Objective: Sociotropy and autonomy have been described as personality styles promoting the occurrence of stress in individuals. The SAS (for sociotropy-autonomy scale) corresponds to a 60-item instrument devoted to sociotropy and autonomy estimation in various populations. However, the different SAS versions display variations in item scoring, in scale structure and in factor composition. Furthermore, there is a lack of scale invariance analysis. Consistent with a larger work dedicated to understanding contributors of students' achievement during their trainings, the objective of this study was devoted to a refinement of French SAS item properties and to a measure of scale invariance in order to ensure that the French SAS be a trustable tool to measure sociotropy and autonomy among first- year students. Because a preliminary analysis invalidated several SAS items, a second objective has consisted of the definition and validation of a new scale estimating social dependency in students., Methods: In all, 2365 students registered in first year of technological training (n=778) and health training (n=1587) were enrolled. They were asked to electronically answer the 60-item French SAS. The responses were collected between October and November 2017 for students registered in technological formation, and between March and April 2018 for health students. Item-score correlation coefficients (r item-score ) and anti-image correlation coefficients (AIC) were calculated for each item, and threshold values (r item-score >0.3 ; AIC>0.6) were considered as acceptance criteria. Factor analyses were run in order to determine scale structure. Internal consistency was deduced from Cronbach's α, McDonald ω and Great Lower Bound (GLB) coefficients. Convergent and discriminant validities were analysed in considering construct reliability coefficient (CR>0.7), average variance extracted (AVE>0.4), mean shared variance and squared correlation coefficient calculated between two factors, as validity criteria. Configurational, metric and scalar levels of invariance were analysed prior to statistical comparisons of the scores obtained by different subgroups., Results: In all, 1223 responses were collected and analysed. GLB and ω coefficients calculated for the full SAS indicated unacceptable internal consistency. Of the 60 items, 40 did not meet the acceptance criteria (i. e. r item-score <0.3 and AIC<0.6). The remaining items had acceptable psychometric properties, and their composition defined a new scale reflecting the measure of social dependency. The overall scale internal consistency was good (ω=0.83, GLB=0.88). Factor analyses resulted in a replicable 5-factor structure including: need of affection (α=0.74, ω=0.74, GLB=0.77), fear of dropping out (α=0.67, ω=0.68, GLB=0.70), fear of loneliness (α=0.61, ω=0.61, GLB=0.67), attention to others (α=0.69, ω=0.69, GLB=0.71) and worry about the disapproval of others (α=0.71, ω=0.71, GLB=0.74). Discriminant validity was satisfied for all factors. Convergent validity was entirely satisfied for need of affection, fear of dropping out and for attention to others, but it was not optimal for fear of loneliness and worry about the disapproval of others. Invariance measurements identified non-invariant items that were discarded from score calculations. Following statistical comparison, it was observed that female students had higher social dependency and fear of dropping out than males. Furthermore, female students registered in health training had stronger attention to others than did their male equivalents. It was also observed that female health students scored better for social dependency, fear of dropping out, attention to others, and importance of others' gaze than their technology-educated counterparts., Conclusion: This work has invalidated the use of the overall 60-item French SAS to measure sociotropy and autonomy in first year students. By contrast, the refinement of the French SAS items led to a trustable 20-item instrument to investigate social dependency. Scale invariance characteristics allow confident statistical comparisons between sub-groups. This work shows that first-year female students registered in health or technological trainings exhibit a higher social dependency than males. Furthermore, first-year female health students show more social dependency than their technological counterparts. Because the conditions of recruitment of first-year students in heath formations in France are planned to change in September 2020, future studies will be warranted to analyse their social dependency., (Copyright © 2019 L'Encéphale, Paris. Published by Elsevier Masson SAS. All rights reserved.)