This talk will present several experiments that reveal a previously unknown relationship between lexical confusability and coarticulation. Lexical confusability refers to the chance that one word will be misheard as another, and is understood here in terms of two factors that play a role in lexical competition: usage frequency and phonological similarity neighborhood. I will present cross-linguistic data that show that speakers produce confusable, or "hard," words with significantly more coarticulation than less confusable, or "easy," ones. And I will show that the effect of confusability on coarticulation is truly lexical, arising at the level of whole words. It does not reflect segment-level confusability and is insensitive to phonological contrast. These results have interesting implications for views of both coarticulation and hyperarticulation and support a functional theory in which talkers use increased coarticulation and/or hyperarticulation to mitigate listeners' lexical access difficulties.