Lerp with a float3 for the weight

Cg has a variation of lerp that uses a vector for the weight, but I can't find any explanation of how that works. Since the weight vector always has to equal the output, I'm assuming that it uses each component of the weight vector for the corresponding component of the two vectors being lerped together; but I've seen it used for blending textures in triplanar shaders, in which case the colors would be munged if it changed each of the RGB components according to a different weight. NVidia's documentation doesn't describe how it works in this case.