My intuition is that onomatopoeia is quite rare and that the letters and sounds of most words are meaningless (e.g. if "chicken" and "fish" were switched all along, no one would think "That word doesn't sound like it tastes").
But I want to know the frequency of onomatopoeic words (or is that a spectrum rather than binary?) both (1) as a proportion of a reference vocabulary, such as a dictionary (e.g. "10% of all words are onomatopoeic"), and (2) as a proportion in spoken or written usage.
Does such a corpus annotated with the property "this word is onomatopoeic" exist and what are those frequencies (if they are public I can calculate that myself also)?