Bird song has been extensively investigated as a model for understanding the physiological basis for animal vocalization.. Juvenile songbirds acquire their songs and perfect them as they transition into adulthood, just as we acquire our native language by exposure and imitation of adult tutors. Scientific investigation of bird song requires the collection of hundreds of hours of audio data containing songs, calls, and cage noise. These data must be sorted into categories of interest for specific research questions, with singing behavior being the dominant behavior of interest. Data categorization is a tedious and time-consuming process, and while current software hastens this process, substantial human effort is still required. This project investigates whether or not machine learning algorithms can be used to more efficiently categorize audio data collected in songbird research. Specifically, we developed a convolutional neural network (CNN) in PyTorch to classify whether or not 0.5 second sections of audio contain bird song. Using a supervised learning paradigm, we trained the CNN using labeled spectrograms (visual representations of audio frequencies across time) acquired from recordings of the zebra finch (Taeniopygia guttata). After training the CNN, we implemented it into an algorithm that identifies song within audio recordings. We then compared the CNN based software to a pre-existing, custom-written LabVIEW template-matching algorithm to determine the relative speed and accuracy of the software. Recordings were taken in both noisy and quiet recording environments to test the strengths and limitations of the two approaches. Our data indicate that the CNN based algorithm achieves comparable levels of accuracy to the pre-existing algorithm and accomplishes the categorization using a fraction of the time required by the template matching program. . These results suggest that machine learning algorithms can effectively be used to automate and rapidly categorize stereotyped vocal patterns. Further development of this software may facilitate rapid analysies of data and be extended to categorization of a broader range of vocal patterns, including human speech.

Hormonal contraceptives can have many negative side effects that deter women from using them. One example that many women are unaware of is alcohol craving. Preliminary studies in our lab have shown that women on hormonal contraceptives have greater alcohol cravings than women who are naturally cycling. Given that this data is mostly survey-based, we aim to add a research manipulation in the current study. We are testing this through a Qualtrics survey, distributed via Amazon’s Mechanical Turk (Mturk), that is designed to prime alcohol cravings. We will ask women to rate their alcohol cravings before and after they watch a video containing alcohol. We expect that when primed with an alcohol video, women’s cravings towards alcohol will increase to a level higher than before they watched the video. Further, we expect to find that women on hormonal contraceptives will have a higher increase in cravings than women who are regularly cycling. If we do find that women on hormonal contraceptives have a stronger reaction to an alcohol cue, women would benefit from being well informed about this effect. For instance, armed with the knowledge that hormones influence craving and behavior, women may be more mindful about their drinking habits. In addition, we eventually hope that this knowledge will influence those who are developing future contraceptives to take these side effects into account.

Recent research has begun to explore the basic misperceptions that underly political divides. For instance, people tend to believe that their political opponents accept objective moral wrongs (e.g., homicide, watching child pornography). These misperceptions then motivate avoidance and dehumanization of political opponents. However, the socio-cognitive processes preceding the formation of these misperceptions are less understood. Across two studies, we examined existential isolation towards political opponents, or the belief that people with a different political orientation than you do not understand your perspective and worldview, as one such social determinant. Study 1 surveyed 194 undergraduate students, and Study 2 surveyed 250 adults via Amazon Mechanical Turk. Results provide consistent support to suggest that individuals feel more existentially isolated from political opponents (e.g., politically liberal individuals report feeling more existentially isolated from politically conservative others). The more existential isolation people felt towards liberal or conservative others, the more they believed that these individuals endorsed objective moral wrongs, the less willingness they were to engage in political discussions with these individuals, and the more they dehumanized them. These findings emphasize the important role of existential isolation in the formation and persistence of political divides, and highlight the need for interventions that target feelings of existential isolation towards one’s political opponents.

Psychological well-being is shaped by an individual’s ability to buffer existential anxiety through self-esteem, cultural worldviews, and close relationships. However, prior research suggests that trauma weakens these mechanisms, increasing vulnerability to distress. Studies indicate that individuals with high trauma exposure struggle to reinforce cultural values in response to mortality salience, leaving them susceptible to psychological disorders. This study examines whether disrupted anxiety-buffering mechanisms contribute to increased fear of death and lower well-being. Participants, which are college students, (N=100) will complete measures assessing childhood trauma (ACE), fear of death (CL-FODS), and well-being (SWLS, PANAS). It is hypothesized that high-trauma individuals will report greater death anxiety and lower well-being compared to their low-trauma counterparts. This research aims to refine models of trauma’s psychological impact and inform interventions designed to restore effective anxiety-buffering mechanisms in survivors.

There is substantial literature exploring public perceptions of police, with many studies focusing on demographic factors such as race, age, prior police encounters, and neighborhood characteristics as key influences. While these factors are important, there remains a gap in research examining the public’s perceptions of law enforcement's abilities in handling public health emergencies, particularly opioid overdoses. This gap is concerning as law enforcement often serves as the first responders to such crises. The current study aimed to address this gap by interviewing residents of Tarrant County regarding their perceptions of law enforcement’s ability to effectively intervene in opioid overdose situations. Participants were recruited from various public locations across Tarrant County (N = 72). As part of the interview process, participants completed a nine-question survey using a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree) to assess their confidence in police responses to opioid overdoses. Results revealed that while there were marginal variations in survey scores, no significant differences were observed based on sex or education level. Overall, the findings suggested a moderate level of confidence among Tarrant County residents in law enforcement’s ability to effectively respond to opioid overdoses. Future research should further explore the factors influencing these confidence levels and develop interventions aimed at strengthening trust-based relationships between Tarrant County residents and law enforcement.