When you have a symbol representing a word, it represents all the defined (by convention) properties of the word. All the semantics, all the the vocalization. T

When you have a symbol representing a word, it represents all the defined (by convention) properties of the word. All the semantics, all the the vocalization. The defined part, that is. You don't get any undefined parts, e.g. if you had separate symbols for "read (past tense)" and "read (present tense)" they would "naturally" include relative speed but not relative pitch. If vowels are significant at one point in the language's use, but the rebus principle allows for mismatching substitutions it would be a strong indication that at some earlier point they were not, either because they had regionally differing conventions (while still mutually understood) or because they were just filled in ad-hoc based on context (e.g. how pitch is decided upon when speaking a language where it is not significant: just make it sound pleasant and/or communicate macro structure). Of course, over time weird things do happen and you might end up with a mess like english where the relationship between phonetic encoding and actual vocalization is incredibly weak and serves only as a hint for narrowing down the actual encoding+context to semantics to vocalization lookup chain.