MacEachern, Margaret R. (1999) Laryngeal Cooccurrence
Restrictions. Garland Publishing, ix+180pp, hardback ISBN
0-815-33267-X, $55.00, Outstanding Dissertations in Linguistics.
Mike Cahill, SIL International
DESCRIPTION OF CONTENTS
This volume is a revised version of MacEachern's 1997 UCLA
dissertation. Besides specific Optimality Theory (OT)
analyses of laryngeal cooccurrence restrictions in 11
languages, MacEachern uncovers an implicational hierarchy
of similarity effects, as well as a "leftness effect" in
several languages.
Chapters One (Introduction) and Two (Outline and Analysis)
lay out the basic structure, claims, and the OT analytical
approach of the work. Three main claims are laid out in
these chapters:
- cooccurrence restrictions fit an implicational hierarchy
- in some languages, identical elements may co-occur,
even if similar elements do not.
- the threshold of acceptable similarity differs from
language to language
The following is one of the cooccurrence tables M presents,
adapted from her (4) in Chapter One. I use the following as
cover symbols, similar to those M employs. Th is any
aspirated voiceless stop, T' is any ejective stop, D and B
are implosive stops. K is a stop heterorganic to T, ? is
glottal stop, and h has its usual value. An 'x' indicates
the occurrence of this cooccurrence pattern within a
morpheme, '*' indicates its absence, and a blank indicates
that the sounds in question do not exist in the language or
are excluded due to distributional restrictions. CQ is
Cuzco Quechua, PA is Peruvian Aymara, BA is Bolivian
Aymara, Tzu is Tzutujil, and Shu is Shuswap. This table
indicates cooccurrences, in either order; thus the first
line with Th <-> ? signifies either Th--? or ?--Th.
CQ PA BA Tzu Shu
1 Th <-> ? x x x
2 T' <-> h x x x x
3 D <-> h x
4 h <-> ? x x x x
5 Th <-> h * x x
6 T' <-> Kh * * x
7 Th <-> Kh * * x
8 ? <-> D x
9 ? <-> T' * * * x x
10 T' <-> B x
11 T' <-> K' * * * * *
12 T' <-> Th * * *
-----------------------------------------------------
13 h <-> h * x x
14 ? <-> ? x x
15 Th <-> Th * x x
16 T' <-> T' * x x x *
In the order that M has arranged the cooccurrences for
these languages, three patterns are evident. First, in the
non-identical pairs 1-12, each language has occurrences of
the pairs higher in the chart, but the lower ones in the
chart do not occur. There is a single discontinuity between
occurrences and non-occurrences. But the location on the
chart of the discontinuity between occurrence and non-
occurrence is different for each language.
Second, in 13-16, where the segments are identical, there
are three patterns. PA, BA, and Tzu allow all identical
segments; M terms this the "complete identity effect." Shu
allows only identical glottals, which lack a Place node; M
terms this the "incomplete identity effect." CQ does not
allow cooccurrence of any identical segment, and so shows
no identity effect.
Third, as one goes down the list, the pairs at the top are
more dissimilar to each other, and the pairs at the bottom
are more similar.
Thus there is an implicational hierarchy in 1-12. If a
given cooccurrence exists, then any cooccurrence higher on
the chart will also exist (assuming the sounds in question
occur in that language). Conversely, if a possible
cooccurrence does not exist, any cooccurrence lower on the
chart will not exist either. M later introduces other
languages (for a total of 11 in all) as well as other
segment cooccurrences (for a total of 24 pairings), but the
same relationship holds.
In Chapter Three (Data), M presents in detail the data for
the 11 languages she analyzes. Besides Cuzco Quechua,
Peruvian Aymara, Bolivian Aymara, Tzutujil, and Shuswap
previously mentioned, she also includes Souletin Basque,
Sanskrit, Ofo (an extinct Siouan language), Gojri (Indo-
Iranian), Hausa, and Old Georgian. For each language, M
presents language basics such as family and segment
inventory, details of the cooccurrence restrictions, and
further descriptive details, including counterexamples. For
some languages, she adds an additional excursus on one
detail or another (which for Bolivian Aymara is the bulk of
that section). Thus the reader has a firm empirical base on
which to judge M's analysis in the following chapters.
Chapter Four (Earlier Analyses) is a brief look at previous
analyses of cooccurrence restrictions, especially McCarthy
1989. M notes two shortcomings of cooccurrence restrictions
based purely on the Obligatory Contour Principle (OCP).
First, it is difficult to capture gradient effects with
these. Second, theory-internal principles are violated. If
the OCP allows completely identical consonants by means of
a branching Root node, then Aymara, where consonants
intervene between identical consonants, must have crossing
association lines.
Chapter Five (Analysis -- Constraints, Tableaux, and
Rankings) is where M provides in-depth analysis of
cooccurrence restrictions in each of the 11 languages in
her sample. She uses McCarthy and Prince's 1995
Correspondence Theory as her model, but for individual
features, uses MAX-IO (Feature) and DEP-IO (Feature) rather
than IDENT (Feature) constraints. For dissimilation, M uses
the terminology of Suzuki 1998, with his Generalized OCP,
which does not depend entirely on feature geometry.
Constraint rankings and illustrative tableaus are given for
all languages except Gojri and Old Georgian, for which she
supplies only the constraint rankings.
M replaces the *SIMILARITY constraints of her dissertation
with conjunctive constraints penalizing cooccurrence of
similar segments. For example, the constraint *2CG
prohibits two segments with [constricted glottis], e.g.
t'at'a or ?at'a. The constraint *IDENTITY prohibits two
identical segments. The conjunctive constraint *IDENTITY &
*2CG is violated if one or both the component constraints
is violated. I have reservations about the necessity of
these conjunctive constraints, which I will discuss more
below.
Interestingly, her analysis in terms of similarity and GOCP
in this chapter does not explain the implicational
hierarchy illustrated by the table reproduced above. She
needs a separate set of principles, discussed in Chapter 7.
In Chapter Six (Additional Analysis), M deals with the
"leftness effect," one of the more interesting empirical
phenomena of the study. In Souletin Basque, Cuzco Quechua,
and Peruvian Aymara, laryngeal stops (either aspirated or
ejective) may only occur as the first stop in the word. M
proposes constraints that penalize laryngeals that do not
occur in leftmost position to account for these. Bolivian
Aymara offers a complication; it may have both an aspirated
stop and an ejective, but the ejective must precede the
aspirated stop unless the first stop is bilabial or uvular,
in which case the reverse order holds. For this, M offers 5
additional constraints, including the arbitrary-appearing
EJECTIVES PRECEDE ASPIRATES. (In my opinion, if one is to
deal with the whole of a language in a finite time, one
will probably eventually come face-to-face with what looks
like an arbitrary pattern. In such cases, the only option
is to posit a constraint that describes the surface
pattern, and trust that future research will uncover a
phonetic, psycholinguistic, or historical rationale for it,
or that it can be decomposed into constraints that have
such a rationale.)
In Chapter Seven (Similarity and Laryngeal Cooccurrence
Restrictions), M shows that a factorial typology of the
main constraints she deals with produce a number of
grammars which are unattested. As a result, the
implicational hierarchy illustrated in the table above is
not generated. She suggests that the actual typology is not
a product of these constraints themselves, but that an
additional factor, acoustic similarity, must be called on.
The pairings at the top of the chart are less acoustically
similar than the pairings at the bottom. She concludes that
the phonetic facts of similarity drive this implicational
hierarchy. Some of the rankings of these are not
empirically attested, and this entire area is a subject for
future experimentation. One wonders how far one could get
in language-specific analysis using only acoustically-based
constraints.
In Chapter Eight (Stochastic Constraints), M briefly
discusses the works of Frisch, Broe, and Pierrehumbert in
accounting for cooccurrence restrictions in Arabic, based
on similarity of segments. She suggests that this approach
is incapable of accounting for languages in which identical
segments can co-occur, but similar ones may not, including
several languages in her survey.
A very brief Chapter Nine (Summary Remarks) concludes M's
text. A Bibliography and Index are included as well. The
Index is almost entirely of languages and authors; only
four subjects are included (Grassman's Law, the identity
effect, the leftness effect, and the Obligatory Contour
Principle).
CRITICAL EVALUATION
The selection of languages that M presents is far from
being totally diverse, Peruvian and Bolivian Aymara being
the most obvious example. Still, these two illustrate that
even closely related languages may have different patterns
of cooccurrence restrictions. Her sample includes languages
from South America, Meso-America, North America, west
Africa, Europe, and subcontinental Asia. This does
represent quite a diverse group of languages, and though
one may always wish for a broader sample (insert SiSwat'i
or your favorite language here), this language sample
represents quite a fair representation. One may also
question the use of three extinct languages (Sanskrit, Old
Georgian, and Ofo), but these are at least as well
documented as many languages still spoken today.
I applaud M's devoting an entire chapter to description of
the details of the data upon which she bases her analysis.
For too many phonological analyses, we are treated to a
fraction of the actual relevant data. M is honest in
telling us what data her analysis does not account for
(which is quite a tiny fraction). It is also commendable
that she tells what her methodology is, e.g. her criteria
for including or excluding forms. M notes that some sources
she consulted were unclear on whether the word cited was
monomorphemic or not. Thus there is some question as to the
domain of the cooccurrence restrictions in several of the
languages. As she notes, this is an area for further
research. Furthermore, she did not depend entirely on
written accounts, but dug up some language study tapes of
Peruvian Aymara (PA) and re-transcribed some forms from
them.
In PA, words that were originally transcribed as vowel-
initial are transcribed by M as glottal stop-initial.
(Unfortunately, M is not explicit about whether the target
words were phrase-medial and thus potentially intervocalic
or not; in a language study tape, we might guess there was
a mix.) She also interprets Hausa's voiceless stops as
aspirated. This is not how they are generally described,
but M does offer some support for this interpretation with
literature references. In each of these cases, the re-
interpretation is more compatible with M's analysis than
the original transcription. When one finds data that do not
fit one's analysis, it is of course a valid procedure to
re-check the data. There is both a trap and an opportunity
in such situations. One may be tempted to re-interpret the
data in a way that has no objective justification at all.
Then your theory survives, but you have not done justice to
the data at all. Or your theory may drive you to a closer
reexamination of the data, and you may actually get a more
accurate transcription or interpretation of the data. In
the case of PA (as well as other instances where she adds a
[?]), I think M has done us a service of better
interpretation of the data. Her case for Hausa is perhaps
not as strong.
I mentioned that M uses MAX-IO (Feature) and DEP-IO
(Feature) rather than IDENT (Feature) constraints. This
requires that the insertion and deletion of a feature are
evaluated separately. Struijke (2000) has analyzed the same
Cuzco Quechua language with a modified model of IDENT (F)
and rejects the use of MAX/DEP for features, and the debate
is still open on the general use of one strategy vs.
another. At this point I am not prepared to judge the
relative merits of these in Quechua, but the use of MAX/DEP
for features still seems appropriate at least for features
that exhibit classical autosegmental behavior, such as tone
and nasality. For cases in which a tone is inserted on an
epenthetic vowel or a toneless suffix, and MAX(HighTone) is
undominated but DEP(HighTone) is very low-ranked, see
Cahill (1999). An example using nasality is the [+nasal]
feature in Yaminahua which originates in a suffix but
surfaces only on the first syllable of the stem (Loos in
prep). For a discussion of other features, see Zoll (1996).
One of the areas in which M is not always consistent is
that of underlying representations, or input forms, which
are included in every tableau. The very need for underlying
representations has been called into question by Burzio
1996 and others, and the uniqueness of inputs has been
somewhat undermined by the notion of the Richness of the
Base (McCarthy & Prince 1995 and others). Nonetheless, in
Correspondence Theory, specific input forms are necessary
to evaluate a tableau. In this work, M is not always
consistent in the type of input form she posits, and does
not tell us what principles, if any, she is using to
determine the input. In many cases, the input is identical
to the optimal form. In other cases, she proposes an input
with two aspirated or two ejective consonants, when the
output contains only one. Presumably this is to show that
it does not matter whether you have an input with two
ejectives or one; the constraints will give the same output
in either case. However, the choice of inputs appears to be
arbitrary in, for example, Cuzco Quechua, where she posits
an input of /kat'a/ to give an output of [kata]. An input
of /kata/ or even /katha/ would also produce the same
output. If we claim some sort of psychological reality for
our input forms (and as one who has developed an
orthography for a previously-unwritten language, I am
concerned about this), then it would be interesting to test
whether Cuzco Quechua speakers have an intuition about what
exactly is the underlying segment in the second syllable
here. Though the theory may be able to handle multiple
inputs, it may not be the case that the native speaker is
so unconstrained.
M notes in her introduction that one difference between her
dissertation and this published version was that she
replaced her *SIMILARITY constraints with local conjunction
of constraints. However, I do not see anywhere in her
analyses where these conjunctive constraints do any crucial
work, as opposed to their components. In the outline in
Chapter Two, for example, the conjunctive constraint *ID &
*2CG could be replaced by its component *ID with no change
in optimal output. In the more detailed analysis in Chapter
Five, conjunctive constraints are not used with Souletin
Basque, Cuzco Quechua, or Shuswap. With PA, BA, and Ofo,
they are unnecessary in the tableaus given. In Hausa and
Tzutujil only one member of the conjunctive constraint is
sufficient to rule out undesirable candidates. (No tableaus
are provided for Sanskrit, Gojri and Old Georgian). I
suspect that in the process of editing the dissertation for
this publication, some explanatory sections had to be cut.
Unfortunately, that leaves us with no explanation in some
cases of the justification for some of the rankings, and
why M felt it necessary to use a conjunctive constraint
rather than just one component of it.
MacEachern has provided us a service in uncovering cross-
linguistic patterns of laryngeal cooccurrence restrictions,
and especially the similarity hierarchy. My main criticism
is with some of the formalities of her OT analysis. But I
suspect that these will, like those in my own dissertation,
be of less interest in a decade than the cross-linguistic
language patterns. Analyses come and analyses go, but the
data remains (relatively) forever.
REFERENCES
Burzio, Luigi. 1996. Surface constraints vs. underlying
representations. In Jaques Durand and Bernard Laks
(eds.). Current trends in phonology: models and methods.
Salford, Manchester: European Studies Research
Institute.
Cahill, Michael. 1999. Aspects of the Morphology and
Phonology of Konni. Ph.D. dissertation, Ohio State
University.
Loos, Eugene. (in preparation). Discontinuous nasal spread
in Yaminahua.
McCarthy, John. 1989. Linear order in phonological
representation. Linguistic Inquiry 20:71-99.
McCarthy, John, and Alan Prince. 1995. Faithfulness and
reduplicative identity. In Beckman, Dickey, and
Urbanczyk (eds.) Papers in Optimality Theory. UMOP 18.
Amherst, MA: GLSA. pp. 249-384.
Struijke, Caro. 2000. Existential Faithfulness. Ph.D.
dissertation, University of Maryland at College Park.
Suzuki, Keiichiro. 1998. A Typological Investigation of
Dissimilation. Ph.D. dissertation, University of
Arizona.
Zoll, Cheryl. 1996. Parsing Below the Segment in a
Constraint Based Framework. Ph.D. dissertation, UC
Berkeley.
ACKNOWLEDGMENT
Thanks to Steve Parker for keeping me from making worse
mistakes than I have. Remaining errors are my own fault.
BIOGRAPHICAL SKETCH
Mike Cahill has done on-site linguistic investigation in
the Konni language of northern Ghana for several years,
including application to literacy and translation work. He
received his Ph.D. from Ohio State University in 1999, and
is primarily interested in African phonology, cross-
linguistic patterns in tone, and labial-velar stops and
nasals. He currently serves as SIL's International
Linguistics Coordinator.
|
