Without, it appears, any trace of irony, the iconic black and white POW–MIA flag flies at Camp Justice at the U.S. prison at Guantanamo Bay. But irony there is. The prison is synonymous with indefinite detention and the abuse of prisoners in a war against stateless terrorists, while the flag honors and agitates for the return of American prisoners of war in conventional interstate wars. Geoffrey P. R. Wallace’s valuable book, Life and Death in Captivity, unites the concern about prisoner treatment with a focus on interstate wars in an effort to explain the significant variation in prisoner abuse across time and space.

Wallace defines prisoner abuse as “a military strategy enacted by political and military authorities that involves the intentional killing or harming, either directly or indirectly, of enemy combatants who have laid down their arms and surrendered” (pp. 16–17). In addition to executions, harming includes five other violations of the Geneva Conventions: torture, hazardous labor, poor housing or nutrition, denial of legal rights, and compulsory military conscription. To fall under Wallace’s definition, these abuses must be intentional government policy, which admits willful ignorance and neglect, rather than spontaneous and nonsystematic abuses at lower levels.

The source for measuring prisoner abuse between 1898 and 2003 is a modified version of the Correlates of War data set, disaggregated into pairs of captor–captive “directed dyads” for each party to the conflict. Thus, for example, World War II includes the –United States–Germany and Germany–United States as two observations. Countries that did not have control over their prisoners due to alliances or simply had no prisoners to (mis)treat are not “capable captors” and are eliminated from the main analysis. Measurement difficulties stemming from wartime chaos and propaganda and the desire to capture the other forms of violence visited upon prisoners led Wallace to classify the different types of abuse into high, medium, and low levels, except for the denial of legal rights and compulsory military conscription, which were measured dichotomously. Drawing on both primary and secondary sources, each directed dyad was coded with the highest level of abuse it received in any of the categories. Wallace’s survey of these data show substantial variation over time and wars, with no level of abuse dominating consistently.

Wallace argues persuasively that conventional explanations relying on cultural differences (on the one hand) or military culture (on the other hand) inadequately account for variation in prisoner treatment within the same war over time, or against different states in the same conflict. Instead, Wallace views states as confronting a real dilemma about how to treat prisoners, a difficult choice to treat them well or abuse them. They resolve this dilemma on the basis of strategic considerations rather than normative commitments. Wallace focuses on two sets of factors. The first is regime type; democracies tend to engage in lower levels of abuse out of fears of retaliation against their own troops and because the desire to end wars quickly makes them more likely to appreciate the strategic advantage that comes from convincing your enemy to surrender to good treatment rather than fight to the death. Since, however, democratic publics and leaders are motivated by interests rather than norms, when circumstances alter these incentives, changes in prisoner treatment follow. When democracies do not fear retaliation against their own prisoners or expect that reciprocity considerations will have no effect on the enemy, then abuse is more likely.

The second set of factors are related to the nature of the conflict being fought. In wars of attrition as well as in wars to capture and hold territory, abuse is more likely. In the latter case, states will be concerned to eliminate potential sources of post–conflict opposition to their rule. In the former case, state leaders view the potential benefits from brutalizing the enemy into submission as outweighing the costs. Moreover, “desperate states will often feel pressured into adopting any strategy that improves their chances of victory, however remote” (p. 69). Alternative explanations include nonterritorial expansive war aims (regime change or unconditional surrender) and cultural differences between the dyad combatants.

The book employs a mixed-method design in Chapters 3 through 5 to test this argument against its competitors. After describing the measurements for the explanatory variables, the quantitative analysis first examines bivariate relationships between levels of prisoner abuse, on the one hand, and the strategic versus alternative factors, on the other. While the hypothesized relationships clearly hold for high levels of abuse—democracies perpetuate fewer such abuses than nondemocracies, and wars of attrition and territory result in more such abuses—the pattern is not consistent. Medium levels of abuse are more frequent than lower levels for both the hypothesized and competing factors.

A multiple (ordered logit) regression also finds support for the book’s claims about high levels of abuse, with the probability of high abuse dropping by 58% if the captor state is a democracy; it rises by 153% if the war is severe and by 232% if the war is for territory. However, a control variable, prisoner abuse within the dyad, also has a dramatic effect, increasing the likelihood of high abuse by 122% (p. 82). The appendix demonstrates that these findings are robust to a very comprehensive set of alternative specifications. In order to tease out the effects of norms from strategic incentives in democracies, the analysis tests for the effects of retaliation in the same war and shows that while democracies tend not to initiate abuse, they are just as willing to retaliate with abuse as nondemocracies. This analysis also examines the interaction between democracy and the nature of the war and shows that democracy reduces the probability of high levels of abuse if both sides hold prisoners, the democracy has expansive war aims, and it has ratified treaties on the laws of war. In contrast to the other evidence, democracy has no statistically significant effect on high levels of abuse in wars of attrition. More surprisingly, democratic regime type actually increases the likelihood of high abuse in wars of territorial annexation.

Although a few other cases are discussed briefly, the qualitative studies all come from World War II, with an entire chapter devoted to the Katyn massacre of Polish officers by the Soviets in 1940. The goal of these chapters is to look for fine-grained evidence of the hypothesized causal mechanisms. So, for example, in the first part of Chapter 4, Wallace adduces documents and draws on the secondary literature to show that policymakers in the United States and Great Britain were concerned about reciprocity and inducing Germans to surrender, rather than preoccupied with any normative concerns about how they should treat prisoners. Later, he shows how the abuse increased after the war when the United States forcibly repatriated former Soviet prisoners as reciprocity vis-à-vis the Germans waned, but that vis-à-vis the Soviets waxed (because the Soviets took over German camps with U.S. POWs in them). Chapter 5 provides a case study of territorial conquest (the Soviet occupation of eastern Poland following the German invasion in September 1939), and the Katyn massacre is interpreted largely on the basis of a memo from security chief Lavrentiy Beria to Stalin, as a direct result of concerns about eliminating any possible Polish resistance.

Life and Death in Captivity has multiple merits. First, the question and implications of its findings are important. Interstate wars remain very much a danger. Moreover, arguments about reciprocity figured prominently in the decision by the United States to mistreat Al Qaeda captives, and so as Wallace argues, some of the mechanisms he identifies travel to nonstate actors. Second, Wallace makes a generally convincing case that normative commitments and international law tend to get trumped by strategic considerations, suggesting the need to think of new strategies for preventing prisoner abuse. Third, the data set alone is a contribution to the quantitative conflict literature.

Theoretical and empirical questions temper these contributions somewhat. With respect to theory, the argument is more a collection of things that matter to states and democratic publics, rather than a set of more or less brittle propositions that might be proven false. At one point, for example, Wallace says that “states can even become socialized into appropriate norms of conduct’’ (p. 49). This may very well be true, but it is very different from the cost–benefit dilemma motivating the rest of the argument. Given that the approach and assumptions are self-consciously strategic and rational, some explicit statement of preferences over relative trade-offs between them might have been helpful. With respect to the empirical test, the description of the middle level of abuse as a residual category between high and low, the high frequency of middle-level abuse for the explanatory variables, and intercoder reliability of “about .7’’ all suggest the possible undue influence of measurement error (the exact test used for intercoder reliability is not identified, nor could I find the online appendix associated with the book). Finally, there is little explicit discussion of why and on what basis the case studies in the qualitative part of the book were selected.

Nevertheless, Life and Death in Captivity tells a cautionary tale about an important but understudied problem that is not likely to go away anytime soon.