Neighborhood safety is defined inconsistently across epidemiologic studies - a conceptual problem that results in incomparable measurements, hampering the design of health interventions. Using child behavior problems (measured via the Child Behavior Checklist) as the outcome of interest, this study directly compared four measures of neighborhood safety: two of experienced safety and two of perceived safety, with each one measured at family and community levels. These included children's direct experience of harm, parental perceptions, community crime statistics, and community perceptions. In a sample of 3291 ten-year-olds from the Generation R cohort (living in municipal Rotterdam, Netherlands, 2013), all four measures were correlated (χ2 ≥ 9.2, P < 0.002 in pairwise chi-square comparisons), but ultimately identified different levels of risk for behavioral health. Direct experiences of harm, parental perceptions, and community crime statistics were all associated with increased child internalizing behaviors (β = 3.12, β = 2.10, and β = 1.77, respectively), while only experiences of harm and parental perceptions were associated with increased externalizing behaviors (β = 2.75 and β = 1.31, respectively). These results provide novel evidence that the conceptual distinctions underlying different measures of neighborhood safety are meaningful for child mental health and should be considered in intervention design.