Background Motor vehicle crashes are the 2^nd most common fatal unintentional injury in young children in the USA. Proper use of car seats reduces risk of death and serious injury by 78%, but 94% of parents instal car seats improperly. Parents must recognise when they instal car seats improperly in order to take corrective actions, but such recognition may be lacking. The objectives of this study were: 1) assess the accuracy of perceived mastery among novice car seat users, and 2) identify predictors of overconfidence among users who instal car seats improperly.

Methods Novice users ages 18–29 (M = 25; SD = 2.4) were randomly assigned to receive installation instructions from the manufacturers’ guide or an expert technician via phone or merged reality app. Participants installed a car seat and strapped a life-like infant doll, then reported perceived mastery and quality of instructions. Trained assessors rated mastery using a structured coding sheet.

Results 4 of 39 total participants (10%) installed the seat and strapped the doll with no errors. 27 (70%) made 2 or more errors with either the seat or straps and were judged to have failed, of whom 17 (63%) were overconfident (OC). 19 (49%) failed the seat, with 13 (68%) OC. 21 (49%) failed the straps, with 14 (67%) OC. Total errors were significantly lower in OC vs. non-OC participants among those who failed the seat and those who failed the straps (t = 3.71, p = 0.002 and t = 2.41, p = 0.03, respectively). Among those who failed the straps, perceived quality of instructions was significantly higher in OC vs. non-OC participants (t = 2.09, p = 0.05).

Conclusions Expert instruction reduces car seat installation errors, but novices continue to unknowingly make life-threatening errors. Overconfidence may be more likely with fewer total errors and higher perceived quality of instructions. Among other strategies, merged reality apps may reduce overconfidence and errors.