Nesting success has been a surrogate to assess environmental changes that affect the reproductive success of birds, and to analyze changes that influence natality. We analyzed the nesting success of Turdus grayi at "El Zapotal" Recreational Ecological Preserve, Central Chiapas. During the 2015 breeding season, we intensively searched for active nests and evaluated habitat characteristics. We located a total of 56 nests of which 27 (48.2 %) were successful and 29 (51.7 %) were unsuccessful. Most unsuccessful nests (19) were depredated, while 8 (27.6 %) were abandoned and 2 (6.9 %) parasitized. Nesting success was 46 %. Depredation was the main mechanism of nest failure, mostly during incubation. We identified operating variables at site and patch scales. Through analysis of binary logistic regression models for each scale, we predicted which variables increased the probability of nesting success. We also performed a multivariate logistic regression analysis to rule out possible interactions among variables. At nest scale we found that nest height, ranging 4.2 - 5.1 m had increased nesting success, while at patch scale, the probability of nest success was greater at sites with a tree density ≥ 12. Our results indicated that the probability of nesting success did not show a relationship between scales. We suggest a decoupling between scales and mechanism. Thus, habitat changes occurring at local scale did not seem to interfere with the patch scale. Also, our results showed that variation in characteristics at nest scale could also influence depredation, when depredation had been presented at random, since nests with medium and high probability of being successful, failed also for this reason. Thus, in order to understand the factors, mechanisms and life history characteristics influencing nesting success, we suggest that in addition to the environmental variables, future studies should also consider the bird nesting behaviour.

Keywords: nesting success, nest depredation, nesting site characteristics, scale decoupling, logistic regression.