Sociology: Proscience or Antiscience?

Author(s): Randall Collins

Source: American Sociological Review, Vol. 54, No. 1 (Feb., 1989), pp. 124-139
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2095666
Accessed: 02/06/2009 08:18

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=asa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
American Sociological Review.

http://www.jstor.org
 SOCIOLOGY:
PROSCIENCE OR ANTISCIENCE?*
RANDALL COLLINS
Universityof California-Riverside

Criticismsof the scientific status of sociology possess some validity when applied
against narrowlypositivist interpretationsof sociological methodsand metatheory,
but do not underminethe scientificproject offormulating generalized explanatory
models. (1) Critics allege that sociology has made no lawfulfindings; but valid
general principles exist in many areas. (2) Situational interpretation,subjectivity,
reflexivity, and emergence are alleged to undermineexplanatory sociology, but
these topics themselvescan be explained by a widened conception of science that
allows informal procedures in theorizing aimed at maximizing explanatory
coherence. (3) Thefact that intellectualdiscourse itself is a historicallychangeable
social product does not invalidate objective explanatory knowledge. (4) The
historicist claim that there can be no principles that hold across particular times
and places is invalidand rests upon a confusionof underlyinggenerativeprinciples
with the complexities of the empirical surface of history. (5) Metatheoretical
criticismof the concept of causality does not underminea sophisticatedconception
of scientific sociology. Sociological knowledge can and does advance, but it
depends upon building the coherence of theoretical conceptions across different
areas and methods of research.

In recent years there have been a variety of sociological knowledge. But the overall thrust
assaults on the conception of sociology as a of the criticisms, that sociology has and can
science. These include the following themes. have no scientific validity, I believe is wrong.
(1) Sociology has failed to produce valid To be sure, science is not the only valid mode
findings or lawful generalizations.(2) Deter-of discourse, or of knowledge. Sociology,
ministic laws do not exist because social like many other intellectual disciplines, can
action consists of situational interpretation,
be concerned with empirical descriptions
based on human subjectivity, reflexivity, andincluding both contemporary social condi-
creativeness. (3) We are locked in a world oftions and historical sequences; it can discuss
discourse; society itself is a kind of text that
moral issues, propose or excoriatepolicies for
we read in different ways at different times.practical action, and compare existing condi-
(4) The foregoing position is often connectedtions against ideals; its can discuss founda-
to historicism, the claim that only historical
tional, methodological, and other metatheo-
particularsexist, and no general laws can be retical issues. But the core activity that gives
found that apply at all times and places. (5)the field of sociology its intellectualjustifica-
Finally, there are various technical criticisms
tion is the formulation of generalized explan-
of scientific methods and metatheory, espe- atoryprinciples, organized into models of the
cially of the conception of causality. The underlyingprocesses that generate the social
philosophy of science today is in a postposi-world. It is these that determine how
tivist mode; and a scientifically oriented
particularconditions result in particularkinds
sociology, it is said, is intellectually out of
of outcomes. It is these generalized explana-
date.
tory modes that constitute a science.
The various criticisms are not necessarily
I will attempt to show that none of the
united. Some of them include important
difficulties raised by the antiscience argu-
points that contribute to the widening of
ments prevent sociology from formulating
valid generalized explanatory modes. We
* I am indebtedto Paul DiMaggio,Richard have the basic structures of several such
Campbell, Robert Hanneman,Arthur Stinch- models already, in areas rangingfrom micro-
combe,andJonathan Turnerfor commentson an sociology, through formal organizations, to
earlierdraftof thispaper. macrosociology. It is not inevitable that
124 American Sociological Review, 1989, Vol. 54 (February:124-139)
PROSCIENCEOR ANTISCIENCE? 125
sociology must be a scientific failure, nor has vided that they are not unequals in power or
it failed. Those who attack scientific sociol- competitorsfor scarce resources. Variations
ogy typically fight against a caricature of on this principle have been formulated
"positivism" at its narrowest. On the other numerous times, based on a wide range of
hand, many practitionersof allegedly scien- research. Homans (1950) stated the basic
tific sociology have assumed just those point, drawing upon studies of informal
narrowconceptions of method and substance groups in industrial settings, as well as
thatmake them vulnerableto the antipositivist anthropological research and experimental
attack. small groups. Durkheim ([1912] 1954), in
In what follows, I will deal in turnwith the analyzing the central dynamic of religious
major criticisms of sociological science but rituals, seized on the similar principle that
also suggest what we can learn from the intensely focused interactionproduces moral
antiscience critiques. The flux of situational solidarity and conformity to the group's
interactions,humansubjectivityandreflexive- symbols; Goffman(1967 pp. 1-136) extended
ness, the dynamism and episodic movement this model to sociable conversations. The
of macrostructures,are all partof sociology's famous Asch (1951) experiments demon-
subject matter. It has been the merit of some strated the effect of group cohesion upon
of the antiscience positions to bring these to pressures for conformity even in visual
our attention, and even to explore their perception. Symbolic interactionist theory
pattern; but this exploration has made it converges on the same point: if a person's
possible to formulate these processes more concepts are derived from the stance of a
generally and hence to widen the realm of generalized other based on his or her social
explanatory models which are the core of experience, then what individuals think must
scientific sociology. be influenced by their patternsof interaction.
Research on self-conceptions (M. Rosenberg
THE ALLEGEDFAILURE OF 1979; R. Turner 1978) can be regardedas a
SOCIOLOGICALRESEARCH variant on this principle, demonstratingthe
influence of group membershipand solidarity
One line of attack dismisses sociology upon beliefs about oneself; in another appli-
because we have no findings. After almost cation, expectation states research (Berger,
100 years of research, we still have come up Wagner, and Zelditch 1983) shows the effects
with no valid generalizations, no laws of of group pressuresupon task performance(as
sociology. This criticism is often made by did W.F. Whyte's famous Street Corner
outsiders to our discipline; for instance, Society, 1943). Studies of networks (Bott
AlasdairMacIntyre(1984) uses it as grounds 1971) provide the equivalent formulation:
for his argumentthat there can be no secular, network cohesion results in homogeneous
nontraditionalbasis of morality; Alexander attitudes.The coherence among these various
Rosenberg (1980) argues that since the social kinds of theory and research constitutes
sciences have, and can have, no laws, any strong evidence that the interaction-
sociological explanation must come from a density/solidarity/conformity principles are
sociobiological level of determinism. Sociol- true.
ogists themselves sometimes also make the (ii.) Human cognitive capacity is limited;
same dismissal, usually in the context of a accordingly, the more complexor uncertaina
discussion of alternative metatheories (e.g., situation, the more that participantsfall back
Spencer 1987). upon a taken-for-grantedroutine and focus
But the charge that sociology knows on the particular area that presents the most
nothing, that we have no valid generaliza- dramatic problems. There has been a great
tions, is patentlyuntrue. Let us review a few deal of convergence on this principle from
of these, starting from the microlevel and very different points of view. HerbertSimon
moving up to the macrolevel. presented this as the principle of "bounded
(i.) The longer, more intensely, and more rationality,"which explains why membersof
exclusively persons interact with each other, organizationsengage in "satisficing" in most
the more that they will identify with one areas while troubleshootingthe most pressing
another as a group, and the more pressure problem at any given time (Simon 1957;
they will exert and feel for conforming to Marchand Simon 1958, pp. 173-71). From a
local patterns of behavior and belief, pro- very different angle, Garfinkel's (1967)
126 AMERICAN SOCIOLOGICALREVIEW
ethnomethodological experiments demon- when a state's apparatusof militarycontrol is
strate that individuals cannot cope with the broken down by internal dissension among
full complexity of social arrangementsand elites: this breakdown is especially likely
their justifications (especially since questions when there is military defeat and/or the
about these justifications are in principle economic strain of long-term military ex-
endless); accordingly, people actively resist penses beyond the organizational capacity of
whenever they are forced to question more the state to collect revenues. This principle,
than a very few taken-for-grantedroutines at as stated is limited: it tells us when a revolt
once. Corroboratingevidence comes from will break out, not who will win, nor what
experimental research on judgments under kind of social transformation, if any, will
uncertainty(Kahneman, Slovic, and Tversky follow. This pattern has many variations:
1982). when the factors causing breakdownoccur in
There are many ramifications of this a centralized state and coincide with popular
general insight regardinghumanstrategiesfor class conflict mobilized by changing property
operating with bounded rationality in a relations, the result is a major social transfor-
complex world; one might say it is one of the mation, a "revolution"in the fullest sense of
common themes of the social sciences in the the term (Skocpol 1979). The interactionof
late 20th century. It affects our efforts to demographic growth, money supply, and
constructmodels of humancognitive process- inflation can be a prior determinantof the
ing. It has implicationsfor understandingthe state's fiscal crisis (Goldstone 1986, 1987);
nature of organizations and of the social the position of a state within the world system
shaping of markets(Williamson 1975; White affects its ability to bring in revenuesaheadof
1981). It explains why a source of power its military expenses (Wallerstein 1974, pp.
within organizationsand occupationsis occu- 133-47); geopoliticalpatternsdeterminewhich
pying a position that has access to a crucial states will become overextended and hence
area of uncertainty,whose occupantsare able unable to sustain themselves militarily (Col-
to define to the rest of the organizationwhat lins 1981b, 1986, pp. 145-209). In particular
kind of nonroutine reality they are facing kinds of military/stateorganization (such as
(Crozier 1964; Wilensky 1964).' The cogni- most premodern empires), the result of
tive limitation principle also implies that geopolitical or fiscal crisis is disintegration
change on the macrolevel would follow a into smaller coercive units. A more complete
patternof long periodsof routinizationbroken theory of state crises, both revolutionaryand
by sudden episodes of restructuring.I would nonrevolutionary, would have to take into
suggest, in this light, that the microprinciple account these kinds of considerations But I
of cognitive limitations is implicated in believe we can accept with confidence the
Perrow's (1967, 1984) macromodel of orga- basic principle of state military/fiscal crisis
nizational systems, in which the combination leading to disintegrationof the apparatusof
of nonlinearity of organizational processes coercive control, and that in turn leading to a
and tight coupling among them results in revolt of subordinates.
episodic "system accidents." My purpose in citing these principles is
(iii.) On the macrolevel of the state, an merely to disprove the argumentthat sociol-
importantprincipleis: A political crisis arises ogy knows nothing, and hence that a social
science is impossible. I have not attemptedto
pick out our most importantprinciples, to set
1 Power can also be based on resource depen-
dence in a network structure (Cook, Emerson,
2
Gillmore, and Yamagishi 1983; Willer 1987); and A fiscal and/or military crisis is not the only
on coercion, exercised with varying degrees of route to the dissension among elites that leads to
effectiveness in differentnetworkstructures(Willer the disintegrationof the coercive apparatus.The
1987; Schelling 1962). Power also depends on the fiscal/militarycrisis theoryis not a complete theory
organizational distribution of control resources of all revolutions and other revolts; but it appears
(Etzioni 1975) and on the conditions of mobiliza- to be true, as far as it goes, and it covers a very
tion and conflict among opposing groups, both importantportion of events. As Paul DiMaggio
civilian (Tilly 1978) and military (Collins 1988b). points out (personalcommunication),this theoryis
Power is a complex phenomenon;we have made related to a more abstract explanatory principle,
progresson a series of partialtheories, without yet applicablein many contexts, about the disintegra-
pulling them all together. tion of an organizationalsystem.
PROSCIENCEOR ANTISCIENCE? 127
forth a systematic theory, or to assess the enact society. Such work includes the partic-
overall state of our knowledge in sociology ipant observation practiced by symbolic
(an effort to do so is in Collins 1988). Hence interactionists,as well as Goffman's efforts to
these principlesmay look eclectic, lacking the map out the natureof everyday life. Without
elegance of an overarching image of the such research, we would be left studying the
social world. But I have given some hints fundamental realities of sociology's subject
how these principles, although they are matter only indirectly, at methodological
chosen almost at randomfrom differentparts arm's length. In the last few decades, there
of the field, may be coherent with one have been further innovations in microre-
another, and I have suggested that such search,includingethnomethodologists'breach-
principles are not trivial but lead to sociolog- ing experiments, and culminatingin perhaps
ical insights into a wide range of important the closest empiricalanalysis ever done in the
questions. There are a good many other social sciences, using audio and video
principles of this sort, especially in organiza- recordings of natural interactions as a basis
tion theory but also in other areas of for developing formal models in conversation
sociology. There is of course plenty of room analysis.
for improvementin our precision and in our Most of this work would have been ruled
understandingof the scope of these theories, out by textbook canons of research 30 years
but enough evidence has built up that we can ago. The sense of alienation form the
be confident we have good approximationsof sociological "Establishment" felt by many
how some importantprocesses work. Many interpretive sociologists is no doubt due to
different social scientists have contributedto memories of having lived through that time.
this knowledge; we do have a core we can The vituperation against "positivism" is
build upon, and a discipline we can be proud partlythe expression of an oppressedintellec-
of. tual minority against their long-standing
Is there anything salutary in the criticism oppressors after finally gaining a foothold in
that sociology knows nothing? It is not true, respectability.
but it should serve to remindus that sociology But we need not assume that all connection
has a serious problem in professional self- between interpretiveand scientific sociology
presentation.And within our own ranks, we is now severed, and that microresearchwith
need to pay more attention to the explicit interpretivemethods should be institutional-
cumulationof what we do know. ized as a kind of "separate but equal"
enclave. On the contrary,the achievementof
the interpretive microsociologists should
SITUATIONALAND REFLEXIVE
INDETERMINISM broaden our sense of acceptable methods for
sociological science. Clearly, a scientific
It is sometimes held that deterministicexpla- methodfor our field cannotrule out studies of
nations are impossible because social action the subjective; sociological science cannot be
consists of situational interpretation,subjec- founded on an exclusionary behaviorism
tivity, reflexivity, and emergence. This is an (although we should avoid going to the
old criticism, going back at least to Dilthey's opposite extreme of ruling out the importance
distinctionbetween the Geisteswissenschaften of behavior, includingunconsciousbehavior).
and the Naturwissenschaften,and ultimately A science does not have to be built out of
to the German Idealists' revolt against the "harddata" in the narrowsense. What makes
Enlightenment. In recent years this line of it scientific is its ability to explain the
criticism has become very prominent, so conditions under which one kind of pattern
much so that one might characterizethe late holds ratherthan another, in whatever realm
20th centuryas a time of neo-Idealistrevival. those patternsmay be found.
It is importantto recognize that subjectiv- Similarly, sociological science cannot be
istic and interpretive schools of thought in equated to a rigid operationalizationof all of
recent sociology have made positive contribu- its concepts. Not only is it legitimate for
tions to sociological knowledge. On the explanatory theories to include nonopera-
methodological side, these approaches have tional concepts at some levels; even a very
fostered microresearchin naturalisticsettings, positivistic model must include generalorient-
empathetically entering into the processes, ing concepts within which its specific hypoth-
feelings, and thoughts of real people as they eses and operationalizedvariablesare located.
128 AMERICANSOCIOLOGICALREVIEW
We always need a model of what the world is so to speak; even numbers and logical
like, a mental picture of what are the relations have some areas of indeterminacy.
fundamentalprocesses and entities and how We encounterthis when number systems are
they link together (see J. Turner1988; Willer extended to infinity or to infinitesimality,and
1987). Specific hypotheses make sense only in nonconverging algebraic series. Many
in terms of some such backgroundassump- systems of equationsareunsolvablemathemat-
tions aboutwhat kind of world we are dealing ically. Even ratherrestricted formal systems
with. The narrower traditional forms of of logic always encounter G6delian incom-
positivism, in demandingtotal operationaliza- pleteness; more complex systems of multi-
tion of all concepts, merely took for granted valued, modal, and other nonclassical logics
an unconscious conception of the world in have even greater areas of disagreement in
which its explicit hypotheses were lodged. interpretation(Lewis 1986). I have expressed
Such researcherscould easily lock themselves this elsewhere (Collins 1988, Appendix A) in
into commonsensical or ideological assump- the following formula:mathematicsis always
tions, which their focus on the foregroundof embedded in words. But notice what conclu-
research techniques kept them from seeing. sion follows: not that mathematics and
The interpretivesociologies do us a service in mathematical science is impossible; on the
forcing the issue, so that we are made contrary,a successful science is possible even
explicitly aware of those backgroundmodels incorporating areas of fundamental uncer-
and thus able to theorize them. tainty, dealt with by tacit and informal
understandings. Tacit knowledge is knowl-
The Place of Informal Concepts and edge too, as long as it works.
Intuitionsin Theory Whateverour kinds of explanatorymodels,
we still need to be concernedaboutvalidating
The notion of a complete and rigid formaliza- our theories. The fact that we are always
tion, operationalization,and measurementof involved in interpretations (and at many
everythingin a scientific theory is a chimera. levels) does not mean that we can accept
There are informal concepts and intuitive every interpretationoffered at face value.
leaps at several points. There is always a Typically we cannot decide these issues by a
metatheoretical stance about what we are simple operationalization,measurement, and
doing intellectually in the first place. Scien- one-shot test. But the physical sciences face
tific theory sketches a model of the aspect of most of these same problems, and their
the world underconsideration;hypotheses are success in many areas shows that some
derived from this, by a process of derivation research programsand theoretical models do
that itself involves intuitive leaps. When prove themselves in the long run over rival
operationalizingconcepts for empirical test, ones; there can be convergence on models
we always make another intuitive leap in that work, that capture the central ways in
deciding that particular measurements or which the world is, even if these scientific
other observations actually bear upon the models are inevitably vague and cluttered
theory. These intuitive or informal leaps are around the edges. Successful heuristics and
areas in which theoretical discussions can intuitionsare possible, and unsuccessful ones
take place (or, in many cases, ought to take which lead us into blind alleys can be ruled
place). But they are not illegitimate. That is out.
simply the way the world is. They do not The crucial criterionis that the best theory
undermineour ability to have a science, for (with its ancillaryassumptionsand heuristics)
all sciences have these places where there are is that which maximizes coherence; it brings
intuitive leaps. If physical scientists some- together the most successful explanatory
times forget this and talk in crude positivistic models into a consistent overall picture of
terms as if they report "nothing but the how the world operates. Methodologically,
facts," that is because they have been empiricism can be part of the coherence
successful at making the right intuitive leaps criterion;the best validated theory is the one
as their scientific procedureshave cumulated, maximally grounded to the empirical world
so that they now have workable models that via the various explanatory submodels it
they know intuitivelyhow to apply to most of incorporates.An extreme, all-or-nothingem-
the things they study. piricism is impossible; but a flexible empiri-
Everythingwhatsoeverhas "fuzzy edges," cism, working with imprecisionsand intuitive
PROSCIENCEOR ANTISCIENCE? 129
concepts where necessary, and makinga great out (Goffman 1974), but Goffman did not
deal of room for theoretical work that ties regard this framing activity as free-floating,
things together, is a central part of science. and he rejectedthe suggestion that it reduced
One needs to work nonpositivistically, so to the world to a kind of psychedelic fantasy.
speak, to be a successful positivist. Transformativereinterpretationsof subjective
It is in just this vein that the interpretive reality are linked together into orderly trans-
schools have contributedsubstantivelyimpor- formations, among "adjacent"frames, so to
tant theories. Among these are the symbolic speak. For Goffman, the bottom level frame,
interactionisttheory of the self (part of which out of which all the others arise, is the
is coherentwith well-establishedprinciple [i.] physical interactionof human animal bodies,
mentioned above); the ethnomethodological an ecological baseline that links Goffman
theory of everyday rationality (which is theoretically to Durkheimian theory of the
coherentwith the limited cognition principle, ritual basis of solidarity and of symbol-
[ii.] above); as well as other existing and construction (for elaboration, see Collins
potential contributionsto sociological knowl- 1988, pp. 188-203, 291-97, 320-34).
edge. Goffman's (1959) model of dramaturgy It is possible, then, to have a structured
in everyday life is a model, in the sense of understandingof subjectivity.In that perspec-
"what the world is fundamentallylike" that I tive, the explorationsof the subjective side of
mentioned above; from this base, one can go human life in the last few decades, notwith-
on to develop specific explanatoryprinciples. standingtheir sometimes extreme pronounce-
I have argued, for instance, that it provides a ments, have contributedelements towards a
base for understandingthe differencesin class much more sophisticated explanatorytheory
cultures, between those wielding power and of mind than would have been possible
those subjected to power (Collins 1988, pp. before.
203-14). How Unpredictable is the Social World?
Many sociologists in the interpretivecamp, Let us confront the issue of unpredictability
however, claim that their major substantive head on. How much of the social world is
finding is the impossibility of deterministic unpredictable?A great many things are highly
theories (e.g., Blumer 1969, p. 60). In their predictable. It is the patternof people going
empiricalinvestigations,they discover, above to work over and over again to the same jobs
all, emergence, unpredictability,situational- that makes up much of formal organizations;
ity, humancapabilitiesfor subjectivelyreflect- repetitive patterns make up households and
ing upon and changing social conditions. families; networks of friends and acquaint-
Here we have an argumentaboutwhat kind of ances similarly are constituted out of behav-
model we arrive at, not about whether it is iors, cognitions, emotions, and communica-
possible to have any model at all. tions that are heavily patterned.For regularity
But is it true that the main feature of the and predictabilityto exist, it need not even be
social world is unpredictability,overwhelm- the same persons who repeatedly interact.
ing any determinateprocesses? I suggest that Most stores have only episodic relationships
it is not true, and that this perceptioncomes with particular customers, but it is the
from selectively focusing on a limited portion predictabilityof a certain number of people
of the social world. Although much (but not coming in to shop that allows businesses to
all) content of sociology consists of human stay open at all. Although microsociology is
subjectivity, it does not necessarily follow the theoretical bastion of indeterminacy, it
that this cognition and feeling are indetermi- should be apparentfrom these examples of
nate. Without pursuing this point into the everyday life that the microlevel has a high
status of theories of cognition and emotion, degree of predictability.
let us think of Goffman, the acknowledged The theory of indeterminacyseems to rest
genius of micro-interpretivesociology. Goff- upontwo suppositions.One is thatthis kind of
man used soft methods, but he believed the predictabilityis banal. It is true, but it is too
world he was studying is hard. His social boring for sociologists to pay attentionto it;
theory of language (Goffman 1981) grounds we should focus on something that everyone
cognition in the social ecology of interaction. does not alreadyknow. So one might say there
The complexity and reflexiveness of human is a built-in bias towardstudyingthe dramatic
subjective worlds come from the many and unpredictable.But I would deny thatwhat
possible "reframings"that actors can carry is banal from a participant'spoint of view is
130 AMERICANSOCIOLOGICALREVIEW
necessarily banal for an explanatorytheory. tions, the emergent events often look highly
On the microlevel, Garfinkeltook the banality patterned.The belief in indeterminismhere is
of everyday life as a productto be explained, a by-product of individual reductionism. If
and uncovered cognitive mechanisms that al- we go to a genuinely interactive level, it is
low it to happen-and that show us the pres- possible to specify the combinationof ingre-
sure points when these mechanisms are inter- dients that go into the situationand determine
fered with. On the macro- and mesolevels, its outcomes. (Such models of chains of
good sociological work consists in refraining interactions are proposed by Heise 1979,
the banalityof taken-for-grantedpatterns.Al- 1987, and by Collins 1981a.)
though it seems naturalto a particularperson Does Knowing a Sociological Law Reflex-
thatshe or he does a job and chats with friends ively OverturnIt? Finally, there is the argu-
every day, thereis much thatsociologists have ment that whateverlaws sociologists may dis-
uncoveredas to why jobs are structuredin this cover will be reflexively overturned.As soon
way ratherthan that, why these persons are as people know what these laws are, they can
friends ratherthan others, and so forth;these act to make them false. But although this
are the contents of, for example, organiza- sounds plausible in the abstract,it is difficult
tional theory, exchange and networktheories, to think of many cases in which it actually
and stratificationtheory. applies. Perhapsthe ThomasTheoremitself is
The other supposition leading us toward the prime example of a principle that can be
theoretical indeterminacyis more valid. It is reflexively undermined.We have been most
the recognition that situations can sometimes interestedin this as a theory of prejudice, of
change very rapidly: that there are negotia- the self-perpetuatingnature of biases against
tions, conflicts, sudden insights, decisions, particularcategories of persons. By becoming
and, on the macrolevel, movements, revolts, aware of these biases, the liberal public in the
and revolutions. All this is true. But do we United States has made efforts to counteract
take this as the end of analysis, or as a the bias. But does this actually violate the
beginning point, a challenge to develop Thomas Theorem?Instead, we seem to be at-
theories to explain when such sudden shifts tempting to avoid the circumstances under
will occur? I have already noted that on the which the theorem gets started in a negative
macrolevel, we know some of the crucial direction;we avoid giving negative definitions
featuresthatmake revolutionspredictable.On of persons, in the hopes thatthe consequences
the microlevel, indeterminacy is typically will be a positive self-fulfilling prophecy in-
grounded in some version of the Thomas stead of a negative one.
Theorem. But if situationsare determinedby Determinativeprincipleson the macrolevel
subjective definitions, we can still ask what look ratherdifficult to evade. For instance, if
determineswhat the definitionof the situation military defeat or fiscal crisis leading to
will be.3 What sometimes makes situations breakdownof the coercive apparatusfosters
seem to have an unpredictable, emergent revolutionaryconflict, one can hardlyprevent
quality is that one looks at them from the this simply by being aware of it; the best a
point of view of a single actor who knows governmentcould do would be to try to avoid
only his or her own intentions. But if we this principle's coming into operation, by
know enough about all the actors in the avoiding circumstancesthat lead to military
situation, and the structureof their interac- or fiscal crisis. Reflexivity may enable people
to try to change the distributionof indepen-
dent variables, but not the relationships
3The findings of ethnomethodologicalresearch between independent and dependent vari-
do not supportthe notion of a great deal of sudden ables. Similarly, principles of formal organi-
emergence and reinterpretation.Clegg (1975), for zational structuregive informationas to what
instance, who set out to study a constructionfirm people can work around,but hardlywhat they
in microdetail armed with tape recorder, soon
can fly in the teeth of. Even on the
found that the banality of everyday repetitiveness
was overwhelming, and he had to shift to conflict microlevel, where an individual would seem
points in management to find more dramatic most capable of reflexively changing an
material. Ethnomethodological theory proposes outcome, I would suggest that when individ-
that makingeverydaylife into a routineis the basic uals actually control outcomes they are doing
process, and that people attemptto avoid and gloss it by applying microsociological laws, not
over disruptionsas much as possible. going against them. For instance, when
PROSCIENCEOR ANTISCIENCE? 131
people deliberately go into an encounter makes inroads into the cosmopolitan side of
group or similar group-dynamics situation, sociology (e.g., Brown 1987; Gottdiener
they are implicitly making use of the 1985).
microprinciple ([i], above) regarding how But the upsurgeof sociological researchon
group solidarity is generated, because they cultureis not always merely relativistic. It has
want its emotional payoff. Their mistake also been carried forward in a deterministic
typically is to overestimate how long such mode: we have rather good researchers and
solidarityand emotionalenergy will last, after theories about the materialand organizational
a temporarygroup of this kind is disassem- base of the productionof culture, both as it is
bled. Knowing the principle does not under- distributedamong social classes, as well as in
mine it.4 more specialized culture-producing institu-
It would be rash to predictthat sociological tions (Bourdieu 1984; DiMaggio and Useem
science will explain everything. There may 1982; Coser, Kadushin, and Powell 1982).
well be a considerableresidue of indetermin- Culture does not simple organize itself; it is
ism even if sociology is successful far into the organizedby social processes.
future. But our intellectual impetus comes Foucault(1969), the most significantof the
from pushing back the frontier. To preach French "discourse" school, has documented
indeterminismand nothing further seems to the social bases of ideas in such practical
me a parasiticalstrategy, since it is intellectu- fields as psychiatry and law. But Foucault
ally of interest only if there is some body of does not attemptto underminethe validity of
theory already existing that one wishes to his own discourse as a historian. Moreover,
counter. The constructivejob is to build as the historical patterns that he emphasizes as
good an explanatorytheory as we can. determinativeof the field of discourse-the
I have tried to establish that a science can
bureaucratizationand specialization of social
operate flexibly, and on all kinds of subject
control agencies, the shift in public/private
matters. Against this backdrop, I will com-
boundaries,the move from ritualpunishments
ment more briefly on the remainingcriticisms
to a reflexively conscious self-are highly
of sociological science.
congruentwith Weberian(Weber [1915]1946)
and Durkheimian/Maussian(Carrithers,Col-
SOCIETYAS DISCOURSE lins, and Lukes 1985) theories of modernity.
The best of this European work reinforces
Another criticism goes as follows. We are rather than departs from the central cumula-
locked in a world of discourse; society itself tive traditionsof sociological knowledge.
is nothing more than a kind of text that we The intellectual popularity of "discourse"
read in differentways at differenttimes. This as a master worldview is bolstered by the
is a populartheme now, derivingfrom French success of the sociology of science in
structuralismand its offshoots. It has created showing how knowledge is a social construct.
a veritable revolution in the world of literary This success is indeed something to cheer
criticism. This is understandableas a profes- about. But we should not forget that the
sional ideology, elevating literary theorists' sociology of science is an empirical research
own field by the assertion that all reality is a
discipline, which has made great stridesin the
piece of literature. "Discourse" has thereby
last 30 years in cumulating sociological
won a wide influence in the largerintellectual
models of what determines the kind of
work (including the publishing business). It
fits well especially with the particularistic, knowledge produced by particularkinds of
organizationalconditions (see Whitley 1984,
descriptive themes of anthropologyand also
for recent summary and synthesis). Notice
what this does for the claim of undermining
4 It is sometimes said that knowing too much scientific knowledge. There is a determinism
abouthow socialrelationshipsoperatemakesthem at the very heart of this alleged indetermin-
go flat. Canan exchangetheoristor someonewho ism; sociology of science itself is becoming a
appliesDurkheimian andGoffmanian ritualtheory
fall in love? Does the theoretical self- scientific success.
consciousnessdestroythe situation?I can attestto This raises some interesting issues of
you it does not; robustsocial processeshave a reflexive self-consciousness. Some sociolo-
quite wonderfulpower, overridinga 'weaker gists of science (e.g., the British school
processlike momentary reflectiveness. aroundMichael Mulkay, HarryCollins, Steve
132 AMERICANSOCIOLOGICALREVIEW
Woolgar, and others; see Knorr-Cetinaand HISTORICISM
Mulkay 1983) go so far as to argue that
science is merely a set of competing power Historicism is the claim that only historical
claims. The only democraticpath is to allow particularsexist, and no general laws can be
no single voice to have a privileged position; found that apply at all times and places. This
hence, Mulkay (1985) and others have taken argumentarises to some extent in conjunction
to writing and presenting papers in a "New with other antipositivistcriticisms, as a kind
Literary Form," in which the author steps of oppositionalunitedfront. But it also has its
aside and lets many competing voices appear. autonomous basis in the actual practices of
The result is entertaining,but it is not clear historical sociology. It is a price we are
why reflexiveness should prevent scientific paying for a very good thing. The last 20
knowledge. Bloor (1976, 1983), who power- years, since the publications of Barrington
fully applies a Durkheimiantheory, argues in Moore and Charles Tilly in the 1960s, have
his "strong program" that a sociology of been a golden age for historical sociology.
science can and should explain not only false We have learneda great deal aboutmacropro-
knowledge claims but also true knowledge. cesses by looking at historical materialswith
From an organizational perspective, the a sociological eye and by making compari-
power claims in scientific discourse that sons across societies and times. This, for
Mulkay and others document are themselves instance, is how Moore (1966) showed the
part of fairly predictable patterns. Different connection between the forms of capitalist
kinds of intellectualdiscourse (i.e., particular agriculture and the differing political struc-
scientific disciplines) are embeddedin organi- tures of modern states. But although this
zations, and they themselves can be under- shows us something about particular states
stood as organizationalforms. Hence, organi- (17th-century England, 19th-centuryUnited
zational theory (especially the theory of how States, etc.), the underlying theoretical con-
various kinds of task uncertainty and of ceptions have a more universal application;it
is for this reason that a family of models
resource dependence affect organizational
related to Moore's (Paige 1975; Skocpol
behavior and structure; see Whitley 1984;
1979; see also Stinchcombe 1961; Weber
Fuchs and Turner 1986) shows scientific
[1923] 1961, pp. 81-94) have been power-
discourse is not a free-floating construction fully applied to other times and places.
but appears in predictableways under given Historical sociologists are under two sorts
circumstances. Furthermore, we know that of pressure to announce themselves as
organizational structures are at least partly historicists. One is that they come into a good
determinedby the task environmentsin which deal of contact with historians.Historicismis
they work (Collins 1988, pp. 467-85). This a kind of professionalideology for historians;
means that the objective natureof the subject they make their living by describing particu-
matteris one of the determinantsof the social lars, and furthermorethe pressuresof intellec-
activity (including the discourse) that makes tual competition within their field lead them
up science. to specialize and to resent intruderson their
Argumentsthat exclusively emphasize dis- turf. Hence, historians tend to dislike any
course are one-sided; although there is a suggestion that there are general processes,
culturally constructed component in any and particularlythat such processes might be
knowledge, it also can be knowledge of uncovered by comparing across times and
something. Indeed, any argument about the places (i.e., across the boundaries of their
social basis of knowledge is self-undermining researchspecialties). Historiansoften espouse
if it doesn't have some external truth- an ideology that would make it impossible to
reference as well-otherwise why should we consciously develop a general explanatory
believe that this social basis itself exists? We theory; and many historical sociologists
need to get beyond polemically one-sided respondto criticism from historiansby giving
epistemologies, of either the subjectivist or in to their ideology.
the objectivistsort;a multidimensionalepiste- But historians'claims are not consistent. In
mology can take account of the way we live interpretingtheirparticularcases, they implic-
in a cultural tunnel of our own history, but itly draw upon some ideas of what general
still we can cumulate objective knowledge structures are and how patterns of social
about the world. motivation and change operate. The analysis
PROSCIENCEOR ANTISCIENCE? 133
of historicalreality can hardlybe approached theoretical level of abstract and universal
with a tabula rasa; historians have theories explanatoryprinciples, and a level of histori-
whether they know it or not. What makes cal particulars. Insofar as our theories are
someone a great historian, one whose work successful, we will become better and better
commands widespread intellectual attention, at showing how the myriad arrangementsof
is typically his or her ability to create a historicalparticularsare generatedby particu-
theory, to show the more general skeleton lar combinations of variables in theoretical
underlyingthe narrativeof particulars.Lesser modes. There will always be tasks for
historiansare usually those who operate with historians, and for historical sociologists, in
naive, taken-for-grantedconceptions, or with making this kind of concrete interpretation.
old theories that have passed into common At the same time, exploring concrete history
discourse. Historians at their best have been is one of the main ways that we make
building sociological theory, although they progress in building and validating our
have not always discussed it as such, and general models-although such theory is built
have typically interwoven it with their and validated by its coherence with all kinds
particular historical description, sometimes of sociological research, contemporary as
with considerableartistryand dramaticstyle. well as historical.
I have no sympathy with the blanket claim It is not true that there are no explanatory
that historical particularism is all that is principles that hold generally across history.
possible; on the contrary,we cannot even-see The three examples that I gave at the begin-
particulars without general concepts. But ning of this paper are completely coherent, as
there is a more valid reason why historical far as I know, with evidence from any histor-
sociologists tend to work at a low level of ical epoch, andthereare many more such prin-
generality with theories embedded in the ciples. Of course, some principlesmay not be
understandingof a particularrange of cases. applicable because the independent variable
Even if one's aim is to develop general does not exist in a given historical situation.
theory, the macromaterials of long-term One cannot predicteither the existence or the
history are extremely complex. Although we nonexistenceof revolutionarycrises if there is
may know something of rudimentary pro- no state. But there is doubtlessa more abstract
cesses, getting any very abstractpictureof the formulationrelatedto principle(iii.) thatwould
overall combinationof conditionsoperatingin apply to sources of crises in political power in
historical change is very difficult. Theoreti- stateless societies. Macroprinciplesin general
cally oriented historical sociologists have are likely to be more complex than microprin-
workedwith intermediateapproximationsto a ciples, involving combinations of many pro-
level of explanatorygenerality. For instance, cesses. But we have at least rudimentaryout-
Weber's massive historical comparisons of lines of promisingprincipleson the macrolevel
the conditions involved in the rise of as well. It is a mistake in criticizing the limi-
rationalized capitalism have yielded many tationsof particulartheories(e.g., in widening
general analyticalpoints, but embeddedin an the referencefromallegedlyindependentsocie-
account of certain concrete sequences of ties to a world system, or overthrowingunili-
world history. This same halfway combina- near evolutionism or development theory or
tion is found in modem work such as that of functionalism),to go to conclude that general
Mann (1986) on the conditions determining theory is impossible. The result of this critical
the history of social power, of Goldstone development is not no theory at all, but im-
(1986, 1987) on the state crises and social proved theory.
transformations;and I will admitthat my own
work (Collins 1986), on such matters as the
extension of Weberiantheories of capitalism THE METATHEORETICALATTACK
or of sexual stratification, is also quasi- ON CAUSALITY
embedded in accounts of particularhistorical
developments. Such works are a challenge for Critics of explanatorysociology like to point
theoriststo attemptto departicularizewhat we out that the consensus in the philosophy of
have learned, to pull out the more fundamen- science has changed since the heyday of
tal theoretical structures underlying these logical positivism. It is generally acknowl-
accounts. edged that programssuch as that of Carnap,
We will always have two levels: a which attempted to construct all scientific
134 AMERICANSOCIOLOGICALREVIEW
knowledge from sensory experience orga- positivist image of their opponentsand do not
nized into statements of formal logic and touch a more realistic image of science.
mathematics,have failed. There is no consen- Criticisms of the concept of causality are
sus now on an alternative epistemology for frequentlyraised in this spirit. Causal theories
science, although most philosophers make are dismissed on the grounds that there is no
room for theoretical preconceptions and such thing as "the cause" of anything;there is
programs, and for pragmaticsin both theory always a complex of conditions, and those in
formulationand research (Quine 1969; Dum- turnhave antecedentconditions, which can be
mett 1978; Putnam 1983). It would be traced backwardsand outwardsin an endless
generally agreed that mathematicsand formal web. Given causes explain something only
logic are not self-grounding, and there is under particularconditions, which are typi-
considerable recognition of the role of cally taken for granted, especially in statisti-
nonformalized statements in any body of cal analyses of survey data that attempt to
knowledge. Along with this, there is a causally explain all the variance in their
loosening up in the conception of what particular sample. Some of the attack on
constitutesknowledge: not merely the ideal of causality, however, has come from within the
classical physics, but widened to include the scientific camp itself (Gibbs 1972) and does
rather different scope of knowledge in the not dismiss the aim of testable, generalized
biological and earthsciences, history, extend- explanatoryprinciples.
ing perhapseven as far as alternativeforms of Certainaspectsof this debateare merelyter-
knowledge embodied in art (Goodman 1978). minological. "Cause"is to come extent a met-
What does this mean for sociology? I aphor,a shorthandfor referringto our focus on
would suggest that it puts sociological science a particularportionof a complex of conditions
more on an even footing, epistemologically that are involved in producing certain out-
speaking, with the established natural sci- comes. Some of these conditionsmay be con-
comitantrelationshipsamong partsof a social
ences. For they too operate under the same
sort of epistemological imprecisions. Sociol- structure,as well as antecedentconditionsthat
determinewhich kinds of outcomes will fol-
ogy will never be a science fitting the old
low. (See general discussions by Klein 1987;
logical positivist ideal, but none of the natural
Walker1987;MeekerandHage 1988. As Wal-
sciences fits that ideal either. We are not
lace [1987] points out, there is a variety of
aiming at the impossible; if we can reach the causalpatterns-continuous, episodic, multile-
degree of approximateand pragmaticsuccess veled, etc.) But it is importantto retain this
the natural sciences have achieved, that concept, whetherunderthe term "causality"or
would be plenty. It is true that some somethingequivalent,for it enables us to dis-
sociologists may continue to uphold a meth- tinguish between explanationsthat work and
odological ideal that is closer to the crude ones that are vacuous. Functionalistanalysis,
induction-plus-mathematical-formalization for example, has turnedout to be a poor mode
model of science. I would suggest that this is of explanation,unless it can be translatedinto
particularlylikely in the applied end of our causalmechanisms(see Stinchcombe1986, pp.
field, where purely descriptive information 80-100). One cannot "explain"somethingby
(e.g., on the success of desegregation pro- giving it a name, even if thatname is "norms"
grams) has some immediate use, and hence or "rules" or "culture"-or for that mattera
straightforwardinduction is more likely to be "problematique"or "discourse"-any more
pursued. But this does not affect the larger than one can explain gravity by referringto a
issue of the methods appropriatefor building "gravitationouspropensity."Here "causality"
a general explanatoryscience. is useful by giving us a mechanism,which tells
Modernphilosophyof science does not de- us what process is operatingand when partic-
stroy sociological science; it does not say that ularoutcomescan be expectedratherthan oth-
science is impossible,but gives us a more flex- ers. "Causality"saves us from reifications,as
ible pictureof what a science is. This is all to well as from ideologicaljustificationsmasquer-
the good in consolidatinga science out of the ading underthe guise of explanations.
materialsthat sociology alreadyhas available. As we have seen above, an explanatory
A numberof the more specific technical criti- theory has as its core a model of "how that
cisms mounted by the antiscience position in part of the world works," what the parts are
sociology seem to me to cling to a narrowly and how they fit together. Specific causal
PROSCIENCEOR ANTISCIENCE? 135
propositionsfit into such a model and are the ical level, Turner seems to assume a rigidly
object of empirical testing, but they depend positivist conception of theory and fails to
upon the backgroundconditions of the whole recognize that any theory involves interpre-
model. Some of the objections to "causal tive leaps and pragmatics, including deciding
theory" in sociology are directedat particular that given observations are appropriately
kinds of statisticalmodels (e.g., in the status connected to a given theory. All theories are
attainmentliterature)built entirely at the level not equally valid; the question is which theory
of propositions. But although such models works in the largest number of contexts that
may be too embedded in a particularbody of are coherent with one another.
data from a given historicalperiod and fail to
make explicit the structuralconditions that
frame these processes, that is not to say that Obstacles to Cumulationand The
such causal processes cannot be incorporated OrganizationalPolitics of Sociology
in a valid theory of the larger social universe I maintainthat when we go looking for it, we
(see Campbell 1983). find bits and pieces of sociological knowledge
Steven Turner (1987) voices a more lying all over the landscape. Our problem is
specific objection, to causal statementsof the recognizing what we have and organizing it
form "the more the X, the more the Y." He so as to maximize its visibility. Why is this so
argues that such propositions are literally difficult?
untrue unless the correlation is perfect; but One reason is the fragmentation and
empirically there are always exceptions, and antagonism within our field. Sociology is
hence such propositions have no logical divided into a large number of specialties.
foundation. Turner denies that imperfect This is hardlysurprising,since there are many
correlationcan be taken as an approximation thousands of researcherswith an interest in
to true causal relations. He holds that there is cultivatingtheir own turf;and since sociology
no path, logically, from general propositions takes the entire social world (including its
(which are always idealized and perfect) to causes and effects) as its target, there is an
the messy world of inexact relationships. enormous range of empirical things that we
Statistics provides no answer to this underly- can study. The sheer amountand diversity of
ing issue. Theory is always underdetermined sociology give us a practical incentive not to
by data, and a wide-open pluralism of pay attentionto work outside one's own area.
theories is the consequence that presumable On top of this, there is a diversityof methods,
will always be with us. whose proponents frequently regard work
Turner's argument, however, leads to produced by rival methods as having no
absurdextremes. Does anyone really believe knowledge value. And there is a furthersplit
that if we had a large number of well- into theoreticalschools, which often denigrate
validatedpropositionsof the form "in a large each other as part of their competition for
proportion[fill in the range of probability]of hegemony. These battles are especially in-
cases, the more the X, the more the Y," we tense when theoreticalfactions have political
would know nothing at all? Turner's argu- overtones, or when it is argued that only
ment cuts against physical science as well as practical knowledge or politically engage
against sociology; again, I would be quite expressions of a particular sort are worth-
happy with the level of approximationand while. All these conditions make it difficult
pragmatic success that these other sciences for us to pick out the places where theoretical
have achieved, whatever a purist argument explanationscome together, and to assemble
like Turner's would say about the logical the bits of evidence thrown up by different
status of such knowledge.5 On the philosoph- approachesinto a coherent pattern.6

5 As Willer (1987, pp. 43, 220) points out, of experimentalerrorwas consideredordinaryand

physical scientists typically carryout experiments, no one even considered statistically testing the
not to achieve statisticalvalidation, but to discover result. . . . [W]hat can it possibly mean to say that
the range of conditions underwhich a theoretically physics is an exact science? . . . [E]xact meantthe
derived relationshipholds. Willer (1987, p. 220) exact use of theory, not necessarily the exact
comments: "physicists were satisfied with an productionof clean results."
experimentif its results came withina factor of ten 6 One of the main advantages of the natural
or more of theoreticallypredictedvalues. This sort sciences may be that there are relatively few direct
136 AMERICANSOCIOLOGICALREVIEW
Perhapseven more importantfor American ing togetherdifferentstudies into theoretically
sociology has been a different type of split, coherentpatterns.
between descriptive and theory-orientedre- Data from different time periods are a
search. The latter is the search for general positive advantagein attemptingto cumulate
explanatoryprinciples; the former is a much general explanatoryprinciples. The change in
more straightforwardinvestigation of ques- empirical distributionsthat we often observe
tions that, at least initially, are fully compre- enables us to formulate principles of scope
hensible to a layperson. This is the class of and to refine our theories. For example,
issues such as: how much social mobility is research from industrialsociology and com-
there? (are we still the land of opportunity? munity studies from the 1930s through the
were we ever?) how much racial discrimina- 1950s led to general principlesabout how the
tion, and what progress towards its diminu- experience of giving or taking orders pro-
tion? These kinds of researches could be duces divergentclass cultures(summarizedin
theory-relevant, but it is not necessary that Collins 1988, pp. 211-14). The knowledge
one have an explanationin termsof genuinely accumulatedremains relevant on the analyti-
analyticalvariablesin orderto produceuseful cal level, even if (as seems to be the case) the
information. Statistical technique without harsher forms of authority have diminished
much theory finds its place most naturally for most working people in recent decades,
here. It is likely this practically oriented which would bring about declining differ-
search for descriptive information that anti- ences in class cultures. A comparison of
positivistic sociologists have in mind when historical differences would thus not merely
they attack the methodological and antitheo- add particularisticdescription, but have value
retical bias they perceive in sociology. for testing and extending the theory;its scope
This division of labor between our different would be further bolstered by bringing in
activitiesneed not poison the atmosphereover even more extreme instances, which we can
whatis a scientificsociology. Practical/descrip- find most easily by going back to another
tive researchcould doubtlessbe done betterif it historicalepoch in which extremely harshand
were groundedin more analyticaltheory, but violent coercive controls were common.
for manypurposesan atheoreticaldescriptionis Analytical theory must be distinguishedfrom
enough. Whatis importantto recognize is that empirical generalizations about trends over
cumulatingan analytical core of explanatory time; only the formercan give us understand-
propositionsis a differententerprise.When we ing of future social patterns, whether or not
attemptto cumulateour generalknowledge, we there is a reversal of empirical trends in the
have to know what we are looking for and to independentvariables.
cull it from the mountainsof descriptivemate- Anotherreason why we fail to cumulateis
rial that make up so much of the researchlit- that we place too much emphasis on the most
erature.Theory-orientedresearchandpractical- recent advances in statistical techniques, and
/descriptive research can live together insufficientemphasison the connectionsof re-
harmoniously, even if they produce an sults acrossdifferentmethods.As Stinchcombe
information-retrieval problem. (1984, pp. 55-56) pointedout underthe head-
More debilitatingis an attitudethat seems ing "WhyareGeneralizationsforgotten?":"the
to carry over from practical/descriptivework biggestcorrelationcoefficientsin a body of em-
into all of sociological research:the tendency piricalmaterialswill ordinarilygive rise to the
to regardonly the latest dataas relevant. If we biggest path coefficients. . . . Similarly, in al-
want to know what the equalityof educational most all cases, the metric coefficients in a re-
opportunityis, the survey with the latest date gressionequationwill be biggestrelativeto vari-
is obviously of most practical significance. ationsin causalforces whenthe pathcoefficient
But this "bias of the present," a kind of is largest.... And, in turn,the big relationsin
headline-chasingin sociology, is one of the loglinear analysis are almost always the same
factors that turn our faces away from pull- ones thathave big metricregressioncoefficients
....What this means is that almost all of our
causalknowledgefrom the time we were using
political implications of their theories. This has correlation coefficients(oreven eyeballingcross-
enabled them to avoid a source of controversies tabulations) is still our causal knowledge
that has clouded the analyticalprojectof the social throughoutthe transitionsto path analysis, to
sciences, and of sociology above all. structuralequationswith metriccoefficients, to
PROSCIENCEOR ANTISCIENCE? 137
loglinear analysis." There are of course ad- a sense, is too much knowledge, especially
vances, in the sense that log linear analysis or since so much of it is on the practical/descrip-
latent variablemeasurementmodels enable us tive side, which becomes unwieldy if it is not
to see certaindifficultpatterns,to model certain boiled down into theoretical generalizations.
complexitiesandtakeaccountof particularpos- We suffer from the limitations of our
sibilities of measurementerror. But these are cognitive capacity, and this cognitive over-
developments,so to speak, on the edges of the- load reinforces the structurallybased tenden-
oretically substantiveparadigms, not at their cies in the intellectualworld to simplify one's
core. We should be buildingon the earlierre- sociological worldview down to one's own
sults ratherthan dismissing them because the researchspecialty, theoreticalcamp, or polit-
researchlacked all of today's technicalrefine- ical faction, and to focus solely on the latest
ments. researchdata or technique.
As current methodological sophistication
(e.g., Campbell 1983; Lieberson 1985; Willer
1987) readily acknowledges, there is no auto- CONCLUSION
matic "methodologicalfix"; explanatorymod- My final point, perhapsthe most important,is
els are always underdetermined by any partic- about the mood of sociology. It concerns our
ular set of data, and theoreticalconsiderations social relationships,our attitudestoward one
always enter into our assumptionsas to which anotherwithin our intellectualfield. Much of
models can be plausiblyexcluded. If empirical what we expresstodayabouteach other'swork
patternsare robustenough-that is, if our the- is negativistic,hostile, dismissive.This faction-
ories are on the righttrack-we will find them alism is debilitatingbecause we need multiple
with a varietyof methods. If the theorieshave approachesin orderto cross-validateour find-
a weak graspof the centralprocesses,thenwhat ings.
we end up with aftergreatstatisticalrefinement For sociology to make progress, we need
is likely to be at best a shadow. some spiritof generosity, insteadof a spiritof
All research methods have their weak- factionalantagonism.This is not the same as a
nesses. We can overcome this problem by policy of "go your own way," toleratingeach
showing a coherent pattern among results otherbut havingnothingto do with one another
found with differentmethods. Thus, although intellectually.Buildingsociologicalknowledge
there are weaknesses in cross-sectional re- is a collective enterprise, and in more ways
search (Lieberson 1985), the coherence of than one. All humanactivities are social, and
theorybased on all sources of evidence can be science itself is a processof organizingthought
used to validate our inferences as to what collectives. As in other humanactivities, con-
factors are operative in a given body of data. flict is inherentwithin the organizationof the
The existence of experimentalevidence helps intellectualworld. This is not bad, since con-
enormously, when tied to a general theoreti- flict is a main source of intellectualdynamics,
cal scheme; so do theoretical connections to including the processes by which we put for-
naturalisticstudies (e.g., of organizationsor wardnew theoriesandcollectivelydecidewhich
of face-to-face interaction), to historical ones lead to the best results. But conflict need
studies of dynamics, and indeed to the entire not be extreme. In no form of intellectuallife
range of sociological research. Coherence, if do we depend more heavily upon each other
it is sufficiently strong, can bail us out of any than in a science. To come togetheras scien-
local methodological jam. And this is a tists, we need to concentrateon the coherence
reason for working to theorize our findings of theoreticalconceptionsacrossdifferentpieces
everywhere. From a methodologist's angle, of research.The personalaspect of that intel-
theory might be regarded as a device for lectual structureis generosityand good will, a
storing our understandingsof this coherence. positive feeling towardseach other's best con-
Hence, ignoring earlierresearchis a vice and tributionsas we grope our way forward to-
not a virtue; the very differences in research gether.
methods give us an opportunity to validate
our findings by "triangulation."
Sociology is engaged in many different REFERENCES
enterprises.The extreme diversity of what we Asch, Solomon E. 1951. "Effects of Group
do tends to keep what we already' know Pressure Upon the Modification and Distortion
hiddenundera cloud of dust. Ourproblem, in of Judgments." In Groups, Leadership and
138 AMERICANSOCIOLOGICALREVIEW
Men, edited by Harold Gaetzkew, Pittsburgh: Durkheim, Emile. [1912] 1954. The Elementary
CarnegiePress. Forms of the Religious Life. New York: Free
Berger, Joseph, David G. Wagner, and Morris Press.
Zelditch, Jr. 1983. "ExpectationStates Theory: Etzioni, Amitai. 1975. A ComparativeAnalysis of
the Status of a Research Program." Stanford ComplexOrganizations.New York: Free Press.
University:Technical ReportNo. 90. Foucault, Michel. [1969] 1972. The Archaeology
Bloor, David. 1976. Knowledge and Social of Knowledge. New York: Random House.
Imagery. London:Routledge and Kegan Paul. Fuchs, Stephan and Jonathan H. Turner. 1986.
. 1983. Wittgenstein: A Social Theory of "What Makes a Science 'Mature'? Organiza-
Knowledge. New York: Columbia University tional Control in Scientific Production."Socio-
Press. logical Theory4: 143-50.
Blumer, Herbert. 1969. Symbolic Interactionism. Garfinkel, Harold. 1967. Studies in Ethnometho-
Englewood Cliffs, NJ: Prentice-Hall. dology. Englewood Cliffs, NJ: Prentice-Hall.
Bott, Elizabeth. 1971. Family and Social Network. Gibbs, Jack. 1972. Sociological TheoryConstruc-
London:Tavistock. tion. Hinsdale, IL: Dryden Press.
Bourdieu, Pierre. 1979/1984. Distinction:A Social Goffman, Erving. 1959. The Presentation of Self
Critique of the Judgmentof Taste. Cambridge: in EverydayLife. New York: Doubleday.
HarvardUniversity Press. _ 1967. Interaction Ritual. New York:
Brown, Richard Harvey. 1987. Society as Text. Doubleday.
Chicago: University of Chicago Press. . 1974. Frame Analysis. New York: Harper
Campbell, Richard T. 1983. "Status Attainment and Row.
Research:End of the Beginning or Beginning of _ 1981. Forms of Talk. Philadelphia:
the End?" Sociology of Education 56: 47-62. University of PennsylvaniaPress.
Carrithers,Michael, Steven Collins, and Steven Goldstone, Jack A. 1986. "State Breakdownin the
Lukes (eds.) 1985. The Category of the Person. English Revolution." American Journal of
Cambridgeand New York: CambridgeUniver- Sociology 92: 257-322.
sity Press. _ 1987. "Cultural Orthodoxy, Risk and
Clegg, Stewart. 1975. Power, Rule and Domina- Innovation:The Divergence of East and West in
tion. A Critical and Empirical Understandingof the Early Modern World." Sociological Theory
Power in Sociological Theory and Everyday 5: 119-35.
Life. London:Routledge and Kegan Paul. Goodman, Nelson. 1978. Ways of World-Making.
Collins, Randall. 1981a. "On the Micro- Indianapolis:Bobbs-Merrill.
foundations of Macro-sociology." American Gottdiener, Mark. 1985. "Hegemony and Mass
Journal of Sociology 86: 984-1014. Culture: A Semiotic Approach." American
_ 198 lb. "Long-term Social Change and Journal of Sociology 90: 979-1001.
the Territorial Power of States." In Randall Heise, David R. 1979. Understanding Events.
Collins, Sociology since Mid-century:Essays in Affect and the Construction of Social Action.
Theory Cumulation. New. York: Academic New York: CambridgeUniversity Press.
Press. _ 1987. "Affect Control Theory: Concepts
. 1986. Weberian Sociological Theory. and Model." Journalof MathematicalSociology
Cambridgeand New York: CambridgeUniver- 13: 1-33.
sity Press. Homans, George C. 1950. The Human Group.
. 1988. TheoreticalSociology. San Diego: New York: Harcourt,Brace.
HarcourtBrace Jovanovich. Kahneman,D., P. Slovic, and A. Tversky. 1982.
Cook, Karen S., Richard M. Emerson, Mary R. Judgment Under Uncertainty: Heuristics and
Gillmore, and Toshio Yamagishi. 1983. "The Biases. London:CambridgeUniversity Press.
Distributionof Power in Exchange Networks." Klein, David M. 1987. "Causation in Sociology
AmericanJournal of Sociology 89: 275-305. Today: A Revised View." Sociological Theory
Coser, Lewis A., Charles Kadushin, and Walter 5: 19-26.
W. Powell. 1982. Books: The Culture and Knorr-Cetina, Karin and Michael Mulkay, eds.
Commerce of Publishing. New York: Basic 1983. Science Observed. Perspectives on the
Books. Social Study of Science. Beverly Hills: Sage.
Crozier, Michel. 1964. The BureaucraticPhenom- Lewis, David. 1986. On the Plurality of Worlds.
enon. Chicago: University of Chicago Press. Oxford:Basil Blackwell.
DiMaggio, Paul and Michael Useem. 1982. "The Lieberson, Stanley. 1985. Making It Count: The
Arts in CulturalReproduction."In Culturaland Improvementof Social Research and Theory.
EconomicReproductionin Education, edited by Berkeley: University of CaliforniaPress.
Michael Apple. London: Routledge and Kegan MacIntyre, Alasdair. 1984. After Virtue. Notre
Paul. Dame, IN: University of Notre Dame Press.
Dummett, Michael. 1978. Truth and 'Other March, James G. and Herbert A. Simon. 1958.
Enigmas. Cambridge:HarvardUniversityPress. Organizations.New York: Wiley.
PROSCIENCEOR ANTISCIENCE? 139
Mann, Michael. 1986. The Sources of Social _ 1968. Constructing Social Theories. New
Power. Volume 1. New York: Cambridge York: Harcourt,Brace and World.
University Press. . 1984. "The Origins of Sociology as a
Meeker, BarbaraF. and JeraldHage. 1988. Social Discipline." Acta Sociological 27: 51-61.
Causality. Allen and Unwin. Tilly, Charles. 1978. From Mobilization to
Revolution. Reading, MA: Addison-Wesley.
Moore, Barrington, Jr. 1966. Social Origins of
Turner, Jonathan H. 1988. A Theory of Social
Dictatorship and Democracy. Boston: Beacon
Interaction.Stanford:StanfordUniversityPress.
Press. Turner, Ralph H. 1978. "The Role and the
Mulkay, Michael. 1985. The Wordand the World: Person." American Journal of Sociology 84:
Explorations in the Form of Sociological 1-23.
Analysis. London:Allen and Unwin. Turner, Stephen P. 1987. "Underdetermination
Paige, Jeffery. 1975. Agrarian Revolution. New and the Promise of Statistical Sociology."
York: Free Press. Sociological Theory5: 172-84.
Perrow, Charles. 1967. "A Framework for the Walker, Henry A. 1987. "Spinning Gold from
ComparativeAnalysis of Organizations."Amer- Straw: On Cause, Law and Probability."
ican Sociological Review 32: 194-208. Sociological Theory5: 28-33.
_ 1984. Normal Accidents. New York: Wallace, Walter L. 1987. "Causal Images in
Basic Books. Sociology." Sociological Theory5: 41-46.
Reason. Wallerstein, Immanuel. 1974. The Modern World
Putnam, Hilary. 1983. Realism and
System. Volume 1. New York: Academic Press.
Philosophical Papers. Volume 3. New York: Weber, Max. [1915] 1946. "Religious Rejections
CambridgeUniversity Press. of the World and their Directions." In From
Quine, Willard V. 1969. Ontological Relativity Max Weber: Essays in Sociology. New York:
and OtherEssays. New York:ColumbiaUniver- Oxford University Press.
sity Press. [1923] 1961. General Economic History.
Rosenberg,Alexander. 1980. Sociobiology and the New York: Collier-Macmillan.
Preemptionof Social Science. Baltimore:Johns Whitley, Richard. 1984. The Intellectual and
Hopkins University Press. Social Organization of the Sciences. Oxford:
Rosenberg, Morris. 1979. Conceiving the Self. ClarendonPress.
New York: Basic Books. White, Harrison C. 1981. "Where Do Markets
Schelling, Thomas C. 1962. The Strategy of Come From?" American Journal of Sociology
Conflict. Cambridge: Harvard University Press. 87: 517-47
F. 1943. Street Corner Society.
Simon, Herbert A. 1957. Models of Man. New Whyte, William
Chicago: University of Chicago Press.
York: Wiley.
Wilensky, Harold L. 1964. "The Professionaliza-
Skocpol, Theda. 1979. States and Social Revolu- tion of Everyone?"AmericanJournal of Sociol-
tions. New York: CambridgeUniversity Press. ogy 70: 137-58.
Spencer, Martin. 1987. "The Imperfect Empiri- Willer, David. 1987. Theoryand the Experimental
cism of the Social Sciences." Sociological Investigation of Social Structures. New York:
Forum 2: 331-72. Gordon and Breach.
Stinchcombe,ArthurL. 1961. "AgriculturalEnter- Williamson, Oliver E. 1975. Marketsand Hierar-
prise and Rural Class Relations." American chies, A Study of the Economics of Internal
Journal of Sociology 67: 165-76. Organization.New York: Free Press.