29
AN APPROPRIATE METHODOLOGY FOR
THE SOCIAL SCIE1ESl
Milton L. Andersen
University of California
Berkeiey
The posltivist doctrine is the philosophical position with which
most American social scientists are acquainted.  Unfortunately, there is
little awareness of the limitations and implications of the positivist
position, or of any viable alternative to it.  There are many difflculties
created by an embracement of the positivist position; some of them are
especially inhibiting to psychological and social research. This paper
will attempt to present a sketch of arn alterative view of proper scien-
tific procedure that would not inhibit us so unduly in cur theoretical and
research efforts.
The discussion presented wlll draw heavily from the works of Karl
Poppero  The Oen S         and Its Enemies, 1950, The Poverty of Historicism,
1957, The Logic of Scientific Discoery, 1959, and Conjectures and Refuta-
tions, 1962. The writings of Popper are extremely relevant to contemporary
styles of thought in the American social sciences. He has provided over
the last thirty years a vigorous and comprehensive critique of the positi-
vist doctrine.   His writings, and also the writings of many others asso-
ciated with the University of London, constitute a strong challenge to some
of our most basic assumptions aDout proper scientific procedure.  Popper's
critique of contemporary practices flows from a consideration of the logic
of scientific discovery and validation, and not simply from what he might
wish scientific actlvrity to be. His viewpoint is derived from a considera-
tion of what is logically entailed in certaln kinds of statements we might
choose to make. An impllcation of his approach is that we need to accept
very few principles from the philosoph7y of science. The two most basic
principles are:  (1) Our theories and statements must always be open to
empirical testing (falsifiability) and to free, critical discussion; and
(2) We must specify the universe to which we wlsh to generalize from our
sample data.
The first requirement, that of submitting our theories to empirical
test, is simply a convention we have adopted. The adoption of this conven-
tion has been very fruitful, and we could adopt other conventions to help
us decide which theory should be regarded as the best. However, I would
argue that we should retain this conventlon and continually strive to put
our ideas, and hypotheses, into empirically testable forms.
In any event, we need not accept a doctrine of positivism which
restricts us in our research, and dampens the sociological imagination.
Besides, much of the philosophy of science talked about today is at least
twenty years out of date. WJe need a methodology relevant to our problems,
and I think that with a new view of what science is all about, we can
begin to develop a methodology that helps instead of hinders.  To present
this new view, we begin with the problem of induction.
30
The Problem of Induction, or: The Non-Recognition
of Using Other Information
A common procedure in the social sciences is to take a sample of
observations or perform an experiment, and from the results obtained, to
make inference about a "population."  This is a process of inductive infer-
ence, and is regarded as proper scientific procedure. We state that on the
basis of our sample, we ca'n mke such-and-such statements about the popula-
tion. However, consider the following example. Suppose that I have col-
lected the traditional epistemological sample of ravens, and all of them are
black. This Information 12 itself does not provide any basis for making
inferences about the blackness of ravens elsewhere, or at different times.
It is on the basis of other information that inferences from a sample of
observations to the unobseFved elements in a population can be made. In
fact9 a great deal of other information is needed simply to determine what
to specify as being the universe or "population."  It is only with other
information that we can generalize beyond our sample of observations.
There is no principle of induction which pemits this great leap. Even
Hume rejected induction as a form of valid inference.  ?Humeg I felt, was
perfectly right in pointing out that induction cannot be logically justi-
fied. He held that there can be no valid logical arguments allowing us to
establish 'that those instances, of which we have had no experience,
resemble those, of which we have had experience'" (Popper, 1962, po 42)o
Often, in place of other information, a certain assumption is made.
This assumption is that the conditions under which the individuals in our
sample were functioning are the same conditions under which the remaining
indivriduals in the population function. It is our failure to recognize
the utilization of this assumption, or the use of other information, that
leads us to believe in the validity of induction as a proper form of gener-
alization.
This use of inductive inference involves an additional, and very
serious disadvantage, namely: that by not specifying the conditions under
which our inductive inference is to hold, and under which conditions it is
not to hold, we are not able to determine functional or ecological relation-
Thlps. We are then confined to making "abstractions," or statements about
the properties of the objects in the sample. For example, if we were to
examine the conditions under which ravens are black, and when they are not
black, we might find that blackness in ravens is a function of climate, or
of diet, or of genetic mechanisms. The inductive procedure, and its
abstractive counterpart, seems to be part and parcel of the Aristotelian
procedure that Kurt Lewin (1935) criticized so severely. This belief in
"abstractioft as a proper form of generalization from a sample to a popula-
tion effectively prevents us from determining the ecological or environmen-
tal determinants of human conduct. A common practice is to obtain a cor-
relation coefficient indicating the degree of relationship between two
processes. This correlation coefficient is a summary statement,, and it
alone provides no information other than an estimate of the degree of rela-
tionship. It would seem more fruitful to determine the situations within
which the relationship holds, and those in which it doesn't hold.
An illustration of the non-generalizability via inductive procedures
would be the "Back from Mars" game. Suppose that an unmanned space ship has
31
probed the surface of Mars. On its return, the shlp was destroyed before
reaching earth, but was able to relay ahead the information that in a set
of n objects, each had the property y. This alone is very meager informa-
tion, and nothing much could be inferred from it. We would need a great
deal of other information to learn anythlng about these objects, such as:
the length of time the ship was on Mars, the extent and nature of the area
probed, conditions in the ship, the length of time the ship had traveled,
other information relayed home, etc.  It is important to notice that the
comparing of different types of information, all focussed on one problem,
is a deductive, analytical procedure, and is certainly not an inductive
process.
I know that the story of the ravens is ridiculously simple, and no
one today would believe in this simple form of inductive inference. Also,
probability theory and statistical procedures are much more complicated
than my arguments here imply0  However, Inductive inference is often re-
introduced through the use of probabli'ty notions.  To illustrate this,
let me use an example taken from Popper ('957a).
Suppose that we have flipped a coln one milllion times, and the
number of heads obtained Is about one-half the number of flips.   If we are
asked to estimate the probability of getting heads on another toss of the
coin, we would most likely say that it is one-half. But this probability
estimate is based or our as-tion that the ecology will remain constant0
That is, we assume an unbiased coin, and a flat landing surface. We could
construct an ecology so that 1the coln always lands on edge, and a head or
tail is never obtained0  A table could be made with deep, narrow slots that
are wide at the mouth, so that the coin always slides into a slot and comes
to rest upon its edgeO  With this ecology, the probabillty of getting heads
is zero. The importance of this example is to illustrate that we cannot
smuggle in inductilve styles of inference via probabilllty notlons0  Our
inferences concerning the behavior of an object are based upon assumptions
or inforration we have about its ecology0  The ecology, or experimental
set-up generates the obtained probabllitieso  LI addition to the invalidity
of simple inductive inference, it often keeps us from undertaking a criti-
cal analysis of ecological structure, even wlth such a simple ecology as
that used in coin-tossing.
However, a certain kind of statement can properly be made from a
sample of observations. We will see that the only proper statement is one
which falsifies or refutes a general statement, but does not verify a
general statement.
Universalistic vs0 Existentialistic Statements,
and Falsifiability vs. Verifiability
We can clarify the type of statements that are permissible from
finite samples by distinguishing universalistic from existentialistic state-
ments. A uni'versalistic statement is here defined as one asserting "for all
places and for all times."
An existentialistic statement on the other hand, is one which
asserts only that somethlng exists, or that something has occurred, or that
:a certain finlite set of objects have certain properties.  For example,
32
suppose that we have collected our favorite sample of ravens aga in, in
which all or some of them are black. Then, we have verified the existen-
tialistic statement that there exist some black ravens, but we have falsi-
fied the universalistic statement that all ravens are white. Existential-
istic statements can be verified, but not falsified. To falsify an
existentialist'ic statement would necessitate an exmination of the entire
universe. That is, to falsify the statement that "there exist white ravens,"
would require, logically, looking into every nook and cranny of the universe
to see if any can be found. On the other hand, universalistic statements
can be falsified, but not verified. The mark of a universalistic statement
is its falsifiability.  This flows logically from its quality of "allness."
This "allness" permits us, as is well known, to falsify a universalistic
statement with a single case.
It is important to mention that although we speak of "universalis-
tic" statements, the universe referred to must be specified. That is, we
might specify universes such as: all females, all Europeans, all English-
speaking persons, all the members of a given culture, all persons exposed
to certain experimental conditions, and so forth. A certain confusion might
occur between stril    universalistic and numerically universalistic state-
ments. The term ''all Europeans" could refer to all Europeans presently
living.  This would be a numerically universal statement.  Or, the term
"all Europeans" could refer to Europeans who lived in the past, those now
alive, and those in future times.  This latter is a strictly universal use
of the term mall Europeans."  In any event, the "allnesst of a universalis-
tic statement refers to all possible cases within the boundaries of the
specified universe.
The term "unmiversalistic" does not refer to how frequently some
event will occur, nor does it 'imply that something occurs throughout the
universe. The term "universalistic" should indicate that under certain
specified conditions (no matter how rare or infrequent these conditions
might be) certain events will universally occur. We can again use our
ravens as an example. The statement that "all ravens are black" implies
that if you find a raven, you will universally find it to be black. The
statement does not assert anything about how many ravens (if any) exist in
the world, nor does it assert that they exist universally throughout the
world.
Many theories in the social sciences are in a universalistic form,
although limits or boundaries are unspecified, or are silently assumed. A
neglected but very enlightening enterprise is to consider     we should ex-
pect an assertion to hold only within specified boundaries. The boundaries
might delimit a universe which includes only college sophomores, enrolled in
Psychology IA,, attempting to act as they think they should in the psycho-
logical laboratory. Or, the universe might include only white, English-
speaking, middle-class, Protestant Americans. Consider some of the many
possible universes one might be referring to when making an assertion about
human conduct. The assertion could refer to: (1) All humans, (2) all
humans within a given historical period; (3) all humans within a given cul-
ture or soclety, (4) all humans within a certain sector of society, say
lower-class persons, (5) all humans with a given characteristic, say those
scoring over 100 on an I.Qo test, or (6) all the behavior of a single indi-
vidual, (7) all the behavior that occurs in a certain type of situation,
33
say the school classroom; and so forth. We should always attempt to specify
the universe.  We should not assume that our assertlon will hold beyond the
specified universe, and we should investigate to find out where our asser-
tion holds and where it doesn't hold. The irnvestigation may prove to be
devastating to our theories, buit should prove illumlnating to ourselves and
to our colleagues.
A theoretical statement., then, is a statement which refers to a
specified uni'verse, and is stated in such a way that operations can be
undertaken in an attempt to falsify the statement. A strong theory is one
that has withstood severe criticism and many attempts at falsif'ication.
Its strength is in having survived many attacks upon It. However, the sur-
viving of numerous critical tests does not verlfy, confirm, or prove the
theory. A scienttific theory is one that remains open to criticism and
attempts at falsification0 Once a theory is regarded as proven or verified,
it is thereby removed from the sclentlfic a-:ena, from the arena  of critical
debate and empirical testing.
It may be very disturbing to realize that a powerful theory asserts
very little, but denies a great deal. An example could be made using again
our sample of ravens. The statement that "all ravens are black" has more
power or informative content than the statement "all ravens are non-white."
In the latter statement we are denylng only that white ravens will be found,
but in assert'ing that "all ravens are black," we are also denying that any
green, red, yellow, blue, brown, purple, or other non-black ravens shall be
found.
We are disturbed by the suggestlon that a theory denies certain
events, because of our mistaken belief that theories can be verified.    It
might prove even more disturbing to realize t.hat a powerful or informative
theory is also very improbable. Conslder this example adapted from Popper
(1962, pp. 217-218).  I might assert that i w Il rain on Friday, and this
assertion has a certain objective probabi-i'ty. That is, we can estimate,
from weather bureau records and reports, the probability of rain on Friday.
I might further assert that it wlll rain on Friday, and it wlll rain between
2 p.m. and 3 p.m.  This assertlon has a smaller objectlve probability, but
greater informative content.   I might further assert that it will rain on
Friday, between 2 and 3 p.m., and that it will rain more than one inch.    As
the content of a theory increases, its objective probability decreases.
It should be clear that the use of operatlonal definitions, although
very useful in specifying what research procedures we have followed should
not be used to define concepts or to define variables. The logic of opera-
tional definition and the logic of inductive inference are the same.    We
cannot validly generalize from a very specific sltuatlon or arrangement of
conditions to a concept or a universal statement. If our goal is the forma-
tion of sci entific theory (iOe.--universalistic statements), then we must
avoid the use of 'inductive procedures. Inductive inference could be des-
cribed as "intellectual provincialism," in that the use of induction assumes
that the limited portion one has seen of the world is true elsewhere, with-
out one having been elsewhere to see if this inference is true or not.
There are very few universalistic statements in a clear, testabl  e
form to be found in the social sciences. These statements are very difficult
34
to construct when one is concerned with problems of social behavior. The
difficulties which arise are too well known to require mention. In fact,
one writer, R. S. Peters, has asserted that psychology will never have
highly general theories (i.e. universalistic statements). The reason for
his assertion is not that practical and experimental problems are overwhelm-
ing, but that these highly general theories are logically impossible.
* . . the role of the Galileo of psychology mus be  orever unoccupied.
Psychology has not soared into its Galilean period as is often thought,
through lack of bright ideas, experimental ingenuity, or methodological
rigour. It has remained earth-bound in mazes and Skinner boxes because the
highly beneral theories which, it was honed, would emerge, are logically
impossible" (R. S. Peters, 1960, p. 1956).
Although universalistic statements may be rare in the social scien-
ces, we are, however, accumulating a fund of existentialistic statements.
In this manner, we are building a body of knowledge. Because these state-
ments are not universalistic ones, we should not belittle them, nor should
we underrate their power. These existentialistic statements can, it must
be remembered, falsify universalistic statements. It is in their role as
potential falsifiers of sweeping general statements that our body of exis-
tential statements can play a most important role. Consider the great mass
of myths, superstitions, prejudices, so-called "common sense," and unexamined
assumptions that flourish outside (and inside) the world of social science.
As examples, take certain beliefs about race or beliefs about "human nature."
The role of social science has been to caution us about making any easy
generalizations about human behavior. We have become aware of the extreme
complexity, variability, and adaptability of human behavior. Many social
scientists are now engaged in research in areas where previously only un-
tested theorizing and unquestioned assumptions held sway. Such areas as
race relations, belief systems, social stratification, social deviance, and
so forth, are the areas now under the critical scrutiny of the social scien-
tist. In particular, the role of social sciences has been a most signifi-
cant one in the investigation of fascistic beliefs during World War II, and,
during the last decade, in producing evidence to support the many desegrega-
tion decisions of the federal courts.
We do not, then, have a tidy set of universalistic statements that
we can parade before the world. But we do have a set of existentialistic
statements, gathered from such diverse sources as the psychological labora-
tory and the anthropological field study. All of our sources indicate to
us the great variety found in human affairs. Our fund of existentialistic
statements comprises a critique of all statements that have the form:
"Man is . l ."  The "man ist" statement, when given without qualification,
has the form of an 'tall," or universalistic statement referring to "all
humans at all times and places." If someone asserts that "tman is rational,"
we can always point to examples of irrationality. The same case holds if
someone asserts in a universalistic manner that *man is irrational." Our
commitment and devotion to inductive procedures in the social sciences has
blinded us concerning the importance of the type of evidence we have been
producing.
The Context of Invention anwd the Context of Testing
There still exists a proper use of inductive procedures in science
or in any human enterprise.  I have discussed and critized the use of induc-
tive procedures in hypothesis testing. However, within the context of
hypothesis formation, inductlve inferences and all forms of intuitlve
processes should be encouraged. The confusion concerning the use of induc-
tive inference stems from a failure to distinguish clearly the context of
hypothesis invention from the context of hypothesis testing.  The two terms,
"discovery" and "justification" are usually utilized in making this dis-
tinction, stemming from ReiXchenbach.  However, both terms seem improper and
misleading.  In the first case, hypotheses are certainly not discovered,
but invented, as George Kelly (1960) has emphasized. In the second case,
hypotheses are not to be justified, but to be testedO  Within the context
of hypothesis invention, we can and shou'id utilize whatever inductive hunch-
es and ideas we might have. But, wlthin the context of hypothesis testing,
there exists no inductive procedure which provldes a degree of validity to
our hypotheses, without the use of certaln assumptions, or the use of other
information.
The context of Oinvention refers to the source and formation of our
Ideas, guesses, hunches, theories, conjectures, and hypotheses. The only
policy we should take regarding the source of our hypotheses is that they
should come from =   source.  WJ;e should use an.d develop whatever inagina-
tion, intuition, creativity, sk-ills, experience, and sensitivities we might
possess to ald us in the creation and formation of0 our hypotheses.  The
production of ideas and hypotheses ls not, a logical or epistemological prob-
lem, but a problem for social and psychological study, the study of scien-
tific and artist'ic creatlv'ity.
The Two Kinds of Empiricismt
There remains one more issue to be discussed before we come to the
summary and conclusion of this paper. This issue Is. What should we
regard as proper empirical method in science?
There are two very different ways in whlch we can interpret the
empirical method: (1) We can regard the source of our knowledge as deriv-
ing solely from sensory experience and observation. Our hypotheses, gener-
alizations, and laws are then derived from abstractions or inferences made
from these observations. This viewpoint will be called "pure empiricism,"
and is part of the positivist doctrlne crlticized in this paper.  (2) In
contrast to the position of pure empiricism, we can regard observation as
the method by which we test our theorles, hunches, conjectures, and hypoth-
eses.  In this view, knowledge can have a   source. However, it is by
empirical methods that our knowledge is tested or criticized. The empiri-
cal procedure is a convention employed to serve as an arblter between con-
tending theorles.  This viewpoint will be called "critical empiricism., and
is the viewpoinlt propounded in thls paper.
The crltical emplric'ist takes the posixtlon that the laboratory (or
the experimental sitution) is not a unique source for the creation and
development of theory. Theorles can have any source, and any relevant em-
pirical method can test themo A critical empiricist would want to confine
36
all discussion concerned with "rigor" and the like, to the context of test-
ing, and leave the scientific worker free to invent and create in whatever
manner he finds to be most comfortAbl7
It may very well be that some of the most exciting scientific theo-
ries in the future will be derived from very unusual sources. But If we are
squeamish about the source of our theories, Instead of with their clarity
and testability, we may poison the wells of creativity. The approach of
critical empiricism, although it demands severe testing of our hypotheses,
pleads for complete freedom in the invention and creation of hypotheses.
There remains, of course! the process of trnsforming the original, crude
hypothesis into a clear, testable form, This process, however, usually
dernands a great deal of imaginations and only a minimum of "rigor,"
ME and Concluding Rerks
This paper has provide4f a mere sketch of an alternative to the posi-
tivist philosophy,  The positivist p sition contains  ra  prescriptions
which are especially restri0tlIVe to sqcial sclence rese rgh,
We have seen the poyer y of inductive procedures, and the impo9ssi-
bility of proving or verifing a theory. We have thrown out the destructive
belief that our hypotheses 'mst 4p 4erived only from observation, rand have
come to see that ideas and theqries cgn derive from a varietir of sources.
It was argued t}~t very f   rev irements need to be met  for an
activity to be labelled as scienoTie.   However, these few :reqdrements are
strict ones, and derive from the logicl im-plications of the assertion being
made, The universalistic assertion is logically falsifiable within its
specified universe, The existentialistic stAtemvnt, which asserts only that
something exists or has opccwre4, can falsify a universal statements but can-
not logically verify a general statement, These strict requirements may
help instead of hindevr,by a34eviating some of the verbal diarrhea that is
so pronounced in the social sciences.
Iny persons may find disturbing the suggestion that social science
will produce very few laws or universalistic statements, This uneasiness
is aroused when we coanpre our results with what has been achieved in the
physical sciences.  Perhaps we will not be able to produce arG laws that
can withstand genuine attempts at falsification.  Some reseiF"hers may wish
to produce only existentiallstic statements to be used as falsifying evi-
dence against certain miyths and  nquestioned assumtions.    is approach
should not be regarded as secondary to theory construction.  In the social
sciences, our role for some time to come my be that of the critic or gad-
fly, subjecting the most cherished beliefs of our society to a critical and
systematic investigation.
There has been lately a rrssive identification of science with its
gadgets, its technology, its Impressive creations, and its white-coated
laboratorians. With such an identification of what to regard as scientific,
the social sciences will always be regarded somewhat contemptuously. Sci-
ence should not be identified as technology but as the acceptance and appli-
cation of the critical attitude. If we see that science is a method, a
method of free and critical discussion (and the utilization of empirical
37
tests when appropriate), we may be able to convince ourselves and others
that freedom of discussion is not a mere frill, but is the very essence of
science. We need a philosophy of science that is appropriate to the inves-
tigation of soclal and psychological problems, and we also need the condi-
tions in which we can examlne these problems freely and critically.
NOTE
This paper is a slightly revlsed verslon of a talk given at the
Eighth Annual Kroeber Anthropologiocal Soc?iety Meetings,9 Berkeley, Cali-
fornia, April 25, 1964. This paper was written during the author's
fellowship (NIMH grant: 1 Tl MH 8104-01) with the Behavioral Science
Project, School of Public Health, University of California, Berkeley.  I
would like to thank my colleagues on the Project, especially Octavio
Romano and Ronald Dillehay, for many exc-Itng discussions which led to the
writing of this paper.
REFERENCES
NManrs constructlion of hi s alternati'veso
motives, New York,, Grove Press0
In Assessment of hurran
Lewin, K.
1935
A &ynamic theory of personality, New York, McGraw-Hillo
Peters, R. S.
1960  The concept, of motivation, London, Rovtledge and Kegan Paul.
Popper, K.
1950
1957
1957a
1959
1962
The open society and its enemies, New Jersey, Prlnceton Univer-
sity Press.
The poverty of historicism, Boston, The Beacon Press0
The propensity interpretation of the calculus of probability,
and the quantum theory. In S. Kodrner (ed.), Observation and
interpretation, New York,'cademic Press0
The loglc of scientific discovery, New York, Basic Books.
Conjectures and refutations, New York, Basic Books.
Kelly, G.
1958