3.1.1.1 Adjective + adjective compounds

Adjective + adjective compounds have been recorded in the languages shown in Table 3.2.

Table 3.2. Adjective + adjective compounding in the study sample

3.1.1.1.1 Formal characteristics

Formally, adjective + adjective compounds can occur without and with a linking element. Examples are given in Table 3.3.¹

Table 3.3. Adjective + adjective compounding with/out a linking element

The members of the compounds can be stems, but also inflected forms, as in Finnish (11) and Udihe (12) and (13). The latter can combine a content element and the copula verb bi- ‘be’ and, sometimes, also ede- ‘become’ in the corresponding form. The content element may be an adverb, a derived adjective or an ideophonic adverb:

(11)

(11)

Finnish

vaalea-n-sininen

light-gen-blue

‘light blue’

(Laakso)

(12)

(12)

Udihe

bei-bi-

in.vain-be

‘simple’

(13)

(13)

Udihe

täsi bi-

very.much-be

‘full’

(Nikolaeva and Tolskaya Reference Nikolaeva and Tolskaya2001: 200)

Adjective + adjective compounds can be, to use Scalise and Bisetto’s (Reference Scalise, Bisetto, Lieber and Štekauer2009) terminology, attributive (14) and coordinate (15):

(14)

(14)

Serbian-Croatian

svetloplav

‘light blue’

(15)

(15)

Spanish

italo-argentino

‘Italian-Argentinean’

Subordinative compounds usually present a structure of modifier + head sequence, but the opposite is also possible, as illustrated in (16):

(16)

(16)

Romanian

galben-auriu

yellow-golden

‘golden-yellow’

3.1.1.1.2 Semantic characteristics

Semantically, three types of meaning have been identified in the study sample: a purely compositional meaning,² a modified meaning,³ and a new concept motivated by the adjectival constituents, as shown in Table 3.4:

Table 3.4. Semantic types in adjective + adjective compounding

Māori presents an ambiguous case: according to Harlow (pers. comm.), there is only one example of adjective + adjective compound in this language (29) and even this seems to be questionable:

(29)

(29)

Māori

hari-koa

happy-happy

‘happy, pleased’

According to W. Bauer (Reference Bauer1997: 77), Māori does not have the word-class adjective, which in this language is rather a class of verbs.⁴ These verbs can occur as modifiers in post-nuclear position in noun phrases, as in (30) (W. Bauer Reference Bauer1997: 303):

(30)

(30)

Māori Ka pātai mai a Pou, “He wāhi pai anō?” tam ask hither per Pou cls plc good again ‘Pou asked, “Is it a nice place, too?”’

3.1.1.2 Compound verbs

Compound verbs, i.e. compounds whose head is verb, as in to spotlight or to stagemanage, are a controversial case in word-formation in languages like English. While the status of nominal and adjectival compounds is relatively clear, the existence of compound verbs in English has been called into question by a considerable number of morphologists, like Marchand (Reference Marchand1960) or Lieber (Reference Lieber2004: 48). The former calls this type of structure pseudocompounds and explains their formation by processes other than compounding. This is common in the literature and a number of sources have been cited for the alleged compound verbs, e.g. back-formation from other compounds (Marchand Reference Marchand1960: 59; Adams Reference Adams1973; Allen 1978: 214), zero-derivation from other compounds (Marchand Reference Marchand1960: 59; Adams Reference Adams1973, Reference Adams2001: 101)⁵ or noun incorporation (Štekauer Reference Štekauer2009: 282).⁶

Interestingly, Adams (Reference Adams1973) cites Pennanen (Reference Pennanen1966: 7.5)⁷ in support of her view that compound verbs may be based on analogical formation, while her 2001 revision (Adams Reference Adams2001: 109) assumes that ‘genuine verb compounding is not likely to develop in modern English’. The controversy on the existence of a verb-forming process of compounding in English is unavoidable when L. Bauer and Renouf (Reference Bauer and Renouf2001: 110) identify neologisms such as dry-burn, test-release, thumb-strum and slow-bake as compound verbs. Kiparsky (Reference Kiparsky and Yang1982a: 16) considers compound verbs with adverbials as left-hand constituents (English hand-pick, sun-dry, etc.) to be a class of systematic exceptions to the First Sister Principle. Kiparsky emphasizes that ‘the process which derives these compound verbs . . . e.g. air-condition, is the same which forms all compounds including synthetic compounds’ (1982a: 19). This process is based on the rule [Y Z]_X, with X being a verb.

Compound verbs are recorded in the languages shown in Table 3.5.

Table 3.5. Verb + verb compounding in the study sample

3.1.1.2.2 Semantic characteristics

As with any other type of compounds, the meaning of verbal compounds can be based on the principle of compositionality or it can be more than a mere sum of the meanings of the motivating constituents, as shown in Table 3.7.

Table 3.7. Compositional and non-compositional meaning in verbal compounding

3.1.1.3 Noun incorporation

Noun incorporation is recorded in the languages shown in Table 3.8. (38.18 per cent in the study sample):

Table 3.8. Noun incorporation in the study sample

Incorporation is considered here a verb-forming process whereby a nominal stem is fused with a verbal stem to yield a larger, derived verbal stem, according to the definition proposed for the purposes of our research by L. Bauer (pers. comm.). Noun incorporation follows after the section of compound verbs (3.1.1.2) because instances of noun incorporation are prevailingly viewed as compound verbs.

While a broad approach to incorporation does not seem to make a distinction between incorporation and polysynthetism, our findings are in accordance with Aikhenvald’s (Reference Aikhenvald and Shopen2007: 6, 12) observation that not all polysynthetic languages have noun incorporation and languages with incorporation need not be polysynthetic.

On the other hand, the non-existence of compounding in these languages is confirmed by the absence of incorporation. This observation corresponds with what is assumed by Leza (Reference Leza, Haspelmath, König, Oesterreicher and Raible2001: 718), according to whom not all Amerindian languages feature noun incorporation even if they are polysynthetic. Noun incorporation is productively used in many non-polysynthetic languages.

3.1.1.3.1 Formal characteristics

While the prototypical case of incorporation is nominal incorporation, specifically one in which the incorporated noun functions as an object argument of the verb, a number of peripheral types of incorporation have been reported in the literature and have been found in the languages sampled here.

Thus, Werner (Reference Werner1998: 58ff.) demonstrates that languages like Ket, Kott and Yugh, which are Yeniseian languages spoken in Eurasia, make use of ‘nominal incorporation’ in which a compound adjective in the function of the modifier and a noun as the head form a unity. The view that these expressions involve noun incorporation is supported by their ability to include possessive prefixes which are usually combined with nouns and, in certain cases, also with verbs. Examples of this kind of nominal incorporation are given in (57) and (58):

(57)

(57)

Ket

qä-γit,

⁴qä-²ki^?t

big-price

‘expensive’

(58)

(58)

Yugh

χεgit

²χε^?-²ki^?t’

big-price

‘expensive’

Adjective incorporation also occurs in Pipil, although it is reportedly rare:

(59)

(59)

Pipil

sek-kalaki

cold-enter

‘get cold’

A similar case occurs in Estonian, where a small group of verbs formed by incorporation usually have as the first component an adjective stem, but a noun is also possible. Estonian syntax prefers to express the meanings of such nominal stems as adverbials. Thus, (60a) is more marked than (60b):

(60)

(60)

Estonian

1. a. sügav-künd-ma

   deep-plough.v-sup

   ‘plough deep.v’

2. b. sügavalt.adr kündma.v

   deeply-plough.v-sup

   ‘plough deep.v’

Examples of frequently used Estonian incorporated verbs are:

(61)

(61)

Estonian

kuri-tarvita-ma

evil-use-sup

‘abuse.v’

(62)

(62)

Estonian

häda-maandu-ma

hardship-land.v-sup

‘force-land, make emergency landing’

Lakhota is a source of insights into various aspects of noun incorporation. The following is based on de Reuse’s (1994) account of this Siouan language. In Lakhota noun incorporation is treated as a type of compounding with which it also shares phonological properties, like the following three different stress patterns:

1. (a) noun incorporations with a single main stress (labelled lexical compounds) are controlled by the Dakota Accent Rule, which places stress on the second syllable of a word regardless of whether this syllable is a part of the first or of the second compound constituent. This type of noun incorporation is, phonologically, a tight unit, often with lexicalized meaning, and its productivity is low compared to so-called syntactic compounds,

2. (b) noun incorporation with stress weakening on the second syllable of the second constituent imposed on it (as well as on the first constituent by the Dakota Accent Rule before the process of incorporation). This type is semantically more transparent and more productive. Since the former, the lexical compound type of noun incorporation, is phonologically very tight, the noun stem undergoes various phonological modifications (truncation), like voicing of the final stop (final /t/ and /č/ become /l/) and devoicing of a final fricative. Cases of coalescence of the final vowel of the first constituent and the initial vowel of the second constituent accompany both types of noun incorporation, as shown in Table 3.9.

3. (c) noun incorporation which maps compounding, labelled noun stripping, because the noun ‘is stripped of the articles, determiners, and case-marking elements that usually accompany it, and then juxtaposed to the verb’ (de Reuse Reference de Reuse1994: 206):

   (65)

   (65)

   Lakhota

   napé yúzA

   hand.ins-take.hold

   ‘shake hands’

   (66)

   (66)

   Lakhota

   phá kaksÁ

   head.ins-cut

   ‘behead.v’

   (de Reuse Reference de Reuse1994: 209)

Table 3.9. Noun incorporation with truncation and with coalescence in Lakhota

Noun incorporation need not be restricted to the left-hand position of the incorporated noun. In Clallam (68) the nominal object, expressed by a lexical suffix rather than by a root, takes right-hand position. Examples where the incorporated noun takes right-hand position can be found in other languages too:

(67)

(67)

Anejom

esjaalak

esjañ-nalak

put.down-roller

‘put down rollers (for a canoe)’

(68)

(68)

Clallam

λ’əmé?q^w

λ’ə’m’-e?q^w

be bumped-head

‘get bumped on the head’

(69)

(69)

Nelemwa

thu-naar-e

do-oven-tr

‘cook it in the oven’

The position of the incorporated noun in these languages matches Mardirussian’s (1975: 384) observation that the incorporated noun attaches to the right of the verb in verb-initial languages. Word order in Clallam is VSO, VOS in Anejom and, in Nelemwa, ‘nominal arguments come after the predicate, either as VS (where [S] is the absolutive nominal argument of an intransitive verb) or VOA (where O stands for the second argument/patient and A for the agent of a transitive verb)’ (Bril Reference Bril and Haspelmath2004: 500). Interestingly, the examples from Anejom, Clallam and Nelemwa contradict the generalization made by Caballero et al. (2006), according to which the order of noun and verb in languages with unproductive noun incorporation is noun + verb, while in languages with productive noun incorporation it follows the order of words in syntax.¹⁰

3.1.1.3.2 Semantic characteristics

The description of noun incorporation mentions that an incorporated noun functions as an argument (usually object) of the predicative verb. Remarkably, in none of the latter examples does the noun function as an object: instead it identifies location. This has been reported in the literature both in general and in a number of languages. According to Gerdts (Reference Gerdts, Spencer and Zwicky1998: 87), while incorporated nouns are typically related to objects or to subjects of inactive predicates,¹¹ they can also express locatives, instruments or passive agents and they do not generally correspond to subjects of active intransitives or transitives, to indirect objects or to benefactives. Mithun (Reference Mithun, Lieber and Štekauer2009: 576) remarks that the variety of relations that hold in Mohawk between the noun incorporation members is much like what we find with noun + noun compounds in English and that ‘the same Mohawk verb root can . . . occur with incorporated nouns with quite different semantic roles’. Similarly, Kageyama (Reference Kageyama1982: 244) claims that the incorporated noun functions as a direct object in about 50 per cent of this type of Japanese compound verbs ((70) and (73)), as an adverbial in about 25 per cent ((71) and (74)), and as a subject of intransitive predicates in about 25 per cent ((72) and (75)) (see Table 3.10).

Table 3.10. Incorporated noun as direct object, as adverbial and as subject in Japanese

In line with the above, the languages sampled here show that it is not only internal arguments of verbs that can be incorporated. In Mandarin Chinese, where noun incorporation is reportedly very productive, there are, beside the default cases of object, numerous instances of the incorporated noun in other functions, as shown in Table 3.11.

Table 3.11. Incorporated nouns as other than object in Lakhota and Mandarin Chinese

Pipil offers examples where the incorporated noun performs the function of an instrumental prefix (83), as well as standard examples (84):¹⁵

(83)

(83)

Pipil

tan-kwa

tooth-eat

‘bite.v’

(Campbell Reference Campbell1985: 96)

(84)

(84)

Pipil

ku:-tapa:na

wood-break.v.open

‘split firewood.v’

(Campbell Reference Campbell1985: 97)

These examples of Pipil lead to the role of locative and instrumental prefixes in noun incorporation in Lakhota. According to de Reuse (Reference de Reuse1994: 210ff.), incorporations of the structure [n+[loc+v]] and [n+[ins+v]] are very common. They usually co-exist parallel to prefixless noun incorporations. The prefixed and prefixless constructions may differ in valence and meaning. Table 3.12 presents various combinations of locative and instrumental arguments and prefixes in Lakhota:

Table 3.12. Combinations of noun incorporation in Lakhota

Constructions with both locative and instrumental prefixes incorporated by a verb (n+[loc+[ins+v]]) are also possible (de Reuse Reference de Reuse1994: 212):

(89)

(89)

Lakhota

wé ayúš ^?e

blood-loc-ins-drop

‘drop blood on (somebody)’

While noun incorporation is connected with incorporating a single noun in the vast majority of cases, Lakhota also allows a rare case of two incorporated nouns, mostly combined with locative and instrumental prefixes. One such example, whose structure is [n+[loc+[n+[loc+[loc+v]]]]], is given in (90):¹⁶

(90)

(90)

Lakhota

xta ^?ómakhiyokpazA

evening-loc-earth-loc-loc-be.dark

‘grow dusk’

In Ket (Vajda, pers. comm.), noun incorporation is limited to a handful of verb stems like eat, make, pour or have:

(91)

(91)

Ket

d-qus-i-bet

1sg.sbj-tent-thm-make

‘I engage in tent-making’

(92)

(92)

Ket

san-nO

²sa^?q-³qO:

squirrel-kill/die

‘hunt squirrels’

(Werner Reference Werner1998: 99)

What makes noun incorporation in Yeniseian languages attractive is that in the finite verb form these nominal incorporations are split into two parts: the determining constituent always assumes position 12 and the basic constituent the zero position. If the determining constituent is a compound itself, the first constituent takes position 13 and the second constituent takes position 12 (Werner Reference Werner1998: 99).¹⁷

Noun incorporation is productive in Tamil. According to Schiffman (1996), the commonest and most general of the incorporating verbs is ஹஊ (pannu) ‘make, do’. It can be added to a noun to make a verb. This is the most common way of making verbs out of borrowed English words. Sometimes Tamil even borrows words for lexical items which are already available in the language:

(93)

(93)

Tamil

·ை ஹஊ

draiv pannu

‘drive (a car)’

(94)

(94)

Tamil

‘ ஹஊ

vaakking pannu

‘take a walk, go walking’

Schiffman explains that ஹஊ (pannu) can be attached to both nouns and verbs (usually borrowed from English), but always with the effect of having been added to nouns, i.e. what precedes pannu is a noun phrase in Tamil, regardless of whether it is a noun phrase or a verb phrase in English. Thus, ·ை draiv in (93), though a verb in English ‘drive.v’, is treated as a noun in Tamil.

Noun incorporation in Slavey may be illustrated with a set of three examples (Rice, pers. comm.). In the first (95), the pre-verb and stem together mean ‘sg.stand’, the incorporate is -tse ‘cry.n’ and d- is a prefix that occurs with an incorporate indicating an oral activity. In the second, (96), the noun stem tthí ‘head’ is incorporated, the verb stem is -chu and the prefix e- is required with this. This verb without the incorporate means ‘handle sg.object around’. Finally, in the third example, (97), the incorporate is shé ‘food’, because this stem does not occur independently:

(95)

(95)

Slavey

ná-tse-de-we

preverb-cry-qualifier-stm

‘S/he stands crying’

(96)

(96)

Slavey

k’e-tthí-e-chu

around-head-qualifier-stm

‘S/he turns his/her head’

(97)

(97)

Slavey

shé-tiN

food-stm

‘S/he eats’

This is similar to Kwakw’ala, where there is a group of incorporating verbs that only occur with their objects incorporated (Anderson Reference Anderson1992: 269).

According to Pingali, in addition to prototypical incorporation in Telugu (98), a set of auxiliary verbs participate in constructing what Pingali calls ‘quasi-compounds’. Auxiliary verbs also attach to English borrowings, as in (99) and (100):

(98)

(98)

Telugu

purugu paTTu

insect-catch

‘be infested by insects’

(99)

(99)

Telugu

dress avvu

dress-happen

‘get dressed’

(100)

(100)

Telugu

swim ceeyi

swim-do

‘swim.v’

Examples of noun incorporation in other languages are shown in (101) to (106):

(101)

(101)

Catalan

cama-trencar

leg-break-inf

‘cause (something) to break its leg’

(102)

(102)

Greek

αφισοκολλώ

afis-o-kolo

poster-stick

‘stick posters (on the wall)’

(103)

(103)

Tibetan

ག་

lug-rdzi

sheep-tend

‘shepherd.v’

(104)

(104)

Totonac

laka-lás-a

face-slap-ipf

‘slap (somebody) in the face’

(105)

(105)

Tzotzil

t^!šút^!š-ul púšil-an

small.pieces-fold.repeatedly

‘fold (something) smaller and smaller’

(Cowan Reference Cowan1969: 95)

(106)

(106)

Wichí

yenlhipeya

o-yen-lhip-ey-a

3sbj-make-piece-pl-acc

make-pieces

‘S/he chops’

3.1.1.4 Noun + noun compounds

Noun + noun compounding is recorded in the languages shown in Table 3.13.

Table 3.13. Noun + noun compounding in the study sample

Noun + noun compounds may appear without a linking element and with a linking element, as shown in Table 3.14.

Table 3.14. Noun + noun compounding with/out a linking element

This section now reviews first noun + noun compounds without a linking element and then noun + noun compounds with a linking element.

3.1.1.4.1 Formal characteristics of noun + noun compounds without a linking element

In nominal compounding, the left-hand position of the modifying noun is a default case. However, the languages sampled offer alternative distributions, like right-hand position for the modifier or either position, i.e with the modifier in the left-hand position or in the right-hand position, as in the Hindi and Vietnamese examples in Table 3.15.

Table 3.15. Modifier position in noun + noun compounding

Modifiers in compounds therefore do not always seem to be restricted to one position. Position is, however, used for a classification in some languages, like Breton, where strict compounds are right-headed (130) and loose compounds are left-headed (131) (see Table 3.16).

Table 3.16. Strict (right-headed) vs loose (left-headed) compounding in Breton

3.1.1.4.2 Semantic characteristics of compounds without a linking element

Section 3.1.1.2 shows that the non-applicability of the principle of compositionality is not a universal feature of compounds. Further examples of noun + noun compounds whose meanings can be computed as a sum of the meanings of their constituents and which are figurative meanings are given in Table 3.17. In the table, (132) and (138) are instances of the so-called pleonastic compounds, where the head is a hyperonym of the non-head.

Table 3.17. Compositional and figurative meaning in noun + noun compounding

Typical dvandva compounds in the field of kinship terms are offered by Georgian (144) and Maipure (145): ²⁰

(144)

(144)

Georgian

დედ-მამ

ded-mama

father-mother

‘parents’

(145)

(145)

Maipure

ani-kiwakané

son-father

‘husband’

3.1.1.4.3 Formal characteristics of compounds with a linking element

Compounding with a linking element is recorded in the languages shown in Table 3.18.

Table 3.18. Compounding with a linking element in the study sample

Perhaps the best-known and theoretically most extensively discussed linking elements of compounding are in German, as in Arbeitslohn, Alterungsprozess, etc. The discussion about this issue usually concerns the status of -s-, -n-, -e and of other linking elements that are historically developed from genitive and plural morphemes (Neef Reference Neef, Lieber and Štekauer2009).²¹

However, linking elements are not an exclusive feature of a particular genetically related group of languages, in this case the Germanic genus,²² or of a particular inflectional type or geographic area. Admittedly, links are not equally frequent in all linking languages. Thus, although they are rare in the Dravidian language Telugu, they can still be found in such languages, as shown in (146):

(146)

(146)

Telugu

penk-u-Tillu

tile-lnk-house

‘tiled house’

Links appear to be mainly single-phoneme elements, either vowels or consonants. If an interfix is formed by a vowel, then it is usually a back vowel (low or mid), while for consonants there does not seem to be any pattern. Examples are given in Table 3.19.

Table 3.19. Vocalic and consonantal links in compounding

3.1.1.5 Exocentric compounds

Exocentric compounding is recorded in the languages shown in Table 3.20.

Table 3.20. Exocentric compounding in the study sample

Compounds have traditionally been divided into endocentric and exocentric. The former are characterized by the binary structure of determinant–determinatum with the compound being a hyponym of its determinatum (head). The latter are said to have no head constituent (cf., among others, Scalise and Bisetto Reference Scalise, Bisetto, Lieber and Štekauer2009) or zero determinatum (i.e. one lying outside the compound; Marchand Reference Marchand1960: 11) and, therefore, the compound cannot be a hyponym of the determinatum, e.g. English paleface, redskin, etc.

This classification has raised much debate. In this respect, the position of this book is one that these compounds are generated in the same way as endocentric compounds. Two arguments can be raised:

1. (a) the psychological reasons can be found in both classical structuralist and onomasiological approaches. Marchand (Reference Marchand1960: 11) points out the general tendency of speakers ‘to see a thing identical with another already existing and at the same time different from it’. This principle, which Kastovsky (Reference Kastovsky1982: 152) calls the ‘identification-specification scheme’ is a key to one of the fundamental principles of Marchand’s and Kastovsky’s theories based on the binary, syntagmatic structure of motivated words: each word-formation syntagma is based on the determinant–determinatum relation, where the latter identifies and the former specifies (cf. 1.1.2.). The same principle underlies the onomasiological conception²⁵ and Natural Morphology, although in a different way in the latter: the most natural coinages are the most diagrammatic (a new meaning is accompanied with a new form), e.g. read-er, where there is ‘a diagrammatic analogy between semantic and morphotactic compositionality (or transparency)’ (Dressler, Mayerthaler, Panagl and Wurzel Reference Dressler, Mayerthaler, Panagl and Wurzel1987: 102), and

2. (b) there is no reason to surmise that there is any other cognitive process underlying a small group of exocentric compounds deviating from the identification–specification scheme, because this way of conceptual analysis is the essence of naming in general. Štekauer (Reference Štekauer1998) explains exocentric compounds as a two-step process in which only the first step has word-formation relevance. It consists in the formation of an auxiliary, onomasiologically complete (i.e. with both the base and the mark included), compound complex word. The second step is based on shortening, which is sometimes not considered to be a word-formation process. By implication, this type of complex word can be analyzed on a par with the underlying full, auxiliary, version, although the latter has not come to be used (institutionalized). Hence, redskin can be analyzed as ‘redskin + person’, sabretooth as ‘sabretooth + tiger’, garde-manger as ‘garde-manger + place’ and killjoy as ‘killjoy + person’.

   Ten Hacken (Reference Ten Hacken, Booij, Lehmann and Mugdan2000: 358) maintains that ‘in view of their problematic nature, it is not surprising that it has sometimes been proposed that exocentric compounds are irregular and unproductive’. Contrary to this assumption, Ten Hacken (Reference Ten Hacken, Booij, Lehmann and Mugdan2000: 358) maintains that the English possessive adjective + noun type of exocentric compound does ‘not constitute a closed class and new coinages can be interpreted on a regular basis’ and the verb + noun type ‘is productive in Romance languages’.

The data used here bear witness to:

1. (a) a relatively strong position of figurativeness in natural languages, and

2. (b) the universal tendency to a speaker-friendly (but not listener-friendly) economy of expression which at all language levels struggles with the listener-friendly and not speaker-friendly tendency towards clarity of expression.

The relative popularity of this word-formation process means that, in this particular case, the tendency towards the economy of expression overpowers the opposite trend towards the clarity of expression, even if the cumulative meaning of the constituents of many exocentric compounds is indicative of their final meaning.

Our questionnaire examined the occurrence of two different types of exocentric compounds:

1. (a) the redskin type (meaning a ‘person with a red skin’ or a ‘potato with a red skin’), in which the expressed constituents identify the feature of the unexpressed head (base, determinatum), and

2. (b) the garde-manger type (lit. ‘keep food’, i.e. ‘pantry’), where a verb and its object are used to denote an entity with a new meaning.

Exocentric compounding of the redskin type is recorded in all the languages listed in Table 3.20 except Maipure and Swahili. Exocentric compounding of the garde-manger type is recorded in the languages shown in Table 3.21.

Table 3.21. Exocentric compounding of the garde-manger type in the study sample

3.1.1.5.1 Formal characteristics

Chung (Reference Chung1994) discusses verb + noun compounds, i.e. exocentric compounds of the garde-manger type, in Chinese, French and Spanish. She argues that Spanish has two predictable and invariable forms of this type of compound: the third person singular indicative form of a verb plus a plural noun (156) and the second element is not in plural. Chung, however, notes that some exocentric compounds of this type have an alternate form with a pluralized nominal constituent:

(156)

(156)

1. a.

   Spanish

   limpiadientes

   cleans-teeth

   ‘toothpick.ins’

   (Chung Reference Chung1994: 4)

   French

   couvre-lit

   covers-bed

   ‘bedspread’

   (Chung Reference Chung1994: 10)

2. b.

   Spanish

   lavaplatos

   washes-dishes

   ‘dishwasher.ag’

   (Chung Reference Chung1994: 7)

   French

   garde-malade

   watches-sick.person

   ‘nurse.n’

   (Chung Reference Chung1994: 10)

3. c.

   Spanish

   saltamontes

   jumps.over-hills

   ‘grasshopper’²⁶

   (Chung Reference Chung1994: 7)

In either case, the principles postulated by level-ordering models are violated, because inflection precedes the word-formation process of compounding. According to Spencer (Reference Spencer, Štekauer and Lieber2005: 74), this is frequent in Romance languages:

(157)

(157)

Catalan

eixugamà

dries-hand

‘towel’

(158)

(158)

Italian

porta-lettere

carries-letters

‘postman, mailman’

(159)

(159)

Portuguese

guarda-roupa

keeps-clothings

‘piece of furniture used to keep the clothes’

(160)

(160)

Romanian

pierde-vară

loses-summer

‘dawdler, slowcoach’

While the number of exocentric compounds of the garde-manger type is much smaller in Mandarin Chinese than in French and Spanish, their semantics map the three basic groups identified for the latter two languages (agents,²⁷ instruments and animals/plants), even if such compounds in Chinese are ‘often semantically more opaque – even to a native speaker – than in e.g. Spanish and French’ (Chung Reference Chung1994: 14) (see Table 3.22).

Table 3.22. Agents, instruments and animals/plants exocentric compounding of the garde-manger type in Mandarin Chinese

According to Berman (pers. comm.), verb + object compounds are common in Hebrew, but restricted to verbs in the so-called benoni form, which functions as a participial and as a present tense verb and is often lexicalized as a noun (either agent or instrument). However, it is more usual for verbs to be in an exclusively nominal pattern.²⁹

Interestingly, as noted by Chung, some Spanish exocentric compounds of this type can become a part of a new compound of the same type:

(167)

(167)

Spanish

portaparaguas

(porta + [par(a) + aguas])

holds-stops-waters

‘umbrella stand’

(Chung Reference Chung1994: 9)

Exocentric compounds are quite common in Indian languages like Malayalam (here, only of the garde-manger type) (168) and Marathi (169):

(168)

(168)

Malayalam

kuttuvaakkə

kuttuka-vaakkə

pierce.v-word

‘taunt.v’

(Asher and Kumari Reference Asher and Kumari1997: 400)

(169)

(169)

Marathi

िपक न

pikale paana

matured-leaf

‘a person likely to die due to age’

Comrie (Reference Comrie1989: 26ff.) maintains that the morphology of a language reflects its syntax. Chung’s analysis confirms this assumption in terms of the order of the constituents of the examined compounds, i.e. the verb is followed by the noun in the SVO type of the languages examined. By contrast, in Burmese, a Sino-Tibetan language spoken in South East Asia and Oceania, and Persian, an Indo-European language spoken in Eurasia, the order of constituents in compounds of this type is reversed, which reflects the Burmese (170) and Persian (171) SOV syntax:

(170)

(170)

Burmese

htamin-hce

rice-cook.v

‘cook.v’

(Chung Reference Chung1994: 23)

(171)

(171)

Persian

jibbor

pocket-take.away

‘pickpocket’

(Chung Reference Chung1994: 23)

Table 3.23 reviews exocentric compounds of the redskin and garde-manger types.

Table 3.23. Exocentric compounding of the redskin and garde-manger type

3.1.1.5.2 Semantic characteristics

The languages sampled indicate that the prototypical semantics of exocentric compounds encompasses first, human beings, animals and plants whose explicit part (i.e. surface name) refers to the characteristic quality and, second, instruments of action. This is in accordance with Chung’s observation (1994) that compounds of the garde-manger type usually denote agents (specialized professions), instruments and animals and plants as in Table 3.23. It follows from her analysis that many exocentric compounds of this type have isomorphic equivalents in other languages. This does not mean that other meanings might not occur (see Table 3.24).

Table 3.24. Semantic diversity in exocentric compounding

3.1.1.6.1 Formal characteristics

Certain formal differences may occur between coordinative and other compound types, as in Malayalam. As pointed out by Fabb (Reference Fabb, Spencer and Zwicky1998: 67), coordinative compounds may have properties by which they differ from other types of compound, e.g. they do not accept gemination.

While dvandva compounds admit, theoretically, the reordering of their members without violating the meaning, the sequence of constituents is practically fixed. The fixed order is, according to Kageyama (Reference Kageyama1982: 236), determined by various linguistic, social and cultural factors.³³

3.1.1.6.2 Semantic characteristics

Semantically, not all coordinative compounds are equal in nature, with dvandva compounds representing a subclass of coordinative compounds. In spite of this, Scalise and Bisetto (Reference Scalise, Bisetto, Lieber and Štekauer2009: 36) point out that the Sanskrit term dvandva is often used inappropriately for the whole class of coordinative compounds. L. Bauer’s (Reference Bauer, Lieber and Štekauer2009b: 351–2) analysis of coordinative compounds distinguishes the following types (cf. also L. Bauer Reference Bauer2008):

1. (a) translative, e.g. Paris-Rome [flight],

2. (b) co-participant, e.g. Russian-Turkish [war],

3. (c) appositional, e.g. owner-director,

4. (d) compromise, e.g. blue-green,

5. (e) generalizing, e.g. Mordvin, a Uralic language spoken in Eurasia, t’ese-toso (lit. ‘here there’, i.e. ‘everywhere’), and

6. (f) dvandva, which may be subclassified as additive, co-hyponymic, co-synonymic, approximate, and endocentric.

A different position is represented by Wälchli (Reference Wälchli2005), who takes inherence-based ‘natural coordination’³⁴ as the criterion of co-compoundhood, and disregards copulative compounds which violate this principle. As a result, Wälchli eliminates from the scope of co-compoundhood types (a) to (d) of L. Bauer’s classification. The analysis of co-compounds presented here is based on the approach represented by Bauer.

Thus, the meaning of coordinative compounds can be completely based on the principle of compositionality or the motivating constituents may develop a new quality. A source of the latter is Hindi, because the meaning of a number of coordinative compounds in Hindi is more than the mere sum of the meanings of their motivating words: they constitute a new conceptual meaning both in nominal and in adjectival compounds (see Table 3.27).

Table 3.27. Non-compositional nominal and adjectival compounding in Hindi (examples by Kachru Reference Kachru2006: 119–20)

A similar technique can be found in other languages too:

As noted by Kachru (Reference Kachru2006: 120) ‘a number of such compounds have one item from Indo-Aryan and one from Perso-Arabic source, both with identical meaning’:

Conversely, the second member of some compounds may have an opposite meaning to that of the first constituent:

Examples of coordinative compounds based on the compositional principle and of those which are not are given in Table 3.28.

Table 3.28. Semantic diversity in coordinative compounding

Asher and Kumari (Reference Asher and Kumari1997: 399) maintain that the compositional principle may be reflected in the attachment of the plural morpheme, as in Malayalam, where the coordinative compounds with the feature [+ human] are always in plural (suffix -maar). If the feature is [– human], the plural suffix -kaì is used:

3.2.1.1 Formal characteristics

Roberts (Reference Roberts and Dutton1991) notes that reduplications like those in (314) and (315) have a reduplicated stem:

This indicates the repeated action (iterative aspect), and then a vowel alternation which indicates the haphazard manner of the repeated action. Without vowel alternation in the reduplicated formant, the meaning is ‘repeated action’, whereas with vowel alternation, the meaning is ‘repeated action + irregular motion’. Thus, a process of vowel alternation indicates a grammatical category of ‘irregular motion’. The meaning of this ‘irregular motion’ category will be different for the context of the action expressed by each verb. The form of the alternation is different depending on the operative vowel in the verb stem. According to Roberts, eight possibilities have been observed (see Table 3.39).

Table 3.39. Vowel alternation in reduplication in Amele

Some languages, like Amele (Roberts Reference Roberts and Dutton1991) or Indonesian, make use of both complete and partial reduplication. The following are some examples of complete reduplication in Indonesian and Karao:

In Amele, complete reduplication can be divided into whole-word reduplication (including inflectional morpheme) and whole-stem reduplication. While the former applies to all major word-classes, the latter only applies to verbs. Examples are given in Table 3.40.

Table 3.40. Whole-word and whole-stem reduplication in Amele by word-class

By contrast, partial reduplication in Amele is primarily leftward from the base form, probably as a compensating device for the absence of prefixation, because regular inflection is by suffixation (Roberts Reference Roberts and Dutton1991: 120).⁵⁹ Roberts demonstrates that leftward reduplication usually copies the first CV of the base form (324) and (325), but V and VC types also exist (326):

Partial infixing reduplication is also present in Amele:

Example (328) illustrates partial preposing reduplication in Karao:

Specific cases can be found concerning reduplication, e.g. reduplication in Nelemwa can also be accompanied by stem modification:

Reduplication is most frequently used with verbs and nouns, but other word-classes are also possible, mainly adjectives. The word-formation reduplication of adverbs, pronouns and numerals is relatively rare, in descending order. Reduplication is productive in Amele (Roberts, pers. comm.) and in Japanese (Kimenyi Reference Kimenyi2008). In the latter it is especially so with stem reduplication of words that consist of two syllables. This complies with the cross-linguistic preference for a disyllabic foot melodic template as suggested in Mutaka and Hyman (Reference Mutaka and Hyman1990). All types of words (verbs, nouns, adjectives, numerals, adverbs, pronouns, prepositions) can occur in the reduplicated form. Examples in (330) and (331) illustrate disyllabic structures and examples in (332) and (333) illustrate double reduplication in Japanese in which words with two independent stems reduplicate them both:

As illustrated by Kimenyi, the Rendaku rule, the voicing of the consonant of the first syllable in Japanese compounds, also applies in reduplication:

Some languages, like Pipil, make use of partial reduplication only. Apart from using reduplication for the expression of plurality, it is frequently used with verbs. As stated by Campbell (Reference Campbell1985: 80), reduplication in verbs can be of two types:

both reduplicate the initial consonant(s) and the first vowel of the root, or only the vowel if no consonant occurs. In both cases the vowel is short. They differ in that one has only this form (i.e. CV-) while the other takes an additional h (i.e. CVh-). The latter is quite productive . . . The former (CV-) is not productive, but many examples of it exist, typically with -ka, or -tsa verbs . . . It means that plural objects or repetition of the action are involved.

Relevant examples are:

In addition, reduplication is also used with the so-called diffusion verbs⁶² to express repetition. If the verb is intransitive, it ends in -ka and, if it is transitive, it ends in -tsa (see Table 3.41).

Table 3.41. Intransitive and transitive verbs with reduplication in Pipil

Kwakw’ala has formally a highly differentiated system of complete and partial reduplication, where partial preposing reduplication includes a number of cases (Boas Reference Boas, Boas Yampolsky and Harris1947: 220).⁶³ Hindi can completely reduplicate nouns (341), adjectives (342), participles (343) and adverbs (344) and participles admit partial preposing reduplication (Kachru Reference Kachru2006: 122):

Ambiguous cases can also be found concerning the existence or not of reduplication in a given language. According to Schiffman (1996) in Tamil, as in other South Asian languages, there is a kind of reduplication process which consists in taking a lexical item and following it with the same item reduplicated, except that the first consonant and vowel are replaced by the CV sequence. The general meaning of this construction is ‘(item) and other things like it’:

Schiffman also gives examples of the distributive function of reduplication aimed at specifying different kinds of things, at linking different things in a certain relationship or at distributing qualities among various members of a set of things. In this case, reduplication is complete (the final vowel of the first constituent and the initial vowel of the second constituent merge):

Another ambiguous case is that of Dangaléat, where the separation inflection vs derivation is involved. According to Shay (pers. comm.), there is some evidence of derivational reduplication in verb stems of Dangaléat with the structure C1VC2C1VC2. These represent rightward reduplication of the root consonants and insertion of a high vowel, which matches the final consonant for the feature [+ round]. The newly formed syllable has the same tone as the original syllable. Reduplicated verb stems show evidence of derivation from both Dangaléat and Arabic:

In Shay’s view, such reduplication may once have been inflectional, e.g. to encode plurality or intensity of an event.

A final case of ambiguity of reduplication involves the distinction between compounding and reduplication. Thái Ân (pers. comm.) notes that sometimes there is no clear distinction between compounding and reduplication in Vietnamese. An example of indisputable reduplication is given in (351):

Finally, reduplication may be recursive. The literature cites several cases: Harrison (Reference Harrison1973: 426),⁶⁶ Moravcsik (Reference Moravcsik and Greenberg1978: 301)⁶⁷ and Wiltshire and Marantz (Reference Wiltshire, Marantz, Booij, Lehmann and Mugdan2000: 559)⁶⁸ speak of ‘triplication’ and Rose (Reference Rose2003: 114) cites examples of Tigré, an Afro-Asiatic language spoken in Africa.⁶⁹ That these cases are not rare also follows from Moravcsik’s (1978: 312) observation of ‘instances of multiple reduplication in many languages and possibly in all’, even if this claim seems to be rather strong. The existence of recursive reduplication has been expressly denied for several languages (e.g. Ilocano, Karao or Telugu). In Hausa, triplication is not common according to Newman (pers. comm.), although it does occur with ideophones. As far as the sample studied here is concerned, examples of recursive reduplication may be found in Slovak diminutives, as in (248).

As to the distribution of the reduplicated material, as maintained by Wiltshire and Marantz (Reference Wiltshire, Marantz, Booij, Lehmann and Mugdan2000: 560), ‘the position of the material in the base that is copied in reduplication may vary . . . material copied from base-initial position may appear in prefix, suffix or infix position. Base-final material may be copied by prefixes, suffixes, or infixes as well.’ In the case of partial reduplication, the most frequent position of the reduplicated material is at the beginning of a base (Rubino Reference Rubino and Haspelmath2005a: 114) (see Table 3.42).

Table 3.42. Some distributional patterns of reduplicated material

One of the formal features of partial reduplication recorded is that the stem material is copied on the side from which it is taken, i.e. prefixes tend to copy stem-initial material, suffixes stem-final material and infixes stem-internal material. This seems to relate to Marantz’s account of reduplication as an affixation process. He argues that the reason why all the diverse morphological processes labelled as reduplication are subsumed under this general label ‘is the resemblance of the added material to the stem being reduplicated’ (Marantz Reference Marantz1982: 436). In his view, reduplication is the affixation of a CV skeleton (a reduplicating morpheme) of an autosegmental tier, e.g. the phonemic melody, from the stem to which the reduplicating morpheme affixes.

Apart from the absolute linear position of the reduplicated material, segmental vowelhood and consonantality are crucial for the formal description of reduplications (Moravcsik Reference Moravcsik and Greenberg1978: 305, 307). Table 3.43 shows the most frequent structures of reduplicated material in our sample.

Table 3.43. Some consonant/vowel patterns of reduplication

A final case concerning formal characteristics is that of echo compounding.

According to Kachru (Reference Kachru2006: 128), echo compounding is characteristic of South Asian languages. It is based on the principle of copying the first constituent except for the first consonant. In Hindi, the first consonant of the echo constituent is always v-. This means that, whatever the initial consonant of the first constituent is, it is changed to v- in the echo constituent. The meaning of the echo constituent (which never occurs on its own) is ‘and the like’:⁷²

Similar examples can be found in other languages:

3.2.1.2 Semantic characteristics

Reduplication has received considerable attention for its importance in non-European languages. Moravcsik’s (1978: 316) review of reduplication in a number of languages finds out that the meanings associated with reduplication strikingly recur across languages. What follows agrees with this statement, but also disagrees with her view that ‘there is no a priori reason why reduplication . . . should serve as the expression of some meanings rather than as that of others’ (Moravcsik Reference Moravcsik and Greenberg1978: 316).

The fact that the most common meanings of reduplicated forms are intensification and iterativity suggests that the role of iconicity, in particular diagrammaticity, is crucial. This is in accordance with the assumptions of Natural Morphology: extended form is accompanied by semantic enlargement, in other words, the reduplication of a particular form is an indicator of a growing quantity of items, actions or quality in general, as in the numeral ‘two’ in Kalkatungu, where doubling the form doubles the meaning:

Thus, while Regier’s (1994: 3–4) radial category model identifies repetition as the core semantic concept, the core semantic concept may be specified more generally, in particular, as increased quantity (of various kinds), with a range of manifestations mentioned by Moravcsik (Reference Moravcsik and Greenberg1978), Nomura and Kiyomi (1993)⁷⁴ and Regier (Reference Regier1994), among others. However, the same process of reduplication may have different and even contradictory semantic effects in the same language with two different word-classes: in Hausa, many adverbs comply with the core meaning of increased quantity when reduplicated (400), but adjectives generally run counter to this core meaning and reduplication leads to reduced quantity (401):⁷⁵

This is not an isolated case of the violation of the iconicity principle. In fact, the quantity-raising meanings of reduplication, like intensity, iterativity, continuity or augmentativity have their counterparts in the quantity-reducing meanings of approximation, attenuation, diminutiveness, distribution, hypocoristics, etc. Thus, one and the same process can have two opposite effects in terms of iconicity, and this has attracted considerable interest.⁷⁶

Mattes (2006)⁷⁷ discusses full reduplication in Bikol, an Austronesian language spoken in South East Asia and Oceania which can produce both augmentative and diminutive meanings of the same form, and presents a different view: these are different realizations of the underlying concept of the change of quantity, with both types of meaning being iconic.

The languages studied here illustrate a wide range of meanings and shades of meaning that can be expressed by reduplication. Among these, the most frequent variants of the basic semantic concept of increased quantity are iterativity (in compliance with Regier Reference Regier1994) and intensification, i.e. the meanings which correspond to the basic natural iconic function of reduplication (see Table 3.44).

Table 3.44. Semantic diversity in reduplication

Among the semantic features of reduplication, the basic semantic concept expressed by reduplication in word-formation apparently is the natural concept of increased quantity, mainly represented as iterativity and intensification of action. This overview of various categories of meaning and diverse shades of meaning indicates the extreme semantic capacity of reduplication as a word-formation process which, outside Indo-European languages, approaches the capacity of affixation processes. In languages like Amele,⁸⁶ in which the semantic changes expressed by reduplication may be accompanied by word-class change, the possibility to distinguish between class-changing and class-maintaining reduplication strengthens the analogy to affixation processes.

4.1.1.1 Recursiveness in affixation

4.1.1.1.1 Recursive suffixation

Recursive suffixation is recorded in the languages shown in Table 4.3.

Table 4.3. Recursive suffixation in the study sample

Recursiveness varies widely from language to language. Slavic and Germanic languages offer a wealth of examples, Tibetan less so, and Slavey shows constraints, in that recursiveness is restricted to possessed augmentatives and diminutives. In some other languages, recursiveness depends on the standpoint taken. In Gã the only possible case of co-occurrence of two suffixes is the combination of the reduplication and suffixation processes. In particular, verbs that have the iterative suffix -mO may reduplicate to show distributive event, often with the additional suffix -i:

(15)

(15)

Gã

tsO-mO-tsO-mO-i

turn-over-turn-over-sfx

‘turn over and over, many things in many places’

The number of permissible suffixes in a word also varies cross-linguistically. Hardman (Reference Hardman2000: 51) characterizes Jaqaru in terms of recursiveness as follows: ‘the processes of nominalization and verbalization are recursive. It is not uncommon for nouns to be verbalized and then renominalized, or vice versa.’ Apart from simpler cases like (16) and (17), there are also much more complex recursive derivations as in (18) and (19)(see Table 4.4).

Table 4.4 Recursive suffixation in Jaqaru

Jaqaru is also interesting for the capacity of some of its suffixes to recur on the same verb root:⁶

(20)

(20)

Jaqaru

yanh-shi-rqay-ishi

yanh-ishi-rqaya-ishi

help-mut-everyone-mut

‘help each other’

Fortescue (Reference Fortescue1980: 261) maintains that there can be up to a dozen affixal morphemes in a West Greenlandic verb form and up to four of the same category. In his view, there are over 400 suffixes in West Greenlandic, ranging from fully productive to fully lexicalized. As in Jaqaru, these can be used recursively to build up complex verbs and nouns, with possible switches back and forth between verbal and nominal base several times within a single word:⁷

(21)

(21)

West Greenlandic

allattu-i-vvi-ssaaliqi-sar-sima-qa-anga

write.down-aps-loc.of-lack-fre-pfv-int-

1sg.ind

‘I was really short of notebooks’

Unlike English, derivational suffixes in West Greenlandic may reverse their relative order, the result of which is a change of meaning, as in the following example (Fortescue Reference Fortescue1984: 313):

(22)

(22)

West Greenlandic

1. a. Urnik-kusun-niqar-puq

   come.to-want-pas-3sg.ind

   ‘Somebody wanted to come to him’

2. b. Urnin-niqa-rusup-puq

   come.to-pas-want-3sg.ind

   ‘He wanted somebody to come to him’

The relative freedom of suffixes in Kwakw’ala is also a source of a specific sort of recursiveness, because it permits alternate orders of suffixes corresponding to distinct meanings. Anderson (Reference Anderson and Shopen1985: 33) gives an example of the suffixes -amas ‘cause’ and -exsd ‘want’. From the verb ne’nak^w ‘go home’, we can form ne’nak^w’exsd ‘want to go home’. By attaching the suffix -amas we obtain ne’nak^w’exsdamas ‘cause to want to go home’. On the other hand, from q’aq’oλa ‘learn’ we can make q’aq’oλaamas ‘cause to learn, teach’ and, from this, q’aq’oλamadzexsd ‘want to teach’ can be formed in turn. In these examples, the same suffixes appear in opposite orders, corresponding to different meanings ‘cause to want’ vs ‘want to cause’. Consequently, new forms of arbitrary complexity can be produced (‘want to cause to want to cause to . . .’).

According to Cowan (Reference Cowan1969: 97), it is quite common for a radical stem in Tzotzil to be formed by a root followed by three derivational suffixes. Sometimes, a derivational affix is preceded by a perfective, referential,⁸ passive or subjunctive morpheme. Cowan considers these morphemes to be derivational, because they may be followed by other derivational suffixes:

(23)

(23)

Tzotzil vó-b-il-al < vó-b-il ‘roasted (things)’ ‘roasted (by someone)’

(24)

(24)

Tzotzil yáx-em-al < yáx-em ‘wound’ ‘wounded’

According to Volpe (pers. comm.), recursive suffixation is also very productive in Japanese, in particular, in cases like (25) and (26):

(25)

(25)

Japanese 言語学者 gengo-gaku-sha < gengo-gaku < gengo ‘linguist’ ‘linguistics’ ‘language’

(26)

(26)

Japanese 社会学者 shakai-gaku-sha < shakai-gaku < shakai ‘sociologist’ ‘sociology’ ‘society’

The dominating word-classes in recursive suffixation are nouns and verbs, but examples of adjectives are also possible. Table 4.5 gives examples of all three cases.

Table 4.5 Recursive suffixation in nouns, verbs and adjectives

Like simple suffixation, recursive suffixation can result in a change of word-class. This is fairly common, especially in the direction from verb to noun, as in (46) to (52), but is also possible in the direction from noun to verb (53) and from verb to adjective (54):

(46)

(46)

Afrikaans

skei-baar-heid

separate-able-ity

‘separability’

(47)

(47)

Diola-Fogny atεk-ɔr-a < -tεk hit-rcp-ag ‘boxer’ ‘hit.v’ (D. J. Sapir Reference Sapir1965: 56)

(48)

(48)

Indonesian

pakai-an-nya

wear-nmr-def

‘the clothes’

(49)

(49)

Maipure

sunua-ta-tí

sit-cau-nmr

‘chair’

(50)

(50)

Marathi

गल-मय-

mangala-maya-taa

holy-full.of/with-abn.mrk

‘holiness’

(51)

(51)

Telugu

prem-inc-aDam

love-vl.sfx-grn

‘the act of loving’

(52)

(52)

Tibetan

ཐ་ག་པ

thag-a-pa

weave-nmr-pef

‘weaver’

(53)

(53)

Kalkatungu

mimi-yan-ati

breast-having-become

‘getting breasts’

(54)

(54)

Belorussian

буд-аў-нiч-ы

budaynichy

build-v.sfx-adj.sfx-adj(m.sg.nom)

‘building.adj’

4.1.1.1.2 Recursive prefixation

Recursive prefixation is recorded in the languages shown in Table 4.6.

Table 4.6 Recursive prefixation in the study sample

While the identification of prefixes in terms of both form and meaning is usually easy (but cf. 4.1.1 above), the analysis is not always simple and it may influence the assumption of the (non-)existence of recursive prefixation. A case in point is the combination of prefixes mem- and per- in Indonesian:

(55)

(55)

Indonesian

memper-kenal-kann

tr.act-know-tr

‘introduce (somebody)’

Müller (pers. comm.) maintains that, although memper- in cases like (55) can be analyzed as consisting of two prefixes (mem- and per-), it does not follow the usual morphophonemic constraints of the prefix mem-(specifically, the [p] following mem- should elide but it does not). So, if memper- is analyzed as two prefixes, then Indonesian has recursive prefixation, otherwise it does not.

Table 4.7 provides examples for nouns, verbs as well as adjectives, showing that recursive prefixation is bound to the same word-class. While suffixal recursiveness is cross-linguistically most characteristic of nouns, prefixal recursiveness is most typical of verbs.

Table 4.7 Recursive prefixation in nouns, verbs and adjectives

Recursive prefixation, like simple prefixation, may have class-changing effects:

(74)

(74)

Anejom

awo-nev-edou

cau-qsn-way

‘how to do (something)’

(75)

(75)

French

dés-en-cadrer(un tableau)

neg-in-frame

‘take (a picture) out of its frame’

(76)

(76)

Lakhota

i-wá-naˋu

prx-prx-hear

‘radio’

(77)

(77)

Māori

kai-whaka-haere

ag-cau-go

‘organizer’

(78)

(78)

Romanian

ne-des-facut

prx-prx-touch

‘intact, untouched’

This analysis takes into consideration that in some polysynthetic languages it is very difficult to speak about word-classes at all:

(79)

(79)

Clallam

nəsx ^wčɬɬqčšɬšá

nə-sx ^w-čɬ-ɬ qčšɬšá

my-cau-adv-affect-fifty

‘Fifty of them got me’

(80)

(80)

Nelemwa

Hli pe-fa-k‚laxi I hli

3dua.rcp-cau-be.ashamed.of-3dua

‘They are making each other mutually ashamed’

Recursive derivational prefixation appears to occur in languages in which there is recursive derivational suffixation, with three exceptions in the study sample: Lakhota, Māori and Yoruba (the last of which does not use suffixation for word-formation processes) show recursive prefixation but not recursive suffixation. Recursive compounding also appears to occur in languages where recursive derivational suffixation occurs and, again, some languages show recursive compounding but not recursive suffixation: Amele and Movima.

4.1.1.2 Base modification in affixation

4.1.1.2.1 Word-formation base modification in suffixation

Base modification in suffixation is recorded in the languages shown in Table 4.8.

Table 4.8 Base modification in suffixation in the study sample

As in prefixation, the changes are mostly caused by assimilation, but other types of changes are also possible, as in Bardi, where suffixes with o cause vowel harmony, and -n initial suffixes cause trill deletion (rr + n > n)(Bowern, pers. comm.). A range of morphophonological stem modifications within the same language is illustrated by Estonian. Most common changes of the stem during the word-formation process of suffixation in Estonian are shown in Table 4.9.

Table 4.9 Base modification and suffixation in Estonian16

According to Laakso (pers. comm.), some suffixes in another Finno-Ugric language, Finnish, consist (in the traditional analysis) of vowels that replace the stem-final vowel by regular (morpho-)phonological processes, such as the labial vowel in action derivatives (historically going back to a *w suffix):

(86)

(86)

Finnish laul-u < laula- ‘song’ ‘sing.v’

(87)

(87)

Finnish näk-ö < näke- ‘sight’ ‘see.v’

Other suffixes may contain satellite vowels which replace the stem-final vowel:

(88)

(88)

Finnish

koivi-kko

koivu-kko

birch.tree.col

‘birch grove’

(89)

(89)

Finnish

pahe-ne

paha-ne

bad-trn.v

‘get worse’

A characteristic change in the base due to the change of intransitive verbs to transitive verbs by suffixation in Malayalam is consonant gemination. Asher and Kumari (Reference Asher and Kumari1997: 275–6) distinguish five types of gemination:¹⁷

1. (a) the result of doubling k > kk, ʈ > ʈʈ , r > rr: kayar- ‘climb’ vs kayarr- ‘cause to climb’,

2. (b) a change from double nasal or homorganic nasal + plosive to the corresponding double plosive: ŋŋ > ŋk, mp > pp: kuump- ‘fold-intr’ vs kuupp- ‘fold-tr’,

3. (c) a change from a lateral to a double plosive: l > ʈʈ, l > rr: cuzȥal- ‘rotate’ vs cuzȥarr- ‘cause to rotate’,

4. (d) the addition of a double plosive: Ø > tt: poʈʈ- ‘break-intr’ vs poʈʈikk- ‘break-tr’, and

5. (e) the replacement of one double plosive by another: kk > tt: nilkk- ‘stop-intr’ vs nirtt- ‘stop-tr’.

In principle, stem modifications may be divided into vowel modifications (Table 4.10) and consonant modifications (Table 4.11).

Table 4.10 Vowel modification and suffixation

Example (104) from Ilocano shows the influence of suffix in a prefixal-suffixal derivation:

(104)

(104)

Ilocano pastreken < pa-serrek-en cau-enter-tr ‘let in’

In (105), the suffix selects the Latin participle emiss- instead of the regular emesso:

(105)

(105)

Italian

emissione

‘emission’

The formation of diminutives is a frequent source of morphophonological change. Example (106) illustrates diminutive formation accompanied by vowel alternation:

(106)

(106)

Tibetan

rtevu

rta-vu

‘horse-dim’

A special variant of vowel change due to diminutive formation is Slovak diphthongization:

(107)

(107)

Slovak

žienka

žen-k-a

‘woman-dim-f.nom.sg’

The category of diminutiveness brings us at the same time to the area of stem consonant changes, because it is a general phenomenon in Slavic languages, as illustrated in Serbian-Croatian and Slovak, but also in some other languages, like Romanian, and in Zulu, where root-medial and final labial consonants become palatal in the diminutive (also in the locative and the passive):

(108)

(108)

Serbian-Croatian

ruč-ic-a

ruk-a

hand-dim-f.nom.sg

‘little hand, handle’

(109)

(109)

Slovak

rúč-k-a

ruka-dim-f.nom.sg

hand-dim

‘little hand’

(110)

(110)

Romanian cărucior < căruţă ‘perambulator’ ‘cart/waggon’

(111)

(111)

Zulu nkony-ana < n-komo ‘calf’ ‘cow/bull’

Stem consonant changes are not limited to the category of diminutiveness. Some other cases are shown in Table 4.11.

Table 4.11 Consonant modification and suffixation

When the Tzotzil suffix -ol ‘act of’ combines with intransitive roots of the shape CVC plus -in, the final n changes to m:

(120)

(120)

Tzotzil

táximol

táxin-ol

play-nmr

‘game, play’

(Cowan Reference Cowan1969: 107)

The stem-final /n/ may change to /m/ also in Udihe and the causative suffix changes to -uAn-:

(121)

(121)

Udihe

lagbam-uan-

stick-cau

‘glue (something)’

(Nikolaeva and Tolskaya Reference Nikolaeva and Tolskaya2001: 301)

With certain verbs (Class II), the stem-final /n/ in Udihe merges with suffix-initial /g/ of the repetitive suffix -gi-, which results in -ŋi-:

(122)

(122)

Udihe

ilaktan-ŋi-

ilaktan-gi

appear-ite

‘appear again’

(Nikolaeva and Tolskaya Reference Nikolaeva and Tolskaya2001: 317)

As in Nelemwa prefixation and in Nelemwa suffixation, stem modification mostly concerns phonological lenition, in this case with transitive (123) and with possessive suffixes on nouns (traces of proto-stems)(124), while another example illustrates extension of stem due to suffixation (125):

(123)

(123)

Nelemwa

cawi

cap-I

escape-tr

‘escape from (something)’

(124)

(124)

Nelemwa

jixela-ny

jixet-pos.1sg

rifle-pos.1sg

‘my rifle’

(125)

(125)

Nelemwa

haroon

aroo-n

husband-pos.3sg

‘marry.v’

As in English, suffixes in Spanish and Totonac may cause stress shift. In the latter case, the nominalizing suffix shifts the stress leftwards:

(126)

(126)

Spanish

horroroso /oroɻóso/

horror-oso /oróɻ/

horror-sfx.adj

‘horrific’

(127)

(127)

Totonac liné’he’ < li:ne’hé: li:-ne’hé:-h li:-ne’hé: ins-fan-nmr ‘fan’ ‘fan with (something)’

Finally, Romanian shows a combination of vowel and consonant alternations ia ~ ie/ t ~ ţ:

(128)

(128)

Romanian

băiat – băieţel – băieţaş – băieţandru – băieţoi

‘boy – ‘urchin’ – ‘a slip of a boy’ – ‘youngster’ –

‘tomboy, hoyden’

4.1.1.2.2 Word-formation base modification in prefixation

Base modification in prefixation is recorded in the languages shown in Table 4.12.

Table 4.12 Base modification in prefixation in the study sample

While prefixation with base modification is reportedly rare in Catalan, English, Hebrew and Russian, it is frequent in other languages, like Datooga, and occurs systematically in Breton, because it participates in the system of mutations (Stump, pers. comm.):

(129)

(129)

Breton

digoulouma~n

di-koulouma~n

‘un-tie’

In Afrikaans, the only change in the base is orthographic: capitalization disappears in the names of languages after prefixation:

(130)

(130)

Afrikaans

ver-afrikaans

vr-Afrikaans

‘make Afrikaans’

According to Carstens (pers. comm.), any other changes in the base are not a result of the prefix with which it combines, but a coincidental occurrence where fossilized ablaut has become associated with the use of a prefix in a particular complex word:

(131)

(131)

Afrikaans ver-swe-e < swyg vr-silent-attr ‘tacit’ ‘remain silent’

The changes recorded are of diverse nature and seem to be highly language-specific, often even affix-specific. Müller (pers. comm.) remarks that, although some processes may exist and indeed be very productive, they may be so only for a very small number of affixes. Müller illustrates this observation with an example from Indonesian, where morphophonemic alternation of the stem when adding a prefix occurs productively with the various allomorphs of the prefixes peng- and meng-, but not with any other Indonesian prefixes:

(132)

(132)

Indonesian

mem-egang

meng-pegang

tr-hold

‘take hold of’

In Slavey, the changes bear on the voicing alternations of stem-initial fricatives determined by the immediately preceding prefix. The so-called D-effect combines a prefix d- with a stem-initial fricative or glottal stop.

As noted by Brainard (pers. comm.), Karao displays a set of complex morphophonological alternations in connection with prefixes and infixes. The processes are governed by the canonical shape of the affix and the root and may affect both the affix and the root. If the canonical shapes of the affix and root are eligible to undergo a particular process, then the process will automatically apply.²⁸ Thus, when a prefix ending in N attaches to a root beginning with a CV (C) syllable, N assimilates to the same point of articulation as the root-initial consonant, and the root-initial consonant is then deleted. When e- attaches to a root beginning with a CeCV (C) root, the prefix changes to iya-, e is deleted and the root-initial consonant changes to its syllable-final allomorph (if it is different):

(133)

(133)

Karao

ena-cha

eN-cha-cha

ag-help

‘help.v’

(134)

(134)

Karao

iyalpek

e-depek

pas.real-wet

‘be wet’

In Bardi, all (C)V- prefixes cause changes in obstruent-initial roots and there is also a set of complex cluster reduction rules (Bowern, pers. comm.):

(135)

(135)

Bardi

ngaambala

nga-jambala

‘1sg-foot’²⁹

In Zulu, the root -mb- ‘dig’ becomes -emb- after a prefix ending in -a:

(136)

(136)

Zulu

basemba

ba-sa-mb-a

‘They are still digging’

The vowel -a of the prefix falls away before e- and so (as assumed by van der Spuy, pers. comm.) this situation could be alternatively interpreted as the root changing a preceding -a to -e. However, there are only about five roots in the language that display this phenomenon. Other changes are shown in Table 4.13.

Table 4.13 Prefixation and base modification

4.1.3.1 Suffixation

The most frequent reason for the existence of suffix variants is phonological conditioning (assimilation). Thus, in Estonian a suffix beginning with a vowel cannot be added to a vowel-final stem. Therefore, the stem is often shortened (230) and, if the stem is too short to be shortened, other variants of the suffix are used, such as -mus, -dus, -vus, -tus(Kilgi, pers. comm):

Another example of assimilation shows that, in Udihe, with certain verbs (Class II), the stem-final /n/ merges with the suffix-initial /g/ of the repetitive suffix -gi-, which results in -ŋi-(Nikolaeva and Tolskaya Reference Nikolaeva and Tolskaya2001: 301):

Further examples of this phenomenon are shown in Table 4.21. They illustrate a number of phonologically conditioned forms in Belorussian ((232) and (233)), Estonian ((239) and (240)) and Malayalam ((243) to (246)) or different realizations of suffixes according to what precedes and/or follows them (Luo (236) to (238)):

Table 4.21. Many-to-one relation in suffixation

Dissimilation is sometimes at play as well. Cowan (Reference Cowan1969: 99, 101) shows that the variants of the Tzotzil intransitivizing suffix -ub ‘developmental’ and the transitivizing suffix -Vn are controlled by the dissimilation principle: the former suffix appears as -ib if the stem contains a back vowel and as -ub after other vowels. But when the stem vowel is o, the suffix vowel is also o(247). The latter suffix occurs with i after back vowels of root and with u after others (248):

In Jaqaru, variants of suffixes result from systematic vowel-drop in combination with certain other suffixes following them. Hardman (Reference Hardman2000: 90) explains that ‘all suffixes with two vowels will drop at least one of the vowels in most constructions’. An example with the diminutive suffix -uña is given below:

Swahili provides an example of allomorphy due to the opposition of vocalic vs consonantal environment. In Contini-Morava’s examples, the variants illustrate post-vocalic allomorphy in the applicative suffix combined with the vowel-harmony principle:

This brings us to another source of suffix variants, such as vowel harmony, which is typical of agglutinative languages. The following examples come from Hungarian:

Pipil illustrates an influence of an affix upon the form of another affix. In particular, the short variant of the adjectival suffix -a:wa-k/-a-k occurs in combination with the following suffix. From the inchoative verb we get the adjective chipa:wa-k ‘clear’ where -wa is an inchoative suffix. If the suffix -nah is attached, the long suffix -a:wak is reduced to -ak(Campbell Reference Campbell1985: 62).

Another source of suffixal variants is the so-called Rhythmical Law. This law imposes a requirement upon Slovak words to avoid two long syllables next to each other. The operation of the law is illustrated for the agentive suffix variants -ík(long variant) in (254a) and short -ik(short variant) in (254b):

In Zulu, there is a co-existence of a productive variant -kazi(phonologically identical with the augmentative) and a rare unproductive variant allomorph -azi:

Finally, Stump (pers. comm.) points out that several Breton suffixes participate in alternations of the form X ~ iX with lexical conditioning, though with some partial phonological regularities:

There are also unpredictable variants. Cowan (Reference Cowan1969: 98ff.) illustrates this possibility with one of eight intransitivizing suffixes used in Tzotzil. This suffix is based on the combination of a V (owel) and x, where V is realized unpredictably as a, e, i, o or u:

The last example, taken from Dutch, is of similar flavour. At the same time, it demonstrates how complex the relation between variants of a morpheme may be in word-formation. Don (pers. comm.) refers to De Haas and Trommelen (Reference de Haas and Trommelen1993: 298), for whom the distribution of ‘-lijk and -elijk is not fully complementary and hence also not fully predictable’.⁵⁷ De Haas and Trommelen discuss five tendencies which are summarized by Don as follows:

1. (a) after a stem-final syllable containing schwa, we get -lijk: open > openlijk, *openelijk ‘openly’,

2. (b) after a stem-final plosive we get -elijk: hoop > hopelijk ‘be hoped’,

3. (c) after a fricative there is preference for -elijk: stof > stoffelijk ‘material’(but also forms in -lijk are found after fricatives),

4. (d) after a stem-final long vowel followed by a nasal or liquid, we find -lijk: natuurlijk ‘natural’, and

5. (e) after a stem-final diphthong or glide, we usually get -elijk: vrouwelijk ‘female’.

4.1.3.2 Prefixation

As with suffixal allomorphy, the dominant reason for the existence of prefixal allomorphs is phonological conditioning. In Hausa, ma- often assimilates to /mu/ when the vowel in the following syllable is /u/(259). Some patronymic prefixes in Totonac also have variant forms, as in (260) and (261):

Phonological condition applies also in the transitive active verbal prefix me- in Indonesian (Mojdl Reference Mojdl2006: 46–8). It exists in five different variants depending on the stem-initial phoneme:

1. (a) the variant mem- is used before stem-initial -b-, -p- and -f-, as in mempakai ‘wear’,

2. (b) the variant men- is used before stem-initial -t-, -d-, -c- and -j-, as in mencari ‘look for, find’,

3. (c) the variant meng- is used before stem-initial -k-, -g-, -h- and before a vowel, as in mengolah ‘cheat.v’,

4. (d) the variant meny- is used before stem-initial -s- and -sy-(the initial -s- and -sy- is dropped), as in menyewa ‘rent.v’(from sewa ‘rent.n’), and

5. (e) the variant me- is used before -m-, -n-, -ng-, -ny-, -l-, -r-, -w- and -y-, as in melompat ‘jump.v’.

   Phonological conditioning occurs in other languages, too (see Table 4.22).

Table 4.22 Phonologically conditioned prefixes

The following example of Breton shows a different type of phonological conditioning in which, unlike the prototypical cases of phonological conditioning, there is no complementary distribution of the allomorphs:

Stump (pers. comm.) explains that the alternation in the prefix ken-/kev- is phonologically conditioned in a rather weak sense: ken- can appear with essentially any stem and kev- appears with stems beginning with oral sonorant sounds. Similarly, the alternation in the prefix ad-/as- is phonologically conditioned in a weak way: ad- can appear with essentially any stem and as- appears with stems beginning with voiceless obstruents. In neither instance is the distribution complementary: ken-, like kev-, can precede an oral sonorant and ad-, like as-, can precede a voiceless obstruent. In this way, as noted by Stump, they are reminiscent of im-/in-(with the meaning ‘in’) in English: im- precedes bilabial sounds (import, immigrate), while in- precedes any sort of sound, including bilabials (inmate, input).

This leads to free variants, as exemplified by the Afrikaans prefix dis-. The choice between the two variants is reportedly arbitrary:

Another important factor is stress. In Clallam, unstressed schwas tend to be deleted (Montler, pers. comm.):

Another different type of allomorphy, i.e. other than phonologically conditioned, is illustrated by Tzotzil: according to Cowan (Reference Cowan1969: 109), the allomorph ax- of the prefix x- occurs after a pronoun:

The following example from Zulu (van der Spuy, pers. comm.) shows a word-class-conditioned selection of allomorphs. Zulu adjectives and certain adverbial forms can be used as predicates. There, the prefix sa- ‘still’ is combined with verbs (278a) and se- with non-verbal predicates (278b):

A completely different condition for the existence of variants of a prefix is mentioned by Hudson (pers. comm.): the use of the causative prefix a-/as- in Amharic is determined by the category of transitivity such that a- is used for intransitives and as- for transitives (however, with many exceptions):

The relations between the asymmetric form-meaning prefixal relations and the influence of prefixation on the form of the word-formation base are very diverse (see Table 4.23).

Table 4.23 Prefixation vs suffixation: discrepancies between one-to-many and many-to-one relations