An alternative theory of imprecision

<div class="page" title="Page 1"><div class="layoutArea"><div class="column"><p><span>Numerical expressions are often used imprecisely or approximately. This paper defends a novel analysis of numerical imprecision based on the notion of scale granularity, construed here in terms of sets of alternatives. I apply this approach to account for new facts relating to the interaction of (im)precision and comparison, in particular the necessarily precise interpretation of measure expressions in compara- tives, and the negative polarity status of overt approximators in comparatives (e.g. </span><span>Mabel owns *(no) more than about one hundred sheep.</span><span>) </span></p></div></div></div>