Word-level 'prosodies' in Chadic languages in synchronic and
diachronic perspective
H. Ekkehard Wolff , University of Leipzig
The Chadic languages constitute a family of about 140 languages in the vicinity of Lake Chad in West-/Central Africa, as such they are part of the Afroasiatic language phylum.
In particular certain Central Chadic languages spoken in North-Eastern Nigeria and Northern Cameroon display heavy constraints on the co-occurrence of vowels within the word without, however, manifesting "vowel harmony" in the usual sense for African languages (e.g. based on the feature ATR). Furthermore, these vowels appear to escape any attempt to establish regular sound correspondences between these languages when applying the comparative method in order to reconstruct the common proto-forms of what obviously appear to be otherwise cognate words.
In the 1970s and 1980s, linguists such as C. Hoffmann and R. Mohrlang for Higi, J. Hoskison for Gude, E. Jarvis & J. Swackhamer for Podoko, D. Barreteau for Higi and adjacent languages, and others have applied to notion of "prosody" in synchronic descriptions (less so in historical reconstructions, with the exception of E. Wolff for the reconstruction of vowels in the Mandara Group languages), in order to account for the at times highly irritating distribution of features of labialisation, palatalisation, (pre-) nasalisation, and even voicing (and possibly glottalisation) and combinations thereof within the word..
The paper will look at some closely related Central Chadic languages, which appear to recognise different "prosodic" domains for, for instance, labialisation and palatalisation, i.e. individual segments, the syllable, or the word. The synchronic distribution of these prosodies will be explained, partly at least, in terms of diachronic processes in which affixes that were added to the lexical root or base play a role. These affixes, however, may be no longer productive and may even have disappeared from the segmental tier of the word in its present-day shape. In a nutshell: "vanished" affixes may still define today what is a "word" in these languages by constraining the effects of labialisation and/or palatalisation prosodies.
Phonetics and word definition in Ahtna Athabascan
Siri Tuttle, University of Alaska Fairbanks
Athabascan verbs are predominantly prefixing, often displaying six or more prefixes before the stem, and one or more suffixes. Recent representations of Athabascan verbs (Rice 2000, among others) claim that their wordhood is prosodic, and that their structure is best understood via a syntactic analysis. This analysis depends on the prosodic unity of the Athabascan verb; it predicts that unambiguous audible diagnostics can be discovered which define word edges.
This study is the third in a series of studies of words in Ahtna, an Athabascan language spoken in Alaska. Tuttle 2002 looked at consonant fortition effects in different positions in complex Ahtna verbs; Tuttle 2003 is a study of the phonetics of stressed Ahtna vowels, based on a recorded text. Both studies found that the beginnings of Ahtna verbs are non-prominent. Final or near-final stems, which are prominent, cannot be depended on for word-beginnings. The present study considers further textual material, returning to the measurement of consonants to look for word-initial or word-final effects. The text is from the same speaker as that in the vowel study: Katie John's narration of "Lazeni 'Iinn Nataelhde Ghadghaande," When Russians were Killed at Roasted Salmon Place (Batzulnetas). The text was recorded and transcribed by James Kari, and appears in published form in Kari (1986). Preliminary results support earlier findings: consonants were not strengthened in initial position unless they were also stem-initial. This suggests that delimiting an Ahtna verb on the basis of prosody alone may be a difficult task.
An alternative suggestion is to consider morphological and lexical facts along with prosody: the identities and prosodic properties of prefixes and suffixes can be considered to work together to delimit verbs. There are several strategies a learner might employ to do this.
- Stems can be spotted based on fortition and stress.
- Since the stem is not always final, the stem-central processing approach requires that the hearer be able to tell prefixes from suffixes—that is, learn lexical items. There are 27 positions in the Ahtna prefix complex (Kari 1990), and all manners and places of articulation are represented in the catalogue. There are only three suffix positions, however. The ten possible inhabitants of these positions can be, and I suggest are, easily learned, and form the next part of the key to finding the beginning of the next word. While there are more prefix positions, the inhabitants of each position form small sets, providing good clues to word position. This strategy should be considered lexical and morphological.
-The third part of the delimitation lies in the syllabification of Ahtna prefixes. Unlike suffixes, prefixes may be vowelless, and their concatenation results in the creation of remarkable consonant clusters unrelieved by epenthesis. (Nukahwkngelh'iilh "I'll look for you" (Kari 1990, p. 86) Such a medial cluster triggers the expectation that a stem is coming up in the next syllable or two; no stems have such a form, and no suffixes. I conclude that prosodic facts can indeed help define the edges of an Athabascan verb; however, they do so in collaboration with morphology and lexicon.
References
Kari, James (1986) Tatl'ahwt'aenn Nenn', The Headwaters People's Country. Fairbanks: Alaska Native
Language Center.
Kari, James (1990) Ahtna Athabaskan Dictionary. Fairbanks: Alaska Native Language Center.
Rice, Keren (2000) Morpheme Order and Semantic Scope. Berlin: Mouton.
Tuttle, Siri (2002) Prosody of incorporated structures in Ahtna Athabaskan. Proceedings of the
Athabascan Languages Conference, Fairbanks, June 2002. Gary Holton, ed. Fairbanks: Alaska Native Language Center.
Tuttle, Siri (2003) Realizations of stress and intonation in an Ahtna text. Paper presented at Society for
the Study of the Indigenous Languages of the Americas annual meeting, Atlanta.
The Fate of Clitics at the Phonology-Morphology Interface
Baris Kabak and Rene Schiering, University of Constance
Host-clitic combinations continue to pose problems for any theory or typology of the "word" (see Dixon & Aikhenvald (eds.) 2002). On the one hand, host-clitic combinations form a single unit from the point of view of phonology. From a syntactic perspective, on the other hand, they are considered to be two grammatical words since they take two distributional slots within a phrase. For instance, the complementizer dass in Ruhrdeutsch forms one phonological unit with the clitic subject pronoun =e and the clitic object pronoun =t (1). While the complementizer appears in the COMP slot, the pronouns belong to syntactic slots in the "middle field" position.
(1) ich glaub nich, dass=e=t in vier wochen schaffs.
I think not that-2sg-3sg(neut.) in four weeks accomplish
"I don't think that you can accomplish it within four weeks." (Schiering 2002: 23-24)
Bavarian provides an interesting case where we observe that the Comp-pro combinations are analyzed as a single grammatical word while the strong form of the pronoun occupies the "middle position" in the same sentence (2). Note that the cliticized pronoun has phonological resemblance to the corresponding full form.
(2) da:s=må ç:Bå mIå I:BårçI dåbaI hannd.
that=we but we allover there are
"that we are in the middle of everything" (Altmann 1984: 205)
Closer inspection of such host-clitic combinations in German dialects reveals that phonology alone cannot always be held accountable for the appearance of person markers on complimentizers. For instance, the person marker -sd on the complimentizer das in (3) has no phonological similarity to the full pronoun [dU:]. It is rather identical to the verbal person suffix on the verb bis.
(3) das-sd ç:Bå dU: I:BårçI dåbaI bIs-sd.
that=you but you allover there are
"that you are in the middle of everything" (Altmann 1984: 205)
In this talk, we will suggest a diachronic scenario that accounts for the development of such inflectional paradigms. Surveying data from various Middle and Upper German dialects, we will show that the cliticization of articles and pronouns to function words starts the grammaticalization cline. As a next step, fully affixal exponents of inflectional categories emerge through analogical extension. As the developmental path from clitics to affixes involves gradual rather than discrete steps, we can develop several diagnostics to capture the distinction between the two. In the context of German dialects, these involve various phonological processes such as stress assignment, glottal stop deletion, flapping, and consonant epenthesis that seem to occur within host-clitic combinations. Taking a cross-linguistic stand, the final part of this talk will comment on how phonology and morphology interact towards the development of full inflectional paradigms on typically "uninflectable" grammatical categories such as complementizers and adpositions through cliticization and discuss the consequences of this interplay for morphological typology.
References
Altmann, Hans (1984). "Das System der enklitischen Personalpronomina in einer mittelbairischen Mundart." Zeitschrift für Dialektologie und Linguistik 51: 191-211.
Dixon, R. M. W. & Alexandra Y. Aikhenvald (eds.) (2002). Word. A cross-linguistic typology. Cambridge: Cambridge University Press.
Schiering, René (2002). Klitisierung von Pronomina und Artikelformen. Eine empirische Untersuchung am Beispiel des Ruhrdeutschen. (Arbeitspapier, 44 (N.F.)). Köln: Institut für Sprachwissenschaft, Universität zu Köln.
Prosodic words, affixes and the typology of reduplication
Sharon Inkelas (UC Berkeley) and Cheryl Zoll (MIT)
Kiparsky (1986) develops a view of reduplication in which a sharp distinction is drawn between compounding reduplication (whose output is two words) and affixing reduplication (whose output is a single prosodic word). The essential division, picked up on also in Generalized Template Theory (McCarthy & Prince 1994a; b; Urbanczyk 1996), involves reduplicant size: short reduplicants are considered affixes, while long (foot-sized or above) are considered (minimal or maximal) words. However, recent research addressing GTT, e.g. Downing in press, has shown that this typology does not match cross-linguistic data well. It is neither always the case that short reduplicants have the phonology of affixes, nor that long reduplicants have the phonology of words. In this paper we propose that there is a distinction to be made between different kinds of reduplication, but that the essential difference is not primarily one of size, or affix vs. prosodic word, but rather of the way in which duplication takes place.
The proposed approach offers two methods of duplication: morphological reduplication and phonological copying. Morphological reduplication yields a morphologically bipartite stem, phonological copying a single stem. The proposed typology is more in line with the morphology and phonology of reduplication than the one offered by Kiparsky and by GTT. Four criteria distinguish between phonological copying and morphological reduplication:
(1) Phonological copying Morphological reduplication
Serves a phonological purpose Serves a morphological purpose
Is phonologically local Is not necessarily phonologically local
Involves single phonological segments Involves morphological constituents
Is driven by a phonological identity Is not driven by a phonological identity
imperative imperative
The clearest examples of morphological reduplication come from apparently reduplicated words where the two halves of the reduplicant are different suppletive allomorphs of a stem; this is found, for example, in a number of the languages of Central and Southern Vanuatu. For example, the Sye root for "fall" has two suppletive allomorphs, amol (occurring in one arbitrary set of morphological contexts) and omol (occurring in the complement set of contexts). In reduplication it is possible to get one copy of each allomorph, e.g. amol-omol (e.g. Crowley 1998). This clearly involves double insertion of the lexeme "fall", rather than insertion of one allomorph and phonological copying thereof. The clearest examples of phonological duplication are essentially epenthetic in nature, involving the copying of a nearby consonant or vowel to supply a syllable onset or nucleus (the sort of constituent often inserted epenthetically). For example, gerundives in Yoruba are formed by prefixing a hightoned /i/ (e.g. Pulleyblank 1988). The syllable this prefix creates requires an onset, and the requirement is met by phonologically duplicating the closest (stem-initial) consonant. Thus the gerund formed from /wçò/ "enter" is
[w-iô-wçô], with phonological duplication of [w].
Our broad cross-linguistic survey of reduplication, has found no phonological limitations on the sizes of elements involved in morphological reduplication. Although the Sye example cited above involved total root reduplication, similar examples involving partial CV reduplication occur in Ambrym (another Oceanic language of Vanuatu; Parker 1968) and Kawaiisu (Uto-Aztecan; Zigmond, Booth & Munro 1990), among other languages. By contrast, there are phonological limitations on the amount of material involved in phonological duplication; just like phonological epenthesis generally, epenthesis of copy elements is limited to single segments. The proposed typology has important implications for the role of phonological identity in reduplication. In Morphological Doubling Theory, phonological identity is not the driving force behind morphological reduplication; in fact, it plays no greater role in morphological reduplication than in any other kind of morphological construction. Removing phonological identity as a possible requirement of morphological constructions in general is consistent with the absence from ordinary morphology of identity constraints such as rhyme, alliteration and assonance.
References
Crowley, Terry. 1998. An erromangan (Sye) grammar: Oceanic linguistics special publication; no. 27. Honolulu: University of Hawaii Press.
Downing, Laura. in press. The emergence of the marked: tone in some African reduplicative systems. In Bernhard Hurch (ed.) Studies in reduplication. Berlin: Mouton.
Kiparsky, Paul. 1986. The phonology of reduplication. Unpublished ms., Stanford University.
McCarthy, John & Alan Prince. 1994a. An overview of Prosodic Morphology. Part I: Templatic form in reduplication. Paper presented at Workshop on Prosodic Morphology, Utrecht University.
McCarthy, John & Alan Prince. 1994b. [An] overview of Prosodic Morphology. Part II: Template satisfaction. Paper presented at Workshop on Prosodic Morphology, Utrecht University.
Parker, G. J. 1968. Southeast Ambrym verb inflection and morphophonemics. Pacific Linguistics Series A 15:27-40.
Pulleyblank, Douglas. 1988. Vocalic underspecification in Yoruba. Linguistic Inquiry 19:233-70.
Urbanczyk, Suzanne. 1996. Patterns of reduplication in Lushootseed. Amherst: GLSA.
Zigmond, Maurice L., Curtis G. Booth & Pamela Munro. 1990. Kawaiisu: A grammar and Dictionary with Texts. Berkeley: University of California Press.
On the Notion of the Word in Isolating Languages:
The Case of Riau Indonesian
David Gil (MPI for Evolutionary Anthropology, Leipzig)
Isolating languages pose a particular problem for the notion of word. In languages with a substantial amount of morphology it is generally possible to identify a set of criterial features, some universal, others language-specific, distinguishing word-external syntactic structure resulting from the concatenation of words and phrases, from word-internal morphological structure resulting from processes such as affixation, compounding and the like. Such criteria form the basis for the existence, within linguistic theory, of autonomous disciplines of syntax and morphology. However, in isolating languages, characterized by a paucity of morphological structure, there may not be enough morphology to support a robust and systematic distinction between morphological and syntactic structure. Accordingly, in isolating languages, there may be relatively little evidence for the existence of words as a viable unit of linguistic structure, as distinct from morphemes.
This paper examines the notion of word in one extreme exemplar of the isolating language type: Riau Indonesian. In general, it is argued that, compared to other, non-isolating languages, the word plays a much smaller role in the grammar of Riau Indonesian. Nevertheless, it is still possible to support a distinction between morphological an syntactic structure in Riau Indonesian.
This paper argues for the following morphological structure underlying words in Riau Indonesian:
Evidence for the above structure is derived from a number of different sources of evidence, as summarized in the following table:
|
CORE FOOT |
INNER WORD
(terminal)
|
INNER WORD
(non-terminal)
|
OUTER WORD |
| focus intonation |
X
|
|
|
|
| no reduction |
X
|
|
|
|
| epenthesis |
X
|
|
|
|
| loanword expansion |
X
|
|
|
|
obligatory si-
|
X
|
|
|
|
| N- realized as nge- |
X
|
|
|
|
| Warasa ludling |
X
|
|
|
X
|
| final k realized as ? |
|
X
|
|
|
| Sabaha ludling |
|
X
|
|
|
| Bahasisa ludling |
|
X
|
|
|
| Pantun rhythm |
|
X
|
|
|
| reduplication |
|
|
X
|
|
| spelling |
|
X
|
X
|
X
|
| Evidence for Word Structure Categories |
In total, the above phenomena provide convincing evidence for the posited underlying word structure, but hardly overwhelming evidence: the extent to which the grammar of Riau Indonesian makes reference to the various categories constituting word structure is still much less than in most other languages.
Romanian between Balkan and Romance: the case of the tense auxiliaries
Paola Monachesi (Utrecht University)
Romanian tense auxiliaries have not received much attention in the literature, the only exceptions being Dobrovie-Sorin (1994), Legendre (2000), Monachesi (2000) and Rivero (1994). However, an investigation of their status from a typological perspective can lead towards a better understanding of the interaction of the various components of the grammar.
I show that Romanian tense auxiliaries pattern like the Balkan auxiliaries (and unlike the Romance ones) with respect to adjacency requirements, but they behave like the Romance auxiliaries (and unlike the Balkan ones) with respect to the distribution of pronominal clitics.
Balkan auxiliaries are usually considered clitics in the literature (Franks and King 2000) while Romance auxiliaries have word status. It becomes of crucial importance to assess the status of Romanian tense auxiliaries, given their hybrid properties. I claim that they are morphosyntactic words, which are phonologically dependent on an adjacent word. They are simple clitics according to the terminology of Zwicky (1977).
Romanian tense auxiliaries can be found in the present perfect, the conditional and the future. They behave similarly to auxiliaries in Balkan languages, such as Bulgarian and Macedonian (Legendre 2000, Franks and King 2000) with respect to adjacency requirements. Nothing can intervene between the auxiliary and the lexical verb: subjects, adverbs or quantifiers are not allowed in that position. Romanian tense auxiliaries differ from those of other Romance languages since in French and Italian, adverbs can occur between the lexical verb and the auxiliary, similarly for quantifiers (at least in French) and for subjects.
Furthermore, Romanian tense auxiliaries pattern like the Balkan (and Slavic) ones since they represent a reduced form. The full forms of the verb a avea "to have" become monosyllabic when they occur as tense auxiliary (i.e. am vs. a.vem) :
(1) am, ai, a, am, at¶i, au (tense auxiliary)
(2) am, ai, a.re. a.vem, a.vet¶¶i, au (a avea, "to have")
In the Romance languages, the tense auxiliary is equivalent to the full form of the verb "to have"/"to be" and no reduced form occurs.
Romanian tense auxiliaries pattern, however, like the equivalent Romance ones with respect to the distribution of the pronominal clitics: object clitics precede the auxiliaries (3a). This is not the case in the Balkan (and Slavic) languages where pronominal clitics generally follow the tense auxiliaries (3b) (Franks and King 2000):
(3) a. le-am va((zut. (R)
CL.ACC AUX seen
"I have seen them."
b. az sum mu go dal. (B)
I AUX CL.DAT Cl.ACC give
"I have given it to him."
I suggest that Romanian tense auxiliaries are morphosyntactic words which combine with the lexical verb in syntax forming a verbal complex. I show, however, that at the phonological level they form a prosodic word either with the verb or with the preceding word given that they are monosyllabic and dependent elements. It is thus possible to account for their adjacency requirements while an analysis in terms of argument composition (Monachesi 1999, 2000) can deal with the distribution of the pronominal clitics in the presence of tense auxiliaries both in Romanian and in other Romance languages. I provide evidence that the tense auxiliaries do not form a morphological cluster with the pronominal clitics, but they act as host for them.
Constraints on Stem Length
Bhavani Saravanan (Stony Brook University)
This paper is devoted to studying some constraints that determine the exact length that simplex words have to be in Tamil: the Minimality, Maximality and Regularity constraints. I argue that given the Minimality and Maximality constraints, words can really only be one foot long. The Regularity Criterion is specific to verbs, and ensures that all regular verbs are disyllabic; the consequence of disobeying this length requirement is irregularity.
In Tamil the prosodic word has to be at least bimoraic (minimality requirement). The two moras may be contained in one syllable as in (1) or in two syllables, as in (2).
(1) [k´j] "hand" (2) [pU.li] "tiger"
This minimality requirement holds of compounding: each member of a compound is required to be bimoraic. And it holds of both strata of the lexicon: native as well as borrowings. We do not see any violators that disobey minimality, in native vocabulary or in loanwords. Subminimal forms simply do not surface.
The Maximality Requirement restricts monomorphemic words in Tamil to containing no more than three moras. Then, irrespective of the exact composition of the stem, it contains exactly one bimoraic foot, as this is the maximum that can be constructed out of three moras. To accommodate the third mora, I adopt the notion of a loose minimal word (McCarthy and Prince 1996, Ito and Mester 2003). A loose minimal word contains the minimal word and possibly more material, but not enough to form another foot. The function of the Maximality constraint is then to act as a restriction on the number of feet that may be accommodated in one word. In Tamil, the minimum is the same as the maximum: one foot. We need only state that stems should contain one foot, without imposing separate minimality and maximality constraints. The native stems are subject to the constraint on stem length, but borrowings violate it.
For verbs, there is a further requirement that they should be disyllabic. A monosyllabic verb that contains a bimoraic foot would not satisfy this criterion, though it would most certainly meet the requirements for stem length. But we do see monosyllabic verbs (monosyllabic in the imperative, the citation form of the verb). Therefore this is certainly not an inviolable constraint. However, all of the monosyllabic verbs are irregular, a seeming penalty for disobeying the Regularity Criterion. Below, an irregular verb [vej] is compared to a regular verb [v´r´j].
| IMP |
FUTURE |
PTCPL |
PAST |
INF |
PRESENT |
vej
scold |
ve-jj--v- |
ve--j |
ve--j(- |
ve-jj-´ |
ve-jj-´-r- |
v´r´j
draw |
v´r´j--v |
v´r´--j(-
|
v´r´--j(- |
v´r´j-´ |
v´r´j-´-r- |
The Regularity Criterion gives the minimum for the verbs. Verb stems are also, however, subject to the same maximum that nouns are: no more than three moras. And these three moras must come from two or three syllables.
Phonological requirements on regularity are not restricted to Tamil. Arabic and Hebrew have disyllabic requirements for verbs. In Arabic, not only do the verbs have to be disyllabic and triconsonantal, but there are also constraints on what consonants make up the triliteral root. Obstruents and liquids are acceptable, but glides are not. If a glide is one of the three consonants that make up the root, then the verb stem is irregular.
References
Ito, Junko & Armin Mester. (1992) / 2003. Weak Layering and Foot Binarity. Festschrift for Dr. Haraguchi. Downloaded from website.
McCarthy, John & Alan Prince. 1986 / 1996. Prosodic Morphology. ROA.
Words Without Vowels in Upper German: Clitics Wanting to Be Affixes
Astrid Kraehenmann and Frans Plank (University of Constance)
Grammatical words commonly exist in two or more forms, one full and the other(s) reduced, with the latter in some sense derivable from the former. Our interest here is in what appears to be straightforward cliticization in Upper German, where several words - including the definite articles die (PL.NOM/ACC and SG.NOM/ACC.FEM) and das (SG.NOM/ACC.NEUT), the personal pronouns es (3SG.NOM/ACC.NEUT), sie (3SG.NOM/ ACC.FEM, 3.PL.NOM/ACC, 2.SG/PL.HON.NOM/ACC), du (2.SG.NOM), and zu, functioning as an infinitive complementizer, adverb, and preposition (all words here cited in their Standard German forms and spelling) — are notorious for losing their vowels and tightly fusing with their hosts. What is particularly worth noting is that vowel loss is not merely optional, as is not uncommon in fast speech in general, but is obligatory in certain well-defined morpho-syntactic contexts even in the most carefully enunciated speech. For example, in Eastern Swiss German Alemannic, die is obligatorily without vowel when it precedes a noun (1), while it has one when it precedes an adjective or an adverb (2).
Alemannic and Bavarian, the two dialect branches that make up contemporary Upper German, would seem to show essentially the same reductions and assimilations, with a few minor differences not crucially affecting the nature of these alternations. A closer look, however, does bring out differences, especially between Swiss German relative to Bavarian. They are subtle, but they point to the reduced forms having a different grammatical status: in Swiss German the alternations are not between full forms and clitics, but the vowel-less forms are in many essential respects like affixes, however intricate their morpho-syntactic distribution (which is instructively compared to that of definiteness markers in Scandinavian), while in Bavarian they are still clitics, however phonologically reduced.
The evidence for assuming a different status includes differences in syllabification and attendant phonetic realizations: in Swiss German, vowel-less forms are fully incorporated into the syllable structure, while in Bavarian (1') they are only adjoined, as reflected in (more or less obligatory) additional phonetic material (stop release; glottal stop insertion) between clitic and host.
Bavarian permits initial clusters which violate phonotactic constraints of lexical words as well as syllables in general (1'c), while Swiss German does not ((1c) is the only possible realization). Third, as is expected of affixes, the vowel-less forms in Swiss German are very selective as to the class of stems they combine with (as seen in (1)/(2), d is limited to nouns), while d in Bavarian combines with a wider range of hosts.
Although there is a lot of phonetic and phonotactic evidence in favour of clitics turning into affixes in certain Upper German dialects, evidence from phrasal syntax speaks to the contrary. If d were an affix beyond doubt, the expectation would be that it would be anchored as a definiteness marker to the noun whether or not the noun was modified, making structures such as the ones in (2') a logical possibility. However, no such development is evident (nor, it seems, imminent): vocalic and vowel-less forms remain dissociated as shown in (1) and (2).
(1).
(1) a. d'Tour /tt-ttu:{/ [ttu:{] 'the tour'
b. d'Uhr /tt-u:{/ [ttu:{] 'the watch'
c. d'Mueter /tt-mu´tt´{/ [/mu´tt´{] 'the mother'
(2) a. di toll Tour /ti ttoll ttu:{/ [ti ttol ttu:{] 'the great tour'
b. di toll Uhr /ti ttoll u:{/ [ti ttol u:{] 'the great watch'
c. di toll Mueter /ti ttoll mu´tt´{/ [ti ttol mu´tt´{] 'the great mother'
(1') a. d'Tour /t-tu:{/ [t tHu:{]/[tHu:{] 'the tour'
b. d'Uhr /t-u:{/ *[tHu:{]/[t/u:{] 'the watch'
c. d'Mutter /t-mu´tå/ [tmu´.tå]/[/mu´.tå] 'the mother'
(2') a. *toll d'Tour */ttoll tt-ttu:{/ *[ttol ttu:{] 'the great tour'
b. *toll d'Uhr */ttoll tt-u:{/ *[ttol ttu:{] 'the great watch'
c. *toll d'Mueter */ttoll tt-mu´tt´{/ *[ttol /mu´tt´{] 'the great mother'
Word-internal morphological structure: the evidence from prosody
Renate Raffelsiefen (Free University of Berlin)
It is widely accepted that prosodic structure is constructed based on morphosyntactic structure and that word-internal morphological constituents can be mapped into separate pwords (cf. the distinction between cohering and non-cohering affixes in Dixon (1977)). A question which is largely neglected is on what basis word-internal morphological boundaries are posited, what labels are associated with those boundaries, and how learners recognize those structures. Several approaches can be distinguished:
(1) "stem-based, intrinsic": the assignment of internal morphological structure is determined by intrinsic stem properties: the decisive question is whether or not the "stem" can stand alone (i.e. CATX), or not (i.e. CATY) (e.g. [[hope]CATX[ful]SUF] versus [[grate]CATY[ful]Suffix], because *grate) (cf. Bloomfield 1933, Newman 1948)
(2) "stem-based, relational": the assignment of internal morphological structure is determined by relational stem-properties: the decisive question is whether or not the "stem" can combine with inflectional markers (i.e. CATX), or not (i.e. CATY), (e.g. German [[red]CATx[est]Suffix] versus [[nied]CATy[lich]Suffix], because nied never combines with inflectional markers) (cf. Matthews 1974)
(3) "affix-based, relational": the assignment of internal morphological structure is determined by relational affix-properties: the decisive question is whether or not affixes can be recognized: if so, whatever remains after the affixes have been removed is labeled CATX. If no affixes are recognized the word is treated as a simplex . Examples are German [[red]CATx[lich]Suff], [[nied]CATx[lich]Suff], but [Wacholder] because -der is no longer recognized as a suffix) (cf. Lyons 1970).
The most influential view in formal morphophonology is (1), which for languages like English is equivalent to (2) (cf. Benua 1997, Giegerich 1999). However, the prosodic evidence clearly supports the approach in (3). That is, it can be shown that boundary signals indicative of non-cohesion are stable for as long as an affix is recognized, whether the stem corresponds to an independent word (cf. column (ia)) or not (cf. column (ib). The relevant boundary signals include moraic structure (cf. row (ia,b)) and subphonemic boundary signals (i.e. glottalization of [k] in row (ic)).
(i) A B C
a. [[Teil]STEM[chen]SUFFIX]WORD [[Veil]STEM[chen]SUFFIX]WORD => (Oxeil)wchen
b. [[hate]STEM[ful]SUFFIX]WORD [[grate]STEM[ful]SUFFIX]WORD => (Oxate)wful
c. [[neck]STEM[less]SUFFIX]WORD [[reck]STEM[less]SUFFIX]WORD => (Oxeck)wless
Approach (3) raises the question under what conditions affixes are recognized. It appears that for as long as an affix occurs in some word-based constructions (cf. the examples in (iA), that affix is recognized in all relevant formations, whether or not the stem recurs (cf (ib). Consequently the morphological structures of all words in (i) are identical according to approach (3) with the result that these words are also mapped into identical pword structures. Diacritic marking to describe non-cohesion of the suffixes in (ib) (cf. Hall 2001) is inadequate in that it fails to capture the systematic nature of the relevant boundary effects.
The "word" in sign language
Annette Hohenberger (MPI for Psychological Research, Munich)
Due to the different modality, sign languages offer an ideal opportunity to test the psychological reality of the "word" as a basic unit of representation in the lexicon, the grammar (phonology, morphology, syntax) and in processing. Sign languages are processed in the visuo-gestural modality; spoken languages are processed in the aural-oral modality.
According to Brentari (2002), sign languages are typologically unique with respect to word shape in that they are polysynthetic but monosyllabic. Sign languages are characterized by "vertical" processing, i.e., linguistic information is displayed simultaneously on various articulators (manual and non-manual) which are represented on various tiers layered upon each other in a vertical fashion. Visual-gestural processing has a high spatial but low temporal resolution. Production time on the word level is long (1:2 as compared to spoken languages). As a consequence, linguistic information is preferably organized on the vertical axis. Spoken languages are characterized by "horizontal" processing, i.e., linguistic information is organized sequentially, on the horizontal axis. Aural-oral processing has a high temporal but low spatial resolution. A trade-off between information load and processing time is achieved insofar sign languages compensate their long production time on the word level by packaging linguistic information into few big chunks with high information load as opposed to spoken languages which package linguistic information into many small chunks with low information load (Hohenberger, Happ & Leuninger, 2002). On the propositional level, production time is the same for both languages.
In a research project on sign and spoken language production, we elicited slips of the hand in German sign language (Deutsche Gebärdensprache, DGS) and slips of the tongue in spoken German (Leuninger, Hohenberger, Waleschkowski, Menges & Happ, in press). We compared them with respect to affected unit (phonological feature/segment, morpheme, word, phrase) and slip category (anticipation, perseveration, exchange, substitution, blend, fusion). As for affected unit, words were affected to a higher extent in DGS than in spoken German (50% vs. 35%), morphemes to a lesser extent (6% vs. 18%), and phrasal slips were almost completely absent in DGS but a respectable unit in spoken German (1% vs. 16%). The low percentage of morphological slips in DGS can be explained by the modality difference. As in DGS morphemes are displayed simultaneously rather than being concatenated, they are harder to detach and will not engage in a slip. In spoken German, however, concatenative morphemes can be detached quite easily. Phrasal slips occur rarely in DGS because what would typically be expressed as a phrase in spoken German, is more likely to be expressed by a single word in DGS. As for slip category, fusions occurred much more frequently in DGS as compared to spoken German (8% vs. 0%) which again reflects the simultaneous/fusional character of DGS.
Words, thus, carry more information in a vertically-processing type of language, as compared to a horizontally-processing type of language. In both languages, however, the word is a major processing unit. The dynamic interplay between processing time and information load leads to typologically different languages which, however, are equally well-adapted to the overall goal of conveying language in real-time.
References
Brentari, Diane (2002): Modality differences in sign language phonology and morphophonemics. In: Richard P. Meier, Kearsy Cormier, and David Quinto-Pozos (Eds): Modality and structure in signed and spoken languages. Cambridge, MA: Cambridge University Press, 35-64.
Hohenberger, Annette, Happ, Daniela, and Leuninger, Helen (2002): Modality-dependent aspects of sign language production: Evidence from slips of the hands and their repairs in German Sign Language. In: Richard P. Meier, Kearsy Cormier, and David Quinto-Pozos (Eds.), 112-142.
Leuninger, Helen, Hohenberger, Annette, Waleschkowski, Eva, Menges, Elke & Happ, Daniela (in press): The impact of modality on language production: Evidence from slips of the tongue and hand. In: T. Pechmann & C. Habel (Eds.) Multidisciplinary approaches to language production. Berlin, N.Y.: Mouton de Gruyter.
Phonological Restriction on Morpheme in Shaping Reduplicated Words
Guanjun Feng (USC)
It has been observed in the literature that size restrictions may be imposed on phonological elements by phonological and morphological constraints. (Walker 2002, Ono 2002). This paper expands on these observations by discussing a reduplication pattern of adjectives in Mandarin (gaoxing “happy” ->gaogaoxingxing “very happy”) which is not phonologically common in other languages. I propose that the pattern is shaped by a morpho-phonological characteristic of the Chinese languages, namely the size of morpheme is always the size of a syllable. Not only does this syllable-size morpheme restriction shape reduplication patterns, it also motivates reduplication itself, as found in the diminutive affixation of the Anxiang Chinese dialect (described below)
Chinese is well known for the isolating nature of its morphology, in the sense that every syllable represents one morpheme and each morpheme is represented by one syllable. This one-to-one relationship between syllable and morpheme and is accomplished via four high-ranked alignment constraints: ALIGN(syllable, morpheme, L/R) and ALIGN(morpheme, syllable, L/R). (McCarthy & Prince 1993). In the reduplication of Mandarin Chinese adjectives, an adjective composed of s1 and s2 is reduplicated as s1s1s2s2
Base Literal Translation Gloss Reduplication Gloss
ganjing "dry + clean" "clean" ganganjingjing "clean" (intensified)
mingbai "bright + white" "clear" mingmingbaibai "clear" (intensified)
The reduplication pattern shows that one RED morpheme shows up as two in the output as two separate syllables. The alignment constraints are perfectly obeyed with the result that one morpheme in the input corresponds to two morphemes in the output. This violates a morpheme integrity constraint, which is a counterpart of the integrity constraint restricting phonological elements, forbidding one morpheme having multiple correspondents in the output. In the meanwhile, s1s2 is not reduplicated as s1s1s2 which would have obeyed both the syllable-size morpheme restriction and kept morpheme integrity. This is explained by the composition of the word s1s2 in which both s1 and s2 work as the morphological head of the word (Revithiadou 1999). Faithfulness to the morphological head determines that both syllables have to be reduplicated, with the violation of morpheme integrity.
In the diminutive affixation of the Anxiang dialect, the affixation of a retroflex consonant occurs together with reduplication.
Stem Diminutive Gloss
pHa pHa.pH´r claw
ke ke.k´r square
I claim that underlyingly there is no RED morpheme in the input and the reduplication is motivated by the syllable-size morpheme restriction. In order to satisfy the requirement of the diminutive morpheme to correspond to one morpheme, material is recruited from the root and reduplication occurs without an underlying RED morpheme. (Tableau 1)
Tableau 1:
/ke/, /r/
|
ALIGN[s] |
Dep-IO |
INTEGRITY |
| a.([ke].[r]) |
*!
|
|
|
| b.([ke].[t´r]) |
|
**! |
|
| +c.([ke]).([k´r]) |
|
|
** |
In conclusion, as an isolating language, Chinese has a strict syllable-size morpheme restriction. Together with faithfulness to morphological heads, it determines the shape and size of the reduplication pattern of adjectives in mandarin. This restriction is also apparent in motivating the reduplication witnessed in the diminutive affixation of the Anxiang dialect. Typological predictions of a ranked and violable syllable-size morpheme restriction are explored in connection with Vietnamese and Bantu languages.
Phonological cues for word domains
Anthi Revithiadou and Kalomoira Nikolou (Aegean University)
Presentation Cancelled
In every language certain phonological rules (p-rules) can be enforced (or blocked) in specific m(orphosyntactic) or p(honological) domains not necessarily isomorphic to each other. For instance, verb-clitic constructions in Greek are mapped onto one prosodic word (PrW) whose edges do not match with the edges of the grammatical word (GW): [{word}-cl]PrW. Moreover, optionality in the application of p-rules is reported to be typical of p-domains (Nespor & Vogel, 1986): s-voicing, for example, is obligatory within the PrW-domain and optional in larger ones (e.g. PPh, IP), izmíni "name", pés mu [pézmu]PrW "tell-IMP.2sg me", éxis maxépsi [[éxiz/s][maxépsi]]PPh "you have dazzled". This paper, however, draws the attention to p-rules which can be optional or not within certain m-structures and, furthermore, proposes that phonology can serve as a diagnostic for different word domains.
Greek has two distinct morphological types of compounding: {stem stem suff} and {stem {word}}, both shown in (1). (1a) forms a PrW and displays the default antepenultimate stress (APU), whereas (1b) is organized into a recursive PrW and preserves the stress of the second component. There is, however, an intermediate "hybrid" type, (1c), which has the morphological structure of (1b) and the prosodic structure of (1a).
(1) a {kuzin-o-máxer-o}GW [kuzinomáxero]PrW < kuzín-a maxér-I "kitchen knife"
b {lemon-o-{Dás-os}GW}GW [lemono[Dásos]PrW]PrW < lemón-I Dás-os "lemon forest"
c {lemon-o-{Dás-os}GW}GW [lemonóDasos]PrW < lemón-I Dás-os "lemon forest"
The -o- is an empty morpheme that connects the two members of the compound. When the following element is V-initial, o-deletion applies to resolve hiatus, e.g. aloƒourá [aloƒurá] "horse tail". An examination of a corpus of 450 compounds revealed that hiatus resolution is nearly obligatory in types (1b) and (1c) where the percentage of o-deletion is 70.56% and 72.53%, respectively. On the contrary, o-deletion is highly disfavored in stem-stem-suff compounds where /o/ is preserved in 68.5% of the data. This entails that o-deletion is sensitive to m-structure and not to p-boundaries since (1a) and (1c) have the same p-structure.
As mentioned above, clitic structures form one PrW with their host because s-voicing applies and stress shifts to avoid violation of the 3s-window requirement (2a). Nasal-stop assimilation (NSA), which applies obligatorily within the PrW, e.g. /paxin-tikós/ [paxindikós]PrW "fattening", fails to apply in (2b) . This implies that NSA is disfavored in an m-defined domain, namely the syntactic word, SW (Di Sciullo & Williams, 1987). It should be noted that the domain in question cannot be a prosodic one, i.e. the Clitic Group, because then the application of NSA between the determiner and the noun remains unaccounted for. In contrast, in (2c) e-deletion (e->Ø/c_tV) is more dynamically enforced in the SW rather than the GW. To conclude, certain p-rules by applyiing more (/less) forcefully in different m-domains provide cues for different m-structures, esp. GW- vs. SW-domains.
(2) a /tus krokóDilus mas/ [tus krokodilúzmas] "our crocodiles"
b /ton krokoDílon tus/ [toNgrokoDílon tus]/tonkrokoDílon dus "of their crocodiles"
c /párete/ [párete]/?[párte] "take-IMP.2pl"
/páre to/ [párto]/?[páre to] "take-IMP.2sg it"
This paper focuses on p-rules that refer to non-prosodically defined domains. In the spirit of Inkelas & Orgun, 1994, et al. we propose that different phonological mapping functions, called co-phonologies, are associated with different word domains. The theoretical significance of the paper relies on the fact that the proposed interface model accounts for optionality effects in the force of p-rules by means of a system of parallel phonologies which function as linking paths between different m-domains.
Lenition, coronal blocking and compounding in Irish
Antony D. Green (University of Potsdam)
Compounding is an important method of word formation in Irish. The best known kind of compound in Irish is head-final, with the first member invariably in the nominative singular, while only the second member is inflected. The initial consonant of the second member undergoes the process of lenition, which among other things, changes stops into continuants. Lenition is indicated in the orthography by an h following the initial consonant. Examples of head-final compounds are shown in (1); the location of lenition is indicated by bold face.
(1) Head-final compounds
a. ainm-fhocal "noun" (lit. "name-word")
ainm-fhocail genitive singular/nominative plural
b. breis-chéim "comparative degree" (lit. "increase-degree")
breis-chéime genitive singular
breis-chéimeanna nominative plural
It is fairly uncontroversial that the prosodic structure of this kind of compound involves a recursive prosodic word (w): w( w(ainm) w(fhocal)), etc. The novel claim of this paper is that Irish has a second kind of compound with the same prosodic structure, but a very different morphological structure. This second kind of compound is head-initial, with the second member in the genitive, and the first member inflected for number. Orthographically these are written as two words. lenition of the second member occurs only if the first member is either (i) a feminine singular noun or (ii) a plural noun ending in a palatalized consonant. Examples are shown in (2).
(2) Head-initial compounds
a. bliainFEM bhisigh "year of-increase" = "leap year"
blianta bisigh plural
b. iascMASC mara "fish of-sea" = "sea-fish"
éis[kj] mhara plural
The evidence that head-initial compounds in Irish also have the recursive structure w(w w) comes not from the fact of lenition, since lenition can be shown to apply across pword boundaries in some instances, but from the fact that lenition is blocked when two coronals (excluding r) come in contact. Ní Chiosáin (1991) refers to this phenomenon as coronal blocking and argues that heteromorphemic coronals share a single place node and are thus immune to lenition because of geminate inalterability.
Coronal blocking applies between a proclitic and a host, between a prefix and a root, and in both kinds of compound, as shown in (3).
(3) Coronal blocking of lenition (Lenition would normally be expected in each case)
a. proclitic + host an tairbh "the bull" (genitive singular)
b. prefix + root an-deas "very nice"
c. head-final compound ard-sagart "high priest"
d. head-initial compound tonnFEM tuile "tidal wave" (lit. "wave of-flood")
In this talk I argue that the domain of coronal blocking is the prosodic word, since it does not apply (as proved by the lenition of a coronal after another coronal) in environments where a pword boundary is clearly present. Such environments include definite adjuncts in the genitive singular ((4)a-b) and attributive adjectives after either a feminine nominative singular noun ((4)c) or a plural noun ending with a palatalized consonant ((4)d).
(4) Lenition even where two coronals meet
a. díon shiopa an bhúistéara "the roof of the butcher's shop"
b. foireann Dhoire "the Derry team"
c. cosFEM thinn "a sore leg"
d. eitleái[nj] dhearga "red airplanes"
Unlike the cases in (3), the forms in (4) involve two (or more) independent pwords with no recursivity. Thus coronal blocking fails to apply, and a coronal is lenited even after another coronal.
References
Ní Chiosáin, M. (1991). Topics in the Phonology of Irish. Ph.D. dissertation, Univ. of Massachusetts.
Observations on the coherence of monoconsonantal prefixes
Diana Passino (University of Padua)
Crosslinguistically the stem-suffix domain, although it may not be isomorphic to a morpho-logical domain, shows a greater phonological coherence than the prefix-stem domain. In many languages, as a matter of fact, prefixes are independent syllabification domains whereas syllabification is, in the majority of cases, blind to the stem-suffix boundary. The independence of the prefix, furthermore, is often implemented through the overapplication or the blocking of phonological processes at the prefix-stem boundary in order to protect the left edge of the stem.
My study concerns word formation with monoconsonantal prefixes1, and it may constitute a little contribution to the understanding of the issue of prefix vs. suffix asymmetries and to the debate on the relationship between grammatical and phonological words.
From my study it emerges that in languages that do not protect the prefix-stem boundary through the blocking of syllabification, the prosodic independence of the prefix is nevertheless enforced at the level of syllable heads.
My research reveals, in fact, that whereas the blocking of syllabification at the prefix-stem boundary is connected with unrestricted word formation with monoconsonantal prefixes (i.e. Polish z-analizowac» "to analyze + perf.", z-mówic» "to speak + perf."), languages where syllabification is blind to the prefix-stem boundary show restrictions (i.e. Italian s-contento "unhappy" but *s-uguale "unequal") or repair strategies (i.e. Abruzzian r-esce "go out again" [arhESS´] where [˙] is an epenthetic segment) in order to keep those kind of prefixes prosodically apart from the stem. This does not happen with non monoconsonantal prefixes because despite resyllabification with the base, they keep their syllable head apart from the base (Italian di.s-u.gua.le). Word formation looks as if it is bound to the satisfaction of a prosodic minimality condition (McCarthy & Prince 1986 et seq.) that prefixes have to respect2. Prefixes must be equal at least to a syllable heading mora, i.e. they cannot share the syllable heading mora with their base. The respect of the minimality condition, from a functional point of view, may be seen as a preservation of the prosodic independence of the prefix from its base3: if the prefix were not linked at least to a mora and shared it with its base, it should be prosodically licensed by the base and this would result in an increase in the coherence between the two.
The behaviour of monoconsonantal prefixes in word formation seems at first sight to lend support to the well documented non coherence of prefixes with respect to their base. It must be stressed however, that a minority of languages can also be found where monoconsonantal prefixes show greater coherence with their bases than other ordinary prefixes do (i.e. Russian s-jes-tj [sjjesjtj] "to eat with" where palatality assimilation applies to the prefix only if it is monoconsonantal). It looks as if the strategy of those languages is not that of enforcing the separation of the prefix but that of extending the stem domain in order to include those items that due to their subminimality cannot stand on their own.
References
Bagemihl, B. 1991. Syllable structure in Bella Coola. Linguistic Inquiry 22 589-646
McCarthy J. H. & A. S. Prince 2001. Prosodic Morphology. Constraint interaction and satisfaction.
Zec, D. 1988. Bulgarian ¥ epenthesis: a case for moraic structure. in J. Carter & R.-M. Dechaine (a cura di) Proceedings of the NELS 19, University of Massachusets Amherst: GLSA
1 The languages, including dialectal varieties, thoroughly investigated at the present state of the research are 10, all of the Indo European family, in particular of the Slavic, Germanic and Romance families. Some non Indo European languages have started to be investigated at the time of writing.
2 My study also provides evidence of the sensitiveness of prefixes to minimality through data from monosyllabic prefixes.
3 The mora has been shown to be able of prosodically license a segment (Zed 1988, Bagemihl 1991 among others)
Semantic and Syntactic Sources of Word Liminality
Daniel L. Everett (University of Manchester)
This talk discusses the widespread phenomenon of
Word Liminality, the sharing of properties of either (i) words and parts of words (clitics) or (ii) words and phrases (periphrastic morphology). I explore several sources of type (ii) liminality in natural languages, namely, linear correlations, communicative function, and the semantics and syntax of synecdoche (part-whole relations). To understand the role of these sources of liminality, we examine possession structures in two languages of Brazil: Portuguese (Romance) and Madi (Arawan). I argue that the best account of liminality is based on conjunctions of syntactic and semantic constraints rather than the strategem common in typological studies of offering definitions and prototypes for entities and constructions. Crucial to my account will be the concepts of
index and
concord features (Wechsler & Zlatic 2000) and the notion of the Referential Head, an adjustment to the RRG entity, NUC (Van Valin and La Polla 1997).
"
Case Suffixes", Postpositions and the Phonological Word in HungarianJochen Trommer (University of Osnabrück)
Descriptive tradition and orthographic convention suggest that Hungarian has two different types of functional items corresponding to adpositions: case affixes and postpositions. The main imperical evidence for this distinction (É. Kiss, 2002: 185) is that case suffixes (1-a) undergo vowel harmony with the preceding head noun while postpositions (1-b) do not:
(1) a. ház-ban kért-ben b. ház alatt kért alatt
house-in garden-in house under garden under
'in a house' 'in a garden' 'under a house' 'under a garden'
In this talk, I argue that case markers are part of the same phonological word as their head nouns, but syntactically independent units, in other words they are postpositions. This is consistent with the independently motivated observation made in Nespor and Vogel (1986) that Hungarian vowel harmony is not operative on the morphological word, but on the phonological word. However, it requires reconsideration of their deifnition of the phonological word as a stem plus all following suffixes.
Postpositions and case suffixes behave identically in almost all respects: they are right-adjacent to the heaad noun, are unstressed and show agreement with a pronominal head. Additionally, both can occur with a pro-dropped pronominal argument (e.g. benn-e, 'in it', alatt-a 'under it') without a preceding stem which shows that neither can be bound morphemes (hence suffixes). But there are also two important differences: apart from the vowel harmony facts, postpositional phrases unlike noun phrases with case markers show elipsis of head noun or the postposition. However, there is a further difference obviously not related to morphological affixhood: "case suffixes" are monosyllabic, propositions are bisyllabic. This suggests to revise Nespor and Vogel's definition of phonological word in Hungarian as follows, where "Stem" denotes the verb root + all derivational suffixes:
(2) Definition of phonological word in Hungarian
Nespor and Vogel: stem + following suffixes
revised definition: Stem + following monosyllabic functional elements
This immediately accounts for the fact that "case suffixes" undergo vowel harmony while postpositions do not: both are functional elements of the same type, but only case markers form a phonological word with the stem. It also gives rise to an explanation why noun phrases with case markers cannot undergo ellipsis which would leave the monosyllabic marker stranded not being able to form a minimal phonological word. The only configuration where a case marker can occur without a head is when it is followed by a agreement marker which results in a two-syllabic word. A potential problem with this account is that Hungarian seems to have bisyllabic inflectional affixes which undergo vowel harmony, namely the 2nd person plural agreement markers (e.g. -játok, -itek). However, as shown in Rebrus (2000) and Trommer (2003), there is independent evidence that these are composed of different (mono-syllabic) affixes. Hence, all functional elements undergoing vowel harmony are monosyllabic.
É. Kiss, K. (2002). The Syntax of Hungarian. Cambridge University Press.
Nespor, M. and Vowel, I. (1986). Prosodic Phonology. Foris Publicatons: Dodrecht.
Rebrus, P. (2000). Morfofonológical jelenségek a magyarban. In Kiefer, Fl, editor Strukturális magyar nyelvtan 1. Mondattan, pages 763-949. Budapest: Akadémiai Kiadó.
Trommer, J. (2003). Hungarian has no portmanteau agreement. In Siptár, P. and Pinon, C., editors, Approaches to Hungarian, volume 9. Akadémiai Kiadó. to appear.