Cartwright on reductionism in science

In hersuperb recent book APhilosopher Looks at Science , Nancy Cartwright revisits some ofthe longstanding themes of her work in the philosophy of science.  In anearlier post, I discussed what she has to say in the first chapterabout theory and experiment.  Let’s looknow at what she says in her second chapter about reductionism, of which she haslong been critical. 

Reductionismdoes not have quite the same hold in philosophy of science that it once did, havingbeen subjected to powerful attack not only from Cartwright, but from PaulFeyerabend, JohnDupré, and many others.  (Idiscuss the anti-reductionist literature in detail in Aristotle’sRevenge.)  Still, the ideathat whatever is real is somehow ultimately nothing more than what can inprinciple be described in the language of a completed physics exerts a powerfulhold on many.  Cartwright cites JamesLadyman and Don Ross as adherents of this view, and AlexRosenberg is another prominent advocate.  As Cartwright notes, in contemporary writingabout science, the lure of reductionism is especially evident in discussions ofthe purported implications of neuroscience for topics like free will.

Cartwrightsets the stage for her discussion by quoting a famous passage from physicistSir Arthur Eddington’s book TheNature of the Physical World:

I have settled down to the task ofwriting these lectures and have drawn up my chairs to my two tables.  Two tables! Yes; there are duplicates of every object about me – two tables, twochairs, two pens…

One of them has been familiar to mefrom earliest years.  It is a commonplaceobject of that environment which I call the world.  How shall I describe it?  It has extension; it is comparativelypermanent; it is coloured; above all it is substantial…  [I]f you are a plain commonsense man, not toomuch worried with scientific scruples, you will be confident that youunderstand the nature of an ordinary table…

Table No. 2 is my scientific table…It does not belong to the world previously mentioned – that world whichspontaneously appears around me when I open my eyes... My scientific table ismostly emptiness.  Sparsely scattered inthat emptiness are numerous electric charges rushing about with great speed;but their combined bulk amounts to less than a billionth of the bulk of thetable itself…

There is nothing substantial about my second table.  It isnearly all empty space – space pervaded, it is true, by fields of force, butthese are assigned to the category of “influences”, not of “things”. (pp.xi-xiii)

Now, reductionismholds that in some sense the first table is really “nothing but” the secondtable – or even that the first table does not, strictly speaking, really existat all, and that only the second table does (though philosophers typicallycharacterize the latter sort of view as eliminativist rather thanreductionist).

Reduced reductionism

The firstconsideration Cartwright raises to illustrate how problematic reductionism isconcerns the way reductionists have, over the last few decades, repeatedly hadto qualify their claims.  The ambitionsof reductionism have, you might say, been greatly reduced.  Bold type-typereductionism gave way first to a weaker token-tokenreductionism, and then to yet weaker superveniencetheories. 

Type-typereductionist theories hold that each typeof feature described at some higher-level science can be identified with some type of feature described at alower-level science, and ultimately at the level of physics.  Perhaps the best-known theory of this kind isthe original mind-brain identity theory,which holds that every type of psychological state (the belief that it israining, the belief that it is sunny, the desire for a cheeseburger, the fearthat the stock market will crash, etc.) can be identified with some specifictype of brain process.  A stock examplefrom the physical sciences would be the claim that temperature is identical tomean kinetic energy.

AsCartwright notes, one problem with this sort of view is that it is difficult tofind plausible cases of successful type-type reductions beyond such stockexamples.  Another is that the stockexamples themselves are not in fact unproblematic.  “Reduction” claims seem really to be eliminativistclaims after all.  For example, given theso-called reduction of temperature, it’s not that what we’ve always understoodto be temperature is really just mean kinetic energy.  It’s that what we’ve always understood to betemperature is not real after all (or exists only as a quale of our experienceof the physical world, rather than something there in the physical worlditself) and all that really exists is mean kinetic energy instead.

A problemwith supposing otherwise is that the laws that govern the features of somehigher-level description and the laws that govern the features of some allegedlycorresponding lower-level description can yield conflicting predictions.  One way to think about this – though notCartwright’s own example – is in terms of DonaldDavidson’s view that descriptions at the psychological level are notlaw-governed in the way that the materialist supposes that descriptions at theneurological level are.  Hence, even if abrain event of a certain type is strictly predictable, the corresponding mentalevent will not be.  Given this sort ofmismatch, there is pressure on the type-type reductionist to treat thehigher-level description as not strictly true.

Anespecially influential consideration that led philosophers to abandon type-typereductionism is the “multiple realizability” problem – the fact thathigher-level features can be “realized in” more than one type of lower-levelfeature, so that there is no smooth mapping of higher-level types on tolower-level types of the kind an ambitious reductionist project aims for.  In the case of the mind-brain identitytheory, the problem is that the same mental state (believing that it israining, say) could plausibly be associated with different types of brainprocess in different people, or even in the same person at differenttimes.  Or consider how an economicproperty like being one dollar can berealized in paper currency, in metal currency, or as an electronic record ofone’s bank account balance. 

This ledphilosophers to embrace less ambitious token-tokenreductionist theories.  The idea here isthat even if types of feature at ahigher level cannot be smoothly correlated with types of feature at a lower level, nevertheless every token or individual instance of a featureat the higher level can be identified with some token or individual instance ofa lower-level feature.  For example, this particular instance of believing that it’s raining is identicalwith that particular instance of acertain type of brain process.

AsCartwright notes, however, token reductions in fact tend to yield, after all, typereduction claims of a sort.  An examplewould involve disjunctive types atthe lower level of description.  Forinstance, a token reductionist view of mind-brain relations may entail that atype of mental state like believing thatit is raining is identical to a “type” of neural property defined as being in brain state of type B1 OR being inbrain state of type B2 OR being in brain state of type B3 OR…  And this will, in turn, open up thepossibility of a conflict between what the laws that govern the higher-leveldescription entail and what the laws that govern the lower-level descriptionentail.

If it isobjected that disjunctive “types” of the kind just described seem artificial, thatis certainly plausible.  But the problem,as Cartwright notes, is that this illustrates how identifying what counts as aplausible type is going to require detailed metaphysical analysis, and cannotbe read off the science, as the reductionist supposes.

In anyevent, token-token reductionism gave way in turn to talk of supervenience.  The basic idea here is that phenomena at somehigher level of description A superveneon phenomena at some lower level of description B just in case there could not be any difference at what happens atlevel A without some correspondingdifference in what happens at level B. 

But exactlywhat this amounts to is not obvious, and debating the meaning of superveniencehas, Cartwright complains, been a bigger concern of philosophers thanexplaining exactly why anyone should believe in it in the first place.  (More on this in a moment.)  As its vagueness indicates, supervenience entailsan even weaker claim than token-token reduction.  Though, in recent years, there has been a lotof heavy going about “grounding,” which, Cartwright notes, is stronger thansupervenience.  The idea is that allfacts are “grounded” in the facts described at the level of physics, in thesense that whatever happens at the higher levels is “due to” what happens atthe lower, physics level.  But here too, why suppose this is the case?

Groundless grounding

Where theclaim that everything supervenes on the level described by physics isconcerned, Cartwright says, there are three basic reasons given for it, none ofthem well worked out or convincing. First, there is a leap from the fact that the lower-level featuresdescribed by physics affect whathappens at the higher levels, to the conclusion that those features by themselvesentirely fix what happens at thehigher levels.  This is simply a non sequitur.

Second,there is a leap from the supposition that successful reductions have beencarried out in a handful of cases, tothe conclusion that reductionism is ingeneral true.  But this too is a non sequitur (and on top of that, thepremise is questionable).  Third, thereis the claim that physicalistic reductionism is in fact the method appliedwithin science.  But this, Cartwrightargues, is simply not true to the facts of actual scientific practice.

“Grounding”accounts of reduction suppose that the level described by physics is the sole cause of what happens at the higherlevels, and also that it is in no way itself caused by what happens at thehigher levels.  These claims too, arguesCartwright, are not supported by actual scientific practice.

Here sheappeals in part to recent work in the philosophy of chemistry, in which twogeneral lines of anti-reductionist argument have been developed.  The first and more ambitious of them arguesthat chemistry as a discipline rests on classificatory and methodologicalassumptions that are simply sui generisand make the features of the world it uncovers irreducible to those uncoveredby physics.  The second does not rule outreductions a priori, but argues on acase by case basis that purported reductions have not in fact successfully beencarried out.  (I discuss this work inphilosophy of chemistry at pp. 330-40 of Aristotle’sRevenge.)

But it isnot just that chemistry and other higher-level sciences are not in fact “allphysics” at the end of the day.  AsCartwright emphasizes, “even physics isn’t all physics.”  For one thing, “physics” covers a range ofbranches, theories, and practices, not all of which have been reduced to the mostfundamental theories.  For another, eventhe fundamental theories themselves are not fully compatible with each other,the notorious inconsistency between quantum mechanics and the general theory ofrelativity being a longstanding and still unresolved problem.  She adds:

The third and to me most importantpoint is that in real science about real systems in the real world, forpredictions and explanations of even the purest of physics results, physicsmust work in cooperation with a motley assembly of other knowledge, from othersciences, engineering, economics, and practical life. (p. 110)

Cartwrightthen goes on to describe in detail the Stanford Gravity Probe Bproject as an illustration of the vast quantity of theoretical knowledge andpractical know-how that are necessary in order to apply and test abstract physicaltheory, yet cannot itself be reduced to such theory.  This recapitulates a longtime theme inCartwright’s work over the decades, viz. that the mathematical models and lawsof physics are idealized and simplified abstractionsfrom concrete physical reality, and do not themselves constitute or captureconcrete physical reality.

In short,reductionism, Cartwright judges, is poorly defined and poorly argued for.  Its lingering prestige is unearned. 

I’ve mainlyjust summarized Cartwright’s arguments here, since I sympathize with them andthey supplement those that I develop in Aristotle’sRevenge.  They give us, though, onlyher case against the views sheopposes, rather than the positive account she’d put in place of them, which isdescribed later in the book.  More onthat in a later post.

Relatedreading:

Dupréon the ideologizing of science

Scientism:America’s State Religion

Theparticle collection that fancied itself a physicist

Fallaciesphysicists fall for

One LongCircular Argument

Science andScientism

Rosenbergroundup