The Level-of-Analysis Problem in International Relations
Author(s): J. David Singer
Source: World Politics, Vol. 14, No. 1, The International System: Theoretical Essays (Oct.,
1961), pp. 77-92
Published by: Cambridge University Press
Stable URL: https://www.jstor.org/stable/2009557
Accessed: 20-09-2018 09:17 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend
access to World Politics
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
THE LEVEL-OF-ANALYSIS PROBLEM
IN INTERNATIONAL RELATIONS
By J. DAVID SINGER
TN any area of scholarly inquiry, there are always sever
1 which the phenomena under study may be sorted and ar
purposes of systemic analysis. Whether in the physical or socia
the observer may choose to focus upon the parts or upon
upon the components or upon the system. He may, for examp
between the flowers or the garden, the rocks or the quarry, t
the forest, the houses or the neighborhood, the cars or the t
the delinquents or the gang, the legislators or the legislative, a
Whether he selects the micro- or macro-level of analysis is os
a mere matter of methodological or conceptual convenience. Yet the
choice often turns out to be quite difficult, and may well become a
central issue within the discipline concerned. The complexity and
significance of these level-of-analysis decisions are readily suggested
by the long-standing controversies between social psychology and
sociology, personality-oriented and culture-oriented anthropology, or
micro- and macro-economics, to mention but a few. In the vernacular
of general systems theory, the observer is always confronted with a
system, its sub-systems, and their respective environments, and while
he may choose as his system any cluster of phenomena from the most
minute organism to the universe itself, such choice cannot be merely a
function of whim or caprice, habit or familiarity.2 The responsible
scholar must be prepared to evaluate the relative utility-conceptual
and methodological-of the various alternatives open to him, and to
appraise the manifold implications of the level of analysis finally
selected. So it is with international relations.
But whereas the pros and cons of the various possible levels of
analysis have been debated exhaustively in many of the social sciences,
the issue has scarcely been raised among students of our emerging
1 As Kurt Lewin observed in his classic contribution to the social sciences: "The first
prerequisite of a successful observation in any science is a definite understanding about
what size of unit one is going to observe at a given time." Field Theory in Social
Science, New York, I95I, P. I57.
2For a useful introductory statement on the definitional and taxonomic problems
in a general systems approach, see the papers by Ludwig von Bertalanffy, "General
System Theory," and Kenneth Boulding, "General System Theory: The Skeleton of
Science," in Society for the Advancement of General Systems Theory, General
Systems, Ann Arbor, Mich., I956, I, part I.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
78 THE INTERNATIONAL SYSTEM
discipline.3 Such tranquillity may be s
cation that the issue is not germane t
dence that it has already been resolve
quietude with a measure of concern. He is quite persuaded of its
relevance and certain that it has yet to be resolved. Rather, it is contended
that the issue has been ignored by scholars still steeped in the
intuitive and artistic tradition of the humanities or enmeshed in the
web of "practical" policy. We have, in our texts and elsewhere, roamed
up and down the ladder of organizational complexity with remarkable
abandon, focusing upon the total system, international organizations,
regions, coalitions, extra-national associations, nations, domestic pressure
groups, social classes, elites, and individuals as the needs of the moment
required. And though most of us have tended to settle upon the nation
as our most comfortable resting place, we have retained our propensity
for vertical drift, failing to appreciate the value of a stable point of
focus.4 Whether this lack of concern is a function of the relative infancy
of the discipline or the nature of the intellectual traditions from whence
it springs, it nevertheless remains a significant variable in the general
sluggishness which characterizes the development of theory in the study
of relations among nations. It is the purpose of this paper to raise the
issue, articulate the alternatives, and examine the theoretical implications
and consequences of two of the more widely employed levels of
analysis: the international system and the national sub-systems.
I. THE REQUIREMENTS OF AN ANALYTICAL MODEL
Prior to an examination of the theoretical implications of the level
of analysis or orientation employed in our model, it might be worthwhile
to discuss the uses to which any such model might be put, and
the requirements which such uses might expect of it.
Obviously, we would demand that it offer a highly accurate description
of the phenomena under consideration. Therefore the scheme must
present as complete and undistorted a picture of these phenomena as is
possible; it must correlate with objective reality and coincide with our
empirical referents to the highest possible degree. Yet we know that
3An important pioneering attempt to deal with some of the implications of one's level
of analysis, however, is Kenneth N. Waltz, Man, the State, and War, New York, I959.
But Waltz restricts himself to a consideration of these implications as they impinge on
the question of the causes of war. See also this writer's review of Waltz, "International
Conflict: Three Levels of Analysis," World Politics, xii (April i960), pp. 453-6i.
4Even during the debate between "realism" and "idealism" the analytical implications
of the various levels of analysis received only the scantiest attention; rather the
emphasis seems to have been at the two extremes of pragmatic policy and speculative
metaphysics.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
THE LEVEL-OF-ANALYSIS PROBLEM 79
such accurate representation of a complex and
phenomena is extremely difficult. Perhaps a use
borrowed from cartography; the oblate spheroid
most closely represents is not transferable to the
of a map without some distortion. Thus, the
exaggerates distance and distorts direction at an
move north or south from the equator, while t
jection suffers from these same debilities as we m
Neither offers therefore a wholly accurate prese
enough to reality to be quite useful for certain
same sort of tolerance is necessary in evaluati
for the study of international relations; if we m
sentational accuracy, the problem is to decide w
dysfunctional and where such accuracy is abs
These decisions are, in turn, a function of the
any such model-a capacity to explain the rel
phenomena under investigation. Here our concer
accuracy of description as with validity of expl
have such analytical capabilities as to treat th
a fashion which is not only valid and thorough,
latter requirement is often overlooked, yet its
strategy are not inconsequential.5 It should b
primary purpose of theory is to explain, an
explanatory requirements are in conflict, the la
priority, even at the cost of some representation
Finally, we may legitimately demand that
offer the promise of reliable prediction. In ment
last, there is no implication that it is the most
of the three. Despite the popular belief to the c
mands less of one's model than does explanati
For example, any informed layman can pred
5For example, one critic of the decision-making mode
Snyder, H. W. Bruck, and Burton Sapin, in Decision-M
Study of International Politics (Princeton, N.J., I954),
searcher could deal with all the variables in that model a
than a very few comparative studies in his lifetime. See He
Strategies for a Science of International Politics," World
pp. 28I-95. In defense, however, one might call attenti
which many of Snyder's categories could be collapsed int
apparently done in the subsequent case study (see not
telling criticism of the monograph is McClosky's comment
of theory is introduced into the proposal and the relations
more concretely, it is likely to remain little more than a
like any taxonomy, fairly limited in its utility" (p. 29I).
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
80 THE INTERNATIONAL SYSTEM
accelerator of a slowly moving car wil
less of the moon will be visible tonight than last night; or that the
normal human will flinch when confronted with an impending blow.
These predictions do not require a particularly elegant or sophisticated
model of the universe, but their explanation demands far more than
most of us carry around in our minds. Likewise, we can predict with
impressive reliability that any nation will respond to military attack in
kind, but a description and understanding of the processes and factors
leading to such a response are considerably more elusive, despite the
gross simplicity of the acts themselves.
Having articulated rather briefly the requirements of an adequate
analytical model, we might turn now to a consideration of the ways in
which one's choice of analytical focus impinges upon such a model and
affects its descriptive, explanatory, and predictive adequacy.
II. THE INTERNATIONAL SYSTEM AS LEVEL OF ANALYSIS
Beginning with the systemic level of analysis, we find in the total
international system a partially familiar and highly promising point
of focus. First of all, it is the most comprehensive of the levels available,
encompassing the totality of interactions which take place within the
system and its environment. By focusing on the system, we are enabled
to study the patterns of interaction which the system reveals, and to
generalize about such phenomena as the creation and dissolution of
coalitions, the frequency and duration of specific power configurations,
modifications in its stability, its responsiveness to changes in formal
political institutions, and the norms and folklore which it manifests as
a societal system. In other words, the systemic level of analysis, and only
this level, permits us to examine international relations in the whole,
with a comprehensiveness that is of necessity lost when our focus is
shifted to a lower, and more partial, level. For descriptive purposes,
then, it offers both advantages and disadvantages; the former flow from
its comprehensiveness, and the latter from the necessary dearth of detail.
As to explanatory capability, the system-oriented model poses some
genuine difficulties. In the first place, it tends to lead the observer
into a position which exaggerates the impact of the system upon the
national actors and, conversely, discounts the impact of the actors on
the system. This is, of course, by no means inevitable; one could conceivably
look upon the system as a rather passive environment in which
dynamic states act out their relationships rather than as a socio-political
entity with a dynamic of its own. But there is a natural tendency to
endow that upon which we focus our attention with somewhat greater
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
THE LEVEL-OF-ANALYSIS PROBLEM 81
potential than it might normally be expecte
to move, in a system-oriented model, away
much national autonomy and independence
more deterministic orientation.
Secondly, this particular level of analysis alm
that we postulate a high degree of uniformi
operational codes of our national actors. By def
room for divergence in the behavior of our pa
the whole. It is no coincidence that our most p
and one of the very few text writers focusing
system-should "assume that [all] statesmen
of interest defined as power."6 If this single-m
preted literally and narrowly, we have a sim
to economic man or sexual man, and if it be
no better off than the psychologist whose hum
realization" or "maximization of gain"; all su
from the same fatal weakness as the utilitarian
ciple. Just as individuals differ widely in what
and pain, or gain and loss, nations may diff
consider to be the national interest, and we
down and refine the larger category. Moreover
finds himself compelled to go still further and
of both motives and ideological preferences i
these represent two of the more useful dimens
among the several nations in our internation
any empirical concern with the domestic and in
the separate nations, the system-oriented appr
sort of "black box" or "billiard ball" concept
By discounting-or denying-the differences
6 Hans J. Morgenthau, Politics Among Nations, 3rd e
Obviously, his model does not preclude the use of po
differentiation of nations.
7 The "black box" figure comes from some of the simpler
in which the observer more or less ignores what goes
concentrates upon the correlation between stimulus an
as empirically verifiable, whereas cognition, perception, an
to be imputed to the individual with a heavy reliance on
variables." The "billiard ball" figure seems to carry the
is best employed by Arnold Wolfers in "The Actors in Inte
T. R. Fox, ed., Theoretical Aspects of International R
I959, pp. 83-i06. See also, in this context, Richard C. Sny
Theory-Continued," World Politics, xiii (January i96
Singer, "Theorizing About Theory in International Politics,
tion, iv (December i960), pp. 431-42. Both are review a
anthology.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
82 THE INTERNATIONAL SYSTEM
positing the near-impossibility of ob
at work within them,8 one concludes
of our nations in the international sy
inadequate foundation upon which to base any causal statements, it
offers a reasonably adequate basis for correlative statements. More
specifically, it permits us to observe and measure correlations between
certain forces or stimuli which seem to impinge upon the nation and
the behavior patterns which are the apparent consequence of these
stimuli. But one must stress the limitations implied in the word
"apparent"; what is thought to be the consequence of a given stimulus
may only be a coincidence or artifact, and until one investigates the
major elements in the causal link-no matter how persuasive the
deductive logic-one may speak only of correlation, not of consequence.
Moreover, by avoiding the multitudinous pitfalls of intra-nation
observation, one emerges with a singularly manageable model, requiring
as it does little of the methodological sophistication or onerous empiricism
called for when one probes beneath the behavioral externalities
of the actor. Finally, as has already been suggested in the introduction,
the systemic orientation should prove to be reasonably satisfactory as
a basis for prediction, even if such prediction is to extend beyond the
characteristics of the system and attempt anticipatory statements regarding
the actors themselves; this assumes, of course, that the actors are
characterized and their behavior predicted in relatively gross and
general terms.
These, then, are some of the more significant implications of a model
which focuses upon the international system as a whole. Let us turn
now to the more familiar of our two orientations, the national state
itself.
III. THE NATIONAL STATE AS LEVEL OF ANALYSIS
The other level of analysis to be considered in this paper is the
national state-our primary actor in international relations. This is
clearly the traditional focus among Western students, and is the one
which dominates almost all of the texts employed in English-speaking
colleges and universities.
Its most obvious advantage is that it permits significant differentiation
among our actors in the international system. Because it does not require
the attribution of great similarity to the national actors, it encour-
8Morgenthau observes, for example, that it is "futile" to search for motives because
they are "the most illusive of psychological data, distorted as they are, frequently beyond
recognition, by the interests and emotions of actor and observer alike" (op.cit., p. 6).
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
THE LEVEL-OF-ANALYSIS PROBLEM 83
ages the observer to examine them in greater
results of such intensive analysis cannot be ov
when the actors are studied in some depth tha
really valid generalizations of a comparative na
systemic model does not necessarily preclude com
among the national sub-systems, it usually eventu
comparisons based on relatively crude dimensio
On the other hand, there is no assurance that the
proach will produce a sophisticated model for t
of foreign policy; with perhaps the exception of
study,9 none of our major texts makes a serious an
describe and explain national behavior in terms of
cant variables by which such behavior might b
lyzed. But this would seem to be a function, not o
employed, but of our general unfamiliarity w
sciences (in which comparison is a major preo
retarded state of comparative government and po
most international relations specialists are lik
experience.
But just as the nation-as-actor focus permits us t
homogenization which often flows from the syst
lead us into the opposite type of distortion-a m
the differences among our sub-systemic actors. W
neither of these extremes is conducive to the deve
cated comparison of foreign policies, and such
balanced preoccupation with both similarity and d
seems to be greatest when we succumb to the tend
tiate; comparison and contrast can proceed only
formities.'0
One of the additional liabilities which flow in tu
to overdifferentiate is that of Ptolemaic paroch
emphasizing the differences among the many
observer is prone to attribute many of what he c
to his own nation and the vices to others, especial
the moment. That this ethnocentrism is by no
borne out by perusal of the major international re
9Ernst B. Haas and Allen S. Whiting, Dynamics of Inte
York, 1956.
10 A frequent by-product of this tendency to overdifferent
"second-image fallacy," in which one explains the peacefu
nation's foreign policy exclusively in terms of its domestic ec
characteristics (op.cit., chs. 4 and 5).
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
84 THE INTERNATIONAL SYSTEM
in the United States since i945. Not only
through the prism of the American nat
degree of attention (if not spleen) is dir
it would hardly be amiss to observe tha
equally well as studies in American fore
quacies of this sort of "we-they" orient
yet they remain a potent danger in any utilization of the national
actor model.
Another significant implication of the sub-systemic orientation is that
it is only within its particular framework that we can expect any useful
application of the decision-making approach.1" Not all of us, of course,
will find its inapplicability a major loss; considering the criticism which
has been leveled at the decision-making approach, and the failure of
most of us to attempt its application, one might conclude that it is no
loss at all. But the important thing to note here is that a system-oriented
model would not offer a hospitable framework for such a detailed and
comparative approach to the study of international relations, no matter
what our appraisal of the decision-making approach might be.
Another and perhaps more subtle implication of selecting the nation
as our focus or level of analysis is that it raises the entire question of
goals, motivation, and purpose in national policy.'2 Though it may well
be a peculiarity of the Western philosophical tradition, we seem to
exhibit, when confronted with the need to explain individual or collective
behavior, a strong proclivity for a goal-seeking approach. The
question of whether national behavior is purposive or not seems to
require discussion in two distinct (but not always exclusive) dimensions.
Firstly, there is the more obvious issue of whether those who act on
behalf of the nation in formulating and executing foreign policy consciously
pursue rather concrete goals. And it would be difficult to deny,
for example, that these role-fulfilling individuals envisage certain specific
outcomes which they hope to realize by pursuing a particular
1"Its most well-known and successful statement is found in Snyder et al., op.cit.
Much of this model is utilized in the text which Snyder wrote with Edgar S. Furniss,
Jr., American Foreign Policy: Formulation, Principles, and Programs, New York,
I954. A more specific application is found in Snyder and Glenn D. Paige, "The United
States Decision to Resist Aggression in Korea: The Application of an Analytical
Scheme," Administrative Science Quarterly, iii (December I958), pp. 34i-78. For those
interested in this approach, very useful is Paul Wasserman and Fred S. Silander,
Decision-Mating: An Annotated Bibliography, Ithaca, N.Y., 1958.
12 And if the decision-making version of this model is employed, the issue is unavoidable.
See the discussion of motivation in Snyder, Bruck, and Sapin, op.cit., pp. 92-I17;
note that 25 of the 49 pages on "The Major Determinants of Action" are devoted to
motives.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
THE LEVEL-OF-ANALYSIS PROBLEM 85
strategy. In this sense, then, nations may b
organisms which exhibit purposive behavior.
However, purposiveness may be viewed in a somewhat different
light, by asking whether it is not merely an intellectual construct that
man imputes to himself by reason of his vain addiction to the free-will
doctrine as he searches for characteristics which distinguish him from
physical matter and the lower animals. And having attributed this
conscious goal-pursuing behavior to himself as an individual, it may be
argued that man then proceeds to project this attribute to the social
organizations of which he is a member. The question would seem to
distill down to whether man and his societies pursue goals of their own
choosing or are moved toward those imposed upon them by forces
which are primarily beyond their control.13 Another way of stating the
dilemma would be to ask whether we are concerned with the ends
which men and nations strive for or the ends toward which they are
impelled by the past and present characteristics of their social and
physical milieu. Obviously, we are using the terms "ends," "goals," and
"purpose" in two rather distinct ways; one refers to those which are
consciously envisaged and more or less rationally pursued, and the
other to those of which the actor has little knowledge but toward which
he is nevertheless propelled.
Taking a middle ground in what is essentially a specific case of the
free will vs. determinism debate, one can agree that nations move
toward outcomes of which they have little knowledge and over which
they have less control, but that they nevertheless do prefer, and therefore
select, particular outcomes and attempt to realize them by conscious
formulation of strategies.
Also involved in the goal-seeking problem when we employ the
nation-oriented model is the question of how and why certain nations
pursue specific sorts of goals. While the question may be ignored in
the system-oriented model or resolved by attributing identical goals to
all national actors, the nation-as-actor approach demands that we investigate
the processes by which national goals are selected, the internal
and external factors that impinge on those processes, and the institutional
framework from which they emerge. It is worthy of note that
despite the strong predilection for the nation-oriented model in most
13 A highly suggestive, but more abstract treatment of this teleological question is in
Talcott Parsons, The Structure of Social Action, 2nd ed., Glencoe, Ill., I949, especially
in his analysis of Durkheim and Weber. It is interesting to note that for Parsons an act
implies, inter alia, "a future state of affairs toward which the process of action is
oriented," and he therefore comments that "in this sense and this sense only, the
schema of action is inherently teleological" (p. 44).
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
86 THE INTERNATIONAL SYSTEM
of our texts, empirical or even deductiv
conspicuously few.'4 Again, one might
methodological and conceptual inadequ
which international relations specialist
any event, goals and motivations are b
variables, and if we intend to explain
cannot settle for the mere postulation o
to go back a step and inquire into the
which they become the crucial variabl
behavior of nations.
There is still another dilemma involv
nation-as-actor model, and that concer
do we examine our actor's behavior in
which allegedly influence that behavior,
actor's perception of these "objective f
proaches are not completely exclusive of
greatly different and often incompat
markedly divergent models of national b
The first of these assumptions concern
causation. One view holds that indivi
quasi-deterministic fashion to the real
the acts or power of other individuals o
and "real" forces or stimuli. An oppos
and groups are not influenced in their be
but by the fashion in which these for
however distorted or incomplete such per
of this position, the only reality is the
cerned by the human senses; forces that
14 Among the exceptions are Haas and Whitin
the chapters in Roy C. Macridis, ed., Foreign Polic
N.J., I958, especially that on West Germany by K
15 As early as I934, Edith E. Ware noted that ". . .
no longer entirely a subject for political science
sociology, geography-all the social sciences-are c
understanding . . . of the international system." S
in the United States, New York, 1934, p. 172. Fo
Karl Deutsch, "The Place of Behavioral Sciences
Relations," Behavioral Science, III (July 1958),
Relevance of the Behavioral Sciences to the Stud
(October i96i), pp. 324-35.
16 The father of phenomenological philosoph
Edmund Husserl (i859-I938), author of Ideas: Gen
ology, New York, 1931, trans. by W. R. Boyce Gib
under the title Ideen zu einer reinen Phdnomen
sophie. Application of this approach to social psych
work of Koffka and Lewin.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
THE LEVEL-OF-ANALYSIS PROBLEM 87
for that actor, and those that do exist do s
they are perceived. Though it is difficult t
individual, a group, or a nation is affected
distance, or a neighbor's physical power on
nized and appraised, one must concede that
affect the manner in which such forces ar
been pointed out, an individual will fall to
out of a tenth-story window regardless of
forces, but on the other hand such perception is a major factor in
whether or not he steps out of the window in the first place."7 The point
here is that if we embrace a phenomenological view of causation, we
will tend to utilize a phenomenological model for explanatory purposes.
The second assumption which bears on one's predilection for the
phenomenological approach is more restricted, and is primarily a
methodological one. Thus, it may be argued that any description of
national behavior in a given international situation would be highly
incomplete were it to ignore the link between the external forces at
work upon the nation and its general foreign policy behavior. Furthermore,
if our concern extends beyond the mere description of "what
happens" to the realm of explanation, it could be contended that such
omission of the cognitive and the perceptual linkage would be ontologically
disastrous. How, it might be asked, can one speak of "causes"
of a nation's policies when one has ignored the media by which external
conditions and factors are translated into a policy decision? We may
observe correlations between all sorts of forces in the international
system and the behavior of nations, but their causal relationship must
remain strictly deductive and hypothetical in the absence of empirical
investigation into the causal chain which allegedly links the two.
Therefore, even if we are satisfied with the less-than-complete descriptive
capabilities of a non-phenomenological model, we are still drawn
to it if we are to make any progress in explanation.
The contrary view would hold that the above argument proceeds
from an erroneous comprehension of the nature of explanation in
social science. One is by no means required to trace every perception,
transmission, and receipt between stimulus and response or input and
output in order to explain the behavior of the nation or any other human
group. Furthermore, who is to say that empirical observation-subject
17 This issue has been raised from time to time in all of the social sciences, but for an
excellent discussion of it in terms of the present problem, see Harold and Margaret
Sprout, Man-Milieu Relationship Hypotheses in the Context of International Politics,
Princeton University, Center of International Studies, 1956, pp. 63-71.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
88 THE INTERNATIONAL SYSTEM
as it is to a host of errors-is any bett
informed deduction, inference, or anal
flows logically from a coherent theoretical model just as reliable as
one based upon a misleading and elusive body of data, most of which
is susceptible to analysis only by techniques and concepts foreign to
political science and history?
This leads, in turn, to the third of the premises relevant to one's
stand on the phenomenological issue: are the dimensions and characteristics
of the policy-makers' phenomenal field empirically discernible?
Or, more accurately, even if we are convinced that their perceptions
and beliefs constitute a crucial variable in the explanation of a nation's
foreign policy, can they be observed in an accurate and systematic
fashion ?18 Furthermore, are we not required by the phenomenological
model to go beyond a classification and description of such variables,
and be drawn into the tangled web of relationships out of which they
emerge? If we believe that these phenomenal variables are systematically
observable, are explainable, and can be fitted into our explanation
of a nation's behavior in the international system, then there is a further
tendency to embrace the phenomenological approach. If not, or if we
are convinced that the gathering of such data is inefficient or uneconomical,
we will tend to shy clear of it.
The fourth issue in the phenomenological dispute concerns the very
nature of the nation as an actor in international relations. Who or what
is it that we study? Is it a distinct social entity with well-defined
boundaries-a unity unto itself? Or is it an agglomeration of individuals,
institutions, customs, and procedures? It should be quite
evident that those who view the nation or the state as an integral social
unit could not attach much utility to the phenomenological approach,
particularly if they are prone to concretize or reify the abstraction. Such
abstractions are incapable of perception, cognition, or anticipation
(unless, of course, the reification goes so far as to anthropomorphize
and assign to the abstraction such attributes as will, mind, or personality).
On the other hand, if the nation or state is seen as a group
of individuals operating within an institutional framework, then it
makes perfect sense to focus on the phenomenal field of those individuals
who participate in the policy-making process. In other words,
people are capable of experiences, images, and expectations, while insti-
18This is another of the criticisms leveled at the decision-making approach which,
almost by definition, seems compelled to adopt some form of the phenomenological
model. For a comprehensive treatment of the elements involved in human perception,
see Karl Zener et al., eds., "Inter-relationships Between Perception and Personality: A
Symposium," Journal of Personality, xviii (1949), pp. I-266.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
THE LEVEL-OF-ANALYSIS PROBLEM 89
tutional abstractions are not, except in the
our actor cannot even have a phenomenal field, there is little point
in employing a phenomenological approach.19
These, then, are some of the questions around which the phenomenological
issue would seem to revolve. Those of us who think of social
forces as operative regardless of the actor's awareness, who believe that
explanation need not include all of the steps in a causal chain, who
are dubious of the practicality of gathering phenomenal data, or who
visualize the nation as a distinct entity apart from its individual members,
will tend to reject the phenomenological approach.20 Logically,
only those who disagree with each of the above four assumptions
would be compelled to adopt the approach. Disagreement with any one
would be sufficient grounds for so doing.
The above represent some of the more significant implications and
fascinating problems raised by the adoption of our second model. They
seem to indicate that this sub-systemic orientation is likely to produce
richer description and more satisfactory (from the empiricist's point of
view) explanation of international relations, though its predictive power
would appear no greater than the systemic orientation. But the descriptive
and explanatory advantages are achieved only at the price of
considerable methodological complexity.
IV. CONCLUSION
Having discussed some of the descriptive, explanatory, and predictive
capabilities of these two possible levels of analysis, it might now be
useful to assess the relative utility of the two and attempt some general
statement as to their prospective contributions to greater theoretical
growth in the study of international relations.
In terms of description, we find that the systemic level produces a
more comprehensive and total picture of international relations than
does the national or sub-systemic level. On the other hand, the atomized
and less coherent image produced by the lower level of analysis is somewhat
balanced by its richer detail, greater depth, and more intensive
portrayal.21 As to explanation, there seems little doubt that the sub-
19 Many of these issues are raised in the ongoing debate over "methodological individualism,"
and are discussed cogently in Ernest Nagel, The Structure of Science, New
York, i96i, pp. 535-46.
20 Parenthetically, holders of these specific views should also be less inclined to adopt
the national or sub-systemic model in the first place.
21In a review article dealing with two of the more recent and provocative efforts
toward theory (Morton A. Kaplan, System and Process in International Politics, New
York, 1957, and George Liska, International Equilibrium, Cambridge, Mass., 1957),
Charles P. Kindleberger adds a further-if not altogether persuasive-argument in favor
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
90 THE INTERNATIONAL SYSTEM
systemic or actor orientation is consider
as it does a more thorough investigati
foreign policies are made. Here we are en
tions imposed by the systemic level and t
the more significant causation. And in te
tions seem to offer a similar degree o
function of what we seek to predict. Th
to prefer predictions about the way in w
to a contemplated move on his own na
will probably prefer either generalized
havior of a given class of nations or tho
Does this summary add up to an over
of the two models? It would seem no
reasons the scholar may be more interes
any given time and will undoubtedly shi
his research needs. So the problem is rea
level is most valuable to the discipline
that it be adhered to from now unto e
realizing that there is this preliminary c
be temporarily resolved prior to any giv
it must also be stressed that we have d
more common orientations, and that
perhaps even more fruitful potentially
here. Moreover, the international syst
prospective change, and it may well be t
will take on new characteristics or tha
their place. As a matter of fact, if incap
leads to the transformation or decay of
steady deterioration and even ultimat
state as a significant actor in the world
However, even if the case for one or
of analysis cannot be made with any cer
maintain a continuing awareness as to
level here and another there, but we can
tion in the midst of a study. And when
of the lower, sub-systemic level of analysis: "The
everything interacting. One can discuss it intell
"Scientific International Politics," World Politic
22 It should also be kept in mind that one cou
model which successfully embraces both of thes
conceptual clarity and internal consistency. In thi
to date, though Kaplan's System and Process in
fairly close.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
THE LEVEL-OF-ANALYSIS PROBLEM 91
selection or replace one with another at approp
so with a full awareness of the descriptive, exp
implications of such choice.
A final point remains to be discussed. Despite
one might still be prone to inquire whether th
exercise in verbal gymnastics. What, it might b
between the two levels of analysis if the em
essentially the same? Or, to put it another way
between international relations and comparat
haps a few illustrations will illuminate the subt
ences which emerge when one's level of analysi
example, postulate that when the international
by political conflict between two of its most p
a strong tendency for the system to bipolarize. T
proposition. A sub-systemic proposition, dealin
empirical referents, would state that when a po
in political conflict with another of approxima
exert pressure on its weaker neighbors to join it
tion, assuming it is true, is theoretically usef
verified by a different intellectual operation
the crucial thing for theoretical development-o
two kinds of statements together to achieve
empirical generalizations.
To illustrate further, one could, at the system
when the distribution of power in the internat
diffused, it is more stable than when the disce
defined coalitions occurs. And at the sub-sys
the same empirical phenomena would produce t
when a nation's decision-makers find it diffi
nations readily as friend or foe, they tend t
more uniform and moderate fashion. Now, t
propositions, how much cumulative usefulne
tempting to merge and codify the systemic pr
illustration with the sub-systemic proposition
versa? Representing different levels of analysis
frames of reference, they would defy theoreti
well be a corollary of the other, but they ar
binable. A prior translation from one level to a
This, it is submitted, is quite crucial for the t
of our discipline. With all of the current empha
empirical and data-gathering research as a prer
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms
92 THE INTERNATIONAL SYSTEM
ing, one finds little concern with the re
and discrete data-gathering activities. Even if we were to declare a
moratorium on deductive and speculative research for the next decade,
and all of us were to labor diligently in the vineyards of historical and
contemporary data, the state of international relations theory would
probably be no more advanced at that time than it is now, unless such
empirical activity becomes far more systematic. And "systematic" is
used here to indicate the cumulative growth of inductive and deductive
generalizations into an impressive array of statements conceptually
related to one another and flowing from some common frame of reference.
What that frame of reference should be, or will be, cannot be
said with much certainty, but it does seem clear that it must exist. As
long as we evade some of these crucial a priori decisions, our empiricism
will amount to little more than an ever-growing potpourri of discrete,
disparate, non-comparable, and isolated bits of information or extremely
low-level generalizations. And, as such, they will make little contribution
to the growth of a theory of international relations.
This content downloaded from 147.251.109.95 on Thu, 20 Sep 2018 09:17:35 UTC
All use subject to https://about.jstor.org/terms