I will analyze the Slave D-effect in the framework of Optimality Theory (Prince &
Smolensky 1993). My analysis will cover the full range of phenomena and will not refer to morphological information. This makes it superior to previous analyses of the D-effect, e.g., Lamontagne & Rice 1994, 1995, which have to refer to morphological information, and which do not account for all D-effect alternations. I will propose constraints guiding the inner workings of coalescence (which features of which input segment are maintained), thus shedding light on the nature of coalescence in general. Finally, I will show that my analysis is more valid universally, as it is compatible with accounts of coalescence in child language (Ganandesikan
1995).