We study the identification of a mixed proportional hazard model with lagged duration dependence when data provide multiple outcomes per stratum. Within strata variation is exploited to non-parametrically identify lagged duration dependence in more general models than in the literature.