2000;321:0
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Dear Editor,
Michael Innis (1) and Peter Morrell (2) raised points that set me
thinking in more detail around this subject (not the least being that I
have presumed myself). This response is long for it is predominantly "me
getting my mind in gear". It is a mixture of passion, opinion, frustration
and (probably) naïvity.
Bio-Medicine (to my view) has sunk into the attitude that speculation
should only be carried out behind closed doors and between consenting
adults in case peer review finds you out. I think speculation has got this
bad image because, in the distant past, so called natural philosophers
(scientists) went into speculative free fall and created a lot of theories
that were accepted lock stock and barrel by gullible disciples. (This
reflects Peter Morrell's sentence - "19th century…//... possible tendency
to over-theorise.") Freud was, possibly, one of the heaviest offenders.
Because his hypotheses seemed attractive they were taken as gospel truths
by his disciples - and did a lot of harm. This underpins the importance of
implying "I have developed this model that helps me to formulate things in
a way that seems to give me some understanding. Until a better model comes
along this might be a fair start for others too". "The truth" is a hell-
hole for a scientist. Even the word science (from scio - to know)
compounds the problem. My Latin dictionary has a word pernoscio - to
thoroughly understand. Perhaps we should be practising pernoscience rather
than science.
People who don't think highly of the speculative (even theoretical!)
approach to science see themselves as empiricists who are sound scientists
(etymologically correct perhaps). They call a spade a spade and don't
pussy foot around in all this dark, speculative stuff. Should you doubt
this, then look out for it. It is highly prevalent in Bio-Medical circles
and is caricatured in the institutions that gives it voice. "We are far
too good as scientists to believe that speculation can tell us anything
important about Bio-Medicine. After all, all the major concepts have been
worked out in detail and all we have to do, to understand things better,
is to cut things up into smaller components and look at them in finer
detail." Now THIS is blinkered, naïve, empirical, inductive, reductionism.
I, personally, do not yet see any need to question the continuing
power of the reductionist approach. This approach seems to imply that
everything, including the human soul, will eventually (even if
complicatedly) be explained through logic and deeper and deeper
fundamental laws of physics. Blinkered reductionism, however, is the
approach where a parallel trend is adopted but now we assume that all our
current science is gospel true, set out in tablets of immutable stone by
the Galileos, Newtons, Darwins, Einsteins, Plancks and many equally or
less renowned scientists of this world. No major surprises are waiting in
the wings - we have already sussed them out. If you doubt that these
people exist then read John Hogan's book, "The end of science."(3) (My
guess is, he was playing devil's advocate.)
To practice pure (100%) induction is a highly skilled ability. So, it
is helpful to get it clear what Induction and Deduction are about.
An expectation, gleaned solely from observation and without any
conceptual interpretation other than through statistics, can only give us
a probability. For the whole of the last 2 or more millennia a sun has
been observed to rise every day. So the probability that one will do so
tomorrow is at least something in the order of a million to one (c 730,000
days in 2 millennia). This is all that empirical induction can tell us.
Inductive reasoning is simply a straight application of statistical
probability. For example, every morning for the last 364 days the turkey
had been fed twice a day. On the 365th day, without having any prior
knowledge of human habits or festivals, he expects to be fed but he is in
for a rude shock. He is an inductivist. He reached the conclusion that
there was at least a 364:1 chance he would get fed in the same manner on
Christmas day. Being also a naive statistician, he had every confidence in
the null hypothesis result (p <.00000etc,etc001) that "being fed twice
a day every day for 364 days in a row was pure chance" and that the 365th
day would, therefore, be like all others. The turkey's smart-Alec friend,
the pig, being a sceptical sort of chap who takes a dim view of humans,
mocked the turkeys optimism. He deductively guessed that one day, soon,
something nasty would happen and this was more likely with every passing
day; he had long ago begun to speculate and hypothesise about a selfish
reason for human benevolence.
Now, when we create a logical model of what is happening (like
Copernicus' de revolutionibus etc and Newton's principia - a
mathematically rich model) and find that it superimposes neatly over
empirical observations, this greatly heightens the statistical
significance of the empirical data. These models not only accurately
predict the correct observations today but those that will occur at other
extrapolated times both in the future or in the past. Putting this
together with the empirical probability that the observation of sunrise
will be repeated tomorrow leads to great confidence that a repeat will be
far more likely than one million to one. We can infer on good extrapolated
evidence that the earth has been spinning to face to and away from the sun
for, perhaps, four billion years. Furthermore, we can start to envisage
the conditions that might suddenly make this stop. This gives us a feel
for the improbability of this happening. This is deductive reasoning.
Meanwhile, the inductivist has yet to establish the probability that the
sun that sets is the same as the one that rises.
It is difficult not to mix induction and deduction. For example,
Newton mixed induction into his largely hypothetical deductive model. He
inductively assumed bodies have mass, they attract one another and that
distance and time are universally fixed constants. He formulated his
hypothesis on top of known empirical data and the excellence of fit made
him confident that his observations about the motions of the planets
around the sun had much higher significance than the "number-of-occasions-
observed-extrapolation-into-a-probability-that-it-will-happen-again".
Similarly, in deducing (on good evidence) that the sun has risen for at
least the last 2 billion years it is impossible for any person who has
been exposed to current herd beliefs NOT to colour this observation with
the knowledge that the earth spins on its axis and so sees both day and
night once every 24 hrs.
When we do an experiment to empirically test something then the
following happens. Let's take an example where one observation could be
seen to have high significance when the underlying law is suspected. Say
we want to know what effect increasing the amount of current that flows
through a resistor has. If we do this in an entirely empirical way then it
will be no good showing just once that doubling the current doubles the
potential difference across the resistor. To be of statistical
significance we will have to do this repeatedly, with many different
currents, until we have so many individual observations that we can
attribute high statistical significance to these repetitious observations.
There is no inductive basis for assuming that the result obtained with 1.5
amps will fall halfway between 1 and 2 amps OTHER than by repeated
observation.
However! It is difficult NOT to bring our logical scheming brains to
bear and consider, if we treble the current we might get three times the
PD. If we use 1.27 times the current we will get 1.27 times the PD. In
doing so we are generating a new law (or theory) of what is happening. We
are throwing years of exposure to (often flawed) logical thinking at the
problem. Indeed, if we had started our experiment having done enough
intellectual groundwork BEFORE we made any observations, then we might
have predicted these figures and been pretty satisfied that one reading at
each predicted level would be highly significant. This would save on the
repetitious readings needed to show an empirical, statistical
significance.
But then, next year, instances are observed that show that the
derived laws aren't quite that accurate! This reminds us that "the truth"
is a hell-hole.
Some people believe that Hume was implying that isolated induction is
only able to imply a probability. And, of course, in the real world no-one
but no-one approaches a problem in an entirely inductive fashion.
Statisticians do their best to achieve this and epidemiology tells us a
lot about the biases that get in their way. It also tells us how hard it
is to act as a pure inductivist.
For me, blinkered inductionism is the situation where investigators
spend little preparation carefully analysing their perceptual frameworks
BEFORE they start making observations - in particular, clearly separating
out the inductive and deductive elements of the process. "Smug
reductionism" is the misguided belief of its proponents that they are
unbiased observers of the empirical truth. My response to that? *****
(Censored). Humes' response to it? "We can't escape it - just get on with
it; but recognise that we can't help observing in a biased way that is
replete with both inductions and deductions that we are unaware we have
made." In this way he gets out of the philosopher's nightmare scenario of
using the logical argument that he claims, by his logic, to have
destroyed.
This is what I was trying to encompass by "Presumption is the
Achilles heel of blinkered, empirical induction."
Perhaps we should leave Charles Darwin to have the last word: "I have
steadily endeavoured to keep my mind free, so as to give up any
hypothesis, however much beloved -- and I cannot resist forming one on
every subject -- as soon as facts are shown to be opposed to it."
Sorry! Can't-resist-own-last-word: "Don't let the 'facts' put you off
too soon! They may not be facts at all but extrapolations of our own
prejudices dressed up as unbiased observation. If your theoretical base is
strong - hang out; all may come good."
(1) Innis M. e-BMJ 2000 http://bmj.com/cgi/eletters/321/7260/0#EL1
and EL4
(2) Morrell P. e-BMJ 2000 http://bmj.com/cgi/eletters/321/7260/0#EL5
Horgan J. The End of Science : Facing the Limits of Knowledge in the
Twilight of the Scientific Age. Little Brown and Co, 1996 (other editions
available).
Competing interests: No competing interests
Sir,
In his letter [1], David Horrobin states that “Speculation that is
disseminated yet is wrong does no harm apart from irritating a few people
who are so unimaginative that they do not know how to speculate.
Speculation that is broadly correct but is suppressed can lead to the
crippling of innovation.” [1]
He also expresses the view that suppression of speculation can only
be a bad thing in biomedicine. Jamie Cunliffe also echoes this viewpoint
in his letter [2]. When he says, “The discipline has sunk into a state of
smug reductionism,” it means the same thing to me as David Horrobin
saying: “people who are so unimaginative that they do not know how to
speculate” [1] As Jamie Cunliffe then states, such editors should “exert
no influence” on the evolution of ideas in biomedicine. I am happy to say
I entirely agree with virtually everything these learned correspondents
say on this matter.
However, I am unsure what Jamie Cunliffe means, in his second letter
[3], by this:
“Presumption is the Achilles heel of blinkered, empirical induction.”
[3]
There is a curious quote from Darwin, which states “Without
speculation there can be no good and original observation.” [4]
I have long pondered over this quote and have always thought Darwin
meant that unless you have some ideas then mere observation alone is not
much use to you in gaining an understanding of our world. Observations and
correlations on their own do not necessarily mean a great deal, but when
harnessed with speculation, we then have a tool better suited to the task.
I can see nothing bad, harmful or disreputable about speculation. It is,
in fact, a major component of the primary way we come to know just about
anything. It is how we make sense of observations made in our world.
Darwin’s quote seems to show that we do not come to know solely through
observing events, but also by speculating on their significance.
Speculation is indeed undervalued in science, just as Jamie Cunliffe
and David Horrobin say. It should be encouraged more widely in
biomedicine; it leads to growth of new ideas and challenges old dogmas, so
pruning the tree of knowledge of “dead wood”, and encouraging new growth
too. Surely, only those who cling tightly to old dogmas can have anything
to fear from change, innovation and speculation?
The only problem I can see with speculation again harks back mostly
to the 19th century, as a possible tendency to over-theorise. In the early
days of science, especially in life sciences [less so in physics], it was
mostly observation-driven and speculation was regarded as a disreputable
tendency, because until you have completed your accurate description of
the world, you should [arguably] reign-in any deductions about it. I
suppose, therefore, Darwin’s quote was made to some extent to counter this
view. Speculation alone is probably bad just as endless description and
observation leads nowhere in particular. But when joined together they can
lead to much new knowledge being generated. Arguably, observation alone
leads to no new knowledge, while speculation unchecked merely leads to an
overabundance of un-tested ideas.
Hence this quote, also of Darwin, which conveys much the same
message:
“About thirty years ago there was much talk that geologists ought
only to observe and not theorise; and I well remember someone saying that
at this rate a man might as well go into a gravel-pit and count the
pebbles and describe the colours. How odd it is that anyone should not see
that all observation must be for or against some view to be of any
service.” [5]
When Darwin saw the various finches on the Galapagos Islands it might
be regarded, for example, as merely a raw observation of fact, but the
unusual nature of the sight soon set going in his mind a veritable ferment
of speculation. I guess the above quote [4] comes from his ‘Voyage of the
Beagle’ or possibly his ‘Diaries’ and therefore relates to the mental
process he engaged when confronted, as he then was, with such unusual
plants and animals, that profoundly challenged the rather cosy view of
nature he had received while at Cambridge – the accepted view of ‘the
unimaginative’ and ‘smug reductionists’, perhaps? Therefore, in his case,
speculation did indeed lead directly to innovation and the generation of
new knowledge – the theory of evolution based on natural selection. In
this sense, therefore, we might wish to elevate speculation to a far
loftier position as a major engine of change of ideas in science.
Sources
[1] BMJ 2000; 321:571 (2 September), Letters, Why is speculation so
awful?
http://www.bmj.com/cgi/content/full/321/7260/571/a
[2] BMJ letter Jamie Cunliffe, Let speculation run riot, 2 September
2000
http://www.bmj.com/cgi/eletters/321/7260/0
[3] BMJ letter, Jamie Cunliffe, Professional Myopia - or definitional
scotoma? 6 September 2000,
http://www.bmj.com/cgi/eletters/321/7260/0
[4] In Biographical Dictionary in Infopedia CD ROM, 1997, no written
source given
[5] Charles Darwin. 1861. Letter to Henry Fawcett, quoted in Steven
J. Gould "Dinosaurs in the
haystack," Natural History 3 (92): 2-13.
Competing interests: No competing interests
Let me assure Jamie Cunliffe that he is mistaken in thinking
“Michael Innis is implying that speculation is bad science” I do no such
thing.
Nor do I pretend to be acquainted with all the important Scientific
speculations from the time of the triumvirate of Milesian Philosophers
Thales, Anaximander and Anaximenes [1] on the subject of what constitutes
cosmic matter, to the speculations of Hartle and Hawking [2] on Imaginary
Time and the fate of Singularities; however I can do without a lecture on
the “interchangeable use of ‘foreign organism’ and ‘pathogen’ by Jamie
Cunliffe.
The point I tried to make (obviously unsuccessfully in the case of Jamie
Cunliffe) is that the speculation of Hume and Russell on the subject of
Inductive generalization was misguided and myopic. I cannot make my reasons any
clearer.
Referring to my definition of ‘day’ Jamie Cunliffe says, “he shows how
easy it is to fall foul of a definitional blind spot”. The definition I used was from ‘The Concise Oxford Dictionary of Current
English’. Your readers can see "how easy it is to fall foul of a definitional blind spot" even for Jamie Cunliffe!
I decline to respond to any further comments from Jamie Cunliffe.
Michael Innis
References.
1. Treasury of World Philosophy (1959) page 54. Ed Dagobert D.Runes
Published by Littlefield, Adams & Co. Patterson, New Jersey.
2. Ferguson K. (1999) Measuring The Universe page 286 Published by
Headline Book
Competing interests: No competing interests
I think - it is not clear - that Michael Innis is implying that
speculation is bad science. He is doing so by defending (bottom up)
empirical induction. But, in doing so he shows how easy it is to fall foul
of a definitional blind spot. Given a choice, most of us would elect to
accept that tomorrow is that period of time starting immediately after
midnight tonight and extending to midnight the following night. Though we
have arbitrarily elected to synchronise 12-o-clock noon (GMT) with the
sun's zenith this is entirely volitional (wilfully ignored during BST) and
can hardly be regarded as cause or effect. Further, on Michael Innis'
definition tomorrow morning before dawn and tomorrow night after sunset
are non-existent. The standard second is unlikely to be changed in line
with the 2 msec/century slowing of the earth's spin.
This, I think, shows how careful we have to be with definitions
before we start extrapolating. It is a bit like the interchangeable use of
"foreign organism" and "pathogen" by those who are fixed in the belief
that the immune system discriminates self from non-self. Look up the
definition of a pathogen and you will see that they have scored an own-
goal for the "danger" protagonists.
Presumption is the Achilles heel of blinkered, empirical induction.
Quotations encompass far reaching principles in a few words. Science
Week ran this earlier this year.
"A hypothesis or theory is clear, decisive, and positive, but it is
believed by no one but the man who created it. Experimental findings, on
the other hand, are messy, inexact things, which are believed by everyone
except the man who did that work." -- Harlow Shapley (1885-1972)
Neither hypothetical deduction (speculation included here) nor
empirical induction - the top down and bottom up approaches respectively -
should be used in isolation. The fast track is to look at things from as
many different perspectives as possible so that we start to get a better
feel for what is going on.
"Let speculation run riot" highlights that speculation has got a bad
press in the erroneous belief that it is automatically "bad (or non-)
science". History will be the eventual judge but, if we heed Thomas Kuhn,
it will reveal little that is new.
Competing interests: No competing interests
Dear Editor,
Re: There is a serious asymmetry in the consequences of suppressing
speculation and allowing it to run riot. (David Horrobin).
Science comes in two parts – trail blazing/pioneering stuff (that
drags principle from a sea of background noise) and consolidation.
Consolidation is essential work. It is not, by nature, adventurous and
needs to be rigorous. It is the foundation of authoritative science. But,
in Bio-Medicine, consolidation has been elevated to the status of being
the only valid way and imagination has been relegated to a minor task -
probably because there are so many parameters that it is difficult to tie
them down in thought experiments. The discipline has sunk into a state of
smug reductionism. Painstaking elaboration of "facts" and the dissection
of detail are no substitute for regularly deconstructing then
reconstructing perceptual frameworks.
So, speculation needs to be freed up from peer review (effectively,
peer suppression). An answer may be to keep speculation and authoritative
science in separate compartments within a paper. Science that pretends to
be authoritative has to respect the consolidated body of accepted belief
about what constitutes "good science" and this conformity gives it its
seal of approval. It is highly unlikely that such a corporate body of
belief and method has been built entirely on a quicksand of gullible
presumption. However, even "good science" needs to face its challenges. It
is only by constantly testing the fallibility of accepted "certainties"
(dogmas) that flaws will be exposed (Paul Feyerabend's "anything goes").
The progress of ideas in science is very much like the process of
evolution itself. We need to produce large numbers of speculative ideas
that enter the next round of testing by authoritative, consolidated
science. The latter needs to be aware of its sensitivity to "heresy",
resist the trap and expect an occasional huge surprise. This evolutionary
process can be left to test many ideas against the potentially flawed
consensus view. Ideas that fail (the majority) still have value, for they
force us to hone our logic. Ideas that stick may expose the flawed logic
of unwarranted presumption and teach us much in the process. Authoritative
science aims for caution and conservatism. Speculative science aims to be
cavalier and loose. We simply have to label which is what whenever it is
presented.
Editors will want to limit speculation to the subject in hand and to
a set number of words but beyond this they should exert no influence.
Speculative comment could be added after peer review has had its swipe and
the body of the article is accepted. But this new section must be clearly
labelled as opinion/speculation/conjecture. Editors and peer reviewers
have a duty to encourage speculation to run riot. Only then we will begin
to see ideas fly in the biomedical sciences and the excitement of the
chase return. After all, if we are seeking out principle that is buried in
the "noise" of countless observations, its first appearance will be little
more than a vague rumour; it will be a long time before it is elevated
into an authoritative certainty.
Competing interests: No competing interests
Speculation - Professional Myopia - Hume
There is a tendency for all disciplines to speculate and exhibit some
degree of professional myopia. Philosophy is no exception.
Following Hume’s [1] criticism of Induction (all inferences from
experience, he believed, were the effects of custom, not of reasoning)
Induction was described by Bertrand Russell [2] as the ‘scandal of
Philosophy’. It has also been called ‘a piece of fraudulent nonsense’ [3]
and it would be true to say that virtually all Philosophers in the Western
tradition, would agree with Russell.
But Russell was mistaken.
He argued [4], “… The fact that two things have been found together
and never apart does not, by itself, serve to prove that they will be
found together in the next case we examine”. Up to this point Russell is
correct. But what he ignores is the view of Claude Bernard [5] who says
the observation must be taken a step further; it must still be established
that when one condition is removed so also is the other. In other words
the two things must be co-variant to prove a universal association.
Western Philosophers are under the illusion that Induction, as
practiced in Science, is as interpreted by Hume and Russell. Some will
not have heard of Claude Bernard of whose work Sir Peter Medawar [6] said,
“the wisest judgments on scientific method ever made by a working
scientist were indeed those of a great biologist, Claude Bernard”.
Russell justifies Hume’s position on Induction when he says, “that
the Sun will not rise tomorrow …implies no more contradiction than the
affirmation that it will rise”. But again, both Russell and Hume are under
a misapprehension.
Consider this.
“Tomorrow” means ‘the next day after today’.
“Day” means ‘the time the Sun is above the Horizon’.
We cannot speak rationally of ‘tomorrow’ if there is ‘no Sun above the
horizon’. The ‘Sun’ and ‘next day’ (tomorrow) are inextricably linked as
cause and effect or effects of the same cause – gravity. If there is ‘no
Sun’ there is ‘no Day’. One cannot occur without the other. Hume’s view is
a contradiction.
The scandal of Philosophy is not Induction, but Western Philosophy’s
myopic view of it.
Michael Innis FRCPath
References
1. Hume D. An Enquiry Concerning Human Understanding (Originally
published 1748) Editor Hendel CW. The Library of Liberal Arts
(1955) page 57. The BOBBS-MERRIL Company Ltd Indianapolis. New
York
2. Paulos JA. ‘Once upon a Number’ (1998 ) page162 Allen Lane
The Penguin Press.
3. Miller D. Skepticism and Relativism. Science, (1999) 285:(Letter
response) pages 201-02
4. Russell B Great Books of the Western World 20th Century
Philosophy and Religion (1990) vol 55 page 262. Editor in Chief Adler
MJ. Encyclopaedia Britannica, Inc. Publisher Robert P.Gwinn
5. Bernard C. An Introduction to the study of Experimental Medicine
Dover Edition (1957). Translated by Henry Copley Green . page 55
Dover Publications, Inc New York
6. Medawar PB Induction and Intuition in Scientific Thought. (1969)
pages 1-2 . Methuen & Co LTD London E.C.4
Competing interests: No competing interests
Re: Let speculation run riot
Editor
To speculate in Science is to pursue an inquiry into experimental or
observational facts. The objective of the inquiry should be to provide a
rational explanation of the facts. A rational explanation involves finding
causal associations (as opposed to non-causal or incidental associations)
between the facts being considered.
Without going into historical details, which date back to the ancient
Greeks, the finding of causal associations is a straightforward exercise
in logic.
If p then q,
And
If not p then not q
Then
p and q are related as cause and effect or as effects of the same
cause.
Substitute any true premises for p and q and the causal relationship
between the premises must hold. This is the quintessence of Scientific
Speculation – A Law of Causation.
Michael Innis
Competing interests: No competing interests