A natural way of understanding (non-epistemic) looks talk in natural language is phenomenalist: to ascribe looks to objects is to say something about the way they strike us when we look at them. This explains why the truth values of looks-sentences intuitively vary with the circumstances with respect to which they are evaluated. But Mike Martin (2010) argues that there is no semantic reason to prefer a phenomenalist understanding of looks to “Parsimony”, the position according to which looks are basic visible properties. He suggests a semantics for looks-sentences that explains their intuitive truth values and is compatible with Parsimony. I argue that there is semantic reason to prefer a phenomenalist understanding of looks to a parsimonious one since there is a simpler semantics compatible with a phenomenalist understanding of looks, but not with Parsimony. This semantics provides a better explanation of the relevant truth value distribution.