Abstract: Work engagement (WE) is one of the most important constructs in both nursing and management literature. The states of nurses' work engagement affect their nursing quality and the level of medical services. However, the level of work engagement reported by nurses was lower than the level of work engagement reported by other occupational groups in the hospital. Considering their working role as highly stressful and specialized, it is necessary to provide new graduate nurses with holding work environments. Despite of the accumulated knowledge, to date the reciprocal relationship between holding behaviors(HB) and work engagement among new graduate nurses remains ambiguous. Some scholars pointed out that holding behaviors from mentors are a key predictor of newcomers’ work engagement (e.g., new graduate nurses). However, other scholars held the opposite view and found mixed or even contradictory findings from a longer time frame. Thus, many scholars have called for future research to unravel the complex relationship between holding behaviors and work engagement. Some studies, using longitudinal correlations or cross-lagged regression, have provided initial evidence about the influence and possible direction of holding behaviors and work engagement. However, these studies ignored a different trajectory per participant as marked by a different (subject-specific) intercept and slope. Without accounting for the overall trajectories of the variables over time, spurious cross-lagged effects might appear while they essentially do not exist. To clarify the effect of holding behaviors on work engagement, this study used autoregression latent trajectory model (ALT) which combines a latent trajectory model (LTM) with an autoregressive (AR). The sample consisted of 215 new graduate nurses from 5 large or medium-sized hospitals in Guangdong province with five measurement occasions over a week period. In this study, there were 110 males (51.16%) and 105 females (48.84%). This distribution of gender was consistent with that of the new graduate group of nurses. The age range of the participants was from 19 to 23 years old (M = 20.93, SD = 1.25). The level of nurse work engagement and perceived holding behaviors from their mentors were measured by Utrecht Work Engagement Scale and Holding Behaviors Index for Mentor, respectively. For our analyses, we conducted the ALT approach in Mplus 8.0 models. First, latent curve model (LCM), autoregressive (AR) and ALT models were tested for holding behaviors and work engagement separately. Second, multivariate LTM (estimating the progression of HB and WE over time for each individual, and the possible association between HB and WE trajectories), AR (estimating time-specific associations of HB and WE and their possible influence on each other from day to day), and ALT models (allowing for the examination of both processes simultaneously) were estimated. As ALT is a complex model, it should always be compared with simpler models to assess if its complexity improves significantly the representation of the data. The findings showed a positively concurrent effect of HB (T) on WE (T) and a positively cross-lagged effect of WE (T) on HB (T+1). Results also showed that the trajectories of HB and WE in new graduate nurses were relatively stable. More importantly, we found that the slope of HB was negatively correlated with the slope factor of WE. It implied that for new graduate nurses, increased in HB related to decreased in WE. Overall, the findings of the present study had three major contributions: First, from a dynamic process perspective, this study clarified the potential negative path of holding behaviors in work engagement, which provided rich empirical evidence for future studies. Second, this study confirmed the reverse relationship between perceived holding behaviors and work engagement and revealed the intra-individual dynamic nature of holding behaviors and work engagement. Third, future research should carefully interpret the between-person level results of work engagement and sincerely infer these findings to the within-person level in order to enhance our further understanding in work engagement research.