Paul Roser said:
> Nordlinger and Sadler distinguish two types of nominal tense
> marking, 'independent nominal TAM' and 'propositional nominal TAM', in the
> former specifying TAM (tense/aspect/mood) information intrinsic to the
> nominal itself - eg. 'my former wife', 'the ex-president'

Assuming that the constituents being marked with 'independent nominal TAM'
really have to be considered nouns, I wonder what they think can be gained
from calling it TAM. Not everything that carries a TAMesque meaning is a
TAM marker. In English, 'today' is not a marker for hodiernal tense: it's
just an adverb, and I don't see how anything can be gained from calling it
a tense marker. Similarly, nominal morphology that has to do with
temporal/eventive or pragmatic relations isn't TAM.

The only instance of morphological TAM *marking* on nouns that I would be
prepared to accept would be when a noun, not the verb, carries the TAM
marking for a clause. I've never encountered a language that does this,
and none of the compendious catalogues of grammatical structure (written
by groups of descriptive linguists who've *really* been around...) include
the possibility IIRC, so I would be surprised if I did encounter this
phenomenon. (Although situations with clitics are surely common enough --
as in the case of English.)

That said, I think it's an *awesome* idea for a conlang. It's different
enough from the way natlangs work to be intriguing, while not so different
that it would prevent usage (unlike, say, unrestricted center embedding).
Lots of natlangs have clause-level markers on nouns, after all -- but they
tend to be involved with valence assignment and/or pragmatic functions
that are more-or-less intimately tied up with the noun being marked.

Semantically, putting TAM on or near the verb falls out very naturally, as
it happens. If 'to eat' is a class of events, then 'to start and finish
eating at some time prior to the time of utterance' is a subclass of that
class of events. It's not a subclass of any of the other parts of the

> A particularly nice example of the former is provided for Iate, a
> Brazilian
> language:
> -Realis-
> present  'that which is a house/serving as a house'
> past     'that which was once a house/stopped being a house'
> future   'that which will be a house/house which is being built'
> -Possible(Irrealis)-
> present  'a possible house/something which has the possibility of
>           being a house'
> past     'something which would have been a house but wasn't/had the
>           possibility of being a house'

Assuming that the roots or stems here have to be considered nouns (which
may or may not be the case), what we have is temporal relations being
encoded on nouns. Assuming that these temporal relations are *not* the
ones pertaining to the clause (and presumably encoded morphologically on
verbs or periphrastically somewhere in the clause), they're not TAM:
they're just temporal relations encoded on a noun.

The same applies to the Somali example:

> the current house vs the past house/former house).

> N&S report that in some languages the number of tenses are essentially the
> same for both nominals and predicates, which would go along with (3) above
> in languages where the distinction between nouns and verbs (Austronesian,
> Salish, etc) is tenuous,

The languages I've seen that have tense spreading or more limited
agreement have always had similar markers on all constituents (modulo some
morphophonological reshaping, as I've understood it). They definitely
encoded the same semantic values in each place, even if there was some
variation on the surface. I remember seeing data for a handful of
Australian languages, but I don't remember their names.

> but there are also cases (I believe Tariana falls
> into this group) where the tenses for nominals form a simpler system than
> for true predicates (as exemplified in the outline I sent to the list
> previously, Tariana nominals have decidedly different though equally
> complex morphology when compared to the predicates).

Right. That's because the "tenses" for nominals aren't tenses. They're
morphologically encoded temporal relations. They're no more an expression
of tense than is the adjective 'former' in English -- or should we be
describing English noun phrases in terms of periphrastic tense categories?

-- Mark