Professional Documents
Culture Documents
Trends in Linguistics
Studies and Monographs 49
Editor
Werner Winter
Mouton de Gruyter
Berlin · New York
Contemporary
Morphology
Edited by
Wolfgang U. Dressler, Hans C. Luschützky,
Oskar E. Pfeiffer, John R. Rennison
Mouton de Gruyter
Berlin · New York 1990
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter & Co., Berlin.
This volume presents selected papers from the Third International Mor-
phology Meeting held, under the auspices of the International Association
of Morphology, in Krems (Austria) from July 4 to July 7, 1988. In
contrast to the relatively small previous meetings in Veszprem (Hungary),
the Krems conference was an open congress which attracted many par-
ticipants from the immediately preceding Sixth International Phonology
Meeting. 1 In effect, the last day of the Phonology conference overlapped
with the first day of the Morphology meeting.
Accordingly our volume opens with the topic "Interface" (mainly
between morphology and phonology) where Spencer's morpholexical
approach to morphophonemics contrasts with Shapiro's semiotic ap-
proach to isomorphism of rule types. Carstairs' paper on suppletion
focuses on phonological triggers, Bayer and Lahiri's on morphosyntactic
constraints on Bengali clitics.
Within the second topic "Word formation" a wide range of subjects is
covered: Corbin presents her own rule-based model of complex words,
whereas Mötsch discusses conflicting proposals for word-structure theory
(e.g., analogy vs. constraints vs. rules). Principles of headedness in com-
pounds are in Di Sciullo's contribution on argument inheritance and in
Vogel's account of shortening of English loans in Italian. Beard argues
against morphemes as lexical items and for a strict separation of meaning
and form in derivational morphology, Dressier and Kiefer deal with the
morphopragmatics of Austrian and Hungarian diminutives, Warren with
types of phonologically modified English compounds. Zwanenburg dis-
cusses the order of compounding and inflection in French, Scalise argues
why Italian adverbs are derivational, not inflectional.
Although belonging to the third topic "Inflectional morphology and
clitics", Zwicky's presentation of his modular approach to both sub-
topics includes their relations to other modules as well; Wurzel's approach
to inflection arrives at comparable conclusions although couched in a
model of markedness. Morin adduces arguments for lexicon-internal
1. The selected papers of that conference will be published under the title Phonologica 1988
by the Cambridge University Press.
VI Preface
Topic 1: Interface 1
Josef Bayer and Aditi Lahiri
Bengali emphatic clitics in the lexicon-syntax interface 3
Andrew Carstairs
Phonologically conditioned suppletion 17
Michael Shapiro
On a universal criterion of rule coherence 25
Andrew Spencer
The advantages of morpholexical phonology 35
Claude Hagege
1. Introduction
joj and j\j are the so-called emphatic clitics in Bengali which mean
something like 'too' and [ + emphatic] respectively. Since only joj intro-
duces a new lexical meaning, we will mostly use joj for the examples, but
/i/ behaves alike in all important respects. At a first glance, /o/ and /i/
seem to adjoin as enclitics to an element of type X° which is then the
focus of the clitic, e.g.,
* We wish to thank Probal Dasgupta for his advice and Jogamaya Bayer for discussing
the Bengali data presented here. We are also grateful to Wim van der Wurff and an
anonymous reviewer for their suggestions. Realizing that proposals independently made
in Dasgupta (1984, 1987, in press) turned out to be similar in spirit to ours gave us
encouragement to pursue this work.
4 Josef Bayer and Aditi Lahiri
Although the clitic can be attached between the stem and the inflection,
it cannot be inserted between affixes. The choice is binary — either the
clitic comes right after the stem, or it must come after all the affixes are
added. Thus the forms in (6) are ungrammatical in Bengali.
(6a) *mer-e-ch-i-o-l-am
b) *mer-e-ch-i-l-o-am
It must be noted that there are no instances of categories other than
verbs in which the emphatic clitic could be "infixed". For instance, (7) is
ungrammatical. joj appears here between a noun stem and the case-
marker -ke, which is arguably an inflectional ending.
(7) *babul chele-o-ke mereche
'Babul has beaten also the BOY'
This restriction also holds for compound-like word formations. Bengali
has a verbal noun, which is derived by attaching the suffix -a to a V-
stem, e.g., por 'read' + a —• pora 'reading'. Similar to German infinitives
such as rad+fahrert 'bike riding', the verbal noun can incorporate an N°-
object into the verb stem involved, e.g., golpo por-a 'story reading'. As
(8a) below shows, joj can adjoin to the verbal noun and select its focus
inside, but as shown in (8b), it cannot adjoin to the focused Ν incorpo-
rated. For these examples, imagine a preceding discourse in which some-
one states that (s)he liked somebody's reading of stories very much.
(8a) tader [kobita por- a]- ο bhalo laglo
their poetry read-ing- too pleased-has
'(I) was pleased by also their reading of POETRY'
b) * tader [kobita-o por-a] bhalo laglo
This also holds true for the so-called dvandva constructions such as in
(9) below. The contrasting sentences with different clitic placements are
given in (10).
(9a) bap ma
father mother
'parents'
b) uttor dokkhin
north south
The fact that (3b) and (4b) are bad but not (5b) can be accounted for by
examining the minimal phonological unit that /o/ can attach to. Observe
that joj can attach to mere but not to mar. We will argue that the host
of the clitics must minimally be a phonological word. Under this view,
mere constitutes a phonological word while mar is merely a stem. In the
following discussion, we will focus on the facts which constitute evidence
for differentiating phonological words from stems, indicating also how
the clitics themselves are different from regular affixes.
Implicit in this view is the claim that affixes can be added to stems as
well as to larger units like words. The compound-like constructions (cf.
9 — 10) also have affixes added at the end ([bap ma]r 'of parents', [*bap-
er ma]). Moreover, the prosodic unit after the addition of a clitic to a
phonological word, is still a word to which an affix can be added.
3.2. Evidence that mere is indeed a phonological word and that there
is less cohesion between word + affix than stem + affix comes from re-
duplicating echo words. In Bengali, an echo word can be formed by
reduplicating the entire word except for the initial consonant which is
usually replaced by a coronal. The echo word could be interpreted as 'X
and so forth' with perhaps a slight pejorative tinge. There is a constraint,
however, in what can be reduplicated. All stem + affix constructions can
be reduplicated, but no stems alone.
(14a) Nominal forms
pa — ta 'leg'
bari — tari 'house'
chele — tele 'boy'
kobita pora — lobita pora 'poetry reading'
b) Verb forms:
mare — tare '[3rd pers] beat'
khae — tae '[3rd pers] eat'
mere — tere '[past part] beat'
3.3. Evidence that the clitics differ from superficially similar affixes can
be obtained from phonological rules of deletion and shortening. Deriv-
ative vocalic suffixes can trigger vowel shortening in stem vowels and
deletion of vowels; clitics, however, do not trigger such processes. The
examples in (16) are taken from Dasgupta (1984).
(16a) na:k 'nose'
b) naki 'nasal'
c) ra:g 'anger'
d) ragi 'angry'
e) na:k-o 'the nose too' (not: *nak-o)
0 ra:g-o 'anger too' (not: *rag-o)
k) pagol 'idiot'
1) pagli 'mad woman'
m) pogol-i 'idiot/[ + emphatic]'
To summarize, we have shown that the clitic jo/ and jij are different from
derivational suffixes and have as their minimal host category the pho-
nological word. The latter fact, however, still does not account for the
fact that sentences such as (7a) are ungrammatical. It seems unreasonable
to suggest that chele does not constitute a minimal phonological word. 4
In the next section, we will therefore explore a completely independent
line of reasoning.
Let us make the assumption that the clitics under consideration impose
quantificational properties on their morphological/syntactic domain, sim-
ilar to only and even in English. As Rooth (1985) and others before him
Bengali emphatic clitics in the lexicon-syntax interface 9
4.1. Bengali does not have a mixed system as German does. It has
exclusively postpositions. Since it is an OV-language, PPs should not lead
to island effects. At least partially, this expectation is borne out.
4.3. Quantified NPs which are adjoined to X max can freely undergo
quantifier raising because they are ungoverned. The X max to which they
belong does not count as a bounding node. The Bengali possessor-NP
appears to be adjoined, i.e., not in [SPEC, NP]-position as is the case in
English or Standard German. For lack of space we have to simplify
somewhat and propose the following structures. 6
(25a) [NP. [NP. amar baba- r] [NP. oi bari ]]
my father [gen] this house
'this house of my father's'
b) [npj [npj my father 's] [N< house]]
c) *my father's this house
Both the grammar of syntactic movement and the grammar of scope
assignment behave accordingly, as we will show next.7
(26) NP-split in the syntax
a) tumi [kon lok- ta- r bari] dekhecho?
you which man- [defj- [gen] house seen-have
'Which man's house did you see?'
b) [kon lok- ta- r] tumi [bari] dekhecho?
This is not possible in English. It would violate what has since Ross
(1967) become known as the "left branch condition". 8
(27a) [ Whose house] did you see?
b) *[ Whose] did you see [house]?
12 Josef Bayer and Aditi Lahiri
5. An apparent paradox
(31) IP
V CLITIC Γ + tense]
L+AGRJ
mer-e- ο -che
beat-en too (past, 3pers)
Notes
8. In the framework of Chomsky (1986) one could account for the left-branch condition
because adjunction to arguments (here NP) for purposes of movement is generally not
allowed in this theory.
9. Notice that post-NP even as well as too is generally acceptable, as in my FATHER even
would agree on that and my father TOO would agree on that. Under the assumption
that the English genitive -s is a clitic which is unselective with repect to its host (cf.
Zwicky — Pullum 1983), the examples in (27) cannot easily be excluded otherwise.
10. The predicate "quantified" here pertains only to the cases under consideration, namely
focusing particles and emphatic clitics.
11. On V as the head of S, see Jackendoff (1977). On the distinction of syntactic and
semantic heads, see Abney (1986). Notice that in Bengali, I cannot take VP as its
complement because this would predict that in babulke mere-o-chilam (Babulfobj] beat-
too-have [I]) Ί/we have also beaten Babul' the clitic can focus on babul. This, however,
is never possible. The clitic can only focus on the verb in such cases.
References
Abney, Steven
1986 "Functional elements and licensing" [Paper presented at the GLOW con-
ference Girona, Spain],
1987 The English noun phrase in its sentential aspect [Dissertation, MIT, Cam-
bridge, MA],
Bashir, Ε. —M. Deshpande —P. Hook (eds.)
1987 Selected papers from SALA 7. (Bloomington: Indiana University Linguistics
Club).
Bayer, Josef
1988 "Directionality of government as a locality constraint for scope assignment"
[Paper presented at the GLOW conference Budapest, Hungary],
1990 "Interpretive islands: evidence for connectedness and global harmony in
logical form", in: G. Grewendorf — W. Sternefeld (eds.).
Chomsky, Noam
1986 Barriers. (Cambridge, MA: MIT Press).
Dasgupta, Probal
1984 "Bangla emphasizers and anchors", Indian Linguistics 45: 102—117.
1987 "Sentence particles in Bangla", in: E. Bashir —M. Deshpande —P. Hook
(eds.), 4 9 - 7 5 .
[ms.]. "The word in Bangla", unpublished paper, Pune, Deccan College.
Grewendorf, G. —W. Sternefeld (eds.)
1990 Scrambling and barriers. (Amsterdam: Benjamins).
Huang, J. —R. May (eds.)
in press Logical structure and linguistic structure. (Dordrecht: Reidel).
Jackendoff, Ray_
1977 X-Syntax. (Cambridge, MA: MIT Press).
Kayne, Richard S.
1981 "On certain differences between French and English", Linguistic Inquiry 12:
349-371.
Koster, Jan
1986 Domains and dynasties. (Dordrecht: Foris).
Longobardi, Giuseppe
in press "In defense of the 'correspondence hypothesis': island effects and parasitic
constructions in logical form", in: J. H u a n g - R . May (eds.).
16 Josef Bayer and Aditi Lahiri
May, Robert
1977 The grammar of quantification [Dissertation, MIT, Cambridge, MA],
Rooth, Mats
1985 Association with focus [Dissertation, University of Massachusetts, Amherst].
Ross, John R.
1967 Constraints on variables in syntax [Dissertation, MIT, Cambridge, MA],
Szabolcsi, Anna
1984 "The possessor that ran away from home", The Linguistic Review 3:
89-102.
Zwicky, Arnold — Geoffrey Pullum
1983 "Cliticization vs. inflection: English n't", Language 59: 502 — 513.
Phonologically conditioned suppletion*
Andrew Carstairs
* Acknowledgements are as for Carstairs (1988) with the addition of Brent de Chene,
Wolfgang Dressier, Brian Joseph, Paul Kiparsky, and Steve Marlett. Faults remain my
responsibility. The work was supported by University of Canterbury English Department
research grant 87/2.
18 Andrew Carstairs
(7) Italian: finire 'finish' (and similarly for many other verbs of the
-ire conjugation): stem fin- when unstressed, finisc- (i.e., [fi'nisk]
or [fi'niJJ], according to context) when stressed (Matthews 1981).
(8) Italian irregular preterit stems: romp- 'break', mov- 'move', prend-
'take', etc., when unstressed; rupp-, moss-, pres-, etc., when
stressed.
(9) Sanskrit: asthi 'bone' (and similarly for three other nouns): stem
asthi word-finally and before consonant-initial case-number suf-
fixes; asthn- or asthan- or astha.n- (according to context) before
vowel-initial suffixes (Whitney 1889: 122, 160).
(10) French oeufs 'eggs': [0] after [z], [oef] elesewhere, e.g., les oeufs,
des oeufs, trois oeufs with [0] versus quatre oeufs, cinq oeufs, huit
oeufs with [cef] (Swiggers 1985).
But grossly suppletive examples like (1) —(5) show that this approach will
not always work. There is no escaping the conclusion that morphological
alternants may have phonologically complementary distributions without
sharing phonological representations at any level. The consequence for
phonological theory is that the existence of a phonologically conditioned
alternation does not by itself prove the existence of some synchronic
phonological process giving rise to it. This consequence is serious. Few
linguists now hold that the main function of phonological rules is to
account for morphological alternations. Even so, many phonologists still
implicitly assume that suppletive alternants can be distributed only on a
grammatical or lexical basis (like English past went versus non-past go,
or regular -es in foxes versus lexically marked -en in oxen), not on a
phonological basis, and consequently that a phonologically predictable
distribution always implies a shared phonological underlier and phono-
logical rules to generate the surface shapes. If the mistakenness of this
assumption had been recognized earlier, phonological theory might have
been spared some of its morphological digressions of the late 1960s and
early 1970s.
Despite that sobering conclusion, there still appear to be interesting
phonological restrictions on where and how phonologically conditioned
suppletion occurs. In all of (1) —(9) the phonological determinants of the
alternation are within the same word as the alternants, and we can
probably say the same of (11) if we interpret "word" as "phonological
word" or perhaps "clitic group" (Nespor — Vogel 1986). We do not find
phonologically conditioned suppletions in which the conditioning factors
reside in other words within the same phrase or clause. One apparent
exception is (10). But there is evidence that French speakers are uncom-
fortable with the distribution of alternants for oeuf described by Swiggers
(1985). In some varieties of colloquial French the [(z) 0] alternant is
maneuvered into all plural contexts, with (for example) cinq oeufs being
replaced by cinq [z] oeufs or cinq beaux oeufs (i.e. [... boz0]). It does not
matter for our present purposes whether this [z] is analyzed as part of
the preceding word, as a plural prefix, or as part of a new stem alternant
[z0]. The important point is that the distribution of the alternants is no
longer inescapably dependent on phonological factors; it can now be
analyzed instead as an 'ordinary' suppletion, based on number ([cef]
singular, [0] or [zo] plural) — a development which may have been
motivated by pressure to eliminate a phonologically conditioned supple-
tion of an unusual and highly marked kind. If so, the examples at (1) —(11)
fortify Plank's (1984) suggestion that "the locality requirement on the
triggering of morphophonemic rules be extended to rules of exponence",
20 Andrew Carstairs
affix proper {-bhih or -ah), which can plausibly be analyzed as added "at
the same time as", rather than "after", the empty intermorph which
precedes it. The conditioning involved here is quite complex, in that the
need for an intermorph and the range of alternants for it must be specified
in the lexical entry for the stem, while the choice of the individual alternant
is determined by phonological and grammatical properties of the case-
number suffix; but none of this conditioning seems to violate the spirit
of hailstone models.
If we accept that phonological factors may affect the incidence and
shape of intermorphs, we will expect to find instances where these factors
reside in the stem rather than in an accompanying affix; and we do find
them. The most plausible analysis of the Turkish example (6) involves
taking -s- as an intermorph which is inserted between vowel-final stems
and the possessive affix proper. An analysis under which -si as a whole
is simply an independent alternant of third singular possessive, alongside
-i, is equally compatible with hailstone models, because the conditioning
involved is outward rather than inward; but the fact that an intermorph
analysis is also available provides some reassurance that the resort to an
intermorph analysis in an instance like (9) is not a resort to an otherwise
unmotivated ad hoc device.
There are a variety of reasons why phonological theory has moved
away from its earlier preoccupation with morphological alternations.
Although much remains to be discovered about the incidence and impli-
cations of phonologically conditioned suppletion, I hope to have shown
that it constitutes one more reason why that move is wise.
Notes
1. The terms "phonetically uninterpretable" and "phonetically incomplete" here are not to
be confused with "underspecified", as used by, e.g., Kiparsky (1982) and Pulleyblank
(1988). Underspecification in underlying representations is intended to avoid duplication
between "morpheme structure conditions" and phonological rules. That issue is not
directly connected with the sort of phonetic incompleteness discussed here.
References
Anderson, Stephen R.
1982 "Where's morphology?", Linguistic Inquiry 13: 5 7 1 - 6 1 2 .
Arnott, D. W.
1970 The nominal and verbal system of Fula. (Oxford: Clarendon Press).
Phonologically conditioned suppletion 23
Blake, Barry J.
1979 A Kalkatungu grammar (Canberra: Australian National University)
(= Pacific Linguistics Series B, No. 57).
Carstairs, Andrew
1988 "Some implications of phonologically conditioned suppletion", Yearbook
of Morphology 1: 67 — 94.
Dimmendaal, Gerrit J.
1983 The Turkana language. (Dordrecht: Foris).
1987 "Drift and selective mechanisms in morphological change: the Eastern
Nilotic case", in: A. G. Ramat et al. (eds.) Papers from the 7th International
Conference on Historical Linguistics, (Amsterdam: Benjamins), 193 — 210.
Dressier, Wolfgang U.
1985 "On the definite Austrian and Italian articles", in: E. Gussmann (ed.) Phono-
morphology: Studies in the interaction of phonology and morphology, (Lublin:
Wydawnictwo Katolickiego Uniwersytetu Lubelskiego), 35—47.
Kiparsky, Paul
1982 "From cyclic phonology to lexical phonology", in: H. van der Hulst —Ν.
Smith (eds.) The structure of phonological representations (I), (Dordrecht:
Foris), 1 3 1 - 1 7 5 .
Lewis, Geoffrey L.
1967 Turkish grammar. (Oxford: Oxford University Press).
Marlett, Stephen Α. —Joseph P. Stemberger
1983 "Empty consonants in Seri", Linguistic Inquiry 14: 617 — 639.
Matthews, Peter
1981 "Present stem alternations in Italian", in: H. Geckeier et al. (eds.) Logos
semantikos: Studia linguistica in honorem Eugenio Coseriu, (Berlin/Madrid:
de Gruyter/Gredos), vol. 4 (Gramätica), 57 — 64.
Mohanan, Karuvannur P.
1986 The theory of lexical phonology. (Dordrecht: Reidel).
Mugdan, Joachim
1986 "Was ist eigentlich ein Morphem?", Zeitschrift für Phonetik, Sprachwissen-
schaft und Kommunikationsforschung 39: 29 — 43.
Nespor, Marina —Irene Vogel
1986 Prosodic phonology. (Dordrecht: Foris).
Plank, Frans
1984 "Romance disagreements: phonology interfering with syntax", Journal of
Linguistics 20: 3 2 9 - 3 4 9 .
Pulleyblank, Douglas
1988 "Vocalic underspecification in Yoruba", Linguistic Inquiry 19: 233 — 270.
Sauvageot, Aurelien
1951 Esquisse de la langue hongroise. (Paris: Klincksieck).
1971 L'edification de la langue hongroise. (Paris: Klincksieck).
Swiggers, Pierre
1985 "How to order eggs in French", Folia Linguistica 19: 63 — 66.
Szymanek, Bogdan
1985 "On intermorphic extensions in English and Polish", in: E. Gussmann (ed.)
Phono-morphology: Studies in the interaction of phonology and morphology,
(Lublin: Wydawnictwo Katolickiego Uniwersytetu Lubelskiego), 177 — 191.
Whitney, William D.
1889 Sanskrit grammar. (Cambridge, Mass.: Harvard University Press).
Zwicky, Arnold M.
1985 "Rules of allomorphy and phonology-syntax interactions", Journal of Lin-
guistics 21: 431—436.
On a universal criterion of rule coherence
Michael Shapiro
Nominals Verbals
[-f] proof [-v] prove
safe save
wife wive
On a universal criterion of rule coherence 27
Notes
References
Andersen, Henning
1966 Tenues and mediae in the Slavic languages: A historical investigation [Un-
published Ph. D. Thesis, Harvard University.]
1969 "A study in diachronic morphophonemics: the Ukrainian prefixes", Lan-
guage 45: 8 0 7 - 8 3 0 .
1972 "Diphthongization", Language 48: 1 1 - 5 0 .
1979 "Phonology as semiotic", in: S. Chatman et al. (eds.) A semiotic landscape
(The Hague: Mouton), 3 7 7 - 3 8 1 .
1980a "Morphological change: towards a typology", in: J. Fisiak (ed.), Recent
developments in historical morphology (The Hague: Mouton), 1 —50.
1980b "Summarizing discussion: Introduction", in: T. Thrane et al. (ed.) Typology
and genetics of language ( = Travaux du Cercle Linguistique de Copenhague
20) (Copenhagen: Linguistic Circle of Copenhagen), 197 — 210.
Anttila, Raimo
1969 Uusimman äännehistorian suunnasta ja luonteesta [About the trends and
character of the newest historical phonology] ( = Publications of the Pho-
netics Department of the University of Turku 5).
1972 An introduction to historical and comparative linguistics. (New York: Mac-
millan).
1977 Analogy. (The Hague: Mouton).
1980 "Language and the semiotics of perception", in: I. Rauch —G. F. Carr (eds.)
The signifying animal: The grammar of language and experience (Blooming-
ton: Indiana University Press), 263 — 283.
Battistella, Edwin
1985 "Markedness isomorphism as a goal of language change: the spread of
periphrastic do in English", Lingua 65: 307 — 322.
1990 Markedness: The evaluative superstructure of language. Albany: S U N Y
Press.
Dressier, Wolfgang U.
1985 Morphonology: the dynamics of derivation (Ann Arbor, Mich.: Karoma
Press).
1987 "Semiotische Grundlagen einer Theorie der Natürlichen Phonologie und
Morphologie", in: Semiotica Austriaca, herausgeg. v. J. Bernard (Wien:
Österreichische Gesellschaft für Semiotik), 165 — 172.
Dressier, Wolfgang U. — Mayerthaler, Willy — Panagl, Oswald — Wurzel, Wolfgang U.
1987 Leitmotifs in natural morphology. (Amsterdam: Benjamins).
Givön, Talmy
1985 "Iconicity, isomorphism, and non-arbitrary coding in syntax", in: J. Haiman
(ed.), 1 8 7 - 2 1 9 .
Haiman, John
1985 Natural syntax. (Cambridge: Cambridge University Press).
Haiman, John (ed.)
1985 Iconicity in syntax. (Amsterdam: Benjamins).
Hjelmslev, Louis
1938 [1970] "Essai d'une theorie des morphemes", Essais linguistiques? (Kobenhavn:
Nordisk Sprog- og Kulturforlag), 152 — 164.
Horn, Wilhelm
1954 Laut und Leben: Englische Lautgeschichte der neueren Zeit (1400—1950),
II, edited by Μ. Lehnert (Berlin: Verlag der Wissenschaften).
34 Michael Shapiro
Itkonen, Esa
1986 "Form-meaning isomorphism, or iconicity, in diachronic linguistics (and
elsewhere)", in: M. Remmel (ed.) Symposium on formalization in historical
linguistics: Summaries, (Tallinn: Academy of Sciences of the Estonian SSR),
38-46.
Jakobson, Roman
1932 "Zur Struktur des russischen Verbums", in: Charisteria Guilelmo Mathesio
oblata (Praha: Prazsky Lingvisticky Krouzek), 74 — 84 [reprinted in Jakob-
son 1971: 3 - 1 5 ] ,
1965 "Quest for the essence of language", Diogene 51: 21—37 [reprinted in
Jakobson 1971: 345-359],
1971 Selected writings, II: Word and language. (The Hague: Mouton).
Mayerthaler, Willi
1981 Morphologische Natürlichkeit. (Wiesbaden: Athenaion).
1988 Morphological naturalness [translated by Janice Seidler]. (Ann Arbor, Mich.:
Karoma Press).
Plank, Frans
1979 "Ikonisierung und De-Ikonisierung als Prinzip des Sprachwandels", Sprach-
wissenschaft 4: 121 — 158.
1981 Morphologische (Ir-) Regularitäten. (Tübingen: Gunter Narr).
Shapiro, Michael
1969 Aspects of Russian morphology: A semiotic investigation. (Cambridge, Mass.:
Slavica).
1971 "Markedness and Russian stress", Linguistics 72: 61 —77.
1974a "Morphophonemics as semiotic", Acta Linguistica Hafniensia 15: 29—49.
1974b "Tenues and mediae in Japanese: A reinterpretation", Lingua 33: 101 —114.
1976 Asymmetry: An inquiry into the linguistic structure of poetry. (Amsterdam:
North-Holland).
1980 "Russian conjugation: theory and hermeneutic", Language 56: 67 — 93.
1983 The sense of grammar: Language as semeiotic (Bloomington: Indiana Uni-
versity Press).
1985 "Teleology, semeiosis, and linguistic change", Diachronica 2: 1—34.
1987 "Sapir's concept of drift in semiotic perspective", Semiotica 67: 159 — 171.
Trubetzkoy, Nikolai S.
1962 Grundzüge der Phonologie 3 (Göttingen: Vandenhoeck & Ruprecht).
Vachek, Josef
1976 "The opposition of voice and tension in modern English paired consonants",
in: J. Vachek, Selected writings in English and general linguistics, (Praha:
Academia), 364 — 372.
Wurzel, Wolfgang U.
1984 Flexionsmorphologie und Natürlichkeit: Ein Beitrag zur morphologischen
Theoriebildung (= Studia Grammatica 21) (Berlin: Akademie).
1985 "Zur Determiniertheit morphologischer Erscheinungen: Ein Zwischenbe-
richt", Acta Linguistica Academiae Scientiarum Hungaricae 35: 151 — 168.
The advantages of morpholexical phonology
Andrew Spencer
Notes
1. I am inclined to say that the postcyclic and postlexical components take the form of
output constraints governed by universal and language-particular principles such as
syllable structure, metrical structure, autosegmental structure, and so on. This would
make phonology look rather like syntax on the "principles-and-parameters" approach.
2. The strict-cyclicity condition also prevents a rule A which counterfeeds a rule Β from
being fed by Β on a subsequent cycle. In a morpholexical framework, this effect can
only be reconstructed using complex diacritics. It seems to me undesirable that a principle
of universal grammar should have the effect of maintaining rules in a marked order, so
I am happy with this situation.
References
Fodor, Jerrold
1983 The modularity of mind. (Cambridge, MA: MIT Press).
Kiparsky, Paul
1985 "Some consequences of lexical phonology", Phonology Yearbook 2: 83 — 136.
1988 "Two approaches to suppletive allomorphy" [Paper presented at the 3rd
International Morphology Meeting, Krems, Austria].
Lieber, Rochelle
1982 "Allomorphy", Linguistic Analysis 10: 2 7 - 5 2 .
Spencer, Andrew J.
1988a "Arguments for morpholexical rules", Journal of Linguistics 24: 1 —30.
1988b "Morpholexical rules and lexical representations", Linguistics 26: 619 — 640.
ms "Morpholexical phonology".
Sproat, Richard
1985 On deriving the lexicon [Dissertation, MIT, Cambridge, MA].
Topic 2: Word formation
Associativite et stratification dans la representation
des mots construits
Danielle Corbin
2.1.1 Four che tie a pour structure morphologique (2a), done pour sens
predictible (2b)9, mais pour sens atteste (2c):
(2a) [[fourche]N(ette)af]N
b) SP « petite fourche » (cf. flechette, hachette, serpette, etc.)
c) SA «1° ustensile de table (d'abord ä deux, puis a trois, quatre
dents), dont on se sert pour piquer les aliments »
(3a) [[[[public]A]N(iste)af]A]N
b) Spcr « N en relation avec le public»
c) SPsp « Ν [ +hum] specialiste du public »
d) SA «3° [...] agent de publicite»
sur le nom publicite, dont le suffixe -ite a subi une troncation devant
-isten.
Dans cet exemple, c'est l'homonymisation de la forme publiciste qui
permet de conserver Γ associativite.
(5a) [[X]v(oir)af]N
effet comparables (9a), et on pourrait penser que les suffixes -aie et -ierj
-iere appartiennent tous deux au paradigme morphologique associe ä la
regle de construction de mots qui construit, sur des bases nominales
designant des vegetaux, des noms dont le sens predictible peut se para-
phraser par «lieu plante de [nom de base]». Mais, si ces proprietes
paraissent bien etre celles de -aie, elles ne conviennent pas ä -ierj-iere,
dont on peut dire tres brievement qu'il construit fondamentalement des
adjectifs ä sens relationnel sur des bases nominales 20 . La structure mor-
phologique (9b) et les sens predictibles de sapiniere (9c) ne sont done pas
directement deductibles de l'apparence:
2.5. Conclusion
Pour conclure cette partie, je rappellerai les principes essentiels qui y ont
ete defendus:
— A toute structure morphologique est associee une interpretation se-
mantique compositionnelle et un seul sens predictible construit par la
regie, qui ne correspond pas necessairement au sens atteste.
— S'il apparait une distorsion entre la structure morphologique et l'in-
terpretation semantique, ou bien la distorsion est reelle (exemples (8),
(10) ä (12)), et des dispositifs modeliques (paradigme morphologique,
regies mineures d'allomorphie et de troncation) permettent de ne pas
remettre en cause l'associativite, ou bien la distorsion n'est qu'apparente
(exemples (2) ä (7), (9)), et il faut deplacer l'analyse.
— II n'y a pas de bijection entre l'ensemble des affixes et celui des regies
de construction de mots.
— Ce n'est done pas l'unite de forme affixale qui delimite une operation
derivationnelle, mais l'association entre un rapport categoriel, une ope-
ration semantique et un Operateur morphologique (affixe, conversion)
appartenant ä un paradigme dont le nombre d'unites est variable.
— II en decoule que la morphologie associative que je preconise est
fondamentalement homonymique, dans la mesure ou le critere d'identi-
fication d'un element (affixe ou mot construit) est l'association de pro-
prietes formelles et de proprietes semantiques. Au contraire, les morpho-
logies dissociatives sont polysemiques.
REGLES
D'INSERTION
LEXICALE
4. Pour conclure
Notes
12. Le signe 0 precede des formes ou des sens grammaticaux quoique non attestes, par
opposition au signe * qui precede des formes ou des sens agrammaticaux.
13. Cette troncation est representee dans le schema (4) par les traits < + T>, signifiant que
-ite est predispose ä subir une troncation dans un contexte approprie, et <T + >,
signifiant que -iste est predispose a provoquer la troncation d'un suffixe predispose a
la subir. Voir sur ce point Corbin (1987: 341 —370).
14. Certains sont developpes dans Corbin (1987: 247 — 248).
15. Encore attestee dans le LXXe et son successeur le GLE avec le sens «Qui peut ou doit
etre invalide», mais plus dans le GDEL, ni aucun dictionnaire frangais strictement
contemporain. Validable est atteste dans le GRLF.
16. Ma proposition differe sur plusieurs points de ce que Zwanenburg (1984) appelle «types
de derivation »:
— Peuvent etre inclus dans le meme paradigme morphologique des prefixes, des
suffixes, la conversion, alors que Zwanenburg n'inclut dans ses «types » que des suffixes.
— Zwanenburg propose qu'ä l'interieur d'un type donne, les suffixes soient exclusifs
Fun de l'autre, c'est-ä-dire que son dispositif lui sert ä expliquer certains blocages, alors
que je ne pense pas que ce principe de blocage soit adequat: si les procedes appartenant
ä un meme paradigme obeissent aux memes contraintes, ils peuvent s'appliquer ä une
meme base remplissant ces contraintes (en temoignent par exemple, dans le lexique
atteste: abonnir / bonifier; clarifier / eclaircir, etc., et dans le lexique possible °laidifier
face ä enlaidir, °sauvagiser face ä ensauvager, etc.).
— Zwanenburg raisonne dans un cadre dissociatif, moi dans un cadre associatif.
17. Certaines de ces contraintes sont explicitees dans Corbin (1987: 450).
18. Ce qui s'oppose ä l'hypothese «One affix, one rule» formulee notamment par Aronoff
(1976).
19. Le genre feminin de sapiniere n'est pas plus predictible que celui de saliere compare a
poivrier.
20. Par exemple, un argument contre l'assimilation du suffixe -aie et du suffixe -ier/iere est
que, contrairement ä ce qui se passe pour -aie, on trouve des noms de lieu en -ierj-iere
sur des noms de base ne designant pas des vegetaux {sablonniere («lieu d'ou Ton extrait
le sable»), heronniere («endroit amenage pour l'elevage des herons»), etc.). On trouvera
une argumentation plus detaillee concernant le suffixe -ierj-iere dans Corbin — Corbin
(ä paraitre).
21. Voir par exemple Jackendoff (1975), Zwanenburg (1982, 1984, 1987: 64).
22. Cette troncation est recurrente, comme le montre l'exemple finalite —> finaliste,, au sens
« Qui croit ä Taction des causes finales et, en general, ä la finalite comme explication
de l'univers » (sens predictible specific: « Partisan de la finalite »).
23. Ce verbe, non atteste aujourd'hui, etait atteste au XIVC siecle (Greimas, 1968).
24. Si ce nom est derive d'anxieux. Je simplifie ici le traitement de ce mot. Pour des details,
voir Corbin (1987: 505-508).
25. Extrait de Corbin (1987: 417).
26. Voir sur ce point Corbin (1987: 476-504).
27. Et, si necessaire, sens predictible specifie.
28. Je m'oppose sur ce point ä des propositions recentes comme celle de Roeper (1988) qui
propose d'inserer des affixes, qui selon lui sont des «tetes», sous des nceuds syntag-
matiques.
29. A l'exception, remarquable, de Botha (1988).
58 Danielle Corbin
References bibliographiques
Aronoff, Mark
1976 Word formation in generative grammar (Linguistic Inquiry Monograph 1).
(Cambridge, Mass.: MIT Press).
Booij, Geert
1979 "Semantic regularities in word formation", Linguistics 17: 985—1001.
Botha, Rudolf P.
1988 Form and meaning in word formation. A study of Afrikaans reduplication.
(Cambridge: Cambridge University Press).
Corbin, Danielle
1987 Morphologie derivationnelle et structuration du lexique, 1—2. (Tübingen:
Niemeyer).
1988 «Une hypothese a propos des suffixes -isme, -ique, -iste du frangais: la
troncation reciproque », in: R. Landheer (ed.) Aspects de linguistique fran-
Qaise. Melanges Q. I. M. Mok (Amsterdam: Rodopi), 63 — 75.
a paraitre «Contre une transposition de la theorie X a la morphologie derivation-
nelle», Acta Linguistica Hungarica 37.
Corbin, Danielle —Corbin, Pierre
ä paraitre «Pour un traitement unifie du suffixe -ier/-iere», Lexique 10.
Dressier, Wolfgang U.
1987 "Word formation (WF) as part of natural morphology", in: W. U. Dressier
et al., Leitmotifs in natural morphology (Amsterdam: Benjamins), 99 — 126.
<GDEL>
1982—1985 Grand dictionnaire encyclopedique Larousse. (Paris: Larousse), 10 vol.
<GLE>
1960 — 1964 Grand Larousse encyclopedique en dix volumes. (Paris: Larousse), Supple-
ments 1968, 1975.
Greimas, Algirdas Julien
1968 Dictionnaire de l'Ancien Frangais. (Paris: Larousse).
<GRLF>
1985 Le Grand Robert de la langue Frangaise. Dictionnaire alphabetique et ana-
logique de la langue frangaise de P. Robert, 2e ed. entierement revue et
enrichie par A. Rey. (Paris: Dictionnaires Le Robert), 9 vol.
Jackendoff, Ray
1975/1977 « Regularites morphologiques et semantiques dans le lexique », in: M. Ronat
(ed.) Langue. Theorie generative etendue (Paris: Hermann), 65 — 108.
<LXXe>
1928 — 1933 Larousse du XX" siede en six volumes publie sous la direction de P. Auge
(Paris: Larousse).
Lieber, Rochelle
1980/1981 On the organization of the lexicon [Dissertation, MIT] (Bloomington: In-
diana University Linguistics Club).
<PR86>
1986 Le Petit Robert. Dictionnaire alphabetique et analogique de la langue fran-
gaise. (Paris: Le Robert) [1(1967), 2(1977)].
Puchulu, Agnes
1987 Etude des suffixes -aire dans le cadre de la morphologie derivationnelle
[Memoire de maitrise <dactylographie>, Universite de Lille III].
1988 Propositions pour une interpretation semantique des adjectifs denominaux
[memoire de DEA <dactylographie>, Universite de Lille III].
Associative et stratification 59
Roeper, Thomas
1988 «Arguments implicites et la relation Tete-Complement», Lexique 7:
121-141.
Selkirk, Elizabeth
1982 The syntax of words (Linguistic Inquiry Monograph 7). (Cambridge, Mass.:
M I T Press).
Toman, Jindrich
1987 Wortsyntax. Eine Diskussion ausgewählter Probleme deutscher Wortbildung.
(Tübingen: Niemeyer).
Zwanenburg, Wiecher
1982 « Types de derivation comme universaux », Recherches de linguistique fran-
ςaise d'Utrecht (Utrecht: Institut d'Etudes frangaises), 57 — 66.
1984 "Word formation and meaning", Quaderni di Semantica 5: 130 — 142.
1987 «Structure et forme des adjectifs denominaux», Recherches linguistiques
d'Utrecht·. 5 5 - 8 0 .
Formal relations and argument structure*
Anna-Maria Di Sciullo
0. Introduction
We will assume that the external argument is the head of the argument
structure. But how can we define the head with respect to the argument
structure in morphological objects which contain more than one argument
structure? A possible definition is found in (1), if we define feature F as
being the argument structure:
(1) Head F (read: head with respect to the feature F)
The head F of a word is the rightmost element of the word marked
for feature F. (Di Sciullo-Williams 1987: 26)
In English, (1) holds for a large class of morphological objects. There-
fore, in (2), for instance, the suffix -ed has no argument feature, and thus,
* This work was conducted as part of the Argument Structure Project at the Linguistics
Department of the Universite du Quebec ä Montreal. Support for the project was
provided in part by the SSHRCC grant no. 4 1 0 - 8 6 - 0 7 6 0 .
62 Anna-Maria Di Sciullo
even if it is the rightmost element of the word, it is not the head with
respect to that feature. Of the two constituents with the relevant features,
it is the rightmost which is the head with respect to the argument structure.
Thus, the external argument of formalize is the external argument of the
verbal suffix -ize and not the external argument of the adjective formal?
(2) formal\A ize]v ed\w (x, y)
ω (*, y)
There are languages where (1) is empirically false. Lieber (1988), for
instance, reports cases of Vietnamese and Breton compounds, as well as
affixed forms in Tagalog, where the head, with respect to the categorial
features, is on the left. Moreover, (1) may fail within the same language.
Consider the fact that, in Italian, the head with respect to the category,
and for our purpose the head with respect to the argument structure, is
on the right in affixed forms and generally on the left in compounds.
2. Control relation
syntactic control, since in the latter case it is only the external argument
that is controlled. The -e (-ee) nominals show that the internal argument
can also be controlled, which is not the case in syntax.
3. Binding relation
4. Function-composition relation
In suffixed forms, the head can functionally combine with the non-head,
as in (13). As defined in Di Sciullo —Williams (1987), when function
composition occurs, the argument of the head becomes the external
argument of the whole, and the argument of the non-head becomes the
internal argument of the whole.
lazi ness
(13a) y)
construct ion (r, x, y)
(*, y) W
the construction of the city by the enemy
66 Anna-Maria Di Sciullo
(15a) * 0 , r)
b) *(x, y, r)
We want to explain this fact and reduce the reference to specific
functors in the system, by proposing that function composition is the
general case for the combination of arguments in suffixed forms, and
that its non-occurrence is attributed to (16):
(16) Specificity
The semantic nature of the external predicate-argument structure
variable of a word is specific to its category.
The semantic nature of an external predicate-argument structure var-
iable is fixed at the lexical-conceptual structure level. For a nominal such
as construction, for example, the r predicate argument structure variable
is projected from a semantic concept which is "the process r" or "the
result r"; for a derived adjective such as fearful, the y variable is projected
from a semantic concept which is "the affected thing y", and so on. 5
According to (16), there are semantic types of arguments which cannot
be the external arguments for certain categories. This is the case for r,
which cannot be the external argument for words other than nouns.
Moreover, r cannot be the internal argument for any other category, as
it is specific to the external argument of nouns. Thus, specificity blocks
function composition in (14), and prevents r from being the internal
predicate argument structure variable of adjectives and verbs.
Formal relations and argument structure 67
5. Summary
Notes
References
Chomsky, Noam
1981 Lectures on government and binding. (Dordrecht: Foris).
1986 Knowledge of language. (New York: Praeger).
Di Sciullo, Anna-Maria
1988 "Argument satisfaction, lexical and syntactic". [Paper presented at the
GLOW 1988 workshop on the interaction of the lexicon with other modules
of the grammar, Budapest],
Di Sciullo, Anna-Maria—Williams, Edwin
1987 On the definition of word (Linguistic Inquiry Monograph no. 14). (Cam-
bridge, Mass.: MIT Press).
Grimshaw, Jane
1987 "Psych verbs and the structure of argument structure". [Ms. Brandeis
University],
1988 "Adjuncts and argument structure", Lexicon Project Working Papers 21.
(Cambridge, Mass.: MIT Occasional Paper no. 36).
Hale, Ken —Keyser, Samuel Jay
1987 "Explaining and constraining the English Middle". [Ms. MIT],
Higginbotham, James
1985 "On semantics", Linguistic Inquiry 16: 547 — 593.
Jackendoff, Ray
1983 Semantics and cognition. (Cambridge, Mass.: MIT Press).
Levin, Beth — Rappaport, Malka
1987 "Non-agentive -er nominals: probe into argument structure". [Ms. MIT],
Lieber, Rochelle
1980 On the organization of the lexicon. [Ph. D. dissertation, MIT].
1988 "Phrasal compounds in English and the morphology-syntax interface",
Papers from the Parasession on agreement in grammatical theory, Chicago
Linguistic Society, PCLS 24-11.
Rappaport, Malka —Levin, Beth
1986 "What to do with theta roles", Lexicon Project Working Papers 11, MIT.
Williams, Edwin
1981 "Argument structure and morphology", The Linguistic Review 1: 81 — 114.
Zubizarreta, Maria-Louisa
1987 Levels of representation in the lexicon and in the syntax. (Dordrecht: Foris).
Austro-Hungarian morphopragmatics
Wolfgang U. Dressier and Ferenc Kiefer
ugliest'. In contrast to the derived forms of nagy 'big', the derived forms
of csünya 'ugly' are not used relatively, the things compared are presup-
posed to be ugly.
What is the difference between the excessive and the superlative? Let
us have a look at the pairs legeslegnagyobb — legnagyobb and leges-
legcsünyäbb — legcsünyäbb. The excessive legeslegnagyobb can only be
used of things which are big, i.e., it introduces a presupposition in contrast
to the corresponding superlative which does not. On the other hand, the
difference between legcsünyäbb and legeslegcsünyäbb cannot be purely
presuppositional since the superlative already introduces the relevant
presupposition.
One important function of the excessive is, then, to introduce a se-
mantic presupposition in cases where it is absent in the superlative. In
order to find out what the pragmatic difference, if any, is between the
superlative and the excessive, let us then concentrate on the pair leges-
legcsünyäbb — legcsünyäbb. Compare the following sentences
(1) Εζ α legcsünyäbb rajz, amit valaha lättam.
'This is the ugliest drawing I have ever seen'.
means of the property in question in an absolute sense. That is, the very
biggest house cannot refer to a small house since the very biggest expresses
the biggest possible. In other words, whenever the semantic presupposi-
tion is not present in the case of the superlative, it has to be introduced
via the excessive. (4) formulates the discourse function of the excessive
which is in principle independent of semantic presuppositions. One may
thus claim that the excessive introduces a discourse function which clearly
belongs to morphopragmatics, it is neither lexically conditioned nor a
property of morphosemantics.
The German correspondence of the Hungarian agglutinating prefix
legesleg- is aller- (originally a genitive plural of all- 'all') preposed to a
superlative, e.g., in Wissen Sie schon das Aller #neu + este? 'Do you know
the most recent news of all?' (neu 'new' superlative neu + est-). Although
less grammaticalized than its Hungarian equivalent (as is typical for
compounding, vs. derivation, cf. Lehmann 1982), it seems to share the
same pragmatic and semantic effects: it expresses the absolutely highest
possible degree of a property and it is only compared with items which
do have this property to a high degree; it is used for emphasis and for
impressing the hearer. Therefore it is often found at the end of a text
chunk, as in the "Auszählreim" (nursery rhyme):
(5) Du, bist schön und duj bist schön und duk die aller schönste.
'YoUi are beautiful and youj are beautiful and youk the most
beautiful of all'.
If aller- is prefixed recursively (e.g., die alleraller allerschön + sie), then
the respective form must be used at the end of the respective text chunk.
An exception is — for diagrammatic reasons — aller φ erst 'first and
foremost' which usually introduces the enumeration of things to be done
first, cf. the Ancient Greek equivalent pän#prötos in Homer (Ilias VII
324, IX 93, XVII 568, Odyssey IV 577, IV 780).
Similarly the excessive can be used as a corrective device in discourse:
(6a) Das ist sehr schlimm.
'That is very bad',
b) Aber das aller φ schlimmste ist, dass X ...
'But the very worst of all is that X ...'.
Emphasis and interactive orientation of expressing the absolutely high-
est possible degree also explain the use of the German excessive as
equivalent to Neo-Latin (and Hungarian!) superlatives in the following
feudal terms:
72 Wolfgang U. Dressler and Ferenc Kiefer
Diminutives
Let us now turn to the Hungarian "diminutive" suffix -i. Nouns are
formed from nouns by means of the suffix -i in two different ways:
(a) -i is added to the truncated form of the initial noun, cf. csok-i
(from csokoläde 'chocolate'), dir-i (from direktor 'director'), ov-i (from
ovoda 'kindergarten'), fagy-i (from fagylalt 'icecream'), zong-i (from zon-
gora 'piano'), szak-i (from szaktärs 'fellow-worker'), nyug-i (from nyu-
galom 'quiet').
(b) the suffix -i is added to the stem of the noun: läb-i (from läb 'foot'),
comb-i (from comb 'thigh'), has-i (from has 'belly'), hus-i (from hüs 'meat'),
ägy-i (from ägy 'bed'). As can be gathered from the examples cited, the
class of nouns in (a) is quite heterogeneous. In contrast, the nouns (b)
seem to belong to the semantic classes of body parts, baby food, and
objects which are important in the small child's life.
The suffix -i has no semantic meaning (it does not denote smallness):
both (a) and (b) are nouns with the same denotations as the corresponding
simplex nouns. If csoki and csokoläde 'chocolate' and läbi and läb 'foot'
are semantically identical, is there any pragmatic difference between the
corresponding forms? The forms in (a) presuppose a certain degree of
intimacy and have often a jocular touch. They are used in school slang
Austro-Hungarian morphopragmatics 73
This clash occurs only if the referent has a more than normal size,
whereas a normal-sized referent may be easily expressed with a diminu-
tive. This also explains the difficulty of accepting potential diminutives
such as Ries-erl 'giant-ie'. This example illustrates also that the presence
of the possessive 'my' in the microcontext conveys the connotation of
endearment. However, the macrocontext may add negative connotations
as in the oral narrative of a Jew who was forced to leave Vienna in 1938:
(12) So hab ich mein Koffer-l gepackt und bin weg.
'Thus I packed my suitcas-ie and went off.'
{Koffer-l is rather a haplology for *Koffer-erl than an instance of an
/-diminutive.)
Consider next the diminutive suffixes -kaj-ke and -cskaj-cske in Hun-
garian. The former occurs mainly after noun stems ending in I, r, n, m,
ny, s or ό, ö, and i. That is, -kaj-ke also occurs after the suffix i: läbi-ka
'small foot', combi-ka 'small thigh', husi-ka 'little meat', etc. Its primary
meaning is diminutive: level-ke 'small letter', asztal-ka 'small table',
leäny-ka 'little girl', tänyer-ka 'small plate', kanal-ka 'small spoon',
väros-ka 'small town', szekreny-ke 'small wardrobe' (some of these are
lexicalized forms). The same with -cskaj-cske, e.g., felhö-cske 'small
cloud', bokr-ocska 'small bush', vödr-öcske 'small bucket', läb-acska 'little
small foot', vaj-acska 'little butter', terd-ecske 'small/little knee'. The
diminutive meaning is exclusive except for the following semantic classes:
body parts, food, and familiar objects (this latter class is yet to be defined
more precisely). In the case of the above-mentioned semantic classes the
diminutive may acquire a purely pragmatic function when it is transferred
to the adult world (in the language of love, or if used jokingly in certain
speech situations). In such cases denotative diminutive meaning disap-
pears. That is, szäjacska 'small mouth', for example, will no longer mean
small mouth but, say, a mouth which I love. Or, läbika may even be a
big foot, in the language of love it means something like a foot to be
caressed. In nursery talk, it may still keep the diminutive meaning while
at the same time expressing affection. That is, depending on the speech
situation, the diminutive suffixes may acquire either a purely pragmatic
meaning or, in addition to the denotative meaning, certain pragmatic
overtones.
To sum up, it would seem that both Austrian and Hungarian dimin-
utives prove essentially the same thing, i.e., that diminutives may have
purely pragmatic functions. All morphological rules which contain a
pragmatic variable in the description of their meaning are morphoprag-
matically relevant.
76 Wolfgang U. Dressler and Ferenc Kiefer
References
Beard, Robert
1986 On the separation of derivation from morphology: toward a lexeme I
morpheme-based morphology. (Bloomington: Indiana University Linguistics
Club).
1987 "Morpheme order in a lexeme/morpheme-based morphology", Lingua 72:
1-44.
Beneveniste, Emile
1948 Noms d'agent et noms d'action en indo-europeen. (Paris: Adrien-
Maisonneuve).
Brandstctter, Alois
1964 "Semantische Studien zum Diminutiv im Mittelbairischen", Zeitschrift für
Mundartforschung 30: 335 — 351.
Austro-Hungarian morphopragmatics 77
3. As to the first implication mentioned above, that is, the type of rules,
adopted to account for properties of compounds and derived words, the
most general device are rewriting rules of the type
X° Y° x°
80 Wolfgang Mötsch
There are more refined versions of this rule scheme, but they may be
neglected for the purpose of my argumentation.
Rewriting rules of this form have some general properties in common
with phrase-structure rules of the X type. They are, however, restricted
in some fundamental respects:
(i) Only a limited subset of lexical categories may enter into these rules.
(ii) Only zero-bar categories may appear in word structures, that is, no
phrasal structures are allowed within word structures.
(iii) Word-structure rules generate binary structures.
(iv) One of the constituents of a word structure is identical with the
dominating category. In our version of the rule scheme it is the
right-hand constituent, which may be defined as the head of the
construction.
All predictions of this rule scheme are empirically more or less proble-
matic. There are many counterexamples to each of them. This, however,
is not necessarily a sufficient argument against the theory, because it may
turn out to be an interesting generalization which represents the unmarked
properties of word structure. This presupposes that deviating cases might
be interpreted as marked exceptions. Unfortunately, a proper theory of
markedness which excludes purely intuitive judgements on markedness is
not available.
Despite the lack of a theory of markedness, I think, there are some
problematic predictions which — as far as I can see — may not be
resolved by markedness assumptions. One general problem of word-
structure theories is the fact that some of the presumably marked cases
cannot be described other than in an ad hoc way. Difficulties arise with:
(i) The prominent role of affixation among other kinds of devices in-
volved in morphological processes.
(ii) Synthetic compounds and other types of words containing phrases,
cf. German Arbeit-nehm-er, Saure-Gurken-Zeit
English car driv-er, peace lov-ing,
generative-grammar approach.
(iii) The restriction to binary branching.
Word-structure rules of the above-mentioned type generate hierarchical
structures with lexical categories as final elements and as dominating
nodes. It is assumed that the lexical insertion rule applies to these
structures. This, in turn, presupposes that affixes — like roots — are to
be analyzed as lexical entries. Affixes, of course, are considered to be
Problems of word structure theories 81
[VP - V NP]J
However, whereas verbs requiring NPs as direct objects are an important
syntactic generalization, no class of "adjective suffixes requiring nouns
as co-constituents" is relevant in word structures. Restrictions concerning
the base word in derivational processes are a purely individual property
of affixes, not of classes of affixes.
A crucial problem for all sorts of word-structure theories evolves from
synthetic compounds which may be observed in many languages. In
Problems of word structure theories 83
English and German, there are many highly productive types of synthetic
compounds, that is, processes which attach an affix to a phrasal base.
However, this phrasal base is restricted to a concatenation of words, that
is, no normal phrase is allowed. As far as I can see, no word-structure
approach has a sufficient solution to the problem of how restricted
phrases may enter into word-structure trees. Even if we agree with an
analysis like
gle
References
Chomsky, N o a m A.
1970 "Remarks on nominalization", in: R. A. Jacobs — P. S. Rosenbaum (eds.)
Readings in English transformational grammar (Waltham, MA: Ginn), 184 —
221.
Dressler, Wolfgang U.— Willi Mayerthaler —Oswald Panagl — Wolfgang U. Wurzel
1987 Leitmotifs in natural morphology. (Amsterdam: Benjamins).
Jackendoff, Ray
1975 "Morphological and semantic regularities in the lexicon", Language 51:
639-671.
Lieber, Rochelle
1980 On the organisation of the lexicon (Cambridge, M A ) [MIT dissertation],
Mötsch, Wolfgang
1988 "On inactivity, productivity, and analogy in derivational processes", Lin-
guistische Studien 179: 1—30.
Constraints on the Italian suffix -mente*
Sergio Scalise
It has been maintained that the suffix -ly in English, being fully produc-
tive, could be considered "inflectional". 1 In this paper I intend to dem-
onstrate that the Italian suffix -mente, which may be viewed as equivalent
to English -ly, although very productive, is not "fully" productive. On
the contrary, it is subject to a number of subtle semantic and morpho-
logical constraints which cannot be interpreted as "inflectional". It is
therefore claimed that -mente (and consequently -ly) are derivational
affixes.
Constraints on simple adjectives and complex adjectives (compounded
and derived) will be examined in that order.
1. Negative Constraints
(10)
*
1. -accio golosaccio — • *golosacciamente 'greedy'
*
2. -acco polacco — • *polaccamente 'Polish'
*
3. -ace seguace — > *seguacemente 'follower'
*
4. -aceo cartaceo — • *cartaceamente 'papery'
Yj* 5. -ale postale — v *postalmente 'postal'
naturale — • naturalmente 'natural'
Y/* 6. -aneo cutaneo — • *cutaneamente 'cutaneous'
temporaneo — V temporaneamen te 'temporary'
*
7. -ano isolano — • *isolanamente 'insular'
Yj* 8. -ante ignorante —• *ignorantemente 'ignorant'
brillante — • brillantemente 'brilliant'
*
9. -ardo savoiardo *savoiardamente 'Savoyard'
YJ* 10. -are immobiliare — • *immobiliarmente 'immovable'
militare — • militarmente 'military'
Yj* 11. -ario funerario — • *funerariamen te 'funerary'
arbitrario — • arbitrariamente 'arbitrary'
*
12. -asco comasco — > *comascamente 'from Como'
*
13. -astro biancastro — • *biancastramente 'dirty white'
*
14. -ate arpinate — • *arpinatemente 'from Arpino'
Yj* 15. -atico asmatico — • *asmaticamente 'asthmatic'
dogmatico — • dogmaticamente 'dogmatic'
Yj* 16. -ato stellato — • *stellatamente 'starry'
dettagliato — • dettagliatamente 'detailed'
Yj* 17. -bile utilizzabile — • * u tilizzabilmen te 'utilizable'
amabile — • amabilmente 'lovable'
*
18. -eccio mangereccio — • *mangerecciamen te 'eatable'
*
19. -ello miserello — • * miserellamente 'quite poor'
*
20. -eno madrileno — • *madrilenamente 'Madrilenian'
Yj* 21. -ente adiacente *adiacentemente 'adjacent'
precedente — • precedentemente 'previous'
*
22. -eo europeo *europeamente 'European'
*
23. -eo argenteo *argenteamente 'silvery'
*
24. -escente fosforescente *fosforescen temen te 'phosphorescent'
Y/* 25. -esco cardinalesco —> * cardinalescamente 'cardinal'
principesco — • principescamen te 'princelike'
*
26. -ese bolognese — » *bolognesemente 'Bolognese'
*
27. -etto furbetto —>· *furbettamente 'cunning'
Y/* 28. -evole pieghevole *pieghevolmente 'pliable'
amichevole — • amichevolmente 'friendly'
Y/* 29. -(i)ano transiberiano —». * transiberianamen te 'trans-Siberian'
cristiano —> cristianamente 'Christian'
*
30. -iccio torbidiccio —> *torbidicciamente 'rather turbid'
*
31. -icello grandicello —>· *grandicellamente 'quite big'
*
32. -icino grandicino *grandicinamen te 'rather big'
*
33. -ico 1 libico —»· *libicamente 'Libyan'
Y/* 34. -ico 2 desertico — • *deserticamente 'desert-'
barbarico barbaricamente 'barbarian'
Constraints on the Italian suffix -mente 91
These data are very clear: 45 out of 65 suffixes never allow the affixation
of -mente·, 20 suffixes sometimes allow and sometimes do not allow the
affixation and only two "always" allow the affixation (including the
superelative suffix -issimo, for which cf. 1.3.1). The quantitative data
alone would suffice to show that the claims regarding the "full" produc-
tivity of the suffix under analysis prove to be empirically unfounded.
-mente "avoids" entire semantic sets: in fact it is not added to adjectives
92 Sergio Scalise
2. Positive constraints
The suffix in point is also subject to positive constraints. One of the
constraints affecting the various suffixal categories has to do with the
notion of "time", -mente may or may not be added to a certain set of
suffixes. If these suffixes form adjectives denoting "time", then -mente
may be added freely:
(14) acciden t+ale 4- men te 'accidental'
episod + ica + mente 'episodically'
quotid + iana + mente 'daily'
regol + are + mente 'regular'
mens + ile + mente 'monthly'
This constraint must be linked to the one seen above for simple adjectives
(cf. 1.1).
4. Conclusions
5. Post-scriptum
A reader of the present paper (whom I want to thank for his observations)
suggested to me that the restrictions on the attachment of -mente can be
explained very easily if one assumes that "A + mente" is possible when
the morphological sequence corresponds to 'in an A way' (e.g., semplice-
mente 'simply' = 'in a simple way'). The reader maintains, furthermore,
that in a language with a case "ablativus modi", the constraints on the
adjectives are the same and therefore there is no point in arguing in favor
of the derivational status of -mente.
Even though such a proposal has some advantages, it cannot be
accepted here for the following reasons. In the first place, the theoretical
framework adopted here, being strictly "lexicalist", rules out the possi-
bility of "matching", so to speak, words with phrases. In the second
place, the offered explanation needs, in my opinion, to incorporate some
diachronic information which is not within the reach of a synchronic
model such as the one adopted here.4 In the third place, not every problem
would be explained anyway. In general, if the offered explanation were
correct, we would expect both the following implicational patterns:
(20a) if 'in an A way' then A+mente
b) if *'in an A way' then *A + mente
This pattern holds true for many cases, but not for all. In fact we also
find the following possibilities:
(21a) 'in an A way' but *A + mente
b) *'in an A way' but A + mente
Examples of (21a) are the following:
(22) in modo rivelatore *rivelatormente
in modo calzante *calzantemente
in modo invidiabile *invidiabilmente
in modo distorto * dist or lament e
Constraints on the Italian suffix -mente 97
Notes
References
Bauer, Laurie
1983 English word-formation. (Cambridge: Cambridge University Press).
Bybee, Joan
1985 Morphology. (Amsterdam: Benjamins).
98 Sergio Scalise
Mutarello, R.
1987 Alcuni aspetti delta prefissazione in italiano. [Unpublished thesis, Universita
di Venezia.]
Scalise, Sergio
1983 Morfologia lessicale. (Padova: CLESP).
1984 Generative morphology. (Dordrecht: Foris).
1988 "Inflection and derivation", Linguistics 26: 561 — 581.
Scalise, Sergio —Irene Zannier
1982/83 "Restrizioni sulle regole di formazione di parola: la condizione di adiacenza
e la condizione atomo", Quaderni Patavini di Linguistica 3: 159 — 210.
Scalise, Sergio —F. Bevilacqua —A. Buoso —G. Piantini
in press "II suffisso -mente", in: Raccolta di saggi in memoria di A. Limentani
(Venezia).
Siegel, Dorothy
1977 "The adjacency condition and the theory of morphology", ΝELS 7:
189-197.
Williams, Edwin
1981 "On the notions 'lexically related' and 'head of a word'", Linguistic Inquiry
12: 145-274.
English compounds in Italian: the question of the head*
Irene Vogel
1. Introduction
In this paper, I will examine, and propose an account of, the fact that
when Italian borrows a particular type of compound from English,
typically only one of its members is retained (e.g., night club —• night). I
will first describe the relevant data and then consider several hypotheses
as to how to account for them. While the phenomenon in question
represents a relatively small area of Italian morphology, it will be shown
that it nevertheless involves a more general issue, the relationship between
compound structure and basic sentential word order.
* I would like to thank Wolfgang Dressier, the anonymous reviewers, and the participants
in the Krems meeting for their helpful comments.
100 Irene Vogel
A number of the items in (1) and (2) also appear in other European
languages (e. g., smoking in French, Dutch, German). In such cases, it is
not always clear whether each language borrowed the word from English
independently or whether one language borrowed it first and other
languages borrowed it from this one rather than from English. As Haugen
(1988: 8) points out, there are serious difficulties "in teasing out the
English models, their origins in speech or writing, their tortuous ways of
reaching the European public, and the current results in the form of local
replicas". While there may be some doubt as to the origin of a number
of the earlier borrowings in Italian which also appear in shortened form
in other languages, Italian differs from a number of other languages
which have borrowed many of the same items. Of the fifteen languages
and their treatment of English loanwords reported in the studies in
Filipovic (1982), none shows as consistent a pattern of shortening of
compounds as Italian. Instead, the full form is usually retained, though
sporadic instances of shortening are also mentioned (e.g., Albanian strip
'comic strip' (Mehmeti 1982); Dutch keeper 'goal keeper' (Gerritsen 1982);
Finnish klosetti 'water closet' (Oresnik 1982)). In Italian, on the other
hand, the reducing of English compounds is quite productive and con-
tinues to contribute new forms to Italian such as those in (3) (T. Cravens,
personal communication).
pound to a single word, but that the word that is retained is precisely
the one native speakers of English would not choose if they were to
shorten the same compounds. Compare the English examples and their
Italian equivalents in (5).
(5a) night club: Let's go to the club.
Andiamo al night.
b) smoking jacket: Give me the jacket, please.
Dammi lo smoking, per piacere.
c) scotch tape: I need the tape.
Ho bisogno dello scotch.
The question that must be addressed now is why Italian reduces the
compounds in the opposite way from the way in which they are reduced
in English. This is not to say, of course, that all English compounds
borrowed by Italian are shortened. Some are retained in their full form
(e.g., T-shirt; pickup 'record player'). Others are used in either their full
form or in the shortened form (e.g., blue jeansjjeans, basketball/basket).
Still others have undergone translation, either retaining the English word
order, particularly in earlier borrowings (e.g., banconota 'banknote', gen-
tiluomo 'gentleman' (originally from French). While these patterns are
rather sporadic in contemporary Italian, the one illustrated in (1) —(4) is
productive and it is this one that will be the focus of the rest of this
investigation.
Thus, water lily "is a" type of lily and blue cheese "is a" type of cheese.
This follows from the fact that the head of complex words in English is,
with few exceptions, on the right (cf. Williams 1981). That is, the category
label and certain other features of the right member of a compound
percolate up to the node that dominates it and its sister, as shown below.
(8 a)
watery
b)
bluet, cheeseN
croceN via^
b) N.
altoi piano N
(11a)
capoN stazioneN
b)
giallof limoneN
member of the compound in (10), they percolate from the left member
in (11). As it turns out, however, only (11) represents a productive pattern,
a point we will return to below.
(13 a)
'ferry boat'
valigiaH armadioM
'wardrobe-valise'
Since Italian is an SVO language, compounds with the head on the left
are precisely what we would expect. The other cases are mostly transla-
tions from other languages with different word order patterns or from
earlier stages in the development of Italian when the basic word order
was not strictly SVO.2
Thus, we can conclude that in retaining the leftmost member of English
compounds, Italian is behaving in a way consistent with other aspects of
the language. That is, it is retaining the element that is in the crucial
position according to its own structure. The fact that the crucial element
(i.e., head) is on the right in English is irrelevant for Italian, as are other
semantic and phonological characteristics.
5. Further considerations
At first glance several other forms also seem to complicate the situation
in Italian:
(16 a) blue jeans —• jeans
b) plum cake —• cake
While these items appear to contradict our generalization that it is the
leftmost element of the English compounds that is retained, further
investigation shows that they do not, in fact, constitute counterexamples
to the above analysis. Jeans was borrowed from English already in its
shortened form, and we can hypothesize the same for (16b), though I
have never encountered an Italian who actually uses the short form cake
(cited in Zolli 1976) rather than plum cake. Thus, it is often essential to
know something about the way in which specific words enter a language.
Other sources of borrowings may also provide relevant data. For
example, in Italian, we find additional compounds of German origin
which are shortened in the same way English compounds are:
(17 a) Volkswagen —• ν ο Iks
b) Blitzkrieg —• blitz
While blitz most likely entered Italian through English, already in its
shortened form, volks appears to be a form Italian developed on its own.
We thus have some confirmation of our prediction since the shortening
phenomenon also appears to work the same way on compounds from at
least one other language, though in general there are far fewer borrowings
from other languages in Italian.
Finally, the extent to which borrowed forms are integrated into the
borrowing language must be taken into consideration. In the case of
Italian, let us consider the small set of recently formed items illustrated
in (18).
(18 a) scuola bus 'school bus'
b) fiera district 'fair district'
In these items, the head is on the right, contrary to the claim made above
that the productive compound formation rule in Italian places the head
on the left. These forms, however, are not felt to be Italian, and their
low degree of integration appears to be determined not only by the fact
that they contain elements that are, indeed, foreign, but also precisely
because of the order of the words (i. e., the head is on the right). Related
to this, we find some speakers who refer to gli Stones and i Floyd (rather
than using the more common i Rolling and i Pink), and we also occa-
108 Irene Vogel
sionally encounter the shortened form club, instead of the more regular
form night. In the case of club, the "non-Italianness" is also usually
marked phonetically. That is, we often hear [kleb], which seems to indicate
that the speaker knows the English"u" is not pronounced [u], as it would
be in Italian. That [ε] is not the correct English vowel either is not relevant
here. What all these cases have in common is the fact that they reflect
an attempt on the part of speakers to exhibit their knowledge of English
by using forms that do not conform to the patterns of Italian. As long
as such items are felt to be "exotic" and external to the system of Italian,
they do not constitute counterexamples to the proposal advanced here.
That is, they are not Italian forms and therefore do not need to conform
to the linguistic patterns of Italian. To the contrary, they are intentionally
used to signal knowledge of a different language.
6. Conclusions
Notes
1. This pattern may, in fact, be fairly common in other Romance languages as well. See
Bauer (1978: 41), where it is reported that French, too, has a strong preference for (at
least NN) compounds to have their head on the left, in contrast with Danish and English.
2. This is, in fact, not an isolated phenomenon. Changes in compound structure related to
changes in basic word order have also been noted, for example, in Irish (cf. Ahlqvist
1985).
References
Ahlqvist, Anders
1985 "The ordering of nominal compounds in Irish", in: J. Fisiak (ed.) Historical
semantics — historical word formation (Berlin: de Gruyter), 1—9.
Allen, Margaret
1978 Morphological investigations. [Ph. D. dissertation. University of Connecti-
cut],
Bauer, Laurie
1978 The grammar of nominal compounding. (Odense: Odense University Press).
Chomsky, Noam —Morris Halle
1968 The sound pattern of English. (New York: Harper and Row).
Filipovic, Rudolf (ed.)
1982 The English element in European languages. (Zagreb: Zagreb University
Press).
Gerritsen, Johan
1982 "English influence on Dutch", in: R. Filipovic (ed.), 1 5 4 - 1 7 9 .
Greenberg, Joseph H.
1963 "Some universals of grammar with particular reference to the order of
meaningful elements", in: J. H. Greenberg (ed.) Universals of language,
(Cambridge, Mass.: M I T Press), 6 1 - 1 1 3 .
Haugen, Einar
1988 "The influence of English: a transatlantic perspective", Folia Linguistica 22:
3-9.
Klatzky, Roberta L.
1975/1980 Human memory. Structures and processes. (San Francisco: Freeman).
Levi, Judith N.
1978 The syntax and semantics of complex nominals. (New York: Academic Press).
Mehmeti, Ismail
1982 "A morphological and semantic analysis of the adaptation of anglicisms in
Albanian", in: R. Filipovic (ed.) 28 — 56.
Oresnik, Brigitta
1982 "On the adaptation of English loanwords into Finnish", in: R. Filipovic
(ed.), 1 8 0 - 2 1 2 .
Scalise, Sergio
1984 Generative morphology. (Dordrecht: Foris).
Selkirk, Elizabeth O.
1982 The syntax of words. (Cambridge, Mass.: M I T Press).
Slobin, Dan I.
1973 "Cognitive prerequisites for the acquisition of grammar", in: Ch. A. Fer-
guson—D. I. Slobin (eds.) Studies of child language development. (New
York: Holt, Rinehart & Winston), 1 7 5 - 2 0 8 .
110 Irene Vogel
Tollemache, Federico
1945 Le parole composte nella lingua italiana. (Roma: Edizioni Rores di Nicola
Ruffolo).
Warren, Beatrice
1984a "Covert connectors", Quaderni di Semantica 5.2: 331—349.
1984b "The functions of modifiers of nouns", Quaderni di Semantica 5.1: 111 —123.
Williams, Edwin
1981 "On the notions 'lexically related' and 'head of a word'", Linguistic Inquiry
12: 245-274.
Zolli, Paolo
1976 Le parole straniere. (Bologna: Zanichelli).
The importance of combining forms*
Beatrice Warren
1. Introduction
* As the reader will detect, the ambiguity of the title is intended: either interpretation being
apposite. I would here also like to thank most heartily Gunilla Malmborn and Gabriele
Stein — the former for pointing out inadvertencies in the original manuscript and above
all for helping me excerpt examples; the latter for generously sending me some relevant
articles.
112 Beatrice Warren
Typically they are neo-classical. A vowel (usually -o- but often -/-) is
inserted as a link between the combining form and the base. They point
out that there are certain bases which are particularly common, viz.
-meter, -graphy, -gram, and -logy.
Bauer (1983) does not hesitate to apply the term to first as well as last
elements, distinguishing simply between initial and final combining forms.
Unlike prefixes, initial combining forms almost invariably end in a vowel.
If they do not, the linking vowel -o- may be added. The crucial difference
between initial combining forms and prefixes, according to Bauer, is that
the former may combine with final combining forms whereas prefixes
may not. Consequently, hyper- in hyperactive could be considered a prefix,
but must be considered a combining form in hypertrophy. Similarly, the
crucial difference between suffixes and final combining forms is that —
unlike suffixes — final combining forms may co-occur with initial ones.
In fact, Bauer (1983: 272) suggests that they invariably do.
Marchand (1969) claims that the term "combining form" is usually
employed for prepositive elements which derive from Latin or Greek
stems of full words, whereas prepositive elements which go back to Latin
or Greek prefixes, prepositions, or particles are termed prefixes. He does
not think that this distinction is justified, however (Marchand 1969:
3.1.5). Nevertheless, he himself distinguishes between prefixes (bound
morphemes that are prefixed to full English words) and prepositive
elements which occur in compounding on a Neo-Latin basis, such as
astro-, electro-, galato-, hepato-, osteo-, which he thinks are of "a purely
dictionary interest". As far as final elements are concerned, Marchand
seems to make the following distinctions: suffixes, semi-suffixes, and
terminal elements. As examples of semi-suffixes he gives -like, -worthy,
-monger, -way/ways, -wise, -wort, -wright. These are like full words in that
their word character is still recognizable. They are like suffixes in that
some of them are used only as second words of compounds (Marchand
1969: 4.80). Suffixes Marchand defines as elements which are not inde-
pendent words and which are tacked on to full English words or to
allomorphs of English words. Last elements which are not combined with
full words or allomorphs of full words, such as, for instance, -scope in
galvanoscope, he considers not to be proper suffixes; they are simply
terminal elements and excluded from consideration in his book. Never-
theless, Marchand (1969: 4.1.8.1—4.1.8.2) does discuss the creation of
last elements such as -athon, -burger, -furter, -rama and calls them suffixes,
which seems inconsistent since they often do not have English words or
allomorphs of English words as bases.
114 Beatrice Warren
It is obvious from this survey that there are certain word components
which linguists intuitively feel are neither affixes nor roots. It is also
obvious that there is as yet no uniform terminology for these elements,
nor any generally valid description of them. 2 For example, some linguists
insist on their classical origin; others include in the category native
morphemes too; some emphasize the ability of these word components
to occur in words without proper roots; others ignore this interesting
aspect of their character. The present paper is an attempt to improve on
the descriptive as well as explanatory adequacy of the account of these
affixlike morphemes. To do so, I have excerpted from Mort (1987) words
containing elements which on intuitive grounds I considered good can-
didates for being formatives of this kind. The excerption yielded 81
different types of likely combining forms. These were of course first
divided into first and last elements. However, both in the case of initial
and final combining forms, my examples naturally divided themselves
into two further groups: one consisting of elements which are allomorphic
variants of some other word, e.g., astro-, which is a variant of Lat.
astrum, and -drome, which is a variant of Gk. dromos; and one consisting
of elements which represent parts of other words, e.g., cyber-, which we
recognize as a part of the word cybernetics and -aholic, which is not a
variant of alcoholic, but a part of it. Among the final combining forms,
there was also a third group of forms, viz., elements which from a purely
formal point of view are not new morphemes, but which have novel
meanings. Gate, for example, is no new morpheme-form, but as a com-
bining form it has a novel meaning, i.e., "political scandal involving a
cover-up". Altogether, there are then five subgroups as illustrated in
Table 1. For additional examples, see lists in the Appendix.
Glancing through the Appendix, we will notice, among other things,
that my examples support the claim that initial combining forms end in
a vowel, more precisely /i/, /o/, or /a/. (Of these, /o/ and probably /i/ —
although my examples happen not to confirm this — can often be looked
as linking phonemes.) In other words, initial combining forms differ from
final ones in that the former have a characteristic phonetic shape, whereas
the latter are amorphous in this respect.
We will also notice differences between the groups, in particular be-
tween Group I, on the one hand, and Groups II and III, on the other.
With a few exceptions, notably debtno- and smello-, both the first and
116 Beatrice Warren
Initial Final
combining forms
4. Secretion
"Suffix" Meaning-particularization
(The dotted line signifies that some complementing semantic feature is obligatory.
Brackets surround relational meanings, which, as is shown in Section 6, are in
principle variable.)
The importance of combining forms 119
be caused by the fact that often there is more than one model word
possible. In principle, as soon as a composite containing a combining
form has been established, it may in turn serve as a model word. As an
example of this, let us consider -naut again. Originally, it was secreted in
the sense "explorer of unknown territories" from argonaut, referring to
a sailor in the legendary Argo (see Adams 1973: 185), to form aeronaut,
which refers to people travelling in balloons. Aeronaut in turn served as
a model word for astronaut, cosmonaut among others, which explains the
existence of the features "especially in space" and "in an enclosed space"
in Bauer's definition.
The question which will be pursued in this section is: are combining
forms a special kind of root or affix or are they in fact neither?
Let us first consider similarities and differences between roots and
combining forms. Combining forms and roots are both morphemes with
lexical content. Roots may be free or bound. Combining forms, I have
argued, are invariably bound. It follows that a free morpheme must be
a root, since — again by definition — it cannot be an affix either. The
difficulty is therefore to distinguish between bound roots and combining
forms. Granted that we know which elements are suffixes, this is possible.
It follows from the following: a bound root must combine with a suffix
to make a word; a single combining form combined with a suffix does
not make a word. Consequently, a bound morpheme in word-final po-
sition is either a suffix or a combining form, but not a root. A single
bound morpheme followed by a suffix is a root. If it is followed by
anything else, it is either a prefix or a combining form.
In this connection, let us deviate from the topic of the section and
compare ordinary compounds and composites containing combining
forms. An initial combining form may co-occur with a final one (e.g.,
neurology) or with a free morpheme (neurophilosophy). A final combining
form may consequently be found together with an initial combining form,
but also with a free morpheme (feature-itis)5 and very exceptionally with
a prefix [prequel). Whatever type of combination, the first element is the
modifier and the last the head. 6 This is also true of regular compounds.
It may appear then that the only difference between ordinary compounds
The importance of combining forms 123
8. Summary
The present study reveals that there are two main ways in which so-
called combining forms may be formed: by phonetic modification of some
existing morpheme or by secretion. It is possible to distinguish between
The importance of combining forms 125
Appendix
Group I
alpino- alpinodrome Lat. alpinus
"place for climbing contests"
angio- (balloon) angioplasty Gk. angeion
"technique for treating angina"
antenno- elect roan tennogram ?E antenna
"chart of electricity in antennae"
laqua- aquatube Lat. aqua
"water slide"
astro- astrodome Gk./Lat. astrum
"stadium with translucent domed
roof'
bio- biocomputer, biohazardous, Gk. bios
bioholonics, biomotor et al.
chrono- chronobiology et al. Gk. chronos
"the study of biological rhythm in the
human body"
circo- circotherm( oven ) Lat. circum
"oven with fan circulating heat"
debtno- debtnocrat Ε debt
"banker dealing with international
debts"
electro- electroantennogram, et al. Lat. electricus
"tracing of electricity in antennae"
ethno- palaeoe thnobo tany Gk. ethnos
Igiga- gigadisc, gigaflops ?Gk. Gigas
"disc capable of storing large
amounts of data"
hetero- heterosexism, (heterosexist) Gk. heteros
"discrimination agabinst homosexu-
al«"
alS
iatro- iatroblast Gk. iatros
"embryonic doctor"
irido- iridology, (iridologist) Gk. iris
"technique of diagnosis involving iris"
litho- lithotripter Gk. lithos
"device for pulverizing kidney stones"
mega- "great" mega-blockbuster, megabrand, Gk. megas
megabuck, mega-city et al.
The importance of combining forms 127
Group II
cyber- cyberphobia cybernetics
"fear of computers"
eco- ecosocialism ecology
"socialism concerned with ecology"
econo- econospeak economics
"the jargon of economists"
Euro- Eurofighter, Eurο feebleness, Europe(o)
Eurosclerosis, Eurowimp et al.
Ipluro- pluro-communism pluralism
"communism and pluralism com-
bined"
vege- vegeboom vegetarianism
"sudden popularity of vegetarianism"
Group I
-blast iatroblast Gk. blastos
"embryonic doctor"
-cracy yobbocracy ?aristocracy
"rule by louts and thugs"
-erat debtnocrat ?bureaucrat
See debtno-
-drome alpinodrome Gk. dromos
"place for climbing contests"
-ectomy lumpectomy Gk. -ektomia
"surgical removal of cancer"
-gram electroantennogram et al. Gk. gramma
"chart of electricity in antennae"
-graphic psychographic (s) Gk. graphos
"market research of attitudes"
-graphy videosomatography Gk. graphia
See somato-
The importance of combining forms 129
Group II
-(a)holic spendaholic alcoholic
"compulsive spendthrift"
-(a) thon bikeathon, duckathon, pedalathon, marathon
readathon, swimathon
-erati slopperati, glitterati lliterati
"the deliberately untidy rich";
"jetsetters", respectively
-gram Tarzangram, potatogram telegram
"greeting delivered by costumed
person"; "message on a potato"
-itis feature-itis Ίarthritis, etc.
"excessive occurrence of feature arti-
cles"
-nik Wappnik Irefusenik
"journalist refusing to cross picket
line in Wapping to go to work"
-nography warnography pornography
"literature or films glorifying war and
violence"
-tro outro intro
"ending of song or musical number"
-rrhoea processorrhoea diarrhoea
"excessive flow of words induced by
word processor"
130 Beatrice Warren
Group III
Comments
Notes
1. Capitalization signifies that the particular word or form has an entry of its own in the
dictionary.
2. I am not the first to have noticed inconsistencies in the treatment of combining forms.
See Stein (1977 and 1984).
3. Barnhart et al. (1980), however, use the phrase "abstracted form" for secreted elements.
4. The fact that combining forms are bound morphemes does not prevent them from
occasionally being converted into proper nouns. Consider, for example, itis, which can
be used as a noun in the sense "bodily condition or disease".
5. Bauer's claim that final combining forms are restricted to the company of initial
combining forms is not correct.
6. There is an exception to this rule, viz., the so-called dvandva compounds exemplified
by Franco-German, Anglo-Polish, Sino-Italian. These compounds consist of two heads
rather than a modifier and a head.
References
Adams, Valerie
1973 An introduction to Modern English word-formation. (London: Longman).
Bauer, Laurie
1983 English word-formation. (Cambridge: Cambridge University Press).
Barnhart, Clarence L., et al.
1980 The second Barnhart dictionary of New English. (Bronxville, Ν. Y.: Barnhart
Books).
Hansen, Barbara — Klaus Hansen —Albrecht Neubert — Manfred Schentke
1985 Englische Lexikologie. (Leipzig: Enzyklopädie).
Jespersen, Otto
1950 Language — its nature, development and origin. (London: Allen & Unwin).
Marchand, Hans
1969 The categories and types of present-day English word-formation. (München:
Beck).
Mort, S. (ed.)
1987 Longman Guardian new words. (Harlow: Longman).
Quirk, Randolph, et al.
1985 A comprehensive grammar of the English language. (London/New York:
Longman).
132 Beatrice Warren
Stein, Gabriele
1977 "English combining forms", Linguistica 9 (= Tartu riikliku ülikooli toime-
tised 437), 140-147.
1984 "Word-formation in Dr. Johnson's dictionary of the English language",
Journal of the Dictionary Society of North America 6: 35 — 112.
Warren, Beatrice
1988 "Ambiguity and vagueness in adjectives", Studia Linguistica 42: 122 — 177.
Compounding and inflection
Wiecher Zwanenburg
0. The general idea behind this paper is the hypothesis that in the
unmarked case languages organize their word structure hierarchically in
such a way that inflection is peripheral to compounding.
I want to discuss here the apparently exceptional relationship between
compounding and inflection in the Romance languages, and more par-
ticularly in French. In French, we find cases where inflection seems to
be inside compounding, as in (la), next to the expected cases where
inflection is peripheral, as in (lb):
(la) des secretaires-generaux 'secretaries-general'
des basses-cours 'poultry yards'
b) des en-tetes 'letter headings'
des apres-midi{s) 'afternoons'
I will argue that the cases of (la) are only apparent exceptions to the
above-mentioned generalization.
Before discussing this point in Section 2, I will consider in Section 1
the possibility to distinguish in a satisfactory way between derivational
and inflectional affixation. This possibility, questioned by quite a number
of linguists, is of course a necessary condition for the above-mentioned
generalization to make sense.
(7) N - -> VP
This leads them to admit more generally, for the periphery of the
grammar, nonmorphological word-creating rules reanalyzing phrases
such as VP as words. Unlike the cases of (4) and like ordinary words
created by morphological rules, syntax may only refer to these words as
a whole and cannot "look into them".
Compounding and inflection 137
Di Sciullo —Williams then argue that in fact there are more possibilities
for this type of words, as shown in (8). And they conclude that the right
generalization is as in (9).
(8a) des trompe-l'oeil 'trompe l'ceil paintings'
b) des boit-sans-soif 'boozers'
c) des bons-ä-rien 'good-for-nothings'
d) des hommes-de-paille 'straw-men'
e) des hors-la-loi 'outlaws'
(9) Ν —• XP
Now, this would constitute a problem for our generalization: in (8c)
and (8d), unlike the other cases, inflection is not on the entire compound-
like word but on a part of it only, contrary to our generalization. But it
is easy to see that precisely this fact allows us to interpret the cases of
(8c) and (8d) as syntactic expressions with idiosyncratic meaning or
listemes, like those in (4). Thus, such expressions turn out to be more
widespread than Di Sciullo — Williams suggest.
In this way we are led to maintain rule (7) instead of (9), but generalized
over nouns and adjectives to account for adjectival cases like (10). The
rule will then take the form of (11).
(10) un enfant brise-tout 'a child that breaks everything'
(12a) N: en-tete
b) V: en-chain-(er) 'to enchain'
(13). The examples given so far suggest that its hierarchical relationship
with inflection is the same as that of compound rules: inflection is outside.
But then Di Sciullo — Williams do not discuss the not infrequent case of
VP nouns where VP contains a plural object, like casse-noisettes 'nut-
cracker', which have the same form in the plural. This fact does not
impair the reasoning above about the ungrammaticality of *nounshood
and * choirs boy as plural forms of nounhood and choir boy. The question
to be examined is to what extent syntax-like nonmorphological word
structure admits inflected elements. And more generally, what are the
restrictions on such a structure other than nonreferentiality discussed by
Di Sciullo —Williams.
References
Anderson, Stephen R.
1982 "Where's morphology?", Linguistic Inquiry 13: 571 —612.
Di Sciullo, Anna-Maria —Edwin Williams
1987 On the definition of word. (Cambridge, MA: MIT Press).
Selkirk, Elizabeth O.
1982 The syntax of words. (Cambridge, MA: MIT Press).
Topic 3: Inflectional morphology and clitics
Arguments against the passive as a universal
morphological category
Paul Kent Andersen
1. Introduction
For quite some time now it has been recognized that constructions
generally regarded as "passives" differ in morphological, syntactic, se-
mantic, and even discourse features; in fact, there is very little that is
shared by all "passive" constructions, cf., among others, the conclusions
reached by Siewierska (1984: 259). Thus we are faced with the problem
that there is no single, generally accepted definition of the "passive" that
does justice to all instances of "passives" attested in the literature. This
is particularly evident in the most recent collection of papers concerning
the "passive" and Voice edited by Shibatani (1988): here we find that the
definitional properties of the "passive" are regarded by some authors,
e.g., Rude and Cooreman, as discourse features, by others as syntactic
142 Paul Kent Andersen
properties, e.g., Siewierska and Kimenyi, and by others still, e.g., Dezsö,
as the "passive" morpheme. Of course, one of the easier ways out of this
problem is to define the "passive" in a rigid manner and then to claim
that instances not covered by the definition are not "passives". Since,
moreover, "theoretical consideration of passive constructions has natu-
rally focused on English" (Langacker — Munro 1975: 789), it therefore
comes as no surprise to find that the English "passive" generally ends up
either as the expression par excellence of "the passive prototype" — cf.,
e.g.,
We shall refer to passives like (lb), John was slapped, as 'basic pas-
sives' Our justification for calling such passives 'basic' is that they
are the most widespread across the world's languages (Keenan 1985:
247)1
and
The characterization in 40 applies to prototypical passives like English
Many soldiers were killed, or its Japanese equivalent Takusan no heitai
ga koros-are-ta (Shibatani 1985: 837)2
or positioned at one extreme along "the passive continuum", cf.
If the correlations (6), (7), and (8) above indeed hold, then one excepts
one major continuum in the typology of passivization, with the English
type representing one extreme of the continuum... . The Ute extreme
of the scale would have just the opposite properties (Givon 1981: 171).
Had theoretical considerations been focused on a language other than
English, we may have found that an entirely different prototypical defi-
nition of the "passive" would have been proposed and that the English
"passive" would have been regarded as an "alternative",
3
"pseudo-passive", "peculiar passive", or the like. In the present paper
an alternative approach to the problems of the "passive" will be taken,
i.e., instead of looking for one single "passive", the possibility of the
existence of a number of distinct "passives" will be investigated.
3. Verbal categories
Bybee (1985) and Dahl (1985) have each devoted an entire monograph
to the investigation of verbal categories based upon two independent
samples of languages especially suited for typological research. Although
the results that were reached do differ in various aspects, they both
demonstrate the existence of a number of prototypical verbal categories.
But because they both concentrate on Tense, Aspect and Mood, they
unfortunately have very little to say about "passive" Voice; both do,
nevertheless, agree that such a morphological category exists. So, for
example, according to her principles of relevance6 and generality,7 Bybee
establishes the following hierarchical order of verbal morphemes relevant
to the verb root: VERB-Valence-Voice-Aspect-Tense-Mood-number/
person/gender Agreement. 8 Since the "passive" is regarded as an instance
of the category Voice, we see that within this framework the "passive"
not only has a specific place in this hierarchy of verbal categories, it also
has a specific function (in terms of relevance and generality) that distin-
144 Paul Kent Andersen
guishes it from the other categories. In this respect it is worth noting the
definitions of the categories for Valence, Voice, and Aspect employed by
Bybee in her investigation:
Valence refers to differences in the number or role of arguments that
the verb stem can take. Voice indicates the perspective from which the
situation described by the verb stem is viewed. Aspect refers to the
way the internal temporal constituency of the situation is viewed
(Bybee 1985: 28).9
Thus we have reason to believe that should there be a single, universal
morphological category for "passive voice", it will exhibit not only a
specific place along the hierarchy of verbal categories, but also a specific
function with regards to Bybee's principles of relevance and generality.
Dahl, on the other hand, gives the following reason why he has chosen
not to investigate the "passive" with his method based upon a compre-
hensive questionnaire:
In addition to the combinatorial explosion, there are additional reasons
why it is hard to study e.g. passive constructions with a methodology
of this type. In some languages, passives may be very marginal or even
not occur at all; other languages may have several constructions that
could be labelled passive, with more or less subtle differences in
conditions of use. This means that by simply giving an English sentence
in the passive you cannot guarantee that what comes out in the
translation is the passive construction you are looking for, if it is a
passive at all (Dahl 1985: 47).
In the following investigation we will assume that it is indeed possible
to classify the distinctive verbal morphology employed in constructions
generally regarded as "passives" according to the principles established
by Bybee and Dahl without recourse to the circular axiom mentioned
above. Our first indication that the morphemes employed in "passive"
constructions are not all instances of one and the same verbal category
can be seen from the fact that morphemes isolated and glossed as
"passives" generally perform functions other than that of appearing in
"passive constructions", i.e., "passive" morphemes are generally em-
ployed in constructions other than the "passive". 10 A closer look at these
other constructions will reveal the fact that they are themselves not
identical, but rather that they do define certain independent sets of
construction types. Thus it is not the case that all of the other functions
of the "passive" are the result of the mere extension of a single, basic
"passive" meaning in all of these instances.
Arguments against the passive 145
4. Agreement
discourse as one where the speaker relates a series of real or fictive events
in the order they are supposed to have taken place" — we will be in the
position to better characterize the actual function of the "middle" mor-
pheme in this type of "passive" construction: one participant of the
narrated event — i.e., the "Actor" — does not participate in the narrative
discourse (/speech act) within its respective narrated context; in the
corresponding "active" the "Actor" participates not only in the narrated
event, but also in the narrated discourse (/speech act). Thus we see that
the narrated event is not altered at all in this type of "passive" construc-
tion.
To sum up, in Classical Greek and other languages the specific mor-
phological means employed in the "passive" can be identified as an
inflectional category for person/number "Agreement". The basic/
prototypical semantic function of the "middle" is furthermore to indicate
that the "subject" is "affected" by the verbal action. The specific function
of this morphological category when employed in the "passive" can be
characterized as indicating that the "Actor" participates in the narrative
event without, however, participating in the narrative discourse (/speech
act). Since reflexive affixes in numerous languages throughout the world
exhibit almost 16 the very same functions as the "middle", there is some
indication that they may also represent the same prototypical verbal
category. 17
5. Perfect
Let us now turn to the type of "passive" found in languages like English,
cf., e.g.:
(2) John was seen (by Mary).
Our very first problem will be to isolate the specific morpheme employed
in this construction that distinguishes it from a corresponding "active"
construction. Unfortunately, it is a general consensus among various
linguists (especially those working within formal frameworks) that the
distinctive morphology in "passives" such as (2) above is the so-called
"passive participle" — or more accurately, the suffix -en which forms the
"passive participle". 18 Nevertheless, it should be rather obvious that this
form of the verb is not in any way restricted to "passive" constructions,
cf.:
148 Paul Kent Andersen
(6) När hon kom ut pä gatan, upptäckte hon, att bilen blev stulen.
'When she came out onto the street, she discovered that the car
was being stolen.'
6. Valence
7. Aspect
the category perfective Aspect. In fact DeLancey (1981), who does dif-
ferentiate between perfective Aspect and Perfect, states that "Perfective
views an event from its terminal point" (1981: 647) and proposes that in
the "passive" there is a "terminal view point with respect to the transitive
vector" (1981: 647). Thus we see that according to DeLancey the "passive"
is an expression of terminal viewpoint and hence perfective Aspect.22
Should the specific verbal morphology that is employed in such "pas-
sive" constructions actually represent the category perfective Aspect, then
we would be able to conclude that this category is again distinct from
the other categories discussed above, cf.:
A PFV verb will typically denote a single event, seen as an unanalysed
whole, with a well-defined result or end-state, located in the past.
More often than not, the event will be punctual, or at least, it will be
seen as a single transition from one state to its opposite, the duration
of which can be disregarded (Dahl 1985: 78).
We would furthermore expect to find that if these morphemes are em-
ployed in constructions other than the "passive", then these construction
types would differ from those construction types that employ other verbal
categories.
8. Voice
The basic assumption of this paper has been that it is possible first of all
to isolate and identify the verbal morphology that is distinctive of con-
structions generally claimed to be "passives" in the literature, and second
of all to classify the specific morphemes as instances of prototypical
verbal categories in accordance with the methods proposed by Bybee
(1985) and Dahl (1985). The preliminary results presented here give
sufficient evidence that the morphemes traditionally regarded as "pas-
sives" do not represent instances of a single, prototypical morphological
category, but rather that they represent instances of the following distinct
categories:
(i) Agreement
(ii) Perfect of result
(iii) Resultative
(iv) increase in Valence (causative)
(v) decrease in Valence ("detransitive")
(vi) perfective Aspect
(vii) "passive" Voice
Each of these prototypical verbal categories furthermore has its own
distinct function, thus giving evidence that the constructions employing
these verbal categories are themselves distinct and not merely variants of
a single, prototypical construction type. Semantically, these functions can
be interpreted as various processes by which a narrated event is trans-
formed into an appropriate discourse event — or state — within the
respective discourse context.
Since the basic function of Agreement is referential in nature (cf.
Lehmann 1988), we see that the specific function of Agreement has to
do with identifying participants of the narrated event with respect to the
154 Paul Kent Andersen
Notes
1. Notice that although this particular construction may be widespread in the world's
languages syntactically speaking, the specific morphology found in this construction is
certainly restricted to a rather small number of languages.
2. Shibatani's characterization 40 of the passive prototype contains the following mor-
phological property: Active = P; Passive = P[ +passive] (837). As I will argue in the
course of this paper, there is no evidence for a single, prototypical passive morpheme,
i.e., P[ +passive],
3. Iconoclastic though it may be, in actual fact the type of "passive" attested in English
should be classified as either a stative or a resultative according to Nedjalkov and
Jaxontov's (1988: 6) definition: "... the stative expresses a state of a thing without any
implication of its origin, while the resultative expresses both a state and the preceding
action it has resulted from". So, for example, in the sentence "When I came at five,
the door was shut, but I do not know when it was shut" the first occurrence of the
"passive" was shut is a stative, whereas the second is a resultative. Note, however, that
not all passives can be classified as states or resultatives (especially those discussed in
Sections 4, 6, 7, and 8 below). For a further discussion see Section 5 below; note,
though, the difference in terminology.
Arguments against the passive 155
4. Cf., among others, Keenan (1985: 245): "Thus consider how, as field workers, we can
tell if a sentence in a language is passive or not. What is it about passives that makes
them observably distinct in surface from basic actives? ... Thus we cannot recognize a
passive in terms of its NPS being marked or positioned in the sentence in ways different
from those used in basic actives... . In fact the only way we know that (4b) above is
passive is by the presence of a specifically passive suffix, -aki, on the verb. And this
observation turns out to be general across languages. That is, in general in a language,
what is distinctive about the observable form of passives is localized within the predicate
or verb phrase (understood broadly enough to cover auxiliary verbs)." See also the
relevant discussion in Haspelmath (1988: 3 — 11). A notable exception to this is Sie-
wierska (1984: 256).
5. Notice that some authors have even attempted to motivate this axiom, cf., e.g.,
Haspelmath (1988: 1): "My view that the nature of passive morphology is inseparable
from the nature of passive syntax..." and (1988: 51, fn. 3): "And that, more generally,
whenever a syntactic construction is associated with morphological marking, the nature
of the morphology is inseparable from the nature of the syntactic construction." Should
the results of this paper be accepted, Haspelmath would then be forced to accept the
existence not of one single "passive" construction, but rather a number of distinct
"passive" constructions.
6. Cf. "A meaning element is relevant to another meaning element if the semantic content
of the first directly affects or modifies the semantic content of the second" (Bybee 1985:
13).
7. Cf. "By definition, an inflectional category must be applicable to all stems of the
appropriate semantic and syntactic category and must obligatorily occur in the appro-
priate syntactic context. In order for a morphological process to be so general, it must
have only minimal semantic content. If a semantic element has high content, i.e., is
very specific, it simply will not be applicable to a large number of stems" (Bybee 1985:
16 f.).
8. Bybee's hierarchy of verbal categories is reminiscent of, and certainly related to, the
distinction made by Foley —Van Valin (1984) between different, structured layers of
the clause and specifically between different operators of the various layers.
9. It should be pointed out that Bybee's (1985: 20) characterization of Voice is based
upon Barber (1975), which in turn represents the characterization of only one single
type of "passive" construction, i.e., those discussed under the heading "Agreement"
below. Should the results of this study be accepted, the characterization of Voice will
necessarily need modification.
10. Cf., e.g., Shibatani (1985: 825): "Cross-linguistic evidence that passives are often related
to other constructions is overwhelming. In fact, it is far more deep-rooted and wide-
spread than has been indicated by casual observations found in many descriptive
materials". See also Keenan (1985: 253): "Moreover, the same types of formal mor-
phological means used in deriving basic passives are often used to derive VPS which
are not passives".
11. See Andersen (1989) for a detailed discussion.
12. As will be discussed below, the Classical Greek clause louo-metha can also mean 'We
are washing ourselves' as well as 'We are washing each other'.
13. Notice, in this respect, the following generalization already mentioned in Greenberg's
discussion of implicational universals: "Universal 30. If the verb has categories of
person-number or if it has categories of gender, it always has tense-mode categories"
(Greenberg 1966: 93).
14. Since there are numerous references in the literature of other "passive" morphemes
that are also employed in reflexive — possibly even reciprocal and anticausative —
constructions, these construction types may prove to be not indicative of this type of
morphological category. Notice, though, that these other "passive" morphemes have
156 Paul Kent Andersen
not been identified as instances of any particular verbal category (other than the
"passive"): a detailed study of these morphemes may reveal the fact that they too are
instances of this prototypical verbal category. At any rate, what is distinctive of the
"middle" is its employment in a number of so-called (intransitive and transitive) "middle
constructions".
15. Although there has been much discussion of this notion of "affectedness" as the primary
function of the "middle" in recent years, it should be noted that this very function has
been recognized in some linguistic circles for more than a century. In fact, the very
term päthos that was employed for this morphological category in the Greek gram-
matical theory has the meaning "affection", cf. Andersen (1989). The characterization
of the function of the "middle" as "affectedness" should also be regarded as a very
useful, but still only a prototypical characterization, because there are some exceptions.
So, for instance, there are examples of the "middle" employed in impersonal "passives"
in which there is no "subject" to be affected. This applies not only to impersonal
"passives" of intransitive verbs such as Latin venitur, curritur, itur (cf. Lehmann 1985:
247 f.), but also to impersonal "passives" of transitive verbs that therefore result in
"non-promotional passive" constructions, cf. the two examples from Plautus discussed
in Wackernagel (1950: 146). Furthermore, there are also other means of expressing the
fact that the "subject" is affected by the verb, cf. the type Latin pudet me and German
Mir ist heiß with "active" morphology.
16. Notice though that in a few languages the reflexive markers are placed closer to the
verb root than categories for Tense and Agreement; they therefore can be regarded as
less prototypical in such instances.
17. Should the semantic characterization of this type of "passive" construction given above
be valid, I see absolutely no reason not to regard constructions with active morphology
that express this very same semantic structure as "passives", cf. Keenan (1985: 247):
"It appears however that languages without passives have somewhat more grammati-
cized means for expressing functional equivalents of basic passives. Perhaps the most
common means is to use an active sentence with an "impersonal" third plural subject.
By impersonal here we mean simply that the third plural element is not understood to
refer to any specific group of individuals." The only reason for not regarding these
constructions as "passives" is by reference to the fundamental — yet totally circular
— axiom discussed above. In fact Keenan (1985: 255) even states that "[o]ne might
expect, then, to find passives VPS which are identical to the transitive verbs they are
derived from." Moreover, there are numerous instances in which such third-person
plural markers do in fact develop into morphemes for the "passive", cf., e.g., Shibatani
(1985: 845 f.) and Haspelmath (1988: 37).
18. Cf., e.g., the discussion in Bresnan (1982: 81 - 8 2 , fn. 5), Hoekstra (1984: 129 ff.), Sells
(1985: 55), Riemsdijk - Williams (1986: 248 f.), Jaeggli (1986: 590 ff.) and McCloskey
(1988: 44 ff.). A notable exception to this rule is Haider (1986).
19. This particular construction contrasts with "passives" such as The glass is regularly
broken by vandals', the morphology of both predicates is, however, identical.
20. According to Nedjalkov — Jaxontov (1988: 6 ff.) the category in question here is the
resultative, or more specifically the objective resultative.
21. Note also the distinction in German between the "Zustandspassiv" with the verb sein
"to be" and the "Vorgangspassiv" with the verb werden "to become".
22. Whether terminal view point does indeed presuppose perfective Aspect is certainly
debatable.
Arguments against the passive 157
References
Klaiman, Μ. H.
1982 "Affectiveness and the voice system of Japanese: satisfaction guaranteed or
your money back", Berkeley Linguistic Society 8: 267 — 281.
Langacker, Ronald W. — Pamela Munro
1975 "Passives and their meaning", Language 51: 789 — 830.
Lehmann, Christian
1985 "Ergative and active traits in Latin", in: F. Plank (ed.) Relational typology
(Berlin: Mouton), 243-255.
1988 "On the function of agreement", in: M. Barlow —Ch. A. Ferguson (eds.)
Agreement in natural language. Approaches, theories, descriptions (Stanford:
Center for the study of language and information), 55 — 87.
Lehtinen, Tapani
1984 Itämerensuomen passiivin alkuperasta [On the origin of the Baltic-Finnish
passive] (Helsinki: Suomalaisen Kirjallisuuden Seura).
Lyons, John
1968 Introduction to theoretical linguistics. (Cambridge: Cambridge University
Press).
McCloskey, James
1988 "Syntactic theory", in: F. J. Newmeyer (ed.) Linguistics: The Cambridge
Survey I. Linguistic theory: Foundations (Cambridge: Cambridge University
Press), 1 8 - 5 9 .
Nedjalkov, Vladimir P.— Sergej Ε. Jaxontov
1988 "The typology of resultative constructions", in: V. P. Nedjalkov (ed.) Ty-
pology of resultative constructions (Amsterdam: Benjamins), 3 — 62.
Reichenbach, Hans
1947 Elements of symbolic logic. (New York: Dover) [Reprint],
Riemsdijk, Henk van —Edwin Williams
1986 Introduction to the theory of grammar. (Cambridge, MA: MIT Press).
Sells, Peter
1985 Lectures on contemporary syntactic theories (Stanford: Center for the study
of language and information).
Shibatani, Masayoshi
1985 "Passives and related constructions: a prototype analysis", Language 61:
821-848.
Shibatani, Masayoshi (ed.)
1988 Passive and voice. (Amsterdam: Benjamins).
Siewierska, Anna
1984 The passive: a comparative linguistic analysis. (London: Croom Helm).
Wackernagel, Jacob
1950 Vorlesungen über Syntax mit besonderer Berücksichtigung von Griechisch,
Lateinisch und Deutsch l 2 . (Basel: Birkhäuser).
The empty morpheme entailment*
Robert Beard
1. Separationist morphologies
The view that affixation and other morphological operations are separate
from those of lexical and inflectional derivation has been articulated
recently by Beard (1981, 1987, 1988), M o o r t g a t - v a n der Hulst (1981),
Pounder (in press), Szymanek (1985), Wood (1985) and for inflection
alone by Anderson (1982) and Matthews (1972). On this view, the addition
and adjustment of lexical categories in lexical bases are carried out by
abstract morpholexical (for word formation) and morphosyntactic (for
inflection) derivation rules. Morphology comprises autonomous phono-
logical "spelling-out" operations such as affixation, reduplication, revo-
calization, or autosegmental recombination.
(1) Lexeme Operations on Lexemes Grammatical
Component
Lexical Lexicon
feature Syntax
inventory
I
/p/ Morphology
Phonology
where /P/ = the phonological matrix of the lexical formant. This model,
the "separation hypothesis", implies that language is based on two discrete
types of basic elements, lexemes and (grammatical) morphemes, each with
its own related set of operations located in different parts of the grammar.
It does not take issue with the claim that lexical morphemes are signs;
lexemes are undisputedly signs, direct associations of sound and meaning.
The separation hypothesis is distinguished by its definition of grammatical
morphemes as semantically empty phonological operations and not lexical
objects.
Fleisch 'meat (beef) broth'. Colloquial forms like eine Weiß (= Weiß-
wurst), eine Brat (= Bratwurst) must be performative wordplay since the
head noun is stylistically optional, not grammatically omitted,
(b) Lexical synonymy is approximate, not exact.
(4) die Fleischbrühe die Fleischbouillon die Bohnensuppe
die Kalbsbrühe (die Kalbsbouillon) die Kalbssuppe
Suppe is distinguished from Brühe and Bouillon by its heavier ingredients.
Brühe also means 'gravy, sauce' besides 'bouillon'.
Finally, (c) while some synonyms of such semantic fields are sporadi-
cally associated with one gender, such associations are only occasional
(5)·1
(5) die Vorspeise das Zwischenessen das Nachtmahl
die Schultasche der Postsack der Sportbeutel
As Bloomfield (1933: 280) put it, "[t]here seems to be no practical criterion
by which the gender of a noun in German, French, or Latin could be
determined".
The separation hypothesis predicts something quite different. Since L-
derivation and affixation are independent, (a) derivation may operate
without affixation, thus generating zero-marked forms. Since separation
allows several affixes to mark an identical L-derivation, (b) it predicts
absolute synonymy among such forms. Finally, in cases of derivational
synonymy, we should (c) find instances of derivationally determined
gender which does not vary with suffix selection. Since derivation deter-
mines meaning and affixes are semantically empty, in most instances
derivation should determine gender, number, and diminution and these
categories should remain consistent across suffix markers.
To test the predictions of the empty morpheme entailment of the
separation hypothesis against the lexical morpheme hypothesis, this paper
examines two Indo-European L-derivations which are marked in each
language by several ostensibly synonymous suffixes. The purpose is to
discover whether suffixes or abstract derivation rules determine gender
among synonymous L-derivations.
162 Robert Beard
This exit is blocked two ways: (a) French -eur and German -nis are not
always feminine and (b) gender is also assigned when no affix at all is
added.
German -nis is an unproductive nominalization suffix. French -eur is
highly productive as a masculine agentive marker.
The empty morpheme entailment 163
+ Noun
0 —* -eur j < + Agent
+ Masculine
+ Noun
4- Feminine
164 Robert Beard
the division of the locative case into locative case functions with the same
two meanings, i.e., those marked by the prepositions na 'on' and u 'in'
(Beard 1981: 171—202). The verbs below are presented in their past-
tense stem form where this form serves as the base for locative nomin-
alizations; otherwise, the citation form is provided. /I/ is regularly replaced
by /o/ unless followed by a vowel.
(12) igral- 'play, dance' igral-ist-e 'playground'
igrao-n-ic-a 'dance hall, casino'
radil- 'work' radil-iste 'work site'
radio-n-ic-a 'workshop'
vezbal- 'exercise' vezbal-ist-e 'training field'
vezbao-n-ic-a 'gymnasium'
Assuming with Zepic (1970) and others that the "in'Mocative means
'small, enclosed place of N' and the "on'Mocative means '(large, open)
place of N', where the latter is the unmarked locative, the separation
hypothesis also predicts the data quite well. Since two derivations are
necessary, either the derivations or their suffixes, e.g., -n(ic)a and -iste,
respectively, could assign gender appropriately. However, the "in'Moca-
tive is marked by a fairly wide range of affixes — all feminine.
(13a) pepeo 'ash' pepel-jar-a 'ashtray'
pil-a 'saw' pil-an-a 'sawmill'
racun-i- 'calculate' racun-ic-a 'arithmetic notebook'
rafin-ir-aj- 'refine' rafin-er-ij-a 'refinery'
The examples show how derived locative nominals associated with the
sense of 'inside', i.e., referring to an enclosed space, all exhibit feminine
gender. Here, even the idiomatic variations (13a) reveal feminine gram-
matical gender. Most striking, again, are those "in'Mocatives which, like
the French examples in -eur, display a masculine agentive suffix, -ar,
declined in the feminine declension I paradigm, that is, -ara, -are, -ari,
..., instead of the masculine, declension II paradigm -ar, -ara, -aru, etc.4
(13b)
mes-o 'meat' mes-ar 'butcher' mes-ar-a, mes-ar-n(ic)a 'meatshop'
pek- 'bake' pek-ar 'baker' pek-ar-a, pek-ar-n (ic) a 'bakery'
knjig-a 'book' knjiz-ar 'bookseller' knjiz-ar-a, knjizar-n-ic-a 'bookstore'
The femininization cannot percolate from any suffix here for the only
suffix is -ar, which is no more feminine than masculine.
The exceptions to this pattern of feminine markers on "in'Mocative
nominalizations themselves form a subregular pattern. (14) exhibits an
166 Robert Beard
array of suffixes, -ac, -n-jak, -in-jak, -ar-n-ik, -n-ik (the masculine variant
of -n-ic-a) plus an unusual occurrence of the agentive suffix -or, kokos-
ar 'chicken coop', marking those "in'Mocatives based on animate nouns.
The Slavic languages distinguish animacy among masculine (and plural)
nouns at the inflectional level; that locative nominalizations are masculine
is hence perfectly consistent. Only the empty morpheme entailment ex-
plains this interlevel consistency.
(14) svinj-a 'hog' svinj-ac 'pig sty'
zee 'rabbit' zec-ar-nik, zec-in-jak 'rabbit warren'
zab-a 'frog' zab-n-jak, zab-ljak 'frog pond'
Exceptions to this subregularity follow the inanimate pattern in exhib-
iting the default, feminine gender: ovc-ar-a 'sheep pen', koz-ar(ic)a 'goat
pen', konj-us-n-ic-a 'horse stable', krav-ar-a 'cowshed', pset-ar-n(ic)a
'kennel'. Milojevic (1934), however, points out that terms like konj-ar-
nik often occur in speech and other masculine forms with -ar-nik are
acceptable to him: ovc-ar-nik 'sheep pen', koz-ar-nik 'goat pen', krav-ar-
nik 'cow shed'. We can only conclude that masculinity is now productively
associated with "in'Mocatives derived from animate nouns according to
the animacy-masculine association in inflectional usage. The gender con-
flict in these forms no doubt arises from the conflict between marked
and unmarked genders for nomina loci.
It does not follow, of course, from the fact that derivation rules can
determine gender that they always do. The diminutive derivation generally
does not determine gender (cf. Italian il bimb-o : il bimb-ett-o, la bimb-
-a : la bimb-ett-a); however, in German it does: der Bart : das Bärt-chen,
die Haut : das Häut-chen; das Mäd-el, Mägd(e)-lein, Mäd-chen. Such
variation is the result of language-specific factors which do not impinge
on the crucial issue raised by separation.
The Pashto deadjectival nominals seem to represent another violation
of the empty morpheme entailment.
(15) äräm 'quiet' äräm-i [F] 'quietness'
sur 'red' sur-tiä [F] 'redness'
arzan 'cheap' arzan-tob [M] 'cheapness'
klak 'firm' klak-välay [Μ] 'firmness'
The empty morpheme entailment 167
5. Conclusion
Notes
1. Zubin —Kopeke (1984, 1986) argue that words referring to "unimageable" superordinate
categories tend to be neuter, e.g., das Ding, das Gut, das Stück vs. der Tisch, die Tür.
However, their arguments (a) are necessary but not sufficient, (b) are based on converse
accidence, underestimating synonyms like die Sache, die Ware, and (c) ignore some
massive regularities which are counterexamples (das Fräulein, Mädchen) while depending
168 Robert Beard
heavily on others, e.g., das Kraut, Unkraut, Heilkraut, Würzkraut (p. 160) as examples
of independent unimageable superordinate categories.
2. This is not unexpected within the lexeme-morpheme based morphology of Beard (1988).
That article explains that the derivation rules are much more regular than affixation
rules, probably because they are universal, while affixation rules are locally determined
by specific languages.
3. The same is true of a handful of absolute French exceptions (e.g., le chaud 'warmth', le
froid 'cold') though not of the dozen or so Slavic exceptions (e.g., bogat-stvo 'wealth',
zdorov-'e 'health'). Here the logical gender, neuter, may be accounted for by affixation
but only exceptionally. Since masculine is the default (unmarked) gender in French, and
all zero marked derivations are generally masculine, lexical morphologists might argue
that the lexicon automatically assigns masculine to derivations not marked by gender-
bearing affixes. However, these derivates are still lexically excluded from affixation rules
and whether they are excluded from affixation by their exceptional masculine gender or
their masculine gender results from their exceptional null morphology is presently a
moot point.
4. Some grammars suggest that forms like those of (15b) are derived from the corresponding
agentives, e.g., mes-ar 'butcher' : mes-ar-a 'butcher shop'. This is unlikely, however,
since (a) agentives are not always available for locatives (ucio-nica 'classroom') and (b)
they are often irrelevant when available, e.g., igl-ar 'needle-maker': igl-ar-a 'pin cushion',
ovc-ar 'shepherd' : ovc-ar-a 'sheep pen, barn'. Such a derivational path would make the
lexical morpheme position even more difficult since (a) /-a/ is not a derivational affix
but a nominative singular declension I desinence and (b) -ar is a masculine agentive
suffix.
References
Anderson, Stephen R.
1982 "Where's morphology?", Linguistic Inquiry 13: 571-612.
Beard, Robert
1981 The Indo-European lexicon: a full synchronic theory ( = North-Holland
Linguistics Series 44) (Amsterdam: North-Holland).
1985 "Is separation natural?", Studia gramatyczne 7: 119 — 134.
1987 "Morpheme order in a lexeme/morpheme-based morphology", Lingua 72:
1-44.
1988 "On the separation of derivation from affixation; toward a lexeme/mor-
pheme-based morphology", Quaderni di Semantica 9: 3 — 59.
1989 Review of Di Sciullo - Williams (1987). Lingua 77: 81 - 9 0 .
Bloomfield, Leonard
1933 Language. (New York: Holt).
Boskovic, G.
1934 "Povodom clanka G. A. Paunova" [On the occasion of an article by G. A.
Paunov], Νas jezik 1: 275 — 278.
Chomsky, Noam —Morris Halle
1968 The sound pattern of English. (New York: Harper & Row).
Di Sciullo, Anna Maria —Edwin Williams
1987 On the definition of word. (Cambridge, MA: MIT Press).
Jackendoff, Ray
1975 "Morphological and semantic regularities in the lexicon", Language 51:
639-671.
The empty morpheme entailment 169
Lieber, Rochelle
1981 On the organization of the lexicon. (Bloomington: Indiana University Lin-
guistics Club).
Matthews, Peter H.
1972 Inflectional morphology. (Cambridge: Cambridge University Press).
Milojevic, V. Z.
1934 "Jedan predlog za nekoliko naziva" [A proposal for some terms], Νas jezik
2: 209-210.
Moortgat, Michael —Harry van der Hulst
1981 "Geinterpreteerde morfologie", Glot 4: 179-214.
Moskovljevic, Milos S.
1934 " Opkladionica' i tim povodom ο imenicama na -ionica, -aonica" [Opkla-
dionica, and on this occasion on substantives in -ionica, -aonica], Nas jezik
2: 7 6 - 7 7 .
Pounder, Amanda
in press "The semantic organization of word-formation paradigms and diachrony",
in: Proceedings of the International Morphology Conference at Veszprem ( =
Acta Linguistica Academiae Scientiarum Hungaricae 36).
Roeper, Thomas
1987 "The syntax of compound reference" [XIV International Congress of Lin-
guists, Berlin, August 1 0 - 1 5 , 1987],
Selkirk, Elizabeth
1982 The syntax of words ( = Linguistic Inquiry Monograph 7). (Cambridge,
MA: MIT Press).
Szymanek, Bogdan
1985 English and Polish adjectives: a study in lexical word-formation. (Lublin:
Wydawnictwo Katolickiego Uniwersytetu Lubelskiego).
Wood, Mary
1985 "The semantics and dynamics of derivation", in: G. A. J. Hoppenbrouwers
et al. (eds.) Meaning and the lexicon (Dordrecht: Foris), 49 — 54.
Zepic, Stanko
1970 "Izvedenice sa sufiksima za tvorbu mjesnih imenica (nomina loci)" [Inves-
tigation of suffixes forming place-names], Jezik 18: 83—90, 195 — 109.
Zubin, David —Klaus-Michael Kopeke
1984 "Affect classification in the German gender system", Lingua 63: 41 —96.
1986 "Gender and folk taxonomy: the indexical relation between grammatical
and lexical categorization", in: C. Craig (ed.) Categorization and noun
classification (Philadelphia: Benjamins), 139 — 180.
The benefits of morphological classification: on
some apparently problematic clitics in Modern
Greek*
Brian D. Joseph
* This work was supported by a Grant-in-Aid from the College of Humanities of the Ohio
State University, and by a Fulbright Research award that enabled me to spend time in
Greece collecting data and consulting with colleagues there. I would like to thank T.
Christides and the other linguists at the University of Thessaloniki and my colleague A.
Zwicky for stimulating discussion on these topics; A. Aikhenvald, V. Bubenik, B. de
Chene, C. Hagege, R. Janda, E. Petrounias, and E. Ternes all contributed useful comments
at the presentation of this paper at the Third International Morphology Meeting in
Krems, Austria, in July 1988.
172 Brian D. Joseph
The nature of this classificatory scale is beyond the scope of this paper,
but it should be clear that although the absolutist view of the criteria is
adopted here, the relativist interpretation is available for analysts who
are so inclined.
With these criteria in place, the two hypotheses about clitic behavior
mentioned above — second position for S-bar clitics and the ban on
endoclitics — can be explicitly tested. More particularly, apparent coun-
terexamples to these claims provided by some facts from Modern Greek
can be subjected to rigorous testing via Zwicky's criteria.
The problems posed by the Greek facts both center on the finite verbal
complex, a unit consisting of the inflected verb plus various elements
traditionally called "clitics", in the case of the weak object pronouns
(1SG ACC me, 1SG GEN mu, 3SG ACC NTR to, etc.), and "particles",
in the case of the modal markers na and as, the future marker θα, and
the negation markers den and min. These elements modify the inflected
verb for tense, mood, negation, and argument structure (see Joseph 1985
for some discussion). The general schema for the verbal complex is
sketched in (2), with some examples of possible expansions given in (3):
with indicative mood, min with subjunctive — pose a problem for the
modified Wackernagel's Law second-position generalization. To focus
just on the indicative negator, discussed further below,3 it need not occur
in second position (note (4a) with sentence-initial Sen) even though its
scope is demonstrably sentential, 4 in that it determines the selection of
the negative polarity indefinite pronoun kanenas as subject, as opposed
to the nonnegative käpjos:
(4a) den irOe kanenas / * kanenas irOe
NEG came/3SG no-one/NOM
'No one came'
b) käpjos irOe / *δen irOe käpjos
someone/NOM came/3SG
'Someone came'
It turns out, however, that there is a solution to these problems. A
close examination of the properties of the morphemes in question — the
weak pronominals and the indicative negator — with regard to the Zwicky
criteria allows for an analysis of these elements as more affixal in nature
and thus not clitics in the strict sense that these criteria now permit.
Crucial to this solution is the ability to classify the problematic elements
on a principled basis as being outside the domain of these generalizations.
In that way, both the negator den and the weak pronominals become
irrelevant for these hypotheses, and so do not constitute counterexamples
to them.
The affixal analysis that provides the key to preserving the above-men-
tioned generalizations is well-supported by facts from both Standard
Modern Greek and various dialects. The dialect evidence cannot of course
bear on the analysis in Standard Modern Greek per se. However, it does
provide "pan-Hellenic" support for the affixal analysis by lending typo-
logical credence; the dialects, as varieties of Greek, are closer to Standard
Greek in all respects — lexically, structurally, grammatically — than any
other language.
With regard to the indicative negator Sen, it is clear first that it must
be a nonword. It is phonologically dependent, unable to stand alone, for
The benefits of morphological classification 175
In addition, they are strictly ordered with respect to other elements they
combine with:6
(9) den ton νΐέρο
NEG him/ACC.WEAK see/1 SG
/ *ton Sen vlepo / *ton vlepo den j *Sen vlepo ton
Ί don't see him'.
On other criteria, moreover, the weak pronouns test out as affixes,
and not clitics.7 They show selectivity of combination, in general occurring
only with verbs, though accusatives also occur with the adverb kalos
'well' in the meaning 'welcome' (e.g., kalos ton 'welcome to him!') and
genitives occur with some prepositions (e.g., brostä mu 'near me') and a
few adjectives (e.g., monos mu 'on my own'). The existence of these
occasional nonverb hosts might suggest lesser selectivity, but these com-
binations are actually extremely limited, among adverbs, for instance,
restricted only to kalos. Also, the weak pronouns show various combi-
natory gaps, a type of selectivity recognized by Zwicky — Pullum (1983)
as typical of affixes, e.g., the absence of first person genitives with second
person accusatives (see Warburton 1977):
(10) se dosane
me/GEN.WEAK you/ACC.WEAK gave/3PL
'They gave you to me'.
In addition, the weak pronouns show several idiosyncrasies. The 2SG
genitive /su/ combined with third person weak accusative forms, which
all begin with /t-/, anomalously undergoes a contraction to [s], e.g., /su
+ to/ 'to you itT —* [sto], even though no regular elision process affecting
/u/ occurs in that context in standard Greek. Also, for some speakers,
the initial /t-/ of the weak third person pronouns may be voiced in
combination with the future marker Θα and the modal na, e.g., /0a to
käno/ —> [0a do käno] Ί will do it', even though intervocalic voicing is
not a regular rule of Greek phonology. Semantic and morphosyntactic
idiosyncrasies are found in the occurrence of weak pronouns in idioms
with otherwise intransitive verbs, e.g.:
(11) ρύ θα tin pesume
where FUT her/ACC.WEAK fall/1 PL
'Where will we go?' (literally "*Where will we fall her")
Not only is the weak pronoun semantically empty, with no interpretation
as an argument, but its co-occurrence with an intransitive verb is entirely
irregular too.
178 Brian D. Joseph
4. Conclusion
What emerges from this brief discussion is that these elements from
Modern Greek, which at first glance seem so problematic for claims
regarding clitics universally, turn out to present no problem, once a
stringent set of classificatory criteria is applied to the data. Rather than
being clitics, they are instead well-behaved affixal elements and as such
are irrelevant to the hypotheses discussed at the outset. A ban on endo-
clisis and second-position for S-bar clitics can therefore be maintained as
viable universals, given present knowledge. More examination of the
Modern Greek verbal complex is surely needed, though the preliminary
indications 9 are that it is a word-level unit of a verbal stem plus clitic
modal markers, tense and negation prefixes, affixal object markers, as-
pectual suffixes, and endings for the person and number of the subject.
Thus Zwicky's principles of morphological classification have enor-
mous benefits for research in this area. To the extent, then, that the
Zwicky criteria lead to satisfying results in dealing with these seemingly
difficult facts, the analyses presented here lend considerable credence to
the overall framework and approach to morphological classification em-
bodied in the interface program.
Notes
1. Some apparent cases of endoclitics have been reported for various languages, but closer
inspection in each case yields a better, nonendoclitic, analysis (e.g., Nevis 1984, Klavans
1983).
2. See Zwicky (1990) for an overview of the theory.
3. I am deliberately ignoring the potential counterexample posed by the modal negator
min, largely for reasons of space; although it is a safe assumption that the analysis of
den given below can carry over into the classification of min, the latter presents other
analytic problems that go beyond the scope of this paper (see Joseph—Janda 1987).
4. Similar problems for the modified Wackernagel's Law would arise, of course, if any of
the M O O D and TENSE markers in (2) are sentential in scope and are clitics (so also
with the modal negator min, see note 3).
180 Brian D. Joseph
References
Joseph, Brian D.
1985 "Complementizers, particles, and finiteness in Greek and the Balkans",
Folia Slavica 7.3: 390-411.
1988 "Modern Greek pronominal affixes: the case against clisis", CLS 24:
203-215.
in press "Diachrony and linguistic competence — the evidence from morphological
change", in: University of Chicago Special Publications in Linguistics vol. 2
(Papers from the Conference on the Theory and Practice of Historical
Linguistics).
Joseph, Brian — Richard D. Janda
1987 "On generalizations as linguistic categories: Of rules, redundancy-rules,
meta-rules, rule-conspiracies, and rule-constellations" [Paper presented at
the Conference on Linguistic Categorization, University of Wiscon-
sin—Milwaukee, April 1987].
Kaisse, Ellen
1982 "Sentential clitics and Wackernagel's law", Proceedings of the 1st West
Coast Conference on Formal Linguistics'. 1 — 14.
Klavans, Judith
1983 "The morphology of cliticization", Papers from the Parasession on the
Interplay of Phonology, Morphology, and Syntax (CLS): 103 — 121.
Malikouti-Drachman, Angeliki —Gabereil Drachman
1988 "On Greek clitics and lexical phonology" [Paper presented at the Sixth
International Phonology Meeting (Krems), July 1988],
Nevis, Joel
1984 "A non-endoclitic in Estonian", Lingua 64.2/3: 209-224.
Pernot, Hubert
1934 Introduction a l'etude du dialecte tsakonien. (Paris: Societe d'Editions "Les
Belles Lettres").
Thavoris, Andonis
1977 "Morfolojikä merikön iöiomäton tis öitikis make8onias" [The morphology
of some Western Macedonian dialects], First Symposium for Northern Greek
Linguistics (Thessaloniki: Institute for Balkan Studies), 75 — 95.
The benefits of morphological classification 181
Tzartzanos, Ahilefs
1909 Pragmatia peri tis sinxronu Oessalikis Sialektu [A treatise on the contem-
porary Thessalian dialect] (Athinai: P. A. Peträku).
Warburton, Irene P.
1977 "Modern Greek clitic pronouns and the 'surface structure constraints'
hypothesis", Journal of Linguistics 13: 259 — 281.
Zwicky, Arnold M.
1985 "Clitics and particles", Language 61: 2 8 3 - 3 0 5 .
1987 "Suppressing the Zs" Journal of Linguistics 23: 133 — 148.
1990 "Inflectional morphology as a (sub)component of grammar", this volume,
217-236.
Zwicky, Arnold M. —Geoffrey K. Pullum
1983 "Cliticization vs. inflection: English η 'Γ, Language 59: 5 0 2 - 5 1 3 .
Case markers and pragmatic strategies: Romanian
clitics
Maria Manoliu-Manea
are not redundant, but serve as case markers, (b) Even when they are
copies of NP's in the same sentence, clitics preserve some of their endo-
phoric function, and as such they may become means of discourse
strategies.
1. Noun declension
but
(5) Latin second declension: lupus 'wolf
NOM lupus lupu lup
GEN lupi lupojad lupu- lup
DAT lup ο lupo lup
ACC lupum lupu lup
As (4) and (5) show, only feminine nouns have inherited two case markers,
while the other nouns have not. New case markers have been sought in
order to replace the ones that have been lost.
(11) Mari-e -i
Mary-GEN/DAT -the-gen./dat
'Mary's/to Mary'
This distribution is probably due to the fact that, when co-occurrent with
proper nouns, the definite article loses its anaphoric value and becomes
a mere case marker. As it has been shown in 1.1, masculine nouns have
not preserved case endings which could attract the new ones, i.e., the
cliticized demonstratives. Feminine proper nouns have preserved a post-
posed case ending and consequently the attraction under discussion took
place. Compare:
In (22b), the possessive clitic /a/ is the only means of marking the
following noun as an attribute, while in (22a), the absence of the posses-
sive clitic leads to the interpretation of lui Ion 'the-gen./dat. John' as an
indirect object.
2.2.2. The anaphor which copies an indirect object may also play the
role of a proclitic case morpheme as shown by the following utterances:
(25) i- ai cerut cartea
her-DAT have-you asked-for book-the-ACC
Ane -it — Da' cuil
Ann-DAT/GEN -the-DAT/GEN? - but whom-DAT?
'You asked Ann for the book? — Who(m) else?'
where i 'her-DAT.' is a copy of the noun following the direct object and
plays the role of identifying the copied noun as an indirect object and
not as an attribute.
When the clitic copy of the indirect object is absent, the noun following
the direct object is interpreted as an attribute referring to the possessor
rather than as an indirect object:
190 Maria Manoliu-Manea
explaining her refusal to make any donation — and it is not meant "to
be seen" by the reader. In such cases, the definite article is sufficient for
marking topical and/or known entities.
3.2.3. With prepositional direct objects, the clitic has become obligatory
in standard Romanian if the common noun is not determined by a definite
description, but is topical and/or known, because pe, as many other
prepositions, does not accept a definite article, unless the noun is accom-
panied by a definite description 6 :
4. Conclusions
4.1. In contemporary Romanian, there are three clitics which can serve
as case markers: (a) the definite article, (b) the possessive article, and (c)
the personal anaphor which copies direct or indirect objects. The definite
article can be the sole bearer of a case marker for the following syntactic
functions: (i) either subject or direct object and (ii) either indirect object
or attribute (see 1.2). The possessive article marks the attribute in op-
position to the indirect object (see examples (22a) and (22b)). The cliti-
cized personal anaphor eliminates the possibility of interpreting the noun
in question as subject (see (23) and (24)).
4.2. Even when they lose their pronominal function, clitics preserve
traces of their endophoric values and, as such, they play an important
role in various discourse strategies. For example, when copying left-
dislocated direct objects and playing the role of case markers, clitics are
always paired with prominent discourse entities (topicalized or focalized
NP's — see examples (32) —(35)). They can also copy postverbal prepo-
sitional direct objects if representing paragraph topics (see (42)), topical/
definite NP's — when not followed by a definite description (see (41)) or
focalized entities (see (43)).
Notes
1. In the literal translation, the dash (-) links morphemes belonging to the same word; for
example, 'child-the-MASC-GEN/DAT' is the morphematic translation of the word
copilului 'of/to the child'. The slash (/) means 'either ... or'. The abbreviations are to be
read as follows: ACC: accusative, DAT: dative; FEM: feminine, GEN: genitive; MASC:
masculine, NOM: nominative; PL: plural, POSS.-CL.: possessive clitic; SG: singular;
SUBJ: subjunctive.
2. With masculine common nouns, the definite article has a special form for marking the
vocative: e.g., omule\ 'man-the-MASC-SG-VOC'.
3. In spoken Romanian, the article -ul is reduced to -u.
Romanian clitics 195
4. The "topic framework" is a compound of those entities derivable from the physical
context and from the discourse domain of any discourse fragment. The "domain of
discourse" is defined by the entities which have been introduced in the preceding co-
text (Brown-Yule 1984: 135-137).
5. When talking about the hierarchy of discourse, current approaches in discourse analysis
have brought into the picture interesting converging concepts such as "staging", "theme",
"relative prominence", etc. "Staging", for example, is a dimension of prose structure
which identifies the relative prominence given to various segments of prose discourse
(Clements 1979: 287). The most prominent entity is usually referred to by the term
"theme", either as the left-most constituent in the sentence or clause (Halliday 1967), or
as the grammatical subject of a series of sentences (Katz 1980: 26). The topic is 'the
event, character, etc. speakers want to talk about'. The concept of "stage distance" refers
to the difference between "close-ups", i.e., events which are foregrounded in order to be
presented as taking place before the addressee's eyes, who is supposed to become a
spectator (cinematographic effect) and backgrounded events.
6. For more details see Manoliu-Manea (1988).
References
Niculescu, Alexandru
1965 Individualitatea limbii romäne intre limbile romanice. Contribufii gramaticale
I [Romanian among the Romance languages. Grammatical contributions]
(Bucure§ti: Editura §tiintificä).
Rosetti, Alexandru
1986 Istoria limbii romäne de la origini ρΐηά in secolul al XVII-lea [The history of
the Romanian language from its origins to the 17th century] (Bucure§ti:
Editura §tiintifica).
Seidel-Slotty, Ingeborg
1940 "'Hypertrophie' der Pronomina im Rumänischen", BL 8: 142-145.
Zwicky, Arnold — Geoffrey Pullum
1983 "Cliticization vs. inflection: English n't", Language 59: 502-513.
Parasitic formation in inflectional morphology
Yves-Charles Morin
2. Defective paradigms
3. Parasitic formation
The past subjunctive is now obsolete in modern French. The fact that
there nonetheless exist some speakers who have internalized this tense
allows us to assume that an account of the verbal morphology of French
which includes the past subjunctive is a reasonable model of a natural
inflectional system and hence relevant to linguistic theory.
Past subjunctive stems are regularly derived from the corresponding
(unmarked) preterit indicative stems by adding the suffix /s/ for all
persons, except for the third person singular for which the past subjunctive
Parasitic formation in inflectional morphology 199
stem is identical to the preterit stem; e.g., from the preterit indicative
stem /dormi-/ of dormir, one derives two past subjunctive stems: (1)
/dormi/ for third person singular and /dormis-/ for all other persons. This
correspondence between preterit indicative and past subjunctive stems is
systematic , whether the preterit stem is regular, as in the case of excuser,
pret. /skskyza-/, past subj. /Ekskyza(s-)/, semi-regular as in the case of
devoir, pret. /dy-/, past subj. /dy(s-)/, or totally idiosyncratic, as in the
case of etre, pret. /fy-/, past subj. /fy(s-)/ or naitre, pret. /naki-/, past
subj. /naki(s-)/. Defective verbs such as extraire which lack a preterit
stem also lack a past subjunctive stem. Defective verbs which lack a
default stem, but have a specific preterit stem have a past subjunctive
stem, e.g., cheoir, pret. /fy-/, past subj. /fy(s-)/. 1
In this account of past subjunctive stems, one "does derive a 'stem'
(i.e., a part of one word-form) from another 'stem' (a part of another
word-form) of identical status", a process Matthews (1972: 86) calls a
parasitic formation. This runs counter the principles for inflectional
morphology proposed by this author, who then develops other tools to
circumvent the problem (Matthews 1972: 175 — 182).
Parasitic formations are equally problematic for a model of lexicon-
external inflection if the basis for the derivation is a stem which sometimes
does, and sometimes does not, belong to the lexicon. To see why, let us
examine how some specific cases could be analyzed in Anderson's model.
The only stem of naitre which can be inserted under a node having
the morpholexical representation [subjunctive past lpl] is the specific
preterit stem /naki-/ to which the regular affixes /-s-/, /-}-/, and /-5/ would
be added by the rules of the "phonological" component to form /nakisjo/.
The same analysis applies to all the verbs which have a specific preterit
stem. In the case of regular verbs, however, the lexicon does not have
access to any such stem — as regularly inflected forms are not generated
within the lexicon. One would thus be required to have two completely
different mechanisms for past subjunctive depending upon the presence
of a specific preterit stem in the lexicon. This is clearly inadequate, as
the formation of past-subjunctive forms is obviously the same in all cases.
One solution to the problems raised by the parasitic formation de-
scribed here for a model of lexicon-external inflection would require that
both preterit-indicative and past-subjunctive forms be derived from a
common intermediate stem, phonologically identical but nonetheless mor-
phologically distinct from the preterit indicative — that is to say, find an
alternate analysis without parasitic formation. For instance, one would
derive from the intermediate stems /*ekskyza-/, /*dormi-/, and /*naki-/
200 Yves-Charles Morin
both the preterit indicative stems /ekskyza-/, /dormi/, and /naki-/and the
past subjunctive stems /ekskyza(s)/, /dormi(s-)/, and /naki(s-)/. The idio-
syncratic stem /*naki-/ would be listed in the lexicon, while the regularly
inflected forms /*skskyza-/, /*dormi-/ would be derived from the default
stems /skskyz-/ and /*dorm-/ outside of the lexicon. The complete par-
allelism between preterit indicative and past subjunctive could now be
accounted for directly by the (lexicon-external) rules of the grammar
which could derive them both in similar ways from the intermediate
stems.
However, there does not appear to be any empirical justification for
such intermediate stems. In particular, the intermediate stems cannot be
interpreted as being "past" stems from which one could derive both past
subjunctive and past (traditionally called "preterit") indicative stems.
Although one can reasonably define a "past" morphological category to
include both past subjunctive and preterit indicative, it would also have
to include at least the imperfect indicative. For instance, congruence
patterns which could be used to justify a common past category, do not
distinguish between imperfect and preterit indicative verbs which can
equally govern past subjunctive embedded sentences: (i) il demanda [pret-
erit indicative] alors qu'on l'excusat [past subjunctive] vs. (ii) il demandait
[imperfect indicative] souvent qu'on l'excusat [past subjunctive] (cf. Grev-
isse 1980: section 2743). The imperfect forms, however, cannot be derived
from such intermediate stems, cf. the imperfect indicative stem /nss-/ of
(il) naissait of naitre vs. the preterit indicative stem /naki-/.
One could perhaps try to justify intermediate stems underlying both
preterit indicative and past subjunctive stems as a matter of principle,
because such stems cannot be morphologically derived from one another,
although they are obviously related. This stance would deny any relevance
to the worries about parasitic formation discussed by Matthews (1972),
and would claim that the distinction between lexicon-internal and lexicon-
external inflection is not an empirical issue. If this is the case, then no
argumentation is possible.
But if we assume that the distinction must be empirical, then there is
but one means to express in a linguistically significant manner the com-
plete regularity by which past subjunctive stems are systematically derived
from the corresponding preterit indicative stems — whether or not these
stems are completely predictable or highly idiosyncratic, and that is, by
allowing all such stems to appear in the lexicon. In other words, the
formation of regular inflected forms must be allowed within the lexicon.
Parasitic formation in inflectional morphology 201
4. Conclusion
Note
1. Bouix-Leeman et al. (1980: 81) only mention the third singular past subjunctive stem
/Jy-/, but my informants also used the general stem /Jys-/, without hesitation.
References
Anderson, Stephen R.
1977 "On the formal description of inflection", CLS 13: 1 5 - 4 4 .
1982 "Where's morphology?", Linguistic Inquiry 13: 571-612.
1986 "Disjunctive ordering in inflectional morphology", Natural Language and
Linguistic Theory 4: 1 — 31.
Bouix-Leeman, Danielle — Colonna-Cesari, Helene — Dubois, Jean — Sobotka-Kannas,
Claude.
1980 Larousse de la conjugaison. (Paris: Larousse).
Bresnan, Joan (ed.)
1982 The mental representation of grammatical relations. (Cambridge, MA: MIT
Press).
Chomsky, Noam
1965 Aspects of the theory of syntax. (Cambridge, MA: MIT Press).
Dell, Frangois
1970 Les regies phonologiques tardives et la morphologie derivationnelle du fran^ais
[PhD dissertation, MIT],
Grevisse, Maurice
1980 Le bon usage ( l l c ed.) (Paris & Gembloux: Duculot).
Halle, Morris
1973 "Prolegomena to a theory of word-formation", Linguistic Inquiry 4: 3 — 16.
Kiparsky, Paul
1982 "From cyclic phonology to lexical phonology", in: H. van der Hulst —Ν.
Smith (eds.) The structure of phonological representations (Dordrecht: Foris),
vol. 1, 131-175.
Lieber, Rochelle
1980 On the organization of the lexicon [Ph.D. dissertation, MIT].
Matthews, Peter H.
1972 Inflectional morphology. (Cambridge: Cambridge University Press).
Morin, Yves-Charles
1987 "Remarques sur l'organisation de la flexion des verbes frangais", ITL-
Review of Applied Linguistics 77/78: 13 — 91.
1988 "Disjunctive ordering and French morphology", Natural Language and
Linguistic Theory 6: 271 —282.
Piatt, Diana
1981 "Old Provencal verb inflection: the balance between regularity and irregu-
larity in morphology", in: T. Thomas-Flinders (ed.) Inflectional morphology:
introduction to the extended word-and-paradigm theory, (UCLA occasional
papers 4, Working papers in morphology), 41—71.
Plenat, Marc
1981 L' "autre" conjugaison ou de la rigularite des verbes irreguliers (Cahiers de
grammaire 3, Universite de Toulouse-le Mirail).
The mechanism of inflection: lexicon
representations, rules, and irregularities
Wolfgang Ullrich Wurzel
1. Lexical representations
The lexical units of a language are stored in the dictionary of the lexicon.
Usually, i.e., in "regular" cases, the lexicon representation of a word
contains only one representative form of the paradigm, its base form.
This base form, roughly corresponding to the so-called citation form, is
represented in modern Indo-European languages generally by the nomi-
native singular of nouns, the infinitive of verbs, and the unspecified form
of adjectives. But, of course, the categories of lexical representation of
individual word classes may differ from language to language. For ex-
ample, in Russian, the base form of adjectives is not their morphologically
unspecified (masculine) short form but their long form; and in Hungarian,
the base form of verbs is not represented by the infinitive but by the
204 Wolfgang Ullrich Wurzel
2. Paradigm-structure conditions
— Fem
(3 a) [e/Pl]
# Σ #
b) [e/Pl] ^ [n/D.P1]
e/Pl
c) [s/ G.Sg]
— Fem
The mechanism of inflection 207
same position in the word (slot) as in the German noun [e/Pl] and [s/Pl],
both concerning plural symbolization in suffix position. 8 The default
principle allows us to evaluate the unmarked class membership as the
predictable normal case and to dispense with explicit lexical specification
in cases where inflection-class membership does not strictly follow from
the extramorphological properties of words but where one out of several
inflection classes clearly proves unmarked (by quantitative relationships
and productivity). In German, monosyllabic feminines can fall into the
η-plural class like Bahn — Bahnen or into the e-plural class (with umlaut)
like Wand — Wände; but only the marked membership of the e-plural
class needs to be specified.
It is a consequence of the default principle that, for certain marked
cases, negative rule features have to be provided. For example, for
German feminines in -er and -el (as for monosyllabic ones), the mem-
bership of the «-plural class is unmarked as in Schwester 'sister' —
Schwestern and Wurzel 'root' — Wurzeln, but the two words Mutter
'mother' and Tochter 'daughter' form their plural without plural suffix,
but with umlaut. To avoid assigning them the feature [«/PI] by the
respective paradigm structure condition, they have to be specified in their
lexical entry by the feature [0Suff/Pl] 'no suffix plural rule'. Unlike other
inflection features, this of course prevents the assignment of the feature
for the unmarked rule but does not lay down the application of a marked
rule instead. Such a feature, again, can be the starting point for paradigm-
structure conditions. In the example given, the feature [0Suff/Pl] together
with [ + Fem] then implies the rule feature [Uml/Pl], compare Mutter —
Mütter.
Paradigm-structure conditions have the status of redundancy condi-
tions, that is, they are structural and not rule-like, transformational
regularities and thus comparable, e.g., to phonological structure condi-
tions. They reflect the existence and the specific structure of paradigms
and thus show that the paradigm is more than the sum of its inflectional
forms. The paradigm-structure conditions for a word class in a language
as a whole reflect the structure of the entire inflectional system concerned.
3. Inflection rules
Output:
a) //tiJ/ BF e/ N
b) //mens/Stae/N
Also previous inflection rules can already have been applied to the lexical
unit, for example, in German the dative plural of nouns is formed on the
basis of the plural forms, compare dative plural Tischen from //tiJ/BFe/N.
Μ is the morphosyntactic representation of the syntactic position into
which the word is inserted. It consists of the corresponding morphosyn-
tactic category features that are partly primary (like tense features in the
verb), partly due to structural or lexically determined assignment (case
features in subjects and objects) and to concord (case features in adjec-
tives). An inflection rule is applied if the feature constellation demanded
by it is given. If several rules relate to the same category/category bundle
and the same position in the word (slot) like the e-, n-, er-, and s-plural
rule for German nouns, then there will be no problems of application,
for the choice of the correct rule is strictly determined by the correspond-
ing rule feature. Then the word covered by an inflection rule will be
210 Wolfgang Ullrich Wurzel
4. Irregularities
singular applying to nouns of the types Bär 'bear' — dem Bären, Mensch
— dem Menschen, and Funke — dem Funken. So, irregular inflection
forms can also definitely correspond to general inflection rules. Therefore,
irregularity in inflection cannot be claimed merely on the basis of non-
applicability of inflection rules.
Of course, for the word Herz the genitive singular in -ens and the
dative singular in -en are marked forms, but this criterion is insufficient
for delimiting the range of irregularity because it already delimits the
range of regular-marked from that of regular-unmarked forms. Never-
theless, it is without doubt crucial for the irregular status of Herz which
inflection rules apply to the word. For the corresponding irregular forms
of the genitive and dative singular to be produced, Herz must have the
rule features [jw/G.Sg] and [«/D.Sg] in its lexicon representation. This
means that, by way of exception, case-rule features have to be referred
to, while otherwise only plural-rule features occur in the lexical represen-
tations of the entire German noun inflection. So, here the canonical
lexicon information is not sufficient to generate the correct inflectional
forms. Thus, we have arrived at the second decisive criterion for delimiting
irregular from regular cases. Now we can define the three ranges of
inflectional morphology that differ in their regularity and markedness as
follows:
(6 a) The inflectional forms of regularly inflecting words can be ex-
plained by inflection rules on the basis of the canonical "Kenn-
formen" of the system.
(i) In regular-unmarked cases exclusively the lexical base form
with its extramorphological (phonological and/or semantic-
syntactic) properties functions as the "Kennform".
(ii) In regular-marked cases the lexical base form and canonical
derived forms specified by rule features together function as
"Kennformen".
b) The inflectional forms of irregularly inflecting words cannot be
explained by inflection rules on the basis of the canonical "Kenn-
formen" of the system.
By means of this definition also regular individual cases and (non-
suppletive) irregular cases can now be distinguished. Irregularities are
always individual cases, i.e., the respective formation occurs only in one
word as in German Stadt — plural Städte or in very few words as in
bringen 'to bring '/denken 'to think' — preterits brachte/dachte. But not
all individual cases are also irregular, compare the German nouns Floß
212 Wolfgang Ullrich Wurzel
'raft' and Kloster 'cloister' as well as Käse 'cheese'. Floß and Kloster are
the only neuters without er-plural but with plural umlaut and their
inflectional behavior can be specified simply by the rule feature [Uml/
PI], i.e., with reference to the canonical plural form. Also Käse, the only
masculine with 0-plural ending in -e requires only the lexical specification
[0Suff/Pl] preventing the application of all suffix rules for plural. Due to
this feature, it is then assigned — quite regularly by the respective
paradigm-structure conditions — the features [s/G.Sg] and [«/D.P1], i.e.,
it functions entirely like a normal 0-plural word, compare Käse — plural
die Käse — gen.sing, des Käses — dat.plur. den Käsen with Sommer
'summer' — plural die Sommer — gen.sing, des Sommers — dat.plur. den
Sommern. So, the words Floß, Kloster, and Käse are — unlike Herz —
not irregular but actually form regular-marked "miniclasses".
Since irregular cases cannot be explained by inflection rules based on
canonical "Kennformen" they require special unsystematic information
for their inflection. Such information can consist in the statement of
unsystematic rule features (not relating to canonical "Kennformen"), of
idiosyncratic category markers and/or of suppletive stems.11 Then, from
these different possibilities there actually follows a typology of irregular-
ities in inflectional morphology. Compare the following types:
Type I: Unsystematic inflection rule features ("Kennformen")
Example: German Herz
Irregular forms: G.Sg. (des) Herzens, D.Sg. (dem) Herzen
Lexicon
representation: //herts/BF/N [+Neutr, njPI; ns/G.Sg, w/D.Sg]
The irregular forms can be explained by (independently motivated) in-
flection rules; the plural form is regular-marked.
Type II: Idiosyncratic categorial marker
Example: English ox
Irregular form: Plur. oxen
Lexicon
representation:
Neither the stem nor the plural suffix of the irregular form can be
explained by (independently motivated) inflection rules.
Type V: Suppletive stem with "incorporated marker
Example: Engl, foot
Irregular form: Plur. feet
Lexicon
representation:
Notes
1. Compare also recent papers by Anderson and Zwicky, especially Anderson (1986),
Zwicky (1985, 1989). The concept of inflectional system advanced here furthermore
shows a number of other characteristics common with Anderson's and Zwicky's
position.
2. In this field, plausibility considerations are largely dominant. There is not yet any
theory of the base form.
3. Compare Wurzel (1984: 51 ff.) where examples of combinations of base form and stem
inflection can be found.
4. For the concept of inflection class markedness see Wurzel (1988).
5. Compare Bittner (1985, 1988).
6. In Wunderlich (1985) and Wiese (1986), it is explicitly or implicitly assumed that s-
plural is generally regular for German nouns and that all other plurals, among them
the «-plural of feminines ending in -e (the only one possible for such nouns), are
irregular. Such an assumption may meet the requirements of the underlying model of
The mechanism of inflection 215
References
Anderson, Stephen R.
1982 "Where's morphology?", Linguistic Inquiry 13: 571 —612.
1986 "Morphological theory", Linguistics: The Cambridge Survey (Cambridge:
Cambridge University Press), 146 — 191.
Bittner, Andreas
1985 "Das 'Stark-Schwach-Kontinuum' der neuhochdeutschen Verben", Acta
Linguistica Hungarica 35: 31 —42.
1988 Starke 'schwache' Verben — schwache 'starke' Verben. Überlegungen zur
Struktur des deutschen Verbsystems im Rahmen der natürlichen Morphologie
[Dissertation A, Akademie der Wissenschaften der DDR, Berlin].
Kiparsky, Paul
1982 "Lexical morphology and phonology", Linguistics in the Morning Calm
(Seoul: Linguistic Society of Korea), 3 — 91.
Wiese, Richard
1986 "Schwa and the structure of words in German", Linguistics 24: 697 — 724.
Wunderlich, Dieter
1986 "Probleme der Wortstruktur", Zeitschrift für Sprachwissenschaft 5/2:
209-252.
Wurzel, Wolfgang U.
1984 Flexionsmorphologie und Natürlichkeit. Ein Beitrag zur morphologischen
Theoriebildung (Studia grammatica 21) (Berlin: Akademie Verlag) [In Eng-
lish: Inflectional Morphology and Naturalness (Dordrecht/Boston/Lancaster/
Tokyo: Foris)].
216 Wolfgang Ullrich Wurzel
dieting, say, the applicability of one inflectional rule from the applicability
of another, and thus describing the clustering of inflectional rules that
goes under the name of "paradigm class"), syntactic properties to syn-
tactic properties (predicting, say, one subcategory of a lexeme from
another), and so on. These can be systematized as in Table 1.
Table 1. Types of rules and their corresponding input and output entities
2. Inflectional rules
3. Some details
3.5. Gaps
but in the Classical language it lacks a Stem 1 (which is used for present
forms: *coepio Ί begin') (Hale - Buck 1903: sec. 199.2).
And there can be paradigm classes defined in part by the inapplicability
of particular inflectional rules: e.g., the English adjective class FA-
THERLY (vs. WORLDLY), VISIBLE (vs. ABLE), CROTCHETY (vs.
DUSTY) ..., for which inflectional rules realizing comparative and su-
perlative are inapplicable: *fatherlier vs. worldlier, etc. A redundancy rule
presumably relates (at least as a default) the phonology of these adjective
lexemes, involving stems with more than two syllables, to this morpho-
logical property of them; but there are also individual adjectives like ILL
that idiosyncratically have the property.
4. The lexicon
Here I would like to encourage a view of the lexicon of a language as
the set of all relevant information about its words, hence as the domain
of facts to be described by morphological rules for the language. The
Inflectional morphology as a (sub) component of grammar 227
6.2. Clitics
The simple picture sketched so far is known to be inadequate. At least
some of the items that have been labeled "clitics" require genuine com-
plications in the scheme of component interfacing.
At the outset, I must discard a collection of "leaning" elements as
being beside the current point. These are elements that are simply pho-
nologically dependent on adjacent material, forming prosodic units —
phonological words or phonological phrases, in particular — with them
(see Kaisse — Zwicky 1987 for a compact discussion and references). The
English complementizer particle lexeme THAT, for instance, can belong
to the phonological phrase of material following it, so that that Chris
can make a phonological phrase within the sentence I know that Chris
has gone. So long as the syntax and morphology of these elements present
no special features, they are not of interest in the present context, though
they do require stipulations via principles of prosodic domain formation,
principles describing the interface between morphosyntax and automatic
phonology.
It is also true that "syntactic dependence" does not in itself necessarily
cause difficulties for the description of the morphology-syntax interface.
Syntactic rules must be able to locate certain classes of lexemes .by
reference to the contents of some host constituent, in particular by
reference to the host's head or edge words — to stipulate, for instance,
that some class of adverbs is limited to occurrence after the first word
(that is, after the first unit of word rank) in a clause. In many cases this
is all that need be said.
What then of "phrasal affixes"? (The contrast is with "bound words",
the terminology being that of Nevis 1986 rather than of my own earlier
Inflectional morphology as a (sub) component of grammar 233
work on clitics.) Consider the English genitive. So far as its syntax goes,
it works in the same way as an affix; it realizes a grammatical category
feature, here with edge location. The fact that the realization of genitive
interacts with the phonological shapes of ordinary inflection leads us to
assign it to an outer layer of inflectional morphology. (This is not Nevis's
treatment, but it is the one advanced by Kanerva 1987 for Finnish
possessives and Zwicky 1987b for English genitives.) It appears that in
general phrasal affixes are to be analyzed syntactically as grammatical
category, features distributed like other such features (though the usual
situation is for phrasal affixes to be located at constituent edges, for
ordinary inflection to be located on heads) and morphologically as
constituting an outer layer of inflectional morphology.
We are left with bound words, like the English reduced auxiliary clitics
(auxiliary reduction: I'd be quiet, It's been noticed). Each bound word
instantiates a lexeme (WOULD, HAS) and so should be treated syntac-
tically as a formative, as a word, rather than as features.
It is sometimes suggested that bound words are just phonologically
dependent words, but several facts suggest otherwise: (a) The set of
bound-word clitics is often lexically idiosyncratic, as when the forms was
and were fail to participate in auxiliary reduction; (b) bound-word clitics
are often subject to surface filters (the "surface-structure constraints" of
Perlmutter 1971 and consequent literature) that limit their combinations
with one another and with their hosts, and constrain the ordering within
these combinations; and (c) bound-word clitics often show special mor-
phophonemics, as in the auxiliary-reduction shape alternations /wud/ vs.
/d/ and /haez/ vs. /z/. Instead, we need to say that bound word clitics in
combination with their hosts make a new sort of word-like unit — what
I will call a "morphosyntactic word" — for the purposes of morphology,
while they are simply independent words for the purposes of syntax. (The
idea that the words of syntax and the words of morphology need not be
coextensive is now a familiar one, thanks to such works as Sadock 1985
and Di Sciullo-Williams 1987.)
The treatment of bound words as part of morphosyntactic words
means that they too fill slots, and we can expect the principles of inflection
to carry over to bound words. The sort of morphosyntactic word that
deserves the label "clitic group" is then in effect an inflected word built
on an inflected word as stem, and the surface filters that constrain the
constituency of these clitic groups and the ordering of their parts are
nothing more than a slot calculus at a new level.
234 Arnold Μ. Zwicky
I must point out that which lexemes are bound-word clitics, or have
bound-word clitic alternants, in a language cannot necessarily be pre-
dicted on the basis of other properties of the lexemes. Bound words can
be spread across the syntactic categories of a language; some particle
lexemes are bound words, but then some (like dative TO and the personal
pronouns in English) are not; some closed (sub)category lexemes are
bound words, but then some (like the directional particle TO and the
auxiliary SHOULD in English) are not; and some minor category lexemes
are bound words, but then some (like the English quantifiers EACH,
ANY, EVERY, ALL) are not.
Bound-word clitics are, of course, not the only phenomena that appear
to require a divergence between the representations appropriate for syntax
and those appropriate for morphology, and so to require that we for-
mulate principles describing the interface between morphology and syn-
tax, principles (whether parochial or universal) constraining the associ-
ations between morphosyntactic words and syntactic words. There is in
fact a panoply of such phenomena, including portmanteaus (French du),
compounds (English apple eater), serial verbs (English go look), clause
unions (French faire partir), and incorporations (West Greenlandic ga-
mutegarpog 'sled-have'). Bound-word clitics are also not the only in-
stances of word-like units (whether in syntax or in morphology) that
properly contain units of similar type; so do compounds, serial verbs,
clause unions, and incorporations.
7. Conclusion
rations about ordering that has been explored in the recent syntactic
literature is carried through to morphology, where I have suggested that
slot filling and slot ordering should be separated as well.
References
Anderson, Stephen R.
1986 "Disjunctive ordering in inflectional morphology", Natural Language and
Linguistic Theory 4: 1—31.
1988a "Inflection", in: M. Hammond —M. Noonan (eds.) Theoretical morphology
(Orlando: Academic Press), 23—43.
1988b "Morphological theory", in: F. J. Newmeyer (ed.) Linguistics: The Cam-
bridge survey (Cambridge: Cambridge University Press), vol. 1, 146—191.
Carstairs, Andrew
1987 Allomorphy in inflexion. (London: Croom Helm).
1990 "Phonologically conditioned suppletion", this volume, 17 — 23.
Di Sciullo, Anna Maria —Edwin Williams
1987 On the definition of word (Linguistic Inquiry Monograph no. 14) (Cam-
bridge, MA: MIT Press).
Donegan, Patricia J. —David Stampe
1979 "The study of natural phonology", in: D. Dinnsen (ed.) Current approaches
to phonological theory (Bloomington: Indiana University Press), 126—173.
Dressier, Wolfgang U.
1985 Morphonology: the dynamics of derivation (Ann Arbor: Karoma).
Gazdar, Gerald — Ewan Klein — Geoffrey Pullum — Ivan Sag
1985 Generalized phrase structure grammar. (Oxford: Basil Blackwell).
Haiman, John
1980 Hua: a Papuan language of the Eastern Highlands of New Guinea. (Amster-
dam: Benjamins).
Hale, William G . - C a r l D. Buck
1903 A Latin grammar [reprinted 1966] (University of Alabama: University of
Alabama Press).
Hinnebusch, Thomas J.
1979 "Swahili", in: T. Shopen (ed.) Languages and their status (Cambridge, Mass.:
Winthrop), 209-293.
Jackendoff, Ray
1975 "Morphological and semantic regularities in the lexicon", Language 51:
639-671.
Kaisse, Ellen M. —Arnold M. Zwicky
1987 "Introduction: syntactic influences on phonological rules", Phonology Year-
book 4: 3 - 1 1 .
Kanerva, Jonni M.
1987 "Morphological integrity and syntax: the evidence from Finnish possessive
suffixes", Language 63: 498 — 521.
Kiparsky, Paul
1982 "From cyclic phonology to lexical phonology", in: H. van der Hulst — N.
Smith (eds.), The structure of phonological representations, part 1, (Dord-
recht: Foris), 131-175.
236 Arnold Μ. Zwicky
Matthews, Peter
1972 Inflectional morphology: a theoretical study based on aspects of Latin verb
conjugation. (Cambridge: Cambridge University Press).
Nevis, Joel A.
1986 Finnish particle clitics and general clitic theory [PhD dissertation, Ohio State
University 1985] (OSU Working Papers in Linguistics 33).
Perlmutter, David M.
1971 Deep and surface structure constraints in syntax. (New York: Holt, Rinehart
& Winston).
Sadock, Jerrold M.
1985 "Autolexical syntax: a theory of noun incorporation and similar phenom-
ena", Natural Language and Linguistic Theory 3: 379—439.
Sadock, Jerrold M. —Arnold M. Zwicky
1985 "Speech act distinctions in syntax", in: T. Shopen (ed.), Language typology
and syntactic description (Cambridge: Cambridge University Press), vol. 1,
155-196.
Sapir, Edward
1921 Language: an introduction to the study of speech. (New York: Harcourt,
Brace).
Scalise, Sergio
1984 Generative morphology. (Dordrecht: Foris).
Spencer, Andrew
1988 "Arguments for morpholexical rules", Journal of Linguistics 24: 1 —29.
Stump, Gregory T.
1984 "Two approaches to predictive indeterminacy", Linguistics 22: 811—829.
Wurzel, Wolfgang U.
1990 "The mechanism of inflection: lexicon representations, rules, and irregular-
ities", this volume, 203-216.
Zwicky, Arnold M.
1985 "How to describe inflection", Proceedings of the Berkeley Linguistic Society
11: 372-386.
1986 "The general case: basic form versus default form", Proceedings of the
Berkeley Linguistic Society 12: 305 — 314.
1987a "Phonological and morphological rule interactions in highly modular gram-
mars", Proceedings of the Third Eastern States Conference on Linguistics
(Columbus: Ohio State University), 523 - 532.
1987b "Suppressing the Zs", Journal of Linguistics 23: 133-148.
1988 "Morphological rules, operations, and operation types", Proceedings of the
Fourth Eastern States Conference on Linguistics (Columbus: Ohio State
University), 318-334.
to appear "Jottings on adpositions, case inflections, government, and agreement", in:
D. Brentari et al. (eds.), The joy of grammar: A Festschrift for James D.
McCawley.
in press "The morphology-syntax interface", Oxford International Encyclopedia of
Linguistics.
Topic 4: Computer morphology
Morphology in LDOCE and in the ASCOT system
Willem Meijs
The main goal of the "ASCOT" research project 1 was the development,
on the basis of the computerized Longman Dictionary of Contemporary
English (LDOCE for short, editor-in-chief Paul Procter, published 1978),
of an automatic tagging system, i.e., a software package which can
"process" an uncoded text word by word and attach one or more gram-
matical code-tags to every word that it can "recognize".
ASCOT consists of three interacting components: a scanner, a lexicon
(Aslex), and a morphological component (Reroot). The scanner inspects
incoming text-sentences word by word. For each word-form it encounters
it makes a search through Aslex to see if there is an entry identical in
form to the text-word. Since ASCOT must make many passes through
the lexicon, Aslex is stored in the form of an L-tree (Skolnik 1980), which
allows very fast access. If an entry (or entries) identical to the text-word
is/are found, the relevant information stored under the entry or entries
is retrieved and copied onto the text-item. We call this a "straight hit".
If no such entry/entries can be found control is passed to Reroot. By
peeling off possible affixes and comparing the remainder with the entries
in Aslex, Reroot tries to establish whether the text-word is perhaps an
inflected and/or derived form based on an Aslex-entry (for instance, a
plural or a past participle, or a form prefixed with non-, pseudo-, etc.). If
this turns out to be the case, information associated with the base-form,
insofar as it is still relevant to the inflected or derived form, is copied
onto the text-item, along with an indication of the inflection or derivation
involved. We call this an "indirect hit". If neither the straight search via
the scanner, nor the indirect one via Reroot is successful, a "no success"
code is attached to the text-word and the scanning passes to the next
text-word.
Often a straight hit does not exhaust all of the information that could
be retrieved. Thus it can happen that some item may be an inflected/
240 Willem Meijs
derived form or a base form. Thus the noun fitter would be found as a
straight hit (since it has an entry of its own), but if the system were then
to pass on straightaway to the next word, the information (retrievable
via Reroot) that this word might also be the comparative form of the
adjective fit would go unnoticed. Similarly, the straight hit moped as noun
(meaning "bike with small engine") would block its identification as past
tense or participle of the verb mope, the form summons could be either a
noun or third person present tense of the verb summon, etc. (cf. Akkerman
et al. 1985:51, for some more examples). Since one cannot predict be-
forehand for which straight hits further morphological analysis via Reroot
might reveal additional information, the scanner should in fact always
make two passes, one to scoop up the straight hits, the other to reap the
indirect ones.2
abacus1"
re + -a-d v -able A
I
iA
se-a N -kale N
I I
e V !V/N. e r N
un + -der + p -mine v
ite v wear v
just A
re-quitedA
I
s-ervedA
I
jV/N
zulu N
Figure I.
3. Morphology in ASCOT
However, assuming a system which does not stop at "straight hits", the
number of perfectly ordinary words that would get an additional com-
pound analysis in a second pass turned out to be astounding. To give a
few examples, the adjective barbed (as in barbed wire) would be analyzed
as a compound of the nouns bar and bed, the adjective paintable as a
table having something to do with pain, the verb dampen as some sort of
pen somehow connected with a dam, while ticking would presumably be
interpreted as "the king of tics"! Similarly disastrous results would ensue
for words like beefish, flushed, rampant, reddish, and innumerable other
ones; cf. the bracketings in (3):
[|pain]^[table]^]N vs [\paint]vable]A
vs [[barb]^ed\A
Pe*>]N[/i's/2]N]N vs [[beef\Nish]A
[[/7w]N[s/ii?i/]N]N vs [[flush]ved]v.e(i
vs [[red(d)]Aish]A
[[ram]N[/?fl«i]N]N vs [rampant]A
vs [[damp]Aen]v
[[fic]N[fci#ig]N]N vs [[tic(k)]ving]v_ing
4. Conclusion
Notes
1. "ASCOT" (which stands for "Automatic Scanning System for Corpus Oriented Tasks")
is the name of a project supervised by the author and funded by the Dutch Organization
for the Advancement of Academic Research (ZWO) under proj. no. 300-169-004. This
project was carried out between 1-3-1984 and 28-2-1987, with a three-month extension
financed by the Amsterdam University Arts Faculty. Research assistants were: Pieter
Masereeuw (until 1-7-1985), Hetty Voogt-van Zutphen, and Eric Akkerman. For details
about ASCOT see Akkerman — Masereeuw — Meijs (1985) and Akkerman — Meijs —
Voogt-van Zutphen (1987, 1988).
2. Actually, we have made the choice between one or two passes a "user option", with two
passes as the default situation. The "one-pass" option is provided for syntactic analysis
systems that have a morphological analyzer of their own.
3. As it turns out, there are three different formats in which morphologically complex
words can occur in LDOCE: (1) in full, as headword, (2) in full, within (usually at the
end o f ) the entry for the base-word, and (3) in truncated form, with hyphen or tilde,
within the entry for the base-word. What motivates the choice of the particular format
adopted is not always very clear.
References
1. Introduction
In her recent book, Words in the mind, Jean Aitchison (1987: 9) makes
much the same point in terms of a very telling analogy. Suppose, she
suggests, that we viewed the mental lexicon as something analogous to a
library, in which information was stored in the forms of books. From
the standpoint of economy of storage alone, the most efficient approach
would obviously be to stack books of the same size together, filling each
room from wall to wall and floor to ceiling with tightly packed piles.
From the standpoint of ease of retrieval, however, this would constitute
perhaps the least efficient approach: to retrieve a book from such a
library, we would need information about its size (rather than its content)
in order to know what room to go to, and then we would have to extract
the books, one-by-one, pile-by-pile, working from the door to the back
wall, until we chanced to hit upon the particular volume of interest. Thus
we find that the real libraries that we are familiar with are organized on
quite different principles, placing a premium on ease of retrieval, with
few concessions to economy of storage.
Likewise, in real mental lexicons, too, we have no reason to think that
a heavy premium need be placed on economy of storage: the human
mind seems capable of storing almost unlimited amounts of information,
certainly far beyond the average person's normal everyday or even lifetime
needs, and in the case of words, can easily manage the 250,000 or so
entries that Diller (1978) estimates for the typical educated English-
speaking adult. (And this is in the monolingual case; the addition of one
or more other languages does not seem to impose any significant new
burden on the normal memory.)
On the other hand, we do have very good reasons for thinking that a
premium must be placed on the process of retrieval, in view of the fact
that the normal processes of speech production and (especially) compre-
hension must operate within time constraints that are extremely severe.
In one study, for example, it has been shown that native speakers can
recognize a word in their language in approximately one-fifth of a second
from its onset, or can reject a sound sequence that is a non-word in as
little as half a second (Marslen-Wilson — Tyler 1980) — and some of that
time involves the motor reaction after the decision is made. I conclude
that there can be no serious question that the human word-store is
primarily organized to enhance rapid and accurate retrieval and so we
should not be in the least bit surprised if a considerable amount of
redundancy in storage might be tolerated in order to expedite this process.
As Aitchison (1987) also reports in her excellent survey, the psycholin-
guistic evidence to date points consistently in this same general direction.
Whether taken from speech errors, aphasic speech, word-games, or from
Morphology and the mental lexicon: psycholinguistic evidence 251
Analysis 1A (Rhyme)
In any event, to check these two alternatives out, I tested some local
children on nonce-forms that didn't have any rhymes and, for the fuller
picture, I systematically varied the nonce-words used in such a way that
only the final consonant was left unchanged (as in the set wug / droig j
oog) and learned that it didn't make any (significant) difference. For each
stem-type, so long as the final consonant (or consonant cluster) was
preserved, the results were generally the same (see Derwing — Baker 1980
for a summary of these results).
At the time, the conclusion my colleagues and I drew from these data
was that the rhyme analogy was wrong and that the final-segment strategy
was right (or was at least still a candidate for the "one ultimate truth").
A new theoretical perspective has emerged in recent years, however, which
puts the preceding discussion in an entirely new light.
This approach goes informally by the designation of "connectionism,"
or, more technically, "parallel distributed processing", the most familiar
version of which is closely associated with the names of Rumelhart —
McClelland (1986). As the name implies, what all connectionist models
have in common — including a recent linguistic alternative to Rumel-
hart — McClelland proposed by Skousen (1989) — is the attempt to
account for morphological productivity through the establishment and
subsequent utilization of a network of analogical connections among full
lexical items.
One interesting innovation of the analogical approach is that one form
of a rule need not be chosen over another. In fact, in the familiar linguistic
sense of the term, there are no rules per se involved in such accounts at
all. As seen in Fig. 1, both the rhyme and final-segment analogies are
supported by sufficient connections to allow for a solution to the problem
in a Berko-type task.
Thus, from this perspective, Berko's subjects could have used either
the rhyming strategy (by tracing the dotted connection lines) or the final-
254 Bruce L. Derwing
Analysis 1C (Feature)
— iz / [+sibilant]
— sibilant
-s/
(PI) < — voiced
— sibilant
-z/ + voiced
256 Bruce L. Derwing
Responses
Cluster Phoneme
Irr. Null /IZ/ /s/ Μ Dupl.
Μ 0 4.0 0 0 96.0 0
Μ 0 0 0 0 100.0 0
Μ 0 0 0 0 100.0 0
/ö/ 0 0 0 0 100.0 0
IV 0 0 0 0 100.0 0
Μ 0 0 0 0 100.0 0
Μ 0 0 0 0 100.0 0
Η 0 0 0 0 100.0 0
Μ 0 0 0 0 100.0 0
AV 0 4.0 0 0 96.0 0
/g/ 0 4.0 0 0 96.0 0
Μ 0 8.0 0 0 92.0 0
Ν 0 12.0 0 0 84.0 4.0
Μ 0 68.0 28.0 0 0 0
Μ 0 92.0 8.0 0 0 0
1)1 0 72.0 20.0 8.0 0 0
ßl 4.0 48.0 28.0 20.0 0 0
Icl 0 44.0 44.0 12.0 0 0
HI 8.0 52.0 36.0 4.0 0 0
Table 2. Object clusters for the " > 2 1 correct" group (n = 18)
Responses
Cluster Phoneme
Irr. Null Μ Ν Μ Dupl.
Μ 0 0 0 0 100.0 0
Λ)/ 0 0 0 0 100.0 0
Ν 0 0 0 0 100.0 0
/»/ 0 0 11.1 0 83.3 5.6
Ν 0 0 0 0 100.0 0
Μ 0 0 5.6 0 94.4 0
/η/ 0 0 0 0 100.0 0
/m/ 0 0 0 0 100.0 0
Μ 0 0 0 0 100.0 0
/d/ 0 0 0 0 100.0 0
/g/ 0 0 0 0 100.0 0
Μ 0 0 0 0 100.0 0
Ν 0 0 0 0 100.0 0
/f/ 0 0 0 100.0 0 0
/Ρ/ 0 0 0 100.0 0 0
Ν 0 0 0 100.0 0 0
Ν 5.6 0 0 94.4 0 0
ßl 0 0 5.6 83.3 0 0
Μ 0 22.2 77.8 0 0 0
Μ 0 33.3 66.7 0 0 0
1)1 0 11.1 83.3 0 5.6 0
β/ 0 0 94.4 5.6 0 0
/δ/ 0 5.6 88.9 5.6 0 0
m 0 0 100.0 0 0 0
Actually, we now have data to suggest that the analysis 1 D is the best
one, which posits two lexical variants and one low-level phonological
rule:
Analysis I D (Hybrid)
— iz / [ +sibilant]
(PI) =
-ζ / Ε
+ obstruent
R l . [ +obstruent] —> [ — voiced] /
— voiced
4. Conclusions
In sum, after a long period of neglect, rapid progress has been made in
recent years in the psycholinguistic investigation of morphology, both in
terms of the character of the mental lexicon and the ways in which
morphological productivity might be accounted for. As would be expected
from such a surge of research activity, new and even potentially revolu-
tionary theoretical ideas have also begun to emerge. At least as often as
not, these new discoveries have flown in the face of received knowledge
in linguistics (such as support for the full-listing hypothesis, as contrasted
to a notable lack of support for anything like the generative idea of a
morpheme-invariant underlying form); we have also seen some of the
pivotal rule proposals considerably weakened, such as English lax-vowel
insertion (or even completely undermined, as in the case of the phono-
logical analysis of English vowel-shift; see Wang —Derwing 1986); finally,
the challenge has even been raised that rule-based systems may be fun-
damentally misguided. Considering the overall scorecard to date, perhaps
the time is ripe to rethink some of our most fundamental and cherished
theoretical assumptions, now that evidence is accumulating from psycho-
linguistic research that they may be dead wrong. To be sure, progress is
not always painless, but it is very exciting nonetheless.
Notes
1. The issue as to how these "stages" were defined, and the justification for them, goes
well beyond the scope of this paper, but is dealt with in some detail in Baker — Derwing
(1982) and Derwing-Baker (1986).
2. Except the /z/-stems, whose special status can be explained (see Derwing—Baker (1979:
212).
3. Linguists often use the word rouges to illustrate this pattern in introductory courses,
but this is a highly marked plural of a mass noun which children are unlikely to know.
The most likely analog in many dialects is the plural form of garage (itself rather rare
a usage), but this is a /j/-stem in the dialect tested.
4. Note that analysis 1 D illustrates the notion of "phonologically conditioned suppletion",
as developed by Carstairs (1990).
5. See Derwing (1973) and especially (1976) for extensive illustrations of the analogy-
drawing capabilities of the normal child.
6. While few, if any, generativists would actually want to incorporate sound-spelling
relations within the same purview of phonological/morphological description, Skousen
shows that the analogical approach is equally appropriate to either domain.
7. Cf. Derwing (1979) and Derwing — Baker (1979) on the "competition" among alternative
formal patterns, and Bates — MacWhinney (1987) and MacWhinney (1987) for a more
general "competition model".
8. Grammaticality/acceptability judgements show a similar indeterminacy, as Ross (1979),
in particular, has emphasized.
Morphology and the mental lexicon: psycholinguistic evidence 263
References
Aitchison, Jean
1987 Words in the mind. (Oxford: Basil Blackwell).
Baker, William J. —Bruce L. Derwing
1982 "Response coincidence analysis as evidence for language acquisition strat-
egies", Applied Psycholinguistics 3: 193 — 221.
Bates, Elizabeth — Brian MacWhinney
1987 "Competition, variation, and language learning", in: B. MacWhinney (ed.)
Mechanisms of language acquisition (Hillsdale, N. J./London: Lawrence Erl-
baum), 157-193.
Berko, Jean
1958 "The child's learning of English morphology", Word 14: 150-177.
Butterworth, Brian
1983 "Lexical representation", in: B. Butterworth (ed.) Language production (New
York: Academic Press), vol. 2: 257 — 294.
Carstairs, Andrew
1990 "Phonologically conditioned suppletion", this volume, 17 — 23.
Chomsky, Noam
1965 Aspects of the theory of syntax. (Cambridge, Mass.: MIT Press).
Cutler, Anne
1983 "Lexical complexity and sentence processing", in: G. B. Flores d'Arcais —
R. J. Jarvella (eds.) The process of language understanding (New York:
Wiley), 4 3 - 7 9 .
Dennis, Diane
1988 Rule governed behaviour in English inflectional morphology [M. Sc. Thesis,
University of Alberta].
Derwing, Bruce L.
1973 Transformational grammar as a theory of language acquisition: a study in the
empirical, conceptual and methodological foundations of contemporary lin-
guistics. (London: Cambridge University Press).
1976 "What kind of rules can children leam?", in: W. von Raffler-Engel —
Y. Lebrun (eds.) Baby talk and infant speech (Amsterdam: Swets & Zeitlin-
ger), 6 8 - 7 8 .
1979 "Psycholinguistic evidence and linguistic theory", in: G. D. Prideaux (ed.)
Perspectives in experimental linguistics (Amsterdam: Benjamins), 113 — 138.
1987 "A cross-linguistic experimental investigation of syllable structure. Part I:
background and methodology", in: S. Delancey —R. M. Tomlin (eds.) Pro-
ceedings of the Third Annual Meeting of the Pacific Linguistics Conference
(Department of Linguistics, University of Oregon, Eugene, OR), 93 — 102.
Derwing, Bruce L. — William J. Baker
1974 Rule learning and the English inflections (Final report to the Canada Council,
File No. S72-0332).
1979 "Recent research on the acquisition of English morphology", in: P.J.
Fletcher—Μ. Garman (eds.) Language acquisition: Studies in first language
development (Cambridge: Cambridge University Press), 209 — 233.
1980 "Rule learning and the English inflections (with special emphasis on the
plural)", in: G. D. Prideaux — B. L. Derwing —W.J. Baker (eds.) Experi-
mental linguistics: integration of theories and applications (Ghent: Story-
Scientia), 247-272.
1986 "On assessing morphological development", in: P. J. Fletcher—Μ. Garman
(eds.) Language acquisition: Studies in first language development (Cam-
bridge: Cambridge University Press), 2nd ed., 326 — 338.
264 Bruce L. Derwing
1. Introduction
In this paper, I will deal with the following two topics: first, the impor-
tance of the analysis of existing words for the study of morphological
creativity and, in relation to that, second, the role of analogy in the
process of the creation of new complex words. By stressing these points,
this paper joins in with the tradition established by (among others)
Humboldt, Paul, and Saussure, according to which existing words, both
simplex and complex, are subjected to a constant, never-ending in-
spection on the part of the speakers of the language. In this conception,
the lexicon must, in the words of Humboldt, not be equated with "eine
fertig daliegende Masse" since it is — even if we leave the coining of
new words outside consideration — "ein fortgehendes Erzeugniss und
Wiedererzeugniss des wortbildenden Vermögens" (Humboldt 1836:
109-110).
A recent rephrasing of this view can be found in Mötsch (1987: 24):
"The creation of new words (...) presupposes rules. But rules need not
have an existence of their own. We may conceive of rules as the result of
a process of analysis operating on similarity of items of the vocabulary".
Motsch's views come close to the stand taken in van Marie — Koefoed
(1980) and van Marie (1985) where it was argued that the speakers
of a language have the capacity to construct rules on the basis of the
existing words, a capacity we called "rule-creating creativity". This
ability, then, directly bears upon analogy as a synchronic morphological
force.
In the following, I hope to make clear that this never-ending inspection
of the words in the lexicon may have different results and may result in
* I am indebted to Geert Booij and Caroline Smits for their critical comments on an earlier
draft of this paper.
268 Jaap van Marie
(a) -in: ap-in, boer-in, cf. aap 'ape, monkey' and boer 'farmer'
-es: prinses, cf. prins 'prince'
-se: kasteleinse, cf. kastelein 'innkeeper'
-e: leerlinge, cf. leerling 'pupil'
-ster: wandelaarster, cf. wandelaar 'walker, promenader'
Sehr häufig ist der Fall, dass eine Ableitung aus einer Ableitung in
direkte Beziehung zum Grundworte gesetzt wird, wodurch dann auch
Rule-creating creativity 271
*
-iste ( < neutral -ist 4- female -e),
cf. emiriste ( = female emir 'id.')
4. Conclusions
In the preceding I have discussed two mechanisms which I consider to
bear upon analogy as a synchronic force. In my view this discussion
justifies two conclusions. First, what these mechanisms have in common
Rule-creating creativity 273
is that they take existing complex words as their starting point. I am, of
course, aware of the fact that the notion "existing complex word" is a
much debated one. Several linguists have doubted the relevance of this
notion to competence theories of word-formation. The consequence of
this stand is clear: such theories may provide insight in several respects,
but they are unfit to deal with analogy and with the type of word-
formation upon which analogy bears. To put this differently, within the
framework of theories which deny the notion "existing complex word"
all phenomena discussed here must be viewed as part of performance.
Secondly, even on the basis of the above we cannot but ascertain that
analogy is a rather heterogeneous phenomenon.
In my opinion it is the joint operation of these two factors which
underlies the fact that so many morphological theories have not succeeded
in coming to grips with analogy as a synchronic process.
References
* The research reported in this paper was supported in part by grant number BNS84 — 18277
from the US National Science Foundation, and the paper was written while the author
was supported in part by a fellowship from the John Simon Guggenheim Foundation.
The support of both of these organizations is gratefully acknowledged.
278 Stephen R. Anderson
types is then the same as the range of available parameter settings. Insofar
as there are not many possible settings, there will not be very many
possible types; and insofar as the languages avail themselves in roughly
equal numbers of the available options, they will distribute in roughly
equal fashion among the types. We can then use the parameter values
themselves as convenient labels for the resulting typological categories.
It makes excellent sense, however, to ask just how likely it is that a
situation comparable to that found in syntax will in fact turn out to
obtain in morphology. How plausible is it, that is, that the system of
word-formation rules in a language will turn out to be globally para-
metrized in the way syntactic structure seems to be, as implied in any
serious search for a morphologically based typology of languages? Some
efforts have been made along these lines by, for instance, those who think
of morphology as the syntax of words and who focus on a program of
extending essentially syntactic insights to domains within the word. Re-
gardless of how one feels about the ultimate productivity of this enterprise
on its own terms, it still leaves most of the more mechanical side of word
formation — getting the segments right and getting them in the right
places — largely untouched.
When one turns to the details of how words are composed of pho-
nological material, my own impression is that the number of substantial
correlations within a language that could give rise to a productive para-
metrization of the sort sketched above is actually very limited. Much of
the reason for this (as stressed by Hagege 1990) is the fact that word
structure is not really an autonomous domain in the same way syntax
and phonology are. Much more of what one finds in morphological
systems is the result of the interplay of other areas of grammar, and
especially of diachronic change; and it is in these other domains that the
basic parameters of a language's structure are presumably to be sought.
However interesting it may be to study, much of morphology may in
some sense be an epiphenomenon.
To shift the emphasis of the discussion somewhat, though, and despite
the rather pessimistic attitude suggested above toward the potential sig-
nificance of any typology of morphological systems, the search for such
a typology has an important role to play; and indeed, when put into
proper perspective, it constitutes another way of posing the most fun-
damental questions there are about word structure. This is because, beside
the search for overall correlations among phenomena that might yield a
minimal specification of the range of languages in the world, there is
another methodology available for pursuing typological studies. This is
282 Stephen R. Anderson
to attempt to ensure that whatever the account that is given of some area
of linguistic structure (morphology, for instance), it is one that will be
adequate to accommodate all of the systems that might be encountered.
One can do "typology", that is, in the sense of exploring the full range
of diversity in the languages of the world, and without necessarily feeling
that the effort has been a failure if that diversity is not ultimately reducible
to some small number of binary- or ternary-valued parameters.
This is actually a perfectly respectable goal to take along in exploring
the typology of morphological systems (among other parts of language),
even though it is not what most people think of as the goal of typology.
But now notice that on this reading, the development of a genuinely
adequate typology is a project which is essentially co-extensive with the
development of a generally applicable morphological theory. That is, a
theoretical descriptive framework which is really adequate to all of the
world's languages, can be seen as constituting in itself the most central
kind of a "typology", even if its descriptive dimensions do not reduce to
some small number of parametric possibilities. Of course, where corre-
lations are to be found, they ought properly to be incorporated into the
theory; and if enough are found, maybe it will be possible once again to
interpret the theory as a set of labels for language types. But this is only
one of many possible outcomes, and in the meantime there seems no
particularly good reason to distinguish fundamentally between linguistic
typology and linguistic theory, construed as a general enterprise.
Let us, then, turn to the most fully developed notion of word-structure
typology in the traditional literature, taking this to be (at least the outline
of) a theory of the range of constructs that are necessary in principle for
the description of the morphological systems of all possible languages.
As I suggested at the outset of this paper, the system outlined in Sapir's
Language goes as far as anything in this direction, and it is that framework
that I would like to examine briefly here.
Following a tradition represented also by his teacher Franz Boas,
Sapir starts from the notion that the word structure system of a language
is composed of two kinds of object: a set of grammatical processes and
a set of concepts expressed by the application of those processes. There
must also be, of course, a basic word stock that serves as the foundation
for morphological elaboration. Now the important thing about these two
sets is the fact that they are in principle quite independent of one another:
b) [Choctaw]
i. bonot-li 'roll up (sg. obj.)'; bon-li 'roll up (pl. obj.)'
ii. bakaaf-li 'split (sg. obj.)'; bak-li 'split (pl. obj.)'
c) [Koasati]
i. atakaa-li 'hang (sg.)'; atak-li 'hang (pl.)'
ii. lataf-ka-n 'kick (sg.)'; lat-ka-n 'kick (pl.)'
Martin shows that the operation involved in these forms is one which
derives the plural (or iterated, etc.) form from the singular by dissociating
the final syllable from the prosodic pattern and then re-associating its
onset consonant melody into the preceding rhyme (if possible, consistent
with the syllabic structures permitted in the language). Such a process
would have been called a "subtractive morph" in Hockett's (1947) guide
to the structuralist morphological zoo; what is important for our purposes
is that it has no apparent (coherent) reformulation as the addition of an
affix.
With the modifications suggested above, then, a list3 like Sapir's can
serve as a description of the grammatical processes available to individual
languages. Each particular word-formational process invokes one of them
(or perhaps more than one at a time, as in the case of German suffixes
accompanied by stem umlaut). As the basis of a typology, such a list can
be used to characterize the subset of what is formally possible that is
actually instantiated in a particular system. Where a language has no
complex words at all, of course, it requires no word-formation processes,
and this can be regarded as the limiting case which is identified by the
term "isolating". Where all of the rules involve affixation, pure and
simple, with no other changes, this is the defining characteristic of an
"agglutinating" language. If at least some of the word-formation rules of
a language involve non-affixational processes (internal change, metathe-
sis, subtraction, accent shift, etc.), this is the basis for calling the mor-
phology of that language "symbolic" (to use Sapir's term).
Finally, the phenomenon that was classically supposed to make Indo-
European languages of the ancient type so special was the presence of
complex affixation: affixes that are accompanied by some morphologi-
cally motivated phonological change. Languages with such internally
complex processes are called "inflecting" in an earlier terminology, one
that seems sufficiently misleading to require replacement, as for instance
by Sapir's term "fusional". The defining characteristic of this type is
primarily the fact that segmentation into neat, discrete morphemes is
rendered more difficult by the extent to which other effects accompany
Sapir's approach to typology 287
in some positions from the semantics of some part of the structure (e.g.,
grammatical number in English, which is grammatically relevant as the
basis of the agreement relation between subjects and verbs, but which is
a function of the semantics of the subject NP).
Sapir then observes that the members of these categories are not all
equally necessary to the structure of language in general. In order to say
anything at all, of course, the speaker of a language must have at hand
some meaningful words, and so every language must express "basic" or
"radical" concepts. Furthermore, these words must be capable of being
put together syntactically, and the representation of syntactic structure is
the business of the "relational" concepts: indicating subject-hood and
object-hood, status as modifier vs. modified, etc. Much of the structure
indicated by pure relational concepts may be reflected by word order
alone, and so is of no particular interest to a theory of word structure.
Recall, however, that Sapir himself includes word order as a "grammatical
process": in consequence, he can claim that every language must of
necessity have some formal expression of his pure-relational concepts,
since otherwise it would have no syntactic structure. Abstracting away
from word order, then, we can see that "pure-relational" concepts will
be represented morphologically exactly where information of a purely
syntactic nature is carried by the forms of words.
As opposed to "basic" and "pure-relational" notions, however, which
must necessarily be formally represented in some fashion in every lan-
guage, there is no necessity for a language to have any "derivational"
concepts at all. It is always possible to represent a complex meaning as
a (syntactically formed) combination of two or more basic meanings, like
'little tree', instead of as a derivational function modifying a single basic
meaning (as in 'tree-let'). Similarly, it is not necessary for a language to
make use of any semantically significant property, like number or (nat-
ural) gender, as one of the devices that indicate grammatical structure
such as the modifier-modified relation or that between a predicate and
its subject. Mixed-relational concepts as a category are thus dispensable
to particular languages as well.
Sapir thus suggests that we can classify languages according to what
elaborations of the basic, irreducible inventory of concept types ("basic"
and "pure-relational") they employ. That is, a language may optionally
make use of derivational concepts; mixed-relational concepts; both of
these; or neither. It might seem that the resulting four classes of language
would provide a rather clear-cut categorial distinction, and that any given
language would belong to exactly one of the four set types; but in fact
290 Stephen R. Anderson
make the others happy too, though. And seeing the typological and the
theoretical enterprise as basically the same should provide a worthwhile
insight for all concerned.
Notes
1. It is not actually obvious that the affixation analysis solves all the problems involved in
reduplication. In particular, the actual mechanics of association of melodic material with
the empty skeletal positions analyzed as an affix of reduplication are not completely
straightforward. It remains to be seen whether all instances of apparent reduplication
can be successfully reduced to the simple addition of an affix which happens to be
melodically underspecified.
2. This is because both ke- and -an, in the limited circumstances in which they are
independently attested as affixes, produce nouns from various sorts of bases.
3. Naturally, we would like to go beyond listing to provide a more explanatory account
of the class of grammatical processes. Along with the apparent seductiveness of the
classical morpheme, this seems to be some of what lies behind attempts to reduce all of
morphology to affixation. It does not seem possible to accomplish that reduction,
unfortunately; but the alternative of saying simply that the class of grammatical processes
is delimited by the possibilities of a rich tranformational formalism is unsatisfactory as
well. One possibility suggested by Martin (1988) is the claim that the operations available
to morphology are exactly those available to the phonology, thus reducing one unsolved
problem to another. A complicating factor is the role of historical change in shaping
synchronic morphologies: this may have the consequence that actually attested systems
contain an unrepresentative (or at least seriously skewed) selection from among the
theoretically possible processes. For some discussion of this, see Anderson (1980) and
Janda (1984).
4. The notion of "meta-rules" over the morphology was suggested in Anderson (1986).
Janda (1982 and elsewhere) has explored in some detail the notion that individual
languages make extensive use of a small number of processes.
5. The word "lexicon" has been used in a number of rather distinct senses in the recent
literature: as the domain of the "lexical" or "cyclic" rules of the phonology; as the set
of surface-grammatical word forms of a language (not necessarily the same as the set
of its surface-phonological words); as the set of its uninflected (but possibly complex)
stems, perhaps including compounds; etc. It is roughly this last sense of "lexicon" that
we would like to invoke here, but given the confusion that exists in the literature about
what it means to say something is "in the lexicon" we would simply avoid using the
word as much as possible. Aronoff (1988) makes some particularly sharp comments
concerning confusions in the use of "lexical" in recent linguistic discourse.
6. Actually, this is already clear from the fact that the same sense can be conveyed either
by a basic concept associated with a root or by a derivational function. Since both basic
and derivational concepts are localized within the dictionary, however, the indeterminacy
involved here is less radical than the point referred to in the text.
7. Although this ethnonym has a securely established status in the linguistic literature, it
is actually not an accurate name for the people Sapir referred to (more accurately known
as the Tsishaath or Tseshaht — [c'lsa-^ath]) or for their language. Since this paper is not
fundamentally concerned with the ethnography of the northwest coast of North America,
however, we will perpetuate Sapir's usage here in the spirit of quotation.
Sapir's approach to typology 295
References
Anderson, Stephen R.
1980 "On the development of morphology from syntax", in: J. Fisiak (ed.) His-
torical morphology (The Hague: Mouton), 51 —69.
1986 "Disjunctive ordering in inflectional morphology", Natural Language and
Linguistic Theory 4: 1 — 32.
1988 "Morphological theory", in: F.J. Newmeyer (ed.) Linguistics: The Cam-
bridge Survey (Cambridge: Cambridge University Press) I: 146 — 191.
Aronoff, Mark
1988 "Two senses of lexical" [Unpublished paper, SUNY, Stony Brook].
Broadwell, George Aaron
1987 "Subtractive morphology in Southwest Muskogean" [Paper presented at
the 40th Annual Kentucky Foreign Language Conference].
Derbyshire, Desmond — Geoffrey K. Pullum
1981 "Object-initial languages", UAL 47: 192-214.
Dressier, Wolfgang Ulrich
1985 "On the predictiveness of natural morphology", Journal of Linguistics 21:
321 - 3 3 7 .
1987 "Subtraction in a polycentristic theory of natural morphology", in: E.
Gussmann (ed.) Rules and the lexicon: studies in word-formation (Lublin:
Wydawnictwo Katolickiego Uniwersytetu Lubelskiego), 67 — 77.
Greenberg, Joseph H.
1963 "Some universale of grammar with particular reference to the order of
meaningful elements", in: J. H. Greenberg (ed.) Universals of language
(Cambridge, MA: MIT Press), 7 3 - 1 1 2 .
Hagege, Claude
1990 "Do the classical morphological types have clear-cut limits?" this volume
297-308.
Hockett, Charles F.
1947 "Problems of morphemic analysis", Language 23: 321 —343.
Janda, Richard
1982 "On limiting the form of morphological rules: German umlaut, diacritics,
and the 'cluster constraint'", NELS 12: 140-152.
1984 "Why morphological metathesis rules are rare: on the possibility of historical
explanation in linguistics", PBLS 10: 8 7 - 1 0 3 .
MacDonald, R. Ross
1976 Indonesian reference grammar. (Washington: Georgetown University Press).
Marantz, Alec
1982 "Re reduplication", Linguistic Inquiry 13: 435-482.
Martin, Jack
1988 "Subtractive morphology as dissociation", Proceedings of the 7 th West Coast
Conference on Formal Linguistics.
McCarthy, John J.
1981 "A prosodic theory of non-concatenative morphology", Linguistic Inquiry 12:
373-418.
Montier, Timothy
1986 An outline of the morphology and phonology of Saanich, North Straits Salish
(Occasional Papers in Linguistics 4, University of Montana).
Sapir, Edward
1921 Language (New York: Harcourt, Brace & World).
Do the classical morphological types have clear-cut
limits?
Claude Hagege
1.1. Hungarian
Consider a form häz-a 'his house', and compare it with häz-a-i 'his houses'
and with häz-a-i-k 'their houses'. Judging only by these three forms, we
could immediately state that there is a perfect application of the "one
form —one meaning" principle typical of agglutinative languages, since
we can posit k for plural of possessor, i for plural of the possessed object,
and a for third person singular (of possessor). However, the same a vowel
also appears in häz-a-m 'my house', häz-a-d 'your (sg.) house', häz-a-i-m
'my houses' and häz-a-i-d 'your (sg.) houses', where the possessor is
singular, but not third person, as well as in häz-a-i-n-k 'our houses', häz-
a-to-k 'your (pi.) house' and häz-a-i-to-k 'your (pi.) houses', where the
possessor is neither singular, nor third person, these forms, along with
häz-un-k 'our house' and häz-u-k 'their house', only providing confir-
mation as to the status of k and i.
Consequently, as far as a is concerned, we are left with a dilemma.
Either we analyze it, in all the forms where it appears except in häz-a,
häz-a-i and häz-a-i-k, as a simple link vowel, and in these three latter
forms as a marker of the third person (singular) of the possessor, with
such an analysis implying, in a very uneconomical way, that the speaker,
within one paradigm, uses two different kinds of a, which have unrelated
functions. Or we posit am, ad as markers of the first and second person
singular of possessor respectively, un, ato, u as first, second, and third
person respectively when there is more than one possessor, these markers
being combined with k, the common marker for plural possessor and
a...n, resp. a...to as variants for first and second person plural possessor.
But since we have already assigned a meaning to i and k and since there
is no distinctive function attributable to a in the first and second person
singular when there is a single possessed object, nor in the whole paradigm
when there is more than one possessed object, we could also say that,
for instance, the markers of the first and second person of the possessor
are m and d respectively, in which case the one for third person singular
is zero. 1 But this conclusion is itself untenable; although it might seem
justified to the extent that we take into account the paradigm and its
pressures, it makes the simple description of a central feature of Hun-
Do the classical morphological types have clear-cut limits? 299
1.2. Turkish
I will briefly describe here some phenomena which all tend to contradict
the widely shared assumption of the "perfect" agglutinative character of
Turkish morphology. Consider the following list of alternations (from R.
Underhill (1986: 14)):
Nominative Definite accusative
sebep sebeb-i cause
kelebek <— kelebeg-i 'butterfly'
hak <r- hakk-i 'right'
§ehir §ehr-i 'city'
zaman ([zamän]) zaman-i ([zama:nui]) 'time'
These examples illustrate various phenomena which occur at mor-
pheme boundaries. These phenomena should not occur if it were true
that Turkish is purely agglutinative. We see that a syllable-final voiced
stop is devoiced when it is followed by zero (sebep j sebeb-i, from Arabic
sabab); corresponding to a velar consonant at the end of a word, we have
a fricative or zero in an intervocalic environment (kelebek/kelebeg-i)·,
two syllable-final geminate consonants alternate with a word-final simple
consonant (hakk-ijhak, from Arabic haq) — if the consonant cluster is
made of two sonorants, then, in word-final position, an epenthetic vowel
will be inserted between them (§ehr-ij§ehir, from Persian sahr); a short
vowel in a word-final closed syllable alternates with a long vowel in the
syllable preceding -ι (zamanjzaman-i, from Persian zamän).
300 Claude Hagege
1.3. Kannarese
The agglutinative character of Kannarese appears clearly in its nominal
system, but there are some irregular facts even there. 2 Thus, given katte
'donkey', -galu plural suffix (preceding the case marker), and -0, -annu,
-a, -ige, and -inda, markers of nominative, accusative, genitive, dative,
and instrumental respectively, we get the following paradigm:
Singular Plural
Nominative katte-0 katte-galu-0
Accusative katte-y-annu katte-gal-annu
Genitive katte-y-a katte-gal-a
Dative katte-ge katte-gal-ige
Instrumental katte-y-inda katte-gal-inda
Several facts deserve to be noticed here. First, a -y- is inserted in the
singular before -annu, -a, and -inda, but the -i- of -ige, which appears in
the plural, is absorbed in the singular at the boundary between katte and
-ige. Second, we observe that the -u of -galu is deleted before the initials
of the case markers in the plural (except for the nominative, which is 0).
One might say that these small "accidents" do not detract much from
the generally agglutinative nature of such a system, as also evidenced by
the transparency of nominal groups with multiple modifications, like
huduga-r-ibbar-ig-öskara-vägi-yü
(boy-pl.-two-dat.-goal-purposive-inclusive)
'for the two boys too'
But if we examine the verbal system, we find a much more complex
situation. The present of the verb ele 'pull', for instance, is as follows:
Person Number Gender
ene 1 sg
eve 1 pl
iye 2 sg
iri 2 Pi
ele-y-utt- äne 3 sg masc.
äle 3 sg fem.
ade 3 Sg neut.
äre 3 pi masc./fem.
ave 3 pl neut.
Do the classical morphological types have clear-cut limits? 301
From this paradigm, we can infer that -utt- is the present tense marker
(linked to the verb ele by -y- since the latter ends in a vowel), and that
First of all, two series of examples can be given here, which shed light
on the reasons for the situation presented above.
2.1. Estonian
2.2. Palauan
This Austronesian language, spoken on the westernmost island of the
Carolines, would be assigned to the agglutinative type considering that
many forms, in the noun and in the verb systems, are derived through
suffixation of clearly identifiable morphemes. However, some of the
suffixes are stressed, or enclitic, which results in strong alterations of the
root vowels of stems combined with such suffixes. Let us take as an
example the noun phrase expressing inherent possession. 6 The possessive
suffixes constitute the following paradigm (V = stressed vowel):
sg. 1. -Vk pi. 1. inclusive -Vö
1. exclusive -(m)äm
2. -Vm 2. -(m)iu
3. -VI 3. -(r)ir
3. Conclusion
Two logically related conclusions can be drawn from the present study.
Firstly, polytypical complexity is the result of phonetic evolution. Con-
sequently, this study shows that morphology is not an autonomous
domain: word structure cannot be analyzed short of having recourse to
historical phonetics, or if it is, the kind of analysis that such a choice
implies remains at the quite insufficient stage of mere description, and
does not teach us what we can expect to learn about the characteristics
of human languages. Secondly, and as a consequence, if we do not try
to delve into the reasons for polytypical complexity, then we cannot meet
the main requirement which prompts every linguist to do linguistics: to
know more about man's nature. Linguistics is a social science, and as
such, it has something to teach us about human qualities.12 Speakers-
listeners, who are far from being "ideal", as was assumed in Chomsky
(1965), build their language, through generations, less unconsciously than
306 Claude Hagege
Notes
1. A zero-morpheme is proposed for the third person singular suffix in the "several
possessions" paradigm {-(j)aimj-(j)eim, -(j)aid/-(j)eid, -(j)ai/-(j)ei, etc.), in Kiefer
(1985: 89).
2. Cf. Aronoff-Sridhar (1984: 7, 9).
3. This principle is violated even in "perfectly agglutinative" languages, like those in the
Turkic family: in Azeri, the -am and the -ssn of kalir-am and kdir-sm, respectively Ί
come' and ' y ° u come' are blendings of person and number, since the plural has forms
which are not analyzable synchronically: -ik and -siniz. Likewise, in Osmanli, for the
same verb, we have, in the present, gel-i-yor + -um, -sun, -uz and -sunuz respectively.
Furthermore, in Salar and in Saryg-Yugur, Turkic languages spoken in the Chinese
province of Kansu, there is no mark for the plural possessive, kak-am being, for
example, in Salar, 'my brother' as well as 'our brother'. In the declension of the personal
pronouns of these languages, we find inflectional features: in Salar, Τ is men, 'of me'
is mi, 'to me' is maga or ma. All these phenomena are not particularly "agglutinative",
to say nothing of the «regular existence» of characteristic features considered by
Austerlitz (1970) as defining an agglutinative type ("developed participial system, the
finite verb as cloture marker at the end of the sentence", etc.).
Do the classical morphological types have clear-cut limits? 307
4. The following data are taken from Hint (1981), and from A l l i k - H e l p - P a k k (1987).
5. The other two processes, early alternation of geminate stops and historically late
quantity alternation, are still productive in the modern language as is evidenced by the
treatment of recent loanwords (cf. Hint (1987)).
6. All the Palauan material comes from my own fieldwork; the results are published as
Hagege (1986).
7. The detail of this process, rule 6a of the ten rules characteristic of Palauan morpho-
phonemics, is presented in Hagege (1986: 30). It is one of the most idiosyncratic and
striking features of this language: the unpronounceable CwC group (where w =
semi-vowel), which results from the deletion of the root vowel whose stress has been
shifted to the suffix, yields a pronounceable word by vocalization of w: taut 'aim' +
el 'his' —> *twt-el —• tut-el.
8. For some exceptions to this rule, see Hagege (1986: 23).
9. By full vocalic grade, I do not mean "underlying form", since I reject this notion,
especially when it is confused with an historical restitution. An abuse of that kind leads
Wilson (1972: 47) and Josephs (1975: 497) to posit, in order to "explain" the i in the
possessive inflection of ker 'question' (cited above), a "derivation" such as keri —> keri
—» ker, by rules of "stress assignment" and "unstressed vowel deletion". As a conse-
quence, Dressier (1974: 139) mentions Wilson and Josephs' "data" as illustrating his
generalization on the vowel types that are the most threatened with deletion by
apheresis, syncope, and apocope. As a matter of fact, this is by no means a productive
rule in modern Palauan, but only a very old and reconstituted historical process.
10. Cf. Hagege (1986: 22).
11. However, there might be some discussion on whether the Hebrew facts are really mere
violations of the "one form —one meaning" principle.
12. The linguistic theory which underlies this view is expounded in Hagege (1988).
References
Palauan 3 0 3 - 3 0 8
Finnish 100, 150, 233, 261, 3 0 1 - 3 0 3 Pashto 166 f.
French 14, 18 f., 4 6 - 5 2 , 5 5 - 5 7 , 63 f., 66, Persian 299
100f„ 106, 108, 1 3 3 - 1 3 8 , 161 - 1 6 5 , Polish 36, 164
168, 1 9 7 - 2 0 2 , 228, 234, 268 Provencal, Old 201
Fulfulde 18, 21
Romanian 1 8 3 - 1 9 4
Georgian 221 Russian 32, 148, 162, 164, 168, 203, 205,
German 5, 9 - 1 1 , 32, 6 9 - 7 6 , 8 0 - 8 3 , 100, 210, 213
107, 148, 156, 1 6 0 - 1 6 4 , 1 6 6 - 1 6 8 ,
2 0 4 - 2 1 5 , 221, 223, 229, 286, 288 Saanich 285
Greek, Ancient 71, 1 1 3 - 1 1 6 , 1 2 6 - 1 2 9 , Salar 306
145, 147, 155, 175 f., 278 Sanskrit 18, 21, 279
Greek, Modern 1 7 1 - 1 8 0 Saryg-Yugur 306
Greenlandic 150, 234 Serbo-Croatian 36, 1 6 4 - 1 6 6 , 168
Gujarati 150 Sen 17, 21
310 Index of languages
extended word-and-paradigm model 20, 293 Kennform 207, 211 ff, 225
extrinsic rule ordering 37 f.
language of love 73 ff.
fallback procedure 252 learnability 37 — 39
feature-node 258 left branch condition 11,15
feminization 268 left-dislocated objects 191, 194
final-segment strategy 253 f. lexeme-morpheme based morphology 168
focalization 188, 190 f., 194 lexical component 54
foregrounded events 193 — decision tasks 254
full-listing hypothesis 251, 261 — insertion 201
function composition 65 f. — phonology 20
function of rules 251 — pragmatics of morphology 69
fusional 286, 297 — semantics of morphology 69
lexical-morpheme hypothesis 264 f.
gaps 222 f. lexicalist hypothesis 228, 290
government 9 — 12 lexicon-external/internal inflection 197 ff.
grammatical concepts 282 f., 288 ff., 294 locality 20
- basic 288 f., 294 locative nominalization 164 ff.
— concrete-relational 288 ff. logical form 3, 8 — 11
- derivational 288 ff., 294 loose compounds 105
— mixed-relational 288 ff.
— pure-relational 288 ff. mapping rules 219, 221 f.
- radical 288 f., 294 markedness 24, 79 ff, 204 f.
grammatical processes (sensu Sapir) 217, markedness isomorphism 31
282 ff., 293 markers of discourse prominence 192
meta-rule 287, 294
hailstone models 20 ff. metaphorization 119f.
head 61 f., 82ff., 102-107, 134f. metathesis 285
homonymy 46 ff., 52, 163 f. middle (voice) 145
minor rule 51, 56
iconicity 25 modification 81, 284
idiosyncrasy 46, 55, 201, 213 modular grammar 217
incidental rules 269 ff. morpholexical phonology 35 — 39
incorporated markers 213 f. morphological creativity 267 ff.
indexicality 185 morphology-free syntax 228
inflection-class membership 206 morphopragmatics 69 — 76
inheritance of argument structure 83 f. morphosemantics 69
integration of foreign compounds 109 f. morphosyntactic representation 209
interactive orientation 71, 74 — word 233
interface program 172, 217 multiple inflection 207
intermediate stems 199 f.
intermorphs 20 ff. negation 174 ff.
intimacy 72 neoclassical compounds 112
irregular(ity) 205, 210-214 nonmorphological word-creating rules
"is a" condition 101 136ff.
island 9 - 1 2
isolating 284, 286 object clitics 191 ff.
isomorphism 25 one form — one meaning principle 297
Subject index 313
Ferenc Kiefer
A Magyar Tudomänyos Akademia, Nyelvtudomanyi Intezete,
Budapest I., Szenthäromsag utca 2, H —1250 Budapest, Hungary
Aditi Lahiri
Max-Planck-Institute for Psycholinguistics, Wundtlaan 1, Postbus 310,
NL —6500 AH Nijmegen, The Netherlands
Maria Manoliu-Manea
Department of French and Italian, Sproul Hall, University of California,
Davis, California 95616, USA
Willem Meijs
Engels Seminarium, Universiteit van Amsterdam, Spuistraat 210,
NL —1012 VT Amsterdam, The Netherlands
Yves-Charles Μ or in
Departement de linguistique, Universite du Quebec ä Montreal,
Case postale 6128, Succursale "A", Montreal, P.Q. H3C 3J7, Canada
Wolfgang Mötsch
Akademie der Wissenschaften der DDR, Zentralinstitut für Sprachwis-
senschaft, Prenzlauer Promenade 149 — 152, 1100 Berlin,
German Democratic Republic
Sergio, Scalise
Dipartimento di Italianistica e Filologia Romanza, Universitä di Venezia,
Dorsoduro 3246, Ca' Foscari, 1 — 30123 Venezia, Italy
Michael Shapiro
Department of Slavic Languages, Brown University, Box Ε, Providence,
RI 02912, USA
Andrew Spencer
The Polytechnic of Central London, Faculty of Languages,
9 - 1 8 Euston Centre, London NW1 3ET, Great Britain
List of contributors 317
Irene Vogel
Department of Linguistics, College of Arts and Sciences,
University of Delaware, 46 E. Delaware Avenue, Newark,
Delaware 19716, USA
Beatrice Warren
Stockholms Universitet, Engelska Institutionen, S —106 91 Stockholm,
Sweden
Wolfgang U. Wurzel
Akademie der Wissenschaften der DDR, Zentralinstitut für Sprachwis-
senschaft, Prenzlauer Promenade 149 — 152, 1100 Berlin,
German Democratic Republic
Wiecher Zwanenburg
Vakgroep Romaanse Taalen en Kulturen, Rijksuniversiteit te Utrecht,
Kromme Nieuwegracht 29, NL —3512 Utrecht, The Netherlands
Arnold Μ. Zwicky
Department of Linguistics, The Ohio State University, 204 Cunz Hall,
1841 Millikin Road, Columbus, Ohio 43210-1229, USA
m Eung-Do Cook * Keren Rice
m Athapaskan Linguistics
m Current Perspectives on a Language Family
1989.15.5 χ 23 cm. VIII, 645 pages. With 1 map. Cloth.
m ISBN 311011166 7
[Trends in Linguistics. State of the Art Reports 16)
m
This volume represents an attempt to show the pre-
m sent state of the art in the study of this group of
Native American languages spoken in a large part of
m Alaska, as well as on the California coast and in the
American Southwest (including such languages as
m Navajo, Apache, and Hupa). These languages are
characterized by a particularly complex verb mor-
m phology.
m
mouton de gruyter
m Berlin · New York
ra George Horn
im] Lexical-Functional Grammar
1983.14,8 x 22,8 cm. IX, 394 pages. Cloth
[m
ISBN 90279 3169 0
(Trends in Linguistics. Studies and Monographs 21)
\m The analysis outlined in this monograph is formulated in
ι the context of the major developments in linguistic theo-
mi ry stemming from the proposal of the so-called Lexicalist
Hypothesis by Chomsky. The most significant product
of linguistic research during this period has been the
development and expansion of the lexical component
and consequent reorganization and reformulation of the
m rules of the transformational-generative model, in which
this component has been assigned many of the tasks for-
ΐγγΐ merly associated with the syntactic component.
More recently, various counterproposals to Chomsky's
analysis have been suggested. Perhaps the most impor-
tant of these was developed by Bresnan, the key feature
of which is the virtually complete reduction of the syn-
m tactic component.
This work is an attempt to extend and reformulate cer-
IYIJ tain of Bresnan's and Chomsky's ideas, combining the
basic organization of Chomsky's model, in which lexical
[m and non-lexical operations are clearly distinguished,
with a non-syntactic account of bound anaphora, con-
trol, and NP movement phenomena. The proposed
ΓΤΊ model provides a framework in which universal general-
izations can be captured, and language variation can be
|γγι accounted for without the complex machinery of
Chomsky's current analysis, at the same time maintain-
ing distinctions that are obscured in Bresnan's purely
mi lexical analysis.
\m\