Tracing Humanity in Final Moments: An Analysis on Death Row Farewells
Authors: Stella Koh (Project Lead), Deshna Govil, Leah Shin, May Nicole Cagulada, Aarya Khana
Regret? Rage? Relief? What do inmates facing death row feel right before their end?
To date, 591 executions have occurred in Texas, one of the 27 states that currently still authorize the death penalty. These inmates have been charged with capital offenses, primarily some specific form of capital murder. Spending an average of 11.22 years on death row, inmates have much time to reflect on their own crimes before their scheduled date of execution.
Right before their execution, death row inmates have the opportunity to make a final statement, often directly to witnesses, family members, and officials present. Some do not speak, some say a few words, and others dedicate long 3-paragraph speeches. These statements are a rare glimpse into the minds of death row inmates. Through a sentiment analysis of these last words, we aim to humanize the data surrounding capital punishment and provide a nuanced perspective on inmate reflections.
Understanding Inmate Demographics
Before profiling the emotional state of death row inmates, we first hope to provide insight into their demographics to put some context for the circumstances of individuals who face capital punishment.
The first chart, “Race and Average Age of Inmates,” highlights the racial diversity of those executed, revealing individuals from Black, Hispanic, White, and other racial groups face similar average ages at the time of execution, at approximately 40 years old. This suggests that despite variations in racial representation on death row, age is not a significant differentiator among these groups. This depiction also reflects the extensive period inmates typically spend on death row, as many enter the system in their 20s or early 30s. By examining these figures, we gain a better understanding of how systemic factors and justice processes converge across racial demographics.
The second graph, “Education Level vs. Inmate Count,” paints the picture of the educational attainment of death row inmates. The majority of individuals executed had education levels that peaked around the 12th grade, indicating a concentration of inmates with a high school diploma or equivalent. Conversely, there are steep declines in representation among those with higher education levels, while a small portion have less than a middle school education. This distribution highlights potential links between limited educational opportunities and pathways to criminal behavior, suggesting systemic inequities may disproportionately affect disadvantaged populations. Understanding these factors can deepen the broader discussion about the role of socio-economic status, education, and the likelihood of receiving the ultimate punishment in the justice system.
Factoring In Statement Lengthiness
Moreover, we took a look at what trends can be revealed from the length of the inmates’ final statements.
In the graph above, we analyzed the relationship between inmates’ ages and the average number of victims, while also exploring how the word count of the speakers’ last statements varies with their age. We see older inmates, particularly between the ages of 65–70, tend to have more victims as well as longer last statements. This suggests older individuals may have been involved in cases with greater complexity, potentially contributing to the higher number of victims and lengthier last statements. The increased length might also reveal a stronger desire from older individuals to deeply reflect on their actions and life experiences. This trend suggests, as individuals age, they have a greater awareness of the weight of their circumstances, seen through the use of their final statements for more significant reflection.
In addition, we analyzed how the length of the inmates’ final statements differs based on the races of their victims. The graph above illustrates the percentage distribution of victim races across various word count ranges in the final statements. For longer statements, we see the victims are primarily White. Victims of other races are much less represented in the higher word count brackets. This suggests the presence of potential systemic biases, where inmates may construct longer statements and express greater remorse when their victims are White. Additionally, it reflects the possibility of broader societal influences on how inmates perceive their actions based on the racial identities of their victims.
Most Common Words and Phrases
To get a better sense of the most frequently mentioned ideas or themes across all the inmates’ final statements, we created a word cloud which displays the most common words, with their size directly correlating to how often they are mentioned, as seen below:
The biggest takeaway from this word cloud is the high frequency of remorseful (“sorry”, “know”, “hope”) and grateful (“love”, “thank”, “God”) language. Being the most common words in final statements, they suggest inmates making a consistent effort to express regret for their actions, seeking forgiveness, or acknowledging any pain they might have caused. Another common theme lies in the expressive appreciation for family, friends, or faith — likely in an effort to find closure in their final moments. This emphasis on gratitude and spiritual reflection points to an effort of leaving behind a message of positivity or connection, despite the severity of their circumstances. Another method of examining common words/phrases is through analysis of the most commonly mentioned two-word expressions, known as bigrams.
This bar graph highlights the increased level of gratefulness the Texas death row inmates display through bigrams such as “stay strong”, “family love”, and “just want”. All of these phrases present attempts at providing reassurance and support to loved ones, with an emphasis on resilience even in the face of personal adversity. This trend appears to be tied to a further sentiment of seeking internal peace in a final attempt to communicate their emotional intentions. Not only do these bigrams reinforce themes of gratitude and remorse as seen in the word cloud analysis, but they also give us insight into valuable layers of emotional complexity within these death row inmates.
Emotion and Sentiment Breakdown
To organize the emotions behind each of the inmates’ last statements, we used sentiment analysis to categorize them into ten groups: anger, anticipation, disgust, fear, joy, negative, positive, sadness, surprise, and trust.
From the above bar graph, most inmates tend to express positive sentiments in their last living moments, with the most persistent emotions being “positive,” “joy,” and “trust.” On the opposite side of the spectrum, negative feelings are not as regular, with the most uncommon emotion being “disgust.” The greater frequency of positive emotions suggests that even in the cases of impending death, most humans tend to reflect on life with a favorable perspective. This may point to a broader human tendency to seek peace or acceptance in the final stages of life.
In this analysis, we categorized inmates based on the number of homicide victims they were responsible for — 1, 2, or 3 or more — to examine how sentiment varied in relation to the severity of their crimes. As depicted in the graph, the emotional distributions within each group followed a relatively consistent pattern, with positive emotions consistently comprising a larger proportion than negative emotions. This suggests that, regardless of the severity of their offenses, the inmates share common human emotional responses. In the face of extreme circumstances, such as the death penalty, these individuals appear to reflect more on the positive aspects of their lives, underscoring an innate behavior to seek solace and reminisce on the positives throughout their lives.
Conclusion
Thus, our findings suggest most death row inmates embrace their end with remorse towards victims, gratitude for all their loved ones, and positivity as they near the end of their life. We see much less rage and negativity in comparison to inner reflection.
In this wrap up, we do acknowledge the limitations of our analysis. Our primary method of analysis relied on assigning sentiments and emotions to unigrams and bigrams of statements. While this lexicon-based approach may still be an accurate reflection of overall trends, words pulled out of context may be subject to incorrect interpretations, especially when sarcasm is at play. Further work can be done on this topic with more modern machine learning approaches better suited to understanding nuances in full statements. Should more data be available about the inmates from prior case evaluations or interviews, future studies can also examine emotional changes over time to better characterize inmates.