Posttraumatic stress disorder (PTSD) co-occurs with substance use disorders at high rates, but the neurobiological basis of this relationship remains largely unknown. Identifying mechanisms that underlie this association is necessary, and recognizing pathologies shared by these disorders may provide pertinent information in understanding their functional relationship. Separate lines of evidence suggest that PTSD and drug addiction may share a common feature, that is, dysregulation of the brain's reward circuitry. We hypothesize that PTSD results in reduced dopaminergic neurotransmission which may contribute to deficient reward function and vulnerability to drug-seeking behavior. To address this hypothesis, we combined single-prolonged stress (SPS), a rodent model of PTSD, with a series of behavioral and neuropharmacological assays to assess dopaminergic reward function and cocaine intake. The results of the studies presented herein extend our understanding of the effects of severe stress on drug reinforcement and consumption, and establish a potential mechanism by which PTSD produces deficient reward function through alterations in the dopamine system. A modified SPS procedure consisting of 2 hours of restraint, 20 minutes of group swimming, isoflurane exposure until loss consciousness, and 7 days of isolation was used to induce severe stress in our studies. Initial studies were conducted to examine the effect of SPS on cocaine-conditioned reward and anhedonia-like behavior in adult male Sprague-Dawley rats. Using a biased conditioned place preference paradigm, unstressed controls demonstrated a significant preference for the cocaine-paired context following four pairings with cocaine (5-20 mg/kg, i.p.). Preference for the cocaine-paired side was significantly lower in rats exposed to SPS, suggesting a deficit in the rewarding properties of cocaine following exposure to severe stress. Anhedonia-like behavior was assessed by a two-bottle choice sucrose preference test. Robust consumption of sucrose solution (0.25-1%) was observed in rats that underwent control handling, however, SPS significantly reduced sucrose intake compared to controls. These results suggest an increase in anhedonia-like behavior or a reduction in the rewarding effects of sucrose as a non-drug reinforcer. Finally, basal behavioral activity in SPS rats was compared to unstressed controls in a 24-hour test. Results indicate a significant reduction in spontaneous nocturnal activity following SPS versus control handling. In contrast, hyperlocomotion induced by an acute cocaine injection (5-20 mg/kg, i.p.) was unaltered between rats that underwent SPS or control handling. These data suggest that deficient behavioral activity may be specific to voluntary movements or behavior, and support an increase in anhedonia following exposure to SPS. Intravenous cocaine self-administration was conducted to examine the effect of SPS on the acquisition, motivation, and escalation of cocaine intake. Acquisition of cocaine self-administration was studied using an escalating dose regimen in which rats had sequential access to 0.1875, 0.375, and 0.75 mg/kg/infusion on a fixed-ratio 1 schedule of reinforcement. Rats exposed to SPS did not significantly differ from control handled animals in the latency to meet acquisition criteria (consumption of 6.75 mg/kg/day for 3 consecutive days) or the general pattern and level of cocaine intake at each dose. A subsequent study assessing the breakpoint for cocaine self-administration using a progressive-ratio schedule of reinforcement determined a dose-dependent increase in motivation to work for cocaine (0-1.5 mg/kg/infusion) across both experimental groups. However, motivation to obtain cocaine was similar between SPS and unstressed rats, as there was no significant difference in breakpoint for cocaine self-administration at any dose of cocaine tested. To evaluate potential differences in the transition to escalated cocaine intake, self-administration was measured using an extended-access procedure in which unlimited cocaine (0.375 mg/kg/infusion) was available for six hours daily. Upon extended-access to cocaine, SPS significantly attenuated cocaine intake compared to control handling over 14 sessions. Despite a significant reduction in cocaine intake, rats exposed to SPS still significantly escalated their cocaine intake over the course of 14 days. These results suggest that escalation of cocaine intake occurred in the presence of lower total doses of cocaine in the SPS exposed animals compared to controls. In addition, SPS rats demonstrated a greater percent increase in cocaine consumption compared to controls. This finding suggests that rats exposed to SPS compensated for a decrease in cocaine reinforcement by escalating their intake to a greater magnitude than controls. These studies indicate that SPS may not alter the acquisition of cocaine self-administration or motivation for cocaine. However, the finding of reduced cocaine intake upon extended-access in SPS rats is consistent with a deficit in cocaine-induced reward. The ability of SPS rats to escalate cocaine intake in the presence of less cocaine, or a greater magnitude of escalated cocaine intake than controls, may reflect mechanisms leading to enhanced vulnerability to cocaine abuse. To understand the mechanisms of reduced reward and behavior in the SPS model of PTSD, a series of neurochemical assays was used to assess the ability of SPS to induce dysfunction of dopaminergic neurotransmission. Using high performance liquid chromatography, tissue levels of dopamine and the dopamine metabolites DOPAC and HVA were measured immediately and one week following SPS or control handling. Tissue obtained from SPS rats demonstrated significant decreases in dopamine, DOPAC, and HVA content in both the nucleus accumbens and caudate putamen immediately following SPS and one week later, suggesting a potential deficit in dopaminergic tone. Quantitative autoradiography was used measure the density of dopamine transporters and dopamine D1 and D2 receptors. [3H]WIN35428 binding to dopamine transporters was higher in the nucleus accumbens of SPS rats compared to controls, suggesting an increase in dopamine transporter density following severe stress. The level of [3H]WIN35428 binding in the caudate putamen was not different between groups. [3H]Raclopride binding to D2 receptors was significantly reduced in both the nucleus accumbens and caudate putamen following SPS versus control handling. These results suggest a decrease in the density of striatal D2 receptors. D1 receptor expression was not significantly altered by SPS, as no significant difference in [3H]SCH23390 binding was detected in SPS rats compared to controls. A preliminary functional assessment of dopamine transporters revealed a significant increase in dopamine uptake in the nucleus accumbens of SPS rats compared to controls, whereas uptake in the caudate putamen was unaltered between groups. Enhanced dopamine uptake following SPS is consistent with the increase in dopamine transporter density observed in the nucleus accumbens of SPS rats. Activation of D1 receptors and G-protein mediated transduction was assessed using an adenylyl cyclase assay with the D1 agonist SKF82958. In the caudate putamen, a significant decrease in D1 receptor-stimulated cAMP production was revealed in SPS rats compared to controls, whereas SKF82958-induced cAMP was unchanged in the nucleus accumbens. Finally, the function of D2 dopamine receptors was assessed by D2 receptor-stimulated [35S]GTPγS binding using quinpirole. In the caudate putamen, [35S]GTPγS binding following stimulation of D2 receptors was enhanced by SPS compared to control handling, whereas no difference was observed between groups in the nucleus accumbens. These results indicate increased D2 receptor-mediated activation of G-proteins in the caudate putamen following SPS. In summary, the studies described herein tested the hypothesis that reduced dopaminergic function may be a mechanism for deficient reward and heightened susceptibility to drug use in PTSD. Results demonstrated a significant reduction in cocaine-conditioned reward, as well as attenuated sucrose preference and spontaneous activity in rats exposed to SPS. These findings are consistent with the presence of a dysfunctional reward system which may contribute to anhedonia-like behavior in PTSD. Furthermore, reward deficits may promote altered patterns of cocaine taking behavior and vulnerability to substance abuse. Results demonstrated significant escalation of drug intake following exposure to SPS, which occurred in the presence of less cocaine than controls. A greater increase in cocaine intake was observed in SPS rats over the course of escalation, which may reflect a mechanism for enhanced vulnerability to the development of a substance use disorder in PTSD. Dopaminergic dysfunction may contribute to deficient reward capacity and an altered pattern of cocaine intake in SPS. SPS-induced alterations in dopamine function included a reduction in striatal dopamine content alongside enhanced dopamine transporter levels and function. Mild alterations in D2 receptor density and the function of D1 and D2 receptors were also observed. These findings support the hypothesis that PTSD results in reduced dopaminergic neurotransmission, which may contribute to deficient reward function and altered drug-seeking behavior. Identifying the pathology of PTSD, such as altered dopamine neurotransmission, may lead to enhanced treatment strategies and interventions to prevent substance abuse in persons with PTSD. / Pharmacology

Nicotine addiction is a multifaceted disease that can be influenced by several factors. Emerging evidence indicates that the neural substrates of nicotine addiction overlap with the neural substrates of learning and memory. Nicotine modulates various types of learning and memory and the ability of nicotine to alter cognitive processes may contribute to its addictive liability. Acute nicotine enhances contextual conditioning in mice, tolerance develops to this effect with chronic administration, and withdrawal from chronic nicotine produces cognitive deficits. While tolerance and withdrawal deficits both occur following chronic administration, it is unknown if they share similar mechanisms. The series of experiments in Chapter 2 were designed to provide evidence that tolerance and withdrawal are dissociable. C57BL/6J mice were implanted with osmotic minipumps that delivered constant nicotine or saline for various durations and then were trained and tested in contextual conditioning either during chronic nicotine administration or 24 hours after pump removal. Chronic nicotine enhanced contextual conditioning in a dose- and time-dependent manner. Tolerance developed quickly to the enhancing effect of chronic nicotine. Furthermore, the duration of chronic nicotine treatment required to produce cognitive deficits upon cessation of treatment differed than that required to produce tolerance, which suggests that tolerance and withdrawal are mediated by separate mechanisms. Chapter 2 concludes by presenting a model that integrates nicotinic acetylcholine receptor desensitization and upregulation to explain the present findings. The model presented in Chapter 2 predicts that there will be enhanced sensitivity to acute nicotine during a period of nicotine withdrawal. Previous research indicates that prior exposure to nicotine enhances sensitivity to acute nicotine injections, but it is unclear if this enhanced sensitivity is due to prior nicotine exposure or enhanced sensitivity to nicotine during withdrawal. Therefore, the experiments in Chapter 3 were designed to determine if prior exposure to nicotine or nicotine withdrawal altered sensitivity to acute nicotine injections. This was accomplished by assessing the effects of acute nicotine on contextual conditioning immediately after cessation of chronic nicotine treatment and two weeks later, a time period not associated with withdrawal-related changes in cognitive function. Results of the study showed that acute nicotine enhanced contextual conditioning across a wide range of doses in both saline- and nicotine-withdrawn mice. However, a greater enhancement of contextual conditioning was observed in mice withdrawn from chronic nicotine treatment for 24 hours than all other withdrawal groups, suggesting enhanced sensitivity during withdrawal. The enhanced sensitivity to acute nicotine suggests altered nAChR function during withdrawal. In addition, the lowest dose of acute nicotine did not enhance contextual conditioning in groups that received chronic nicotine but did in other groups. The simultaneous observation of a hyper and hyposensitive nAChR system during withdrawal suggests that there may be a phasic response to chronic nicotine. Together, the results of the present study suggest that tolerance and withdrawal operate under separate mechanisms, and that there is overall enhanced sensitivity to nicotine during periods of nicotine withdrawal. / Psychology

The naked mole-rat (Heterocephalus glaber) is one of the most social mammals on the planet. These animals live in underground colonies consisting of a breeding female (the “queen”), 1 to 3 breeding males and up to 300 nonreproductive “workers” organized in a dominance hierarchy. In addition to their eusociality, naked mole-rats have evolved many extreme biological characteristics including an exquisite sense of touch and insensitivity to certain types of pain. However, the overlap between the social and somatosensory abilities of naked mole-rats remains mysterious. In this work, we show that naked mole-rats exhibit caste-like behavioral signatures linked to eusociality, predominantly utilize snout-to-snout interactions in social behavior and possess a somatosensory profile to mechanical stimuli distinct from mice. In the first chapter of this work, we leveraged machine learning and molecular biology tools to create a behavioral atlas of naked mole-rat behavior. We first utilized a computational pipeline of pose-tracking using SLEAP and behavioral segmentation using keypoint-MoSeq to identify 20+ behavioral syllables. This showed that the queen naked mole-rat has a distinct behavioral phenotype from the workers, consisting of faster movements, less anxiety-like freezing, and less colony maintenance behaviors such as digging. We next showed that behavioral differences exist between the workers and that the dominance rank of a given individual could be predicted by its spontaneous behaviors in an open field. Relative rank differential appeared to have relevance for social behaviors: during a tube test for dominance, pairings of a high-ranked and low-ranked animal occurred more quickly and were won more frequently by the dominant animal. Snout touch played an integral role in these dominance tests (perhaps in the communication of individual ID information). This prompted us to investigate snout-to-snout interactions in the open field. We found that two familiar naked mole-rats from the same colony engaged in hundreds of snout interactions in a 10-minute period, and the number of interactions was nearly twice as high between two foreign animals. Follow-up experiments explored the molecular basis of this snout touch and showed that mechanosensory channels (e.g. Piezo2) are likely involved in social snout-to-snout interactions. Furthermore, trimming the sensory whiskers of naked mole-rats disrupts their ability to recognize conspecifics and alters their relative dominance relationships. These findings uncover face touch as a prominent social behavior in naked mole-rats that is intimately linked to social recognition. In the second chapter of this work, we more deeply investigated naked mole-rat somatosensation by using high-speed videography. We began by determining how naked mole-rats respond to both innocuous stimuli (cotton swab, dynamic brush) and noxious stimuli (light and heavy pinpricks) when given to the hindpaw. Compared to mice, naked mole-rats showed a distinct hindpaw phenotype, never responding to innocuous touch stimuli, responding to light pinprick, but rarely responding to heavy pinprick (normally the most noxious stimulus that elicits the strongest pain response in mice). Interestingly, naked mole-rats do respond to brush stimuli to the back skin with a dorsiflexion posture. These animals also exhibit an idiosyncratic withdrawal response to a brush applied to the snout skin which appears to be highly aversive. Interestingly, the velocity of this aversive snout withdrawal appears to be socially modulated and is decreased in the presence of another naked mole-rat in the testing chamber. This phenotype does not occur in mice and provides additional evidence that the naked mole-rat snout is not only extremely sensitive but plays a role in processing socially relevant information.

Restoration of the oyster reefs has become increasingly crucial due to great population declines around the globe. Intertidal oyster reefs provide essential foraging and loafing grounds to many faunal species, including several threatened/endangered wading bird species. Biodegradable oyster reef restoration materials have been introduced to avoid potential plastic pollution from traditional materials. Studies have shown success regarding oyster recruitment rates to these materials. However, their impacts on fauna using restored oyster reefs are unknown. This project aims to evaluate oyster reef restoration using biodegradable materials to increase faunal diversity, abundance, and foraging behaviors. Camera traps were deployed to observe fauna on reefs of the eastern oyster (Crassostrea virginica) in summer 2021, winter 2022, and summer 2022 in Mosquito Lagoon, FL. Treatments included Biodegradable EcoSystem Engineering Elements (BESE) shell mats, cement-jute tiles, and cement-jute rings. Unrestored, live reefs were used as positive controls, and unrestored, dead reefs (piles of disarticulated shell) were used as negative controls with three replicates of each treatment. A total of 11,458 vertebrates were observed out of 82,261 video clips. These comprised 44 species, including seven species of birds listed as threatened in the state of FL. There was a significant interaction between timeframe and treatment for non-foraging behaviors, such as loafing, grooming, and walking. Restoration materials did not decrease counts of foraging. However, foraging counts significantly varied over time, based on bird migratory patterns and time since restoration. This research provides essential information on the faunal use of restored and unrestored oyster reefs and highlights the importance of a mosaic of oyster reef types in estuarine systems.

Animals coordinate their behavior with each other during cooperative and agonistic social interactions. Such coordination often adopts the form of “turn-taking”, in which the interactive partners alternate the performance of a behavior. Apart from acoustic communication, how turn taking is coordinated, is not well known. Furthermore, the neural substrates that regulate persistence in engaging in social interactions are poorly studied. Here, we use Siamese fighting fish (Betta splendens), to study visually-driven turn-taking aggressive behavior. Using encounters with real conspecifics and with computer animations, we discover the visual cues from an opponent and the behavioral dynamics that generate turn taking. Through a brain-wide screen of neuronal activity during aggressive behavior, followed by targeted brain lesions, we then discover that the caudal portion of the dorsomedial telencephalon, an amygdala-like region, promotes continuous participation in aggressive interactions. Our work highlights how dynamic visual cues shape the rhythm of social interactions at multiple timescales and points to the pallial amygdala as a region controlling the drive to engage in such interactions.

Neuroscience as a field is collecting more data than at any other time in history. The scale of this data allows us to ask fundamental questions about the mechanisms of brain function, the basis of behavior, and the development of disorders. Our ambitious goals as well as the abundance of data being recorded call for reproducible, reliable, and accessible systems to push the field forward. While we have made great strides in building reproducible and accessible machine learning (ML) systems for neuroscience, reliability remains a major issue. In this dissertation, we show that we can leverage existing data and domain expert knowledge to build more reliable ML systems to study animal behavior. First, we consider animal pose estimation, a crucial component in many scientific investigations. Typical transfer learning ML methods for behavioral tracking treat each video frame and object to be tracked independently. We improve on this by leveraging the rich spatial and temporal structures pervasive in behavioral videos. Our resulting weakly supervised models achieve significantly more robust tracking. Our tools allow us to achieve improved results when we have imperfect, limited data while requiring users to label fewer training frames and speeding up training. We can more accurately process raw video data and learn interpretable units of behavior. In turn, these improvements enhance performance on downstream applications. Next, we consider a ubiquitous approach to (attempt to) improve the reliability of ML methods, namely combining the predictions of multiple models, also known as deep ensembling. Ensembles of classical ML predictors, such as random forests, improve metrics such as accuracy by well-understood mechanisms such as improving diversity. However, in the case of deep ensembles, there is an open methodological question as to whether, given the choice between a deep ensemble and a single neural network with similar accuracy, one model is truly preferable over the other. Via careful experiments across a range of benchmark datasets and deep learning models, we demonstrate limitations to the purported benefits of deep ensembles. Our results challenge common assumptions regarding the effectiveness of deep ensembles and the “diversity” principles underpinning their success, especially with regards to important metrics for reliability, such as out-of-distribution (OOD) performance and effective robustness. We conduct additional studies of the effects of using deep ensembles when certain groups in the dataset are underrepresented (so-called “long tail” data), a setting whose importance in neuroscience applications is revealed by our aforementioned work. Altogether, our results demonstrate the essential importance of both holistic systems work and fundamental methodological work to understand the best ways to apply the benefits of modern machine learning to the unique challenges of neuroscience data analysis pipelines. To conclude the dissertation, we outline challenges and opportunities in building next-generation ML systems.

Animal behavior has important impacts on animal populations and the ecosystem at large, but the impact of such behavior on many ecological phenomena is understudied. For example, behavior drives transmission between wildlife disease hosts. Space use and resource selection determines where hosts will make contact, movement determines how pathogens may spread over the landscape, and other fine-scale behaviors determine the rate of contact and transmission. Spatial and movement data from GPS telemetry are useful for studying the causes and consequences of many behavioral processes. One particular focus of such spatial analyses is the behavioral responses of prey to predation risk. While many studies have highlighted the broad impacts of these antipredator behaviors, few studies have emphasized how predation risk may impact the behavioral drivers of disease transmission. White-tailed deer (Odocoileus virginianus) are an excellent system to study these questions for three reasons. First, deer exhibit a fission-fusion social structure, so contacts are dependent on numerous interacting factors. Second, deer face varying predation risks and respond to these risks with varying strategies including spatial avoidance, foraging, and grouping behavior. Third, deer are host to many important diseases with differing transmission mechanisms. In this dissertation, I had three main objectives; 1) to evaluate the factors that produced variation in deer-to-deer contact, 2) to evaluate multiple behavioral responses of deer to predation risk and, 3) to use these behavioral patterns to make predictions of the relative risk of deer-to-deer contact.In chapter one, I evaluated population variation in contact and tested the impact of variation in contact-related behavior on inferences from social network analysis. I used camera trap recordings of visits and behaviors by deer to scrapes throughout DeSoto National Wildlife Refuge, Nebraska from 2005 and 2006. Based on 2,013 interactions by 169 unique identifiable males and 75 females, I produced social networks based on indirect contact among deer at scrapes, with edges weighted based on the frequency, duration, and types of behaviors. Social networks based on scrape-related behavior were highly connected and dependent upon the frequency, duration, and type of behavior exhibited at scrapes (e.g., scraping, interacting with a scrape or overhanging branch, rub-urinating, grazing) as well as the age of the deer. Including behavior when defining edges did not preserve the network properties of simpler measures (i.e., unweighted networks) confirming that heterogeneity in behaviors that affect transmission probability are important for inferring transmission networks from contact networks. In chapters two through five, I evaluated the behavior of deer using movement data from GPS collars. I captured and collared white-tailed deer (Odocoileus virginianus) at two sites: Shelbyville, IL, and Carbondale, IL from January 2020 to March 2022. I collared a total of 156 deer across both sites, 71 in Shelbyville and 85 in Carbondale. Of these deer, 45 in the Shelbyville sample were female and 26 male, and in Carbondale, 54 deer were female and 31 male. Deer were tracked with remotely-sensed GPS telemetry collars for periods of roughly one year on average, resulting in a total of 1,933,465 GPS locations. In chapter two, I used this GPS data to develop a method to relate resources to the relative probability of encounter based on a scale-integrated habitat selection framework. This framework integrates habitat selection estimates at multiple scales to obtain an appropriate estimate of availability for encounters. Using this approach, I related encounter probabilities to landscape resources and predicted the relative probability of encounter. Additionally, I further tested the usefulness of this approach by applying this framework to two other systems representing social contact and predator-prey contact respectively. This predicted distribution of encounters was more accurate when predicting novel encounters than a naïve approach or any individual scale alone. In chapter three, I improved estimates of the drivers of movement by developing novel methods for step selection analysis (SSA). To determine the impact of long-term behavior on local selection from SSA, I simulated movement trajectories including bias toward locations simulating different types of long-term behavior. Based on these simulated trajectories, I evaluated the impact of long-term behavior by identifying frequently reused locations based on a three-dimensional kernel density estimate including latitude, longitude, and time of day. Following this, I developed two approaches to account for spatial and temporal patterns of long-term behavior. I then compared estimates of known values of selection from models using these correction methods to previously established methods based on factors such as spatial memory. In chapter four, I applied this method to estimate local-scale step selection of deer in response to sources of risk. Additionally, I evaluated the impact of risk variables on behavioral states using hidden Markov models (HMMs) and determined state-specific estimates of selection. I found that deer avoided human modification but were more likely to change behavioral state in response to mesopredators. Since different sources of risk induce different behavioral responses, it is likely necessary to account for all of these behavioral responses when estimating the impacts of predation risk and its potential consequences. In chapter five, I used inferences from the preceding three chapters to build a mechanistic model of home range selection and movement that can be used to infer contact distributions. This approach could include varying levels of complexity including local-scale step selection, behavioral state transitions, and antipredator response. I ran models with varying levels of complexity and compared the performance of those models to the approach in chapter 2 for predicting contacts. I found that this method could predict contacts accurately even with limited data, but still had difficulty when transferring predictions to new locations.

Contemporary machine learning models exhibit unprecedented performance in the text, vision, and time-series domains, but at the cost of significant computational and human resources. Applying these technologies for science requires balancing accuracy and resource allocation, which I investigate here via three unique case studies. In Chapter 1, I present a deep learning system for animal pose estimation from video. Existing approaches rely on frame-by-frame supervised deep learning, which requires extensive manual labeling, fails to generalize to data far outside of its training set, and occasionally produces scientifically-critical errors that are hard to detect. The solution proposed here includes semi-supervised learning on unlabeled videos, video-centric network architectures, and a post-processing step that combines network ensembling and state-space modeling. These methods improve performance both with scarce and abundant labels, and are implemented in an easy-to-use software package and cloud application. In Chapter 2, I turn to the Gaussian process, a canonical nonparametric model, known for its poor scaling with dataset size. Existing methods accelerate Gaussian processes at the cost of modeling biases. I analyze two common techniques -- early truncated conjugate gradients and random Fourier features -- showing that they find hyperparameters that underfit and overfit the data, respectively. I then propose to eliminate these biases in exchange of increased variance, via randomized truncation estimators. In In Chapter 3, I investigate continual learning, or "finetuning", in large language models (LLMs) with billions of weights. Training these models requires more memory than typically available in academic clusters. Low-Rank Adaptation (LoRA) is a widely-used technique that saves memory by training only low rank perturbations to selected weight matrices in a so-called "base model'". I compare the performance of LoRA and full finetuning on two target domains, programming and mathematics, across different data regimes. I find that in most common settings, LoRA underperforms full finetuning, but it nevertheless exhibits a desirable form of regularization: it better maintains the base model's performance on tasks outside the target domain. I then propose best practices for finetuning with LoRA. In summary, applying state-of-the-art models to large scientific datasets necessitates taking computational shortcuts. This thesis highlights the implications of these shortcuts and emphasizes the need for careful empirical and theoretical investigation to find favorable trade-offs between accuracy and resource allocation.

No description available.

The purpose of this experiment was to compare the effects of combining negative reinforcement and positive reinforcement during teaching with the effects of using positive reinforcement alone. A behavior was trained under two stimulus conditions and procedures. One method involved presenting the cue ven and reinforcing successive approximations to the target behavior. The other method involved presenting the cue punir, physically prompting the target behavior by pulling the leash, and delivering a reinforcer. Three other behaviors were trained using the two cues contingent on their occurrence. The results suggest that stimuli associated with both a positive reinforcer and an aversive stimulus produce a different dynamic than a situation that uses positive reinforcement or punishment alone.