OUTGROWING PHYSICS ENVY:

RECONCEPTUALIZING SOCIAL
RESEARCH
David P. Nichols

ABSTRACT: There is a pervasive sense of unease among social sci-

entists concerning the status of social research. This unease is rooted
partly in a false dichotomy between objectivity and subjectivity and a
belief that an idealized positivist version of classical physics should
be the model for all sciences. Experimental methodology is one of
many valid ways of obtaining knowledge and carries with its use a
particular set of problems, particularly when social phenomena are
studied. A reconceptualization of social research is needed, in which
experimental and quasi-experimental methods are used with more
caution and are supplemented by a more thorough conceptual appa-
ratus.

What are appropriate goals and methods for research in the so-
cial sciences? The fact that this question does not have obvious an-
swers for many social scientists is borne out by the existence of two
edited volumes produced as the result of conferences at the Univer-
sity of Chicago, in 1979 (Kruskal, 1982) and 1983 (Fiske & Shweder,
1986). The range of opinions expressed in these volumes is demonstra-
tive of the deep divisions that exist in the social science community
over basic issues.

David P. Nichols, AM, University of Chicago, Department of Psychology, Commit-

tee on Research Methodology and Quantitative Psychology, also is with SPSS, Inc.,
Chicago, IL. Reprint requests should he addressed to the author at SPSS Inc., 444 N.
Michigan Avenue, Chicago, IL 60611.
Contemporary Family Therapy, 15(1). February 1993
O 1993 Human Sciences Press, Inc. 51
52

CONTEMPORARY FAMILY THERAPY

THE CURRENT STATUS OF SOCIAL RESEARCH:

A PERVASIVE SENSE OF UNEASE
Texts' such as Kerlinger (1964) and Bock (1975) presented views
of social scientific research firmly rooted in the paradigm of logical
positivism. According to Kerlinger, "Scientific research is systematic,
controlled, empirical, and critical investigation of hjrpothetical propo-
sitions about the presumed relations among natural phenomena" (p.
13). The hypothetico-deductive framework borrowed from classical
physics has long been seen by many as the key to the building of a
successful social science, a physics of social phenomena.
However, the idealized version of science presented by the logical
positivists has been under siege for several decades, due to events
both internal and external to research in modem physics. The inter-
nal events center around relativity and, even more importantly, the
radical uncertainties introduced by quantum mechanics (Capra, 1975,
1982). Shweder (1986) views this as an important lesson for social
science. "It is perhaps worth noting that modem physics, since at
least the work of Niels Bohr, has moved away from mechanistic imag-
ery and the idea of objective predetermination and toward what has
aptly been labeled the physics of possibility, ambiguity, and uncer-
tainty" (p. 175).
Nevertheless, perceptions of physics and science have remained
static in many quarters. "It would seem that our mj^hic idealizations
of science change quite slowly" (Shweder, 1986, p. 175). Part of the
reason for the slow evolution of ideas about how science works is that
the people who choose to pursue careers in the physical and natural
sciences are not predisposed to challenging the dominant idealization.
As Harding (1991) puts it, "given the qualities that make them 'good
scientists,' natural scientists are the last people to suppose it desir-.
able to examine the limits of their minds to mirror nature or make
rational scientific choices, and of their wills to bring about their
choices" (p. 95). This makes them "psychologically the wrong people
to provide causal accounts of science" (p. 95).
According to Harding, a better understanding of science can be
gained by the application of social theoretic skills to the natural sci-
entific enterprise. "What is needed are people trained in critical social
theory: that is, in locating the social contexts—psychological, histori-
cal, sociological, political, economic—that give meaning and power to
historical actors, their ideas, and their audiences" (p. 95). Harding's
 53

DAVID P. NICHOLS

thesis is that the idealized model of physics not only is not a good
model for the social sciences, but that it is not even a good model for
physics.
Harding's work points to the fact that some of the same kind of
questioning of basic principles and assimiptions that is taking place
in the social sciences is also occurring in the natural sciences. Much
of the controversy was prompted by the publication in 1962 of
Thomas Kuhn's The Structure of Scientific Revolutions. This bomb-
shell from an historian of science was the external complement to the
events in modem physics previously discussed. Kuhn's challenge to
conventional views of scientific progress was profound enough to be
entertained most seriously by some of the world's most prominent
philosophers and historians of science. A 1965 London symposium on
Kuhn's ideas produced a volume (Lakatos & Musgrave, 1970) that
matches the later social science volumes in the variety of expressed
perspectives.
It is safe to say that the positivist conception of scientific progress
still reigns among the majority of physical scientists. While also a
major force among social scientists, belief in the search for a physics
of social phenomena is no longer near universal. As Cronbach (1986,
pp. 84-85) explains,

A "social physics" is an unlikely development. The Pill dates

back less than one generation; widespread advanced educa-
tion of women dates back less than two; about four genera-
tions back in Western democracies, free public education
spread; five generations back, the Industrial Revolution.
Such rapid change severely limits generalization about social
structures and relationships.

This sentiment is echoed by D'Andrade (1986), who criticizes the

application of the covering law model of classical physics to the sub-
ject matter of the social sciences. "A more adequate model would
stress that the way the generalizations develop in any domain should
be appropriate to the kinds of order and regularity found in that do-
main" (p. 26). Since social phenomena do not display the same kind of
order or regularity found in physical phenomena, the search for the
same level of fundamental laws makes little sense, and the relative
lack of generality of the descriptions and explanations to be found in
social scientific accounts should not be seen as evidence of a lack of
54

CO^^^EMPORARY FAMILY THERAPY

progress, but rather as an accurate refiection of the status of the sub-

ject matter.
Both Cronbach and D'Andrade see the norms of progress found in
the physical sciences Eis inappropriate for the social sciences because
of the real nature of the phenomena found in the social arena. There
are those (e.g., Gergen, 1985, 1986) who go even farther, denying the
reality of social phenomena independent of our descriptions, or con-
structions.

Social actions, as matters of common concern, owe their exis-

tence to the social process whereby meanings are generated
and events indexed by these meanings. There are no inde-
pendently identifiable, real-world referents to which the lan-
guage of social description is cemented (1986, p. 143).

Though we may think that we construct theories to describe real

events, and may therefore at times alter our theories because they
fail to express the nature of the social world, according to Grergen, "in
the case of describing human action, we confront the possibility that
theory is fundamentally closed to empirical evaluation" (p. 159). The
result is that rather than accumulating knowledge about how the world
works, "the sciences [should] chiefly be viewed as sources of intel-
ligibility or vehicles for the conceptual construction of reality" (p. 159).

THE ROOTS OF OUR DISCONTENT:

ABSOLUTISM AND A FALSE DICHOTOMY
Gergen's position represents the opposite side of a dichotomy
formed with the traditional position of logical positivism. This opposi-
tion between objectivism and relativism was appealingly laid out in
Pirsig's (1974) Zen and the Art of Motorcycle Maintenance in the form
of a dilemma posed to the protagonist. The importance and pervasiv-
ness of this dichotomy are hard to overstate. As Bernstein (1983)
says, it is "the central cultural opposition of our time" (p. 7).
Bernstein sums up the standard view of the positivist or objecti-
vist to the relativist argument: "One cannot consistently state the
case for relativism without undermining it" (p. 9). In other words, the
statement that everjrthing is relative is itself an absolute statement
and is therefore logically self-refuting. This response is, however, too
pat and too easy. This is so because the objectivist position begins by
 55

DAVID P. NICHOLS

assuming the rules of logic; the game has been rigged from the start.
The onus of responsibility for establishing basic assumptions has
been placed on the relativist, and he or she will surely fail under
these rules. However, should we reverse the responsibility by refus-
ing to assume the rules of logic, the position of the objectivist will fare
no better.
This is so because no system of thought is capable offiiUyjustify-
ing its own assumptions. As Lakatos (1970) puts it, "All scientific
research programmes may be characterized by their 'hard core'. The
negative heuristic of the programme forbids us to direct the modus
tollens at this 'hard core"' (p. 133). This hard core "is 'irrefiitable by
the methodological decision of its protagonists" (p. 135). For logic, a
part of the hard core is the existence of Absolute Truth. Pirsig (1974)
illustrates the resulting fundamental problem with the example of
the Socratic method: "Once it's stated that 'the dialectic comes before
anj^thing else,' this statement itself becomes a dialectical entity, sub-
ject to dialectical question" (p. 353).
The assumption that formal logical principles are the basis of all
scientific thought is so deeply ingrained in modem western culture
that the way out of this dilemma may be unclear to many scientists.
As Shweder (1986) says, "the mythic idealization in our culture of the
physical or natural sciences may have led us to draw an all-too-sharp
contrast between what is hard and what is soft, between what is ob-
jective and what is subjective" (p. 175). In other words, it may not be
the case that one either accepts the existence of Absolute Truth or
else winds up accepting the idea that all views have equal claims to
validity and one therefore cannot with justification argue against any
perspective.
This point is by no means a new one. Bernstein (1976, 1983),
Hubner (1983), Laudan (1984), Pirsig (1974), Rorty (1979), Shweder
(1986), and Toulmin (1972) all explicitly make this claim. What is
important to understand is that the problems involved in the adher-
ence to this dichotomy stem from the fact that the objectivist and
relativist positions are actually both sides of the same coin—an abso-
lutist interpretation of nature that has troubled us for centuries. As
Bernstein (1983) puts it, "Relativism ultimately makes sense (and
gains its plausibility) as the dialectical antithesis to ohjectivism. If we
see through objectivism, if we expose what is wrong with this way of
thinking, then we are at the same time questioning the very intel-
ligibility of relativism" (p. 167).
56

CONTEMPORARY FAMILY THERAPY

BEYOND THE FALSE DICHOTOMY

The way out of the dilemma begins with the realization that "in-
ductive and deductive logic cannot account for every example of sys-
tematic, constrained thinking, and there are many examples of im-
personal constraints that are not logical rules—not the least of which
are the rules of language" (Shweder, 1986, p. 179). In Toulmin's
(1972) words, we must "abandon . . . the philosophers' traditional as-
sumption that rationality is a sub-species of logicality" (p. 486). If we
"reject the commitment to logical systematicity which makes absolut-
ism and relativism appear the only altematives available" (p. 84) we
will find that "it is still possible to steer a middle way between the
absolutist and relativist extremes" (p. 497).
Shweder's (1986, pp. 177-178) discussion of this third way is
worth quoting at length:

There is, however, a third reaction to that idealization of ob-

jective science discussed earlier. I will call it the science-of-
subjectivity reaction, even though the expression 'science of
subjectivity' is a contradiction in terms from the perspective
of either hermeneutics or positivism. The basic idea is that
we must revise our conception of the relationship between
nature and culture, objectivity and subjectivity. The idea is
that as we have learned more from historians and histori-
cally oriented philosophers about the actual workings of sci-
ence . . . objectivity-seeking science has come to seem inher-
ently subject dependent. And, as we have learned more from
linguists, sociologists, anthropologists, and psychologists
about the workings of subjectivity . . . human subjectivity
has come to seem more objectlike than imagined. The real
world, it seems, is populated with subject-dependent objects
and objectlike subjectivity, two types of phenomena for which
there is no place in the mutually exclusive and exhaustive
realms of the symbol-and-meaning-seeking hermeneuticist
and the automated-law-seeking positivist.

One basic implication of this view for social research is that "There is
a soft side to all hard data, or perhaps the crucial point is that with-
out the soft side there is no hard side" (p. 174).
For those engaged in research centered on issues of rationality,
the implications are perhaps even more profound: "other conceptions
that rest on a neat and clean contrast between what's objective and
 57

DAVID P. NICHOLS

what's subjective ought to be revised, which means, to say the least,

we may have to rethink for a bit our conception of 'rationality' and
our concept of'meaning'" (p. 178). Echoing Pirsig's (1974) belief that
"reason may be expanded to include elements that have previously
been unassimilable and thus have been considered irrational" (p.
230), Shweder (1986) reminds us that "A science of subjectivity re-
quires a broadened conception of rationality" (p. 178).
Shweder's version of a science of subjectivity fits in well with
qualitative research traditions in the social sciences, particularly
those emanating from anthropology. Given my background in more
traditional quantitative methods, rather than discussing the conduct
of qualitative research, I feel more comfortable discussing some of the
pitfalls and limitations of quantitative social research that make the
qualitative approach an important complement.

SCIENCE AS CONDUCTED BY MORTALS

The traditional positiyist conception of science obscures a number

of important elements of scientific endeavor. One major fiaw of this
conception is the failure to realize that science never has been and
never could be value free or value neutral (Harding, 1986, 1991). Per-
haps to some social scientists this realization is commonplace. How-
ever, a simple list of the rather obvious uses of dubious "objective
scientific facts" to argue for the inevitability of inequitable social and
economic institutions would comprise more space than this paper per-
mits. Skeptics are referred to Bayer (1981), Fausto-Sterling (1985),
Gould (1981), Kamin (1981), Lewontin, Rose & Kamin (1984), Mar-
mor, Mashow, & Harvey (1990), Mensh & Mensh (1991), Orfield &
Ashkinaze (1991), Ryan (1971), Schiff & Lewontin (1986), Schwartz
(1986), Schwarz (1988), and Tuana (1989) for a small sampling of the
ways in which class, gender and race based arguments for superiority
and/or privelege are couched in terms of so-called objective science. A
particularly profound analysis of the transformation of political econ-
omy into economic science via removal of the moral dimension is pro-
vided by Lux (1990).
Though prejudices, power politics, and assumptions play major
roles in all forms of science, they are particularly pronounced in so-
cial research. The relative lack of prestige enjoyed by social scientists
as compared with physical scientists stems partly from the fact that
social reseEirch is generally more directly connected to everyday life.
58

CONTEMPORARY FAMILY THERAPY

and is therefore more accessible to nonscientists. Few people have an

opinion about quarks; everyone has an opinion about abortion or so-
cial welfare policy. Research in the social sciences proceeds under
much closer scutiny by the public, and with much more expected,
than does natural scientific resejirch. When I say more is expected, I
mean that in general people have more emotion invested in the find-
ings of social research than they do in physical science research. Rich-
ter (1986, p. 285) notes the implications:
Once one recognizes that measures of success, in the absence
of an absolute standard, are socially negotiated quantities, a
special disadvantage of the social sciences becomes clear. The
scientific fields that can most easily claim to be successful
are those in which the measures of success are internalized,
sometimes even self-fulfilling. Astronomy defines the mea-
sures of success and succeeds. The social sciences, in con-
trast, are often faced with measures of success beyond their
control. In many cases, the larger society defines the rele-
vant problems for the social sciences and has strong preju-
dices as to what constitutes an acceptable answer. This ex-
ternalized valuation system seems to be a source of much of
the present unease in the social sciences.
The treatments of the status of social science by Cronbach (1982,
1986), D'Andrade (1986), Gergen (1985, 1986) and Shweder (1986),
among others, are attempts to begin a renegotiation of the expecta-
tions for social research. Though some (e.g., Gergen) would seem to
prefer to reverse the traditional roles and set the social sciences up in
the position of lecturer, Richter and others point to the benefits of a
more symmetric dialogue. Critical study of the conduct of physical
and natural sciences has much to offer social scientists. The impor-
tant point is to make this study truly critical, to avoid the extremes of
pretending that these fields offer us nothing of relevance or that they
establish valid norms of success that we can hope to approach only by
slavishly mimicking their methods.

EXPERIMENTAL METHODOLOGY AS
A SET OF HEURISTICS
An excellent example of the potential of this attitude is provided
by Wimsatt (1986) in his study of heuristic strategies. Though much
of his paper centers on biological topics, there is much here to be
 59

DAVID P. NICHOLS

learned relevant to social research without denying the differences of

the biological and social domains. Heuristics are particularly impor-
tant in the study of social phenomena because they are ubiquitous.
Indeed, the term heuristic originated among researchers of human
thought.
Heuristics are strategies for problem solving (in any formal or
informal domain) that are characterized according to Wimsatt by four
primary features. First, they differ from the traditional deductive
model of problem solving strategies in that they do not guarantee
correct solutions. In other words, unlike an algorithm, given valid
input and a correctly applied heuristic, invalid conclusions may re-
sult. Second, heuristics are nonetheless widely used because they are
easier to produce and apply than are perfect solutions. Third, when
they fail, they do so in systematic ways, due to the biases in their
simplifying assumptions. These biases lead to systematic errors in
certain classes of problems. Finally, application of an heuristic to a
problem alters or reformulates the question asked to a different but
intuitively related one. The important implication here is that an-
swers to the new problem may not be answers to the original one.
Among the interesting phenomena that can be at least partly
characterized as heuristics are biological, psychological, and social
adaptations. Much of the focus of Wimsatt's analysis is on the reduc-
tionist biases prevalent in population genetic research on the units of
selection controversy. The relevance of his discussion to concerns of
those who see social phenomena as often being unnaturally reduced
to collections of individual actors is made obvious by Wimsatt's com-
ments about what he terms perceptual focus: "If groups are thought of
as merely 'collections of individuals,' . . . then the description of pro-
cesses is referred to the individual level, and one cannot see that as-
sumptions that appear benign at that level may be dangerous over-
simplifications when viewed at a higher level" (p. 305).
The complexity of the social world overwhelms us. We are not
Laplacean demons, so we have to adopt some strategies for dealing
with the mass of available information. We must select certain things
to pay attention to and others to ignore. Our biological endowment
has provided us with some heuristic devices in the form of senses.
Social and cultural institutions provide us with others. In the subcul-
ture of traditional western science, the experimental method is an-
other important heuristic.
Experimental methodology comprises a powerful set of heuristics
that can be very useful when its assumptions are met. However, the
60

CONTEMPORARY FAMILY THERAPY

power comes in exchange for a monumental assumption when we

seek to establish causal relationships. This assumption is that we can
manipulate at will all of the causal factors, so as to be able to produce
all possible combinations (though we sometimes employ only a ran-
dom sample of them). When we choose not to vary some factors we
must be aware that we have reframed the general causal question
and that linkages established at one level of a causal factor must be
assumed conditional upon that state. Cronbach (1982) provides an ex-
cellent brief summary of some of the errors that can result from naive
use of the experimental method in social research.
Experimental methods also assume that we have the power to
randomly assign units of analysis to different treatment conditions.
Unfortunately, social scientists often do not have the ability to ran-
domly assign participants to conditions. We are often forced to deal
with preselected units because this is the way things work in the
domains we wish to study. If we insist on random assignment, we
often introduce another problem: our randomly assigned conditions
are not representative of the conditions to which we wish to make
inferences. As Cronbach says, "randomization almost always comes at
the expense of representativeness" (p. 64). Though we may get volun-
teers to consent to randomly selected therapeutic modalities for pur-
poses of a study, we cannot make people live this way in the world at
large.
People do not live in laboratories. Attempting to replace the
question of what happens in the social world with what happens in a
contrived experimental situation is an extremely h£izardous thing to
do. Even when we are lucky enough to be able to sufficiently approxi-
mate controlled conditions long enough to study a particular set of
interactions in a particular context, we are oflen frustrated when we
attempt to transport these findings to other contexts. As Cronbach
puts it, "Social knowledge is used in circumstances other than those
originally studied. New sites and new clienteles have to be served,
and the original site changes with time. Extrapolation is speculative
even when the effect is established beyond question in the original
context" (p. 66).
Of course most social research in the experimental tradition is
actually what is known as quasi-experimental (Cook & Campbell, 1979)
because we are explicitly aware of our inability to control some poten-
tially major causal factors. Though we have at our disposal a variety of
statistical tools that we can use in order to attempt to account for these
factors, the effort is a perilous one. Cronbach notes (p. 66).
 61

DAVID P. NICHOLS

Disregarding of causally relevant conditions is inherent in

all the time-hallowed invocations of ceteris paribus. Post hoc
matching, partialing, and covariance adjustment try to tell
us what relations would be if we could wipe out certain cor-
relations of background factors with events. In evaluations of
treatments, these adjustments seek to describe a counterfac-
tual world in which initial characteristics of persons do not
influence the treatment they get . . . To create such a world
would require intervention radical enough to denature the
treatment.
In sum, all other things generally are never equal when dealing with
social processes.
One of the most common frustrations of social researchers is the
failure of effects established in one context to replicate in other con-
texts. This lack of generalizability of effects is a major part of the
reason for the belief of some scientists that the social sciences have
not made substantial progress. A large part of this problem is due to
the use of causal reasoning that places an emphasis on particular
events at the expense of broader conditions. Cronbach (1982) notes
that "An event in a community is likely to interact with other fea-
tures of the social scene. For students of an innovation to speak of an
effect 'of the treatment' is therefore shortsighted. The circumstances
surroimding the intervention are part of the cause" (pp. 66-67).
Given that a number of conditions Eire associated with a particu-
lar outcome, the selection of a small subset of these as the important
causal factor(s) is an heuristic device. In the case of most causal
thinking in science and the everyday world, the bias in this heuristic
tends strongly toward choosing as small a number of "causes" as pos-
sible (often only one) and to select these from among the unusual
features of the situation; that is, to ignore that which is normal and
ordinary. Though very common, this is by no means generally justi-
fiable (Salmon, 1984). To Cronbach "The asjnnmetric emphasis on the
treatment as the cause, rather than on the combination of treatment,
units, setting, and so on, arises from the intent of social scientists to
aid in the manipulation of human affairs" (1986, p. 94).

STATISTICAL CONTROLS AND THE SELECTIVITY

OF SOCIAL PROCESSES
We have seen that the direct application of the pure experimen-
tal method to most social research is impossible; quasi-experimental
62

CONTEMPORARY FAMILY THERAPY

methods are thus thought to be the next best thing. If we can't control
all the important factors directly, we can at least control those of in-
terest directly and control for others statistically, or so the reasoning
goes. In other cases we look to "natural experiments," where changes
in purported causes of interest happen without our intervention, and
again attempt to control for nuisance factors through statistical con-
trols. But just how valid is this reliance on statistical control? In a
monograph that should be required reading for all students of social
processes, Lieberson (1985, p. 19) demonstrates that this approach is
fi-aught with peril.

What is wrong with this procedure? If there is any feature of

social life about which a high degree of confidence exists, it
would be this simple principle: social processes are selective
processes. As a consequence, the reason for taking into ac-
count the differences found between the populations is also
the very same reason for doubting whether such efforts can
be successful very often. For selective processes are probably
operating within the control variables themselves.

Most social scientists, if asked, would say that the application of

control variables in quasi-experimental or field research is at worst
benign, and generally to be required. As Lieberson shows however,
this belief is based on mistaken assumptions about the effects of con-
trol variables in the presence of unmeasured selectivity. The conse-
quences of applying the control method "are quite contary to the stan-
dard assumptions in social research. Under some conditions, the
application of controls generates results that are actually farther re-
moved from the truth than would occur if no controls were applied
whatsoever" (p. 22).
This problem, that applying controls is not guaranteed to be help-
ful, or even benign, is not linked to the use of any specific statistical
procedures, but results from a more fundamental problem: without
knowing the reasons for the interrelationships among the indepen-
dent variables in a functional model, we have no way of knowing the
implications of applying controls. Controls can only be assumed to be
benign when they have no effect on the apparent relationship be-
tween the independent and dependent variables. Thus when they are
of the most potential usefulness, they are also the most dangerous.
One of the meiinstays of the scientific enterprise is replication. It
is widely assumed that if a causal relationship exists, repeated
 63

DAVID P. NICHOLS

studies will confirm it. Unfortunately, this is not necessarily the case.
According to Lieberson, replication won't always tell us what's real
and what's chance or artifact because unmeasured selective forces
will often differ across contexts and replications and thus produce dif-
ferent results. "Further tests, follow-up studies, newer statistical pro-
cedures, exchanges in the journals, bigger samples, more controls,
and the like, may or may not support the original conclusion, but it
does not really matter if they are all based on the same underlying
false premise" (p. 39).
Even the latest and greatest of the super statistical methods such
as structural equation modeling are not up to the task in the presence
of unmeasured selective processes. The current rage for such models
belies the fact that the concept of path models has been around for
more than 70 years (see Wright, 1921) without solving the basic prob-
lems associated with drawing causal inferences from associations or
correlations. Firm causal inference comes from theory and design, not
from statistics. As Cronbach (1982) says, "These advances should in-
crease the yield from field data, but I cannot share the optimism that
superadjustments will warrant causal conclusions. Curve fitting can-
not lift us into orbit unless fueled by substantive insight" (p. 67).

OTHER DANGERS IN
QUASI-EXPERIMENTAL RESEARCH:
CONTAMINATION, ASYMMETRY AND
LEVELS OF ANALYSIS
At the risk of some redundancy, I think it is important to sum-
marize further points made by Lieberson (1985). He lists four other
problems that haunt quasi-experimental research. These by no means
exhaust the threats to validity of quasi-experimental research (see
Cook & Campbell, 1979), but considered in conjunction with the selec-
tivity issue they provide a powerful argument for the need to recon-
ceptualize traditional quantitative social scientific research meth-
odology.
The first hazard, contamination, occurs when the presence of or
change in an independent variable has effects where it is not present
or does not change. Examples would be the impact of tax laws or
professional licensure requirements in some jurisdictions on the be-
havior of governing bodies or professionals in other places. Effects of
the measures on behavior within the original jurisdictions would not
64

CONTEMPORARY FAMILY THERAPY

be ascertetinable by comparison with areas without the measures, be-

cause the baseline areas have also been affected.
The second problem is that of asymmetric causality. In general,
social researchers assume that changes in a dependent variable re-
sulting from changes in one direction of an independent variable can
be reversed by moving the independent variable back to the original
state. However, this is quite often not the case with social processes.
For example, the conditions which give rise to social movements can
and often do change without the social movements disappearing. Pro-
cesses tend to take on a life of their own once set in motion, and often
C£in be only partially reversed if at all.
The third issue is the confusion of proper level of analysis or the
assumption that processes functioning in a certain manner at one
level will function analogously at another level. The most common
version occurs when reductionist assumptions are made, as discussed
by Wimsatt (1986), and systemic properties are ignored in favor of
concentration on individual parts of a system. A great deal of psycho-
logical theory and research falls prey to the fallacious assumption
that social processes can be accounted for by studying isolated indi-
viduals. The field of family therapy is in large part a result of the
realization that humans as social beings exist within larger social
systems, the family being a crucial one.

ANALYSIS OF VARIATION AND

ANALYSIS OF CAUSES

The fourth issue, that of the role of variation in causal and statis-
tical analysis, is so important and so widely misunderstood as to war-
rant extensive attention. As anyone who has had an introductory
course in statistics knows, statistical methods cannot be applied
where variation does not exist. The almost universal requirement in
empirical social research to apply statistical methods if one wishes to
have work seriously considered has important implications for our
selection of research problems and our conceptualization of causal
processes. As Lieberson says, "one must ask whether problems are
sometimes posed in a certain way more because they meet statistical
criteria and needs than because they are the most appropriate re-
search question in terms of either substantive or theoretical issues"
(p. 89).
He goes on to point out that issues of variability should be sec-
ondary to more basic issues of underlying causal forces.
 65

DAVID P. NICHOLS

It is premature to think about variability in an event before

knowledge is developed about the fundemiental cause of the
event itself Explanation of a variable's variation should not
be confused with an explanation of the event or process itself.
Contrary to the assumption commonly made in social re-
search, we cannot learn the fundamental cause of an event
by studying the factors affecting its variation; at best, we can
only learn how secondary characteristics might modify the
consequences of a fundamental cause (p. 115).

Choosing research problems focused on accounting for variation in a

variable of interest will not necessarily lead to knowledge about fun-
damental causes because "there is no intrinsic reason for expecting
the properties of the system to be logically deduced from observing
variation in the behavior of the individual units" (p. 116).
Standard reporting of research results often centers around the
attribution of variation in the dependent variable as being explained
by variation in one or more independent variables. However, the as-
sumed implications of the observed associations are often (if not al-
ways) in error.

It is erroneous to assume that the importance of causal

forces—in either an absolute or relative sense—can be deter-
mined by the amount of variation "explained" by them in
some empirical study. The variance and range of both the
dependent variable and the independent variables will al-
most certainly vary from setting to setting, and in turn this
will affect the relative importance attributed to each inde-
pendent variable (p. 117).

Contrary to the assumption underlying most empirical social re-

search, "The analysis of differences or variability is an analysis of
just that—it is not to be confused with the analysis of causes, espe-
cially those that are so widespread that they help explain very little
of the variation" (p. 119). These same points are also well illustrated
by Lewontin (1974).
The fiaw in current practice is due I think not so much to the
basic underlying causal model as to the assumptions made about the
applicability of the model in practice. That is, the model being as-
sumed says that if we could look at the state of the dependent vari-
able of interest under all possible combinations of states of the caus-
ally relevant factors, we would be able to associate a particular state
(under a deterministic model) or distribution of states (under a proba-
66

CONTEMPORARY FAMILY THERAPY

bilistic model) with each combination of factors. Under such circum-

stances of omniscience, we could thus determine all the possible vari-
ations in states of the dependent variable and how they are linked
with combinations of the causal factors.
The error occurs when we make one of two assumptions about
our ability to infer from observed combinations of factors. The more
general assumption is that we can in essence infer the entire causal
system from the nonrandom subset that we observe. This assumption
is always incorrect and not likely to be defended for long by those to
whom it is accurately portrayed. The more common case is to assume
that we can at least infer from the observed combinations to the
larger subset of plausibly produceable combinations. While more
likely to be defended than the more general assumption, the underly-
ing logic is the same and therefore similarly fallacious.
As Cronbach (1982) was quoted as saying earlier, our desire to
intervene in human affairs leads to an asymmetric focus on that
which we deliberately manipulate, pushing off as background the set-
ting or context within which our manipulations occur. However, as he
also says, these conditions are just as much a part of the cause of a
particular outcome as are the "treatments" we apply. Should we find
ways of varying these "background" constants, we can produce differ-
ent outcomes.
I have spent so much time on this problem because I believe that
this erroneous approach to understanding causal relationships is ex-
tremely widespread among social scientists. As evidence I offer the
following passage from an article discussing the relative importance
of genes and environment for a variety of behaviors (Plomin, 1989).
Note how the focus on attempting to account for differences in behav-
iors via differences in causal factors produces a completely nonsensi-
cal statement:
The importance of nonshared environmental factors suggests
the need for a reconceptualization of environmental influ-
ences that focuses on experiential differences between chil-
dren in the same family. That is, many environmental fac-
tors differ across families; these include socioeconomic
status, parental education, and childrearing practices. How-
ever, to the extent that these environmental factors do not dif-
fer between children growing up in the same family, they do
not influence behavioral development [italics added]. The crit-
ical question becomes, Why are children in the same family
so different from one another? The key to unlock this riddle
is to study more than one child per family. This permits the
 67

DAVID P. NICHOLS

study of experiential differences within a family and their

association with differences in outcome (p. 109).

The mistake here of course is the belief that because we cannot

associate differences in a factor with differences in an outcome, the
factor plays no part in the outcome. This expresses itself in the ital-
icized statement above, the obviously erroneous logic of which infers
that any treatment, however profound, that is applied identically to
all of the children in a family will have no effect on their behavioral
development.
Here we have methodology dictating theory, producing truly ab-
surd results. What is worse is that this article appeared in a special
child development issue of the widely read American Psychologist,
and produced no comment (at least none that was published) concern-
ing this fundamental logical blunder. It would appear that the fallacy
of explaining events only through observed variations, while ignoring
observed constants that could vary and would produce different re-
sults if they did vary, is widespread among far more social scientists
than I would have hoped (unless of course a great many people also
noticed this and decided it was not worthy of a response, or as I did,
decided to see if anyone else noticed it). I had long feared this, but
never expected such a graphic illustration.

WHITHER VALID QUANTITATIVE

SOCIAL RESEARCH?
So far we have seen an overwhelming array of threats to the val-
idity of quantitative social research (some of these are of course rele-
vant to broader classes of research methodology). Are we forced to
surrender and simply give up on quantitative social research? I don't
think so. As Lieberson (1985) says, "a rigorous science of society is an
enterprise that is well worth pursuing. The goal is not to show that
social science is intrinsically unobtainable. Rather, it is to eliminate
the unthinking imitation of a crude physical science model that char-
acterizes so much of social research" (pp. 171-172).
Some obvious implications are that we should not pretend that
we have the equivalent of controlled experimental conditions when
we do not, or that context specific relationships are context free. This
means that many current research programs should be reconcep-
tualized as exploratory and descriptive, rather than confirmatory.
There is nothing wrong with this; accurate description, both qualita-
68

CONTEMPORARY FAMILY THERAPY

tive and quantitative, is exceedingly important in any scientific pro-

gram, and contrsuy to popular opinion, we do not have to do experi-
ments to be doing science. To practice science relevant to society re-
quires moving beyond the constricting bounds of the hypothetico-
deductive framework. "To grapple with loosely bounded problems—
that is, with almost any problem that connects up with community
concerns—we need to blur the lines that separate 'values' from 'facts,'
'humanities' from 'sciences,' and 'quantitative' from 'qualitative' or
'applied' from 'basic' research" (Cronbach, 1986, p. 102).
Another recommendation Lieberson makes is to look more care-
fully for what he calls transformational causal principles—basic
causes that produce the same outcomes under a wide variety of condi-
tions. This is obviously much more easily said than done, but as we
have seen, the current method of trying to derive the basic causal
mechanisms from observed associations in particular contexts does
not offer much hope. Lieberson believes this to be due to the fact that
we are looking at superficial linkages that often change while basic
causal principles do not.
His most important suggestion is "a shift in focus from the depen-
dent variable to the independent variables. In effect, one needs a the-
ory or model that accounts for the nonrandom assignment of the con-
trol variables themselves" (p. 211).
We need a theory of data, not merely in the conventional
epistemological sense, but one that applies to the specific
problem. If comparisons are made, we cannot assume that it
is mere chance that some comparative possibilities exist and
others do not. If X varies from one place to another in its
level, it is just as necessary to ask why it varies as to ask
what its apparent infiuence on Y is. For it is almost certain
that the level of X found in each setting is not due to chance.
It is part of the phenomenon under study—every bit as much
as the apparent dependent variable (p. 229).
Following these prescriptions will require a great deal of "soft"
work, constructing theories to account for why we see certain combi-
nations of various entities and not others, or why some are common
and others are rare, as opposed to the "hard" science work of testing
deductions from such theories. This does not mean that we never test
hypotheses, only that we do a lot more work than is the current norm
before proceeding to the confirmatory stage. Paraphrasing Shweder
(1986), without the soft side, there is no hard data.
The need for "soft" qualitative work to accompany "hard" quan-
 69

DAVID P. NICHOLS

titative work is the result of an important hasic principle: since we

must always view the processes of life from some particular perspec-
tive, we must vary that perspective if we are to escape its inherent
hiases. Those relationships that persist as we vary our perspective
more and more are the most certain, the most "rohust" (Wimsatt,
1986). "By using a variety of different models, approaches, means of
detection, or of derivation and comparing the results, we can hope to
detect and correct for the biases, special assimiptions, and artifacts of
any one approach" (p. 308).
This search for robustness via "triangulation" (Campbell &
Fiske, 1959) means that no single academic or professional discipline
can he counted on to provide adequate understanding of social phe-
nomena (Cronhach, 1986). "We can rarely see a topic in proper per-
spective if our inquiry employs resources from only one discipline" (p.
98).
Moreover, even within disciplines, we must have diversity in the
perspectives that are hrought to bear on social phenomena. As Fiske
(1986) points out, "Each investigator can be viewed as a measuring
procedure, as an instrument for obtaining knowledge" (p. 67). Differ-
ences in individual perspectives as a result of differing experiences
and backgrounds are essential in order to remove the hiases resulting
from relative homogeneity of characteristics of researchers.
This is not to say that no standards are to be applied to argu-
ments for the importance of different perspectives or interpretations.
As noted earlier, relinquishing Absolute Objectivity does not mean
embracing total relativism. Or as Cronhach (1986) puts it, "To advo-
cate pluralistic tolerance of alternative accounts is in no way to advo-
cate tendermindedness" (p. 104).
As Wimsatt (1986) warns, however, we must understand that tri-
angulation is itself an heuristic device. It will fail to the extent that
we do not include sufficient variation in the hiases and assumptions
associated with our employed perspectives. We must he careful to en-
sure that hidden hiases common to multiple approaches do not under-
mine the validity of our supposedly rohust observations.

CONCLUSION: NOT A SOCIAL PHYSICS, BUT

RATHER A SOCIAL SCIENCE OF THE POSSIBLE
Earlier I discussed the renegotiation of the role of social research
suggested hy several authors and the fact that social researchers are
often researching questions that are dictated largely hy societal con-
70

CONTEMPORARY FAMILY THERAPY

cems. An importemt issue here is that of consideration of what is "do-

ahle" and what is not. Lieberson (1985) addresses this issue in his
introduction. "Are there questions currently studied that are ha-
sically unanswerable even if the investigator had ideal nonexperi-
mental data? If so, what are the alternative questions that can be
dealt with successfully by empirical social research, and how should
they be approached" (p. 8)?
I think the answer is obviously that there are a great many ques-
tions of interest which we cannot answer. Identifying the alternative
questions that are answerable and clarifying their relationships to
other questions are important steps. This will sometimes lead us to
the point where we are able to answer important but previously un-
answerable questions. In other cases it will demonstrate to us that
some questions, while important, as simply not answerable. As Lie-
berson concludes, "we have to develop a better grasp of what we want
to explain and what is potentially explicable at a given point of
knowledge . . . What is it that we want to explain? But first, what is it
that can be explained?" (p. 235).

REFERENCES
Bayer, R. A. (1981). Homosexuality and'American psychiatry: The politics of diagnosis.
Princeton, NJ: Princeton University Press.
Bernstein, R. J. (1976). The restructuring of social and political theory. Philadelphia:
University of Pennsylvania Press.
Bernstein, R. J. (1983). Beyond objectivism and relativism: Science, hermeneutics, and
praxis. Philadelphia: University of Pennsylvania Press.
Bock, R. D. (1975). Multivariate statistical methods in behavioral research New York-
McGraw-Hill.
Camphell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation hy the
multitrait-multimethod matrix. Psychological Bulletin, 56, 81-105.
Capra, F. (1975). The tao of physics. Toronto: Bantam Books.
Capra, F. (1982). The turning point. Toronto: Bantam Books.
Cook, T. D., & Camphell, D. T. (1979). Quasi-experimentation: Design & analysis issues
for field settings. Boston: Houghton Mifilin Company.
Cronhach, L. J. (1982). Prudent aspirations for social inquiry. In W. H. Kruskal (Ed.),
The social sciences: Their nature and uses (pp. 61-81). Chicago: University of Chi-
cago Press.
Cronhach, L. J. (1986). Social inquiry hy and for earthlings. In D. W. Fiske & R. A.
Shweder (Eds.), Metatheory in social science: Pluralisms and subjectivities (pp.
83-107). Chicago: Universi^ of Chicago Press.
D'Andrade, R. (1986). Three scientific world views and the covering law model. In
D. W. Fiske & R. A. Shweder (Eds.), Metatheory in social science: Pluralisms and
subjectivities (pp. 19-41). Chicago: University of Chicago Press.
Fausto-Sterling, A. (1985). Myths of gender. New York: Basic Books.
 71

DAVID P. NICHOLS

Fiske, D. W. (1986). Specificity of method and knowledge in social science. In D. W.

Fiske & R. A. Shweder (Eds.), Metatheory in social science: Pluralisms and subjec-
tivities (pp. 61-82). Chicago: University of Chicago ftess.
Fiske, D. W. & Shweder, R. A. (Eds.). (1986). Metatheory in social science: Pluralisms
and subjectivities. Chicago: University of Chicago Press.
Gergen, K. J. (1985). The social constructionist movement in modem psychology.
American Psychologist, 40, 266-275.
Grergen, K. J. (1986). Correspondence vs. autonomy in the language of understanding
human action. In D. W. Fiske & R. A. Shweder (Eds.), Metatheory in social science:
Pluralisms and subjectivities. Chicago: University of Chicago Press.
Gould, S. J. (1981). The mismeasure of man. New York: W. W. Norton & Company.
Harding, S. (1986). The science question in feminism. Ithaca, NY: Cornell University
Press.
Harding, S. (1991). Whose science? Whose knowledge?: Thinking from women's lives.
Ithaca, NY: Cornell University Press.
Huhner, K. (1983). Critique of scientific reason (P. R. Dixon, J. & H. M. Dixon, Trans.).
Chicago: University of Chicago Press. (Original work 2nd ed. puhlished 1979)
Kamin, L. J. (1974). The science and politics oflQ. Potomac, MD: Lawrence Erlhaum
Associates.
Kerlinger, F. N. (1964). Foundations of behavioral research. New York: Holt, Rinehart
and Winston, Inc.
Kruskal, W. H. (Ed.). (1982). The social sciences: Their nature and uses. Chicago: Uni-
versity of Chicago Press.
Kuhn, T. S. (1962). The structure of scientific revolutions. Chicago: University of Chi-
cago Press.
Lakatos, I. (1970). Falsification and the methodology of scientific research pro-
grammes. In I. Lakatos & A. Musgrave (Eds.), Criticism and the growth of knowl-
edge (pp. 91-196). Camhridge: Camhridge University Press.
Lakatos, I., & Musgrave, A. (Eds.). (1970). Criticism and the growth of knowledge.
Camhridge: Camhridge University Press.
Laudan, L. (1984). Explaining the success of science: Beyond epistemic realism and
relativism. In J. T. Cushing, C. F. Delaney, & G. M. Gutting (Eds.), Science and
reality (pp. 83-105). Notre Dame, IN: University of Notre Dame Press.
Lewontin, R. C. (1974). The analysis of variance and the analysis of causes. American
Journal of Human Genetics, 26, 400-411.
Lewontin, R. C , Rose, S., & Kamin, L. J. (1984). Not in our genes: Biology, ideology,
and human nature. New York: Random House.
Lieherson, S. (1985). Making it count: The improvement of social research and theory.
Berkeley: University of California Press.
Lux, K. (1990). Adam Smith's mistake: How a moral philosopher invented economics
and ended morality. Boston: Shamhhala.
Marmor, T. R., Mashaw, J. L., & Harvey, P. L. (1990). America's misunderstood welfare
state: Persistent myths, enduring realities. New York: Basic Books.
Mensh, E., & Mensh, H. (1991). The IQ mythology: Class, race, gender, and inequality.
Carhondale, IL: Southern Illinois University Press.
Orfield, G., & Ashkinaze, C. (1991). The closing door: Conservative policy and black
opportunity. Chicago: University of Chicago Press.
Pirsig, R. M. (1974). Zen and the art of motorcycle maintenance: An inquiry into values.
Toronto: Bantam Books.
Plomin, R. (1989). Environment and genes: Determinants of hehavior. American Psy-
chologist, 44, 105-111.
Richter, F. M. (1986). Non-linear behavior. In D. W. Fiske & R. A. Shweder (Eds.),
Metatheory in social science: Pluralisms and subjectivities (pp. 284-292). Chicago:
University of Chicago Press.
72

CONTEMPORARY FAMILY THERAPY

Rorty, R. (1979). Philosophy and the mirror of nature. Princeton, NJ: Princeton Univer-
sity Press.
Ryan, W. (1971). Blaming the victim. New York: Vintage Books.
Salmon, W. C. (1984). Scientific explanation and the causal structure of the world.
Princeton, NJ: Princeton University Press.
Schiff, M., & Lewontin, R. (1986). Education and class: The irrelevance oflQ genetic
studies. New York: Oxford University Press.
Schwartz, B. (1986). The battle for human nature: Science, morality and modern life.
New York: W. W. Norton.
Schwarz, J. E. (1988). America's hidden success (rev. ed.). New York: W. W. Norton.
Shweder, R. A. (1986). Divergent rationalities. In D. W. Fiske & R. A. Shweder (Eds.),
Metatheory in social science: Pluralisms and subjectivities (pp. 163-196). Chicago:
University of Chicago Press.
Toulmin, S. (1972). Human understanding: The collective use and evolution of concepts.
Princeton, NJ: Princeton University Press.
Tuana, N. (Ed.). (1989). Feminism and science. Bloomington: Indiana University Press.
Wimsatt, W. C. (1986). Heuristics and the study of human hehavior. In D. W. Fiske &
R. A. Shweder (Eds.), Metatheory in social science: Pluralisms and subjectivities
(pp. 293-314). Chicago: University of Chicago Press.
Wright, S. (1921). Correlation and causation. Journal of Agricultural Research, 20,
557-585.