[…] the identification of grammatical and semantic units has met with a number of objections [...]. The most important are these: [...] (c) There may be meaningful submorphemic segments (“phonaesthemes”, e.g. 'fl'-'ow', 'fl'-'it', 'fl'-'y', 'fl'-'oat', etc.). [...] But none of these objections seems sufficient. [...] (c, d) Such phenomena as “phonaestheme” and “idiom” are indeed definable as many-to-one correspondences between grammatical and semantic units.
- Weinreich (1963), a pag.117