Brit. J. Phil. Sci. 49 (1998), 397-424
Galileo
the
and
Indispensability of
Thought
Scientific
Experiment
TamarSzaboGendler
AB STRACT
By carefully examining one of the most famous thoughtexpenments in the history of
science that by which Galileo is said to have refuted the Aristotelian theow that
heavierbodies fall fasterthanlighterones I attemptto show thatthoughtexperiments
play a distinctiverole in scientificinquiry.Reasoningaboutparticularentitieswithinthe
contextof an imaginaryscenariocan lead to rationallyjustifiedconclusionsthat given
the same initial information would not be rationallyjustifiable on the basis of a
straightforwardargument.
reconstruction
1 Argumentative
TheEliminationThesis
1.l
of terminology
1.2 Clarification
1.3 Thenegativeargumentandthepositiveargument
ThesisandtheDerivativityThesis
1.4 TheDispensability
andits reconstruction
experiment
2 Galileo'sthought
2.1Galileo'sthoughtexperiment
of the Galileocase
2.2Reconstruction
2.3Fourwaysoutfor theAristotelian
misses
2.4Whatthe reconstruction
3 DenyingtheDispensabilityandDerivativityTheses
does
whatthe thoughtexperiment
3.1Rejectingreconstruction:
3.2Rejectingthepositiveargument:whatmakesthesebeliefsnew?
3.3Rejectingthenegativeargument:whatmakesthesebeliefsknowledge?
andthe contrastwithNortonandBrown
3.4Constructivism
4 Conclusion
reconstruction
1 Argumentative
Philosopherswho are opposed to all things spooky tend to think that thought
experimentsin science are (at least in principle)eliminable,and thatwhatever
demonstrativeforce they have is the result of their being sound arguments
dressed up in heuristically appealing clothing. On such a view, a scientific
thought experiment'sjustificatory force comes from the fact that it can be
reconstructedas an argumentwith explicit premisesthatmake no referenceto
imaginaryparticulars.
) OxfordUniversity Press 1998
398
TamarSzabo Gendler
My goal in this articleis to challenge this view by carefullyexaminingone
of the most famousthoughtexpenmentsin the historyof science:thatby which
Galileo is said to have refutedthe Aristoteliantheory that heavier bodies fall
fasterthanlighterones. I will try to show thatthe thought-expenmentalformat
of Galileo's presentationplays an indispensablerole in the persuasivenessof
his case against the Aristotelian,and that a similar degree of persuasiveness
could not be obtainedon the groundsof explicit argumentalone.
1.1 The EliminationThesis
The view that thoughtexpenments lead to justified conclusions because they
are argumentsfinds clear articulationand powerfuldefence in a recentpair of
papersby JohnNorton.l In those papers,Nortonputs forth a hypothesisabout
thought experiments which he calls the Elimination Thesis. Paraphrasing
slightly, his thesis is this:
The Elimination Thesis: Any conclusion reachedby a (successful) scientific
thoughtexpenment will also be demonstrableby a non-thought-expenmental
argument.2
As initially formulated,the thesis is ambiguous;it is compatiblewith both a
weakerreading,whichI will call theDispensabilityThesis,anda stonger reading,
which I will call the DerivativityThesis.3In orderto formulatethese versions,
however,I need firstto clanfy whata numberof termsin the thesismean,andto
describebrieflythe sortsof argumentswhich mightbe offeredin its favour.
1.2 Clarificationof terminology
A numberof termsin the thesis requirefurtherelaboration.Let me begin with
'thought experiment' and 'non-thought-expenmentalargument'.To draw a
conclusion on the basis of a thoughtexperimentis to make a judgementabout
what would happenif the particularstate of affairsdescnbed in some imaginary scenano were actually to obtain.4One might then use that judgement in
1 Norton [1991] and [1996].
Cf. Norton [1991], p. 131 and Norton [1996], p. 336. I should make it clear at the outset that
althoughI am using Norton's articulationanddefence of this thesis as a convenientjumping-off
point,the specifics of his position arenot my target.For a detailedcomparisonof my views with
Norton's, cf. Section 3.4 below.
3 Cf. Norton [1996], pp. 354-8.
4 For other recent characterizations
of (scientific) thoughtexperimentssee, for instance:Bealer
[1998]; Brown [1986, l991a, b, 1993a, b, 1995]; Carrier[1993]; Gooding [1990, 1992, 1993,
1994]; Hacking [1993]; Humphreys [1993]; Irvine [1991]; Janis [1991]; Kujundzic [1993];
Laymon [1991]; Lipton [1993]; Massey [1991]; Nersessian [1984, 1992, 1993]; Norton [1991,
1993, 1996]; Prudovsky[1989]; Shepard[1994]; Sorensen[1992a, b]; Wilkes [1988]; as well as
other articles in Horowitz and Massey [1991]. The locus classicus is, of course, Mach [1883/
1960, 1897/1976];for otherearlydiscussionscf. also Hesse [1966]; Koyre [1960]; Kuhn[1964];
Popper [1959/1992]. For currentdiscussion, cf. Souder [undated].
2
Galileoandthelndispensability
of ScientificThoughtExperiment399
developing a more generaltheory,just as one might use the resultof an actual
experiment.By contrast,to draw a conclusion on the basis of a non-thoughtexperimental
argumentis to be led by a process of inductive or deductive
reasoningfroma set of explicit premiseswhich make no referenceto particular
hypothetical or counterfactualstates of affairs to a correspondinglygeneral
conclusion.Again, one mightuse thatconclusionas the basis for endorsingone
or another general theory about the phenomena in question. So thought
experimentsdiffer from non-thought-experimentalargumentsin two crucial
respects:first,they are not presentedas arguments,but ratheras invitationsto
contemplatea way thatthe world might (have) be(en); and second, they make
essential reference to particularhypothetical and counterfactualstates of
affairs.5
Whatthe EliminationThesis says is that any good scientificthoughtexperiment can be transformedinto a non-thought-experimentalargumentwithout
loss of demonstrativeforce. Given the characterizationjust offered, what an
elimination will involve is first a process of argumentativereconstructionin
which the narrativepresentationis replaced by a series of explicit premises
sufficient to establish the desired result, and then a process in which
those premises that make reference to hypotheticals, counterfactuals,and
particularsare replaced by premises in which no such reference is made. If
the EliminationThesis is correct,such a process will preservecompletely the
thoughtexperiment'sdemonstrativeforce.
What is meant by 'demonstrativeforce'? I will suggest two problematic
readings and then one that I will endorse. If the claim that 'any conclusion
reached by a good thought experimentwill also be demonstrableby a nonthought-experimentalargument'meansno morethanthatin the reconstruction
of a maturescience, the conclusions that were (as a matterof fact) reachedby
thought experiments can be derived from more fundamentalprinciples by
means of inference schemes licensed within the science, then the thesis is
trivially true. Even if the development of Newtonian mechanics relied on a
series of crucial thought experiments, its textbook presentationmight well
establish particularconclusions on the basis of more conventional forms of
argument.
On the other hand, there is a reading of 'demonstrativeforce' according
to which the thesis is trivially false. If the claim is taken to mean that-as
a matter of psychological fact-any conclusion which was reached by a
good thought experiment might also have been demonstratedto the person
who reached the conclusion by means of a non-thought-experimental
argument,then the Elimination Thesis is certainly false. None would doubt
the importantheuristicand illustrativerole played by thoughtexperimentsin
5 Cf. Norton [1991], p. 129 and [1996], p. 336.
400
TamarSzabo Gendler
scientific exploration, and the crucial tasks they play in instruction and
informaldemonstration.
The properreadingof 'demonstrativeforce' makes the EliminationThesis
epistemologically interesting.On this reading, demonstrativeforce concerns
the role thatthoughtexperimentsplay in living bodies of knowledge:afterthe
moment of discovery and before the end of inquiry. It concerns whether a
particularconclusion based on a particularprocess of reasoning (thought
experiment) is thereby justified whether if such a process leads to true
beliefs, those beliefs should count as knowledge. So the issue raised by
the Elimination Thesis is this: can reasoning about (reasonably) specific
entities within the context of an imaginaryscenariolead to rationallyjustified
conclusionsthat given the same initialinformation would not be rationally
justifiable on the basis of a straightforwardargument?In the next section, I
consider two reasons that one might think such eliminationsare possible.
1.3 The negative argument and the positive argument
The Elimination Thesis (that is, the thesis that thought experiments are
dispensable)can be defendedwith two arguments,one negative, one positive.
So, forinstance,JohnNortonargues(negatively)as follows:thoughtexperiments
mustbe argumentsbecausethereis nothingelse for themto be.
Thoughtexperiments
in physicsprovideorpurport
to provideus informationaboutthephysicalworld.Sincetheyarethoughtexperiments
rather
thanphysical experiments,this informationdoes not come from the
reportingof newphysicaldata.Thusthereis onlyone non-controversial
sourcefromwhichtheinformation
cancome:it is elicitedfrominformation
we alreadyhaveby anidentifiable
argument
(Norton[1991],p. 129).
Nortonconsidersthis position almost trivial: 'the alternative,'he writes, 'is to
suppose that thought experiments provide some new and even mysterious
route to knowledge of the physical world' (Norton [1991], p. 129).6 So the
negative argumentcontends that if we have obtainednew informationabout
the empiricalworld withouthaving obtainednew empiricalinformationabout
the empirical world, the only way we couldhave done so is by means of an
argument.
The positive reason that one might think that thought experiments are
just argumentsin disguise is that 'the analysisand appraisalof a thought
experimentwill involve reconstructingit explicitly as an argument',so that 'a
good thoughtexperimentis a good argument,a badthoughtexperimentis a bad
argument'(Norton [1991], p. 131; [1996], p. 335). So the positive argument
amounts to saying that if a thought experiment can be reconstructed
as an
6 For an endorsementof this alternative,cf. Brown [199la, b, 1993a, b, 1995].
of ScientificThoughtExperiment401
GalileoandtheIndispensability
Evenif the reasonI
argument,thenwhatit was all alongwasan argument.7
scenario
a thought-expenmental
come to know somethingby contemplating
doesn'tseemto be becausethereis an argumentinto which the thought
is
it is. The reasonmy belief is justified
experimentcan be reconstructed,
allalong.
wasa disguisedargument
because,in theend,thethoughtexperiment
1.4 The Dispensability Thesis and the Derivativity Thesis
Thesisas originally
We arenowin a positionto recognizethattheElimination
anddefendedactuallyinvolvestwodistinctclaims.Thesemightbe
formulated
statedas follows:
The Dispensability Thesis: Any good scientificthoughtexperimentcan be
force,by a non-thought-experimental
replaced,withoutloss of demonstrative
argument.
forceof anygoodscientificthought
The Derivativity Thesis: Thejustificatory
experimentcanonlybe explainedby the factthatit canbe replaced,without
argument.
force,by a non-thought-experimental
loss of demonstrative
Thesissaysthatwe canalwaysgetfromhere
Looselyput,theDispensability
to therewithoutappealto a thoughtexperiment.If a thoughtexperiment
us fromone stateof belief to another,a non-thoughtlegitimatelytransports
argumentcouldtoo. Thoughtexperimentsmay be convenient
experimental
andefficientwaysof reachingconclusionsaboutthephysicalworld,butthey
theygetus wherewe want
thata carhasoverwu,lking;
haveonlytheadvantage
to go muchmorequickly,buttheydon'tgetus anywherewe couldn'treachby
morepedestrianmeans.
The DerivativityThesis says that not only can any good scientific
force, by
thoughtexperimentbe replaced,withoutloss of demonstrative
argument,but that to the extent that a good
a non-thought-experimental
force, it is because, deep
scientificthoughtexperimenthas demonstrative
down, the thoughtexperimentis an argument.We may be misledby the
is
surfacefeaturesof the case to thinkthatsomethingnon-argumentational
doingjustificatorywork,but we are wrong.The reason the Dispensability
at play was something
Thesisis trueis thatall thatwas ever justificatorally
Whatlookedlikea carturnedoutto be propelledby footpower
argumentative.
all along(likea child'sgo-car,or one of the vehicleson The Flintstones). So
Thesissayswe can get by withoutwhatwe commonlycall
theDispensability
the DerivativityThesistells us thatwe alreadydo.
thoughtexperiments;
7
'The workingsand achievementsof any thoughtexperimentcan be revealed and capturedfully
in an explicit argumentwhich employs the same resources'(Norton[1996], p. 339; cf. also ibid.,
pp. 357-8).
402
TamarSzabo Gendler
In the next section, I challenge the DispensabilityThesis by showing thatit
does not hold true of a widely acclaimed thought experimentof Galileo's. I
choose this case for two reasons. First, since this particularexample is generally treatedas the paradigmof an effective thoughtexperiment,diagnosing
the sourceof its success is itself a worthwhileendeavour.8Second, challenging
the DispensabilityThesis in this way allows me to shed light on the Denvativity Thesis as well. Obviously, if the Dispensability Thesis is false, the
Derivativity Thesis is too; the more interesting question is whether some
alternativeexplanationcan be offered of the thoughtexpenment's success. I
try to say somethingpositive about this question in Section 3.
2. Galileo's thought experiment and its reconstruction
2.1 Galileo's thought experiment
Perhapsthe most famous thoughtexpenment in the historyof Westernscience
is the thoughtexpenment with which Galileo is creditedwith having refuted
the Aristotelian view that the speed with which a body falls is directly
proportionalto its weight.9 The thought experiment appearsin his last and
most mature work, the Discourse Concerning Two New Sciences, in the
context of a more general discussion of the possibility and natureof motion
in a void.l? Galileo's goal in the section as a whole is to establishthat 'if one
were to remove entirely the resistance of the medium, all materials would
For some of the manydiscussionsof this andrelatedthoughtexperimentsof Galileo, cf. Brown
[1986, l991a, b, 1993a, b, 1995]; Cargile [1987]; Clement [1983]; Koyre [1939/1979, 1960];
Kuhn [1964]; Norton [1996]; Prudovsky[1989]; Sorensen [1992a]. In my discussion below, I
follow the somewhat unfortunatepractice of consideringthis thoughtexperimentoutside of
both its historicaland textual contexts. As a partialremedy to this misleadingpresentation,I
referthe readerto some of the many generaldiscussionsof Galileo's workand its context;one
might fruitfullybegin with: Butts and Pitt [1978]; Claggett [1959]; Clavelin [1974]; Cooper
[1935]; Damerowetal. [1992]; Dijksterhuis[1961];Drake[1978,1989,1990]. See also sources
listed in the next three footnotes.
9 Challenges to the Aristotelian thesis both empirical and conceptual had appeared in
a number of mid- and late sixteenth-centuryworks. (For relevant passages, see: (Cardan)
Cooper[1934], pp.7-77; Damerowetal. [1992], p.365; (Tartaglia)DrakeandDrabkin[1969],
pp. 63-143, esp. 120ff.; Damerow et al. [1992], p. 378; (Benedetti) Drake and Drabkin
[1969], pp. 147-237, esp. 206; Drake and Drabkin [1969], pp. 31-41; Dijksterhuis[1961],
pp. 269-71; Drake [1989], pp. 27-30; (Stevin) Cooper[1935], pp.77-80; Dijksterhuis[1961],
pp. 324-9.)
Galileo himself had produceda less conclusive version of the famous thoughtexperimentas
early as 1590 in an unfinishedand unpublisheddialogue on motion; cf. Cooper [1935], pp.
80-90; Drake and Drabkin [1969], pp. 331-77; Galilei [1590/1960], pp. 26-38, esp. 29
(National Edition, p. 265). For an interesting discussion of whether Galileo ever actually
performed such an experiment, cf. Cooper [1935]; Drake [1978, 1989, 1990]; Drake and
Drabkin [1969]; Koyre [1960]; Segre [1989]. Since my primarypurpose in this paper is not
historical, I will focus only on Galileo's 1638 presentationof the refutation,bracketingthe
interestingquestion (a question not without philosophicalinterest)of why it was that such a
simple and obvious mistake remained part of the West's scientific world view for nearly
2000 years.
8
GalileoandtheIndispensability
of ScientificThoughtExperiment403
descend with equal speed' (Galilei [1638/1989], p. 116);ll the thoughtexperiment in question leads to the weaker conclusion that 'both great and small
bodies, of the same [material],are moved with like speeds' (Galilei [1638/
1989], p. 109, italics added,bracketedword replaced).
The view thatGalileo is challengingis that 'moveablesdifferingin heaviness
are moved in the same medium with unequal speeds, which maintainto one
anotherthe same ratioas theirweights [gravita]'(Galilei [1638/1989], p.106).
That is, he is challenging the view that heavier bodies fall faster than lighter
ones, andthatthey do so in directproportionto theirheaviness. On the version
Galileo takes himself to be opposing,the proportionalityis linear;'a moveable
ten times as heavy as another,is moved ten times as fast' as the other (Galilei
[1638/1989], p 106) 12
The famous thought experiment, rephrased slightly, is the following.
Imagine that a heavy and a light body are strappedtogether and dropped
from a significant height.l3 What would the Aristotelian expect to be the
natural speed of their combination? On the one hand, the lighter body
should slow down the heavierone while the heavierbody speeds up the lighter
one, so theircombinationshouldfall with a speed thatlies between the natural
speeds of its components.(Thatis, if the heavy body falls at a rateof 8, andthe
light body at a rateof 4, then theircombinationshouldfall at a ratebetweenthe
two (cf. Galilei [1638/1989], p. 107).) On the other hand, since the weight of
the two bodies combined is greaterthan the weight of the heavy body alone,
theircombinationshouldfall with a naturalspeed greaterthanthatof the heavy
body. (Thatis, if the heavy body falls at a rateof 8 andthe light body with a rate
of 4, their combination should fall at a rate greater than 8.) But then the
combined body is predictedto fall both more quickly, and more slowly, than
the heavy body alone (cf. Galilei [1638/1989], pp. 107-8). The way out of this
paradox is to assume that the natural speed with which a body falls is
independentof its weight: 'both great and small bodies ... are moved with
like speeds' (Galilei [1638/1989], p. 109).
2.2 Reconstruction of the Galileo case
Transformed into an argument that conforms to the strictures of the
Elimination Thesis, Galileo's reasoning can be reconstructedas follows.l4
l 1 Here andelsewhere, I have made use of Drake's 1974/1989 translationof the Discorsi. Foreasy
cross-referencingwith the more widely available (though less reliable) Crew and De Salvio
translation,page referencesare to the National Edition (except where noted).
12 Cf. Aristotle, Physics, 215a24-216a21; On the Heavens (De Caelo), 301b.
13 Note thatin the remarksthatfollow, all referencesto bodies shouldbe understoodas referringto
bodies of the same material.For the purposesof my discussion, this constraintis irrelevant.
14 For Norton's reconstructionof this case, cf. Norton [1996], pp. 340-5; I discuss his version in
Section 3.4 below.
404
TamarSzabo Gendler
The first claim of the Aristotelianis that:
(1) Naturalspeed is mediative.
That is, naturalspeed is a propertysuch that if a body A has naturalspeed sl,
anda body B has naturalspeed s2,the naturalspeed of the combinedbody A-B
will fall between sl and s2.
The second premise of the reconstructionis that:
(2) Weight is additive.
Thatis, weight is a propertysuch thatif body A has weight wl, andbody B has
weight w2, the weight of the combinedbody A-B will be equal to the sum of
wl and w2.
From these two premises (plus the assumption that not all weights and
naturalspeeds are either zero or infinite),it follows that:
(3) Naturalspeed is not directly proportionalto weight.
Forthe firstis a mediativeproperty,whereasthe second is an additiveproperty,
and a mediativepropertycannotbe directlyproportionalto one thatis additive.
Furthermore,the only way to maintain(1), (2), and (3) simultaneouslyis to
assumethatall naturalspeeds arethe same. Then weight mightbe additiveand
naturalspeed (in a vacuous sense) mediative, with no contradictionthereby
implied. Thus naturalspeed is shown to be independentof weight.
2.3 Four ways out for the Aristotelian
If the DispensabilityThesis is true, then Galileo's thoughtexperimentshould
be replaceableby some non-thought-experimentalargumentwithout loss of
demonstrativeforce. My goal in the next two sections (2.3 and 2.4) is to show
that the reconstructionpresentedin Section 2.2 is not such an argument.l5
I begin my case by pointingout thatthereare a numberof 'ways out' for the
defenderof the view that naturalspeed is directly correlatedwith weight a
view which, for the sake of convenience, I will call the Aristotelian view.
These ways out involve denying premises (1) and (2) by proposinga series of
alternativehypothesesaboutthe physicalpropertiesof strapped-bodies,thatis,
bodies of the sort describedby the thought experiment.The point of talking
aboutthese ways out is to show thatthereare ways to maintainthe negationof
(3) by adoptingalternativesto (1) and (2), and adoptingthese may well be less
disruptiveto the Aristotelianpicture than adopting (3). What I will suggest
below is thatthese ways out, thoughlogically available,runcounterto certain
tacit knowledge about the physical world. It is for this reason that, when the
Below I suggest reasonsfor thinkingthatthis will be truefor any argumentativereconstruction
that conformsto the stricturesof the eliminationthesis.
405
of ScientificThoughtExperiment
GalileoandtheIndispensability
case is presentedas a thought experiment,they do not even occur to us. To
block them as moves in a straightargument,however, requiresmetaphysical
commitments that seem not to be at work in the thought experimentitself.
What these commitments are, and what role I think they actually play in
Galileo's reasoning,is a point I will turnto afterpresentingthe four ways out.
The firstways out would be for the Aristotelianto deny thatthe propertiesin
question are determinatefor strapped-togetherbodies in one of the following
two ways.
(4) Naturalspeed is not physically determinatefor strapped-bodies.l6
(5) Weight is not physically determinatefor strapped-bodies.
That is, she might reject (1) or (2) on the groundsthat they presupposethat
naturalspeed and weight are propertiesthat apply universally,even to bodies
that are in some way monstrous.l7Since strapped-bodiesare odd entities, she
might say, they need not be governedby the sortsof laws that govern ordinary
objects. In particular, they need not have determinate natural speeds or
weights.
The thirdway out for the Aristotelianwould be to avoid the conflictbetween
(1) and (2) by saying thatthereis a fact of the matteraboutwhethera strappedbody is one body or two, and thatits physical propertiesin falling will depend
on the answerto this question. She might say:
(6) Natural speed and weight are mediative for strapped-bodiesthat are
united. Naturalspeed and weight are additive for strapped-bodiesthat are
unified.
That is, sometimes when two bodies are strappedtogether, they are merely
united and remain,as a matterof fact, two objects; sometimes, when they are
strappedtogether,they are unifiedandform, as a matterof fact, a single object.
In the firstcase, both weight and speed will be mediative;so thatthe combined
body will have a weight intermediatebetween those of the two originalbodies,
and fall with a naturalspeed that lies between the two original speeds. In the
second case, both propertieswill be additive;so that the weight of the unified
body will be equal to the sum of the weights of its componentparts, and its
naturalspeed correspondinglyequal to the naturalspeeds of the two combined.
Since the mediativityof the propertiesholds only with respectto unitedpairsof
objects,andthe additivityonly with respectto unifiedsingle objects,thereis no
16
17
For a version of this 'way out', cf. Koyre [1960], p. 51.
Galileo pre-emptivelydeals with this by getting a concession from Simplicio straightaway that
'for every heavy falling body there us a speed determinedby naturesuch that this cannot be
increased or diminished except by using force or opposing some impedimentto it' (Galilei
[1638/1989], p. 107); cf. also Drake's fn. 40 in Galilei [1638/1989] at (modernpagination)
p. 66.
406
TamarSzabo Gendler
way thatthe Aristoteliancan be forcedto a contradiction.Whatshe is forcedto
accept,however, areradicaldiscontinuitiesin nature.A body, united,mightbe
falling steadily at a rate of, say, six, and suddenly, should its partshappento
become unified, begin falling at a rate of, say, twelve.
But the Aristoteliancan avoid the problem of discontinuity.A fourthway
would be for her to say thatgiven two bodies thatfall togetherthereis a fact of
the matterabouttheir degree of connectedness,and that this determinestheir
physical propertieswhen falling. The claim would be:
(7) Naturalspeed andweight for strapped-bodiesaredeterminedby a degree
of connectedness(C) such thatthe speed/weightof Bl-strapped-to-B2where
B1 has wl and B2 has w2 will be: (C)(wl + w2) + (1 - C)((wl + w2)/2.l8
We let C measurethe degree of connectednessbetween the two bodies; thatis,
we let it be a numberbetween zero and one that correspondsto the degree to
which two bodies that fall together are unified:if the bodies are completely
unified,C will take a value of one; if the bodies are completely disunified(that
is, united),C will take a value of zero. For intermediatecases, the value will be
between these two, and the speed and weight of the combined body will lie
betweenthe mean andthe sum of the two initialvalues. So if the two bodies are
completely united, the additive law will apply completely; if the bodies are
merely unified,the mediative law will apply throughout;and for intermediate
cases, some proportionalaverage will be found between them. Thus the
assumptionthat naturalspeed is a function of weight can be maintained,and
it can be maintainedwithoutviolation of continuity.How? Underthe assumption thatdegree of connectednessis a relevantphysicalproperty.Thatthis way
out too seems not to be a live optionbringsus to the point where I will suggest
my alternativeexplanationof what is going on.
2.4 What the reconstruction misses
I want to begin by thinkingabout ways in which the four ways out might be
blocked. And I want to start with the most obvious: appeal to two broad,
defeasible, tacit assumptions,each of which capturesan importantfeatureof
our representationof experienced reality. One is that, for any body that one
might encounter,there is a determinatefact concerningits weight and natural
speed. That is:
(8) Naturalspeed and weight are physically determined.
18
To keep the equation minimally complicated, I have made a number of trivial simplifying
assumptions.I have assumedthat the units for measuringweight and naturalspeed correspond
so that the numberrepresentingan object's weight is the same as the numberrepresentingits
naturalspeed; and I have assumed that the naturalspeed of two merely unified bodies is the
mean of their individualnaturalspeeds.
GalileoandtheIndispensability
of ScientificThoughtExperiment407
The other is that there is no determinatefact whetherstrapped-bodiesare one
object or two. That is:
(9) Entificationis not physically determined.
What (8) says is that a particularquestion about natural properties has a
determinateanswer. Any body, no matter how oddly shaped, will have a
particularweight and a particularnaturalspeed that are fixed by the world.
What (9) says is that a particularquestionof entificationhas an indeterminate
answer.Whetherwe considera strapped-together
body to be a single object, or
two objects held togetherby a strap,or indefinitelymany objectsheld together
by internalforces, is merely a questionof the aspectunderwhich we choose to
view that object. The answer to the question 'how many objects?' does not
follow from anyphysical propertywe might discover;it is a questionaboutour
words, not a question aboutthe world.
These two premises are sufficient to eliminate the 'ways out' enumerated
above. If (8) holds, then (4) and (5) are not available as lines of escape; if (9)
holds, then neither (6) nor (7) can be appealedto as a means of avoiding the
Galilean conclusion. And the way in which they eliminate (4)-(7) is very
different from the way that a simple reassertionof (1) and (2) would. They
show not only thatthereis somethingwrongwith (4)-(7) as descriptionsof the
way the worldis, but whythereis somethingwrongwith them.They show what
aboutourtacitunderstandingof physicalreality,andof ourinstinctsconcerning
plausiblecandidatesfor physicallyrelevantand irrelevantproperties,is missed
by someone who appealsto (4) or (5) or (6) or even (7).
What this reveals is that the initial reconstructionof the Galilean thought
experiment(as presentedin Section 2.2) fails to capturewhat is really doing
the work in the case. As hypotheses about the ways strapped-bodiesmight
behave in fall, (4), (5), (6), and especially (7) are in principle available as
alternativesto (1) and (2); just as naturalspeed might be mediativeand weight
additive,it mightbe thatboth naturalspeed andweight dependon the degreeto
which the two bodies are connected.So if (1)-(3) were trulycapturingwhat is
going on in the Galileanthoughtexperiment,there would be ways out for the
Aristotelianthatwould allow her at least to shift the burdenof proofback to the
Galilean.
That these ways out do not seem available when the thoughtexperimentis
presentedin its unreconstructedform shows that this eliminative reconstruction has failed to captureits originaldemonstrativeforce. (Whathas been lost
is the way in which, by evoking tacit knowledge aboutthe how falling bodies
actually behave, the thought experimentpre-emptivelyprecludes such ways
out.) Accordingly, I tried to come up with a reconstructionsufficientlystrong
to rule out (4)-(7) in a similarly categorical and decisive manner. This
involved appeal to two rather comprehensive and metaphysical-sounding
TamarSzabo Gendler
408
principles,namely (8) and (9). (8) and (9) give backgroundsupportto (1) and
(2) and therebyhelp to establish (3).
But just as (1) and (2) are too weak to capturethe way in which alternative
hypothesesconcerningfall areruledout by the thoughtexperiment,(8) and (9)
aretoo strong.They representapproximatearticulationsof defeasible assumptions about the physical world. But as they stand, they articulateprinciples
that have less certaintythan the conclusion they are taken to support.Prior
to contemplation of the case Galileo describes, the Aristotelian may be
committed to somethinglike (8) and somethinglike (9), but he is certainly
not committed to them unmodified. To reconstructthe case as an explicit
argumentwith some version of (8) and (9) among the premiseswould require
enumeratingoutrighttheir defeasibility conditions. But this is something he
does not know how to do. Contemplationof the case Galileo describesbrings
him to see that these principles are not defeated in this case. And it is
this recognition that serves as the basis for the case's power. No austere
argumentativereconstructionwill be able to do this, because part of the
thought experiment's function is to bring the Aristotelian to accept certain
premises.Inthenextsection,
I willdiscusswhatmakesbeliefinthesepremises
new,and what makes it justified.
3 Denying the Dispensability and Derivativity Theses
3.1 Rejecting reconstruction:what the thoughtexperimentdoes
If the DispensabilityThesis were correct,then the conclusion establishedby
Galileo's thoughtexperiment that 'bothgreatandsmall bodies . . . aremoved
with like speeds' (Galilei [1638/1989], p. 109) shouldbe demonstrableby
means of a non-thought-experimental
argument.'Demonstrable'here means:
rationallyjustifiableon the basis of the same backgroundconditions.So let us
spell out what the backgroundconditions in question are.
For the Aristotelian, daily experience seems to confirm the theory that
heavier bodies fall faster than lighter ones. Just as the Galilean sees cases
where lighter bodies fall more slowly than heavy ones as exceptional, so the
Aristoteliansees as cryingout for explanationthose cases wherethe rateof fall
is (nearly) equal. That is, when gold beaten into a very thin leaf reaches the
groundmore slowly than a solid lump of the same material,the Galileanmust
posit some factorthatexplains the divergenceof this resultfrom the generally
expected outcome.l9 Similarly,when the Aristoteliansees two stones of very
differentweights fall to the groundwith like speeds, the circumstancerequires
diagnosis and explanation.
lg Cf. Galilei [1638/1989], p. 109.
of ScientificThoughtExperiment409
GalileoandtheIndispensability
So far all I have pointed out is the possibility of maintainingtheoretical
commitmentsin the face of apparentcounter-evidence.This can be done by
appealingto additionalprinciplesthatexplainaway anomalousdataby showing
thatthe phenomenaat issue are subjectto the fundamentalprinciplein question,
butthatthe world'scomplexityhas preventedthemfrommanifestingthis. So the
Galileanmightappealto airresistance,the Aristotelianto the fact thatthe bodies
have not been droppedfroma height sufficientlygreat.20Such explaining-away
of recalcitrantexceptionsis not a desperatemove by a failing paradigm;it is a
fundamentalelementof doing normalscience in a non-idealworld.2l
The point of this discussion is to give a better sense of the background
conditionsunderwhich the argumentativereconstructionmust be demonstratively forceful if the DispensabilityThesis is to be shown to hold in this case.
The thesis tells us that some non-thought-experimentalreconstructionof the
case Galileo presents should be able to do the same thing that the thoughtexperimentalversion does: lead the Aristotelianfrom the same background
assumptionsto the same rationallyjustified conclusions. The standardreconstructionpresentedin Section 2.2 fails in this regard,as does the strengthened
version presentedin Section 2.4. And, I contend,any argumentsatisfying the
EliminationThesis is likely fail in the same way.
Why? Because priorto the thoughtexperiment,the Aristotelianis explicitly
committedto the negation of (3), and this backgroundcommitmentserves as a
filterthroughwhichapparentlycontraryevidencewill inevitablybe reinterpreted.
Any argumentwhich satisfiesthe EliminationThesis (thatis, any argumentwith
explicitpremiseswhich makeno referenceto particularhypotheticalor counterfactualstatesof affairs)can be reframedby the Aristotelianas a reductio.(1) or
(2) or (8) or (9) will simplybe denied,once theirimplicationsaremadeevident.
What is remarkableabout the thought-experimentalpresentationis that it is
able to underminethis framework-shapingassumption.
20
21
In Galileo's dialogue, when Salviati points out that a rock of two poundsand a rock of twenty
poundswill strikethe groundnearly simultaneouslywhen droppedfrom a height of one or two
hundredfeet, Simplicio retortsthat this may just be a result of not having given the objects
enough falling-timefor the differencesto become apparent(see Galilei [1638/1989], pp. 10910). Simplicio says: 'Perhapsfrom very greatheights, of thousandsof bracchia[discrepancies]
would follow which [are] not seen at these lesser heights.' That is, Simplicio suggests that
perhapsthe reasonwe do not observethe sortsof differenceswhich Aristotle's theorypredictsis
that we are making observations under non-idealized circumstances,and that were we to
eliminate distortingelements-like the fact that the fall is only a few hundredfeet the true
phenomena would reveal themselves. Galileo provides Salviati with a rather sharp retort
(Galilei [1638/1989], p. 110). See also (paginationhere refers to the modern text) Galilei
[1638/1989], fn. 44 at p. 69.
Of course, it is a commonplacein the history of science that at a certainpoint the burdenof
positing literal or metaphoricepicycles becomes too great, and the theory especially when
thereis a simpleralternativeavailable collapses underthe weight of its own internalcomplexity. But it is a similarly well-establishedcommonplacein philosophy that theories are underdetermined by evidence, and that extra-scientific considerations do some of the work in
determiningtheory choice.
410
TamarSzabo Gendler
So I conclude, tentatively,that the Dispensability
Thesis (and a fortiori the
DerivativityThesis) is false. Suppose,however,that
somehow it were possible
to come up with an argumentativereconstruction
that almost exactly captures
the strengthand limits I have attributedto
Galileo's thought experiment.22
Would we then be justified in accepting the
deeper methodological claim
put forth in the Derivativity Thesis: that the
justificatory
force of whatever
beliefs I hold via thoughtexperimentis a functionof
the thoughtexperiment's
argumentationalessence? In the next section (3.2), I offer
reasonsfor thinking
that the answerto this question is 'no'.
3.2 Rejecting the positive argument: what
makes these beliefs
new?
TheDerivativityThesis says thatthe justificatory
force of any good scientific
thoughtexperimentcan only be explained by the fact
that it can be replaced,
withoutloss of demonstrativeforce, by a
non-thought-experimental
argument.
So rejection of the thesis can take two forms:
denying that the justificatory
forceof a particularscientific thoughtexperiment
can be explained this way,
anddenying thatit can only be explainedthis way.
The simplestway of doing
theformer, of course, would be to show that
the Dispensability Thesis is
nottrue of some thought experiment;obviously,
if there is no non-thoughtexperimentalargumentwith which the thought experiment
can be replaced
withoutloss of demonstrativeforce, then the
justificatoryforce of the thought
experiment
cannotbe explainedby the possibility of such
replacement.But for
thesake of argument,I am supposingthatthe
DispensabilityThesis is (at least
approximately)
true. This leaves two avenues for denying the
Derivativity
Thesis:denying that the argumentativereconstruction
explains the justificatory
force of some thoughtexperimentat all, and denying
thatit explains such
justificatory
force entirely.
As I discussed above (Section 1.3), two sorts of
defence are offered for the
Derivativity
Thesis. The negative argumentcontendsthatthought
experiments
arearguments because there is nothing else for
them to be; the positive
argument
contends that thought experiments are arguments
because their
'analysisand appraisal' involves explicit
argumentativereconstruction.In
this
section I will addressthe positive thesis, suggesting
that it fails to get at
what
is most interesting about thought-experimental
reasoning; in the final
section,
I offer some thoughtsaboutwhat sorts of
alternativejustificationsare
available
such that the negative thesis too is untenable.
It is a mistake, I contend, to think that the
reason conclusions to thought
Suppose,
for instance,we were to enumeratethe
defeasibilityconditionsof (8) and (9), and we
includedthese modified versions as premises alongside
(1) and (2).
22
of ScientificThoughtExperiment411
GalileoandtheIndispensability
experimentsarejustifiedis because thoughtexperimentshave argumentational
analogues (if indeed they do). Rather,I want to suggest that the Aristotelian
comes to have novel justified true beliefs about the empincal world not
because he has (whetherhe knows it or not) followed the path of a recognized
argumentform,but ratherbecause he has performedan act of introspectionthat
beings to light heretofore inarticulatedand (because he lacked a theoretical
framework in which to make sense of them) heretofore implausible tacit
beliefs. There are two things that I need to show myself able to explain. The
first is how it is that knowledge has been gained. In what way is it that the
Aristotelianhas come to believe somethingnew? The second is how it is that
knowledgehas been gained. In what way is it thatthe Anstotelian has come to
believe somethingjustified? I will answer the first in the remainderof this
section, and the second in Section 3.3.
In addressingthe issue of novelty, I will begin with brief remarksaboutwhat
I thinkis not at issue, andthen say somethingaboutwhereI thinkthe important
questionsrest. One might say thatthe beliefs are not new since, in some sense,
the Aristotelianhad access to them before. After all, he has acquiredno new
knowledge of the externalworld; all he has done is reshuffle tacit beliefs he
alreadyheld, coming to see their implications. But this view of what makes
knowledge new is surely too stnngent;it would, among other things, rule out
all mathematicalreasoning as a potential source of new knowledge. On the
otherhand,it seems too weak to say thata belief is new if it merelyresultsfrom
puttingtogethertwo explicitly held beliefs thathave not, for the individualin
question, been previously connected. If I believe that snow is white and I
believe thatcrows areblack, but I have neverthoughtaboutthe two at the same
time, it seems wrong to say that the belief that snow-is-white-and-crows-areblack should count as a new belief for me.23In any case, without spelling out
precisely what it is thatmakes some beliefs new and othersmere implications,
there is a simple reason to think that the Aristotelian's belief that the speed at
which a body falls is independentof its weight shouldcount as a new belief for
him: Until recently he was explicitly committed to the truthof its negation.
This alone suggests that whateverthe implicatoryrelationbetween his pnor
commitmentsand this view about naturalspeed the belief should count as
new.
But thereis a deeperandmoreinterestingway thatthe belief is new, andthat
23
Except undervery odd circumstances.SupposeI believe thatcrows areblack because I live in a
village where thereare crows, and I have seen many of them. I also believe that snow is white,
because I have read aboutit in books. In my village, it is taboo to think aboutblack things and
white things at the same time, because this is thoughtto allow the evil spiritaccess to the soul.
One day, I leave my village for the north,and I observe a crow circling above a field of snow. I
findthe image aestheticallystriking,andI ask myself why this is so. In analysingmy responseto
the visual experienceI am having,I realize with a start:'Snow is white andcrows areblack.' In
such a case, it seems plausible to suggest that this is a new belief.
412
TamarSzabo Gendler
is the following. The thought experiment
that Galileo presents leads the
Aristotelian to a reconfigurationof his conceptual
commitments of a kind
thatlets him see familiarphenomenain a novel
way. Whatthe Galileandoes is
provide the Aristotelianwith conceptualspace for a
new notion of the kind of
thing naturalspeed might be: an independently
ascertainableconstantrather
thana functionof somethingmoreprimitive(that
is, ratherthanas a functionof
weight). It is in this way, by allowing the
Aristotelian to make sense of a
previouslyincomprehensibleconcept, thatthe thought
experimenthas led him
to a belief that is properlytaken as new.
Whatthis suggestsis thatthe DerivativityThesis is
missingthepoint of what
makes the Galileo case work as it does. The
recognitionthat naturalspeed is
independentof weight comes not from tracingthe
implicationsof antecedent
commitmentsto (1) and (2), which, afterall, lead to the
denial of a position to
whichthe Aristotelianis explicitly committed(and
thence to retreatssuch as
thefour ways out). The recognitioncomes from
the suddenrealization,on the
partof the Aristotelian, of the conceptual
possibility of a certain sort of
physicalproperty.Prior to contemplationof the case,
there was no room on
theAristotelianpicturefor the thoughtthatnatural
speed mightbe constant,not
varying thatit mightbe dependentnot on some
specificfeaturesof the body in
question,but only on the fact thatit is a body at all.24
Aftercontemplationof the
case,thereseems to be no conceptualspacefor the
view thatit mightbe variable.
Tempting as it might be to digress, this is not the
place for a general
discussionof incommensurabilityacross
theory-change.What is important
forour purposesis only this: one of the things
that enables this ratherstriking
shiftin the representationof physical realityis that
the Aristotelianrecognizes
that
there are experientiallypossible objects
strapped-togetherbodies for
which
the defeasibilityconditionsof (8) and (9) are not
met (thatis, objects of
which
(8) and (9) hold true),andthatthese are objects
for which his old notion
ofnaturalspeed simply does not make sense.
If entificationis arbitraryand
natural
speed and weight are fixed by the world, then a
feature-dependent
notion
of naturalspeedis just plainincoherent.So one
way of thinkingabouthow
the
thoughtexperimentworksis this: it bringsthe
Aristotelianto recognizethe
inadequacy
of his conceptualframeworkfor dealingwith
phenomenawhichthrough
the contemplationof this imaginarycase he
comes to recognize as
always
havingbeen partof his world.
To get a sense of how odd this transitionis, try
thinkingaboutweight as somethingdependent
not on the specifics of the body in question,but as
somethingconstantfor all bodies. (Thatis, to
ascertaina body's weight, we would not need to know
anythingmore about it that the simple
factthatit is a body.) Clearlythis would be a
majorconceptualreadjustment;one mighteven be
inclinedto say thatwe aren't talkingaboutweight any
more,since whateversortof thingweight
is, it is surely somethingthat dependsupon
specific featuresof the bodies to which it applies.
Theanalogy is not perfect, since partof what
happensas a result of thinkingaboutthe Galileo
caseis thatit becomes apparentthatthereis no
physical applicationfor the Aristotelianidea of
naturalspeed; like phlogiston,it disappearsinto the
ether of abandonedconcepts.
24
of ScientificThoughtExperiment413
GalileoandtheIndispensability
Whatthis suggestsis that 'theanalysisandappraisalof a thoughtexpenment'
need not 'involve reconstructingit explicitly as an argument',in such a way
that 'a good thoughtexpenment is a good argument,a bad thoughtexperiment
is a bad argument' (Norton [1991], p. 129; [1996], p. 335). After all, the
argumentfrom (1) and (2) to (3) is no betteror worse thanthe argumentfrom
(not-3) to (not-l) or (not-2). Like an experiment,part of what makes a thought
experiment good or bad is the validity of the procedureby which the same
result call be repeatedly obtained. But anotherthing that distinguishes good
thoughtexpenments from bad is their ability to directthe reader's attentionto
inadequaciesin her conceptualschemethatshe herselfrecognizesimmediately,
as soon as they arepointedout to her.It is this,I wantto suggest,thatgroundsher
new beliefs. Of course,I have said nothingso far aboutwhat might make these
beliefsjustified.It is to this issue thatI turnin the Section 3.3.
3.3 Rejecting the negative argument: what makes these beliefs
knowledge?
Thoughtexperimentswork in a vanety of ways. By descnbing appropnately
selected imaginaryscenanos, they providecontexts within which sense can be
made of previouslyincomprehensibleconceptualdistinctions.25This happens
when two featuresthat are constantly conjoined in our representationsof all
actualcases are imaginativelyseparatedin the thought-expenmentalscenano
in a way that shows them to have been isolatableall along. And by descnbing
specific situations,thoughtexperiments,like analogical reasoningin general,
can justify conclusions about particularcases without explinit or implicit
appeal to more general absolute pnnciples.26 Many of the higher-level
principlesby whichwe negotiatethe worldaredefeasible,andthe determination
of their applicabilityto particularsituationsmust be made on a case-by-case
basis. By bringingthe readerto focus on particulars,thoughtexpenments can
help the reader distinguish warrantedfrom unwarrantedapplicationsof the
principlein question.27
After all, as Norton
So far, however, this has little to do with justification.
would ask, if the thoughtexpenment is not an argument,why should we put
faith in its conclusion? What I want to explore in this closing section is one
possible answer:thatthoughtexperimentsrely on a certainsortof constructive
25
26
27
By this I mean they do something like what the answer to a riddle does in making suddenly
intelligible what previously appearedto be a nonsensical description.Cf. Cavell [1979], pp.
156-7.
For an articulationand defence of such a view of analogicalreasoning,cf. Sunstein[1993] and
[1996].
This often happenswhen the particularsare sufficientlywell-sketchedto evoke practicalas well
as theoreticalresponses.Cf. also the literatureon mentalmodelling,referencesto which can be
found in Nersessian [1993], as well as at Thomas [undated].
414
TamarSzabo Gendler
participation on the part of the reader,and that the justificatoryforce of the
thoughtexperimentactuallycomes fromthe fact thatit calls uponthe readerto
performwhat I will call an experiment-in-thought.
An experiment-in-thoughtis an actual experiment;the person conducting
the expenmentasksherself:'WhatwouldI say/judge/expectwere I to encounter
circumstancesXYZ?' and thenfindsout the (apparent)answer.This technique
is common in linguistics, where the methodology is used to ascertain
the grammaticalityof sentences, the meanings of phrases, the taxonomic
categories of words, and so on.28 And it is, on one view at least, a central
element of moral reasoning: we think about particular imaginary cases,
observe the judgements that they evoke in us, and use these judgements as
fixed points in developing our moral theones.29
How does this connect with the Galileo case? Whatkind of experiment-inthoughtplays a role there?Answer:by thinkingaboutthe case in question,we
discover what sorts of motions and objects we thinkare possible in the world.
Do we thinkobjects can be strappedtogether?Yes, we do. Do we thinkobjects
fall with radical discontinuitiesin speed? No, we think they do not. Do we
thinkentificationis somethingthatis fixed by the world?No, we do not. Do we
thinkweight andnaturalspeed are fixed by the world?Yes, we do. We come to
recognize that we have these beliefs by contemplatingthe imaginarycase in
question;thinkingabout the case is what brings us to the realizationthat we
believe what we do. And-and this is where thejustificatoryworkcomes in
the fact that we have these beliefs gives us primafacie warrantto think that
they are true.
But why? Why should we think that our pre-theoreticalbeliefs about
the structureof the physical world are reliable?In the Science of Mechanics,
a few pages after coining the expression Gedankenexperiment,
Mach writes:
Everything which we observe imprints itself uncomprehended
and
unanalyzedin our percepts and ideas, which then, in their turn,
mimic the process of nature in their most general and most stnking
features. In these accumulatedexperiences we posses a treasure-store
which is ever close at hand, and of which only the smallest portion is
embodied in clear articulate thought. The circumstance that it is far
easier to resort to these experiences than it is to nature herself, and
that they are, notwithstandingthis, free, in the sense indicated, from
all subjectivity, invests them with high value (Mach [1883/1960],
p. 36).
28
29
Cf. Thomason[1991], p. 247: 'When linguists want to test hypothesesaboutthe structureof a
particularlanguage, their methodology crucially involves thoughtexperimentsin a . . . literal
sense: real experimentscarriedout in thinking.'
Cf., to choose an example nearlyat random,Thomson [1986], p. 257: 'it is . . . our moralviews
aboutexamples,stories,andcases whichconstitute. . . datafor moraltheorizing.'I do not intend
to be taking a standhere on questionsof moralepistemology.
GalileoandtheIndispensability
of ScientificThoughtExperiment415
So one possible explanation is the one Mach gives. We have stores of
unarticulatedknowledge of the world which is not organized under any
theoreticalframework.Argumentwill not give us access to that knowledge,
because the knowledge is not propositionally available. Framed properly,
however, a thought experimentcan tap into it, and-much like an ordinary
experiment allow us to make use of informationaboutthe world which was,
in some sense, thereall along, if only we had known how to systematizeit into
patternsof which we are able to make sense.30
This, of course,is the beginningnot the end of an answerto the question.But
it is sufficient for the modest aim of this article. What I have been trying to
show is that somethingbesides argumentmight give justificatory force to
thought-experimentalreasoning. The alternative I have proposed has been
this: by focusing on imaginaryscenarios and makingreferenceto particulars,
thoughtexperimentscan providea fulcrumfor the reorganizationof conceptual
commitments;this explains the way in which they can provide us with novel
inforrnation
without empiricalinput. And by bringing the readerto perform
experimentsin thought,thoughtexperimentscan lead us to reject shaky (and
ultimatelyfalse) theoreticalcommitmentsin light of newly systematizedbut
previouslyinarticulableknowledge
aboutthe way the world is.
The justificatoryforce of thoughtexperimentsis thus parasiticon the extent
to which the messy twisted web of backgroundbeliefs that underpin our
navigation of the world are rightly consideredknowledge. To establish this,
on coherentistor evolutionaryor empiricist grounds,would be an enormous
undertaking,and one which I will not even begin to endeavourhere.3l But I
hope I have given you some sense, at least, of why I am not convincedthateven
if the Dispensability Thesis is true, the Derivativity Thesis must be true as
well. For even if it couldbe replaced by an equally effective argument,the
justificatory
force
of a thoughtexperimentmight still be based on its capacity
to make availablein a theoreticalway those tacit practicalcommitmentswhich
enable us to negotiate the physical world.
3.4 Constructivismand the contrastwith Nortonand Brown
Before concluding, let me try to articulatemore clearly how the position I
have been advocating differs from those of James Robert Brown and John
Norton, both of whom offer detailed but conflicting discussions of precisely
this thought experiment.32Brown's discussions of the case can be found
30
31
32
For furtherdiscussion of these themes, cf. Kuhn [1964].
Forone such (evolutionarilybased) attempt,cf. Shepard[1994]; a similarexplanationis offered
in Sorensen [1992a]
I thanka non-anonymousrefereefor BJPS JohnNorton,in fact-for suggestingthatI include
such a section.
416
TamarSzabo Gendler
throughouthis works,33but I will focus on his most
explicit version, which
appearsin his book (Brown [199la]), and which is
framedpartlyas a critique
of Norton. Norton's own discussion of the case,
which is framedpartly as a
critiqueof Brown, can be found in his 1996 article
(Norton [1996]).
Brown contends that the Galileo thought
experiment is what he calls a
platonic thoughtexperiment,that is 'a single thought
experiment which
destroys an old or existing theory and simultaneously
generates a new one'
(Brown[199la], p. 77, cf. pp. 43-5). Accordingto
Brown,thoughtexperiments
such as Galileo's give us 'aprioriknowledgeof
nature'(Brown[199la], p. 77);
throughconsiderationof such thoughtexperiments,
Brown maintains,we gain
directquasi-perceptualknowledgeof relations
betweenindependentlyexisting
abstractentities (Brown [199la], p. 76). In the case
of Galileo's thought
experiment,the platonic law of natureto which we gain
access is that 'all
bodiesfall at the same rate'.34
Bracketingquestions concerning the ontological status
of laws of nature,
Brown's basic contention is that Galileo's thought
experiment brings the
heretoforemistaken Aristotelianto see the truthof a
certainplatonic law of
nature,and that it does so neither(a) by introducing
the Aristotelianto novel
empiricaldata, nor (b) by showing the law to be a
logical truth,nor (c) by
makingminimal modificationsto the Aristotelian's
earliertheory, nor (d) by
derivingthat law from antecedentcommitmentswhich
the Aristotelianis in a
positionto maintain35.Of these, (a) and (b) are
apparentlyuncontroversial,36
andBrown argues for (c) whose truthI also
accept for reasons enumerated
inSection 2.3 above by contending that
the post-thought-experimental
justificationfor belief in the Galilean theory is higher
than pre-thoughtexperimental
justificationfor belief in its Aristoteliancounterpart.The
crucial
claim,then, is (d): that the Aristotelian cannot be
brought to the Galilean
conclusion
by argumentalone.
It is precisely this, of course, which Norton
denies. So before returningto
Brown'sown positive account of the thought
experiment, let us consider
Norton'
s proposedargumentativereconstructionof the
case.37Nortonsuggests
that
Galileo'sreasoningis as follows (Norton1996, pp.
341-2). (The comments
in
bracketsthatfollow each of Norton's steps are mine.)
Cf. Brown [199la, b, 1993b, 1995]. In [199la] he
speaksof this as 'the best (i.e. my favourite)
[thoughtexperiment]'( p. 1).
34
ForreasonsthatGalileo himself addresses,this law
is trueonly of bodies falling in a vacuum;cf.
Galilei [1638/1989], p. 109, which is discussed above.
35Cf.
(1), (2), (3) in Brown [199la], pp. 77-9.
36Concerning
(a), it is clear thatthe thoughtexperimentinvolves
no new sensoryinput,noris the
Aristotelianprovided with novel data concerningthe
behavior of objects in the world. Concerning(b), it is not logically impossible (simpliciter)
thatthe speed at which bodies fall should
dependon some non-universalfeature(s).
37Norton
writes: 'there can be little question that what
Galileo gives us here is simply an
argument'(Norton [1996], p. 341).
33
of ScientificThoughtExperiment417
GalileoandtheIndispensability
(N1) Assumption [for reductioproop: the speed of fall of bodies in a given
medium is proportionateto their weights.
[This is equivalentto the negation of (3) in my reconstructionin Section 2.1;
the bracketsaround'for reductioproof' are my own .]
(N2) From (N1): if a large stone falls with 8 degrees of speed, a smaller stone
half its weight will fall with 4 degrees of speed.
[This is an application of the (N1)-law to a particularcase. On my understandingof the EliminationThesis (cf. Section 1.2 above), invocation of this
particular makes Norton's reconstruction-strictly speaking a thoughtexperimentalargument.But since his reconstructionrelies essentially only
on the incompatibilityof (N1), (N3), and (N5), I will neglect this ad hominem
point. The readershould be careful, however, not to focus on (N2), (N4), and
(N6) in thinkingabout the forcefulness of Norton's reconstruction.]
(N3) Assumption:if a slower falling stone is connectedto a fasterfalling stone,
the slower will retardthe faster and the faster speed the slower.
[This is equivalentto (1)
tion in Section 2.1.]
thatnaturalspeed is mediative in my reconstruc-
(N4) FromN3: if the two stones of (N2) areconnected,theircompositewill fall
slower than 8 degrees of speed.
[This is an applicationof the (N3)-law to the example presentedin (N2).]
(N5) Assumption:the compositeof the two weights has greaterweight thanthe
larger.
[This is equivalent to (2)
Section 2.1.]
that weight is additive- in my reconstructionin
(N6) From (N1) and (N5): The composite will fall faster than 8 degrees.
[This is an applicationof the (N1)-law to the example presentedin (N2) as
governedby the (N5)-law.]
(N7) Conclusions (N4) and (N6) contradict.
[As the reconstructionin Section 2.1 demonstrates,a parallel result can be
obtainedfrom (N1), (N3), and (N5) alone, withoutreferenceto the particular
applicationsthat generate (N4) and (N6).]
(N8) Therefore,we must reject assumption(N1).
[Since (N7) is derivedfrom (N1), (N3) and (N5), then one of these threemust
be rejected.Nortontakes (N3) thatnaturalspeed is mediative- and (N5)
that weight is additive as given, so the only option is to reject (N1). As my
418
TamarSzabo Gendler
discussion in Section 2.3 shows, however, thereare ways of denying (N3) and
(N5) which are at least primafacie not completely outrageous.]
Norton then goes on to discuss a final step, which he attributesto Brown:
(N9) Thereforeall stones fall alike.
[Norton attributesthis conclusion to Brown in part on the assumptionthat
Brown accepts that (N8) follows straightforwardlyfrom (N1)-(N7). For
reasons that I discuss below, I think this is a misreadingof Brown's view.
But in any case, as both Norton and Brown recognize, (N9) does not follow
from (N8) withoutadditionalpremises.All that(N8) tells us is thatthe speed of
fall of bodies in a given medium is not proportionateto their weights (thatis,
the negationof (N1)); it has no bearingon the issue of whetherit might depend
on some other factor.38Galileo himself is trying to show that 'if one were to
remove entirely the resistance of the medium, all materials would descend
with equal speed' (Galilei [1638/1989], p. 116), but since precisely what is
at issue in the section as a whole is the possibility and natureof motion in a
void, this idealization is not permitted as one of the assumptions in the
reconstruction.]
Clearly,the move from (N8) to (N9) dependsupon an auxiliaryassumption:
(N8a) Assumption:the speed of fall of bodies dependsonly on their weights.
If (N8a) is accepted,then (N9) follows from (N3) and (N5),39and if (N8a) is
not accepted, then (N9) does not. But, Norton notes, Galileo's charactersare
not in a positionto accept (N8a), nor, if we areconsideringthe fall of bodies in
media, arewe. Forthe speed at which a body falls in mediais dependentin part
on its shape: gold leaf falls more slowly thana gold nugget of the same weight
(cf. Galilei [1638/1989], p. 109), just as a closed parachutefalls more quickly
than an open one. So if we wish to denve (N9) from (N8), we will need to
appeal to (N8a).
In short,Nortonconcludes: 'ourdegree of belief in the final outcome of the
thoughtexperimentturnsoutto dependon ourdegreeof belief in the assumption'
(N8a). How so? Well, 'the transitionfrom (N8) to (N9) is, at worst,a fallacious
inferenceto a falsehood [if (N8a) is denied];or, at best, valid only insofaras it
is invoked in special cases in which assumption[(N8a)] holds, such as the fall
of very heavy compact objects in very raremedia. This final step looks more
like a clumsy fudge or a stumblethan a leap into the Platonic world of laws'
(Norton 1996, p. 345). And this outcome, Norton contends, supports the
previously-citedthesis that 'a good thoughtexperimentis a good argument,a
38
39
Cf. Brown:it is not a logical contradictionto supposethat 'bodies mightfall at differentspeeds
dependingon their colours or on their chemical composition' (Brown [199la], p. 78).
Norton presentsthis argumentas a second reductio;cf. Norton [1996], p. 343.
of ScientificThoughtExperiment419
GalileoandtheIndispensability
bad thought experimentis a bad argument'(Norton [1991], p. 131; [1996],
p. 335). For the strengthof the thought-experimentalconclusion correlates
directlywith the strengthof the correspondingargument.
Now, I think the discussion of (N8a) turnsout to be a bit of a red herringin
the dispute between Brown and Norton because, as far as I can tell, Brown's
concession that 'the thoughtexperimentis an argumentinsofar as it destroys
the Aristoteliantheory' (Norton [1996], p. 342; cf. Brown [199la], p. 78) is
not- or at least should not be a concession by Brown that (N8) follows
logically from (N1)-(N7). Like any reductio,all that the argumentshows is
thatat least one of the premisesneeds to be rejected;it does not show us which
premise.(Thepointof my discussionin Section 2.3 was to show preciselythis.)
So Brownneed not concede as Nortontakeshim to that(N8) follows from
(N1) (bracketsomitted)-(N7). But, of course,Brown is committedto the view
that (N8) is true;indeed, Brown is committedto the view that (N9) is true. So
how does Brown think that we know (N8) and (N9), if not by argumentfrom
(N1)-(N7) and (N8a)?
Brown's answer,in so far as I understandit, is that thinkingabout the case
forces us (or the Aristotelian)to use our platonic perceptualcapacities to see
that (N9) is a platonic law of nature(that is, a connection between independently existing abstractentities). But since it seems to me thata perfectlygood
explanationof what is going on can be offered without appeal to such laws,
ontological parsimony suggests that Norton may be partly wrong without
Brown being fully right. So let me say what I think is right and wrong in
both views as a way of putting my own in context. To give my view a label
that contrastsit with Norton's self-proclaimedempiricismand Brown's self(Kantian
proclaimedplatonism, let me call my own position constructivism
associations intended).
I agree with (my readingof) Brownthat(N1)-(N7) does not get us (N8); the
reductiotells us that somethingis wrong with the Aristotelianpicture,but it
does not tell us whatis wrong.And I agreewith Nortonthatsomethingis being
left out of Brown's storyby his neglect of (N8a); indeed, I don't thinkthatthe
thoughtexperimentin questionshows anythingmorethanthatnaturalspeed is
independentof weight. But what my argumentin Section 2.4 aims to show is
that the sorts of considerationsthat allow the conclusion of (N8) from (N1)(N7) are not available to the Aristotelianas premises before she has followed
Galileo's instructionsfor guided contemplation.It is by focusing on some
imaginaxyparticularstrapped-bodythatshe comes to realize her non-defeated
defeasible commitmentsto my own (8) (the physical determinationof natural
speed and weight) and (9) (the lack of physical determinationof entification).
And thereinlies my answer to both Norton and Brown.
Now, if Norton wants to call this sort of guided contemplationargument,
and Brown wants to call it platonicintuition,then we're all in agreement.I
420
TamarSzabo Gendler
prefer to describe it as a reconfigurationof internal
conceptual space. By
introducing novel categories by which we make sense of
the world, this
reconfigurationallows us to recognize the significance of certain
previously
unsystematizedbeliefs. For the reasons that I discussed in Section
3.2, these
beliefs arenew; andfor the reasonsI discussedin Section
3.3, they arejustified.
Thus the thoughtexperimentbrings us to new
knowledge of the world, and it
does so by means of non-argumentative,non-platonic,
guided contemplation
of a particularscenario.
4 Conclusion
In this article, I have offered reasons for thinking
that a certain view about
thoughtexperimentsin science is false. The view is thatany
scientificthought
experimentcan be reconstructedas a non-thought-experimental
argumentwithout loss of demonstrativeforce.40In Section 1, I
explainedthe philosophical
motivationsfor adoptingsuch a view, and distinguishedtwo
versions of the
position.The first the DispensabilityThesis concernsthe
replaceabilityof
thoughtexperiments;the second the Derivativity Thesis
concerns their
justificatory
force. In the remainderof the article,I offeredreasonsfor
thinking
thatboth of these theses are false. Througha detailed
discussion of a thought
experimentof Galileo's, I tried to show that the standard
argumentative
reconstructionof the case fails to capture its justificatory
power, and I
suggested reasons to think that any other argumentative
reconstruction
wouldbe likely to fail in similar ways. I then argued
that even if one were
toprovide an argumentativereconstructionthat did
almost perfectly capture
thethought experiment's demonstrativeforce, this
would not show that the
reason the thought experiment is successful
is because, deep down, it is
nothingmore than an argumentin disguise. I suggested
that, to the contrary,
thesuccess of the thoughtexperimentmay be a result
of the way in which it
invitesthe reader'sconstructiveparticipation,depicts
particularsin ways that
makemanifest practical knowledge, and describes an
imaginary scenario
wherein
relevant features can be separatedfrom those that are
inessential to
the
question at issue.
Acknowledgements
For
commentson previousdraftsof this article,I am gratefulto
RichardBoyd,
Michael
Glanzberg,Steven Gross, Ned Hall, the late Thomas Kuhn,
Norman
Kretzmann,
John Murdoch, Robert Nozick, Hilary Putnam, Derek
Parfit,
In Section 3.4 I also briefly offered reasons for
rejecting another alternative that thought
experimentsoffer us special access to platonic laws of nature.
40
GalileoandtheIndispensability
of ScientiflcThoughtExperiment421
Shem Roush, Simon Saunders,Roger Shepard,Alison Simmons,Zoltan
GendlerSzabo,andaudiencesat HarvardandCornellUniversities.
Departmentof Philosophy
Syracuse University
Syracuse,NY13244-1170
USA
References
Bealer, G. [1998]: 'Intuition and the Autonomy of Philosophy', in M. DePaul
and W. Ramsey (eds), RethinkingIntuition, Savage, MD, Rowman and Littlefield
Publishers(in press).
Brown,J.R. [1986]: 'ThoughtExperimentsSince the ScientificRevolution',International
Studiesin the Philosophyof Science, 1, pp. 1-14.
Brown,J. R. [199la]: TheLaboratoryof the Mind: ThoughtExperimentsin the Natural
Sciences, New York and London, Routledge.
Brown,J. R. [199lb]: 'ThoughtExperiments:A PlatonicAccount', in T. Horowitzand
G. Massey [1991], pp. 119-28.
Brown, J. R. [1993a]: 'Author's Response (to Norton [1993])', Metascience (new
series), 3, pp. 38-40.
Brown,J. R. [1993b]: 'WhyEmpiricismWon't Work',in D. Hull et al. [1993], pp.2719.
Brown,J. R. [1995]: 'CriticalNotice: Roy Sorensen's ThoughtExperiments',Canadian
Journal of Philosophy, 25, pp. 135-42.
Butts, R. E. and Pitt, J. C. (eds) [1978]: New Perspectives on Galileo, Dordrechtand
Boston, Reidel.
Cargile,J. [1987]: 'Definitionsand Counterexamples',Philosophy, 62, pp. 179-93.
Carrier,M. [1993]: 'Review of Horowitz and Massey (eds), ThoughtExperimentsin
Science and Philosophy', Erkenntnis,39, pp. 413-19.
Cavell, S. [1979]: The Claim of Reason: Wittgenstein,Skepticism, Morality and
Tragedy,New York and Oxford,ClarendonPress.
Claggett,M. [1959]: TheScience of Mechanicsin theMiddleAges, Madison,University
of Wisconsin Press.
Clavelin, M. [1974]: The Natural Philosophy of Galileo: Essays on the Origins and
Formation of Classical Mechanics, Cambridge, MA, MIT Press, trans.
A. J. Pomerans.
Clement,J. [1983]: 'A ConceptualModel Discussed by Galileu andUsed Intuitivelyby
Physics Students',in D. Gentnerand A. L. Stevens (eds), Mental Models, Hillsdale,
NJ, LawrenceErlbaumAssociates, pp. 325-40.
Cooper, L. [1935]: Aristotle, Galileo, and the Tower of Pisa, Ithaca, NY, Cornell
University Press.
Damerow,P. et al. [1992]: Exploringthe Limitsof Pre-classical Mechanics:A Studyof
Conceptual Development in Early Modern Science: Free Fall and Compounded
422
TamarSzabo Gendler
Motionin the Workof Descartes,Galileoand Beeckman,New York, SpringerVerlag.
Dijksterhuis, E. J. [1961, repr. 1986]: The Mechanization
of the WorldPicture:
Pythagoras
toNewton,Princeton,NJ, PrincetonUniversityPress,trans.C. Dikshoorn.
Drake,S. [1978]: Galileoat Work:
HisScientificBiography,
Chicago.IL, Universityof
Chicago Press.
Drake,S. [1989]:HistoryofFreeFall:Aristotle
toGalileo,Toronto,Wall andThompson.
Drake, S. [1990]: Galileo:PioneerScientist,Toronto,University of TorontoPress.
Drake S. and Drabkin,I. E. [1969]: Mechanicsin SixteenthCentury
Italy:Selections
from Tartaglia,Benedetti,Guido Ubaldo,Galileo,Madison, WI, University of
Wisconsin Press.
Galilei, G. [1590/1960]: OnMotion,Madison,WI, Universityof WisconsinPress,trans.
I. E. Drabkin.
Galilei, G. [1638/1914, repr. 1954]: DialoguesConcerning
TwoNew Sciences,New
York, Dover, trans.H. Crew and A. de Salvio.
Galilei, G. [ 1638/1974, rev. 1989]: TwoNewSciences,Including
Centersof Gravityand
Force of Gravityand Force of Percussion,Toronto, Wall and Thompson, trans.
S. Drake.
Giere, R. (ed.) [1992]: MinnesotaStudiesin the Philosophyof Science:Cognitive
Modelsof Science,XV,Minneapolis,MN, University of MinnesotaPress.
Gooding, D. C. [1990]: Experiment
andtheMakingof Meaning,Dordrecht,Kluwer.
Gooding, D. C. [1992]: 'The ProceduralTurn; or Why do Thought Experiments
Work?', in R. Giere [1992], pp. 45-76.
Gooding,D. C. [1993]: 'Whatis Experimental
aboutThoughtExperiments?'in D. Hull
et al. [1993], pp. 280-90.
Gooding, D. C. [1994]: 'ImaginaryScience' (review of Horowitzand Massey [1991],
Sorensen[1992a], Brown [199la]), BritishJournalforthePhilosophy
of Science,45,
pp. 1029-45.
Hacking, I. [1993]: 'Do ThoughtExperimentsHave a Life of Their Own?' in D. Hull
et al. [1993], pp. 302-8.
Hesse, M. [1966]: ModelsandAnalogiesin Science,Notre Dame, IN, University of
Notre Dame Press.
Horowitz, T. and Massey, G. (eds) [1991]: ThoughtExperiments
in Scienceand
Philosophy,Savage, MD, Rowman and LittlefieldPublishers.
Hull, D., Forbes, M. and Ohrulik,K. (eds) [1993]: Proceedingsof thePhilosophyof
ScienceAssociation,2, East Lansing, MI, Philosophyof Science Association.
Humphreys,P. [1993]: 'Seven Theses on ThoughtExperiments',in J. Earman,A. Janis,
and G. Massey (eds).Philosophical
Problemsof theInternalandExternalWorlds:
Essayson thePhilosophyof AdolfGrunbaum,
Pittsburgh,PA, University of PittsburghPress, pp. 205-27.
Irvine,A. D. [1991]: 'On the Natureof ThoughtExperimentsin ScientificReasoning',
in T. Horowitz and G. Massey [1991], pp. 149-65.
Janis, A. I. [1991]: 'Can ThoughtExperimentsFail?' in T. Horowitz and G. Massey
[l991],pp. 113-18.
GalileoandtheIndispensability
of ScientificThoughtExperiment423
Koyre, A. [1939/1979]: GalileoStudies,Atlantic Highlands,HumanitiesPress, trans.
J. Mepham.
Koyre, A. [1960]: 'Galileo's Treatise De MotuGravium:
the Use and Misuse of an
ImaginaryExperiment',in Metaphysics
andMeasurement,
London, Chapmanand
Hall [1968], pp. 44-88.
Kuhn,T. [1964]: 'A Functionfor ThoughtExperiments',reprintedwith new pagination
in TheEssentialTension,
Chicago,IL,Universityof ChicagoPress[ 1977],pp. 240-65 .
Kujundzic,N. [1993], 'How Does the Laboratoryof the Mind Work?Criticalnotice of
James RobertBrown, TheLaboratory
of theMind',Dialogue,CanadianReviewof
Philosophy/Revue
Canadienne
de Philosophie,32, pp. 573-7.
Laymon, R. [1991]: 'Thought Experiments of Stevin, Mach and Gouy: Thought
Experiments as Ideal Limits and as Semantic Domains', in T. Horowitz and
G. Massey [1991], pp. 167-92.
Lipton, P. [1993] 'Review of Horowitz and Massey (eds), ThoughtExperiments
in
ScienceandPhilosophy',Ratio(new series), 6, pp. 82-6.
Mach, E. [1883/1960]: TheScienceof Mechanics,sixth edition, La Salle, IL, Open
CourtPublisherstrans.T. McCormack.
Mach, E. [1897/1976]: 'On Thought Experiments', in Knowledgeand Error,sixth
edition, Dordrecht,Reidel, trans.T. McCormackand P. Foulkes, pp. 134-47.
Massey, G. [1991]: 'Gedankenexperimente:
Where Science and Philosophy Meet',
unpublishedmanuscript.
Nersessian, N. [1984]: Faradayto Einstein:Constructing
Meaningin Scientific
Theories,Dordrechtand Boston, Kluwer.
Nersessian, N. [1992]: 'How Do Scientists Think? Capturing the Dynamics of
ConceptualChange in Science', in Ronald Giere [1992], pp. 3-44.
Nersessian, N. [1993]: 'In the Theoretician's Laboratory:Thought Experiment as
Mental Modeling', in D. Hull et al. [1993], pp. 291-301.
Norton, J. [1991]: 'Thought Experiments in Einstein's Work', in T. Horowitz and
G. Massey [1991], pp. 129-48.
Norton, J. [1993]: 'Seeing the Laws of Nature:Review of James RobertBrown, The
Luboratory
of theMind',Metascience(new series), 3, pp. 33-8.
Norton, J. [1996]: 'Are Thought ExperimentsJust What You Thought?' Canadian
Journalof Philosophy,26, pp. 333-66.
Popper, K. R. [1959/1992]: TheLogicof ScientiJic
Discovery,New York, Routledge,
trans.K. R. Popper.
Prudovsky,G. [1989]: 'The Confirmationof the SuperpositionPrinciple:On the Role of
a ConstructiveThought Experimentin Galileo's Discorsi',Studiesin Historyand
Philosophyof Science,20, pp. 453-68.
Segre, M. [1989]: 'Galileo, Viviani, and the Tower of Pisa', Studiesin Historyand
Philosophyof Science,20, pp. 435-51.
Shepard,R. [1994]: 'MindandWorld:FromNewton, Einstein,andDarwinto Principles
of Mind', unpublishedWilliam JamesLectures,HarvardUniversity.
Sorensen,R. [1992a]: Thought
Experiments,
New York and Oxford,OxfordUniversity
Press.
424
Gendler
Szabo
Tamar
Sorensen,R. [1992b]: 'ThoughtExperimentsandthe Epistemologyof Laws', Canadian
Journal of Philosophy,22, pp. 15-44.
Souder,L. [undated]:'A Home Page for ThoughtExperiments',http://astro.ocis.temple.
edu/ souder/thoughtlindex.html.
Sunstein,C. [1993]: 'OnAnalogicalReasoning',HarvardLawReview,106, pp. 741-91.
Sunstein, C. [1996]: 'Political Conflict and Legal Agreement', Tanner Lectures on
Human Values, 18, Salt Lake City, UT, University of Utah Press.
Thomas, N. J. T. [undated]. 'Imagination, Mental Imagery, Consciousness, and
Cognition: Scientific, Philosophical and Historical Approaches', http://web.
.htm.
calstatela.edu/faculty/nthomas/index
Thomason, S. [1991]: 'Thought Experiments in Linguistics', in T. Horowitz and
G. Massey [1991], pp. 247-57.
Thomson, J. J. [1986]: Rights, Restitution and Risk, Cambridge, MA, Harvard
University Press.
Wilkes, K. V. [1988]: Real People: Personal Identitywithout ThoughtExperiments,
Oxford,ClarendonPress.