What evidence is there that 'to' belongs to any particular part of speech?

Reopen note:

There is a quite finite and modest amount of evidence in the literature about this issue, which members can record here as they see fit. Less than there is for example about what a noun is or what a verb is, or how to tell what the Subject of a sentence is. It's a simple question, that people need to be able to get answered on a site such as this. It is of continuing interest as more recent questions such as this attest to. The fact that people disagree about the outcome of the evidence does not mean that there is too much evidence to be documented. Quite the reverse. They disagree because there is relatively little evidence altogether. However, it is seriously interesting evidence.

The question

What part of speech is to as in:

• I need to know.
• To err is human, to forgive divine.
• What am I to do?

This question is not really about the difference in meaning between the examples. It is a question about what grammatical or syntactic reasons we have to classify this item as a particular part of speech. If arguing that it is 'its own part of speech', I'd like to know what grammatical - but not historical - reasons you have for your claim.

Just to clarify, I'm not asking about what assignment any particular authority or publication gives it. I'm interested to know the reasons why. Your own personal insights are also very welcome.

If you're a grammar anorak - like me - then you may want a list of the contenders out there from other grammarians. In no particular order:

• preposition
• subordinator
• non-finite auxiliary verb
• modal verb
• inflectional element

Can you help?

You can rather easily argue that to belongs to its own part of speech. The only evidence you need is that it does not fit into any of the major part of speech categories. You will just construct grammatical tests showing that it is not a noun, not a verb, not a preposition, etc. Following examples suggest that to is not a verb or noun, and that other prepositions cannot precede verbs in their base form (respectively).

*Nobody shall to.
*The to is ready.
*It is hard to see with run.

Of course, you can go on with this type of argument with pretty much any high frequency word until you have so many parts of speech that the concept is no longer useful. Of possible interest is Croft's discussion of Word Classes in Radical Construction Grammar.

One option that doesn't appear on your list is that it isn't even a word. Perhaps it's merely an open form bound morpheme.

I do see some evidence against its being an affix. If it is an affix, it's the only affix I know with an open form. Other affixes have closed or hyphenated forms. If it is an affix, it's the only affix I know that is occasionally suppressed. For example, except for the to, "I'll help you to do it" and "I'll help you do it" are identical in grammar and meaning.

I also see some evidence for its being an affix. It only appears at the beginning of an infinitive. The to marks the infinitive form of a verb in much the same way as the -ing marks the gerund. The infinitive form is created in other languages by using affixes.

It seems much more likely to me that the infinitive marker is a slightly odd example of inflectional morphemes than that it is the sole example of an entire category of grammatical word function. That may be slim and subjective, but I count that as evidence against its being a full-fledged word.