LINGUIST List 13.3169

Tue Dec 3 2002

Review: Morphology: Bendjaballah et al. (2001)

Editor for this issue: Naomi Ogasawara <>

What follows is a review or discussion note contributed to our Book Discussion Forum. We expect discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for review." Then contact Simin Karimi at


  1. Mike Maxwell, Bendjaballah et al. (2001) Morphology 2000

Message 1: Bendjaballah et al. (2001) Morphology 2000

Date: Tue, 03 Dec 2002 00:50:32 +0000
From: Mike Maxwell <>
Subject: Bendjaballah et al. (2001) Morphology 2000

Bendjaballah, Sabrina, Wolfgang U. Dressler, Oskar E. Pfeiffer, and
M. D. Voeikova, ed. (2001)
Morphology 2000: Selected Papers from the 9th Morphology Meeting, Vienna,
24-28 February 2000. John Benjamins Publishing Company, $95.00,
Current Issues in Linguistic Theory, 218.

Book Announcement on Linguist: 
Announced in Linguist List:

Mike Maxwell, Linguistic Data Consortium

As the title says, this book contains a selection of papers from a
meeting, the main topic of which was ''comparative morphology''. The
editors mean by this term ''cross-linguistic analysis, including
typology, dialectology and diachrony'' (p. 1). An additional topic was
the psycholinguistic of morphology. But in fact a good third of the
papers do not clearly fall into any of these categories. As one might
expect in a conference proceedings, most of the papers are 10-15 pages
long, 24 in all. Unlike most conference proceedings, however, there is
an index of language and subjects (but not of authors).

In the remainder of this review, I will briefly summarize each paper,
adding my own commentary where appropriate. Extended abstracts of all
the conference papers, including some which do not appear in the book,
can be found at the following site.

Adam Albright (''The lexical bases of morphological well-formedness'')
examines how native speakers evaluate the well-formedness of nonce
formations in Italian (much like the hypothetical past tense forms of
the non-existent English verb 'spling' which were the subject of
debate some years ago). Some linguists (e.g. Bybee) would claim that
native speakers evaluate such forms by their conformance to other
patterns (in the case of this English example, real verbs like 'ring',
'sing' and 'fling') on the basis of the type frequency of the
patterns, i.e. the number of words conforming to a given pattern. A
connectionist, on the other hand, might claim that it was the token
frequency of the words belonging to a given pattern that was
important. Since patterns with low type frequency tend to contain
words having high token frequency (presumably because words with low
token frequency tend to regularize, leaving only higher token
frequency words in low type frequency patterns), the predictions
should diverge. Albright's tests (and other tests which he cites)
support the predictions based on type rather than frequency, although
the results are not as clear-cut as one might hope.

Mark Baker (''On category asymmetries in derivational morphology'')
demonstrates on the basis of a sample of languages that verbalization
of adjectives is much more common than verbalization of nouns, even in
languages where there nouns and adjectives behave otherwise quite
similarly. While one might attempt to account for this on the basis of
a feature counting metric, this would not account for similar
asymmetries between other categories. For instance, feature counting
could not account for the fact that nominalization of verbs is much
more common than verbalization of nouns, since both should involve the
same number of feature changes. To account for the adjective-noun
asymmetry, Baker proposes his own feature system, based on the
syntactic properties of categories, and shows that a feature counting
metric then explains the adjective-noun asymmetry. But while it works
for this particular asymmetry, Baker's system is still not capable of
explaining the asymmetry between the nominalization of verbs and the
verbalization of nouns.

In a paper which is nicely complementary to Baker's, Laurie Bauer
explores the varieties of category changes that derivational affixes
make across languages. Her paper takes a typological perspective, and
she does not try to explain the results.

For many years, Andrew Carstairs-McCarthy has explored constraints on
the behavior of inflectional affixes. On the face of it, the
conditioning of stem alternations by inflectional properties seem much
less restricted. His paper, ''How stems and affixes interact'',
compares two possible constraints governing stem allomorphy (including
one by Wurzel). Both proposals face apparent counter-evidence, and
Carstairs-McCarthy closes with a hope that other linguists will
explore these claims further.

Many tests have been proposed for distinguishing unaccusative and
unergative verbs, and it is not uncommon to find conflicts among these
tests. Bozena Cetnarowska explores one such diagnostic, the derivation
of resultative adjectives from telic verbs in Polish. Cetnarowska
attributes the absence of certain otherwise expected resultative
adjectives to diachronic changes which rendered the adjectives in
question unacceptable for reasons not having to do with telicity,
thereby explaining the partial failure of this diagnostic. That is,
while the existence of a resultative adjective is sufficient evidence
for the telicity (hence unaccusativity) of a verb, it is not necessary
evidence, for the absence of such an adjective may be attributable to
other factors.

Bernard Comrie asks whether morphophonological alternations can take
on new morphosyntactic meanings over time. As test cases, he chooses
Celtic 'mutations', and gemination in Italian and Maltese. The
evidence is less than conclusive.

In an invited paper, Greville G. Corbett, Dunstan Brown and Nicholas
Evans demonstrate how computational tools can be used to validate (or
invalidate!) a linguistic analysis. Their test case is the generation
of paradigms; their admission that the computer showed their initial
analysis to be wrong should be a lesson for all of us. The authors
also show how computer science can throw light on linguistic problems
by bringing out hitherto overlooked distinctions; the specific example
here is the notion of defaults, which turns out to have more than one

Wolfgang U. Dressler and M�ria Lad�nyi investigate whether
derivational affixation can be said to be more semantically
transparent in an agglutinating language than in a fusional
('inflecting') language; they use Hungarian and German as exemplars of
these two language types. While the answer is difficult to quantify
(and two languages is a small sample), they conclude that there is not
a significant difference.

Hilke Elsen studies the acquisition of the various plural markers in
German, using data collected from a single child, arguing that the
results support an 'associative learning mechanism' (under a
connectionist approach), as opposed to a rule-based generative
approach. But her characterization of the generative model of
acquisition is a caricature: ''The development of inflection is
independent of the lexicon. Steps of development are irreversible.'' I
know of no generativist working in acquisition who would make such
claims; indeed, the point of the U-shaped curve controversy is that
children change their minds (twice). In short, while her data is
interesting, it is probably insufficient to refute all but the most
na�ve model.

A second article on acquisition is that by Steven Gillis and Dorit
Ravid, who examine how learning spelling influences children's
morphologies, and vice versa. Specifically, they look at the
acquisition of spelling of Hebrew and Dutch words, where the spelling
makes distinctions that are neutralized in the spoken language. In
both languages, the children make use of allomorphy to recover
neutralized sounds. Differences between how the children treat
morphology and spelling in the two languages surface as well, which
the authors attribute to differences between the languages; but it
seems to me that they might just as well be caused by differences in
the way spelling is taught in the two school systems (in Israel and

In an experimental study, Laura M. Gonnerman and Elaine S. Andersen
attempt to disentangle effects of semantic relatedness and
phonological similarity on priming. (An example of two words that are
phonologically similar but semantically unrelated are 'corn' and
'corner'.) Perhaps unsurprisingly, their experiments show that varying
degrees of semantic relatedness correlate with varying degrees of
priming; and likewise with regard to phonological similarity. On the
other hand, prefixing and suffixing do not seem to differ
significantly in their effects on priming. Gonnerman and Anderson
suggest that these gradient results favor connectionist approaches
over ''traditional decomposition'' theories, as well as over dual
mechanism models which allow either lexical storage or
decomposition. However, it appears from the presentation of the data
that the results from the 58 participants in the study were lumped
together for analysis. If so, then the gradient results would be
compatible with a dual mechanism model in which some speakers stored
certain words as wholes, while others analyzed those same words into
their constituent morphemes.

A second experimental study, by Georgi Jetchev and Pier Marco
Bertinetto, compared reaction time in Bulgarian perfective and
imperfective verbs, under the assumption that the members of a
perfective and imperfective pair of verbs are derivationally, rather
than inflectionally, related. As the authors note, the issue of
whether such aspectual pairs in Bulgarian are indeed derivationally
related is controversial, and in my view they do not make a strong
case for the derivational nature of the relationship (assuming of
course that the inflection vs. derivation division is
well-founded). In any case, their experimental results did not
demonstrate a significant difference in priming between inflectionally
related forms and so-called derivationally related forms (aspectual

The last paper in the book also concerns aspectual pairs in a Slavic
language. The authors (Marina Roussakova, Serguei Sai, Svetlana
Bogomolova, Dmitirij Guerassimov, Tatiaia Tangisheva, and Natalia
Zaika) investigate whether aspect in Russian is a ''classifying
category'' or an inflectional category, by which question they mean
''Are members of aspectual pairs stored and processed as separate
lexemes or are they rather the forms of one lexeme?'' Again, this is
an experimental study, in which subjects (ranging from three-year olds
to adults) were given the present tense form of verbs (either
imperfective or perfective) and asked to provide the past tense
forms. Naturally, this was a difficult task for the youngest subjects,
and the authors briefly describe how the task was modified for them
(but not for the adults). These modifications resulted in the youngest
children scoring the best, and the adults scoring the worst!

With regard to the kinds of errors the subjects made, there was a wide
variation in the number of errors across verbs. But since only eleven
verbs were used, the authors admit that it is difficult to decide what
caused this variation. In addition, subjects based their answers more
often on the imperfective stems than on the perfective stems,
regardless of whether the form presented to them was perfective or
imperfective, which they argue is consistent with the lexically stored
form being the imperfective. The authors conclude that the ''members
of aspectual pairs are acquired separately and that their mental
representations are combined during the course of the development of
language competence.'' However, this conclusion seems rather tenuous
to me.

Gary Libben and Roberto G. de Almeida use experimental methods to
study whether native speakers of English analyze words into their
morphological constituents, even though the words may be relatively
common. Previous studies had given inconsistent answers to this
question. Libben and Almeida postulate that morphological analysis
might occur either ''pre-lexically'' (before lookup provides semantic
clues) or post-lexically; the results might then be expected to differ
for words whose meaning is non-compositional. For example, a ''dumb''
(pre-lexical) parser might parse 'humbug' into 'hum' and 'bug', but
the semantics should block such a post-lexical analysis. Unlike
previous work where these two effects were not distinguished, Libben
and de Almeida's experiment was designed to tease apart these two
sorts of parsing. The authors interpret the results (as well as a
study of errors by an aphasic patient) to indicate that parsing occurs
both before and after lexical lookup, regardless of whether the words
being parsed are familiar.

Generative linguists have sometimes been accused of forcing their
analysis of other languages into the mold of English. In a reversal of
this, Peter Hallman re-analyzes the English passive on the model of
his analysis of the Arabic passive. Specifically, he proposes that the
English passive suffix turns a verb into an adjective; valence
reduction is caused by another (zero) affix. (Hallman mentions, but
does not discuss, the claim by Wasow 1977) that English has both
verbal and adjectival passives, and that their properties are

Marit Julien argues that the positions in which tense and aspect
markers can and cannot appear are explicable under certain assumptions
about underlying syntactic structure. The argumentation is
theory-internal (Julien assumes a recent version of MIT-style syntax),
and counter-examples (only a few, to be sure) are re-analyzed. The
evidence for re-analysis is not presented-the reader is instead
referred to Julien's doctoral dissertation.

Elena Kalinina gives examples from a number of languages in which
stems which otherwise appear to be verbs take nominal inflectional
morphology, or in which nouns can be inflected as if they were
verbs. Kalinina reviews the explanations which have been proposed for
this phenomenon, and argues in the end for a syntactic resolution,
with the inflectional morphemes in question attaching at the phrasal
level-as clitics, apparently, although Kalinina does not use this
term. I confess to finding her arguments hard to follow (an unparsable
sentence at a crucial point on page 194 did not help).

A. E. Kibrik outlines the agreement system for transitive verbs in
Alutor. A system which at first glance appears complex and arbitrary
turns out to be relatively straightforward when a deictic hierarchy is
taken into account; the resulting system is reminiscent of the
well-known Potawatomi agreement system (Hockett 1948). Kibrik's
analysis, with its ordered list of rules and overrides, strongly
resembles the analysis one would propose in a realizational theory of
morphology (see e.g. Stump 2001, particularly the discussion of rule
competition in chapter 3).

Michele Loporcaro gives a historical analysis of the origin of
syncretisms among clitics in a dialect of Italian. A formalist
explanation of the synchronic system might attribute the syncretism to
language-particular constraints on a universal morphosyntactic system.
While accepting this as a synchronic explanation, Loporcaro sketches
an account of the historical origin of the syncretism, an account
which includes elements of sociolinguistic causation.

Igor Mel'cuk lays out a set of criteria to ''constrain the use of zero
signs by linguists.'' To a generative linguist like myself, this seems
an odd throwback to the American structuralists of the 1950s, and
their agnosticism about the mental reality of the linguistic
structures they studied. With that caveat in mind, Mel'cuk's criteria
seem fairly reasonable. It would, however, be interesting to study the
extent to which they are rendered irrelevant by an approach such as
Realizational Morphology (see e.g. Stump 2001).

Thomas Menzel attempts to ''examine the iconic relation between Slavic
[specifically, Polish and Russian] noun and adjective/pronoun
markers'', and he claims that the ''oppositions [are] organized in
certain levels of iconic structure'' (pp. 261-262). Again, I must
confess my bias as a generativist; there simply seem to be too many
loopholes in this theoretical approach, in that any counterexample can
be relegated to the marked categories of ''non-iconic'' relations or
''counter-iconic'' relations (both of which are
acknowledged). Granted, the claim is that marked categories will be
replaced by unmarked categories diachronically-but only ''if they
change at all and if the change does not come from outside the
system.'' Menzel finds factors which allegedly prevent iconicity, even
under change, among them a ''proposed tendency to establish
'phonologically similar' markers'' across paradigms (p. 268). In
short, if there is such a thing as iconicity in linguistics, it is
only one of many factors that affect diachronic change-and for all I
can tell, not a very important one.

Tore Nesset and Hans-Olav Enger also investigate iconicity, as it
affects morphological splits (the situation in diachrony where two
morphological markers exist where only one did before, with the two
markers dividing up the semantic domain which was previously the
meaning of the one marker). Their claim is that of the two markers,
the longer will be more 'informative'. Again my bias as a formalist
comes up. For example, the use of a longer allomorph for marking
plural vs. a zero marking is held to correlate with unexpectedness,
since (among other uses) the longer allomorph is ''preferred'' with
polysyllabic nouns, which are ''often abstract''. Furthermore, since
''many'' abstracts are non-countable, they are less likely to be
pluralized (p. 274). The use of a longer allomorph of the plural
marker with a polysyllabic form is therefore ''unexpected''. But in
view of all the qualifiers (''preferred'', ''often'', and ''many''),
the conclusion seems rather tenuous. Moreover, what does 'more
informative' mean? Information Theory (as in Claude E. Shannon's
work) seems to be standing in here for a real theory of the semantics
of grammatical meaning (Nesset and Enger suggest that iconicity
applies to affixes, but not to stems). Perhaps information theory is
appropriate here (there is something intuitively right about it), but
it would be more convincing if it were made explicit, and testable.

Michel Roch� surveys the genders assigned by modern Romance
derivational suffixes which are descended from a small set of Latin
suffixes. In some cases, the reason for the present-day gender is
clear from a historical perspective, while in others it seems
arbitrary. One such apparently arbitrary process Roch� labels 'gender
inversion', in which a single affix derives masculine nouns from
feminine ones, and vice versa.

The next paper invites a historical introduction. Years ago, when I
was a grad student, there was a theory called generative semantics,
which tried to push transformational syntax down into the sub-word
level. One of the arguments against that theory was that sub-word
constituents to fail to act as antecedents at the syntactic level. The
classical example was ''John is an orphan, and he misses them,'' where
''them'' is intended to refer to his dead parents, on the assumption
that the word 'orphan' means something like 'child whose parents have
died.' The idea that sub-word constituents cannot serve as syntactic
antecedents has been labeled the ''lexical integrity hypothesis.''

As originally conceived, the issue of independent reference (or lack
thereof) was actually applied at the sub-morphemic level. More
recently, the argument has been used to bolster the claim that noun
incorporation in polysynthetic languages must be morphological, not
syntactic, since the incorporated noun cannot have independent
reference. In the present volume, David S. Rood applies the argument
in reverse, using data from the polysynthetic language Wichita (a
nearly extinct language now spoken by a few elderly people in
Oklahoma, in the United States).

Rood first seeks to demonstrate that certain long sequences of
morphemes are words, then shows that personal affixes inside those
sequences can have independent reference. Crucially, Rood must
demonstrate that those personal affixes are not agreement morphemes
(since if they did mark agreement with a possibly covert pronoun, one
might argue that it was the pronoun that made independent
reference). Rood makes this argument on the basis of the fact that
there are no independent pronouns in the language with which the
affixes could agree. This reasoning seems to be undermined by the fact
that in Spanish (a so-called 'pro-drop' language, i.e. a language
where subject pronouns are optional), a verbal subject agreement
suffix has independent reference even when there is no subject pronoun
in a particular sentence to agree with.

At any rate, Rood argues further that noun incorporation in Wichita
must be syntactic, under the assumption that independent reference
implies syntactic independence at some level (the lexical integrity
hypothesis in reverse).

In summary, the articles in this book cover a diverse range of topics
in morphology-as I said at the beginning of this review, more than
just the ostensible topic of ''comparative morphology.'' This leads to
the question of whether you should buy the book. Given the price, and
the fact that this is a conference proceedings with a very wide range
of topics, I suspect that few linguists will want to buy this for
their personal libraries. The publisher has made a later book of
conference proceedings in the same series available on-line (viewable
for free, with individual articles downloadable and printable for a
price). That seems a reasonable thing to do for this book, too.


Hockett, Charles. 1948. ''Potawatomi''. International Journal of American
 Linguistics 14:1-10, 63-73, 139-149, 213-225.
Stump, Gregory T. 2001. Inflectional morphology : a theory of paradigm
 structure: Cambridge studies in linguistics, 93. Cambridge ; New York:
 Cambridge University Press.
Wasow, Thomas. 1977. ''Transformations and the Lexicon''. In Formal
 Syntax, eds. Peter W. Culicover; Thomas Wasow; and Adrian Akmajian,
 361-377. New York: Academic Press.


Mike Maxwell works in computational morphology at the Linguistic Data
Consortium of the University of Pennsylvania. He holds a Ph.D. in
linguistics from the University of Washington.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue