The accepted result for the Kramers-Kronig relation for the phase spec
trum Psi(omega) of the complex reflection coefficient at normal incide
nce r(omega) is shown to contain an error. Previously, phase shifts co
mputed from reflectance data using Kramers-Kronig analysis have in som
e cases been shifted by +/-pi with no basic justification other than a
claim that this provides agreement with accepted values and consisten
cy with the incomplete Stokes treatment. The theoretical basis for thi
s shift is provided in this short note. We derive the correct formula,
and show that well-known model dielectrics are consistent with our re
sult and not the accepted result.