[Lingtyp] nominal classification (gender and classifiers)

Thu Mar 30 13:15:40 UTC 2017

Dear all,

I'd like to support Walter's point (as I understand it). To me, it makes
sense to foreground agreement as the notion to which the definition of
gender (and distringuishing gender, or noun classes, from classifers)
should refer, and to consider rigid assignment as its corollary.

Indeed, consider two properties commonly observed in what typologists
usually call gender. This are (a)  being involved into marking syntactic
cohesion (what Walter meant by "indicator of syntactic units") and rigid
assignment, as opposed to sets of classifiers, sometimes (or even often)
assined on flexible, referent-based grounds. These two properties are
sometimes in contradiction, but they seem to be connected in general in the
following way: Whenever you are an agreement device, you tend to be rigid.

Grev and Sebastian suggest that canonical gender is gender based on rigid
assignment. They make rigid assignment the central part of the definition
of gender. An alternative solution is to say, together with Walter (I
think) that canonical gender is involved in agreement*. Which is best?

As hinted already in Walter's message, we can explain why you need to be
rigid if you are an agreement device. If you are too flexible, you would be
of less help in establishing syntactic units. Being an agreement device
explain why you need to be relatively rigid. The category of the controller
has to be continuously activated through processing of a syntactic unit. To
make it easier, I think it makes all sense to store the gender category of
the controller in the lexicon. Otherwise, to track syntactic relations, you
should store both the contrller noun and its category in the context
(reference-based, non-rigid category). If this is a rigid cateogorization
device, you only have to store the controller noun.

As it often happens, the reality is not always (or never) fully compatible
with our expectations. In Russian, there is a vast class of bi-gender
(human) nouns. In Archi, there are referent (and discourse) based gender
assignment phenomena, the most productive (but still rare) being putting
the same lexical item into Class (Gender) 3 or Class 4 depending on the
size of the referent. Still, on the whole, gender assignment is rigid, and
gender serves as a good device for agreement as syntactic relations
tracking device. Both languages are considered to be typical gender
languages. Incidentally, both exibit quite rich and complex system of
agreement.

Looing at it from the opposite persoective, I see no way to explain why
rigid assignment should entail agreement.

I am unaware of any language that has a reference-based (ovrewhelmingly
non-rigid) noun categorization device used for what is an obvious
typological instance of agreement. If you are aware of such example, it
would be interesting to know more about it. On the other hand, Walter's
message indicates that there are consistently rigid systems that are not
used for agreement - his Thai example.

I know much less about classifiers - I never worked on a classifier
language. I agree that the mere number of categories should not be a
criterion. Classification of classifier systems into more and less rigid
(as entailed from the approach in Walter's message and above) is absolutely
necessary; this seems to be an important parameter of variation in what I
would call classifier systems. But, unlike the case of gender, the
orientedness towards referents rather than nouns does not seem to be
motivated by whatever I know (second-hand) about the function of numeral or
possessive classifiers. On the other hand, as explained above, I think
that, in defining gender, it makes all sense to refer to involvement in
agreement as the main property and rigid assignment as its corollary.

Finally, Martin indicated that the non-rigidness of gender in pronouns
remains unexplained in rigid vs. flexible approch to the distinction
between gender and classifier systems. This is one point in which approach
to gender as outlined by Grev and Sebastian seems to fail
cross-linguistically and in a rather consistent way. Eva explicitly but
arbitrarily excluded pronouns from her statement. Why should personal
pronoun be treated differentle in so many languages? I think, very
tentatively, that this is connected to their indexality. The presence of
the speaker and the addressee themselves support agreement category
tracking. This is why the strong requirement of the rigid assignment,
motivated by the needs of syntactic mechanisms of agreement, is very
typically relaxed with first and second person pronouns. But this is
probably a very vague consideration, with further (probably experimental?)
support required.

What I said above suggests that we know what agreement is. Probably, there
will be few typologists who would fully agree on the definition of
agreement in every detail, and even fewer would agree on treatment of
specific (marginal) cases. But this is a very common situation in typology
- a fuzzy definition is based on another hotly debated notion. Still, as
agreement is a phenomenon more general than gender. You can agree not only
in gender but also in other categories. So it does not seems circular to
put the discussion about agreement outside the discussion of what gender
is, and to refer back to it from the discussion of gender. At least, this
is a theoretically feasible approach.

Incidentally, Grev and Sebastian's approach needs to establish some kind of
treshold in terms of rigid / flexible opposition - you cross it and you are
closer to classifier system. I think it will have all kinds of problems.
Dear Grev and Sebastian, you mentioned ongoing research - I'd love to have
a look.

Michael Daniel

2017-03-24 18:47 GMT+03:00 Bisang, Prof. Dr. Walter <wbisang at uni-mainz.de>:

> I couldn't agree more about nominal classification being an interesting
> topic.
>
>
> But I am not so sure about the role of flexibility (see Grev Corbett's
> message):
>
> <<< We would argue that the canonical gender system has rigid assignment.
> >>>
>
>
> This would mean that Thai has a canonical gender system and that an
> example like the following (see my previous message) is similar to Swahili:
>
>
> rót  [khan  yàj]  [khan  níi]
>
> car  CL       big     CL       DEM
>
> ‘this big car’
>
>
> To me the most important distinction is rather that a gender system does
> not add information on grammatical function. In the above Thai example, the
> classifier marks contrastive focus with stative verbs and (mostly) singular
> with the demonstrative. Moreover it is not obligatory (in contrast to its
> occurrence with numerals).
>
> In contrast, a Swahili class marker on its modifiers does not add any new
> information on grammatical function. In my view, this is due to the
> agreement status of gender/noun-class markers. If one sees agreement as an
> indicator of syntactic units, the semantics of their markers become
> irrelevant. What matters is that the markers do their job reilably, i.e.
> they are obligatory and they lack flexibility. This function can also be
> exhibited by markers of other categories (e.g. number, case, ...) as long
> as these markers are semantically general enough to be compatible with the
> meaning of any of the potential modifiers (see Bisang 2002). Let me add
> that  it was probably not such a good idea to abandon agreement after all.
>
>
> As for arbitrary restrictions on the number of distinctions, I fully agree
> with Grev. We may lose important information.
>
>
> All the best,
>
> Walter
>
>
>
> ------------------------------
> *From:* Lingtyp <lingtyp-bounces at listserv.linguistlist.org> on behalf of
> g.corbett at surrey.ac.uk <g.corbett at surrey.ac.uk>
> *Sent:* Friday, March 24, 2017 3:43:32 PM
> *To:* lingtyp at listserv.linguistlist.org
> *Subject:* Re: [Lingtyp] nominal classification (gender and classifiers)
>
>
> Good to see so much interest in nominal classification. Consider this
> example:
>
>
>
>  mi-*na*-si-na                            *na*-yu               *na*-manabweta
> vivila
>
> DEM-FEMALE-PL-DEM     FEMALE-two  FEMALE-beautiful    girl
>
> ‘these two beautiful girls’
>
>
>
> Easy! Plenty of agreement, obviously a gender system. We agree. But not so
> fast. Gunter Senft, THE expert on Kilivila, calls this a classifier system
> (the example is from Senft 1986: 69, that’s his Mouton Grammar). And there
> are reasons for that: for instance, there is flexibility in the choice of
> values.
>
>
>
> Nominal classification is complex. One difficulty is that it’s easy to get
> hung up on sterotypes (French gender, Mandarin classifiers) and we think
> it’s time to pull apart the different criteria in nominal classification
> systems. The particular clusterings that we find in French, Chichewa, Archi
> and Burmese, aren’t the whole story.
>
>
>
> Please let’s not define by the number of values: there are tiny classifier
> systems and large gender systems. (By the way, the Kilivila system has at
> least 177 members.) Biologists don’t say that legs must come in twos or
> fours, and bar millipedes from having legs because they have too many.
> Linguists allow for large tense systems and small consonant inventories.
>
>
>
> We would argue that the canonical gender system has rigid assignment.
> That’s a principled baseline to measure from. Then we find some that are
> much more flexible (like Mawng as mentioned by Ruth Singer, or Savosavo,
> Wegener 2012). And there are “classifier” systems where the speaker has
> little choice. Do check your favourite system using Gerald Gazdar’s orange
> test. If orange takes the ‘sphere’ classifier, step on one and then ask
> your speaker. You may be surprised to find it’s still classified as
> ‘sphere’, because the lexeme for orange takes that classifier (sounds like
> agreement?). Mike Franjieh’s work on relational classifiers in North Ambrym
> is significant here (see *Oceanic Linguistics* 2016).
>
>
>
> So nominal classification systems can be more or less close to canonical
> gender. Exotica are still turning up, even in Europe (see Paciaroni &
> Loporcaro’s recent work on Ripano). As we weaken the constraint of
> agreement (and Steele’s definition is still a helpful one), we get to the
> exotic system of Dutch (see Jenny Audring on this) and to noun classifiers
> (which Nick Reid has shown nicely grammaticalizing into genders in
> Ngan’gityemerri). The markers may be more bound (as in French and Kilivila)
> or less bound (as in Ngan’gityemerri). There is a wonderfully rich variety
> of systems out there, since the different criteria vary independently of
> each other (but not absolutely so, of course, or where would we typologists
> be?). And in many respects, the different types of classifier are more
> different from each other than they are from canonical gender.
>
>
>
> The issues have occupied us (Sebastian Fedden and Greville Corbett) for
> quite some while. If anyone has time to comment, we’ll gladly send you a
> paper or two. It’s work in progress.
>
> Very best
> Grev and Sebastian
>
>
>
> On 24 Mar 2017, at 08:36, Martin Haspelmath <haspelmath at shh.mpg.de> wrote:
>
> On 23.03.17 19:21, Alan Rumsey wrote:
>
> Those of us who have worked on languages with 2-5 such classes (in my case
> Ungarinyin) have sometimes called them ‘genders’, while those who have
> worked on languages with more have called them ‘noun classes’.
>
>
> I had presupposed in my earlier messages that there is no distinction
> between these two types, and that they should be called "genders" – I took
> this as established by Corbett (1991). As Johanna Nichols noted, the term
> "noun class" is vague, so for cross-linguistic purposes, "gender" is surely
> better.
>
> (One might feel that neglecting the sex-based vs. non-sex-based
> distinction is not such a good idea, as in Bernhard Wälchli's message, but
> it seems to me that one really shouldn't use the term "gender" anymore for
> sex-based distinctions, at least in typology. I take Corbett (1991) as
> foundational for all of us.)
>
> But the problems with Corbett (1991) are
>
> – that his definition of gender is based on the notion of "agreement" (for
> which there is no clear definition, cf. Corbett (2006), who only provides a
> definition of canonical agreement)
>
> – that the distinction between "gender" and "numeral classifier" is (in
> part) based on the idea that gender markers are affixes and numeral
> classifiers are free forms, but there is no clear definition of "affix"
> (there is a definition of "free form", as occurring on its own in a
> complete utterance – and numeral classifiers are surely bound by this
> criterion)
>
> – that the distinction between "features" (like gender) and markers (like
> classifiers) is far from clear-cut
>
> Moreover, Corbett himself has given up the distinction between gender and
> other classifiers (there's only a canonical definition of gender now), as
> have others such as Ruth Singer, Sasha Aikhenvald, and Frank Seifart. But I
> still want to talk about "gender" as a comparative concept (as well as
> about "numeral classifiers" – a student of mine just wrote a nice MA thesis
> about this topic).
>
> Guillaume Segerer points out that some Atlantic languages have up to 31
> classes, and it would seem odd to exclude them from having gender on the
> basis of a definition that arbitrarily stops at 20. I agree that this would
> seem odd, but I need to point out that *it wouldn't matter*. Comparative
> concepts are not designed to give the same results in all cases that seem
> similar enough to us (or some of us), but *to allow rigorous,
> intersubjective cross-linguistic comparison*. Comparative concepts must
> sometimes be arbitrary, because the world consists of many continuities,
> and if we still want to discuss differences with words, we need to make
> arbitrary cuts (think of the importance of SMEs in economics – small and
> medium-size enterprises, defined arbitrarily as having fewer than 250
> employees).
>
> Maybe it will turn out that some other, less arbitrary concept will give
> even better cross-linguistic generalizations. But for the time being, we
> have the term "gender" as a comparative concept (especially in legacy works
> such as Corbett's WALS maps), and my definition ("A *gender system* (= a
> system of gender markers) is a system of genifiers which includes no more
> than 20 genifiers and which is not restricted to numeral modifiers") seems
> to be the only definitional proposal currently available.
>
> Best wishes,
> Martin
>
>
> —
> Greville G. Corbett
>
> Surrey Morphology Group
> English (I1)
> Faculty of Arts and Social Sciences
> University of Surrey
> Guildford
> Surrey, GU2 7XH
> Great Britain
> email: g.corbett at surrey.ac.uk
> www.smg.surrey.ac.uk
>
>
> *Features*
> Available now through all good bookshops,
> or direct from Cambridge University Press at: www.cambridge.org/
> 9781107661080
>
> <http://www.cambridge.org/gb/academic/subjects/languages-linguistics/grammar-and-syntax/features>
> *Canonical Morphology and Syntax.* Also available through all
> good bookshops, or direct from Oxford University Press at:
> http://ukcatalogue.oup.com/product/9780199604326.do
>
>
>
>
> _______________________________________________
> Lingtyp mailing list
> Lingtyp at listserv.linguistlist.org
> http://listserv.linguistlist.org/mailman/listinfo/lingtyp
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20170330/ed806570/attachment.htm>