First, we utilized multilevel mixed model analyses with a random intercept in R, as described in the preregistration. However, we did not use the additional growth cure modeling and DSEM in MPLUS to analyze the data. With latent growth curve modeling we could have searched for the moment where the level of oxytocin or cortisol peaked for each participant separately. However, this did not correspond with our research questions which is why we did not do this in the end. Moreover, R is free downloadable software while MPLUS is not. Second, we did not include the first preregistered research aim of whether maternal care after distress decreases the release of cortisol in the child. Originally, the current study was designed to measure oxytocin changes over time and different situation in a within-subject design. The decision to subtract cortisol levels as well from the saliva samples was made later. After further reasoning, the current study's design was not fit to adequately measure cortisol responses to the different situations since cortisol secrection can only be detected in saliva after 15 minutes (Miller et al., 2016). Therefore, the measurement points of saliva were not corresponding to the cortisol responses to the different situations. For example, cortisol recovery during free play should be visible after 15 minutes after the end of the recovery phase. However, we do not have an accurate data point that goes far enough in the future to measure this cortisol level. Additionally, after inspection of the data, we found that baseline levels of cortisol were very high. It might have been that children showed anticipatory stress because they did not know what was coming. This would have influenced the analyses since cortisol might not have been able to increase even more in response to the TSST-C compared to baseline. Therefore, a better baseline measure of cortisol would have been after children were in the room for 60 minutes (Houbrechts et al., 2021). However, in the prospect of the current study, this would not have been feasable since the data collection already took around two hours during a school day.