On pg. 138 I write that an e becomes i "in an open syllable before an i in a following syllable". This formulation works for the first two examples (cinis < *kenis and sine < *seni) but it doesn't work for the second two examples (similis < *semlis and vigilis < *weglis) because these do not contain e in an open syllable. It might be better to assume *semilis > similis, but one would then have explain the survival of e in other forms in -ilis. This doesn't seem to be too hard to do: senilis, which has a long medial i, can be analogical to senex. Some, like Leumann, think this assimilation mainly operated over sonorants other than r. Additional examples of this sort would be milium 'millet' (cf. Gk. μελίνη) and tilia 'lime-tree' (cf. Gk. πτελέα 'elm-tree').