It is commonly assumed that *CLASH and *LAPSE evaluate syllable-sized constituents: a sequence of two adjacent stressed syllables violates *CLASH, while a sequence of two stressless syllables violates *LAPSE (see e.g., Prince 1983, Gordon 2002 for *CLASH; Nespor & Vogel 1989, Green & Kenstowicz 1995, Gordon 2002 for *LAPSE). In this paper I argue, based on patterns of secondary stress in American English -ization, that there exist rhythmic constraints that are evaluated with respect to (normalized) duration. I present evidence consistent with this claim from corpus, judgment, and production studies.