This book uncovers properties of focus association with 'only' by examining the interaction between the particle and bare (or “evaluative”) gradable terms. Its empirical building blocks are paradigms involving upward-scalar terms like 'few' and 'rarely', and their downward-scalar antonyms 'many' and 'frequently', an area that has not been studied previously in the literature. The empirical claim is that associations of the former type give rise to unexpected readings, and the proposed theoretical explanation draws on the properties of the latter type of association. In presenting the details, the book deconstructs the so-called scalar presupposition of 'only' and derives it from constraints against its vacuous use. This view is then combined with a semantics of the evaluative adjectives 'many' and 'few' to explain why the unavailable (but expected) meanings of the given constructions are unavailable. The attested (but unexpected) readings of 'only+few/rarely' associations are derived from independently motivated LFs in which the degree expressions are existentially closed. Finally, the book provides new findings, based on the core proposal, about 'only if' constructions, and about the interaction between 'only' and other upward-scalar modified numerals (comparatives, and 'at most'). The book thus provides new data and a new theoretical view of the semantic properties of 'only', and connects it to the semantics of gradable expressions.