OBJECTIVE: In prior studies of febrile seizures (FSs), prolonged FSs were defined, absent empirical evidence, as lasting 10 or 15 minutes or more. We assessed the distribution of FS duration in a cohort with first FSs, and the association between FS duration and baseline characteristics of the children. METHODS: We calculated the observed cumulative probability, S(t), that a FS would last at least t minutes, S(t) = exp(-t/tau). Data were also fit using a model obtained as the sum of 2 exponential distributions (S[t] = alphaexp[-t/tau(1) ] + [1 - alpha]exp[-t/tau(2) ]). After assessing the best fit, the cutoff defining long FS was determined. Logistic regression was used to examine associations between long FSs and baseline characteristics, behavior, and development. RESULTS: In 158 children with a first FS, median duration was 4.0 minutes. Duration of FS was best fit by a 2-component mixture exponential model. Using this model, we identified 1 population that accounts for 82.3% of FSs and has a mean duration of 3.8 minutes (short FS) and a second population that accounts for 17.7% of FSs and has a mean duration of 39.8 minutes (long FS). Long FSs were significantly associated with developmental delay (p = 0.010) and delays and younger age at first FS (p = 0.048). INTERPRETATION: Like the distribution of afebrile seizure duration in children, the distribution of first FS duration is best modeled by assuming 2 populations. Developmental delay and younger age are associated with prolonged FSs. Our data lend further support to defining 10 minutes as the upper limit for a simple FS.