Introduction
The constitutionality of capital punishment in the United States is premised on the claim that the legal framework the capital jury uses to decide whether to impose the death penalty does not produce sentences that are ‘arbitrary and capricious’ or based on discriminatory factors such as race and ethnicity (Furman v. Georgia, 1972). Yet, research suggests that jurors’ capital punishment decisions are influenced by extra-legal characteristics of the trial, such as the race of the victim (Baldus et al., Reference Baldus1998) and the stereotypicality of black defendants (Eberhardt et al., Reference Eberhardt2006). Here, we examine another factor that may underlie capital punishment decisions: variation in jurors’ social cognition.
The process of mentally simulating the perspectives of other people has been theorized to play a central role in how people make punishment decisions. Prior work has made predictions about the antecedents (Preston and De Waal, Reference Preston and De Waal2002; Batson et al., Reference Batson2005; Ku et al., Reference Ku, Wang and Galinsky2015; Bloom, Reference Bloom2017) and consequences of perspective-taking for punishment decisions (Kogut, Reference Kogut2011; Skorinko et al., Reference Skorinko2014; Okonofua et al., Reference Okonofua, Paunesku and Walton2016). This past research has been limited to laboratory settings or field contexts concerning relatively minor offenses. Using data from real capital punishment trials, we test a series of predictions from previous research about the role of perspective-taking in punishment decisions. In doing so, we document how perspective-taking may introduce extra-legal variability into capital punishment decisions (Kahneman et al., Reference Kahneman, Sibony and Sunstein2021).
Methodologically, whereas previous work has primarily focused on experimentally manipulating perspective-taking (for a review, see McAuliffe et al., Reference McAuliffe2020), the current research design aims to describe jurors’ perspective-taking patterns. With this goal in mind, our analysis reports theoretically informed correlations and conditional relationships between variables; the design is not causal, and thus, we only speculate about causal interpretations of the data in light of existing theory. Prior investigations of perspective-taking have distinguished between (a) activation (do people attempt to perspective-take for the target?) vs (b) calibration and adjustment (how accurately do people perspective-take for the target?; Epley and Caruso, Reference Epley and Caruso2012; Eyal et al., Reference Eyal, Steffel and Epley2018). Our data only allow us to observe the former – whether and for whom jurors perspective-take – and we cannot observe the latter on the quality of their perspective-taking.
In the following sections, we discuss hypotheses derived from the prior literature about the predictors of perspective-taking and the relationship between perspective-taking and capital punishment decisions.
Predictors of perspective-taking
The first aim of the current research is to test two hypotheses grounded in psychological theory about the predictors of perspective-taking in punishment decisions. Researchers have theorized that, given that people hold various stereotypes and biases about others, willingness to perspective-take may depend on the demographic characteristics (e.g., ethnicity) of the people they are judging (Xu et al., Reference Xu2009; Cikara and Fiske, Reference Cikara and Fiske2011; Bloom, Reference Bloom2017). People may selectively refrain from perspective-taking for groups toward whom they hold negative stereotypes to avoid exposure to stereotype-disconfirming evidence (Hamilton et al., Reference Hamilton, Sherman and Ruvolo1990) or to avoid noticing similarity with such groups (Davis et al., Reference Davis1996). Relatedly, people may be less likely to perspective-take when they are in a position of power, relative to the group they are judging (Tjosvold and Sagaria, Reference Tjosvold and Sagaria1978; Galinsky et al., Reference Galinsky2006, Reference Galinsky, Rucker and Magee2016).
Differential use hypothesis: People’s likelihood of perspective-taking for trial participants (i.e., defendants and victims) will depend on the demographic characteristics of the trial participants.
Whereas previous research has primarily focused on how perspective-taking can mitigate racial bias (Vescio et al., Reference Vescio, Sechrist and Paolucci2003; Todd et al., Reference Todd2011, Reference Todd, Bodenhausen and Galinsky2012), the current research will document racial disparities in who people perspective-take for.
As a more specific instantiation of differential use, prior research has theorized that perceived similarity is a key antecedent of perspective-taking (Preston and De Waal, Reference Preston and De Waal2002; Batson et al., Reference Batson2005). That is, people are more likely to perspective-take for targets that are more similar to themselves (Cialdini et al., Reference Cialdini1997; Williams et al., Reference Williams, Parker and Turner2007; Ku et al., Reference Ku, Wang and Galinsky2015). Convergent evidence from social neuroscience has found that people exhibit greater empathy – observed in their neural responses – for same-race individuals compared with other-race individuals (Avenanti et al., Reference Avenanti, Sirigu and Aglioti2010; Han, Reference Han2018). Perspective-taking itself is often conceptualized as an ‘ego-centric anchoring’ process in which when people begin with their own viewpoint and then adjust toward the target’s (Epley et al., Reference Epley2004, Reference Epley, Caruso and Bazerman2006; Ames et al., Reference Ames2008). Because such adjustments are cognitively taxing (Lin et al., Reference Lin, Keysar and Epley2010), people may be more likely to adopt the perspectives of people who are similar to themselves because it requires less effortful adjustment.
Similar resemblance hypothesis: People are more likely to perspective-take for trial participants who are demographically similar to themselves.
We will test these hypotheses by analyzing how the demographic characteristics of jurors, defendants and victims uniquely predict jurors’ perspective-taking patterns.
Perspective-taking and punishment
The second aim of the current research is to test three hypotheses about the relationship between jurors’ willingness to perspective-taking and capital punishment decisions. We do so by analyzing how jurors’ perspective-taking patterns predict whether they vote for a life sentence or the death penalty.
Broadly, perspective-taking may improve liking and prosociality toward the target (Krebs, Reference Krebs1975; Coke et al., Reference Coke, Batson and McDavis1978; Toi and Batson, Reference Toi and Batson1982; Cialdini et al., Reference Cialdini1987, Reference Cialdini1997; Schaller and Cialdini, Reference Schaller and Cialdini1988; Dovidio et al., Reference Dovidio, Allen and Schroeder1990; Maner et al., Reference Maner2002). In punishment decisions, in particular, previous work has theorized and shown that perspective-taking for perpetrators leads to greater pity (Kogut, Reference Kogut2011), lower perceived culpability (Skorinko et al., Reference Skorinko2014) and greater forgiveness (McCullough et al., Reference McCullough, Worthington and Rachal1997), which all may reduce punitiveness toward perpetrators.
Perspective-taking for perpetrators predicts leniency hypothesis: People who perspective-take for defendants are less likely to vote for the death penalty (vs life without parole).
However, other work has argued that perspective-taking may instead increase moral condemnation in situations where the perpetrator is deemed to have malevolent intentions – which may often be the case in capital punishment trials (Lucas et al., Reference Lucas, Galinksy and Murnighan2016). Consistent with this account, Small and Loewenstein (Reference Small and Loewenstein2005) show that providing identifying (vs general) information about perpetrators may intensify anger toward the perpetrators and increase punitiveness. These results could lead to the following prediction:
Perspective-taking for perpetrators predicts punitiveness hypothesis: People who perspective-take for defendants are more likely to vote for the death penalty (vs life without parole).
Our research design allows us to adjudicate between these competing hypotheses.
Meanwhile, taking the perspective of victims may elicit compassion toward victims (Pfattheicher et al., Reference Pfattheicher, Sassenrath and Keller2019) and anger toward perpetrators (Kogut, Reference Kogut2011), which can intensify punitive inclinations toward perpetrators. Perspective-taking for victims may also strengthen perceptions of the perpetrator’s culpability (Skorinko et al., Reference Skorinko2014) and, more broadly, activate retributive goals during punishment (Carlsmith, Reference Carlsmith2006; Gromet et al., Reference Gromet2012).
Perspective-taking for victims predicts punitiveness hypothesis: People who perspective-take for victims are more likely to vote for the death penalty (vs life without parole).
This psychological process is central to extant legal arguments that victim impact statements may bias juries toward harsher punishments (Harris, Reference Harris1991; Bandes, Reference Bandes1996; Nadler and Rose, Reference Nadler and Rose2003). Importantly, however, recent lab experiments found limited evidence that perspective-taking for defendants and victims influences hypothetical capital punishment decisions (Skorinko et al., Reference Skorinko2024).
Our research tests these hypotheses using the Capital Jury Project (CJP), which is the largest dataset on real capital punishment decisions collected to date (N = 1,198 jurors). By describing how jurors perspective-take in real capital punishment trials, we highlight new implications of social cognition for criminal justice theory and policy.
Data
The CJP includes data from in-person interviews with 1,198 jurors (51.4% female; mean age = 46.2, SD age = 12.6) from 353 capital trials across 14 states (Bowers, Reference Bowers1995). Northeastern University’s Institutional Review Board granted the project approval. Trials in the CJP took place between 1988 and 1995. The inclusion of trials within states was random, with some additional constraints: there was an attempt to obtain roughly equal numbers of trials that ended in life sentences and death penalties, and there were geographical limits imposed in large states (CA, FL and TX) so that traveling between interviews was feasible for the interviewers. In the final data, 514 (43%) of the 1,198 jurors voted for life sentences (rather than the death penalty).
The CJP sought to recruit 4 randomly selected jurors per trial, but in some cases, recruited fewer (39 trials included just 1 juror) and in other cases, recruited more (14 trials included 6 or more jurors). Interviews included closed-ended survey questions about the juror, victim, defendant and the trial.Footnote 1 Results reported in the manuscript include data from jurors who had non-missing data for all variables included in the respective regressions. All findings are robust to imputing missing values (see Supplementary Materials; Buuren and Groothuis-Oudshoorn, Reference Buuren and Groothuis-Oudshoorn2011). The sample size and voting decisions are broken down by demographics in Table 1. The demographic composition of defendants and victims is shown in Table 2. Note that similar summary statistics as in Tables 1 and 2 have been reported in previous publications using the CJP data (e.g., Bowers, Reference Bowers1995; Devine and Kelly, Reference Devine and Kelly2015).Footnote 2
Demographic composition of jurors and voting decisions
Demographic composition of victims and defendants
The survey included four questions that measure perspective-taking for defendants and five questions that measure perspective-taking for victims (Table 3).
Perspective-taking questions
We constructed a defendant perspective-taking index (the share of four questions for which the juror reported perspective-taking for the defendant or the defendant’s family) and a victim perspective-taking index (the share of five questions for which the juror reported perspective-taking for the victim, the victim’s friends, or the victim’s family). For instance, a juror who responded with ‘Yes’ to all four questions about perspective-taking for defendants would have the max value of 1 for the defendant perspective-taking index, whereas a juror who responded ‘No’ to all four of those questions would have the minimum value of 0.
Descriptively, jurors took the perspective of victims (mean = 0.43; SD = 0.34) more than defendants (mean = 0.23, SD = 0.26; t(1145) = 19.67; p < 0.001), and the victim and defendant perspective-taking indices were positively correlated (Pearson r = 0.40, p < 0.001). Figure S1 shows the correlation and joint distribution of the two indices.
Results
Predictors of perspective-taking
To investigate whether the demographic characteristics of defendants and victims predict jurors’ perspective-taking patterns, we estimated the following two specifications:
\begin{equation}perspective\_taking\_for\_defendan{t_{ig}} = \alpha + \gamma 'defendant\_demographic{s_g} + {\varepsilon _{ig}}\end{equation}
\begin{equation}perspective\_taking\_for\_victi{m_{ig}} = \alpha + \delta 'victim\_demographic{s_g} + {\varepsilon _{ig}}\end{equation} where i indexes jurors and g indexes trials. 
$perspective\_taking\_for\_defendan{t_{ig}}$ reflects the defendant perspective-taking index and 
$perspective\_taking\_for\_victi{m_{ig}}$ reflects the victim perspective-taking index. 
$defendant\_demographics_g$ and 
$victim\_demographics_g$ reflect vectors of demographic variables – race/ethnicity, gender and age – for defendants and victims, respectively. All specifications use OLS regressions with standard errors clustered by trial. Note that all significant p-values testing the main hypotheses reported in this section and the rest of the manuscript are robust to Benjamini and Hochberg (Reference Benjamini and Hochberg1995) and Holm (Reference Holm1979) corrections unless otherwise specified (corrected p-values are reported in section 3 of the Supplementary Materials). The hypotheses and corresponding empirical support from our analyses are summarized in Table 4. Note that all regression tables reporting results on the predictors of perspective-taking are in section 1 of the Supplementary Materials.
Empirical support for main hypotheses
Consistent with the differential use hypothesis, jurors were more likely to perspective-take for white victims relative to black victims (B = 0.11, clustered SE = 0.04, p = 0.002, standardized b = 0. 25).Footnote 3 We did not find, however, that people perspective-take differentially based on any other demographic characteristics of victims or defendants. See the left panel of Figure 1. These results hold when re-estimating models (1) and (2), except simultaneously including 
$defendant\_demographic{s_g}$ and 
$victim\_demographic{s_g}$ as independent variables in both models, as well as controlling for juror demographics (race/ethnicity, gender, age, education, employment status and income) and trial state.
Whose perspectives get taken and who takes others’ perspectives. The top left panel displays how jurors’ perspective-taking for victims depends on victim demographics. The bottom left panel displays how jurors’ perspective-taking for defendants depends on defendant demographics. The right panel displays how jurors’ perspective-taking for victims (top) and defendants (bottom) depends on juror demographics. The black points reflect sample means and the error bars reflect 95% confidence intervals, with standard errors clustered by trial. The translucent blue points reflect each juror’s value, where darker areas reflect greater density in the distribution, marginalized over the entire sample. Note that the survey instrument only asked people whether they were male or female, limiting the ability to identify non-binary people in the data.
As an exploratory analysis, we examined whether jurors differ in their propensity to perspective-take based on their own demographics. We found that older jurors, relative to younger ones, were less likely to perspective-take for both defendants and victims (B = −0.002, clustered SE = 0.001, p < 0.001, standardized b = −0.11; B = −0.004, clustered SE = .001, p < 0.001, standardized b = −0.14); and, women, relative to men, were more likely to perspective-take for victims (B = 0.07, clustered SE = 0.02, p < 0.001, standardized b = 0.16). See Tables S3 and S4 in the Supplementary Materials for more information on this exploratory analysis, and see the results visualized in the right panel of Figure 1.Footnote 4
To what extent, though, are differences in perspective-taking driven by similarity between the perspective-taker and the trial participant? To examine whether demographic similarity between jurors and trial participants predicts perspective-taking, we estimated the following two OLS regressions with standard errors clustered by trial:
\begin{align}perspective\_taking\_for\_defendan{t_{ig}} &= \alpha + \eta 'juror\_defendant\_similarit{y_{ig}} \nonumber\\
&\quad + \theta 'juror\_demographic{s_{ig}} + {\varepsilon _{ig}}\end{align}
\begin{align}perspective\_taking\_for\_victi{m_{ig}} &= \alpha + \omega 'juror\_victim\_similarit{y_{ig}}\nonumber\\
&\quad+ \theta 'juror\_demographic{s_{ig}} + {\varepsilon _{ig}}\end{align} where 
$juror\_defendant\_similarit{y_i}$ and 
$juror\_victim\_similarit{y_i}$ each reflect a vector of three indices that measure similarity between the juror and respective trial participant in terms of race/ethnicity, gender and age. For instance, 
$juror\_defendant\_similarit{y_{ig}}$ includes race/ethnicity and gender indices that are simply binary variables indicating whether the defendant’s reported race/ethnicity and gender match the juror’s. The age similarity index was constructed by taking the absolute difference between juror and defendant age, dividing it by 60, and subtracting that value from 1 (i.e., jurors and defendants who are the same age have a similarity index of 1, while those who are 60 years apart have a similarity index of 0).Footnote 5 We used analogous methodology to construct the three variables that index similarity between victims and jurors (reflected by the vector, 
$juror\_victim\_similarit{y_{ig}}$). The models also control for juror race/ethnicity, age, and gender.
Using these models, we did not find evidence that demographic similarity between the perspective-taker and target predicts perspective-taking. Specifically, none of the similarity indices statistically significantly predicted perspective-taking for defendants or victims (Tables S5–S6). As a robustness check to assess demographic similarity, we used interaction terms, rather than similarity indices and found similar results (Tables S7–S10).
Perspective-taking and punishment
To estimate the relationship between perspective-taking and capital punishment decisions, we estimated the following linear probability model with standard errors clustered by trial:
\begin{align}capital\_punishment\_vot{e_g} &= \alpha + {\beta _1}perspective\_taking\_for\_defendan{t_{ig}} \nonumber\\
&\quad + {\beta _2}perspective\_taking\_for\_victi{m_{ig}} + {\varepsilon _{ig}}\end{align}where the outcome is a binary variable indicating whether the juror’s final vote was for life (vs death), and the defendant and victim perspective-taking indices are simultaneously included as the independent variables of interest. The coefficients reported in the manuscript are from models without additional covariates. The regression tables for the primary specifications and aforementioned robustness checks are reported in section 2 of the Supplementary Materials.
Jurors who took the victim’s perspectives were more likely to vote for the death penalty rather than a life sentence (B = −0.19, clustered SE = 0.06, p < 0.001, standardized β = −0.08). The coefficient represents a 19-percentage point difference in the probability of voting for a death sentence between jurors who always perspective-take for victims and jurors who never perspective-take for victims. These effects hold after controlling for demographics of jurors, victims and defendants; the trial state; and jurors’ perceptions of the aggravating and mitigating circumstances in the trial (Devine and Kelly, Reference Devine and Kelly2015). This suggests that trial characteristics, such as the severity of the crime, may not fully explain the relationship between perspective-taking and voting.
We also found initial evidence that jurors who took the defendant’s perspectives were more likely to vote for a life sentence rather than the death penalty (B = 0.19, clustered SE = 0.07, p = 0.004, standardized β = 0.05). Note, however, that this estimate is not robust to a Holm–Bonferroni correction, nor is it robust to the inclusion of the aforementioned control variables. Due to these mixed results, we classify this evidence as weakly supporting the perspective-taking for perpetrators predicts leniency hypothesis. We disaggregate the indices and display the relationships between each of the nine perspective-taking items and capital punishment decisions in Figure 2.
Perspective-taking and capital punishment decisions. The x-axis represents the difference in the likelihood of voting for a life sentence over the death penalty (e.g., jurors who imagined being like the defendant were 13 percentage points more likely to vote for a life sentence instead of the death penalty, compared to those who did not imagine themselves being like the defendant). All coefficients are estimated with linear probability regressions without additional covariates, with standard errors clustered by trial. The coefficients for the defendant and victim perspective-taking indices were estimated in the same regression. The defendant vs victim perspective-taking index coefficient reflects the difference in life voting between jurors who always perspective-take for defendants and never for victims, and jurors who always perspective-take for victims and never for defendants. The benchmark effects at the bottom of the graph show two seminal effects from prior research: the differences in life voting between (1) jurors who perceived that the defendant was remorseful and those who did not and (2) black and white jurors.
To differentiate jurors along a single measure, we constructed a combined perspective-taking index by taking the difference between the defendant and victim indices and re-scaling so the index ranges from 0 (always perspective-take for victim, never for defendant) to 1 (always perspective-take for defendant, never for victim).Footnote 6 We estimated the following simplified linear probability model with standard errors clustered by trial:
\begin{align}capital\_punishment\_vot{e_g} =& \alpha + \vartheta perspective\_taking\_for\_defendant\_vs\_victi{m_{ig}} \nonumber\\
&\quad+ {\varepsilon _{ig}}\end{align}As depicted in Figures 2 and 3, higher scores on the combined perspective-taking index were associated with a higher likelihood of voting for life over death (B = 0.38, clustered SE = 0.10, p < 0.001, standardized β = 0.08). The coefficient represents an estimated 38-percentage point difference in the probability of voting for a life sentence between: (a) jurors who always perspective-take for defendants and never for victims (perspective-taking index = 1) and (b) jurors who always perspective-take for victims and never for defendants (perspective-taking index = 0). In Figure 4, we plot the results of 32 different specifications to highlight how the estimated relationship depends on modeling decisions (Simonsohn et al., Reference Simonsohn, Simmons and Nelson2020). There, the standardized β estimates, all of which are statistically significant, range from 0.05 to 0.08 depending on the specification.
Defendant vs victim perspective-taking index and capital punishment decisions. The bottom panel shows the distribution of the defendant vs victim perspective-taking index. The top panel shows the relationship between the perspective-taking index and voting for life sentences. The black slope and gray band reflect the predicted values and 95% confidence intervals from a linear probability model with standard errors clustered by trial. Each black dot reflects the average proportion of people voting for life within one of four evenly spaced bins of the perspective-taking index (0–0.249, 0.25–0.499, 0.5–0.749 and 0.75–1). The error bars reflect 95% confidence intervals, using standard errors clustered by trial.
Specification curve on the relationship between the defendant vs victim perspective-taking index and final voting. The top panel includes the standardized coefficients, ordered by magnitude, from 32 regression specifications that each estimate the relationship between the defendant vs victim perspective-taking index and final voting. The bottom panel indicates what was included in each of the 32 specifications. Specifically, each specification is represented by a column that contains two ‘x’ marks; the upper ‘x’ indicates the covariates that were included in the model and the lower ‘x’ indicates whether the data excludes observations with missing values or imputes missing values. ‘Juror demographics’ (C1) include jurors’ race/ethnicity, age, gender, education, income and employment status; ‘Def. & victim demographics’ (C2) include victims’ race/ethnicity, age and gender; and defendants’ race/ethnicity, age and gender; ‘State’ (C3) refers to the location of the state where the trial took place; and ‘Agg. & Mit. Circumstances’ (C4) includes jurors’ perceptions of aggravating and mitigating circumstances. Confidence intervals account for clustering by trial. To interpret the plot, the first column’s estimate, for instance, is from a regression that controls for defendant and victim demographics, state and aggravating and mitigating circumstances; and uses data that excludes observations with missing values. The upward trend in the ‘x’ marks in the bottom panel reflects that the estimated coefficient generally attenuates as additional controls are added into the model.
Despite including a large set of controls, unobserved differences between trials could still pose an omitted variable bias.Footnote 7 To address this possibility, we analyzed jurors’ first votes at the start of deliberation before consensus was reached, so we could compare jurors within the same trial. Using trial fixed effects, we found consistent evidence that perspective-taking for the defendant rather than the victim (greater values on the combined perspective-taking index) positively predicted initial life voting (B = 0.22, clustered SE = 0.09, p = 0.014, standardized β = 0.05). Breaking this down into its components – by including both victim and defendant perspective-taking indices in the same regression instead of the combined index – we found that perspective-taking for victims predicted marginally lower likelihood of initially voting for a life sentence (B = −0.09, clustered SE = 0.05, p = 0.057, standardized β = −0.04),Footnote 8 and perspective-taking for the defendant predicted higher likelihood of initial voting for a life sentence (B = 0.15, clustered SE = 0.06, p = 0.019, standardized β = 0.04). Perspective-taking predicted capital punishment votes among jurors within the same trial, thinking about the same defendants, victims, crimes and trial performances.
Even after controlling for differences between trials, there may be differences between jurors’ psychological processes that pose omitted variable bias. Namely, jurors’ sympathy for defendants and victims could affect both their perspective-taking patterns and punishment decisions. After controlling for whether jurors’ felt sympathy toward defendants and victims, the relationship between the perspective-taking index and voting remained significant (B = 0.32, clustered SE = 0.10, p = 0.002, standardized β = 0.07). Similarly, the relationship holds after controlling for jurors’ perceptions of the defendant’s remorse, which has previously been identified as one of the strongest juror-level predictors of capital punishment decisions (B = 0.38, clustered SE = 0.10, p < 0.001, standardized β = 0.08; Bowers, Reference Bowers1995; Eisenberg et al., Reference Eisenberg, Garvey and Wells1998).
Although we were able to partially mitigate the risk of omitted variable bias, our analytical approach alone is unable to disambiguate the causal direction of the relationship between perspective-taking and capital punishment. It is possible that a) perspective-taking patterns affect jurors’ death penalty decisions or b) death penalty decisions affect jurors’ reported perspective-taking patterns. There are theoretical reasons why the former causal story is highly plausible. Namely, previous research has theorized and shown that exogenously increasing perspective-taking for victims increases anger and punitiveness toward perpetrators (Kogut, Reference Kogut2011), and increasing perspective-taking for perpetrators increases pity and leniency (Kogut, Reference Kogut2011; Okonofua et al., Reference Okonofua, Paunesku and Walton2016). Our results are consistent with these causal theories.
There are two possible reverse causal theories related to biased recall and initial decisions. Biased recall refers to the possibility that jurors’ final decisions affect their recall of their perspective-taking patterns. For instance, jurors may selectively recall perspective-taking patterns that are consistent with their capital punishment decisions and selectively forget patterns that are inconsistent with their decisions (Mather et al., Reference Mather, Shafir and Johnson2000). That said, the data-gathering approach was deliberately designed to capture jurors’ authentic reflections on their decision processes during the trial. Specifically, the interviews took place shortly after the trials, and jurors did not extensively communicate with each other between the trials and interviews (fast internet communication was not available). Further, interviewers were also trained to make clear that there were no ‘right’ and ‘wrong’ answers to their questions and all responses would be anonymized. The second reverse causal theory refers to the possibility that jurors made up their minds about whether they would vote for life or death early in the trial, entirely independent of perspective-taking, and that initial decision independently affected both their subsequent perspective-taking patterns and final votes. To address this possibility, we found that, among jurors who did not initially vote for a life sentence at the start of deliberation, perspective-taking for defendants, rather than victims, predicted whether they changed their mind to vote for a life sentence by the end of the trial (see Table S18 in the Supplementary Materials).
Discussion
The current research describes perspective-taking patterns in capital punishment trials. The results first shed new light on the predictors of perspective-taking. We found evidence consistent with theorizing that people perspective-take differentially depending on the demographic characteristics of the people they are judging (Xu et al., Reference Xu2009; Cikara and Fiske, Reference Cikara and Fiske2011; Bloom, Reference Bloom2017); Jurors are more likely to perspective-take for white, rather than black, victims. While previous research has primarily focused on how perspective-taking can combat racial bias (Vescio et al., Reference Vescio, Sechrist and Paolucci2003; Todd et al., Reference Todd2011, Reference Todd, Bodenhausen and Galinsky2012), our results provide new evidence that racial bias may affect who people choose to perspective-take for. Of note, other features of the trial – such as attorneys’ defense strategies or victim impact statements – may shape perspective-taking and be deployed differentially based on the victim’s race (Harris, Reference Harris1991; Bandes, Reference Bandes1996; Nadler and Rose, Reference Nadler and Rose2003). Further research is needed to disentangle whether jurors exhibit spontaneous bias in whose perspective they adopt, or whether such disparities stem from contextual features of the trial itself.
In contrast to predictions from previous research, however, we did not find evidence that jurors are more likely to perspective-take if the trial participant is demographically similar to themselves (Preston and De Waal, Reference Preston and De Waal2002; Batson et al., Reference Batson2005; Williams et al., Reference Williams, Parker and Turner2007; Ku et al., Reference Ku, Wang and Galinsky2015). One interpretation of our findings is that stereotypes guide perspective-taking in capital punishment decisions, and in-group distinctions matter less than previously theorized (Greenwald and Krieger, Reference Greenwald and Krieger2006). As an alternative interpretation, when engaging in social categorization, jurors may not view demographic characteristics as a relevant dimension for comparing themselves to capital punishment defendants (Tajfel and Turner, Reference Tajfel and Turner1979); put plainly, people may not view themselves as similar to an atypical person – in this case, one who committed extreme acts of violence – merely because they share the same race, age, or gender. Rather, in contexts with atypical perpetrators like capital punishment trials, similarity on less visible characteristics – such as life experience – may be stronger predictors of perspective-taking.
We further discovered new evidence that jurors’ propensity to perspective-take for others may depend on their own demographic characteristics, including their age and gender. Specifically, we found that older jurors perspective-take less for both perpetrators and victims, and women are more likely to perspective-take for victims than men. We hope these new discoveries spur future psychological research to better understand why these demographics relate to perspective-taking.
In investigating the relationship between perspective-taking and punishment, we found evidence that jurors who take the perspectives of victims are more likely to vote for death sentences (Carlsmith, Reference Carlsmith2006; Kogut, Reference Kogut2011; Gromet et al., Reference Gromet2012; Skorinko et al., Reference Skorinko2014). However, we found mixed evidence for the hypothesis that jurors who take the perspectives of defendants are more likely to vote for life sentences. These mixed findings are broadly consistent with prior theory arguing that the relationship between perspective-taking for perpetrators and moral condemnation may be moderated by other factors (e.g., perceived intentions; Lucas et al., Reference Lucas, Galinksy and Murnighan2016).
Notably, we found that perspective-taking for victims (rather than defendants) predicted initial votes for the death penalty among jurors within a given trial. This suggests that – beyond trial characteristics such as victim race – there is also variation in perspective-taking tendencies across people on the same jury bench that may influence punishment judgments. A defendant’s fate may thus be partially determined by the composition of the jury – specifically, the proportion of jurors inclined to adopt the victim’s perspective rather than the defendant’s.
Implications for criminal justice and policy
Our findings speak to critical issues that sit at the intersection of psychology and policy. The constitutionality of capital punishment in the United States is premised on the claim that the legal framework the capital jury uses to decide whether to impose the death penalty does not produce sentences that are ‘arbitrary and capricious’, or based on discriminatory factors such as race and ethnicity (Furman v. Georgia, 1972). Our findings raise questions about whether the death penalty is being applied in a way that aligns with these tenets.
Namely, we found that jurors’ propensity to take the perspective of victims (rather than defendants) is associated with the decision to impose the death penalty (rather than life in prison), even after controlling for aggravating and mitigating circumstances. Whether such an influence of perspective-taking on jurors’ decision-making would constitute a violation of legally permissible standards remains an open question. On the one hand, attorneys are often legally permitted to present evidence that may affect perspective-taking (e.g., the impact of the crime on the victim’s family). From this view, some scholars may still consider perspective-taking within legal standards. On the other hand, perspective-taking for victims may preclude jurors from being able to give meaningful consideration to mitigation, which is constitutionally required by Lockett v Ohio (Sundby, Reference Sundby2014). More generally, perspective-taking may introduce subjective factors – such as perceptions of a victim’s ‘worthiness’ – into the decision-making process, reflecting possible deviations from the constitutional expectation that jurors use rational and consistent principles to make their decisions (Sundby, Reference Sundby2002). Jurors’ perspective-taking patterns, therefore, represent an additional crucial dimension in the ongoing debate of whether the jury’s death penalty decision can ever be structured to avoid the influence of ‘arbitrary’ factors.
Jurors also perspective-take differentially based on race: they are more likely to perspective-take for white victims than for black victims. Notably, prior work has found that defendants are more likely to receive the death penalty if their victims are white rather than black (Baldus et al., Reference Baldus1998), particularly when defendants are more stereotypically black (Eberhardt et al., Reference Eberhardt2006). This ‘white victim effect’ formed the basis for a major challenge to capital punishment. The Supreme Court narrowly rejected this challenge, in part because the majority believed the capital juror’s decision involved so many factors that one could not identify how race was influencing the juror’s decision (McCleskey v. Kemp, 1987). Our results, however, may offer insight into a possible mechanism: defendants with white victims may be more likely to receive the death penalty because jurors are more likely to perspective-take for white victims, a finding that would suggest that the victim’s race may be influencing the decision in a constitutionally forbidden manner.Footnote 9
Limitations
Given the infrequency of capital punishment decisions, the current research uses the largest dataset on real capital punishment decisions collected to date. This gives rise to several important limitations. First, the hypotheses we can analyze are limited by the items that were included in the CJP survey. Notably, the survey did not elicit emotional empathy, so we cannot separately investigate perspective-taking and emotional empathy (Galinsky et al., Reference Galinsky2008; Longmire and Harrison, Reference Longmire and Harrison2018). Furthermore, while we can examine whether jurors perspective-take, the CJP does not provide insight into the quality of that perspective-taking (Epley and Caruso, Reference Epley and Caruso2012; Eyal et al., Reference Eyal, Steffel and Epley2018). For instance, upon perspective-taking for defendants, it is possible that jurors relied on stereotypes related to criminality that distorted their simulation of the defendant’s perspective (Devine, Reference Devine1989; Eberhardt et al., Reference Eberhardt2004). Future research should examine what jurors imagine when taking others’ perspectives to clarify how this process influences capital punishment decisions.
Second, jurors were only surveyed in 14 states, during one time period from 1988 to 1995. When the CJP was conducted, support for the death penalty was substantially higher than it is now (>70% in favor compared to 53% today; Gallup, 2025), and the number of executions per year was increasing (reaching its maximum of 98 in 1999, compared to 25 in 2024; Death Penalty Information Center, 2025). Testing the replicability of the CJP findings today is difficult given the infrequency of capital punishment decisions. Regardless, we hope future research will again survey capital jurors and test whether the effects generalize over time and across sociopolitical contexts.
Finally, as noted earlier, we caution that our current research is limited by its non-causal design. That said, our findings are consistent with and contribute to a body of work that has provided causal evidence regarding how and why perspective-taking influences punitive judgment (Kogut, Reference Kogut2011; Skorinko et al., Reference Skorinko2014; Okonofua et al., Reference Okonofua, Paunesku and Walton2016), albeit in lower stakes contexts. Considering that field experiments cannot be implemented in death penalty trials, future work must identify naturally occurring exogenous variation in jurors’ tendency to perspective-take to estimate the causal effect on capital punishment decisions.
Conclusion
Reflecting on the death penalty after nearly 25 years on the Supreme Court, Justice Harry Blackmun concluded (Callins v. Collins, 1994):
It seems that the decision whether a human being should live or die is so inherently subjective – rife with all of life’s understanding, experiences, prejudices, and passions – that it inevitably defies the rationality and consistency required by the Constitution.
The current results are consistent with the idea that subjective processes – in this case, perspective-taking – may influence a juror’s capital punishment decisions.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/bpp.2025.10014.
Acknowledgments
We dedicate this article to the late Samuel Sommers, whose undergraduate course in Psychology and Law first inspired this project and whose incisive feedback strengthened the manuscript. The authors also thank Susan Bandes, Eugene Caruso, Marla Sandys and David Tannenbaum for their helpful comments on earlier drafts of this paper, Keith Chen for his helpful feedback about the research design, and Marla Sandys and the researchers who conducted the CJP for collecting and sharing the data used in this research. The authors are grateful to Nicholas Burns, Celia Gleason, Haley Karchmer, Bruce Mei, Akshatha Prabhu, and Mier Shao for their outstanding research assistance.
Funding statement
This material is based upon work supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-1650604. The CJP was originally funded by the National Science Foundation Law and Social Sciences Program under Grant No. SES-9013252.
Competing interests
The authors declare that they have no competing interests.
Data availability statement
The survey instrument from the CJP is available in our ResearchBox page. The CJP data are not available in a public depository due to the original data-sharing agreement with the jurors. To access the data, researchers must submit a proposal to the CJP’s acting administrator. The code used for all analyses reported in the manuscript is available on the project’s ResearchBox page. The ResearchBox page is available here: https://researchbox.org/1148&PEER_REVIEW_passcode=PNUFEJ.
Open access
