“…The first is that, like many distributional learning mechanisms (though see Frank, Goldwater & Keller, 2013 Given the obvious prosodic, pragmatic and structural differences between declaratives and questions, this feature of the model might be regarded as somewhat implausible, particularly in view of recent evidence that even very young children can distinguish between declaratives and questions in the input (Seidl, Hollich & Jusczyk, 2003;Homae, Watanabe, Nakano, Asakawa & Taga, 2006;Soderstrom, Ko & Nevzorova, 2011, Geffen & Mintz, 2012Frota, Butler & Vigário, 2014). For example, Geffen and Mintz (2015) show that by 12 months children can distinguish between declaratives and polar interrogatives even in the absence of prosodic cues, and argue that although infants initially use phonological information to distinguish between sentence types, they have already begun to learn generalisations about the corresponding word-order patterns before the onset of multi-word speech.…”