心理学报 2012, Vol. 44, No.1, 133−137
Acta Psychologica Sinica DOI: 10.3724/SP.J.1041.2012.00133
133
What Is a Good Theory? A Perspective from Theoretical Psychology
YE Hao-Sheng1; Henderikus J. Stam2
(1 Center for Psychology and Brain Science, Guangzhou University, Guangzhou 510006)
(2 Department of Psychology, University of Calgary, Calgary, Canada)
Abstract: We discuss the status of theories in psychology by addressing ourselves to a recent paper published in
this journal by Hong, Chao, Yang & Rosner (2010). We argue that Hong et al. provided a restrictive version of
theory in psychology that is limited by the implicit adoption of logical empiricism as the basis for their views of
theory. In addressing the limitations of logical empiricism we address the broader question of just what
constitutes theory and how theory might be more usefully and widely applied in the discipline.
Key words: theory; theoretical psychology; post-empiricism
With examples from their research on hierarchical
identity, identity hegemony, and bridging identity, Hong,
Chao, Yang & Rosner (2010) illustrated a four-step
approach to building and testing psychological theories
(selecting phenomena, finding critical commonalities,
abstracting and hypothesis testing). We agree with the
authors that “it is never too excessive to emphasize the
value and importance of theories in social psychology” (p.
22) and consider the proposed four-step approach to be of
use for psychological researchers to build and test theories
as a first step in the research enterprise. The place of theory
in psychology however is contested still and in our
comments on their paper we wish to consider various ways
in which we can understand theory in research and practice.
The proposed four-step approach by Hong et al. (2010)
represents one such perspective but one based upon
traditional logical empiricism, sometimes referred to as just
logical positivism. This perspective, while long open to
debate within the philosophy of science, is still more or
less widely accepted in the psychological community and
constitutes normal practice for many psychologists. We say
“more or less” since it is often adopted implicitly rather
than explicitly. It is implied in many of the methodologies
propounded by psychologists and taught to our students.
There are important historical reasons for this, the
discussion of which is far beyond this short commentary.
However, we can sum these up by noting that the adoption
of a neo-positivist viewpoint in psychology coincided in
the early 20th century with the development of logical
收稿日期: 2011-03-14
* 羊城学者首席科学家资助项目(10A030S)。
通讯作者: 叶浩生, E-mail: yehaosheng@yahoo.cn
positivism in philosophy, Fisher’s experimental designs
and inferential statistics in the field of mathematics and the
development of behaviorism in psychology (e.g., Danziger,
1990; Mills, 1998). Each of these has been influential in
turn in creating the methodological and theoretical
orientation so well described by Hong et al. (2010). We
have no comment on the research program that Hong et al.
describe since we think it should be judged on its own
merits. However, the version of theory espoused by the
logical empiricist project has limitations. Its limitations are
such that they restrict the nature and uses of theory and
hence we wish to (a) discuss why we believe this concept
of theory is limited and (b) consider a number of
alternatives to the description of theory that Hong et al.
have presented.
Hong et al.’s (2010) define theory in the following
manner, “theory is defined in science ‘as a comprehensive
explanation of an important feature of nature supported by
facts gathered over time’” (p. 23). Furthermore, theories
make it possible to predict “yet unobserved phenomena” (p.
23). This simplified (but not simplistic) definition is useful
but perhaps makes a host of complexities about theory
invisible. These complexities concern just what counts as
theory in the first place, and avoids the question of how we
are to treat theory in science (are theories representations
of reality, instrumental tools, models that we don’t expect
to be correct, and so on). Unfortunately the authors’
approach is one that relies heavily on one tradition only, as
we have noted, the logical empiricist tradition. Although
we should add that the authors do not commit themselves
explicitly to any particular understanding of theory, they
appear to implicitly adopt what is often referred to as the
‘received view’ in the philosophy of science (Suppe, 1989).
134 心 理 学 报 44卷
On the received view, a good theory is one that is
based on empiricism (hence observation) and logic (the
generalization and induction of empirical observations).
This perspective has the following characteristics: First,
theory comes from the induction and abstraction of
empirical facts. It is built upon facts and is an induction
and generalization after (psychologists’ in this case)
arduous gathering of empirical observations over time, as
is stated by the authors: “for us, social psychological
theories are best grounded in social phenomena” (Hong et
al., p. 24). From the analysis and induction of the empirical
facts thus gathered, psychologists generalize some basic
principles like concepts, theories and, eventually it is
hoped, laws, which may combine systematically to form
some preliminary theories. This inductivist view however
is itself an idealized view of science and the history of the
sciences informs us otherwise. It was Karl Popper (1963)
who noted the problem of induction by showing that it is
far too simple to verify a theory by searching for
confirmatory evidence. What is crucial for the purposes of
science is the falsifiability or refutability of a theory.
Knowledge argued Popper is the outcome of conjecture and
criticism, not induction. Even if one disagrees with the
falsifiability criterion, induction is no sure way from ‘data’
to ‘theory’.
Second, in this tradition, theory is fundamentally
concerned with the explanation and collection of ‘facts.’
What constitutes a fact however is not clear. From the time
of Norwood Hanson’s (1958) important work, we have
come to understand that data are at least to some extent
‘theory-laden.’ Although we may disagree precisely to
what degree any data gathering episode is theory-laden, it
is more or less given that we cannot collect data in the first
instance without some strong theoretical guidance for
telling us just what we deem is important. This does not
mean that the data are determined beforehand. It does
mean that the data-gathering enterprise, the search for
‘facts,’ is constrained and enabled by our conceptual
scaffolding.
Furthermore, according to the ‘received view’ what
theoretical psychologists embark on is largely mediational
work that requires them to establish some relationship
among various empirical facts to explain contradictions
and conflicts. That is, theory is a kind of supplementary
work conducted after the arduous labor of data gathering
has been completed. And of course, on this view theory is
in turn generative of experimental hypotheses. A
hypothesis presumably exists at the very outset before
psychologists carry out an experiment, which is then taken
as a trial to attest to the value of the hypothesis. We assume
this is what Hong et al. (2010) refer to when they argue that
they engage in four steps, “(1) selecting phenomena, (2)
finding critical commonalities, (3) abstracting (theorizing),
and (4) hypothesis testing” (p. 42). A hypothesis will take
on some truth value, after having being verified by
experimental testing, while if not verified, it will be cast
away. Finally, on this view the development of theory is a
cumulative process. Theory is tentative, not a finished
product. But with the ever-continuing development and
perfection of a theory, and also the constant growth of
related theories, the empirically based theory will by
definition come close to the truth.
1 Alternative Views of Theory
Unfortunately, this neat picture of the relationship
between theory and data was long ago shattered not only by
historians and philosophers of science but by scientists
themselves. This picture of theorizing, like the logical
empiricist view on which it is based, restricts theorizing to
a few well-defined activities in which few scientists ever
engage. As Quine (1953/1980), and Duhem (1906/1954)
before him argued, scientists are very reluctant to give up a
treasured hypothesis because their data do not confirm the
theory at hand. Instead, it is much more likely that
adjustments will be made elsewhere in the theory so that
the work of confirmation can continue. As Duhem had
already noted in the early 20th century, scientists never base
their predictions on a single hypothesis but on an entire set
of assumptions and rules of inference. Quine furthered the
case by noting that we cannot separate out a linguistic
component and a factual component in determining the
truth of any given statement; “science has its double
dependence upon language and experience; but this duality
is not significantly traceable into the statements of science
taken one by one” (Quine, 1953/1980, p. 42). In other
words, we do not speak of the empirical content of
individual statements but theories as a whole. And theories
as a whole are always open to revision in the face of
conflicting evidence (see Stam, 1992).
This of course does not mean that scientific progress
cannot occur or that we cannot evaluate the claims of the
scientist. What it does mean however is that such
evaluations simply do not follow the simple prescriptive
guidelines found in our textbooks or in descriptions such as
those in Hong et al. (2010).
1期 What Is a Good Theory? A Perspective from Theoretical Psychology 135
2 What is a theory?
We should note that there is no canonical standard to
which we can turn to explain theory to us. In the 20th
century scholars have presented a number of important
proposals as alternatives to the dominant version of logical
empiricism. As the difficulties with logical empiricism
became clearer, for example following the acceptance that
the theory-observation distinction is much more complex
than logical empiricists believed, some have argued that we
need to return to theories as realistic claims about the
world, rather than seeing them as sentences that are
verifiable. Realists often use what Hilary Putnam once
referred to as the ‘no miracles’ argument, namely, that
‘realism is the only philosophy that does not make the
success of science a miracle’ (Putnam, 1975, p. 73). What
this seeks to recover is the intuition that the success of
science has something to do with the way the world
actually is. However, realism remains controversial within
the philosophy of science because of the problem posed by
Laudan (1981) and others, now referred to as the
‘pessimistic induction’ or ‘pessimistic meta-induction’
problem. It argues that science has been wrong in the past
and what we hold to be true today will likely turn out to be
false. Indeed, on Laudan’s account realism cannot account
for the fact that theories which contain theoretical terms
that are non-referential can nonetheless turn out to be
successful (such as the vibratory theory of heat). This does
not mean that realism is wrong merely that it is
unwarranted.
Another alternative to logical empiricism is
constructive empiricism (van Fraassen, 1980). According
to this account a theory can still be understood to refer to
literal entities, however it is more concerned with empirical
adequacy. This is a subtle distinction meaning that it is not
theories or hypotheses but models of reality that are tested
in science and these models must have some relationship to
a broader theory. The theory is empirically adequate
according to van Fraassen if appearances “are isomorphic
to the empirical substructures of some model of the theory”
(Monton & Mohler, 2008).
The social sciences in general and psychology in
particular however suffer from an additional problem. The
phenomena that we investigate are accessible to us
precisely because we are apt participants in the world that
entails them. Sometimes this is referred to as the
distinction between human kinds and natural kinds
(Hacking, 1995; 2002). Human kinds are distinguished by
the fact that they are not inevitable – they are, according to
Hacking, kinds of people, action, behavior, and so on.
Their most important feature is that they have an effect on
people since people are capable of self-knowledge and they
are moral agents who will respond to the ‘kinds’ they are
taken to be. Hence human kinds are subject to what
Hacking (1995) called “looping effects”, later he called this
“interactive kinds”, namely we respond to the knowledge
of what we are by possibly becoming something different.
Unlike quartz, we respond to the theories that are posited
about us (see also Gergen, 2010).
The uniqueness of psychology among the family of
sciences then is already reflected in the varieties of theory
taken up in the post-empiricist era. These include such
radical departures from empiricism as phenomenology,
hermeneutics and social constructionism. First, theory is
not only the generalization and induction of empirical facts,
but in addressing practical problems psychologists have
generated numerous alternatives to a sterile empiricism.
Second, theory is capable not only of explanation and
prediction, but more importantly, can serve the function of
criticism and reflection. On the empiricist view, theory is
regarded either as a kind of tool capable of providing
hypotheses that can be validated, or as consequences of
logical reasoning after systematic observations. Today we
accept that in the variety of the human sciences, the role of
theory can also be to evaluate and criticize, and to reflect
upon the kinds of categories that have come to exist in
empirical work more generally. There is in fact a venerable
tradition of this in psychology. Watson’s (1913) critique of
the structuralism of the early 20th century was based on a
conception of behaviorism that did not have an empirical
program nor a viable research methodology until much
later. Freud’s (1895/1977) early development of
psychoanalysis was done in the absence of both patient
cases and the neurological evidence Freud might have
hoped for. Early attempts to produce a cognitive
psychology were also more of a hope than a reality and
were based on conceptual work done on computers by such
early cognitive theorists such as Newell and Simon.
Third, theory is not just a resource for experimental
hypotheses - it could be an “invention” rather than a
“discovery” derived from empirical facts. Empirical data
are not independent of theory as we have noted, on the
contrary, theory lays restriction on what facts can be
observed and different data may be acquired in different
theoretical frames. Experimenters are inclined to verify
while they might not notice or just intentionally neglect
other inconsistent evidence, thus always enabling opposing
parties to find their own supporting data. Fourth, theory’s
136 心 理 学 报 44卷
development is not a cumulative process. A theory is not an
empirical model, but is based upon a host of assumptions
and presumptions, what Kuhn (1962) called a paradigm and
later called a ‘disciplinary matrix.’ Science moves, on this
view, from “normal” through “revolutionary periods.”
After a so-called ‘scientific revolution’ however Kuhn
famously (and controversially) argued that scientific
paradigms were largely incommensurable. Science is
generally conservative in its approach to theory change
argued Kuhn in so far as most ‘normal’ science proceeds by
solving puzzles within a disciplinary matrix. Although
Kuhn has been roundly criticized, his work began a
systematic evaluation of the importance of historical, social
and psychological influences on the sciences, including the
human sciences (e.g., Smith, 1997).
We hope it is now clear that from the perspective of
the many different ways of approaching theory that,
depending on one’s views of the place of theory in science,
there are at least several different standards for validating
and evaluating a good theory. In the view of logical
empiricism, theory is supported only after being validated
by experience for its compatibility with empirical facts.
Theory is restrained by empirical methods, as stated by the
authors: “the derived hypotheses would constrain the choice
of method because some methods are better able to test the
hypotheses than are other methods” (Hong et al., 2010, p.26).
As we hope we have demonstrated, this version of theory was
long ago discarded for a host of alternatives.
For post-empiricism, coherence with empirical
evidences is not any longer the only and critical criterion
while the evaluation of a theory can be carried out at
different levels and standards. The first is the standard of
conceptual logic, with which it is analyzed, to wit, whether
the inner factors of a theory are logically consistent with
each other, whether a theory is compatible with other
background theories, and whether a theory is compatible
with the proposer’s epistemology and methodology. The
second standard that can be used is value and ideology. All
psychological theories are ideologically embedded in some
sense, drawing our attention to some aspect of life rather
than other aspects, and in this process the research subject
is overlaid with certain values and meaning. Thus, a theory
is judged by analyzing its social value and ideological
function. Rhetorical and narrative analysis is the third
criterion. It is believed that theory is not a truth only
compatible with empirical facts, but a kind of cultural and
historical construction. The scientifically acquired theory is
just, in effect, a rhetorical or narrative method and, as an
important part of social construction, the application of this
method also holds the status of epistemology and lays a
solid foundation for choosing and comparing theories.
Rhetorical and narrative analysis thus becomes another
important criterion in evaluating theory.
Finally, social practice and application may function
as another standard. Empiricists believe that theory comes
from practice while, in the meantime, practice is guided by
theory, which lost its foundation in the post-empiricist era.
Here, theory is no longer considered to be a generalization
and abstraction of experience, but a kind of discursive
practice; theory usually reflects a specific form of life and
the changes in this discourse would potentially bring about
changes to traditional models of action. In this sense,
theory is closely related to other human activities (Gergen
& Zielke, 2006). Theoretical discourse is a coherent part of
effective actions, and a theory can be evaluated by its
coherence with practice, that is, whether the theory has
promoted social practice to be more effective, more
meaningful and more theoretical.
Of course, we do not intend to deny the contributions
of empirical perspectives. In fact, there are often two
categories of theory in theoretical psychology: substantive
theory in specific areas and metatheory, which concerns
itself with foundational question and is hence more general.
The theory illustrated by Hong et al. (2010) is a substantive
theory, which is also necessary for the development of
theoretical psychology. But what we have tried to
emphasize is that substantive theory should keep pa