Previous research has demonstrated great variability in the rates of scalar inferences across different triggers (Doran et al., 2009; van Tiel et al., 2016). In the current study, we show that variation is more systematic than previously thought. In particular, we present experimental evidence suggesting that endorsements of scalar implicatures (i) are anti-correlated with the degree of negative strengthening of the stronger scale-mate (e.g., whether John is not stunning is interpreted as conveying that John is rather ugly) and (ii) are affected by the scale structure and the underlying scalar semantics of gradable adjectives (in particular boundedness, polarity, and adjectival extremeness). Overall, our research suggests that scale structure should be taken into account in theories of implicature.

It has been generally assumed that certain categories of numerical expressions, such as 'more than n', 'at least n', and 'fewer than n', systematically fail to give rise to scalar implicatures in unembedded declarative contexts. Various proposals have been developed to explain this perceived absence. In this paper, we consider the relevance of scale granularity to scalar implicature, and make two novel predictions: first, that scalar implicatures are in fact available from these numerical expressions at the appropriate granularity level, and second, that these implicatures are attenuated if the numeral has been previously mentioned or is otherwise salient in the context. We present novel experimental data in support of both of these predictions, and discuss the implications of this for recent accounts of numerical quantifier usage.

The adjectives of quantity (Q-adjectives) many, few, much and little stand out from other quantity expressions on account of their syntactic flexibility, occurring in positions that could be called quantificational (many students attended), predicative (John's friends were many), attributive (the many students), differential (much more than a liter) and adverbial (slept too much). This broad distribution poses a challenge for the two leading theories of this class, which treat them as either quantifying determiners or predicates over individuals. This paper develops an analysis of Q-adjectives as gradable predicates of sets of degrees or (equivalently) gradable quantifiers over degrees. It is shown that this proposal allows a unified analysis of these items across the positions in which they occur, while also overcoming several issues facing competing accounts, among others the divergences between Q-adjectives and 'ordinary' adjectives, the operator-like behavior of few and little, and the use of much as a dummy element. Overall the findings point to the central role of degrees in the semantics of quantity. '' (4) a. I much prefer wine to beer. adverbial b. I slept little. '' c. much loved/little known; much/little alike/different '' Finally, much occurs in a context that has come to be known as much support (Corver 1997), where an adjective has been pronominalized with so: (5) Fred is generous; in fact, he is too much so / so much so that it worries me.

Six tests of the spontaneous speech of twenty-one English-speaking children (1 ; 10 to 2 ; 8; MLUs 1.53 to 4.38) demonstrate the presence of the syntactic category determiner from the start of combinatorial speech, supporting nativist accounts. Children use multiple determiners before a noun to the same extent as their mothers (1) when only a and the or (2) all determiners are analyzed, or (3) when children and mothers are matched on determiner and noun types and determiner+noun tokens. (4) Overlap increases as opportunity for overlap increases: children use multiple determiners with more than 50% of nouns used at least twice with a determiner and with 80% of nouns used at least six times with a determiner. (5) Formulae play a limited role in low-MLU children's determiner usage, increasing with MLU. (6) Less than 1% of determiner uses are errors. Prior results showing no overlap are likely a sampling artifact.