It is uncontroversial that, in many varieties of English, the realization of /l/ varies depending on whether /l/ occurs word-initially or word-finally. The nature of this effect, however, remains controversial. Previous analyses alternately analyzed the variation as darkening or lightening, and alternately found evidence that the variation involves a categorical distinction between allophones or a gradient scale conditioned by phonetic factors. We argue that these diverging conclusions are a result of the numerous factors influencing /l/ darkness and differences between studies in terms of which factors are considered. By controlling for a range of factors, our study demonstrates a pattern of variability that has not been shown in previous work. We find evidence of morpheme-final darkening and morpheme-initial lightening when compared to a baseline of morpheme-internal /l/. We also find segmental effects such that, in segmental contexts which independently darken /l/, one can observe /l/ lightening, and contexts which independently lighten /l/ can make lightening effects undetectable. Morphological and prosodic effects are hence sometimes trumped by segmental context. Once contextual effects are controlled for, there is evidence both for morphologically-conditioned /l/-darkening and for morphologically-conditioned /l/-lightening, both of which can be understood as a result of prosodic differences reflecting morphological junctures.