Constraints in Word Formation

言語研究（Gengo Kenkyu）135: 000–000（2009） 1 he Bare-Stem Constraint in Greek Compound Formation Angela Ralli Athanasios Karasimos University of Patras University of Patras Abstract: In this paper, we address the issue of constraints in word formation. We claim that the absence of derivational suixes within Modern Greek compounds is due to the operation of the so-called Bare-stem constraint, which applies to output conigurations. Our analysis builds on diferent types of compounds from Standard Modern Greek and its dialects. However, we focus mostly on dvandva [V V] compounds, which are unique in Modern Greek from all Indo-European languages. We also discuss a limited number of counter-examples, and show that they are only apparent exceptions to the operation of the constraint. We argue that most of them result from a reanalysis procedure, or refer to lexicalizations and loan words, which do not usually obey the rules of the language. The paper also adds to the discussion about the interaction between derivation and compounding. It is argued that the two processes intermingle in such a way that compounding cannot be treated separately from derivation. his conclusion is advocated by the postulation of a morphologically proper constraint restricting the form of compounds with a derived item as left-hand constituent, as well as by the unclear order according to which the two processes occur.* Key words: Constraints, dvandva compounds, derivation, morphology, Greek 1. Introduction Constraints constitute an eicient device for restricting grammars and iltering out ungrammatical structures. hey have become popular in phonological theory, and occupy a prominent position in the constraint-based framework of Optimality heory (Prince and Smolensky 1993). However, constraints may also apply to the other levels of grammar, for instance, to morphology, where they may elucidate why certain word structures are possible while other structures are not. Although there is no extended literature on this topic in morphology, hints about the operation of constraints can be found in Rainer (2000), where he mentions the occurrence of morphological restrictions on the input, which he calls ‘rule- or processspeciic constraints’. * Parts of this paper have been presented at the 136th Meeting of the Linguistic Society of Japan (Tokyo, Gakushuin University: June 2008). Angela Ralli is particularly grateful to Taro Kageyama for having invited her, Hideki Kishimoto for his most constructive comments, and the audience of this meeting for insightful remarks and criticism. he authors would also like to thank Geert Booij for his precious observations on an earlier draft of the paper. 2 Angela Ralli Athanasios Karasimos In this paper, we deal with the operation of constraints on compounding. In particular, we propose the existence of a constraint which afects the form of Modern Greek (hereafter Greek) compounds with a derived item in the left-hand position. By dealing with data that involve the application of both derivation and compounding, we also add the discussion about the interaction of these two processes, an interesting topic in the recent literature (see, among others, Bauer 2005, Booij 2005, ten Hacken 2000, Ralli and Dimela to appear, etc.). We provide evidence in favor of the thesis that compounding intermingles with derivation in such a way that it cannot be treated separately from aixation, contrary to Anderson (1992) who proposes that compounding should be taken care of by syntax. he paper is structured as follows. We start by presenting the general background and hypotheses about Greek compounding, its order of application with respect to derivation, and describe the problem of not having overtly realized derivational suixes within compounds even though these suixes seem to afect the semantic interpretation of these morphological constructions. Next, we propose the existence of a constraint (the Bare-stem constraint), which requires the surfacing of stems in the left-hand position of compound words to be as bare as possible. Signiicant evidence for this constraint is provided by dvandva [V V] formations which are described and analyzed in Section 4. A small number of counterexamples is thoroughly examined in the subsequent section, where evidence is provided that they do not constitute real counterevidence to the correctness of the constraint. In Section 6, we return to the question of the order of application of derivation and compounding, in conjunction with the operation of the Bare-stem constraint, in order to show the close interaction of the two processes, a phenomenon which requires a place of compounding within morphology. he paper ends with a summary of our conclusions in Section 7. 2. Background and Hypotheses Within a lexical morphology framework (Kiparsky 1982), Ralli (1988) has claimed that in Greek, derived items appearing in compounds are formed before compounding takes place, and that the stratum/level of derivation precedes that of compounding. his proposal seems to be borne out as far as the second member of a compound word is concerned, which, in several instances, constitutes a derived item belonging to one of the three major grammatical categories, noun (1a), verb (1b), and adjective (1c):1 (1) a. nixokoptis1 lit. nail cutter ‘nail clipper’ < [nix]-o-[kop-ti-s] [nail]N-CM-[cutV-DERN-INFL(NOM.SG)]N ‘nail’ ‘cutter’ 1 he glosses should be read as: CM=compound marker, INFL=inlectional suix, DER=derivational suix, NOM=nominative, SG=singular, PRES=present, 1P=irst person, SG=singular. See Ralli (2008a) for details about the compound marker, which does not show up when the second constituent of the compound begins with a vowel, as in (1c). he Bare-Stem Constraint in Greek Compound Formation 3 b. krifoxorevoο < [krif ]-o-[xor-ev-o] lit. secretly dance [secretly]ADV-CM-[danceN-DERV-INFL(PRES.1P.SG)]V ‘dance secretly’ ‘secretly’ ‘dance’ c. aksiaγapitos < [aksi]-o-[aγapi-t-os] ‘worth loving’ [worth]A-CM-[loveV-DERA-INFL(NOM.SG)]A ‘worth’ ‘loving’ In the examples listed above, the two constituents are linked together by a linking vowel –o- (‘compound marker’ for Ralli 2008a). he irst constituent is a morphologically simple stem,2 while the second constituent is a derived item, which contains a stem, a derivational suix (-ti-, -ev-, -t-), and the appropriate inlectional ending (-s, -o, -os). We assume that in these words, derivation occurs before compounding, since compounds such as the verbal *nixokovo ‘cut nails’ (< nix(i)3 ‘nail’ + kovo ‘cut’), the nominal *krifoxoros ‘secret dance’ (< krif(os) ‘secret’ + xoros ‘dance’), and the verbal *aksiaγapo ‘worth to love’ (< aksi(os) ‘worth’ + aγapo ‘love’) are not generally acceptable for native speakers of Greek. Corroborating evidence for the claim that the derivation of the second constituent occurs before compounding is also provided by the position of stress: as argued by Nespor and Ralli (1996), a derived word at the right-hand side of a compound blocks the application of a compound-speciic stress rule, which places stress on the antepenultimate syllable of Greek compounds. Consider the following examples: (2) a. thalasodarménos < thálas(a)4 dar-mén-os vs. *thalasodármenos ‘sea beaten’ ‘sea’ beat-DER-INFL(NOM.SG) ‘beaten’ b. pagóvuno < pág(os) vun-ó lit. ice mountain ‘ice’ mountain-INFL(NOM/ACC.SG) ‘ice berg’ ‘mountain’ We see that in (2a) the position of stress of the compound as a whole is identical to the position of stress of the second member, which is a derived word. On the 2 A bare stem coincides with what is usually called ‘root’. Following Ralli (1988, 2005), we assume that in Greek morphology there is no structural diference between a stem and a root, since stems can be morphologically simple (in this sense, they correspond to roots), or morphologically complex. he latter may contain derivational aixes (derived stems) or more than one stem (compound stems). his position is also diachronically justiied because Ancient Greek stems were formed out of roots with the adjunction of a thematic vowel. Today, thematic vowels have lost their original role and are not recognizable as distinct units. See also Kiparsky (to appear) for the use of stem as the base for the formation of verbal derivatives and compounds. 3 In this paper, segmental material, which is not relevant for the discussion, e.g. inlection of the irst constituent, will be included in parentheses. 4 thálasa and págos are the forms of the fully inlected words in the nominative singular. In this paper, stress is noted only if it is relevant for the argumentation. 4 Angela Ralli Athanasios Karasimos contrary, the stress of the compound in (2b), which contains two morphologically simple stems, falls on the antepenultimate syllable, that is on a diferent position from that in the two members when taken in isolation. However, the proposal that derivation precedes compounding is not conirmed as far as the irst constituent is concerned. As noticed by Karasimos (2001) and Ralli (2007), usually derivational suixes do not appear in the irst constituent of compounds, which is generally a bare (morphologically simple) stem. In the examples listed below, the irst constituent behaves like a derived item from the semantic point of view, and its lexical category is not the one that is predicted by its overt form. However, no derivational suix is overtly realized: (3) Compound noun a. sideroporta lit. Iron made door ‘iron door’ b. krifotragudo lit. secretly sing ‘sing in secret’ c. xoropidο lit. dance – jump ‘jump like dancing, bob’ derived irst constituent siderN-eniaA iron-DER ‘iron made’ krifA-aADV secret-DER ‘secretly’ xorN-evVdance-DER ‘dance’ second constituent porta ‘door’ tragudo ‘sing’ pido ‘jump’ For instance, while a compound such as krifotragudo means ‘sing in secret’, a semantic interpretation which reveals the presence of the adverb krifa ‘secretly’ in the position of the irst constituent, the form of this item is similar to the one of the adjectival stem ‘secret’ (krif-), since it does not bear the adverbial suix –a which is usually added to adjectival stems in order to form adverbs. Similar considerations apply to the other two examples, (3a) and (3c), as well. he non-occurrence of derived items as irst constituents of compounds could be used as an argument against a linear ordering in which compounding follows derivation. Since we have already seen evidence in favor of the opposite ordering, the question, is why derivational aixes are absent from compound-internal constituents. Note that with respect to derivation, we restrict our attention to suixation: it is generally known that the derivational status of several preixes is not clear, and that, several preix-like morphemes behave like the left-hand constituents of compounds. Suice it to mention the characteristics of stress subordination and category- neutrality that are shared by the so-called Class II preixes in English (e.g. pro- and en- as in the words proclitics and enclitics), and the left-hand constituents of compounds (see, among others, Stekauer 2005).5 3. he Bare-Stem Constraint In our opinion, a plausible answer to the question above should be looked for in 5 hese two characteristics carry over to the corresponding Greek preixes as well. he Bare-Stem Constraint in Greek Compound Formation 5 the operation of constraints on word structure. We would like to claim that the absence of derivational suixes within compounds is only supericial, and that it is independent from the order according to which the processes of compounding and derivation occur. We propose that derivational suixes within the irst constituent of compounds become invisible because of the operation of a morphological constraint, which applies to output conigurations, and restricts the surface form of compounds with derived items in the left-hand position. Let us call it the Barestem constraint. We will see below that Greek compounds are generally subject to this constraint, which modiies their structures by not permitting derivational suixes to surface word-internally, and requires the irst stem component to be as bare as possible, i.e. without any suixal material.6 Since constraints should not apply at random, but for a particular reason, we further propose that the Bare-stem constraint ensures a better cohesion of the internal structure of compounds, i.e. a strong structural bond between their two basic components: (4) Bare-Stem Constraint he cohesion of a compound is better guaranteed if the irst stem is as bare as possible.7 We believe that the existence of this constraint is justiied by the general structure of the vast majority of Greek compounds that have a stem in the position of the irst constituent, i.e. an item with its inlectional ending stripped of, tightly combined with a following stem or a word. As shown by a number of authors (Drachman and Malikouti-Drachman 1994, Nespor and Ralli 1996, MalikoutiDrachman 1997, Revithiadou 1997, and Ralli 2005, 2007), with few exceptions, Greek compounds are mainly built on two patterns: [stem stem] (5a) and [stem word] (5b): (5) a. [stem stem] ambeloxόrafo < ambél(i) ‘vineyard ield’ ‘vineyard’ b. [stem word] domatosaláta < domát(a) ‘tomato salad’ ‘tomato’ xoráf(i) ‘ield’ saláta ‘salad’ he criteria according to which compounds are assigned to one of these categories are the position of stress and the form of the inlectional ending. Compounds which are subject to a compound-speciic stress rule (that places stress on the antepenultimate syllable), and inlect diferently from their second constituent, when used as an autonomous word, are assumed to have the [stem stem] structure (see 6 According to Booij (p.c.) a constraint according to which the left-hand constituent must be simplex may also be found in certain compounding patterns of Dutch, which combine an adjective with a noun or another adjective. 7 In an Optimality-heory framework, this constraint should be ranked higher than the faithfulness constraint, in order to make its efect visible in compound formation. 6 Angela Ralli Athanasios Karasimos (5a)). hose which preserve the stress and the inlectional ending of the second constituent (in endocentric constructions this constituent has the role of the head) are assumed to have a [stem word] structure (see (5b)). Following the Structure Preservation Principle, as proposed by Emonds (1985), Nespor and Ralli (1996) have argued that the structure of a word constituent that appears in the position of the right-hand head, is preserved in a compound, since it constitutes a fully speciied entity from all points of view: it is an autonomous item on structural grounds, one phonological word, and it bears the appropriate morphosyntactic features that are needed for syntactic purposes. Unlike words, stems have no structural autonomy, are not complete phonological words, and are underspeciied with respect to some morphosyntactic features (e.g. case, number, person, etc.).8 herefore, [stem stem] compounds may display properties that do not belong to those of their members, when these members are used as autonomous items. As an illustration, consider the compound meronixto ‘day (and) night’, which shows a diferent gender value, a diferent stress position, and a diferent inlectional ending from those of its two components: (6) merόnixto < mér(a) lit. day-night.NEU day.FEM ‘day and night’ ‘day’ níxt(a) night.FEM ‘night’ he fact that a stem, i.e. a non-autonomous constituent, appears at the left-hand side of Greek compounds makes their internal structure to exhibit a stronger structural cohesion than the internal structure of compounds which would have a fully speciied word as left constituent. We, thus, suggest that this desire for structural cohesion justiies the operation of the Bare-stem constraint, which does not allow for the overt presence of material other than the segments of the bare stem in the irst position of compounds. However, the degree of internal structural cohesion may vary from one compound type to another. It does not depend only on the morphological category of the constituent parts (stem or word), but also on the kind of structural relation that holds between them. For instance, there are compounds with a weak structural relation between their members, the so-called loose compounds. he absence of a strong structural bond between the constituents of loose compounds should not normally forbid the overt presence of any suixal material within their structure. If this is not the case, the constraint inds robust support: it would prove that requirements for internal structural cohesion hold across compounds, and apply even to those whose members are not tightly bound. 4. Dvandva [V V] Compounds Signiicant evidence for the Bare-Stem Constraint comes from the domain of dvandva [V V] compounds, which are also called copulative or coordinative 8 Only gender is a fully speciied feature of noun stems, as claimed by Ralli (1999, 2002). he Bare-Stem Constraint in Greek Compound Formation 7 (Bloomield 1933), or co-compounds (Wälchli 2005).9 hese constructions are an innovation of the language, since they did not exist in Classical Greek (5th–4th c. BC). hey are unique in Modern Greek within the family of Indo-European languages, but are frequently used in the East and South East Asian languages as, for instance, in Japanese (Kageyama 2009), Chinese (Packard 2000), Korean (Sohn 1999), and Vietnamese (Nguyen 1997). Dvandva [V V] compounds have appeared during the late medieval period (around the 14th c. AD), as shown by Manolessou and Tsolakidis (2007). hey belong to the productive structures of Greek compounds: (7) a. anigoklino ‘open – close’ b. anavozvino ‘switch on – switch of (the light)’ c. benovgeno ‘go in - go out’ d. trogopino ‘eat – drink’ < anig(o) ‘open’ < anav(o) ‘switch on’ < ben(o) ‘go in’ < trog(o) ‘eat’ klino ‘close’ zvino ‘switch of ’ vgeno ‘go out’ pino ‘drink’ Structurally, these compounds combine a stem and a word (they are [stem word] compounds, see Ralli 2009), and semantically, the two coordinated verbs express compatible (often synonymous) or opposite meanings. According to the semantic relationship that holds between the irst and the second verb, dvandva [V V] compounds can be classiied into three groups, additive (8a), synonymic (8b) or antonymic (8c), while most of the times it is diicult to distinguish additive from synonymic ones.10 If the two verbs are synonymous the compound denotes the joint activity over some period (Kiparsky to appear), and one of the verbs is used to reinforce the meaning of the other. On the other hand, compounds involving antonymic verbs express an iterative alternation (Nicholas and Joseph 2007, to appear, Kiparsky to appear), and occur more often than the constructions whose constituents are of compatible meanings: (8) a. zimomagirevo < zim(ono) ‘knead – cook’ ‘knead’ b. klidomadalono < klid(ono) ‘lock – bolt’ ‘lock’ c. pigenoerxome < pigen(o) lit. go - come ‘go’ ‘come and go’ magirevo ‘cook’ madalono ‘bolt’ erxome (iteration) ‘come’ As opposed to subordinative verbal compounds, for instance, [N V] formations (e.g. aisokolo ‘stick posters’ < ais(a) ‘poster’ + kolo ‘stick’) and [Adv V] ones (e.g. 9 he term ‘dvandva’ comes from the Sanskrit tradition, but is adopted by a number of linguists, including (Bauer 2008) 10 For Wälchli (2005: 137–139), additive compounds are the most prototypical. 8 Angela Ralli Athanasios Karasimos kalotroo ‘eat well’ < kal(a) ‘well’ + troo ‘eat’), which are generally right-headed, in dvandva [V V] compounds it is not clear whether the second constituent has the role of the head: the two internal members are of the same grammatical category, they display parallel argument structures, and their meaning is a conjunction of the meanings of their subparts. Since neither of the components dominates the other, we could adopt Kageyama’s (to appear) suggestion about similar Japanese constructions, that they are double-headed. However, the form of their inlectional paradigm, that is their inlection class (IC) implies that the second verb has a more prominent role, at least formally. When two verbs of diferent inlection classes combine in order to form a dvandva [V V] compound, the construction adopts the inlection class of V2.11 As an illustration, consider the examples vrodoastrafto ‘thunder - lighten’, from Standard Modern Greek, and vromomirizo ‘stink - smell’, from the Asia-Minor dialect of Krini, in (9). In both cases, the compound as a whole inlects according to the inlection of V2: (9) compound. IC a. vrodoastrafto.IC1 ‘thunder – lighten’ b. vromomirizo.IC1 ‘stink – smell’ V1.IC vrod(o).IC2 ‘thunder’ vrom(o).IC2 ‘stink’ V2.IC astrafto.IC1 ‘lighten’ mirizo.IC1 ‘smell’ he question, though, is whether headedness can be identiied only on the basis of the criterion of inlection class, since V1 and V2 have an equal status with respect to the rest of their features. Since headedness is not clear-cut in dvandva [V V] compounds, neither of the verbs has a more prominent role over the other, and they express a conjunction of events, we conclude that these formations display a weaker structural relation between their components than that shown by compounds whose members are in a subordinative (or even attributive) relation.12 Additional proof for this conclusion comes from the fact that dvandva compounds generally display structural and semantic transparency, as opposed to subordinative and attributive compounds, which are easy to lose structural transparency and to develop an unpredictable meaning. herefore, they could be considered as a kind of loose compounds. Returning now to the issue of the Bare-stem constraint, we have seen in (3) instances of its application to a number of subordinative (3b, c) and attributive (3a) compounds. Nevertheless, as already stated, the existence of the constraint would be better motivated if dvandva compounds are also submitted to its operation, since the supericial absence of word-internal derivational suixes would show that the need for structural cohesion in compounds also applies to loose structures. In fact, there are dvandva [V V] compounds, the irst member of which does 11 Matsumoto (1996) has claimed that V2 is the head in Japanese dvandva compounds, since it shows the inlectional pattern of the compound. 12 See Bisetto & Scalise (2005) for a classiication of the compounds according to the relation that holds between their basic components. he Bare-Stem Constraint in Greek Compound Formation 9 not have any overt derivational suixes. However, it has the meaning of a derived stem. Consider the examples below, from Standard Modern Greek (SMG) and its dialects, where this type of compounds really abounds. hey are taken from Andriotis (1960) and the Dialectal Data Base of the Laboratory of Modern Greek Dialects at the University of Patras. he origin of each example is listed in parenthesis: (10) Compound a. alonotherizo ‘thresh – reap’ b. klidabarono ‘lock – bar’ c. kuklustsipázumi ‘wrap up – cover’ d. magirukinónu ‘cook – pour’ e. kseromarenome ‘dry – wither’ Derived Const. 1 < alonN-izVthreshing-DER ‘thresh’ < klidN-onVkey-DER ‘bar’ ‘lock’ < kuk(u)lN-ónVhood-DER ‘wrap up’ < magirN-évVcook-DER ‘cook’ < kserA-enVdry-DER ‘dry’ Const. 2 therizo (Crete) ‘reap’ abarono (SMG) stsipázumi (Lesbos)13 ‘be covered’ kinónu (Imbros) ‘pour’ marenome (Skiros) ‘wither’ Like in other typical dvandva compounds ([N N] and [A A] ones, see Ralli 2007, 2008b), in these examples, stem constituents like aloniz(o) ‘thresh’, klidon(o) ‘lock’, etc. are juxtaposed to words of the same category, in this particular case to verbs, and express a compatible or an opposite meaning. It is important to note that examples such as the ones reported in (10) do not constitute blends, and should be distinguished from them. he segments that do not surface in these examples are those of the derivational suixes, which are normally attached to the irst derived constituent, when taken in isolation (with the appropriate inlectional ending). In blends, on the other hand, portions of the two constituents may be subtracted, and this subtraction may also involve segments of the stem, other than those of the suixal part. For instance, in Hatzidakis (1905–1907) and Koutita and Fliatouras (2001), we ind blends of coordinative verbs such as malafo ‘massage and touch’ (< malas(o) ‘massage’ + psilafo ‘touch’), and korojelao ‘mock and laugh’ (< korojδev(o) ‘mock’ + jela(o) ‘laugh’).14 Crucially, the derivational suix, which is not overtly realized in the dvandva compounds of (10), is responsible for the grammatical category (verbal) and the semantics of the irst constituent. In fact, it is always present 13 he examples from Lesbos and Imbros are given in their dialectal phonological form, where unstressed /o/ and /e/ become /u/ and /i/ respectively. 14 See Koutita & Fliatouras (2001) for detailed information on Greek blends, mostly with respect to the dialects. 10 Angela Ralli Athanasios Karasimos when the constituent is used as an autonomous word, as shown by the examples in (11), where for clarity purposes, the word internal constituents are separated by a hyphen, and their lexical category is marked: (11) a. alonN-izV-o threshing-DER-INFL(PRES.IP.SG) ‘I thresh’ b. klidN-onV-o key-DER-INFL(PRES.1P.SG) ‘I lock’ c. magirN-evN-o cook-DER-INFL(PRES.1P.SG) ‘I cook’ etc. It is worth noticing that Andriotis (1960: 55) has tried to explain the nonappearance of the word-internal derivational suix as a syllable erasure afecting verbs with more than two syllables, since, according to him, disyllabic verbs are easier to pronounce than trisyllabic ones. However, this is not always the case. Andriotis himself notes that the use of trisyllabic verbal constituents in compounds is not unknown in Greek. As an illustration, see, for instance, the examples anigοklino ‘open-close’ < anig(ο) ‘open’ + klino ‘close’ and pigenoerxome ‘go - come’ < pigen(o) ‘go’ + erxome ‘come’, etc. It is crucial to stress that the part which is systematically absent from the examples of (10) is not any particular syllable, but the derivational suix itself. herefore, dvandva [V V] compounds are afected by the operation of the Bare-stem constraint, which applies to their structure in order to maximize the bound between V1 and V2, and in spite of the fact that these compounds constitute loose structures. 5. Speciic Cases In this section we examine a small number of compounds with internal derivational suixes, which are not afected by the Bare-stem constraint. We provide a detailed study of these formations, and try to show that they do not provide counter evidence to the application of the constraint. 5.1. he verbal suix –en– here are few counter-examples to the Bare-stem constraint, which do not allow for any suixal material within compounds, namely those containing the verbal stems pigen(o) ‘go’ and ben(o) ‘go in’. hese stems keep their –en- segments in formations like pigenoerxome lit. ‘go - come’ ‘come and go’, pigenoferno ‘go - bring’, and benovgeno ‘go in (and) out’ (12bcd). As opposed to these formations, other compounds with –en-, for instance, anevokateveno (12a) do not display an overt –en-, as predicted by the operation of the constraint: he Bare-Stem Constraint in Greek Compound Formation (12) Compound a. anevokateveno lit. go up - go down ‘go up and down’ b. pigenoerxome lit. go - come ‘come and go’ c. pigenoferno lit. bring forth - bring back ‘bring forth and back’ d. benovgeno lit. go in - go out ‘go in and out’ Derived Const. 1 < anev-en‘go up’ Const. 2 kateveno ‘go down’ < pig-en‘go’ erxome ‘come’ < pig-en‘bring forth’ ferno ‘bring back’ < b-en‘go in’ vgeno ‘go out’ 11 In order to explain the examples of (12b, c, d), a solution would be to suppose that –en- is a suix in the case of aneven(o) ‘go up’ (12a), and as such, it loses its overt form in compounding, while it has no suixal character but is part of the morphologically simple stem, in the cases of pigen(o) ‘go’ and ben(o) ‘go in’. However, the question is whether there is any supporting evidence in favor of this hypothesis. It is important to note that –en- is not a category-changing derivational suix but rather a morpho-syntactic marker, since its main function is to add the [−perfective] aspectual value to a verbal stem. Substantial proof for this interpretation is ofered by verbs which show –en– in the [−perfective] forms (e.g. in the present tense), but have a stem form without –en– in the [+perfective] forms, for instance in the past tense (aorist): (13) Present [−perfective] a. anev-en-o ‘I go up’ b. pig-en-o ‘I go’ c. b-en-o ‘I go in’ Aorist [+perfective] anev-ik-a15 ‘I went up’ pig-a ‘I went’ b-ik-a ‘I went in’ his explains why in the [+perfective] context morphologically simple stems such as anev-, pig-, and b-, do not belong to a diferent grammatical category, and do not have a diferent semantic interpretation from their correspondent stems aneven-, pigen-, and ben- in the [−perfective] context. However, while a verb like aneveno ‘ go up’ is afected by the Bare-stem constraint, and shows only the bare stem anev(the one without the –en- suix), when used as irst constituent of dvandva [V V] compounds, beno ‘go in’ and pigeno ‘go’ behave diferently. In these verbs, the stem forms ben- and pigen- are not only unafected by the Bare-stem constraint, but are 15 -ik- is one of the overt markers of the morphosyntactic-feature of [+perfective]. See Ralli (1988) for more details. 12 Angela Ralli Athanasios Karasimos also used in both the [+perfective] and [−perfective] contexts, in spite of the fact that –en- is the [−perfective] marker. Compare the examples of (14a, b, c, d) with those of (14e, f ), where –en- is underlined: (14) a. To pigenoferni/*pigοferni arketes fores prin apofasisi na mas to xarisi lit. it brings.back.and.forth several times before decides to us it give ‘(S)he brings it back and forth several times before (s)he decides to give it to us’ b. To pigenoefere/*pigoefere arketes fores prin apofasisi na mas to xarisi lit. it brought.back.and.forth several times before decided to us it give ‘(S)he brought it back and forth several times before (s)he decided to give it to us’ c. Benovgeni/*bikovgeni apo to proi os to vradi lit. (S)he comes.in.and.out from the morning till the evening ‘(S)he is coming in and out from morning to night’ d. Benovgike/*bikovgike arketes fores apo to proi lit. (S)he came.in.and.out several times from the morning ‘(S)he came in and out several times from the morning’ vs. e. Anevokateveni/*anevenokateveni ta skalia arketes fores ti mera lit. Climbs.up.and.down the steps several times the day. ‘(S)he climbs up and down the steps several times a day’ f. Anevokatike/*anevenokatevike ta skalia arketes fores simera lit. climbed.up.and.down the steps several times today. ‘(S)he climbed up and down the steps several times today’ With respect to beno ‘go in’, it is important to note that if –en- does not surface (because of the Bare-stem constraint), the stem is reduced to one consonant b-. We would like to suggest that in order to preserve its form integrity the particular stem escapes the application of the constraint, and that the internal structure of the stem [b-en] has been reanalyzed as a morphologically simple stem. As a result, a compound like *bovgeno (< b-CM-vgeno) ‘go in (and) out’ is impossible, and benis used in the [+perfective] context as well (14d). A reanalysis procedure reducing a morphologically complex stem to a simple one seems to have been applied to the internal structure of the verb pigeno ‘go’ too. In this way, we could explain not only why -en- appears inside dvandva [V V] compounds, but also why the form pigen- is used in the [+perfective] forms of the aorist, as the sentence of (12b) illustrates, in spite of the fact that –en- has been described as a [−perfective] marker. Additional proof for this claim is ofered by the free alternation of pigen- with the bare stem form pa- in the paradigm of the present tense of Modern Greek, where pa- does not contain any overt [−perfective] marker: he Bare-Stem Constraint in Greek Compound Formation (14) a. pigeno pigenis pigeni pigenume pigenete pigenun / / / / / b. pao pas pai pame pate pane 13 ‘I go’ ‘you go’ ‘(s)he goes’ ‘we go’ ‘you go’ ‘they go’ If -en- in pigen- has lost its role as a [−perfective] marker, and its contribution to the formation of the verb stem is not morpho-syntactically transparent, it follows that it cannot be afected by the operation of the Bare-stem constraint. 5.2. he nominal suixes he validity of the Bare-stem constraint is also put into doubt by the presence of certain nominal suixes, which are found at the end of the irst stem constituent of nominal compounds. Consider the following examples, which display a wordinternal derivational suix regardless of the operation of the constraint: (15) a. kinisiotherapia ‘kinesiotherapy’ b. c. d. e. f. < kiniV-siN move-DER ‘movement’ klistofovia < klisV-tA ‘claustrophobia’ close-DER ‘closed’ aeriagοgοs < aerN-iN‘gas-pipe’ wind-DER ‘gas’ agrotospito < agroN-t(i)N‘farmer’s house’ land-DER ‘farmer’ anixtomialos < anixV-tA‘open-minded’ open-DER ‘open’ ikonomikopolitikos < ikonomN-ikA‘economic-political’ economy-DER ‘economic’ therapia ‘therapy’ fovia ‘phobia’ agogos ‘pipe’ spit(i) ‘house’ mial(o) ‘mind’ politikos ‘political’ he fact that these compounds are nominal, and that their left-hand stem belongs to the nominal category, is crucial to our argumentation. As is the case for nouns and adjectives, nominal compounds difer from verbal ones, in that they can be loanwords or ‘calques’.16 If compounds like those in (15) belong to a speciic register of words, and if only these compounds display a word-internal derivational suix, we could claim that they are not real counter-examples to the operation of the Bare- 16 he existence of a big range of verbal compounds makes Greek distinct from other European languages, where these formations are either rare or not productive (see Booij 1992, among others). 14 Angela Ralli Athanasios Karasimos stem constraint, which only afects ordinary Greek compounds, both verbal and nominal, i.e. compounds which do not belong to a particular language register. Depending on the origin and their structure, the examples in (15) are marked for certain speciic characteristics, which can classify them into three categories: a) loans, calques and pure translations from other languages, b) compounds which keep the word-internal derivational suix in order to avoid a meaning confusion, and c) compounds which originate from lexicalized phrases. 5.2.1. Words like kinisioθerapia (15a) and klistofovia (15b) are calques, or translations of terms from other European languages, in this particular case, from the English kinesiotherapy and the French claustrophobie.17 As is well-known, the form of loans and calques may deviate from the usual formations of the target language, and thus, may not be afected by the Bare-stem constraint. In fact, kinisiotherapia contains the compound-internal suix –si, which also appears in the English kinesiotherapy but without being identiied as such in the source language. Furthermore, at the moment of the adoption of the French term claustrophobie, the latinate claustro- was translated into the Greek derived adjective klisto- ‘closed’, which can be transparently analyzed into the verbal stem klis- ‘close’, the adjectival suix –t- and the compound marker/linking element –o-. With respect to (15f ), we should point out that the violation of the Bare-stem constraint is not due to the speciic type of –ik-, since there are similar compounds, i.e. dvandva [A A] ones, whose irst component is a derived item in –ik-, and this –ik- is not overtly realized. Consider the following examples, which display a juxtaposition of ethnic names, and a lexible order between constituents: (16) a. anglogermanik(os) English-German b. italorosik(os) Italian-Russian c. rinolaringik(os) rhinolaryngic d. kiklokilindrik(os) cyclocylindrical / germanoanglik(os) German-English / rosoitalik(os) Russian-Italian / laringorinik(os) laryngo-rhinic / kilindrokiklik(os) cylindro-cyclic In these examples –ik- has no overt form when the constituent is at the left-hand position, but is morphologically present when the same constituent is used as second member of the compound. herefore, the reason why there is –ik- in (15f ) should be searched elsewhere. Note that (15f ) belongs to a small group of dvandva [A A] compounds, like ikonomikopolitikos ‘economic-political’, politikokinonikos ‘political-social’, iθikoθriskeftikos ‘etchical-religious’, etc, which have been created during the 19th century in order to fulill speciic scientiic needs (see Babiniotis 2002). Like the examples of the previous category, some of them constitute simple 17 According to the most recent Greek dictionaries, i.e. Babiniotis (2002) and Idryma Triantaphyllidi (1998). he Bare-Stem Constraint in Greek Compound Formation 15 calques from French (e.g. isikoximikos < Fr. physicochimique, attested in 1821 according to the Idryma Triantaphyllidi Dictionary), while others have been created by analogy, more or less at the same period (e.g. politikokinonikos in 1825, ikonomikopolitikos ‘economic-political’ in 1894). Again, words of this type, which are constructed for speciic purposes do not constitute suicient evidence to cast doubt on the validity of the Bare-stem constraint. 5.2.2. In certain formations, the presence of the derivational suix seems to be necessary in order to disambiguate the meaning of the compound. For instance, in the examples aeriaγογοs ‘gas pipe’ (15c) and aγrotospito ‘farmer’s house’ (15d) the alternative forms without the derivational suixes –ti(s) and –i(o), would be aeraγογοs and aγrospito, which are also possible in Greek, but have a diferent meaning, ‘air-hole’ and ‘country-house’, respectively. herefore, semantic ambiguity can be avoided if the compounds in (15c, d) keep the suix in their surface morphological form. 5.2.3. he occurrence of the derivational suix –t- within compounds like anixtomialos ‘open-minded’ (15e), is restricted to cases where the irst component slot is illed by the deverbal adjective anixt(o) ‘open’.18 We would like to propose that compounds with anixt(o) as their irst constituent originate from phrases, in this particular case, from the phrase anixto mialo ‘open mind’, the structure of which has undergone lexicalization, and, as is well-known, lexicalized structures may be diferent from the ones which are built within morphology. However, compounds with anixt(o) at the left-hand side are generally considered to be structurally transparent exocentric formations, and their structure can be analyzed according to the rules of Modern Greek compounding. Following Ralli (2007), we further suppose that after lexicalization, items like (15e) have been submitted to a structural reanalysis as compounds, analogically to other exocentric compounds of a similar structure, i.e. to compounds containing the combination of an adjective and a noun (e.g. oligomelis ‘few membered’ < oliγ(o) ‘few’ + mel(os) ‘member’). It is crucial to note though that this reanalysis has afected only the functional elements of the construction, i.e. the inlectional ending –o of the adjectival word anixto ‘open’, which got reinterpreted as the compound marker –o-, and the inlectional ending –o of the noun mialo ‘mind’, which was replaced by the adjectival inlectional ending –os, as seen in (15e). Items with a lexeme status, such as the verbal stem anix- and the noun stem mial-, as well as the derivational adjectival suix –t-, did not lose their identity. As a consequence, the derivational suix –t- is overtly present within the structure of the compound anixtomialos. 6. Ordering between Derivation and Compounding Revisited In Section 1, we tackled the issue of the order of application of derivation and 18 Other similar occurrences with anixt- as irst constituent are the examples of anixtoxeris ‘open- handed’, anixtokardοs ‘open-hearted’, and anixtomatis ‘open eyed’. 16 Angela Ralli Athanasios Karasimos compounding. In the subsequent sections, we showed that the absence of compound-internal derivational suixes is only apparent, since the non-surfacing of derivational material is due to the operation of the Bare-stem constraint, which renders invisible material other than the segments of the irst component’s bare stem in order to ensure a better structural cohesion between the two components of a compound. herefore, it may be misleading to conclude that the absence of compound-internal derivational suixes provides arguments in favor of an ordering of compounding after derivation. Since we have seen examples advocating the opposite order (see (1)), should we deduce that there is a linear order which requires derivation to occur irst? It is important to point out that there is no positive answer to this question. On the one hand, there are derived words which feed derivation, as shown by the examples in (1), but on the other hand, there are compound structures which are subject to derivation. For instance, consider the adjectival compound xartopektikos ‘gambling’ and the noun peδerastia ‘pederasty’. hese words are built on the basis of the combination of a compound stem with a derivational aix, as depicted in (16): (16) a. xartopektikos < xart-o-pekt(i)-ik-os lit. card-playing card-CM-player-DER-INFL(NOM.SG) ‘gambling’ ‘card’ playing’ b. peδerastia < peδ-erast-ia-Ø lit. child-loving child-lover-DER-INFL(NOM.SG)19 ‘pederasty’ ‘child’ ‘loving’ In (16ab), there are no actual de-adjectival words *pektikos ‘playing’ and *erastia ‘loving’, which would imply a linear order in which derivation occurs before compounding. Moreover, the existence of compounds like xartopektis ‘gambler’ (< xart(ia) ‘cards’ + pektis ‘player’) and peδerastis (< peδ(i) ‘child’ + erastis ‘lover’) ofers arguments in favor of the opposite order, i.e. compounding preceding derivation. It should be noticed that there are also occurrences of verbal compounds, like alatopiperono ‘put salt and pepper’ (17a), where native speakers cannot take a clear decision in favor of one particular order: (17) a. alatopiperono < lit. put salt - put pepper ‘to salt and pepper’ b. alatopiperono < lit. put salt - put pepper ‘to salt and pepper’ c. alatopipero < lit. salt-pepper ‘salt and pepper’ [[[alat-o-piper]-on]-o] [[[salt-CM-pepper]-DER]-INFL(PRES.1P.SG)] ‘to salt’ ‘to pepper’ [[[alat-iz]-o-[piper-on]]-o] [[[salt-DER]-CM-[pepper-DER]]-INFL(PRES.1P.SG)] ‘to salt’ ‘to pepper’ [[alat-o-piper]-o] [[salt-CM-pepper]-INFL(NOM/ACC/SG)] ‘salt’ ‘pepper’ 19 In this compound, there is a zero inlectional ending. Moreover, there is no compoundinternal marker –o- because the second member begins with a vowel. See Ralli (2008a) for more details. he Bare-Stem Constraint in Greek Compound Formation d. alatizo ‘to salt’ e. piperono ‘to pepper’ 17 alat-iz-o salt-DER-INFL(PRES.1P.SG) ‘to salt’ piper-on-o pepper-DER-INFL(PRES.1P.SG) ‘to pepper’ In (17), the very frequent dvandva [N N] compound alatopipero ‘salt-pepper’ (17c) provides an indication for a subsequent derivational formation alatopiperono ‘to salt and pepper’ (17a), on the basis of the compound noun stem alatopiper- ‘salt and pepper’ and the derivational suix –on- (-o being the inlectional ending). However, this is only an indication borne out by the dictionaries, which view the derived verb alatopiperono as a secondary compound formation on the basis of the primary nominal compound alatopipero. heoretically, we could suppose that the structure is built on the combination of two derived verbal stems, the most common alatiz‘to salt’ (17d) and the less common piper-on- ‘to pepper’ (17e), a hypothesis which would denote exactly the opposite order, according to which derivation takes place before compounding, as in (17b). Moreover, in accordance with our argumentation at the previous sections, we should also suppose that the structure is afected by the operation of the Bare-stem constraint, which renders the overt form of the derivational verbal suix –iz- of the verbal stem alatiz- ‘to salt’ invisible (17d). To conclude, there is no clear evidence for an extrinsic linear ordering of the two processes. We have seen that a derived item may be used either as irst or as second member of compounds, but the operation of the Bare-stem constraint hides the overt form of derivational suixes within their structure. We have also seen that derived items can be created after compounding takes place. hus, the interaction of the two processes provides arguments for compounding being a wordformation process, which should be accounted for in the same way as derivation, i.e. within morphology. More crucially, the existence of a speciically morphological constraint, the Bare-stem constraint, which has a speciic domain of operation, i.e. compounding, and afects speciic morphological units, i.e. derivational suixes, stresses the close interaction of the two processes and also implies a morphological account of compounding. 7. Conclusions In this paper we have shown that there are morphological constraints that have an impact on the form of morphologically complex items. We have proposed the existence of the so-called Bare-stem constraint, which afects the output form of compounds with a derived item in the position of the left component. In order to preserve structural cohesion, this constraint renders invisible the derivational suix, and makes the stem component as bare as possible, even though its category and semantics are those of a derived item. he few problematic examples that exist do not provide suicient evidence against the postulation of this constraint. Unless they keep the derivational suix for purposes of disambiguation, or to maintain 18 Angela Ralli Athanasios Karasimos integrity, it is shown that these occurrences result from reanalysis or originate from foreign formations and lexicalized phrases. Finally, our paper comments on the place of compounding within the grammar. By examining the order of application between derivation and compounding, in conjunction with the operation of the Bare-stem constraint, we have shown the close interaction between the two, which argues in favor of an account of compounding in morphological terms. References Anderson, Stephen (1992) A-morphous morphology. Cambridge: Cambridge University Press. Andriotis, Nikolaos (1960) Ta parataktika rimatika sintheta stin Elliniki glossa [Coordinative verbal compounds in the Greek language]. In Memory of Manolis Triantphyllidis, 43–61. hessaloniki: Aristotle University of hessaloniki. Babiniotis, George (2002) Leksiko tis Neas Ellinikis glossas [Dictionary of Modern Greek language], 2nd ed. Athens: Centre of Lexicology. Bauer, Laurie (2005) he borderline between derivation and compounding. In: Wolfgang Dressler, Dieter Kastovsky, Oscar Pfeifer and Franz Rainer (eds.), Morphology and its Demarcations, 97–108. Amsterdam: John Benjamins. Bauer, Laurie (2008) Dvandva. Word structure 1: 1–20. Bisetto, Antonietta and Sergio Scalise (2005) he classiication of compounds. Lingue e Linguaggio IV: 319–332. Bloomield, Leonard (1933) Language. New York: Holt. Booij, Geert (1992) Compounding in Dutch. Rivista di Linguistica 4: 37–61. Booij, Geert (2005) Compounding and derivation: Evidence for construction morphology. In: Wolfgang Dressler, Dieter Kastovsky, Oscar Pfeifer and Franz Rainer (eds.), Morphology and its Demarcations, 109–132. Amsterdam: John Benjamins. Booij, Geert, Christian Lehmann, and Joachim Mugdan (eds.) (2000) Morphologie/ Morphology. Berlin: Mouton de Gruyter. Drachman, Gaberell and Angeliki Malikouti-Drachman (1994) Stress and Greek compounding. Phonologica 1992: 55–64. Dressler, Wolfgang, Dieter Kastovsky, Oskar Pfeifer and Franz Rainer (eds.) (2005) Morphology and its Demarcations. Amsterdam: John Benjamins. Emonds, Joseph (1985) A uniied theory of syntactic categories. Dordrecht: Foris. Hatzidakis, Georgios (1905–1907) Meseonika ke Nea Ellinika [Medieval and Modern Greek]. Athens. Kageyama, Taro (2009) Isolate: Japanese. In Rochelle Lieber and Pavol Stekauer (eds.) he Oxford handbook of compounding, 512–526. Oxford: Oxford University Press. Karasimos, Athanasios (2001) Allomorphy in inlection and compounding of Modern Greek. Ms. University of Patras. Koutita-Kaimaki, Mirto and Asimakis Fliatouras (2001) Blends in Greek dialects: A morphosemantic analysis. In: Angela Ralli, Brian David Joseph and Mark Janse (eds.) Proceedings of the irst international conference of Modern Greek dialects and linguistic theory, 117–130. Patras: University of Patras. Kiparsky, Paul (1982) Lexical morphology and phonology. In: he Linguistic Society of Korea (ed.) Linguistics in the morning calm, 3–92. Seoul: Hanshin. Kiparsky, Paul (to appear) Verbal co-compounds and subcompounds in Greek. MIT working he Bare-Stem Constraint in Greek Compound Formation 19 papers in linguistics: MIT workshop on Greek syntax and semantics. Leksiko tis Koinis Neoellinikis [Dictionary of Modern Greek Koine] (1998) hessaloniki: Idryma Manoli Triantaphyllidi. Lieber, Rochelle and Pavol Stekauer (eds.) (2009) he Oxford handbook of compounding. Oxford/New York: Oxford University Press. Malikouti-Drachman, Angeliki (1997) Prosodic domains in Greek compounding. In: Gaberell Drachman, Angeliki Malikouti-Drachman, Celia Klidi and John Fykias (eds.) Proceedings of the 2nd international conference of Greek linguistics, 87–96. Graz: Neugebauer Verlag. Manolessou, Io and Symeon Tsolakidis (2007) Modern Greek coordinative compounds: Diachrony and synchrony. Electronic proceedings of the 8th international Meeting of Greek linguistics. Ioannina: University of Ioannina. Matsumoto, Yo (1996) Complex predicates in Japanese: A syntactic and semantic study of the notion ‘word’. Stanford: CSLI. Nguyễn, Ðình-Hoà (1997) Vietnamese. Amsterdam: John Benjamins. Nespor, Marina and Angela Ralli (1996) Morphology–phonology interface: Phonological domains in Greek compounds. he Linguistic Review 13: 357–382. Nicholas, Nick and Brian D. Joseph (to appear) Verbal dvandvas in Modern Greek. MIT working papers in linguistics: MIT workshop on Greek syntax and semantics. Nicholas, Nick and Brian D. Joseph (2007). Verbal dvandvas in Greek: What, when and why? Paper presented at the 8th international meeting of Greek linguistics. University of Ioannina. Packard, Jerome (2000) he morphology of Chinese. Cambridge: Cambridge University Press. Prince, Alan and Paul Smolensky (1993) Optimality theory: Constraint interaction in generative grammar. Rutgers University Center for Cognitive Science. [Published by Blackwell in 2004.] Rainer, Franz (2000) Produktivitätsbeschränkungen. In Geert Booij, Christian Lehmann, Joachim Mugdan (eds.) Morphologie/ Morphology, 877–885. Berlin: Mouton de Gruyter. Ralli, Angela (1988) Eléments de la morphologie du grec moderne: La structure du verbe. Unpublished doctoral dissertation, Université de Montréal. Ralli, Angela (1992) Compounds in Modern Greek. Rivista di Linguistica 4: 143–174. Ralli, Angela (1999) Inlectional features and the Morphological Module Hypothesis. Working Papers in heoretical and Applied Linguistics 6: 111–142. hessaloniki: English Dept. of Aristotle University of hessaloniki. Ralli, Angela (2002) he role of morphology in gender determination. Linguistics 40: 519–551. Ralli, Angela (2005) Morfologia [Morphology]. Athens: Patakis. Ralli, Angela (2007) I sinthesi lekseon: Diaglosiki morfologiki prosengisi [Compounding: A cross-linguistic morphological approach]. Athens: Patakis. Ralli, Angela (2008a) Compound markers and parametric variation. Language Universals and Typology and Universals (STUF) 61: 19–38. Ralli, Angela (2008b) Greek dvandva [V V] compounds: A linguistic link between Greece and East/South-East Asia. Paper presented at the 136th Meeting of the Linguistic Society of Japan. Tokyo: Gakushuin University. Ralli, Angela (2009) Hellenic compounds. In: Rochelle Lieber and Pavol Stekauer (eds.), he Oxford handbook of compounding, 243–253. Oxford: Oxford University Press. Ralli, Angela and Eleonora Dimela (to appear) On the borderline between preixation and compounding: he sa-adverbs in the dialectal varieties of Lesbos, Kydonies (Aivali) 20 Angela Ralli Athanasios Karasimos and Moschonisia. In: Mark Janse. Brian David Joseph, Pavlos Pavlou and Angela Ralli (eds.) Proceedings of the 3rd international conference of Modern Greek dialects and linguistic theory. Nicosia. Revithiadou, Anthi (1997) Prosodic domains in Greek compounding. In: Gaberell Drachman, Angeliki Malikouti-Drachman, Celia Klidi and John Fykias (eds.) Greek Linguistics 95. Proceedings of the 2nd international conference on Greek linguistics, 107–116. Graz: Neugebauer Verlag. Sohn, Ho-Min (1999) he Korean language. Cambridge: Cambridge University Press. Stekauer, Pavol (2005) Compounding and aixation: Any diference? In: Wolfgang Dressler, Dieter Kastovsky, Oskar Pfeifer and Franz Rainer (eds.) Morphology and its Demarcation, 151–159. Amsterdam: John Benjamins. Ten Hacken, Pius (2000) Derivation and compounding. In: Geert Booij, Christian Lehmann, and Joachim Mugdan (eds.) Morphology/Morphologie, 349–359. Berlin: Mouton de Gruyter. Wälchli, Bernhard (2005) Co-compounds and natural coordination. Oxford: Oxford University Press. [Received 29 October 2008; Accepted 7 December 2008] Authors’ contact information University of Patras University campus, 26504, Rio Greece Ralli: [email protected] Karasimos: [email protected] ［要 旨］ ギリシア語の複合語形成における裸語幹制約 Angela Ralli Athanasios Karasimos （アンジェラ・ラッリ） （アタナシオス・カリシモス） パトラ大学 本稿は語形成における制約の問題を扱い，現代ギリシア語の複合語の内部に派生接辞が現 れないのは，複合語形成の出力に適用する「裸語幹制約」が働いているからであると主張する。 この主張は現代標準ギリシア語と諸方言における種々のタイプの複合語によって支持される が，とりわけ，インドヨーロッパ諸言語には存在せずギリシア語独特の動詞＋動詞型の並列 複合語（dvandva）が重要なデータとなる。この分析に対して一見反例となる現象もあるが， それは制約の適用に対する見かけの例外に過ぎない。すなわち，これらの反例と思われる現 象は，再分析によって生じたものか，あるいは，通常の言語規則には従わない特殊な語彙化 現象ないし借用語であると見なされる。 本稿は更に派生と複合の相互関係についても議論を進め，複合操作は派生操作と峻別でき ず，両者が相互に入り混じって適用することを論じる。この結論は，派生語を左側要素とし て含む複合語の形式を規制する適切な形態論的制約を設け，複合と派生が適用する順序を明 示しないことで達成される。