In some scientific work describing Discrete Fourier Transform-algorithms, such as the well-known Cooley-Tukey algorithm, I came across the term 'Butterfly operations' and 'Butterfly combinations', which seem to refer to the fact that an algorithm takes elements that are half an array away (where the array itself is a power-of-two in size), which mean that if you want to write an iterative implementation, that you will have to bit-reverse the indices of the input (or the output).
Is this indeed what is called 'Butterfly combinations'? And why are these called 'Butterfly combinations'? What is the origin of this name?