Wikipedia says that:
If a mesh covers more pixels in screen space than it has vertices, interpolating colour values from samples of expensive lighting calculations at vertices is less processor intensive than performing the lighting calculation for each pixel as in Phong shading.
Why does it matter that there are more pixels than there are vertices? In the quote it also says that Gourand shading performs lighting calculation "at vertices" while Phong does it in pixels.
I have always thought that both shading schemes are the same, just that Gourand interpolate colours along vertices, while phong creates a reflection vector to interact with the light sources for specular reflection. Is this correct?