Re: Which part of speech?

Ray responding to Chris:
<<
In short, no arguments so far presented dissuade me from the notion that
'today', 'tomorrow', 'yesterday', 'tonight' may function in _English_
either as nouns or as adverbs. As for 'last night' in "I was reading the
CONLANG list last night.", my empirical approach says just this:
- 'night' is a noun
- 'last night' is a noun phrase with the adjective 'last' qualifying the
noun 'night'
- the phrase 'last night' in "I was reading the CONLANG list last night."
functions as an adverb (of time).
>>
I agree completely.

One thing that the whole discussion might have brought to light is
the fact that syntacticians and conlangers are often (not necessarily,
but often) working at cross purposes.  Quoting Ray who was quoting
Gregory:

<<
> I think the
> best solution is to have "today", etc. as nouns, with qualitative forms
> derived from them.
I think that may well be the best solution in your conlang.
>>
That's the real point.  No matter what any syntactic theory calls "today"
in English, the fact remains that the distribution of these guys is
different, and that different languages can deal with them differently.
Simply calling something like "today" an NP or PP with a null P is
missing the point from a conlanging point of view.  To put it in
simplistic terms, it's better for a conlanger to see things as different,
whereas it's better for a syntactician to see things as the same.

Now let me qualify that simplistic statement.  I can't access
Chris's original reply to my reply, but he made the point that no
syntactic theory accounts for everything (which is true), and
that a syntactic theory not account for everything is not
unforgiveable.  This is also true--from the point of a syntactician.
For a conlanger, though, there aren't going to be areas of a
language which are going to be leveled because they don't fit
with a syntactic theory.  The solution needs to make sense
from the standpoint of the language itself.  If a particular
syntactic theory can predict that, great.  If it can't, that's not a
problem for the language creator.  It's also not a problem
for the syntactician, from this standpoint.

Now, instead of thinking about creating a language to which
a syntactic theory is applied (which would be no different
than applying a syntactic theory to a natural language), think
about creating a language *from* a syntactic theory.  Everything
that that theory can't account for, or for which it has a poor
account, will be reflected in the language.  Additionally, every
assumption it makes will also be made in the language.  This
is a problem when working with something like Chomskyan
syntax, which is notoriously Anglocentric.  Different people
have applied Minimalism to non-English languages (for a
conlang-related example, see Matt Pearson's work on
Malagasy, which, I'm afraid to say, I couldn't make heads or
tails of), using that framework as an example, but changes
need to be made to it, which are not obvious from the standpoint
of the theory itself.  Occasionally these changes morph into
new theories, and the set of assumptions changes.

The main point, I think, is this: Blind assumptions about the
way language works in general are not inherently useful to a
language creator.  Every syntactic theory makes assumptions
about the way language works.  So when examining natural
language data, it's not useful to say what a given syntactic
theory says about that data.  Rather, distributional, descriptive
evidence is most useful, so that the conlanger can take away
whatever generalizations they formulate on their own, without
having to accept both a syntactic analysis and a syntactic
framework.

For that reason, I think Ray's answer, which I quoted at the
beginning, is the most useful.  Besides, while one syntactic
analysis may seem right to some, it may seem really, *really*
wrong to others, in which case the questioner is left with an
argument about abstractions, rather than descriptive
generalizations (note: *generalizations* not, rules) derived from
actual language data.

-David

On 5/11/05, Ray Brown <ray.brown@...> wrote:
> On Tuesday, May 10, 2005, at 10:55 , Gregory Gadow wrote:
> [snip]
> > Ray, Mark, Christopher, thanks for the insights.
>
> You're welcome   :)
>
> > The question came up with me trying to decide how these and other time
> > words would be classed in my conlang. There are some strict rules in how
> > nouns and qualitatives (a mostly interchangeable class of adjectives and
> > adverbs) can be used, so it's not really a trivial decision.
>
> Indeed not - see the continuing discussion below.
>
> > I think the
> > best solution is to have "today", etc. as nouns, with qualitative forms
> > derived from them.
>
> I think that may well be the best solution in your conlang.
> ==================================================
>
> On Wednesday, May 11, 2005, at 03:20 , David J. Peterson wrote:
> [concerning "last night" in: "I was reading the CONLANG list last night."]
>
> > Damian wrote:
> > <<
> >> So then how do you parse "last night" as an adverb?  Wouldn't
> >> it just be simpler (in the Occam's Razor sense) to assign a
> >> zero-derived case to every noun naming a day?
> > >>
> >
> > Simpler in what sense?  It would actually seem needlessly complicated
> > to me.
>
> Yes, I agree. English nouns simply do not have case markings (the
> possessive 's is a clitic which attaches to whole phrases). To have
> zero-derived cases then poses the question of how many zero-derived cases.
>   Does the sentence "John gave Mary a kiss" have zero-marked nominative,
> dative and accusative cases?
>
> I agree with David - this approach IMO makes things more complicated.
>
> > Chris wrote:
> > <<
> > Zero-derived case or null preposition--it depends on your theory of case
> > assignment, mostly. Minimalism prefers the latter,
> .....
> > >>
>
> Yes, a zero preposition could, I guess, be argued. David gives several
> pairs of sentences showing how NPs are used adverbially when they are the
> objects of prepositions. Long years ago (the 1950s) in the terminology
> used in the UK, we called such phrases 'adverbial phrases', naming them by
> their function; today, they are normally called 'prepositional phrases'.*
>
> (*In our 1950s terminology, 'prepositional phrase' meant a phrase that
> functioned as a preposition, such "in accordance with"; but these are
> generally called 'complex prepositions' nowadays.)
>
> Another argument might be that 'night' cannot be used adverbial by itself.
>   We have to say things like 'by night' or 'at night', cf:
> "Owls normally hunt by night".
>
> Therefore we could, I guess, say that when 'night' is qualified by an
> adjective, the preposition is may be omitted (or 'understood'), i.e.
> [during] last night.
>
> So if PPs function as adverbs (they may also function as adjectives "the
> man _in the moon_"), then it could be argued that "last night" in the "I
> was reading the CONLANG list last night." has a zero preposition. But then
> that would make "last night" a PP - umm....
>
> > So while positing a null preposition does seem "attractive", it doesn't
> > seem to account for the data.
>
> I agree.
> ===============================================
>
> On Wednesday, May 11, 2005, at 02:11 , Christopher Wright wrote:
>
> [snip]
> >
> > But yes. In English, these are nouns; in Latin, they're adverbs.
>
> Not correct. While _hodie_ (today), _heri_ (yesterday) and _cras_
> (tomorrow) are certainly adverbs in Latin, _nox_ (genitive: noctis)
> certainly is not. It is a noun.
>
> The Latin for "I was reading the CONLANG list last night." is:
> _proximo nocte_ tabulam Conlangicam legebam.
>
> 'proximo nocte' is the ablative case which, among other uses, is used for
> "time when".
>
> In fact, comparing or contrasting '(last) night' with 'today' & 'tomorrow'
>   is a bit misleading. The word that should be considered is _tonight_ -
> and that, according to the dictionaries, is:
> - a noun meaning 'this night: the night of the present day';
> - an adverb meaning 'on this night: on the night of the present day.'
>
> > There's
> > nothing that says a [semantic] word must have the same part of speech
> > assigned to it in every language. Right?
>
> Obviously not, and if you read my previous email carefully you will see
> that I agree. For while I argued (and still do) that _today_, _tomorrow_
> and _yesterday_ may function either as a noun or as an adverb in English,
> in Latin _hodie_, _cras_ and _heri_ may function *ONLY as adverbs*.
>
> BUT -
> There is nothing that says a [semantic] word has only one part of speech
> assigned to it in English. Right?
>
> Certainly _night_ is a noun in English. Whether in "He's working on the
> night shift" it is an adjective or an 'epithet noun' is arguable. However
> in the phrase "last night", the word night itself is a noun as is shown by
> its being qualified by the adjective 'last'. But the _phrase_ 'last night'
>   functions *adverbially* in "I was reading the CONLANG list last night."
>
> It is not necessary for any word in the phrase to be an adverb for the
> whole phrase to function as an adverb - such phrases used to be called
> _exocentric_ (The X-bar system - at least in most versions - prohibits the
> existence of exocentric constituents, and the use of the term is now
> somewhat debatable).
>
> In a language like English (or modern Chinese) which is largely isolating
> with a few grammatical inflexions, IMO the maxim of my English teacher way
> back in the 1950s still holds good:
> "By their deeds shall ye know them"
>
> In short, no arguments so far presented dissuade me from the notion that
> 'today', 'tomorrow', 'yesterday', 'tonight' may function in _English_
> either as nouns or as adverbs. As for 'last night' in "I was reading the
> CONLANG list last night.", my empirical approach says just this:
> - 'night' is a noun
> - 'last night' is a noun phrase with the adjective 'last' qualifying the
> noun 'night'
> - the phrase 'last night' in "I was reading the CONLANG list last night."
> functions as an adverb (of time).
>
> Ray
> ===============================================
> http://home.freeuk.com/ray.brown
> ray.brown@freeuk.com
> ===============================================
> Anything is possible in the fabulous Celtic twilight,
> which is not so much a twilight of the gods
> as of the reason."      [JRRT, "English and Welsh" ]
>