Using cross-linguistic evidence from the relationship of epenthetic vowel choice to the vowel system of a given language, this paper proposes a set of context-free markedness constraints on vowel features.  Universally ranked constraints yield the result that front vowels are more marked than back and that round vowels are marked; languages may vary in how they rank the markedness of low and nonlow vowels.  If low vowels are unmarked the epenthetic vowel is [a].  If nonlow vowels are unmarked, the epenthetic vowel is the least marked possible given the vowel system of the language. [barred-i] is epenthetic if present; [schwa] is epenthetic if it is present and [barred-i] is not; these vowels are unmarked on the backness and rounding dimensions.  If neither ‘central' vowel is part of the system, the epenthetic vowel will generally be [i]: constraint conflict forces the choice of a more marked front vowel in order to avoid violations of markedness on the mid and round dimensions.