5.1.1 Conversion

The term conversion is connected, in the majority of English-written literature, with the prototypical case of English conversion as a process of forming a new word which belongs to a different word-class without any formal change. From the point of view of constructional iconicity, conversion is not a natural word-formation process: unlike compounding and affixation, the new meaning is not expressed by an additional form (Dressler Reference Dressler, Štekauer and Lieber2005: 269).

This definition of conversion is, however, tricky for the vagueness of notions like word-class and lack of formal change. As noted by L. Bauer and Valera (Reference Bauer, Valera, Bauer and Valera2005: 8), ‘virtually all of this has been questioned at one point or another and yet the concept of conversion remains in use, very much as the conventional system of word-classes does in languages for which it is theoretically inadequate’.¹ This remark shifts the focus to the issue of word-classes for cross-linguistic description. A case in point is Māori (W. Bauer, pers. comm.):

Therefore, while conversion may seem justified from the Indo-European linguistic perspective, it may not be so in other language families. The message of the motto of this chapter is in no way exceptional:² it can also be said of other Polynesian languages (W. Bauer Reference Bauer1997: 65)³ and is apparent from Spencer’s (Reference Spencer, Booij, Lehmann and Mugdan2000: 316) examples of inflected verbs used as nouns in Navajo, a Na-Dene language spoken in North America:

More importantly, this gives support to the view that lexical entries are neutral as regards word-classes (Farrell Reference Farrell2001). The implication for the theory of conversion is, as pointed out by L. Bauer and Valera (Reference Bauer, Valera, Bauer and Valera2005: 9), ‘that the relationship between nouns and verbs of related form (e.g. [English] a bridge and to bridge) is no more than a matter of inflection’. This view has found support in the literature (Myers Reference Myers1984; Josefsson Reference Josefsson1997; Giegerich Reference Giegerich1999), but it is admittedly rejected more often than not.

Conversion is recorded in the languages shown in Table 5.1 (61.82 per cent of the study sample).

Table 5.1 Conversion in the study sample

Conversion can be found in a number of languages: Serbian-Croatian dobro may be an adjective (‘good’), a noun (‘property’) and an adverb (‘well’). In Ket it is common for nouns, adjectives and sometimes verbal infinitives to have the same form (cf. (18)), and in Maipure, the same entry may also be a stative verb, an adjective and an adverb (Zamponi Reference Zamponi2003: 46) (cf. (20)). Similarly, in Amele abul-doc can mean ‘struggle.v’ as well as ‘struggle.n’ and ihan-ec can mean ‘sacrifice.v’ as well as ‘sacrifice.n’.⁴ In Hausa, simple adjectives have the same form as nouns, create feminines and plurals essentially like nouns and use the same genitive linker as nouns; in fact, many words exist in Hausa both as nouns and adjectives (Newman, pers. comm.).

Thái Ân (pers. comm.) explains the high productivity of conversion in Vietnamese by referring to Spencer (Reference Spencer, Orgun and Sells2004: 3), who maintains that ‘in so-called “isolating languages” it is common for a single word to have the syntax of a noun or a verb indiscriminately, but arguably we are better talking of categorical indeterminacy here rather than mixing’. This must be distinguished from the situation in Jaqaru, where there is a common pro-root which ‘functions as a stand-in for all other roots. The pro-root {inchi} may be a pro-noun or a pro-verb or simply a filler. It may carry any or all suffixes of nouns, verbs or sentence suffixes or it may stand alone as a particle’ (Hardman Reference Hardman2000: 8).⁵

Probably as a consequence of the different word-class systems that can be found in different languages, individual cases of conversion may feature various degrees of transfer to a new word-class. Nikolaeva and Tolskaya (Reference Nikolaeva and Tolskaya2001: 166–8) illustrate different degrees of nominalization of verbs in Udihe:

These converted nouns show some verbal properties: they preserve the valency of the corresponding verb and can be modified by an adverbial. The function of converted nouns is to fill the object valency of certain verbs.⁶ By contrast, other converted nouns, as (4) and (5), have undergone further nominalization and take all inflections typical of nouns:

Apart from the prototypical cases of non-homonymous conversion or, rather, conversion as it is understood in Indo-European languages, examples of other subtypes can also be found, e.g. intra-categorial conversion or secondary word-class conversion. In Swahili, the noun sauti means ‘voice’ if accompanied by Class 9 concord and ‘thick/harsh voice’ if accompanied by Class 5 concord.⁷ Examples (6) and (7) illustrate this process in other languages too:

Another borderline case can be found in Bardi, where, according to Bowern (pers. comm.), all adjectives and many nouns can be used as coverbs in complex predicates:

However, it seems better not to consider these and similar cases of conversion for their dependence on another verb, i.e. they cannot occur independently and, in this respect, they resemble clitics.

All these facts bring the description back to Hockett’s (Reference Hockett1958: 221) almost forgotten rejection of the traditional concept of word-classes and to the proposal of new categories like AV, NA, VN and NAV, depending on whether the respective lexeme functions both as an adjective and a verb, a noun and an adjective, etc.⁸ This book does not pursue this issue and limits itself to providing cross-linguistic evidence which might contribute to answering the question of the correctness or falsity of this direction of consideration.

In our sample, the majority of languages where conversion is recorded allow conversion within the categories adjective, noun, verb and, less markedly, adverb. No records of other word-classes have been cited. It should be emphasized, however, that the following analysis reflects the limited scope of data bound to a single item in our questionnaire examining the existence of a productive word-formation process of conversion.

5.1.1.1 Formal characteristics

Homonymous conversion is perhaps the most canonical view of conversion. It occurs in examples (9) to (27):

(9)

(9)

Amele ben > ben ‘big’ ‘big thing’

(10)

(10)

Catalan finlandès > finlandès ‘Finnish’ ‘Finn’

(11)

(11)

Clallam swqa > swqa ‘man’ ‘be a man’

(12)

(12)

Cirecire dau > dau ‘ash, burning’ ‘burn’

(13)

(13)

Dangaléat káté > káté ‘go.v’ ‘going.n’

(14)

(14)

Dutch feest > feest ‘party’ ‘have a party’

(15)

(15)

Hausa tsōhō > tsōhō ‘old’ ‘elder.n’

(16)

(16)

Ilocano aso > aso ‘dog’ ‘be a dog’

(17)

(17)

Karao Emahay daba. e-maha-i daba pass.real-dry-abs laundry ‘The laundry was dry’

(18)

(18)

Ket i’l > i’l > i’l ‘song’ ‘sing’ ‘singing.adj’

(19)

(19)

Luo pan > pan ‘curse.n’ ‘curse.v’

(20)

(20)

Maipure maisuíni > maisuíni > maisuíni ‘be bad’⁹ ‘bad.adj’ ‘badly.adr’ (Zamponi Reference Zamponi2003: 46)

(21)

(21)

Mandarin 熟悉 > 熟悉 Chinese shóuxī shóuxī ‘familiar’ ‘become familiar with’

(22)

(22)

Māori te pai > te pai ‘goodness’ ‘det.good’

(23)

(23)

Nelemwa uya > uya ‘arrive.v’ ‘arrival’

(24)

(24)

Romanian frumosul > frumosul ‘beautiful’ ‘beauty’

(25)

(25)

Serbian- mlada > mlada Croatian ‘young’ ‘bride’

(26)

(26)

Totonac páxni‘ > páxni‘ ‘washed’ ‘pig’

(27)

(27)

Zulu kahle > kahle ‘well’ ‘fine.adj’

As far as the identical form criterion is concerned, this rather general definition lends itself for analytic, non-inflecting languages, but can hardly be applied without modification to inflectional ones. It has been argued, however, that the form of the stem – rather than the form of the word – is important. In the majority of cases, it remains intact. Thus, the situation in Slavic languages shows that, from the formal point of view, this phenomenon is far from being as simple as it might seem when we limit our focus to English. Smirnickij and Achmanova (1952) and Smirnickij (Reference Smirnickij1953, Reference Smirnickij1954, Reference Smirnickij1956) point out that conversion in Russian is based on the change of paradigm. Smirnickij (1953: 24) maintains that conversion in Russian is one of the so-called morphological word-formation processes, where morphological is synonymous with inflectional paradigm. From this it follows that, while no derivational affix is added, formal changes occur. The paradigm thus fulfils the function of a derivational affix (cf. Dokulil Reference Dokulil and Isačenko1968: 218).

Consequently, unlike in English, where in the vast majority of cases conversion entails formal identity, and only a minor group of conversion pairs bear some formal change (re’cord.v vs ’record.n), conversion is not homogeneous cross-linguistically:

(28)

(28)

Czech dobr-ý.adj > dobr-o.n ‘good.adj’ ‘good.n’ (Dokulil Reference Dokulil and Isačenko1968: 221ff.)

(29)

(29)

Estonian julge.adj > julge-ma.v ‘brave’ ‘dare.v’

(30)

(30)

French vol(er).v > vol.n ‘fly.v’ ‘flight’

(31)

(31)

German schlafen.v > Schlaf.n ‘sleep.v’ ‘sleep.n’

(32)

(32)

Hindi चमक > चमक cəmək cməkna ‘shine.n’ ‘shine.v’ (Kachru Reference Kachru2006: 116)

(33)

(33)

Japanese 泳ぐ > 泳ぎ oyogu oyogi ‘swim.v’ ‘swimming.n’

(34)

(34)

Slavey shoN.n > joN.v (D¹⁰ + shoN) ‘old age’ ‘be old’

(35)

(35)

Spanish aceite.n > aceit-ar.v ‘oil.n’ ‘oil.v’

Another subtype can be illustrated by German schneiden ‘cut.v’ vs Schnitt ‘cut.n’, which shows that one of the stems need not be a citation form.¹¹ Such cases are not unique cross-linguistically. A relatively frequent source of conversion are participles. This poses serious theoretical problems as to the status of e.g. -ing participles/adjectives in English, which, in turn, comes back to the issue of the definition of conversion, especially as participles, both present as in (36), (37), and past (38), take part in conversion in a number of languages:

(36)

(36)

Hebrew מנהל vs להנמ menahel menahel nihel-prs.ptc ‘managed, organized’ ‘boss, director’

(37)

(37)

Romanian tremurândă vs tremurândă ‘shaking’ ‘shaky’

(38)

(38)

Romanian Plimbatul de dimineață e sănătos < plimbatul ‘The walk in the morning is healthy’ ‘walked’

In Udihe, resultative participles derived with the suffix -ktu are converted to adjectives that can be modified by degree adverbials, like c’o ‘most’ or belem ‘even more’, and can head a comparative adjective phrase (Nikolaeva and Tolskaya Reference Nikolaeva and Tolskaya2001: 197):

(39)

(39)

Udihe

koŋo-ktu

‘lean, thin’

(40)

(40)

Udihe

soŋo-ktu

‘crying, whining’

Similarly, passive is used in Karao (41) and past tense forms may be used in Ilocano (42):

(41)

(41)

Karao emahay daba e-maha-i daba pas.rea-dry-abs laundry ‘The laundry was dry’

(42)

(42)

Ilocano immay vs immay ‘came’ ‘one who came’

The status of imperative as a converting form in Telugu, as in (43), is ambiguous according to Pingali (pers. comm.):

(43)

(43)

Telugu చదవ vs చదవ caduwu caduwu ‘read.v’ ‘education’

For Pingali, verb roots in Telugu are bound forms and they become words with the affixation of at least a suffixal vowel. It is unclear how conversion should be captured as a process, whether the -u in derived caduwu should be seen as an epenthetic vowel (since words cannot end in consonants) or whether conversion is to be seen as converting the imperative form, which is caduwu with the imperative suffix -u.

Another formal variant of conversion is one which encompasses the addition of a stem-forming morpheme (theme), as illustrated by the following example from Marathi:

(44)

(44)

Marathi माेठ्याने माे ा moThyaane moThaa moth-yaa-ne elder.adj-saamaanyarup.sfx-by-erg¹² ‘elder.adj’ ‘by elder brother/person’

Czech examples are zelenat, zelenět, zelenit ‘be green’, for which Dokulil (1968: 225) postulates stem-forming grammatical morphemes, i.e. thematic morphemes (-a-, -e/ě-, -i-) rather than a derivational suffix.

Conversion may also be accompanied by vowel/consonant alternations, as in Czech and Hindi:

(45)

(45)

Czech silný > silně [n > ň] ‘strong’ ‘strongly’ (Dokulil Reference Dokulil and Isačenko1968: 231)

(46)

(46)

Hindi < mela milna ‘fair’ ‘meet.v’ (Kachru Reference Kachru2006: 116)¹³

Conversion may also be a part of a combined word-formation process and so, for Slovak, a range of conversion variants can be distinguished (see Table 5.2).¹⁴

Table 5.2. Conversion combined with other processes in Slovak

The productivity of conversion may differ both cross-linguistically and inside one particular language, if various conversion subtypes are taken into account. In Udihe ‘virtually every adjective may receive certain nominal properties within headless noun phrase’ (Nikolaeva and Tolskaya Reference Nikolaeva and Tolskaya2001: 169). In Hindi, the infinitive suffix -na yields an abstract noun and therefore, except for the invariable cahiye ‘should, ought to’, Hindi verbs have infinitive forms which can act as abstract nouns (Kachru Reference Kachru2006: 115). In addition, the root of the verb may be used in Hindi as an abstract noun. By contrast, conversion is reportedly rare in other languages, like Finnish, even if some examples can be found:¹⁵

(56)

(56)

Finnish tihku vs tihku ‘drizzle.v’ ‘drizzle.n’

(57)

(57)

Finnish kulu vs kulu ‘expend.v’ ‘expense.n’

In Finnish, Laakso (pers. comm.) notes that there are some cases where the stem-final vowel of a noun coincides with the one-vowel suffix of a verb. In Laakso’s view, from a synchronic perspective, these could be classified as cases of conversion because they share the stem paini- ‘wrestle.v’ and paini ‘wrestling’ and both ultimately go back to paina- ‘press.v’. Laakso notes that there is also a handful of noun–verb ambiguous stems (tuule-, nom.sg.tuuli ‘wind’, tuule- ‘blow.v (of the wind)’). As there is no clear morphosyntactic boundary between adjectives and nouns, this adjective–noun ambiguity might be considered a kind of conversion.

Given the above-mentioned difficulties concerning the limits of conversion, our cross-linguistic research examines conversion as the formation of new complex words by shift of categorial meaning (following Cetnarowska Reference Cetnarowska1993: 86) in the framework of the conventional system of word-classes, prototypically, without formal change in the stem.¹⁶

5.1.1.2 Semantic characteristics

Conversion offers a wide range of meanings which, despite the scepticism of Clark and Clark (1979), seem to be fairly well predictable for the individual novel converted complex words. Štekauer’s (2006) data show that, with the majority of this kind of neologisms, there is usually one, rarely two dominant meanings, much more predictable than the other potential readings of the coinage. Table 5.3 shows a hint of the wide semantic capacity of conversion as a word-formation process.

Table 5.3. Semantic diversity in conversion

5.1.2 Stress

Stress is recorded in the languages shown in Table 5.4 (7.27 per cent of the study sample).

Table 5.4. Change in stress in the study sample

Even in these languages, stress does not play any significant role in word-formation, or at least its function as an independent word-formation device is questionable. Such is the case of English, where examples of stress as a word-formation process are traditionally associated with conversion:

However, in these and similar examples of word-class-conditioned stress change, the position of stress results from the main word-formation process (conversion) which shifts stress to the status of a secondary phenomenon. Štekauer (1996: 55–95) argues that stress difference as in (72) and (73) does not result from an independent word-formation process and that such pairs should be treated as a specific subgroup within conversion.

The value of stress as a word-formation device is perhaps better appreciated comparing English and Luganda: while stress shift may be viewed in English as a by-product of certain cases of conversion (disyllabic nouns, verbs and adjectives), conversion does not exist in Luganda and yet stress shift has the same derivational effect:

In yet other languages, like Hebrew, the examples are diverse in nature: some noun–verb pairs parallel the above examples of English and therefore raise the same doubts. Hebrew does not seem to make relevant use, if any, of stress as a word-formation device. Semantic unrelatedness occurs in similar examples of Romanian and Ukrainian, where the total absence of any semantic relation argues against the word-formation status of stress:

These are different from the example taken from Ukrainian (77), which is a case of semantic divergence from one and the same source word, i.e. a diachronic process rather than one of word-formation:

This is supported by Slovak and Czech equivalents, both of these pairs having the same form, i.e. they are homonyms: Slovak zámok and Czech zámek. However, the two meanings (identical to those in Ukrainian) are not distinguished by stress.

Another borderline case includes examples for Belorussian and Vietnamese, where stress shift is at the border between inflection and derivation:

(78)

(78)

Belorussian paccьιnáь vs paccbınaь rassypac’ rassypac’ ‘spill.pfv’ ‘spill.ipf’

5.1.3 Tone/pitch

Like stress shift, the role of tone/pitch in word-formation seems comparatively minor, except in tonal languages. Tone/pitch is recorded in the languages shown in Table 5.5 (12.73 per cent of the study sample).

Table 5.5. Tone/pitch in the study sample

Chebanne (pers. comm.) explains that Cirecire has two fundamental tones, high and low (H and L, respectively), and two derived tones that are super-high (SH) in the context of consecutive HH tones in sentence-final position and super-low (SL) in the context of consecutive LL tones in sentence-final position. Tone has a derivational function in the following examples:

Tonal properties of verbal and nominal derived forms are given in (82) to (84):

Tone in Cirecire interacts with other word-formation processes, as illustrated in Table 5.6.

Table 5.6. Tone and other word-formation processes in Cirecire

Tone is an important aspect of Datooga inflection, which also has repercussions in derivational morphology. With respect to tonal behaviour, we have to identify tone-integrative suffixes, i.e. suffixes that impose tone patterns onto the noun as a whole, overriding lexical tone patterns. All the nominal plural suffixes do so. There is also a process of tone conversion by which plurals are derived from singulars: generally speaking, a switch from tone class 2 (H(H)L) to tone class 1 (L(L)H) derives a plural form from singulars in 0 and in -èe. Nouns with the primary suffix 0 that display tone conversion are shown in Table 5.7.

Table 5.7. Tone conversion in Datooga (nouns with primary suffix 0)

Nouns with the primary suffix -èe display tone conversion. Some are shown in Table 92.

Table 5.8. Tone conversion in Datooga (nouns with primary suffix -èe)

Newman’s (pers. comm.) examples for Hausa are verbal nouns derived from stems of a particular grade: monosyllabic H-tone verbs ending in short i have verbal nouns that end in a long -ī and have a falling tone:

In Dangaléat, tone can distinguish gender (114) and in Mandarin Chinese tone is sometimes used as a basis for morphological class (115):

Interestingly, tone does not function as a word-formation device in Vietnamese, even though it is a tone language¹⁹. It can only have a meaning-distinctive function in homonymous lexemes, as in (116):

The same can be found in Konni (117):

5.1.4 Word-formation by internal modification

5.1.4.1 Stem vowel alternation

Stem vowel alternation is recorded in the languages shown in Table 5.9 (23.64 per cent of the study sample):

Table 5.9. Stem vowel alternation in the study sample

5.1.4.1.1 Formal characteristics

Stem vowel alternation is frequent in Arabic and Hebrew. Vowel alternation is their fundamental word-formation process and is generally labelled root-and-pattern. According to Schwarzwald (2001: 23), ‘the number of roots [in Hebrew] runs somewhere between 3,000 and 4,500 roots. The number of patterns is limited to approximately 200. A single root may be inserted into many patterns’, e.g. g-d-l in nouns, verbs and adjectives (see Table 5.10).

Table 5.10. An example of root-and-pattern in Hebrew: g-d-l

At the same time, one and the same pattern can be used for the derivation of many new words (Schwarzwald Reference Schwarzwald2001: 23):

(131)

(131)

Hebrew CiCaCon > ביטחון bitaxon ‘security’ > דיבאון dikaʾon ‘depression’ > היגיון higayon ‘logic’ > זיכרון zikaron ‘memory’ > חיסרון xisaron ‘fault’ > שיגעון šiga^دon ‘insanity’

According to Zwarts (pers. comm.), the derivation in the following examples from Luo consists of merely a floating [+ atr] feature that autosegmentally attaches itself to the vowels of the verb root. Thus, the derivation process is based on changing a vowel from [- atr] to [+ atr]. Remarkably, the resulting agentive noun is in plural:

(132)

(132)

Luo choor > chöör ‘steal.v’ ‘thieves’

(133)

(133)

Luo rum > rüm ‘murder.v’ ‘murderers’

There are also combined types of stem modification, in which a vowel change accompanies the main word-formation process. Vowel modification is often combined with other processes, as with reduplication in Konni and Tibetan, with stem vowel modification in Marathi or with suffixation in Breton and Malayalam (see Table 5.11).

Table 5.11. Vowel modification in combination with other word-formation processes

5.1.4.1.2 Semantic characteristics

The range of functions of vowel modification as a word-formation process is broad and no generalizations seem to be possible. It is used to derive verbal nouns in Dangaléat (139), German (140) and Hausa (141):²⁰

(139)

(139)

Dangaléat εyìwì vs εyàawe ‘walk.n’ ‘walk.v’

(140)

(140)

German Gang < gehen ‘walk.n’ ‘go.v’

(141)

(141)

Hausa guugàa < googèe ‘rubbing.n’ ‘rub.v’

Other semantic changes caused by vowel alternation are shown in Table 5.12.

Table 5.12. Semantic diversity of vowel alternation

5.1.4.2 Stem consonant alternation

Stem consonant alternation is recorded in the languages shown in Table 5.13 (7.27 per cent of the study sample).

Table 5.13. Stem consonant alternation in the study sample

This word-formation process seems to play an important role only in Malayalam derivation. However, even this depends on the interpretation, because processes like vowel alternation and consonant alternation in Malayalam can also be interpreted as instances of affixation (Mohanan, pers. comm.):

(147)

(147)

Malayalam maraccu < maran~n~u ‘covered.tr’ ‘disappeared’

In English (148) there are cases of denominal verb formation by consonant alternation, but this process is not productive (Carstairs-McCarthy, pers. comm.). Similar examples can be found in Datooga (with change of q to g) (149) and Slovak (150):

(148)

(148)

English wreathe < wreath

(149)

(149)

Datooga gwalgwal-ees < qwalqwal-aad ‘steal upon ‘tiptoe’ (somebody)’

(150)

(150)

Slovak tíš < tich-ý ‘quietness’ ‘quiet’

5.2.1 Back-formation

Back-formation is recorded in the languages shown in Table 5.14 (16.36 per cent of the study sample).

Table 5.14. Back-formation in the study sample

This process may be regarded as a truly peripheral one. This is especially so as the question may be raised whether back-formation is not relevant only from a diachronic point of view, as assumed by Marchand (1960). From the synchronic point of view, Marchand (1960: 3) proposes the following equation for English back-formation: peddle : peddler = write : writer. This means that, synchronically, back-formation is analyzed analogically with suffixation. This makes sense because, logically, the cutting off of an affix postulates the prior attachment of this affix, even if the corresponding word-formation base was not in use before.

It has already been noted that, unlike Marchand, Kiparsky (1982a) explains the process of forming verbs like air-condition.v or spotweld.v (traditionally explained by back-formation) as compounding, based on the rule [Y Z]_x, with X being V. Similarly, Štekauer (Reference Štekauer1998, Reference Štekauer, Štekauer and Lieber2005) explains this type of example on a par with other word-formation processes, based on the Morpheme-to-Seme-Assignment Principle.

All these approaches, from the synchronic perspective, call into doubt the process of back-formation, and further undermine its status among word-formation processes from both cross-linguistic and language-specific points of view. The scarcity of back-formation does not, however, preclude it from ranging over various categories in some languages, as illustrated by Romanian (see Table 5.15).

Table 5.15. Back-formation in Romanian

In the majority of cases, the direction is from nouns to verbs, which express the action contained in the meaning of the motivating noun. This applies to several languages:

Sometimes the back-formation process does not reach beyond the boundaries of a particular word-class, as in the following examples from Serbian-Croatian and Slovak:

Aside from denominal verbs like (161), Finnish provides examples of an opposite direction, i.e. from verb to noun (162):

Clearly, back-formation is a typical European (plus all types of Englishes) matter, covering Germanic, Romance and Slavic. Back-formation in Finno-Ugric languages is non-existent or very rare. According to Kilgi (pers. comm.), there are some examples of back-formation in the history of Estonian, but it is not a productive word-formation process nowadays (cf. however (163) and (164)) and in Finnish it depends on the account of relatively rare compound verbs traditionally explained either as calques (165) or back-formations (166) (Laakso, pers. comm.):

There are few exceptions to the Eurocentric nature of back-formation in our sample. An example is taken from Marathi:

(167)

(167)

Marathi कर कर kara karaNe ‘hand’ ‘do.v’

5.3 Summary

This chapter reviews word-formation processes which do not involve the addition of derivational material, or which involve subtraction. The former type refers to conversion, stress, pitch/tone and stem alternation. Of these, conversion best exemplifies how the description of certain linguistic concepts is, probably unavoidably, based on the theoretical framework developed for Indo-European languages. As a result, it relies on concepts which otherwise would naturally not be used. By contrast, stress and, more clearly, pitch/tone, are rather foreign to this tradition. Back-formation is presented as the only subtractive word-formation process considered in this book. Its theoretical implications are briefly discussed and it is illustrated in some languages.