Question on agentive nominalizations

[Nicholas Ostler]

Well, Sanskrit as ever provides counter-examples to *(4) - if we accept 
(as ever) that it can be counted among the natural languages of the 
world. Here the agentive is signified with a formative in -tr.-, 
nominative -ta: E.g. (examples from Speijer 1886 - Sanskrit Syntax, p. 40):

Pancatantra iii.71:
narapatir neta: prajña:s
kingNOM.SG leaderNOM.SG subjectACC.PL
king leader of his subjects

Daçakuma:racarita 199:
sambha:vayita: budha:n prabha:vayita: sevaka:n ud.a:vayita: bandhu:n 
nyagbha:vayita: çatru:n
honourerNOM wiseACC.PL promotorNOM servants ACC.PL raiserNOM 
kinsmenACC.PL lowererNOM enemiesACC.PL
honourer of the wise, promotor of servants, raiser of kinsmen, 
bringer-low of enemies

There are some complications. There are two possible accentuations of 
these agentives, oxytone and barytone. According to Panini [3.2.135] 
only the barytone have this construction (otherwise genitive being 
required to mark object dependents). The particular sense of barytone 
agentives is 'lasting and inherent qualities'.

Speijer comments (p. 40):
The acc. with the barytona in -tr.-, though not rare in the earlier 
period, seems to protract but an artificial life in classical Sanskrit 
[from which the two above examples come - NDMO], as it is met with only 
in the refined style and even there side by side with the genitive... On 
the other hand, the examples given by Ka:çika: on Panini 3.2.135 prove 
that, at the time, they were applied at first, the construction with the 
acc. was obvious and natural.

These examples include
karta: kat.am
makerNOM matACC.SG
'(skilful) maker of mats'

Complicating the picture, the ending -tr.- is also used with another 
meaning, viz periphrastic future, e.g.
Malavagnimitra i, p. 15
mukta: ma:ghavasenam tatah. 'ham
freeAGENT.NOM MaghACC then I.NOM
then I shall free Maghavasena
and there behaves like a regular finite verb. Formally it seems to be on 
its way to reanalysis as quite  separate from the noun - since in the 
1st and 2nd person (plural and dual), and in the feminine singular too,  
the MASC.NOM.SG-looking -ta: is still used, instead of corresponding 
dual and plural (or fem). Nevetherless, masculines do select agreeing 
forms of the form, as if it were an agentive noun:  -ta:rau (Dual), 
-ta:rah. (Plural)

We are told (Speijer p. 259) that
"the tense in ta: cannot be used of every future, but only of such 
actions as will not occur soon... it is therefore a remote future." So 
it is not a marginal pattern, but one well established in the literature 
and so interpreted.

In fact, many other deverbal nominals in Sanskrit also allow accusative 
direct objects: e.g.
desideratives in -u-, -ishlu-
agentives in -aka-
some in -in (e.g. çatam da:yi: 'owing 100')

These are all totally distinct from the present participle (formed in 
-ant-), by the way.

So a student of Sanskrit would not naturally come to think of your 
generalization as valid, Mark.

Best

Nicholas

Mark Baker wrote:
> Dear Typologists:
>
> I have a question that perhaps people out there can help me with.  Two of
> the more common kinds of deverbal nominalization are event/action-denoting
> nominals like (1) in English, and agent-denoting nominals like (2) in
> English:
>
> (1)	The finding of the wallet (?so quickly) [was a relief.]
> (2)	The finder of the wallet (*so quickly)  [received a reward.]
>
> These can be reasonably similar in their gross syntax, apart from the
> difference in meaning, as in the English case.  Now alongside (1) is the
> gerundive nominal in (3), which is another way of denoting the
> action/event.
>
> (3)	Finding the wallet (so quickly) [was a relief.]
>
> (3) is semantically similar to (1), and like (1) it acts like a noun
> phrase in the larger clause, but its internal syntax has many verbal
> features: the object of “find” is an unmarked accusative NP, adverbial
> modifiers are possible, the definite article is not used, etc.  The
> literature contains many discussions that compare and contrast examples
> like (1) and (3) in various languages.
> 	What I am interested in is the fact that there seems to be no more verbal
> construction that denotes an agent.  For example, in English there is
> nothing like (4), which would be parallel to (2) in much the way that (3)
> is parallel to (1) [but see qualification below].
>
> (4)	*(The) finder the wallet so quickly [received a reward].
>
> I have good evidence that there is nothing like (4) in three unrelated
> languages I am studying (English, Sakha, Mapudungun) and I don’t recall
> seeing examples like (4) in my general reading.  I am thus interested in
> the possibility that examples like (4) are impossible universally. 
> However, I am finding it a bit hard to evaluate this by looking at
> standard grammars, since many of these simply list a few examples of
> agentive nominalizations in isolation, without saying one way or another
> if they combine with direct objects, adverbs, etc.  Can anyone point me to
> relevant cases, pro or con, that could be worth looking into?  Any
> references to general/theoretical discussions of this pattern would be
> most welcome too.
>
> Thank you!
> Mark Baker
>
> Qualification: One challenge in evaluating this hypothesis that I am aware
> of is the need to distinguish true agentive nominalizations from active
> participles that are used as headless subject relatives—e.g. constructions
> like “The one who is seeking the wallet anxiously”, where “one who is” may
> be null, leaving only “the - seeking the wallet anxiously”.  I know how to
> tell the difference between active participles and true agentive
> nominalizers in the languages I know, but this might not be
> straightforward in other languages.  I don’t necessarily expect people to
> sort this out for me before mentioning possible cases.
>
>   

-- 
Nicholas Ostler
Chairman, Foundation for Endangered Languages
Registered Charity: England & Wales 1070616
172 Bailbrook Lane, Bath BA1 7AA, England
Phone: +44 (0)1225-852865 Mobile: (0)7720-889319
www.ogmios.org
nostler at chibcha.demon.co.uk