Pre-main-sequence and main-sequence binary systems are observed to have periods, P, ranging from 1 day to 1010 days and eccentricities ranging from 0 to 1. We pose the problem of whether stellar-dynamical interactions in very young and compact star clusters will broaden an initially narrow period distribution to the observed width. N-body computations of extremely compact clusters containing 100 and 1000 stars initially in equilibrium and in cold collapse are performed. In all cases, the assumed initial period distribution is uniform in the narrow range 4.5 ≤ log P ≤ 5.5 (P in days), which straddles the maximum in the observed period distribution of late-type Galactic field dwarf systems. None of the models lead to the necessary broadening of the period distribution, despite our adopted extreme conditions that favor binary-binary interactions. Stellar-dynamical interactions in embedded clusters thus cannot, under any circumstances, widen the period distribution sufficiently. The wide range of orbital periods of very young and old binary systems is therefore a result of cloud fragmentation and immediate subsequent magnetohydrodynamic processes operating within the multiple protostellar system.