Wood's anomalies are discussed in detail by P.G. Murdin in ING La Palma Technical Note No. 76 in the context of INT IDS gratings, and a summary of the physical explanation for Wood's anomaly is repeated here.
Consider a reflection grating which produces a range of diffracted light in successive orders diffracted away from the normal. In some order, at some critical wavelength, the diffracted light lies in the plane of the grating. It is not possible for light beyond this point to be diffracted behind the glass of the grating. The power which would be sent into the forbidden region is redistributed back into the allowed orders. The power appears as an addition to the spectral response, with a sharp cut on at the critical wavelength and a steep decline to the red. This additional efficiency is almost entirely polarised perpendicular to the grating rulings.
Wood's anomalies occur at wavelengths:-
where d is the grating groove separation; is the grating angle (the angle between the grating normal and the collimator axis); and n is a positive or negative integer.
There are known to be Wood's anomalies in ISIS at 7200 Å (1200 grating, n=2), 6400 Å (600 grating, n=4) and 4400 Å (600 grating, n=-2).