[Lexicog] Digest Number 1050

Tue Nov 18 03:08:51 UTC 2008

Wayne Leman wrote:

"Ron, are you searching for explanations other than cultural saliency?"

Yes. Cultural salience is just one factor. I would say cultural salience is
one aspect of what I call the universality of human experience. All
languages will have words for 'eat' and 'sleep' because all of us have to do
these things. (The exact meaning of 'eat' may vary slightly from language to
language, but will have the primary sense of 'ingest food through the
mouth'.) Some languages have specific areas of experience that are important
to them but are not universal. 'Domesticated elephants' is an important
domain in some areas of south Asia. You mention a few that are important for
the Cheyenne. So human experience results in a cline from the universal to
the culturally unique. For instance every culture builds houses. Even nomads
build temporary/transportable shelters. But igloos are unique (I think) to
Eskimo culture.

Another explanation is what I call semantic universals. Natural Semantic
Metalanguage (NSM) lists about 60 semantic primitives. If the NSM claim is
true, then all complex concepts are composed of simpler concepts. But I
believe there is more to it than this. (The proponents of NSM would probably
agree.) I would say that there are other fundamental semantic building
blocks (such as agency). One complicating factor is that all concepts seem
(to me) to be complex, even the NSM primitives. Some concepts appear to be
universal, others common, others rare. I would like to sort out why. Is it
merely that common human experiences result in universals and that unusual
cultural practices will result in rare lexemes? Or are there fundamental
semantic concepts (e.g. NSM's primitives) that tend to combine in common
patterns?

Another complicating factor is that all concepts appear to be related to
other concepts. Steyvers and Tenenbaum claim that the three semantic
networks that they investigated each contained a giant network in which
almost all of the words were related. One of the three had a single network
that contained 96% of all words. The other two had networks that contained
99% of all words. Steyvers and Tenenbaum also claim that the network is not
uniformly distributed. Words tend to cluster around hubs. This is the basis
of a semantic domain-a hub with a cluster of related words. As I've studied
semantic domains, I've come to several conclusions. For instance I've known
for a long time that words were not uniformly distributed, but tend to
cluster. The question is why. Steyvers and Tenenbaum also claim that words
tend to be connected to few other words, but the hubs tend to have many
connections. This results in a very dense network. So it only takes an
average of four or five links to get from any given lexeme to any other
lexeme. This is not something that is predictable or true of other kinds of
networks. So why does the mental lexicon have this kind of structure?

Another puzzling feature of the lexicon is that high-frequency,
mono-morphemic lexemes tend to be hubs. A few of them are super generics
(e.g. thing, person, do,). But in my list of semantic domains they tend to
be mid-level domains. The very general and very specific domains tend to be
complex lexemes (e.g. 4 Social behavior, 4.2.1.4.2 Show hospitality).
Mono-morphemic lexemes tend to be hubs and complex lexemes tend to be
specific types (e.g. chair versus highchair, rocking chair, armchair).
Consequently I have felt for a long time that my hierarchical organization
is not a good reflection of the semantic network. It may be convenient, but
it is linguistically wrong. So I am very interested in knowing what the
semantic network actually looks like. What are the major hubs? Why are some
lexemes major hubs and others not? How are the hubs related? How do we model
the semantic network? What implications does all this have for practical
lexicography? The DDP list of domains works really well as a word collection
tool, but could it be improved? Could we construct a truly etic list of
semantic domains/hubs that could be used to investigate the semantics of any
language? To what degree do the hubs of one language correspond to the hubs
of another language? I think this is a huge area needing lots of research.

Ron Moe

  _____  

From: lexicographylist at yahoogroups.com
[mailto:lexicographylist at yahoogroups.com] On Behalf Of Wayne Leman
Sent: Monday, November 17, 2008 5:47 PM
To: lexicographylist at yahoogroups.com
Subject: [Lexicog] Digest Number 1050

<snip>

> I just read Steyvers and Tenenbaum. I don't know anything about 
> statistics,
> so I didn't gain much from it. But the article raises a number of issues
> that I am very interested in. Their semantic networks and central nodes 
> are
> very similar to semantic domains and other perspectives on semantics. I've
> been trying to get a handle on why the mental lexicon tends to clump and
> cluster around key words.
> Ron Moe

Ron, are you searching for explanations other than cultural saliency? I've 
noted what others have, that lexicons tend to be highly developed for those 
semantic domains which reflect a highly developed area of the culture of the

people speaking that language.

So, for instance, Cheyennes, with whom I have worked, have many terms for 
different kinds of horses. Horses have been very important to the Cheyennes 
since their introduction via Mexico quite some time ago.

There are many terms for cuts of meat. That reflects the hunting part of 
their traditional hunter-gatherer culture, where it was important to butcher

a buffalo or deer well and to use the different parts in traditional ways.

There used to be a number of terms for growing corn, when Cheyennes were 
stationary long enough to do some planting. I've seen some of the terms in 
older lists of Cheyenne words. But the corn growing terms are nearly extinct

now that Cheyennes no longer grow corn. (It's easier to buy corn at the 
grocery store.)

Wayne
-----
Wayne Leman
Cheyenne dictionary online:
http://cheyenne. <http://cheyenne.110mb.com> 110mb.com 

No virus found in this incoming message.
Checked by AVG - http://www.avg.com
Version: 8.0.175 / Virus Database: 270.9.4/1792 - Release Date: 11/17/2008
8:48 AM

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lexicography/attachments/20081117/0823934e/attachment.htm>