ABSTRACT. Atmospheric intensity scintillation of stars on milli-and microsecond time scales was extensively measured at the astronomical observatory on La Palma (Canary Islands). Scintillation statistics and temporal changes were discussed in Paper I, while this paper shows how scintillation depends on optical wavelength. Such effects originate from the changing refractive index of air, and from wavelength-dependent diffraction in atmospheric inhomogeneities. The intensity variance aj increases for shorter wavelengths, at small zenith distances approximately consistent with a theoretical X -7/6 slope, but with a tendency for a somewhat weaker dependence. Scintillation in the blue is more rapid than in the red. The increase with wavelength of autocorrelation time scales (roughly proportional to >/\) is most pronounced in very small apertures, but was measured up to 0 20 cm. Scintillation at different wavelengths is not simultaneous: atmospheric chromatic dispersion stretches the atmospherically induced "flying shadows" into "flying spectra" on the ground. As the "shadows" fly past the telescope aperture, a time delay appears between fluctuations at different wavelengths whenever the turbulence-carrying winds have components parallel to the direction of dispersion. These effects increase with zenith distance (reaching -100 ms cross-correlation delay between blue and red at Z = 60° ), and also with increased wavelength difference. This time delay between scintillation in different colors is a property of the atmospheric flying shadows, and thus a property that remains unchanged even in very large telescopes. However, the wavelength dependence of scintillation amplitude and time scale is "fully" developed only in small telescope apertures ( ^ 5 cm), the scales where the "flying shadows" on the Earth's surface become resolved. Although these dependences rapidly vanish after averaging in larger apertures, an understanding of chromatic effects may still be needed for the most accurate photometric measurements. These will probably require a sampling of the (stellar) signal with full spatial, temporal, and chromatic resolution to segregate the scintillation signatures from those of astrophysical variability.