When children have to select one of two structures, do they start with the universally unmarked structure or with the one preferred by the ambient language? This question is directly relevant to metrical systems, which often employ either iambs or the universally unmarked trochees. We argue that children start with the universally unmarked trochaic foot, unless their ambient language provides them with sufficient data to arrive at the language-specific preferred foot prior to the onset of speech. We show that Hebrew-acquiring children, unlike French ones, are exposed to ambiguous data, which do not allow them to determine the type of foot the language's stress system employs. Our quantitative data provide evidence that in such a case, children adhere to the trochaic foot during the very early stage of acquisition (in the case study presented here, the early stage of acquisition refers to the first 100 cumulative target words). Later on, children follow the frequency-based preference in Hebrew, where final stress, and thus the iambic foot is employed in about 75% of the nouns.