Such expressions of phonemes in terms of components are not in themselves of value to linguistics. The advantage they offer in reduction of the number of elements may be more than offset if connecting them with the distributional and phonetic facts requires more complicated statements than are required for regular phonemes. We consider the possibility of such analysis into components only because, as will be shown below, we can select the components in a way that will enable us to give simpler statements of the facts about phonemes. […] We can now say in general terms what we must do when we analyze phonemes into components. We take a list of phonemes, each with its phonetic and distributional description; we select a number of components; we select some method for combining these components simultaneously (e.g. not more than three components at a time), in such a way that each combination permitted by the method will identify a phoneme, and that the grammar becomes simpler and briefer when written in terms of the components.
- Harris (1944), Pag. 186

The greatest advantage from the analysis into components comes from the components with a length of two or more phonemes. These components enable us to express situations which could not be symbolized by the fixed-length phonemes. We shall investigate these components first. The first technique we shall use will yield the syntactic contours. The second will yield a way of treating the limitations of distribution of phonemes. The third will yield special cases of segmental phonemes. Finally we shall consider the components whose length is that of only one phoneme.
- Harris (1944), Pag. 188