One of the principal limitations of phase-shift interferometry is the shift
non linearity. We propose two methods for measuring this non linearity, on
e applicable to the registration of a limited number of interferograms (3 o
r 4) and the other applicable to many interferograms (typically 16). We see
two main advantages in these methods: first, they require no extra instrum
entation, and, secondly, shift non linearity is measured in situ, that is p
articularly important in the case of mechanical phase shift. Experiments co
nfirm the validity of the corrections, allowing in our case reduction of sy
stematic errors by 30 to 45%.