I can think of one instance of this in actual languages, and one in a well-known conlang.
1. Irish (dialectally)
Modern Irish is in the process of transitioning synthetic to analytical verbal morphology. Old Irish was fully synthetic, with each verbal form always showing its subject, similar to Romance languages; but over time, this system has collapsed. In the standard language, there are now only synthetic forms for the first person singular and plural in most tenses (plus second person singular in the conditional and past habitual); the rest use an unchanging base form + explicit subject (pronoun or noun phrase). Using the present tense of beir ‘carry, bear’ as an example, these are the only forms found in Standard Irish:
This process is not uniform throughout the dialects, however. In some dialects, it’s gone further (with analytic forms like beireann mé and beireann muid), while in others, it’s not progressed this far and is still optional.
Particularly in the southwest of the country (Munster Irish), there is much variation; you’ll hear fusional and analytic forms used side by side, with little or no difference in meaning. The third singular has no synthetic form anywhere, and the second plural only in a few tenses, like the simple past which has synthetic forms for all but third singular. Possible Munster past-tense forms of beir (suppletive) are then:
||rugas / rug mé
||rugamair / rug muid
||rugais / rug tú
||rugabhair / rug sibh
||rugadar / rug siad
This shows a clear pattern of using an unchanging form rug (traditionally called the ‘bound’ form) with overt subjects, but various different forms with implicit subjects.
With noun phrases, the bound form is always used:
Rug an páiste barróg ar a mháthair
‘the child gave his/her mother a hug’
Rug na páistí barróg ar a máthair
‘the children gave their mother a hug’
Forms like *rugadar na páistí do not occur.
2. Quenya (conlang)
If you’ll accept non-natural languages, a better case would be J.R.R. Tolkien’s conlang Quenya.
The general system of conjugation in Quenya is that the ‘verbal endings’ that show person and number are really just clitic and optionally shortened forms of the personal pronouns. They are thus not really verbal endings at all—they can also be added to other things, like prepositions. They have standalone forms (with optional lengthening of the vowel when stressed) and clitic forms (long and short); nominative/accusative is not distinguished.
To give a few pronouns to work with (listing standalone, stressed standalone, long clitic, short clitic):
- 1sg.: ni, ní, -nyë, -n ‘I/me’
- 2sg. informal: lyë, lyé, -lyë, -l ‘thou/thee’
- 3pl. (animate/inanimate): te/tai, té/tai, -ltë/-ntë, -lt/-nt ‘they’
The standalone forms act as nouns and can take cases, while the clitic forms can be suffixed to words like verbs and prepositions, but take no clitics themselves:
- ni + -n (dative) → nin ‘to me’
- mi ‘in’ (prep.) + -nyë → minyë ‘in me’
- tul-i- ‘come’ (vb., aorist stem) + -n(yë) → tulin or tulinyë ‘I come’ (aorist)
When a standalone pronoun is used as the subject of a verb, however, the verb takes no personal ending of its own: the pronoun is only expressed once. Instead, a simpler, endingless form of the verb is used, which only distinguishes number (the singular and plural forms are known; the dual form is not known). The singular is the bare inflectional stem, and the plural is formed by adding -r.1
- ní + tul-i- → ní tulë2
- té + tul-i- + -r → té tulir
Quenya also has specifically emphatic pronouns that presumably really emphasise the subject. These are made up of the stem e- with the (non-shortened) clitic form of the pronoun suffixed; so the 1sg. form is enyë. These are also used with the endingless verb form, as can be seen in this direct quote from the song Namárië (‘Farewell’) from The Lord of the Rings, which uses the 2sg. future (-uva-) of the verb hir- ‘find’:
Namárië! Nai hiruvalyë Valimar.
Nai elyë hiruva. Namárië!
Farewell! Perhaps thou shalt find Valimar.
Perhaps even thou shalt find it. Farewell!
There are not many examples of verbs with actual third-person subject endings, but there are many with explicit subjects, and these all have the endingless forms as well: [laurië] lantar lassi (‘[like gold] fall the leaves’, aorist of lanta- ‘fall’), mornië utúlië (‘darkness has come’, perfect of tul- ‘come’), etc.
So even though I haven’t been able to find any sources that say this in so many words, I think it’s safe to say that the system of only adding personal pronoun clitics to the verb when there is no explicit subject holds when the subject is a noun phrase rather than a pronoun, too.
1 This endingless form is also used when negating verbs. Generally, Quenya is loosely based on Finnish, of which Tolkien was a great fan, and this is no exception: as in Finnish, negating a verb involves conjugating a ‘negative verb’ and leaving the main verb in an impersonal form that only distinguishes number (in some tenses); for example Finnish ole-n ‘I am’ vs. e-n ole ‘I am not’ and Quenya cára-n(yë) ‘I make’ vs. ua-n(yë) cára ‘I do not make’. In Finnish, however, an explicit subject does not cause the verb to be in the endingless form, so it is minä olen, not *minä ole.
2 Final -i regularly yields -ë in Quenya, so tulë is the expected outcome of *tuli. This historical sound change is also taken piecemeal from Finnish.