I tend to reject the classification of languages onto the analytic-synthetic spectrum. I don’t think such a spectrum exists beyond a superficial level. Recall that this classification depends on the morphemes-per-word ratio, and as such, depends on the putative existence of “words”, i.e., that there are so-called bound morphemes making up multi-morpheme words in synthetic languages. What makes these morphemes especially more “bound” in synthetic languages? The fact that even in a purely analytic language, free morphemes are not really free in the sense that they must be arranged in some order with some grammar to make meaning, suggests that there is really no fundamental difference between analytic and synthetic languages. All morphemes are bound to one degree or another. Whether they can be uttered in isolation is more about function.

For one, morphemes in compound words are not any more bound in a synthetic language than in similar situations in an analytic one. Each compound part can be uttered with meaning, and therefore are technically words. The only difference might be orthography. On the other hand, grammatical morphemes that show up in, e.g., inflected forms in synthetic languages might seem more bound, but analogous grammatical particles in analytic languages are really not any more free — they just appear so due to the more recognizable grammaticalization to linguists. Uttering particles in isolation (for their grammatical meaning) is as odd as uttering inflectional affixes. In fact, inflections may very well be advanced/contracted forms of previously unbound particles and auxiliary words, and serve the same purpose as they do in analytic languages: they appear in some fixed order in relation to other morphemes like the root, and they are relatively constant across words for the same meaning.

Finally, how many morphemes exist in an utterance is not even clear, either, and often an artifact of the vagaries of phonetic segmentation or the corresponding orthographic tradition. The contraction “I’m'ona < I’m gonna < I am going to” might be analyzed as a single morpheme (I’m'ona) or three (I-am-gonna) or five (I-am-go-ing-to). If “I’m'ona” be a word, then it would appear fairly synthetic with a number of bound morphemes, but is it really? The fact that “words” (beyond “morphemes”) only exist to describe synthetic languages and that word boundaries are things that are tracked really only by alphabet orthographic traditions that transcribe small phonetic atoms (rather than morphemic atoms) should be indication enough that the “analytic-synthetic spectrum” is fiction.