Not seldom, the same source feeds into two or even more different grammaticalization paths. This may produce the synchronic presence, at a later stage of the language, of a set of grammatical formatives which are either homonymous or similar in their significans and have different functions which sometimes may be related only diachronically. Such a branching development of a source is called polygrammaticalization.

While the example of the Chinese perfective particle may be conceived as one of synchronic polysemy, polygrammaticalization may also lead to homonymy. Mandarin zài (be.LOC) 'be [in a place]' functions both as a local preposition () and as a progressive auxiliary ().

Latin habere ‘have’ has already been seen as an auxiliary in the Portuguese perfect. As a matter of fact, this is a rather extreme case of polygrammaticalization. Variants of Port. haver show up in at least four constructions:

1. In the combination [ V-INF haver ], lit.: ‘to have to V’, haver becomes enclitic to V-INF and starts indicating future tense, thus ‘I will V’, as in the reflexive form of .

   .	arrepender	-me	-ei
   Port	repent:INF	-1.SG.ACC	-FUT.1.SG
   	I will repent

2. In the combination [ haver V-PTCP.PRF ], lit. ‘have Ved’, haver functions as the perfect auxiliary, as in hei envelhecido (have.1.SG grow.old:PTCP.PRF:M) ‘I have grown old’.
3. In the combination [ haver de V-INF ], lit. ‘have of V’, haver is a debitive auxiliary; thus the construction means ‘have to V’, as in .

   .	hei	de	ir
   Port	have.1.SG	of	go.INF
   	I have to go

4. In the impersonal combination [ ha N ], lit. ‘it has N’, haver is the existential verb, so the construction means ‘there is N’, as in ha diferências ‘there are differences’.

It is important to note that all of these four constructions in which haver plays a pivotal role are structurally different. Only #4 comprises an NP in complement function. The other constructions have a non-finite verb form depend on the auxiliary. This is an infinitive in #1, a participle in #2 and a prepositional phrase with an infinitive as its lexical head in #3.

Moreover, it is not the case that haver was recruited for all of these grammaticalization paths at the same time. #1 originates already in Vulgar Latin. The same goes for #2, only a few centuries later. #4 occurred at some time in western Romance, and #3 is an Ibero-Romance development. This is obviously a renewal of #1 as this forfeited its initial debitive meaning when grammaticalized to a future marker. Finally, it is worth noting that Portuguese has meanwhile replaced the verb haver by ter in #2 and #3, and replaced it in #1 by ‘go’ in an original motion-cum-purpose construction.

For the sake of completeness, it may be noted that most of these developments occurred in all of Romance. Very similar examples could be adduced – and have been repeated ad nauseam in the literature since Schlegel 1818 – from Spanish, French and Italian.

Of course, the case of Engl. have is similar in many respects. The verb is used, among other things, in the perfect tense, but also in the debitive construction ‘X has to V’. As may be seen, have produces very different construction meanings when combined with a perfect participle and when combined with an infinitive.

Two aspects of polygrammaticalization are noteworthy:

• If only the original lexeme is considered, a single source appears to be at the origin of the branching evolution. However, this lexeme is used in different constructions. These are responsible for the divergence. Thus, polygrammaticalization is convincing evidence for the recognition that it is not the meaning of a lexical item alone which determines the target of its grammaticalization, but instead the construction in which it is used.
• In such cases, the meaning of the source item has low specificity from the outset. This is what renders it usable in different constructions and allows for the development of resulting constructions in which the proper meaning of the formative in question plays a minor role if compared with the holistic meaning of the construction.