Previous studies demonstrated that productive longevity (PL, age of culling minus age at first calving) in dairy cows can be considered as a proxy for their resilience. A key aspect to apprehend the underlying drivers of resilience of an individual cow is to observe how she copes to exposure to environmental perturbations (EP). A perturbation is hereby defined as any deviation from environmental normality (e.g., heat chock, feed depression, infectious agent). Breeds of dairy cows genetically selected for high milk production potential, such as Prim ‘Holstein and Montbeliarde, are considered to be more sensitive to EP. Such a higher sensitivity may be reflected in the lactation function by deviations from an expected normal curve. However, to our knowledge, few studies have been reported so far in dairy cows that make the link between exposure to perturbation and their consequences on PL. We hypothesized that cows that are less affected by perturbations detected in their lactation curves (i.e. less total milk yield losses) may have a higher PL. We used daily data collected over 20 years (1996 to 2016) concerning daily milk yield from 489 lactations of 129 Prim’Holstein and Montbeliarde cows that were reared in similar conditions at an experimental farm located in the French Massif Central region. This experimental farm was designed to mimic a low-input mountain grassland-based system. These time series data described the entire PL of cows with a weekly time step. A previously developed algorithm called perturbed lactation model was fitted to time series data to decompose each lactation with a theoretical unperturbed curve and a perturbed curve made of all detected deviations from the theoretical potential. This allowed to calculate for each lactation the number (N) of detected deviations and the percentage of milk loss (%MYloss). At the lifetime scale, average N and average %MYloss were expressed in a per-lactation basis, to avoid the bias that cows being exposed to more perturbations the longer they stay in the herd. A data subset was then classified in two different groups respectively of low and high PL using the 1st or 3rd quantile of PL as thresholds (1045 ± 3 d and 1156 ± 4 d, respectively). The number of deviations detected for cows from the shorter PL group was significantly fewer than for those from the longer PL group (3.7 ± 0.1 and 4.3 ± 0.1 deviations·lactation-1 , respectively; P =0.0025). However, the former presented slightly higher %MYloss than the later (12.3 ± 0.4 and 11.1 ± 0.4%, respectively, P = 0.05), despite the fact that cows from both groups presented statistically similar cumulative milk yields up to a period of 305d (5123 ± 140 and 4858 ± 145 kg·lactation-1, respectively, P = 0.386). In other words, assuming that cows from both PL groups were exposed to a similar number of perturbations, cows with less MY loss despite showing more deviations events in their lactation curves were associated with higher PL. On-going analyses are in place to verify if different PL groups subside and recover differently from the detected perturbations. We are also investigating whether or not a cause-effect relationship can be established between the responses to perturbations and PL. This will be done by verifying if the same results above remain true for 1st lactation only, in other words, if the number of deviations detected at first lactation as well as their consequent MY losses are related to PL.