In a seminal study, Cameron-Faulkner et al. made two important observations about utterance-level constructions in English child-directed speech (CDS). First, they observed that canonical in/transitive sentences are surprisingly infrequent in child-direct speech (given that SVO word order is often thought to play a key role in the acquisition of English syntax). Second, they found that many CDS constructions are introduced by a lexical frame (such as "Let's. . .," "There is . . .," or "What do you . . .?"). Using a much larger and more diverse dataset than Cameron-Faulkner et al., this study shows that CDS constructions vary with two important factors: (1) the interactive situation and (2) children's age. While canonical in/transitive sentences are not particularly frequent in free toy play sessions, they are predominant in other social situations (e.g. during mealtimes and shared book reading sessions) and they increase in frequency as children get older. Furthermore, our data show that the different CDS constructions occur with different types of frames that vary in length and structure. Many CDS constructions include a short lexical frame consisting of one or two words, but questions are often introduced by extensive frames formed from a small set of lexical items that follow a power-law distribution. Considering these findings, we argue that the lexical and structural properties of CDS constructions are likely to facilitate the acquisition of English grammar and, in particular, the acquisition of questions.