A Cognitive Theory of Deterrence

02berejikian (ds) 17/12/01 10:59 am Page 165
© 2002 Journal of Peace Research,
vol. 39, no. 2, 2002, pp. 165–183
Sage Publications (London, Thousand Oaks,
CA and New Delhi)
[0022-3433(200203)39:2; 165–183; 000000]
A Cognitive Theory of Deterrence*
J E F F R E Y D. BE RE JIKIA N
De p a rtment of Political Science, Un i versity of Ge o r g i a
Prospect theory is an empirical model of choice that stands as the leading alternative to rationality for
explaining decisions under conditions of risk. While many still defend the assumption of rationality as
an appropriate starting point for the construction of international relations theory – deterrence
theory especially – there is growing support for models of international politics grounded in the actual
capacities of real-world decisionmakers. This article accepts that standard depictions of deterrence incen-
tives capture much of the essential character of deterrent relationships. However, it substitutes cogni-
tive assumptions in place of traditional rational choice. Using prospect theory, the article reconsiders
three typical deterrence games. The new model of military deterrence put forth unearths a set of con-
ditions that are required for successful deterrence and uncovers a set of causes for deterrence failures
that run counter to conventional understanding.
Introduction
international politics grounded in the
capacities of actual decisionmakers.
Few decisionmaking environments are
Initially developed by Daniel Kahneman
fraught with greater risk than the anarc h y
and Amos T versky (1979), prospect theory
of the international system. The fact that
is an empirical model of decisionmaking
bad foreign policy choices hold the poten-
that stands as the ‘leading alternative to
tial for disaster (e.g. war) is alone justifi-
expected utility as a theory of choice under
cation for the considerable time spent by
conditions of risk’ (Levy, 1996: 179). It is
students of international politics on the
t h e re f o re well suited to a re-examination of
logic of military deterrence. Most re s e a rc h
d e t e r rence. In contrast to rational choice,
on deterrence assumes that states are
p rospect theory finds that decisionmakers
rational. Meanwhile, accumulating empiri-
d o n’t maximize in their choices, are apt to
cal evidence from laboratory experimen-
ove rweight losses with respect to com-
tation
suggests
that
decisionmakers
parable gains, and tend to be risk ave r s e
systematically violate the strict behavioral
when confronted with choices betwe e n
expectations of rationality. As anarc h y
gains while risk acceptant when confro n t e d
places state surv i val at risk, these findings
with losses. Considerable re s e a rch on
f rom cognitive psychology similarly
p rospect theory outside political science
t h reaten international relations theory and
exists, and scholars studying international
s u p p o rt the growing call for new models of
politics have recently taken it up. Still, the
* I am indebted to the comments of the anonymous review-
g rowing recognition that actors choose as
ers and to the generous help of Robert Grafstein. Please
p rospect theory suggests, and not as our
direct correspondence via e-mail to jberejik@arches.uga.
edu
theories of deterrence often assume, has not
165

---

02berejikian (ds) 17/12/01 10:59 am Page 166
1 6 6
j o u rn a l o f P E AC E RE S E A RC H
volume 39 / number 2 / march 2002
yet forced a thorough re v i ew of deterre n c e
games. Finally, I highlight the unintended
t h e o ry and practice.
consequences of deterrence threats that are
This article presents a new theory of deter-
revealed under prospect theory but hidden by
rence that is grounded in the actual capacities
rational choice.
of decisionmakers. It accepts the standard
depiction of deterrence incentives while sub-
Relative Strengths: Rationality
stituting a cognitive model of decisionmak-
versus Cognition
ing for rational choice. Using prospect theory
as a cornerstone, three types of deterrence are
While the behavioral expectations of
reconsidered: mutual deterrence, unilateral
prospect theory differ from rational choice,
deterrence, and extended deterrence.
prospect theory retains a principal advantage
Prospect theory identi?es a new set of con-
of rational analysis. Actor decisions are stable
ditions that are required for successful deter-
and predictable. Earlier psychological
rence, and uncovers a new set of reasons for
approaches to the study of international
the failure of deterrence that run counter to
relations depended heavily upon the personal
conventional understanding.
traits of individual decisionmakers, upon
The academic literature on deterrence is
information ?ows unique to the decision set,
large, but the number of core arguments is
or upon a unique group dynamic in the
small. This article addresses the core while
decision setting.2 By contrast, prospect
setting aside the important body of work that
theory permits the construction of hypoth-
extends or modi?es the basic set of premises
eses that cut across contexts; it thereby co-
that constitute deterrence theory.1 The
opts simplicity as the principal bene?t of
purpose here is to return to the core and
rational models. This is likely the reason for
begin rebuilding a foundation that incorpor-
its growing recognition in political science.
ates an empirically supported model of
Broadly, research on cognition and decision-
human decisionmaking. Accordingly, the
making demonstrates that individual choices
argument addresses (and adopts) some of the
are as much a function of consistent heuris-
simpler examples of game theory in the study
tics and biases as they are the result of calcu-
of deterrence. Working at this elemental level
lated costs and bene?ts. Prospect theory is
is a necessary ?rst step toward the full expres-
part of this larger investigation into the struc-
sion of a formalized theory of cognitive deter-
ture of human choice.3 It is the most well-
rence.
traveled ?nding to emerge, and it has been
The argument below unfolds as follows.
successful at penetrating other ?elds in which
After a brief discussion of rationality and cog-
human decisionmaking is important (e.g.
nition, I present a short summary of typical
political science and economics). Deviations
concerns about traditional deterrence theory.
from rationality are not treated as exceptions;
Prospect theory is then introduced as an
they are instead important phenomena for
alternative foundation for understanding the
which an explanation is possible. There is
dynamics of deterrence behavior. A general
already considerable empirical evidence that
deterrence model is developed and then used
to re-examine three common deterrence
2 For a discussion of psychology and foreign policy
decisions see Vertzberger (1990). For examples of psychol-
1 Such extensions are too numerous to mention, but
ogy in the deterrence literature see Lebow (1989) and
include topics like: the logic of ?rst strike (Wagner, 1991);
Jervis, Lebow & Stein (1985).
deterrence in the developing world (Kraig, 1999); the
3 For a summary of the major ?ndings see Kahneman &
potential value in promoting proliferation (Bueno de
Tversky (1982) and Bell, Raffa & Tversky (1988). For a
Mesquita & Riker, 1982); critiques of US policy (Jervis,
good non-technical summary of this literature see Plous
1984b).
(1993).

---

02berejikian (ds) 17/12/01 10:59 am Page 167
Je f f re y Be re j i k i a n
A C O G N I T I V E TH E O RY O F D E T E R R E N C E
1 6 7
prospect theory can capture important
science. It lies, for example, at the center of
aspects of state behavior (Levy, 1992;
Elinor Ostrom’s Presidential Address to the
Farnham, 1994; Levy, 1996; Berejikian,
American Political Science Association,
1997; Farnham, 1997; McDermott, 1998),
which challenged political science to develop
suggesting that it is a solid foundation from
a behavioral theory of collective action.
which to build a theory of deterrence.
According to Ostrom, theories of politics
Increasingly, the use of classic rationality
should be ‘based on models of the individual
as a foundation for international relations
consistent with empirical evidence about
theory – or any political theory – is con-
how individuals make decisions in social-
tested. Time constraints, huge amounts of
dilemma situations’ (Ostrom, 1998: 1). The
information, uncertainty, combined with
application of ?ndings from cognitive psy-
cognitive limitations make it dif?cult for
chology to the study of deterrence is con-
foreign policy actors to evaluate all possible
sistent with Ostrom’s call to the discipline.
scenarios and make a universally rational
The most productive theories of politics will
choice. One response to this critique is that
be those built upon the actual characteristics
classic rationality can be replaced with ‘satis-
of individual choice.
fying’ or a ‘bounded’ conception of ration-
Still, rational deterrence has established
ality. The idea here is to utilize an assumption
itself as a powerful anchor for both theory
closer to the actual process of decisionmaking
and practice, producing a framework that has
while retaining the bene?ts of rational choice
‘dominated postwar academic thinking on
models, namely their parsimony and predic-
strategic affairs’ and lies at the core of the
tive breadth. However, scholars in the ?eld of
‘intellectual framework of Western military
deterrence have largely adopted a second
policy’ (Achen & Snidal, 1989: 143). Ulti-
position. This view argues that all theories are
mately, some blend of rational and cognitive
abstractions from a complex reality. What is
models will provide scholars with a compre-
important is not the truth of the assumptions
hensive understanding of military deter-
imbedded in the theory but the explanatory
rence. Recent developments in formal theory
power of the model.4
are suggestive.6 Rational and cognitive deter-
But the explanatory power of rational
rence can coexist as complementary pieces of
choice explanations is itself often questioned.
a larger whole. For example, the model to be
For example, Simon (1983) ?rst noted that
offered below treats threat credibility as
much of the explanatory work of rational
exogenous: it is assumed that deterrence
models comes not from the ‘Herculean’
threats range from highly credible to largely
assumption that people strictly maximize,
incredible. Using formal tools, Nalebuff
but from a rich set of auxiliary claims about
(1991) demonstrates that a central com-
what people want and how they view the
ponent of threat credibility lies in rival per-
world. Similar arguments are found in the
ceptions about the hidden costs of action.
international relations literature. The danger
The lower the hidden costs, the greater the
of emphasizing rationality lies in ‘drawing
actual strength of a state and the greater its
attention away from the areas that contain
willingness to implement threats. The key to
much of the explanatory “action” in which
maintaining credibility is the proper manipu-
we are interested’ (Jervis, 1988: 325).5 This
lation of rival perceptions about hidden costs
argument penetrated to the heart of political
through the judicious use of threats and
aggression. Too much or too little of either
4 For a general defense of this position see Waltz (1990).
5 For a discussion of formal rational choice in security
6 The debate over the utility of formal models in security
studies, also see Walt (1999).
studies is ongoing (Brown et al., 2000).

---

02berejikian (ds) 17/12/01 10:59 am Page 168
1 6 8
j o u rn a l o f PE AC E RE S E A RC H
volume 39 / number 2 / march 2002
can damage a state’s reputation and invite
Figure 1.
Deterrence as Game of Chicken
aggression. This model of credibility for-
Player X
mation and maintenance thus supplies
important pieces of the larger deterrence
Cooperate
Defect
puzzle not provided by cognitive deterrence.
Player Y
Cooperate
3, 3
2, 4
Conversely, cognitive deterrence can ?ll gaps
Defect
4, 2
1, 1
existing in some formal models. Fearon’s
(1994) interest-signaling model suggests a
sequence game wherein challengers select
outcomes is given in Figure 1. If both co-
themselves into or out of deterrence crises.
operate, the status quo continues. If one state
Interestingly, defenders with solid repu-
defects, deterrence has failed. If both defect,
tations will nonetheless often fail to deter
war is the result. The standard assumption is
highly motivated challengers, a ?nding con-
that each side prefers an advantage over its
?rmed by the analysis below. However, the
rival to the cooperative status quo, though
origins for motivation in the signaling model
each prefers the status quo to open con?ict.
are exogenous: states are assumed to possess
Stripped to its core, successful deterrence
high or low levels of motivation. Under the
rests in effectively communicating a willing-
framework of cognitive deterrence offered
ness to impose heavy costs upon a rival if it
here, a state’s motivation will be explained by
pursues unacceptable policies.
its level of risk acceptance.
The dilemma for states in such games is
More immediately, the question posed by
that each has an initial incentive to move off
cognitive psychology to the study of inter-
the status quo. This undermines the credi-
national politics is compelling. Rather than
bility of deterrent threats because ‘having to
asking if traditional rationalist assumptions
carry out this threat if deterrence fails hurts
are useful for theory construction, we should
the threatener, even though the threat itself is
ask whether or not it still makes sense to con-
what is supposed to prevent deterrence from
tinue with a model of choice that we know to
failing in the ?rst place’ (Brams & Kilgour,
be beyond the capacity of real-world actors.
1988: 46). We should then observe an
The question becomes more pressing given
immediate race to defection by both states
that we have at our disposal cognitive models
because the ?rst to defect gains a strategic
that are in fact accurate descriptions of the
advantage. But deterrence often does work.
actual process of decisionmaking.
In response, proponents of traditional deter-
rence theory offer several explanations for the
stability often observed in deterrence
Problems with Deterrence Theory
relationships.
The logic of military deterrence is commonly
One line of reasoning holds that states can
depicted as a game of chicken.7 Here each
deliver credible commitment threats. States
state has two choices. The ?rst is to support
rationally commit to a deterrent strategy if in
the status quo through cooperation. The
the long run such a commitment brings a
second is to overrun the status quo through
higher expected payoff than no commitment.
military action. The universe of possible The assertion boils down to the following:
7 While this metaphor is commonly associated with deter -
It is rational to execute the intention [to
rence, other incentive structures may also capture the
retaliate] if and only if it is utility maximizing
dynamics of deterrence games. For example, Zagare (1987)
to form it. From this follows that since it may
suggests that deterrence games are best represented by the
prisoner’s dilemma. For simplicity of presentation, the
be rational to form the intention to execute the
game of chicken will be assumed throughout.
threat, it may also be rational to carry it out?if

---

02berejikian (ds) 17/12/01 10:59 am Page 169
Je f f rey B e re j i k i a n
A C O G N I T I V E TH E O RY O F D E T E R R E N C E
1 6 9
the intention has been formed, and deterrence
the assumption that states are rational;
fails, then a rational agent who intends to
without that assumption prediction is lost.
retaliate should do so because acting upon this
Similarly, empirical studies have produced
intention is part of the behavior required of an
expected utility maximizer. (Zagare, 1990:
widely different results and little or no con-
252)
sensus on a set of variables that predict
successful deterrence (Jervis, 1989; Harvey,
The argument here is that deterrence suc-
1995). Early ?ndings demonstrated that the
ceeds when both states are able to communi-
relative size of the threatening state is of little
cate the fact that each intends to undertake
importance. Military inferiority is not a
action that is irrational at the time it must be
suf?cient deterrent (Zinnes, North & Koch,
taken, but rational at the time the commit-
1961). Indeed, evidence suggests aggressors
ment is made.
often understand that they are militarily infe-
Many ?nd this an unsatisfying solution
rior, that they would lose an all-out war, and
that clouds rather than clari?es the logic of
that aggression on their part signi?cantly
deterrence. If deterrence should fail, ‘it is
increases the probability of a war (Russett,
irrational to retaliate, even on a probabilistic
1967). One possible explanation is that
basis, because retaliation leads to a worse
defender states will sometimes ?ght to
outcome for the threatener, having to carry
manipulate their reputation for toughness.
out his threat, as well as for the player who
However, there are theoretical reasons to
preempted and thereby provoked retaliation’
believe that such attempts are likely to prove
(Brams & Kilgour, 1988: 46).8 Alternatively,
counterproductive (Jervis, 1984a), and
states might deploy a tripwire as a signal of
empirical results con?rm that states that have
resolve in order to increase the credibility of
fought in the past are no better at deterring
deterrence threats. However, the compara-
rivals then those states that have not (Huth &
tively small losses incurred by sacri?cing the
Russett, 1984). Additionally, many factors
tripwire troops should not, under rational
associated with deterrent behavior are them-
deterrence, lead to a decision to accept the
selves not often predicted by the theory. For
much larger costs of war.9 Others suggest that
example, the greater the level of target dis-
states can form preferences for irrational
satisfaction with the status quo the less effec-
actions – like responding to an attack – and
tive deterrence threats are. One explanation is
that once formed such preferences are usually
that dissatisfaction offsets ‘perceptions of
carried out (Gauthier, 1984). Critics cor-
insuf?cient capability’ thus leading to aggres-
rectly note that this opens a Pandora’s Box
sion even in the case where smaller states face
wherein states are free to create preferences
more powerful rivals (Zinnes, North &
for all kinds of irrational goals. Once states
Koch, 1961: 470). But contemporary deter-
are permitted irrational goals, there are no
rence theory would deny this result. A state
theoretical restrictions on the sort of
dissatis?ed with the status quo should not,
irrational behavior they can undertake. The
under rational deterrence, undertake a course
deductive power of deterrence theory rests in
of action that would cause further deterio-
ration or decline in position.
8 Nalebuff (1991) uses formal logic to demonstrate that
Most deterrence scholars argue that an
states may ?nd retaliation rational because it enhances
empirical consensus will follow with further
reputation. My analysis assumes that states are not playing
reputation games.
theoretical development (e.g. Achen &
9 It is possible that tripwires are one way in which states
Snidal, 1989), and agreement over the oper-
might create so-called ‘audience costs’ in the domestic
ationalization of key terms (e.g. Huth &
polity, creating an incentive to respond (Fearon, 1997). But
this has not been empirically established.
Russett, 1990). Meanwhile, critics remain

---

02berejikian (ds) 17/12/01 10:59 am Page 170
1 7 0
j o u rn a l o f P E AC E RE S E A RC H
volume 39 / number 2 / march 2002
convinced that deterrence theory must be
the value function for losses is convex. The
recreated, rather than simply patched, if it is
principal distinction between prospect
to provide important guidance to states
theory and rational choice lies in this asym-
hoping to avoid con?ict (George & Smoke,
metrical relationship between gains and
1989; Lebow, 1989). The growing call for
losses. Expected utility theory holds that
behavioral theories of politics ?rmly
decisionmakers evaluate the desirability of
grounded in the actual capacities of real-
outcomes against their net asset position, and
world decisionmaking, combined with the
have a single function against which both
theoretical and empirical dif?culties found in
gains and losses are evaluated. By contrast,
standard deterrence theory, suggests the need
the core ?nding under prospect theory is that
for a reworking of military deterrence theory
decisionmakers evaluate each choice anew
along cognitive lines.
and against a neutral reference point (Kahne-
man & Tversky, 1979; Thaler, 1980). Rather
than a single utility function against which all
Prospect Theory
outcomes are evaluated, decisionmakers
Critics of contemporary deterrence theory
instead have two such functions, one for
also argue that often states simply fail to
gains and one for losses.
make rational choices. Empirical work
This relationship is given in Figure 2. The
supports this claim (Jervis, Lebow & Stein,
graph’s origin represents the reference point
1985). For some, this suggests that the foun-
against which gains and losses are evaluated.
dations for rational deterrence theory are
The s-shaped form of the value function cap-
suspect, while others argue that any deduc-
tures the diminishing relationship between
tive model of deterrence is inherently ?awed
objective gains and losses with subjective
(Steinbruner, 1983). However, that second
value. Note also that the value function for
position is correct only if non-maximizing
losses is steeper than for gains. Losses hurt
behavior is random, so that deterrent threats
more than a gain feels good, or as Kahneman
elicit no predictable response from the target
and Tversky put it, this steeper losses curve
and therefore that states have no way to
re?ects the ‘observation that a loss has a
establish a good strategy. Indeed, both sup-
greater subjective effect than an equivalent
porters and critics of deterrence theory tend
gain’ (Kahneman & Tversky, 1982: 166).
to equate non-rational behavior with unpre-
There are two important implications of
dictable choices. By contrast, prospect theory
this model for the study of international
suggests that non-rational choices are both
politics. The ?rst is the so-called framing
stable and predictable. This makes a deduc-
effect. The essentials of framing can be
tive theory of deterrence possible.
described with an example lottery.
The cornerstone for prospect theory lies in
the observation that, subjectively, there is a
•
Gains frame: imagine a choice between
diminishing return to continually increasing
two options that imply personal gains.
gains. For most people, an initial windfall of
Option 1: a sure gain of $80. Option 2:
$1,000 is more highly valued than is the same
a risky venture with an 85% chance of
$1,000 when added to an initial gain of
winning $100 and a 15% chance of
$10,000. That is, the value function for indi-
winning nothing. When confronted
viduals with respect to gains is concave, rep-
with this choice, most prefer the sure
resenting a curvilinear relationship between
gain to the risky venture although the
increasing gains and subjective value. A
expected monetary outcome is $5 less.
similar relationship exists for losses such that
•
Losses frame: now imagine a second set of

---

02berejikian (ds) 17/12/01 10:59 am Page 171
Je f f rey Be re j i k i a n
A C O G N I T I V E T H E O RY O F D E T E R R E N C E
1 7 1
Figure 2.
Subjective Utility Functions under Prospect Theory
options that imply personal losses.
individuals might be either risk averse or risk
Option 1: a sure loss of $80. Option 2:
acceptant across choices, but not both.
an 85% chance of losing $100 and a
Pre f e rence re versal also breaks with the
15% chance of losing nothing. Here
rational model in that the decisional determi-
individuals prefer the gamble to the sure
nant is not the expected outcome, but how the
thing although the expected monetary
choice is perc e i ved re l a t i ve to the status quo.
loss is $5 greater.
Individuals don’t maximize even when they
h a ve perfect information. Indeed, in an
This ?nding is quite robust. Studies deliber-
i m p o rtant sense, objective outcomes become
ately designed to refute it have re c o n ?rmed the
u n i m p o rtant because ‘the same decision can be
basic result (Grether & Plott, 1979). Te r m e d
framed in several different ways and differe n t
p re f e rence re versal, this ?ip in pre f e rences is
frames lead to different decisions’ (Kahneman
not predicted under rational choice because it
& T ve r s k y, 1982: 165). Fi g u re 3 plots the s-
violates the principle of invariance. It should
shaped utility function over the set of gains for
not matter how prospects are framed; under
the hypothetical lottery given above. No t i c e
rational choice, individuals ought to be con-
that in the gains quadrant an individual would
sistent in their pre f e rences no matter how they
p refer a sure gain of $80 to an 85% chance to
a re presented. Because expected utility theory
win $100.1 0 Prospect theory predicts, and
assumes that individuals evaluate gains and
10
losses against net asset positions and along a
In the losses quadrant, the same individual would prefer
the gamble offering an 85% chance of losing $100 to a
single utility function, it predicts that certain loss of $80.

---

02berejikian (ds) 17/12/01 10:59 am Page 172
1 7 2
j o u rn a l o f P E AC E RE S E A RC H
volume 39 / number 2 / march 2002
empirical results con?rm, that decisionmakers
owns, but the pain of forsaking it’ (Nincic,
a re risk averse when confronted with choices
1997: 99).
over gains and risk acceptant when confro n t e d
Prospect theory thus models the subjec-
with choices over losses.
tivity of actual decisionmaking. An indi-
The second implication of the s-shaped
vidual’s attitude toward risk is not simply a
value function is loss aversion. In Figure 2,
function of expected outcomes. Rather, risk
the negative value of a $500 loss is subjec-
disposition is in?uenced by ‘whether the out-
tively greater than the positive value of an
comes are perceived as gains or losses’ relative
equivalent gain. Loss aversion implies that
to a reference point (Quatrone & Tversky,
decisionmakers will act more aggressively to
1988). The important idea here is that our
avoid a loss than to secure an equal gain, and
theories of deterrence lag behind our know-
will pursue loss aversion beyond a rational
ledge of human decisionmaking. Risk accep-
expectation of bene?ts. The tendency here is
tant and non-maximizing behavior is not
to place a greater value on that already pos-
integrated into models of deterrence, despite
sessed compared to equivalent goods not yet
the fact that this is sometimes how decision-
acquired. This ?nding is consistent with an
makers act. For example, under prospect
observed endowment effect, which is also
theory, the deployment of deterrence threats
supported empirically. Here, individuals con-
can have several unintended consequences
sistently overvalue that which they already
that are not predicted by rational deterrence.
possess such that the endowment effect
Credible threats might generate a losses
‘enhances not the desirability of what one
frame for the target state and thereby push it
Figure 3.
The Framing Effect

---

02berejikian (ds) 17/12/01 10:59 am Page 173
Je f f re y Be re j i k i a n
A CO G N I T I V E TH E O RY O F D E T E R R E N C E
1 7 3
into risky behavior. The target of a threat
cooperation, so long as there is also some
might then risk breaking off negotiations
smaller probability of loss. This means that
when the threatening state was simply
under a gains frame even less than perfectly
attempting to secure a better deal, or threats
credible deterrence threats will provide a
may induce the target to attack even if there
suf?cient deterrent. By contrast, in a losses
is only a low probability of success. This is
frame, states will risk defection even if the
exactly the sort of behavior the initial threat
expected value of this strategy is less than
was intended to deter. All of this is explored
continued cooperation, so long as there is a
in detail below. The point here is that our
chance that this course of action will return
understanding of decisionmaking under risk
them to a condition of normalcy.
is quite different with prospect theory as the
Prospect theory does not predict the
assumed model of choice.
content of the decisional frame; that is,
whether a state views the status quo of deter-
rence as acceptable or unacceptable. A state’s
Deterrence and Decision Frames
assessment of the status quo is an empirical
The framing effect identi?ed under prospect
question. As a guide, one could look to recent
theory suggests that state assessments about
changes in a state’s strategic condition.
the attractiveness of the status quo play a
Assuming a near-perfect correlation between
central role in explaining deterrence behav-
a state’s objective condition and its percep-
ior. If continued mutual deterrence is accept-
tion of that condition, changes in the status
able, then states confront a choice between
quo would then provide a guide to the
selecting the known bene?ts of the status
content of the decisional frame.11 However,
quo, or pursuing a gamble intended to
decisionmakers also possess unique world-
further improve their position through defec-
views. Documenting the content of the
tion. When this gamble contains an expected
frame therefore requires evidence about the
value of further gain but also some proba-
state’s actual condition and a reconstruction
bility of loss (i.e. open con?ict), the state
of decisionmaker perceptions. Political sub-
would be in a gains frame. Prospect theory
jectivity thus plays an important role in an
predicts that the state would forgo the oppor-
analysis using prospect theory. The possi-
tunity to improve its position and opt for the
bility of a cognitive frame – that is, where
certainty of the status quo. Conversely, a state
only the perception of the state is operative
operating under an unacceptable status quo
regardless of objective circumstances – is con-
also has an opportunity to improve its
sistent with the observation that the reference
position through defection. When this
against which outcomes are evaluated is the
gamble contains an expected value of further
condition to which ‘one has become adapted’
loss, but also some probability of returning to
and that there are many cases in which it is
an acceptable status quo, the state would be
in a losses frame. Prospect theory predicts
11 Depending upon the nature of change, a gains or losses
that the state would accept the gamble.
frame will exist for some period after change has occurred.
The speed with which updating takes place, that is ‘how
This produces a general theorem for deter-
long one retains an original point of reference’ in assessing
rence: deterrence is more likely to be effective
the value of recent change is an important question for
when both states are in a gains frame, and less
prospect theorists, but one that is little understood (Sha?r,
1992: 316). Jervis, for one, suggests that it takes longer to
likely to be effective when either or both are in
incorporate losses into the reference point than gains
a losses frame. Under gains, a state will not risk
(Jervis, 1992: 200). This tendency is at best a guide to
empirical research, not an ironclad theory of updating.
disturbing the status quo even if the expected
Ultimately, a state’s assessment of the status quo is an
value of defection is greater than continued
empirical question; it cannot be pre-speci?ed theoretically.

---

02berejikian (ds) 17/12/01 10:59 am Page 174
1 7 4
j o u rn a l o f P E AC E RE S E A RC H
volume 39 / number 2 / march 2002
Figure 4a. Mutual Deterrence in a Gains Frame
Figure 4b.
Mutual Deterrence in a Losses Frame
Player X
Player X
Cooperate
Defect
Cooperate
Defect
Player Y
Cooperate
3, 3
1, 4
Player Y
Cooperate
–1, –1
–2, 1
Defect
4, 1
–1, –1
Defect
1, –2
–3, –3
‘determined by events that are only imagined’
represent dissatisfaction with outcomes: the
(Kahneman & Tversky, 1982: 171–172).
larger the negative number the further a state
Thus, states may have a view of their circum-
is from an acceptable condition. We can now
stances that has few connections to reality.
represent the same set of relationships given
However, the existence of a purely subjective
in Figure 1 and introduce the notion of a
frame would require thorough documen-
reference point.12 In standard deterrence
tation.
theory, open con?ict is the worst possible
outcome, and the one both sides wish to
Prospect Theory and Deterrence
avoid most. It is possible then that states
?nding themselves here would view their
Mutual Deterrence
condition as unacceptable. The value –1
Traditionally, the incentives for a deterrence
given to this cell is meant to represent that it
game like that given in Figure 1 are rep-
is unattractive compared to the deterrent
resented using ranked preferences. Unfortu-
status quo. Of course, a larger negative value
nately, ranking outcomes does not tell us if
is possible, say, if we were comparing con-
either state is satis?ed with the deterrent
ventional deterrence to nuclear deterrence.13
status quo as is required for an analysis using
But this would not alter the underlying
prospect theory. The matrix tells us that
dynamics of the game. Figure 4b also rep-
mutual cooperation is preferred to relative to
resents the logic of mutual deterrence, only
unilateral cooperation and mutual defection,
here the status quo of cooperation is
but not if it is a set of conditions with which
unacceptable to both states. Consistent with
the state is ultimately satis?ed.
the standard game, open con?ict is worse
Instead of ranked preferences, interval
still, while unilateral cooperation sits
payoffs allow the introduction of status quo
between open con?ict and mutual deter-
assessments while retaining the incentive
rence. In this game, the only acceptable
structure given in Figure 1. This permits a
outcome for each state is unilateral defection.
direct comparison of traditional and cogni-
13
tive deterrence. Consider Figure 4a. Cell
It has been suggested that at the extremes of ‘very small
probabilities or for catastrophic losses’ the framing effect
payoffs depict state assessment of the various
may break down because it is overwhelmed by the distri-
outcomes in comparison to what each per-
bution of costs and bene?ts attendant to the choice set
(Levy, 1995: 3). In nuclear deterrence, it may be that the
ceives to be an acceptable condition. Positive
potential costs of aggression are so large that they do in fact
values represent increases in satisfaction over
overwhelm the framing effect. This raises an important dis-
an acceptable status quo. Negative values tinction between total versus limited nuclear war as deter-
rents. The notion of a limited nuclear war suggests that the
costs of con?ict are not beyond consideration. Therefore, it
12 The transition to interval payoffs offered here retains the
may be that cognitive deterrence is appropriate for analyz -
preference order for chicken games as given in Figure 1.
ing limited war deterrents while rational choice is a better
From this point forward, the analysis assumes interval
guide to understanding total war threats. This issue clearly
payoffs for both cognitive and rational deterrence.
begs further examination.

---