Abnormal ingestive behavior in male zoo gorillas: investigating the dietary and activity patterns associated with regurgitation and reingestion

Regurgitation and reingestion (R&R) is a widely observed, yet poorly understood, abnormal ingestive behavior among zoo-housed apes, including gorillas. While frequently characterized as a stereotypical behavior, emerging hypotheses suggest R&R may serve metabolic or self-regulatory functions in response to altered feeding patterns in captivity. In this study, we investigated which patterns of activity and food intake are associated with increased likelihood of R&R events in 21 male western lowland gorillas across five zoological institutions in the United States. Behavioral observations were recorded using focal and instantaneous sampling to generate activity budgets of R&R behavior. R&R was found to be negatively correlated with the time spent feeding and to occur more frequently in the afternoons. Predictive modeling identified high-risk behavioral sequences that may be associated with increased likelihood of R&R. R&R was frequently preceded by travel, feeding, and inactivity, suggesting a potential link to satiety signaling or boredom. Nutrient data were also analyzed to determine whether certain food items were disproportionately associated with the behavior. Commercial primate chow, fruit juice, cabbage, nuts, and carrots were among the foods most likely to be regurgitated and reingested. These results support the hypothesis that both behavioral context and dietary composition contribute to the expression of R&R in captive gorillas. This research may help identify opportunities for refining feeding strategies and enrichment protocols to mitigate the occurrence of R&R in managed care populations.

Captive environments including zoological facilities, laboratories, and sanctuaries can significantly alter the behavioral patterns of non-human animals, often leading to the emergence of behaviors commonly described as unnatural or abnormal as they are either not typically observed in the wild or are maladaptive (Jensen, 1986; Lutz et al., 2022; Mason and Latham, 2004; Pereira Bengoa and Manteca, 2025; Rushen and Mason, 2006). Animals may be attempting to regulate their affective, or emotional, states in response to these environments by engaging in avoidance behaviors or coping strategies, such as self-grooming, ear cupping/covering, and hair plucking, or by increasing inactivity (Goldsborough et al., 2022; Jacobson et al., 2016; Miller et al., 2020). Many of these behaviors, particularly when exaggerated or context-inappropriate, may be divorced from their original function or appear functionless altogether (Garner, 2006), leading to their classification as “stereotypical” (Mason and Latham, 2004; Swaisgood and Shepherdson, 2005; Rushen and Mason, 2006).

One such abnormal behavior is regurgitation and reingestion (R&R), where the animal self-induces regurgitation by bringing up recently ingested food out of the stomach into the mouth, hand, or onto the floor, and reingests the contents (Barrett et al., 2025; Gould and Bres, 1986; Hill, 2018; Lukas, 1999; Ruempler, 1992; Tennant et al., 2022). While R&R is rare in wild populations (e.g., proboscis monkeys: Matsuda et al., 2011; indris: Randrianarison et al., 2022), the behavior has been observed to be prevalent in several species in captivity, including all species of great ape (chimpanzees: Baker and Easley, 1996; orangutans: Cassella et al., 2012; bonobos: Miller and Tobey, 2012), dolphins (Yeater and Kuczaj, 2010), and California sea lions (Winship et al., 2024). This behavior is both repetitive and abnormal and has been widely observed in zoo populations of gorillas, with approximately 60-65% of the population in North America exhibiting the behavior (Tennant et al., 2022). R&R is likely a learned behavior that immature individuals master by observing conspecifics and practicing themselves (Gould and Bres, 1986; Lukas, 1999; Hill, 2018). Previous studies have found that sex is a predictor of R&R, with females more likely to exhibit the behavior, while age and rearing history were not reliable predictors (Tennant et al., 2022). Environmental features (e.g., enclosure design and climate) and management practices (e.g., feeding schedules, seasonal changes in staffing, etc.) have not yet been thoroughly investigated but may also influence R&R behavior (Hill, 2018).

Several studies have found a relationship between behavioral facets of the diet and R&R (Gould and Bres, 1986; Cabana et al., 2018; Lukas et al., 1999; Less et al., 2014; Ruempler, 1992). Where wild gorillas spend 40-50% of their time feeding and foraging (Harcourt and Stewart, 1984; Masi et al., 2009; Ostrofsky and Robbins, 2020; Watts, 1988), zoo-housed individuals dedicate a maximum of 20-25% of their time to feeding (Lukas, 1999). Food items in captivity are also nutrient-dense, low in fiber, and likely not as satiating as wild gorilla diets (Cabana et al., 2018; Hill, 2005; Less et al., 2014; Plowman, 2013; Smith et al., 2014). The continued act of eating brought on by repeatedly regurgitating and reingesting food may signal satiety in the body by prolonging total feeding time (Hollis, 2018). R&R in gorillas may thus represent a behavioral demonstration of disrupted satiation or prolonged feeding drive and may parallel metabolic abnormalities observed in humans (Less et al., 2014; Lukas, 1999).

Early investigations in R&R suggested that boredom might be a driver of the behavior and found that decreased feeding time initiated by processing food in a blender was associated with a shorter time to regurgitation after consumption, while increasing feeding time by providing browse reduced R&R overall (Gould and Bres, 1986). Larger adjustments to captive diets including the removal of milk (Lukas et al., 1999; Ruempler, 1992) eggs, meat, porridge, bread, and sugar (Ruempler, 1992) while simultaneously increasing browse and other roughage, or increasing the amount of specific nutrients such as fiber while decreasing others like sugar (Cabana et al., 2018), have led to reductions in R&R behavior. Later research built on the idea that feeding time may be driving R&R behavior, which prompted a continuous feeding hypothesis (Hill, 2005) positing that gorillas in zoos do not have the same continuous feeding opportunities that wild populations do and thus may regurgitate and reingest their food as a way of increasing their feeding time (Hill, 2005). The link to feeding time, specifically in diet processing time, has continued to be supported in recent scholarship (Barrett et al., 2025), but the complexity of and variation in the R&R behavior remain poorly understood (Cabana et al., 2018; Hill, 2018; Mason and Latham, 2004; Tennant et al., 2022).

Some scholars have proposed that R&R may serve a self-regulatory or a metabolic function, potentially in addition to extending feeding time (Cabana et al., 2018; Hill, 2018; Lukas, 1999; Mason and Latham, 2004). In human infants, the behavior is thought to increase feeding time or potentially be a response to stress, lack of stimulation, or neglect (Chatoor and Einhorn, 1984; Lukas, 1999). Similar rumination disorders have been shown to play a self-soothing role in humans but also can induce the hormonal feedback processes in the body that result in satiation (Lukas, 1999), especially the release of cholecystokinin and glucagon-like peptide 1(GLP-1) that slow gastric emptying and promote the feeling of fullness (Asarian and Bächler, 2014). If R&R is driven by multiple factors, investigating the dietary, environmental, and behavioral patterns associated with it may be key to understanding and mitigating it.

There has been little direct testing of the behaviors that precede R&R, as well as the complete nutrient composition represented in commonly regurgitated and reingested food items. Furthermore, due to the time requirements of behavioral data collection and the population distribution of gorillas across North America, sample sizes in R&R studies are often limited, or projects rely on surveying the secondary behavioral accounts from animal care staff (Hill, 2018; Tennant et al., 2022). Directly observing and analyzing the behavioral sequences preceding R&R may help identify the underlying motivational or physiological triggers that have been overlooked in previous diet-focused studies.

The main objective of this study is to provide a more thorough description of R&R in captive gorillas by using direct behavioral observations and nutritional intake data from a large sample size of individuals. We hypothesized that captive gorillas engage in abnormal ingestive behavior (i.e., R&R) to signal satiety and increase feeding time in response to energetically dense diets and altered activity patterns. Thus, we predict that R&R would be more likely to occur at times distant from the morning provisioning period (first and primary feeding event typically occurring between 7-9am) after they had experienced extended times of inactivity and that R&R frequency would be inversely correlated with total time spent feeding. Furthermore, because levels of inactivity are higher in gorillas diagnosed with heart disease than in healthy adults, we predict that gorillas diagnosed with heart disease would be more likely to exhibit R&R. We hypothesized that certain food items would be disproportionately associated with R&R due to their nutrient content, structure, and/or palatability (Cabana et al., 2018; Lukas et al., 1999), and predict that gorillas with a higher energetically dense diet would have a higher frequency of R&R. Lastly, we hypothesized that activity sequences preceding R&R would differ from randomly occurring behavior, and predict that R&R would increase in frequency during extended inactivity immediately after feeding cessation as a way to stimulate satiety response in the body. Thus, we expect that predictive models could be used to highlight sequences or conditions under which R&R risk is elevated and can be used to mitigate potential negative welfare implications.

Study site and subjects

Data were collected from May 2023 to January 2025 on 21 male western lowland gorillas housed in five zoological facilities in the United States that are accredited by the Association of Zoos and Aquariums (AZA). Males were selected for this study because data on these subjects were collected as part of a larger study focusing on heart disease across developmental stages in gorillas, and heart disease disproportionately affects males. Subjects were selected for this study without prior knowledge of their regurgitation and reingestion history. AZA-accredited zoos were chosen for this study because these facilities receive guidelines on nutrition and other husbandry aspects from the AZA Gorilla Species Survival Plan Committee and share a baseline of minimal standards for diet and space (AZA Gorilla Species Survival Plan, 2018). We included gorillas aged 5–40 years, excluding younger and older individuals to avoid potential confounds from juvenile or geriatric stages. Subjects were grouped into four categories for behavioral comparison: adolescents (5–9 years, n = 5), pubescents (12–15 years, n = 4), healthy adults (18–40 years, n = 7), and adults with cardiovascular disease (18–40 years, n = 5). Gorillas aged 16–17 were excluded from the study to ensure distinction between the pubescent and adult life stages.

Behavioral observations and activity budget analysis

RLS collected behavioral data using focal animal sampling (Altmann, 1974) from public exhibits and off-exhibit ‘bedroom’ areas with binoculars, a spotting scope, and digital recording devices. To estimate daily activity budgets, instantaneous sampling was used at 5-minute intervals during daytime hours (approximately 8–12 hours per day for each gorilla) for a total of 1,034 hours of observation. Each 5-minute scan was coded for one activity, including feeding/foraging, resting (above ground or terrestrial), social behavior, abnormal repetitive behavior, and travel (above ground or terrestrial), using behavioral definitions adapted from Ross et al. (2011); Less et al. (2012), and Chelluri et al. (2013). The proportion of time spent in each activity was calculated as the number of scans per activity divided by the total number of scans where the focal was visible, resulting in a proportion of observed time for each behavior (Altmann, 1974; Martin and Bateson, 2007).

To examine how activity budgets varied among male western lowland gorillas of different life stages and health statuses, a generalized linear mixed model (GLMM) was applied. Gorilla subjects were categorized into four groups: adolescents (5–9 years), pubescents (12–15 years), healthy adults (18–40 years), and adults diagnosed with cardiovascular disease (18–40 years). The GLMM was developed using the glmer function in R version 4.1 (R Core Team, 2021), with fixed effects for gorilla group, activity, and their interactions. Random intercepts for each individual and zoological institution were included to account for repeated measures. A Type III ANOVA with Satterthwaite’s approximation for degrees of freedom was then performed to assess significance. Post hoc pairwise comparisons were conducted using the emmeans package in R to calculate estimated marginal means and Tukey-adjusted pairwise contrasts (R Core Team, 2021).

To evaluate whether demographic factors predicted the likelihood of R&R behavior (0 = no R&R, 1 = R&R) during each 5-minute scan, we fit a GLMM using the glmer function, with a binomial error distribution and a logit link function (R Core Team, 2021). The model included the demographic category of gorillas (i.e., adolescent, pubescent, healthy adult, or adult with heart disease) as a fixed effect, and, to account for repeated measures, gorilla identity as a random effect. The statistical significance of model predictors was evaluated using Wald z-tests, and model output was converted back to the target scale for interpretation.

Regurgitation and reingestion analysis

All occurrences of R&R were recorded continuously (Altmann, 1974). Data included food consumed prior to R&R events, total time spent engaging in R&R, and whether events were cyclical (repetitive and uninterrupted or isolated).

To assess whether R&R behavior differed across developmental stages, potentially indicating a behavior learned as a subadult that becomes entrenched in adulthood, all adult gorillas were combined into one category and compared to adolescents and pubescents. This allowed us to isolate age-related patterns from those potentially influenced by health status. We then fit a GLMM with a binomial error distribution and logit link to model the probability of R&R occurrence as a function of age, including individual identity as a random effect to account for repeated observations. To evaluate differences across all three age groups, we used the emmeans package in R to compute estimated marginal means and conducted pairwise comparisons using Tukey-adjusted p-values and predicted marginal probabilities (R Core Team, 2021).

To examine whether R&R may be related to satiety, we ran a GLMM with a binomial error distribution and a logit link function to evaluate whether the duration of daily feeding activity influenced the likelihood of R&R. The model predicted the proportion of R&R observations from the proportion of time spent feeding on a given day, with individual gorillas included as a random effect to account for repeated measures. We also ran a second GLMM to assess whether feeding bout length influenced the likelihood of R&R events. This model included feeding bout length (minutes, scaled) and time of day (categorical: early morning, late morning, early afternoon, late afternoon) as fixed effects, with the response variable defined as whether an R&R event occurred within 10 minutes after the end of each feeding bout (0 = no event, 1 = event). The individual gorilla was included as a random intercept to account for repeated observations within individuals.

To examine temporal patterns in R&R behavior, observational data were categorized into four time-of-day intervals: 6:00–9:00 AM, 9:01 AM–12:00 PM, 12:01–3:00 PM, and 3:01–6:00 PM. A GLMM with a binomial distribution and a logit link was used to model the probability of R&R as a function of time of day. The model included time-of-day category as a fixed effect, with zoological facility, date, and individual subjects as random effects. Predicted probabilities and confidence intervals were extracted and visualized to assess temporal differences in behavior.

We fitted GLMMs with a Poisson error distribution and log link function to assess whether certain foods were disproportionately associated with R&R. First, we modeled R&R at the level of individual food items, with food item included as a fixed effect and gorilla individual as a random intercept to account for repeated measures. We then combined items into broader food categories (e.g., fruit, chow/biscuits, root vegetables, browse) and fit a parallel model with category as the fixed effect and individual gorilla as a random intercept. These models provided incidence rate ratios (IRRs) with 95% confidence intervals, estimating the relative likelihood that a given food or food category was associated with R&R compared to a reference category.

Nutrient values for food items cultivated for human consumption and observed to be reingested were sourced from USDA FoodData Central [United States Department of Agriculture (USDA), 2025]. Where published nutrient values were unavailable, samples of food items were gathered and dehydrated for 12 hours at 40 °C before being shipped for proximate nutritional analysis, which included measures of crude protein, gross energy, neutral detergent fiber, and mineral content for each sample. All laboratory analyses were conducted by the University of Georgia Agricultural and Environmental Services Laboratories in accordance with their established standard operating procedures. After samples were received, they were dried at 105 °C for at least 2 hours and then, if needed for analysis, were ground through a 1mm screen. Total nitrogen was determined using the Dumas combustion method on an Elementar Rapid N Cube analyzer, in which samples were combusted at 960 °C and nitrogen quantified via thermal conductivity detection, with crude protein calculated using the appropriate conversion factor (Elementar Americas, 2012). Nitrate concentrations were quantified using a rapid colorimetric salicylic acid method in plant extracts, where the nitro-salicylic acid complex was measured using a spectrophotometer at 410 nm (Cataldo et al., 1975). Neutral detergent fiber (NDF) was analyzed using the ANKOM2000 Fiber Analyzer according to the ANKOM method, in which samples sealed in filter bags were digested in a neutral detergent solution containing sodium lauryl sulfate, EDTA, sodium sulfite, and heat-stable α-amylase (Ankom Technology, 2025). Mineral concentrations were determined following dry-ash digestion: samples were ashed at 550 °C, extracted with 3 N HCl, and analyzed by inductively coupled plasma (ICP) spectroscopy (AOAC International, 1996).

To compare the nutritional composition of foods that were and were not associated with R&R, individual nutrient differences were evaluated using permutation-based mean difference tests (9,999 permutations per comparison) with Benjamini–Hochberg correction for multiple testing in R. Effect sizes were estimated using Cliff’s delta to assess the magnitude and direction of group differences. Overall nutrient profiles were compared using a permutation-based multivariate analysis of variance (PERMANOVA-style test) on scaled nutrient data (9,999 permutations).

Behavioral scans immediately preceding R&R were analyzed to identify patterns and potential triggers for the behavior at the event (single occurrence), bout (all occurrences within a certain timeframe), and episode (all R&R events after one ingestion act and before another occurs) levels (Table 1). These categories describe different analytical scales and are not mutually exclusive; for example, multiple R&R events may constitute a bout while simultaneously occurring within a single feeding-associated episode. R&R events were defined as any occurrence of R&R regardless of temporal, cyclical, or feeding context. To control for R&R events that are likely not independent of each other, the inter-frequency period of R&R events was determined and used to define a “bout” of R&R as any R&R events that happened within the same 20-minute period. To our knowledge, an R&R bout has not been previously defined in the literature, so the robustness of our 20-minute definition was tested by determining and comparing the behavioral sequence permutation test results for various bout time length definitions (5, 15, 20, 30, 40, and 60 minutes). An episode of R&R was defined as the combination of all R&R that occurred within 30 minutes after a specific ingestion event (either feeding or reingestion of regurgitant). This time window was chosen based on the gastric lag phase for solid food in humans, which ranges from 25 to 35 minutes (Hellström et al., 2006; Ziessman et al., 1996). The lag phase time window was selected because this process precedes the linear emptying phase in digestion (Hellström et al., 2006; Ziessman et al., 1996), during which food items can still be easily regurgitated from the stomach. Human values were chosen because, although the average gut passage rate is known for western gorillas (~50 hours; see Remis, 2000), rates for specific processes, such as gastric emptying, are not yet available.

To identify behavior chains associated with elevated R&R risk, sequences of one, two, and three behavioral intervals (5–15 minutes total) preceding each R&R event were extracted and compared to randomly selected sequences not followed by R&R. A Markov-style transition matrix was constructed to quantify the frequency of each sequence in R&R vs. non-R&R contexts (Martin and Bateson, 2007). A chi-square test was used to evaluate global differences in sequence distribution. To determine which sequences were disproportionately associated with R&R, pairwise Fisher’s exact tests were conducted using simulated p-values with 10,000 permutations and Bonferroni-adjusted alpha thresholds.

We also constructed a decision tree model using logistic regression to explore predictive patterns in behavioral transitions. The model used sequences of up to three behaviors prior to each observation as predictors of whether R&R would occur. Observed individual behaviors were sorted into behavioral categories which included Feeding (active foraging or ingesting of food), Inactivity (resting or not locomoting), Locomotion (travel by an individual including walking, climbing, and running both terrestrially and above ground), Abnormal (any behavior not part of the species’ behavioral repertoire in the wild or one divorced from their original function or that appear to have no specific function; includes R&R), and Social interactions (active affiliative and/or agonistic interactions). The decision tree identified high-risk transitions and cyclical behavior patterns. These results were used to visually summarize predictive sequences and flag behavior chains associated with elevated R&R risk.

R&R behavior in male zoo gorillas

Of the 21 gorillas observed, ~48% of the population (n = 10) exhibited the regurgitation and reingestion behavior at least once during this study (see Appendix A). Of the population observed to exhibit the behavior, adolescent males, pubescent males, and healthy adults each represented 20% of the population (n = 2 gorillas in each category), while males with heart disease represented 40% (n = 4). Analysis of individual variation in R&R behavior revealed substantial differences across gorillas. When the proportion of R&R events was calculated relative to total observations per individual, certain test categories of gorillas emerged as displaying disproportionately high rates of R&R (Figure 1). On average, adolescents spent 3 minutes engaged in R&R (SD = 6.7), while pubescents averaged 8.75 minutes (SD = 11.8). Healthy adults averaged 29.3 minutes (SD = 77.5), and gorillas diagnosed with heart disease showed the highest mean R&R time at 80 minutes (SD = 109.0). Median values reveal a similar pattern, with adolescents and healthy adults both having a median of less than one minute, pubescents at 5 minutes, and adults with heart disease at 15 minutes. The high standard deviations and interquartile ranges reflect substantial individual variability within each group.

Analysis showed no significant main effect of gorilla demographic group on R&R activity (F(3, 98) = 0.16, p = 0.921). GLMM-predicted probabilities showed no significant difference in R&R between healthy adults (≈15%) and gorillas with heart disease (≈7%); wide, overlapping confidence intervals (e.g., CI for heart disease vs. pubescents: −0.54 to 0.98) indicate high individual variability and no clear effect of health status.

Time of day

We found a significant effect of time of day on the likelihood of R&R behavior (χ²(3) = 33.67, p <.001). Predicted probabilities remained low across the day but were lowest during the early morning (6–9am: 0.001%) and late morning (9am-12pm: 0.002%), increased in the early afternoon (12–3pm: 0.003%), and peaked in the late afternoon (3-6pm: 0.007%). This pattern may suggest a temporal escalation in R&R occurrences throughout the day. The effect of time remained significant after controlling for individual variation, indicating a robust relationship between time of day and R&R probability. Moreover, there was a significant negative relationship between the proportion of time spent feeding and the likelihood of R&R behavior (Estimate = -7.493, SE = 1.718, z = -4.36, p < 0.001) (Figure 2). Gorillas that spent more time feeding during a given day were significantly less likely to engage in R&R.

No significant relationship was found between feeding bout length and the probability of subsequent R&R events (Estimate = 0.25, SE = 0.17, z = 1.45, p = 0.146). Although the trend suggested that longer feeding bouts were associated with a slightly higher likelihood of R&R, the effect was not statistically significant. This indicates that while daily-level analyses showed that individuals spending more time feeding engaged in less R&R overall, the length of individual feeding bouts did not predict whether R&R occurred immediately afterward. At the individual level, mean feeding bout length was positively associated with R&R frequency (β = 10.96, SE = 3.74, p = 0.033, R² = 0.63), indicating that gorillas with longer average feeding bouts engaged in more R&R overall.

Nutritional patterns of R&R behavior

Primate biscuits were the most common food item regurgitated for 30% of the gorillas during the total 7-day observation window per gorilla (Table 2; Appendix A). Juice was the second most common amongst subjects (20%) (Table 2; Appendix A). When analyzing individual food items, we found that chow had the strongest effect (Estimate = 3.35, SE = 1.10, z = 3.03, p = 0.002), indicating it was disproportionately associated with R&R events. Grapes (Estimate = 3.36, SE = 1.11, z = 3.04, p = 0.002), cabbage (Estimate = 2.98, SE = 1.08, z = 2.76, p = 0.006), juice (Estimate = 2.93, SE = 1.12, z = 2.62, p = 0.009), mixed nuts (Estimate = 2.56, SE = 1.07, z = 2.39, p = 0.017), and carrots (Estimate = 2.04, SE = 1.03, z = 1.98, p = 0.048) were also significant predictors. Among browse items, only Sycamore (Platanus occidentalis) was significant (Estimate = 3.23, SE = 1.37, z = 2.36, p = 0.018).

Other items such as spinach (Estimate = 1.95, SE = 1.09, z = 1.79, p = 0.074), iceberg lettuce (Estimate = 1.99, SE = 1.17, z = 1.70, p = 0.089) and romaine (Estimate = 1.59, SE = 1.08, z = 1.47, p = 0.141) showed elevated effects but did not achieve statistical significance in this model. Foods that were not associated with increased R&R risk included banana (p = 0.733), broccoli (p = 0.628), kale (p = 0.608), and popcorn (p = 0.915), suggesting their role in R&R is either minimal and/or context dependent.

When examining the proximate nutrient profiles of foods most associated with R&R (Table 2), we found that they fell at opposite ends of both nutritional and structural ranges. On one end, chow and juice were notable for their high palatability and nutrient density in different forms: chow was exceptionally rich in protein (21.17% DM) and sodium (2527 mg/g DM), while juice was almost entirely devoid of fiber and protein but provided a readily available sugar/energy source. On the other hand, cabbage, carrots, and Sycamore browse were notably fibrous (NDF = 26.7%, 23.9%, and 46.1%, respectively), with cabbage and carrots also being mineral-rich, exhibiting very high potassium levels (2556 and 2735 mg/g DM, respectively). Grapes offered high available carbohydrate content with moderate mineral levels, while mixed nuts provided the lowest available carbohydrate content (22.90% DM) alongside substantial protein (19.9%) and magnesium (232 mg/g DM). When comparing the nutrient content of foods that were commonly regurgitated (n=7) with those that were not observed to be regurgitated by any gorillas during this study (n=10), a permutation MANOVA revealed significant overall differences (F = 4.16, p = 0.004). Univariate permutation tests with Benjamini–Hochberg correction indicated that R&R-associated foods contained higher phosphorus, potassium, calcium, and magnesium concentrations (q ≈ 0.03 for each) and tended to have lower neutral detergent fiber content, although not statistically so (q = 0.054). Effect-size estimates (Cliff’s δ > 0.3 for minerals, δ ≈ –0.57 for NDF) confirmed that these differences were moderate to strong in magnitude.

When we analyzed food items sorted by category, browse was strongly underrepresented in R&R sequences (estimate = −3.49 ± 1.00 SE, z = −3.47, p < 0.001; IRR = 0.03, 95% CI: 0.004–0.219), indicating gorillas were ~97% less likely to regurgitate browse than the baseline category (Figure 3). Several food categories were disproportionately associated with R&R, including chow/biscuits (estimate = 1.21 ± 0.21 SE, z = 5.74, p < 0.001; IRR = 3.34, 95% CI: 2.22–5.05), where R&R was over three times more likely, and root vegetables (estimate = 1.82 ± 0.14 SE, z = 13.4, p < 0.001; IRR = 6.19, 95% CI: 4.74–8.09), which were over six times more likely than the baseline. Enrichment food items also predicted R&R (estimate = 0.60 ± 0.13 SE, z = 4.51, p < 0.001; IRR = 1.82, 95% CI: 1.40–2.37), nearly doubling the risk (Figure 3). Fruit (estimate = 0.44 ± 0.17 SE, z = 2.69, p = 0.007; IRR = 1.56, 95% CI: 1.13–2.15) increased the likelihood by ~56%, and nuts/seeds (estimate = 0.67 ± 0.25 SE, z = 2.72, p = 0.006; IRR = 1.95, 95% CI: 1.21–3.16) nearly doubled it (Figure 3).

Activity patterns associated with R&R behavior

Assessing the behavioral sequences preceding R&R in gorillas revealed that the frequency of R&R events occurred predominantly after periods of at least 5 minutes of inactivity, especially if that period of inactivity had been preceded by rest or feeding (Figure 3, 4). The most common 15-minute sequence of behaviors preceding R&R was Rest → Rest → Feed/Forage (Figure 5).

A chi-square test revealed a statistically significant difference in the frequency of behavioral sequences preceding R&R events versus those that did not lead to R&R (χ² = 201.48, df = 16, p < 0.001). Figure 5 shows the top three behavior sequences observed before R&R events. Pairwise Fisher’s exact tests with simulated p-values (10,000 permutations, Bonferroni corrected) further identified sequences such as Rest → Rest → Feed/Forage as significantly overrepresented (simulated p = 0.006), indicating that these chains are not randomly distributed and are significantly overrepresented prior to R&R onset.

Across all individuals, R&R bouts ranged from a single isolated event to extended sequences of up to 21 events. Most bouts were relatively small (median = 1.5 events; 70% of bouts contained four or fewer events), although the average was higher (mean = 3.45 events) due to occasional large bouts. Approximately 10% of bouts contained eight or more events, indicating that while isolated R&R events were most common, some gorillas engaged in extended periods of repeated R&R. Permutation tests revealed that bouts were significantly less likely to follow extended resting sequences (Rest → Rest → Rest, observed = 6, expected = 14.1, 95% CI [8, 20], p =.002). In contrast, repetitive feeding sequences (Feed/forage → Feed/forage → Feed/forage, observed = 4, expected = 4.06, 95% CI [1, 8], p =.62) and short travel–feeding transitions (Travel ground → Feed/forage → Feed/forage, observed = 2, expected = 1.16, 95% CI [1, 2], p =.14) occurred at chance levels. These results held even when variations in the time associated with R&R bouts were adjusted (Appendix B), suggesting that, regardless of how about is defined in terms of time, the difference between single-event level and clustered R&R events is robust.

When R&R behavior was considered at the episode level (all occurrences related to the same ingestion act) rather than the event (single occurrence) or bout (all occurrences within a certain timeframe) level, only feeding was observed in the 30 minutes preceding R&R episodes. Of the 4,051 30-minute feeding sequences observed, only 115 (2.84%) led to an episode of R&R. None of the 107 sequences that included rest, social behavior, or locomotion were associated with an R&R episode onset.

Predictive modeling for R&R events

When identifying which behaviors may predict R&R events, the logistic model indicated sequences with a >50% predicted probability of R&R as high-risk. Common high-risk sequences included ‘Feeding → Inactivity → Inactivity’ and ‘Locomotion → Feeding → Inactivity’. The classification tree (Figure 6) further illustrated that behavioral sequences ending in ‘Inactivity’, ‘Abnormal’, ‘Feeding’, or ‘Locomotion’ were associated with an increased predicted probability of R&R events, while those ending in ‘Social’ or ‘Other’ were least likely to result in R&R. Several high-risk paths included prolonged inactivity, indicating that R&R was more frequently associated with low arousal contexts. Several branches leading to R&R included repeated or escalating states of inactivity (e.g., ‘Inactivity → Inactivity → Abnormal’), suggesting that idleness may represent a context in which R&R can be prompted. Sequences with earlier signs of ‘Feeding’ or ‘Locomotion’ transitioning into ‘Inactivity’ also appeared in high-risk paths, potentially indicating motivational transitions (e.g., from goal-directed behavior to idle or unengaged states). Additionally, R&R itself emerged in cycles as a predictor of future occurrences, suggesting that R&R may be an episodic behavior and potentially an abnormal repetitive behavior in some individuals.

This study revealed that R&R behavior among zoo-housed western lowland gorillas was significantly influenced by time of day, time spent feeding, individual variability, and food category. Rates were lowest in the morning, highest in late afternoon, and more likely on days with reduced feeding. Descriptively, adults with heart disease and pubescents showed higher R&R rates than healthy adults or adolescents, although demographic group was not a significant predictor of R&R and substantial individual variation was observed.

Although individuals who spent more time feeding overall engaged in less R&R, those who structured feeding into longer, continuous bouts showed higher frequencies. Longer consolidated bouts, especially when followed by long gaps until the next feeding, were associated with clustered R&R, whereas shorter, fragmented bouts were less likely to do so. This suggests that both the number of feeding opportunities and the structure of food items may be important factors for R&R, with implications for management. Interrupting the day with more frequent, smaller feeding opportunities, ideally accompanied with browse, may reduce the formation of prolonged feeding bouts followed by prolonged inactivity and, in turn, could therefore decrease the likelihood of clustered R&R. Such a strategy would align with naturalistic foraging patterns, in which gorillas encounter food items throughout their day (Fuh et al., 2022; Masi et al., 2009; Ostrofsky and Robbins, 2020) and could represent a practical approach to mitigating this abnormal behavior in captivity.

Health and demographic patterns

The results of this study demonstrate that although observable differences exist between demographic groups regarding R&R, this distinction alone is insufficient to explain the patterns of R&R. Adults with heart disease and pubescents appeared to show the highest levels of R&R, but there was considerable variation between individuals. One possible explanation for the elevated R&R observed among pubescents and adults with heart disease is variation in energetic demands. In contrast, healthy adults exhibited less R&R, although whether this difference relates to energetic balance, nutritional intake, or other demographic or individual factors cannot be determined from the present study. Pubescents demonstrated relatively high rates of R&R behavior despite none of these subjects being diagnosed with heart disease at the time of this study. One hypothesis consistent with this pattern is that R&R may be associated with metabolic or physiological factors, potentially including mechanisms related to satiation or feeding duration. One possibility is that R&R could be associated with physiological processes related to satiation or feeding drive, particularly during periods of altered energetic demand (Pontzer et al., 2021; Rothman et al., 2007; Speakman, 1997). Engaging in R&R may extend the behavioral experience of feeding in pubescents, which they could use to meet energetic demands in wild contexts, but whether this has any physiological consequences related to satiation or energetic state remains unknown. The diet provided in zoos is largely developed and adjusted based on absolute body weight and body condition score, but body weight and body condition alone may not capture variation in energetic and micronutrient requirements associated with growth and maturation. Since other studies found that female gorillas are ~1.5 times more likely to exhibit R&R (Tennant et al., 2022), if diets are based solely on body mass and composition, without taking into account changes in metabolic needs throughout the ovulation cycle, not just during pregnancy and lactation, then variation in energetic requirements across reproductive states represents another potential factor warranting investigation in relation to R&R (Arhin et al., 2018; Barr et al., 1995; Bisdee et al., 1989). Because R&R has not been reported in wild gorillas experiencing similar metabolic changes, the behavior in captivity may reflect environmental constraints on food intake. One possibility warranting longitudinal investigation is that physiological or feeding-related factors contribute to the initial development of R&R, while reinforcement or learning may contribute to its persistence later in life.

Nutritional links and food associations

Previous studies have suggested that fiber may be a key factor in the triggering of R&R due to an observed decrease in the behavior when more fibrous items such as browse are offered in the provided diet (Barrett et al., 2025; Cabana et al., 2018). It was predicted that food items with higher energy-to-fiber content would be the most commonly regurgitated and reingested items. While most of the food items commonly reingested tended to have a relatively high available carbohydrate level, which would be a source of accessible energy, the relationship of fiber to R&R appears to be more complex. The fiber content of reingested food items varied substantially, with some high-fiber items, such as chow and Sycamore browse, ranking among the most likely to be regurgitated. While our hypothesis was not supported in this study, fiber has a well-established role in stimulating satiety (Farooqui, 2015). Thus, the influence of increasing dietary fiber on R&R may vary among individuals and may depend on the mechanisms underlying the behavior.

However, fiber is still known to play an important role in the health of the cardiovascular systems including helping to reduce blood cholesterol levels (Farooqui, 2015), lower high and maintain healthy blood pressure (Appel et al., 1997; Farooqui, 2015), both of which reduce the risk of cardiovascular disease (Appel et al., 1997; He et al., 2022; Farooqui, 2015; Samson and Garber, 2014). Additionally, dietary fiber is needed to effectively move waste through the gastrointestinal system, to support healthy gut microbiota, to maintain the mucosal lining of the intestinal tract, and can lower the risk of and treat active ulcerative colitis (Jha et al., 2019), one of the most common causes of morbidity and mortality in captive apes (Strong et al., 2016).

Zoo-housed gorillas are consuming notably less fiber than free-ranging populations (Cabana et al., 2018; Less et al., 2014; Remis et al., 2001; Smith et al., 2014). The term fiber is used to generally refer to both soluble and insoluble plant carbohydrates that are nondigestible or slowly digestible (National Research Council, 2003). Neutral Detergent Fiber (NDF) is comprised of hemicellulose, lignin, and cellulose, while Acid Detergent Fiber (ADF) is a measure of lignin and cellulose (National Research Council, 2003). These are important distinctions because each measure provides information about the digestibility of the food item and how much protein and energy may be available from it (National Research Council, 2003). Wild gorilla food items can seasonally contain up to 67% NDF and 50% ADF (Calvert, 1985; Remis, 1997; Remis et al., 2001), meaning that the foods they consume are composed of upwards of 50% indigestible fiber. The fiber recommendation for zoo-housed gorillas is 10-30% NDF and 5-15% ADF (AZA Gorilla Species Survival Plan, 2018; National Research Council, 2003; Remis et al., 2001) which is substantially lower than wild food ranges (Lodwick and Salmi, 2019; Masi et al., 2009; Remis et al., 2001). While it is recommended that future diet developments for zoo gorillas increase the recommended fiber level to support cardiovascular and gastrointestinal health benefits, simply increasing fiber content may not serve as a prescriptive intervention for R&R specifically.

Foods commonly regurgitated and reingested had higher concentrations of phosphorus, potassium, calcium, and magnesium, but it is unclear how these may impact palatability of the regurgitant and thus influence reingestion of the food item. Many of the higher-fiber foods on the list of most regurgitated foods also had relatively high sodium content, which may influence palatability. These results suggest that while fiber may be a protective factor, its influence on R&R may depend on individual satiety pathways and/or predispositions. Foods associated with R&R in this study tended to be nutrient-dense, with readily digestible energy and/or relatively high mineral content. Furthermore, considerations such as sodium content, sugar content, palatability, novelty, and texture should also be taken into account when selecting foods to add or omit from a diet designed to influence R&R in individual gorillas. Provisioning decisions should emphasize items such as browse that are reingested disproportionately to the rate at which they are offered and have been observed to reduce R&R behavior in other species (Struck et al., 2007). However, the nutrient balance of those items should mirror wild diets as much as possible which are naturally sodium-poor (Grueter et al., 2018; Rothman et al., 2006), so offering lower-sodium browse, such as pear (Pyrus communis), American beech (Fagus grandifolia), or hackberry (Celtis spp.) may provide texture and processing benefits without supplying excess sodium. The balance of nutrients should always be considered, as this balance influences physiological responses to food and may also contribute to individual food preferences. Caretakers often assume fruit is the main driver of R&R due to its sugar content, texture, and palatability (Less et al., 2014; Plowman, 2013). Reduced-fruit diets sometimes lower R&R (Barrett et al., 2025; Tennant et al., 2022) but rarely eliminate it (Less et al., 2014), suggesting that fruit alone is not the driver of R&R. Typically, reduced fruit diets also contain higher fiber foods that require longer processing time as easily consumable fruits are replaced with vegetable and browse in the provided diet (Barrett et al., 2025; Hill, 2018). This makes it difficult to tease apart whether the reduction in R&R is due to a reduction of fruit or due to the addition of fibrous foods that increase feeding and foraging time through processing (Hill, 2018).

Individual variation in R&R behavior is noteworthy, especially concerning which food items are commonly regurgitated. For example, the two individuals who demonstrated the highest R&R frequencies in this study did so in association with different foods. One individual regurgitated and reingested disproportionately more after consuming cabbage, and another did so after consuming oatmeal, which was not among the top foods reingested overall. This indicates that although food items may be sorted into food categories such as greens, fruit, browse, etc., for the purpose of diet development and enrichment, categorizing foods with greater specificity is recommended as the individual food items may provide more informative indicators of associations between consumption and subsequent R&R. These findings challenge anecdotal reports and common assumptions among caretakers that fruit is the food category most strongly associated with R&R behavior and supports the development of diets based not only on the nutrient content of specific foods offered but also on their palatability to the individuals who are receiving them. It may also be advisable to pair highly palatable and nutrient-dense foods like fruit and chow with food items like browse that have higher processing times to ensure variation in food categories ingested during each feeding. In other words, nutrient content, food texture, processing time, and individual predisposition should be considered in the development of diets to address the role that nutrition plays in R&R.

Behavioral predictors and sequences preceding R&R

This study is unique in that we analyzed behavioral patterns directly preceding regurgitation and reingestion, revealing distinct temporal scales indicative of a consistent dichotomy in the behavior. At the event level, R&R was most likely to follow periods of inactivity or rest, suggesting an association between isolated instances and low-arousal behavioral contexts. This pattern is consistent with previous hypotheses linking R&R to boredom or unmet motivation to forage, which some previous scholars proposed as the main driver of the behavior (Barrett et al., 2025; Cabana et al., 2018; Hill, 2005, 2018; Lukas, 1999; Mason and Latham, 2004). At the bout level, clustered R&R was closely tied to feeding-related sequences. At the episode level, which was defined relative to an ingestion event, only a small proportion of feeding sequences ultimately resulted in R&R, indicating that recent ingestion alone was insufficient to explain its occurrence. This pattern suggests that temporally clustered R&R may represent a feeding-associated mode of the behavior. One possible explanation is that repetitive R&R extends feeding-related activity and may interact with physiological processes involved in satiation, as suggested by previous work (Cabana et al., 2018; Hill, 2005, 2018; Lukas, 1999). Together, these findings indicate that R&R in zoo-housed gorillas may manifest in at least two behaviorally distinct modes: a feeding-related, clustered form and an isolated, stimulation-seeking form (Figure 7). It may also be possible for a single individual to exhibit both forms when exposed to different environmental contexts, which further underscores the complexity of this abnormal behavior (Figure 7).

Combined, these findings highlight that R&R is not a uniform behavior but may occur in distinct forms depending on the individual and context. While further work is needed to clarify underlying physiological and motivational drivers, our results suggest that R&R occurs within multiple behavioral contexts in gorillas, with episodes closely associated with feeding, isolated events associated with low-arousal behavioral states, and highly repetitive cycles potentially representing abnormal repetitive behavior or stereotypies in some individuals. These patterns raise the possibility that R&R may have different underlying motivations or functions among individuals and contexts, although these cannot be established from behavioral associations alone. Recognizing both feeding-driven and stimulation-driven R&R patterns may allow for more targeted, individualized interventions through adjustments to nutrition, behavior, or the environment. The insights gleaned from gorillas may also inform our understanding of the behavior in other species and help better curate interventions to curb this behavior for all captive populations.

Using predictive modeling, we found that the strongest predictors of R&R events were behavior chains characterized by repeated rest (e.g., potentially boredom-like states) and the expression of other abnormal behaviors (e.g., hair plucking, ear cupping/holding). These results demonstrate the potential predictive utility of behavior sequencing and suggest that behavioral sequencing may help identify broader behavioral contexts in which R&R is more likely to occur. This highlights the potential utility of structured behavioral monitoring in identifying “risk windows” for intervention and guiding proactive intervention, such as enrichment or social opportunities, during periods when R&R is likely to emerge. However, given the temporal resolution of the present sampling method, individual variation in behavioral sequences, and practical constraints on continuous monitoring, these patterns should not be interpreted as validated early-warning indicators of individual R&R events. Furthermore, this technique can be useful for general management considerations, but it does not account for individual variations in behavior and should only be used as a guide, rather than a prescription, for intervention.

This study provides a comprehensive analysis of R&R using direct behavioral observation in zoo-housed gorillas, but it is not without notable limitations. The sample size of 21 gorillas, albeit relatively larger than in previous studies, still consisted only of males and contained four to seven individuals in each test category. Although Institution was included in the statistical models, more refined differences in enclosure design, climate, social grouping, and feeding schedules could potentially impact R&R behavior or dietary intake results. Furthermore, the use of 5-minute interval sampling to record activity may not be effective in capturing shorter-duration behaviors or rapid transitions in behavior that are relevant to understanding the behavior sequences leading to R&R. While regurgitation was interpreted in relation to satiety and feeding patterns, physiological indicators of hunger, satisfaction, or reward were not assessed. Because this study was observational, the identified associations cannot establish causal relationships between feeding, activity, nutritional characteristics, and R&R. Lastly, this study only analyzed current R&R behavior and health status, not longitudinal behavioral, social, or health data that may substantially influence the present patterns of R&R in individuals.

Although not the primary focus of the present study, previous research has linked nutrition and activity patterns in zoo-housed gorillas to broader metabolic health issues, including obesity, leptin resistance, and insulin resistance (Less et al., 2014; Strong et al., 2018). Captive gorillas share many risk factors with humans diagnosed with metabolic syndrome, including imbalances in energy intake and expenditure, as well as diets high in energy and low in fiber. These physiological effects have been implicated in the development of fibrosing cardiomyopathy, the most common form of heart disease in male zoo gorillas (Meehan and Lowenstine, 1994; Lowenstine et al., 2016). The potential relationship between R&R and metabolic state, therefore, should be considered within this broader physiological context, and further research is needed to determine if there is a link between health state and R&R behavior.

Overall, this study demonstrated that R&R is a complex behavior that necessitates consideration of event-, bout-, and episode-level patterns, as well as individual-specific traits, when developing management and welfare strategies to mitigate it. Both behavioral monitoring and dietary strategies can potentially be used to mitigate R&R behavior; however, these factors should be considered on an individual basis rather than at the demographic level. Future studies should survey this behavior more deeply during the developmental stages of both males and females to better understand how the presence of R&R in conspecifics, especially close family members, may impact the prevalence of the behavior later in life.