An argument, based on Chumash sibilant harmony, against Kiparsky's
proposal that Strict Cycle effects are derivable from underspecification.
Includes an argument that feature-changing harmony results from distinct
delinking and spreading rules, since another rule must intervene.
As published in Ellen Kaisse and Sharon Hargus (eds.)
Lexical Phonology and Morphology
San Diego: Academic Press. pp. 315-321.
but formatted and paginated differently.
A critique of the claim by Joseph Greenberg and Merritt Ruhlen that comparative study of
the Hokan family has suffered from emphasis on binary comparison and that this illustrates
the advantage of multilateral comparison over binary comparison.
As published in International Journal of American Linguistics
61.1.135-144. (1995), but formatted and paginated differently.
Presents evidence, citing examples from Japanese, English, and Basque,
that lexical items can block phrasal constructions. Proposes that this effect
is restricted to syntactic constructions that instantiate what in other
respects are morphological categories.
As published in Ivan Sag and Anna Szabolsci (eds.)
Lexical Matters, Stanford, California: Center for the Study of Language and Information.
pp. 111-130. (1992), but formatted and paginated differently.
Athabaskan languages have extremely complex and productive morphology based
overwhelmingly on verbal roots. There are very few basic nouns. Monosyllabic
noun stems are either unanalyzable, and therefore presumptively old,
or, where analyzable, reflect very old derivational processes.
It has therefore been suggested that they can provide
a window into the deeper layers of culture history. However, this project
has not to my knowledge previously been carried out. In this paper I report
on an analysis of the monosyllabic noun stems of Carrier,
one of the few languages of the region for which really extensive lexical
information is available.
Approximately 360 monosyllabic noun stems are attested, representing
about 10% of those phonotactically possible. A handful, such as
"cat" and "berry" are loans (from English and Gitksan
respectively), thus demonstrating that monosyllables are not invariably
old. In general, the monosyllables do reflect what are probably very
old aspects of the culture. By far the most heavily represented semantic
fields are anatomical terms and kinship terms; most of the basic terms
in these areas are monosyllables. Similarly, much of the terminology
for describing the natural world (e.g. island, river, fire, earth, sand)
is monosyllabic. The monosyllabic technological terms are suggestive
of a cultural emphasis on water and on trapping; in general they
reflect a very old layer of technology. Perhaps surprising is
the relatively small amount of monosyllabic biological terminology.
Even items for which Athabaskan speakers have surely had words for a
very long time sometimes are morphologically complex. For example,
most dialects of Carrier use a deverbal form for "porcupine", literally
meaning "the quilled one".
Argues on the basis of the Tongan definitive accent that a morpheme may
have a syntactic distribution (in this case, it apears at the right edge of the NP,
attached to whatever word may appear there) but nonetheless be lexically attached.
This situation may be dealt with by generating the forms in the lexicon and checking
foot features in the syntax.
The paper as presented here is a slightly reformatted version (with footnotes rather than
endnotes) of the original, which appeared in Michael Wescoat et al. (eds.)
Proceedings of the Fourth West Coast Conference on Formal
Linguistics (Dept. of Linguistics, Stanford University,
Palo Alto, California.) (1985) pp. 262-272.
In Carrier, it is impossible for the source and goal of motion to be specified in the same
clause. This constraint applies at the level of argument structure. The constraint results from the
fact that Carrier motion verbs have a single directional argument position.
This is a goal by default, but may be changed into a source by suitable morphology.
This is a revised and expanded version of a paper presented at the Winter meeting of
the Society for the Study of the Indigenous Languages of the Americas,
New York City, 11 January 1998.
To appear in The Nature of the Word: Essays in Honor of Paul Kiparsky,
edited by Kristin Hanson and Sharon Inkelas, MIT Press.
A discussion of a variety of phonological phenomena related to the D-Effect
in Southern Dakelh (Carrier) dialects, including a case of dialectal variation in rule
ordering and the appearance of multiple D-valence prefixes in sequence.
The Double-O Constraint (Harada 1973), is intended to account for the
ungrammaticality of clauses containing two accusative Noun
Phrases. It has been discussed by numerous authors in various formulations.
This paper attempts to clarify the phenomena involved and to reduce the class
of possible analyses. Five main points are made:
(a) that there are actually two constraints, the Deep Double-O Constraint,
violation of which produces gross ungrammaticality, which is not
subject to variation among speakers, and does not require the presence
of two surface accusatives, and the Surface Double-O Constraint,
violation of which frequently fails to produce outright ungrammaticality,
which is subject to considerable variation among speakers, and which
arises only when two accusatives are present on the surface.
(b) Several formulations of the DDOC are untenable, namely
those based on: (i) valency; (ii) surface case; and (iii) thematic roles.
(c) The DDOC must be stated on argument structure.
(d) there are four classes of accusatives that do not count for the
DDOC: (i) path accusatives; (ii) body-part accusatives;
(iii) tokoro complements; and (iv) ablatival accusatives.
The last are exempt only for a minority of speakers, reflecting a nearly
complete historical change.
(e) The status of the accusatives that do not trigger the DDOC
is unclear. They pass certain putative tests for object status.
That is, they may be passivized, and they may float quantifiers. However,
neither of these now appears to be a clear test for object status. It is thus
possible to treat these accusatives as oblique.
Carrier, an Athabaskan language of the central interior of British Columbia,
was first written in 1885 in a derivative of the Cree syllabics, in which, for
a time, there was mass literacy. This writing system, including deviations from
the ``official'' version, is here for the first time described in detail. In spite
of its name, it is shown to be an alphabetic writing system.
The usage of the system is discussed, and the differences between it and the
antecedant Cree and Northwest Territories Athabaskan writing systems are elucidated.
Carrier lexical semantics makes use of a concept of "effective uncountability",
that is, of sets of individuals of such a nature that their members are not normally,
individuated. This notion characterizes one of the categories
of the classificatory verb system, and is part of the definition of one of the verbs
of eating, which describes the marked situation in which an "effectively uncountable"
set is eaten in such a way as to individuate its members.
In most dialects of Carrier the subject marker /id/ (Stuart/Trembleur Lake
dialect) or /id@d/ (Southern dialects), the reflex of the
Proto-Athabaskan first person duo-plural subject marker, has become
restricted to the dual, evidently due to partial blocking by the extension of
/ts'/, the old indefinite subject marker, to first person plural subject.
In the Ulkatcho dialect, however, while only /ts'/ can be used in the true plural,
/id@d/ retains its original duo-plural usage; its use in the plural is not
blocked by /ts'/.
Ulkatcho dialect thus presents a counterexample to the strongest
formulations of the blocking principle, under which the movement of
the indefinite into the first person plural role ought to result in
the immediate restriction of the old duo-plural to the dual.
Ulkatcho presumably reflects the intermediate stage in the
historical development. The indefinite first took on the added role
of first person plural, resulting in competition between the two
forms, as in Ulkatcho. Eventually, the two forms became fully
differentiated, with the restriction of the old duo-plural to the
dual, as in the majority of the dialects. A survey of the
languages with which Ulkatcho Carrier has been in contact
indicates that it is unlikely that the existence of competing
forms is due to influence from another language as has been
suggested by Kroch (1994).
One element of the recent controversy over historical
methodology set off by Greenberg (1987)'s classification of
American Indian languages has been his reliance on superficial
lexical resemblances, with no attempt to establish phonological
correspondences and no evidence from submerged morphology. Proponents
of this methodology argue that this is the methodology used to establish
the Indo-European language family, and that the success of these
methods in the Indo-European case shows them to be reliable.
We argue that this view of the history of Indo-European studies
is seriously flawed, in two ways:
(a) for the most part, neither the recognition of languages as
IE nor their internal classification have been based primarily on
superficial lexical resemblances;
(b) where such methods were employed, they frequently led to
Written with Lyle Campbell. In
Proceedings of the Eighteenth Annual Meeting of the
Berkeley Linguistics Society. (1992) pp. 214-236.
Japanese periphrastic verbs, consisting of a verbal noun together
with the verb suru `do', are generally considered to be
lexically incorporated. I argue that in fact they are not only not
lexically incorporated, they are not incorporated at all, but
remain analyzable at every level of representation. While it is possible
to account for the Japanese facts without positing any sort of
constituency of the verbal noun and suru, there are theoretical
grounds for preferring an analysis in which the periphrastic complex
consists of a V dominating N V, including the proposal that the
introduction of non-head lexical categories is restricted to the expansion
of lexical categories. Distinguishing this case and others like it from
true incorporations makes possible the claim that true lexical incorporations
are syntactically opaque.
Research on Carrier, a dialectally diverse Athabaskan language of the
central interior of British Columbia, has resulted in a fairly large
amount of information stored on-line
The databases in which this material
is kept are used both for research and to generate printed dictionaries.
The database system is home-brew, developed incrementally over a decade, primarily
using UNIX tools. A similar system, though different in detail, underlies the Montana
Salish (Flathead) dictionary. This paper describes this system and explore its virtues
This is the revised and expanded text of the paper I had prepared, but due to illness
was unable to give, for the workshop
"New Methods for Creating, Exploring and Disseminating Linguistic Field Data",
organized by Steven Bird under the auspices of the Talkbank Project at the University of
Pennsylvania, on 6 January 2000 in Chicago, Illinois.
Two periphrastic constructions, combining a verbal noun and suru
"do", are generally recognized in Japanese. In one, the so-called
"unincorporated" construction, the verbal noun is marked accusative
and the object, if any, appears as a genitive complement within the
full NP headed by the verbal noun. This construction is on all
accounts unequivocally phrasal. In the other, the so-called
"incorporated" construction, the verbal noun is not case-marked and
the direct object, if any, is marked accusative. Although this
construction has often been taken to involve lexical incorporation of
the verbal noun into suru, there is considerable evidence that
no incorporation takes place and that the construction is actually
The "incorporated" construction actually conflates two
subtypes: in addition to the periphrastics that exhibit phrasal
behaviour, there is a subset that exhibit truly lexical behaviour.
These fail all eight tests for phrasal status
and differ from phrasal "incorporated" periphrastics in
another eleven properties. Recognizing this third construction
eliminates a number of hitherto mysterious irregularities. Which nouns
form lexical periphrastics and which phrasal is predictable
phonologically: those verbal nouns that are underlyingly monosyllabic
form lexical periphrastics. No explanation for this restriction is
Msort is a program for sorting text files in sophisticated ways,
intended especially for linguistic databases. It allows
arbitrary sort orders to be specified, with ranks defined for
large numbers of multigraphs of effectively unlimited length.
Records need not be single lines of text but may be delimited in a
number of ways. The entire record may used as the sort key, or a
particular field may be used. Key fields may be selected either
by position in the record or by matching a regular expression to a tag.
msort is capable of sorting on several keys, so that when
two records tie on one key, the tie may be broken on another. Each key
may have its own sort order. Any or all keys may be optional.
In addition to lexicographic sorting, sorting by numerical value,
date or time is supported.
For each key a distinct set of characters may be excluded from
consideration when sorting in any combination of initial, final, and medial
position in the key field. Lexicographic keys may be reversed, allowing
the construction of reverse dictionaries.
Designing a dictionary for an Athabaskan language presents unusual
difficulties. Because of the enormous
complexity of the verb, it is impossible to list every form of
every verb. Because Athabaskan languages combine extensive prefixation with
complex stem variation, and because the components that contain the basic
meaning of the verb are distributed throughout the form,
intercalated with grammatical morphemes, there is no
straightforward, easily extracted and manipulated, citation form.
Using a fixed member of the paradigm
is also problematic because the user must have substantial knowledge
of the language to be able to convert other forms to the citation form.
As a result, dictionary designers have had two unpleasant choices.
One is to use fully inflected forms. These are easy to use, but
necessarily far from complete.
The other possibility is to produce root-based analytic dictionaries.
Such dictionaries may be comprehensive but are almost impossible to use for
anyone without considerable meta-knowledge of the language.
The way between the Scylla of incompleteness and the Charybdis of
unusability is an on-line dictionary, internally analytic,
with a morphological parser as front end. This will allow the
user to enter a fully inflected word to be analyzed by the parser.
However, difficult problems arise as to how to present the information
generated by such a system. Just as finding a word in an analytic dictionary
is not trivial, so is making use of the output from one.}
This version differs from the published version in having a title page and
abstract, but has the same page divisions and text.
First Nations languages are referred to by an often puzzling variety of names.
Many people are confused by the fact that apparently authoritative sources
differ in the names they use.
This paper provides details on the names of the languages of British Columbia,
explains the factors underlying variation in nomenclature,
and discusses the controversy over appropriate names.
A description of a strongly grammar-based university-level
Carrier curriculum, which enabled the students to produce
and understand complex sentences that they had not previously heard.
I argue that curricula based on culturally salient vocabulary and
fixed expressions are not based on any authentic native tradition.
Rather, they are ultimately the result of colonialism.
A discussion of certain alternations in the Rayciska
dialect of Ainu, proposing a a simple phonological account in place of Murasaki's
account, which involves lexically conditioned allomorphy. Includes a comparison with
Hale's discussion of Maori passives.
A review of Smalley et al.'s book about the Pahawh Hmong writing system and its development.
Among other things, argues that this is an instance of onset-rime based writing rather than
a truly segmental system.
In Carrier, an Athabaskan language of the Central Interior of British Columbia,
the negative particle 'aw as well as some adverbs has rightward
scope. As a result, material within the scope of negation other than the verb
must follow 'aw, though topicalized constituents fall within
scope even though they precede 'aw
Other scope-bearing elements, including cha "also" and z
"only", have leftward scope. These particles underlie the constructions
S za Vaux "to keep on S-ing" and
S cha Vaux "to S also", in which the choice of
dummy verb Vaux is determined by the event type, valence, and aspect
of the main clause.
It is suggested that the use of the dummy verb results from the interaction
of leftward scope with the requirement that the clause be verb final.
In Marion Caldecott, Suzanne Gessner, and Eun-Sook Kim (eds.)
University of British Columbia Working Papers in Linguistics
(Proceedings of the Workshop on Structure and Constituency in
Languages of the Americas)
2.107-115. Vancouver, British Columbia. (November 1999).
Sir Thomas Young was the first, in 1819, to make explicit statistical calculations
in support of genetic affiliation of languages. His mathematics was correct,
but his application of it incorrect. He nonetheless deserves credit for
his recognition of the problems inherent in this technique.
Chumash sibilant harmony, perhaps the strongest example of a feature-changing
harmony rule, has been dismissed by Russell (1993) and Bird (1995),
proponents of declarative approaches to phonology, as a `phonetic process'.
This characterization does not stand up to analysis.
Chumash sibilant harmony is indeed a phonological rule and must be dealt
with by phonological theory.
A detailed description of the status of documention
(grammars, dictionaries, collections of text, and university-level language
textbooks) for the native languages of British Columbia, with bibliographic
references. Includes information
on the linguists knowledgable about BC native languages, and which ones
are currently active. Prepared at the request of Grand Chief Edward John
and first circulated in January 1999, this has been repeatedly updated and
revised as new information has become available.
[Yinka Dene Language Institute Technical Report #2.]