> Given that 'I don't know' isn't a lexical item but a phrase composed of at least four lexical items,
Well, the three words "I don't know", if fully realized, are only three lexical items.
English has three words that can be spoken without opening your mouth, with the meanings "yes", "no", and "I don't know". Yes and no are conventionally spelled "uh-huh" and "uh-uh", and consist of voicing interrupted by an [h] (in the case of yes) or a glottal stop (in the case of no).
The "I don't know" sequence consists of voicing broken into three tonal segments. It is a single three-syllable lexical item in which every syllable is "mm" (if your mouth is closed) or "uh" (if it's open), but which uses a tone sequence copied from the phrase "I don't know".
It's a good model for how languages develop tone in the first place.
I see what you mean. I wonder if that's really the only item of that type in English.
> It's a good model for how languages develop tone in the first place.
I'm not fully convinced for that. It's a really specialised instance and it's still standing in for a phrase (a whole proposition in fact) rather than a lexical item as such. Full-blown lexical tone, at least in some cases, seems to develop as the result of reinterpreting tonal correlates of some other phenomenon, e.g. Punjabi tone as resulting from reinterpretation of aspiration.
> Full-blown lexical tone, at least in some cases, seems to develop as the result of reinterpreting tonal correlates of some other phenomenon
I would have said that the "I don't know" sequence matches this description -- it's just the tonal contour of the original phrase, with the other details stripped out. I don't see how that differs from Punjabi syllables preserving their (originally) allophonic tone while losing their (originally) phonemic aspiration, such that the tone becomes phonemic.
> I wonder if that's really the only item of that type in English.
Obviously I can't guarantee that, but I think it's pretty likely for several reasons:
- I can't think of anything similar.
- The yes/no question answers (including the "I don't know" sequence) are a clear conceptual group and differ from normal English in a few ways. The "no", "uh-uh", is also (I conjecture) unique, being the only English lexical item to contain a phonemic glottal stop. (Plenty of English words feature an obligatory glottal stop, but that stop is allophonic, as when the /t/ of "kitten" is transformed into a glottal stop by the /n/ that follows it.) It is of interest that the only phonemic glottal stop and the only phonemic tone should occur in these two closely related words. I strongly suspect that this group of answer words has these unusual features because of the constraint that you can pronounce them without opening your mouth.
- It's a single word that contains a lot more information than one English word should. For the Latin word "nescio" to mean "I don't know" is unsurprising; all Latin verbs are marked with subject agreement and in the case of a first person subject that is generally the only subject marking. English subject marking is almost always mandatory. This is just more evidence that the tone sequence is quite different from normal words.
- English has a non-vestigial system of grammatical tone at the sentence level; tonal marking is obligatory for yes/no questions. But sentence-level grammatical tone can't really coexist with lexical tone. So I'd be very surprised to see other instances of lexical tone outside very special cases. (For example, the phrase "I don't know" can't ever be a yes/no question, and so can't conflict with the English interrogative tone.)
> it's just the tonal contour of the original phrase, with the other details stripped out. I don't see how that differs from Punjabi syllables preserving their (originally) allophonic tone while losing their (originally) phonemic aspiration, such that the tone becomes phonemic.
because, as you elaborate, the English example you mention is really isolated - it's not enough to spawn a system of tone. Even in this case (the "I don't know" sequence), the tone isn't really contrastive.
> the phrase "I don't know" can't ever be a yes/no question
Sure it can, though the contexts wouldn't be that frequent. And some speakers have final rising tone even on non-yes/no questions.
> English has a non-vestigial system of grammatical tone at the sentence level
And of course there's more than just question-associated tone, there's also focus-associated tonal prosody.
> Even in this case (the "I don't know" sequence), the tone isn't really contrastive.
It's not contrastive in that there is no minimal pair that would demonstrate it. But good luck being understood if you don't produce it correctly. The lesson I would draw here is more "minimal pairs aren't sufficient to demonstrate every phoneme" than "it's not a phoneme until a minimal pair exists".
> the English example you mention is really isolated - it's not enough to spawn a system of tone
I didn't claim and don't believe that English is developing lexical tones. But I do believe that the development of lexical tone in this single unusual word is fundamentally similar to the larger-scale development of tone in other languages, and that the still-obvious relationship between the phrase and the tone sequence makes this a good example to help English speakers understand how it can happen.
> And of course there's more than just question-associated tone, there's also focus-associated tonal prosody.