Wildlife physiological responses to environmental and human-related stressors provide useful clues on animal welfare. Non-invasive biomarkers, such as fecal glucocorticoid metabolites (fGCM), allow researchers to assess whether variations in habitat quality, behavior, and climate influence the animals' physiological stress. We examined the role of fragment size, ambient temperature, ripe fruit availability and consumption, percentage of records moving, sex, female reproductive state, and group composition as predictors of the level of fGCM in adult brown howler monkeys (Alouatta guariba clamitans) inhabiting three small (<10 ha) and three large (>90 ha) Atlantic Forest fragments in southern Brazil. We collected bimonthly behavioral data and fecal samples from adult individuals over three years, and used a multimodel inference framework to identify the main predictors of fGCM. We found that the mean (±SD) fGCM in the study groups ranged from 57 ± 49 ng/g to 93 ± 58 ng/g, which were within the known range for howler monkeys. We found 10 best models including five of the 17 tested variables. Sex and reproductive state were the only variables included in all these models. We found that fGCM was higher in nursing females (mean ± SD = 104 ± 73 ng/g) than in non-nursing females (64 ± 55 ng/g) and males (53 ± 40 ng/g, P < 0.05) and that it decreased with increasing ripe fruit consumption and minimum temperature. However, fragment size did not predict fGCM concentration (groups in small fragments = 71 ± 58 ng/g vs. groups in large fragments = 63 ± 54 ng/g, P > 0.05). We conclude that factors related to the energetic balance of individuals play major roles in modulating the physiological stress of brown howler monkeys. Future studies should investigate the consequences of higher levels of stress hormones on howler monkey health and demography.