Polysynthetic Tendencies in Modern Greek^*

Chariton Charitonidis (Köln)

1 Language types and the index of synthesis

I would like first to present the main typological classes usually acknowledged in the literature and the current position of MG within this system. As a basis of argumentation, a version of the morphological classification of languages by Philippaki-Warburton (1992: 79–82) will be used.

The morphological differences that are found in languages classify them into two large categories: isolating ("analytic" in Philippaki-Warburton's terms) and synthetic languages. In isolating languages, the vast majority of words consist of free morphemes. That means that in these languages each morpheme constitutes a separate word, and for this reason the distinction between morphemes and words is not valid. Chinese and Vietnamese are typical examples of this language type. (1) is an example from Vietnamese, taken from Comrie (1981).

In (1) there are no bound morphemes. Every word is an invariable free morpheme.

In synthetic languages, words consist of more than one morpheme. Synthetic languages are subdivided into agglutinating and fusional ("inflectional" in Philippaki-Warburton's terms).¹ Agglutinating languages have two characteristic traits: (a) one can easily discern the morphemes that participate in the word form. In other words, the morphological structure of the words is transparent and immediately accessible, and (b) the relation between form (= morpheme) and meaning is one to one.

A representative agglutinating language is Turkish, as becomes apparent from (2) and (3):

On the other hand, in fusional languages (a) the division of a word into morphemes is problematic, and (b) the relation between the elements of form and elements of meaning is indirect and complex, as in the case of some polymorphemic words in MG. For example, the verbal ending ‑es in éghrapses 'you wrote' encodes 2^nd person and singular and past, the verbal suffix ‑tik in ghráftike 'it was written' encodes past and nonactive and perfective, and the nominal ending ‑us in anthrópus encodes accusative and masculine and plural (these examples are taken from Joseph 2002^a: 5f.).²

In (4) there is an overview of the typological classification by Philippaki-Warburton:

The classes A and B in (4) are differentiated on the basis of the number-of-morphemes-per-word whereby the subdivision of B into Agglutinating and Fusional is based on the criterion of the ease-of-segmentation of the word. However, these two criteria do not always operate independently. Τhere is a kind of languages called "polysynthetic," in which each sentence usually consists of only oneword which can have agglutinative and/or fusional traits. For example, the polysynthetic languages Chukchi and Eskimo can be regarded as agglutinative with a certain amount of fusion (cf. Comrie 1981: 42f.). It is obvious that the division in (4) is not adequate because it asymmetrically applies the criterion of the number-of-morphemes-per-word which definitely operates in the A-class.

In line with Comrie's suggestions, two parameters have to be asserted, i.e. the "index of synthesis" (number of morphemes per word) and the "index of fusion" (ease of segmentation of the word). The index of synthesis has the two extremes isolating and polysynthetic and the index of fusion has the two extremes agglutination (with straightforward segmentability) and fusion (with no segmentability). "What are traditionally called polysynthetic languages become languages with a high index of synthesis... The traditional class of agglutinating languages corresponds to those with a low index of fusion... the traditional class of fusional languages corresponds to those with a high index of fusion..." (cf. Comrie 1981: 43). In (5) I give an overview of Comrie's classification.

Following these suggestions, we have to abandon Joseph's (2002^a) assumption that the Greek verb is not polysynthetic because individual endings do not encode "several different grammatical categories all at once" in an "extensive degree" (cf. Joseph 2002^a: 5f.). In line with Comrie's argumentation, the attested patterns refer to a moderate "index of fusion" without excluding MG from polysynthesis. In other words, one can decide this issue only on the basis of the number of morphemes per word, i.e. on the basis of the "index of synthesis." The question which immediately arises is what is word in MG, an issue pursued in the next section.

2 Word in Modern Greek – Slot Patterns

The verb form in polysynthetic languages may be organized (i) in a templatic way or (ii) by scope ordering and that independently of affixal or compositional polysynthesis (see section 6). Greenlandic is an example of a language of the type (ii) in which different orderings of suffixes yield different meanings (cf. Mattissen 2003: 286f. and the examples therein). In organization type (i) there are several slots or "positions" before or after a verb root but not every slot need be filled in a single verb form, cf. Abkhaz, Navaho, Cayuga, etc.

In Cayuga, a Northern Iroquoian language spoken in Ontario, eight major parts of the verb form can be distinguished. From left to right, these parts are (1) the prepronominal prefixes, (2) the pronominal prefixes, (3) the reflexive/semireflexive, (4) the incorporated noun root, (5) the verb root, (6) the derivational suffixes, (7) the aspect suffixes, and (8) the so-called extensions (cf. Table 1, taken from Sasse 1999: 81).

The base of the verb form is constituted by positions (3) through (6). The addition of an aspect suffix (position 7) yields the aspect stem. As Sasse (1998) argues, "everything having to do with the lexical meaning of the verb is in the base."

In MG we get similar patterns, whereby the main difference is that position 4 can be occupied by an adverb or a noun (verb forms with both an incorporated adverb and a noun, such as sixno-kraso-píno³ 'often'-'wine'-'drink' are not grammatical). Table 2 shows how the Greek verb form is organized in Indicative and Subjunctive. NEG stands for the prepronominal negation marker dhen, CON for the prepronominal contrastive (negation) marker mi(n), FUT for the prepronominal future marker tha, MOD for the prepronominal modal marker na, PM₁ and PM₂ for the pronominal markers ('object pronouns' or 'weak pronominals')⁴, VR for the verb root, DS for possible derivation suffixes, AS for the aspect suffix, AGR for agreement.

(6a) and (6b) exemplify the patterns in (a) and (b) of Table 2, respectively.

Some remarks on the slot patterns in Table 2: the preverbal elements in various contexts are optional (cf. section 7), the verb base being VR-(DS)-AS. In IND and SUBJ, the prenominal marker in genitive (goal⁵) appears first, and the pronominal marker in accusative (theme) appears second.

One cannot have both a referential object marker and a referential incorporated noun root as theme. Sentences such as (7) are ungrammatical (i is the coreference index).

The same is also true for Mohawk, a Northern Iroquoian language spoken in United States and Canada (cf. Baker 1996: 22). In this case the object agreement morpheme must be lost, cf. the grammatical (8a) with the ungrammatical (8b).⁶

FUT and MOD are in complementary distribution in MG and define the position and form of the negation particle, which in (b) of Table 2 is named CON (=mi(n))⁷, cf. a similar pattern in Cayuga, whereby the CONTRASTIVE prefix thi-/thẽ-/tha'-/tha:- "regularly occurs as a substitute for the NEGATIVE prefix in combination with the modal prefixes where the NEGATIVE prefix is not allowed to occur" (cf. Sasse 1999: 83f.).

In pattern (b) of Table 2, the hortative prefix as- can appear instead of na-, expressing advice/exhortation, cf. (9).

Another monosyllabic element that may appear before na- (cf. 1b) is ghia- 'to', 'so as to', 'so that', etc. expressing purpose, cf. (10).

I am inclined to consider ghia- as a pre-pronominal prefix. In Cayuga, there is a group of affixes known as the DISLOCATIVE (DIS) suffix group which always occur in the suffix position 6, meaning 'go to do something,' e.g. -atawẽ- 'swim': ‑atawẽ-hne/a 'go there to swim' (cf. Sasse 1999: 90), i.e. with a fixed position in a slot pattern as in the case of the MG ghia-. We must thus extend the subjunctive pattern in (b) of Table 2 with the two patterns in Table 3.

Joseph has argued in an number of works (cf. Joseph 1990, 2002^a, 2002^b) for the benefits of a morphological classification in relation to clitics and other little elements or "particles" that tend to cluster in front of (or sometimes behind) a verbal head, esp. the modal markers na, as, the future (and modality) marker tha, the negation markers dhe(n) and mi(n), and the elements that mark argument structure ('object pronouns' or 'weak pronominals').

In his view, an affixal analysis of these elements is considerably favoured, cf. for example dhen which cannot be an independent word, is unable to occur as an one-word answer, i.e. it has an idiosyncratic behaviour which is characteristic of affixes. Dhen occurs only with indicative finite verbs, but not subjunctive finite verbs or nonfinite verbs (cf. examples in Joseph 1990: 174–176). It also shows semantic idiosyncrasy, e.g. in the expression dhen mu lés 'tell me...', in which it does not have its usual negative value (evidence for a nonsyntactic, noncompositional treatment of this element). In a similar way, the 'object pronouns' cannot occur as one-word answers and are strictly ordered with respect to the other preverbal elements (cf. Tables 2 and 3). Other criteria which advocate an affixal analysis of these pronominal markers are their high selectivity of combination, since in general they occur only with a verb, and the gaps in their combination, a phenomenon which is assumed to be typical of affixes, cf. (11) with the illegal order 1p-2p-verb.⁸

The morphophonological idiosyncrasies in the combination of these little elements also advocates an affixal analysis. For instance, when a weak pronoun in 3SG.ACC follows a weak pronoun in 2SG.GEN, the final ‑u of the first pronoun is deleted, as (12) exemplifies. This process does not fall in the domain of the general phonological rules in MG (cf. Joseph 1990, 2002^a and 2000^b for details).

Morphophonogical idiosyncrasies are very usual in polysynthetic languages, where affixal markers appear before a verbal head, i.e. within a word. For example, in Cree, an Algonquian language spoken in Canada, the pronominal markers ni- and ki- (cf. section 4.4) show idiosyncrasies such as the insertion of -t- before most vowel-initial stems, a phenomenon which (i) does not take place with the full forms of these reduced elements and (ii) does not comply with the general rules of Cree phonology (cf. Joseph 2002^b: 95 – further details and examples are given in (g) of section 3).

In this section I have shown that the verb form in MG has many polymorphemic patterns (patterns with a high index of synthesis according to Comrie 1981) with two features often appearing in polysynthetic languages: (a) slot patterns (b) morphological idiosyncrasies in the combination of morphological elements. But is MG a polysynthetic language? Section 3 deals with a general description of polysynthetic languages by Michael Fortescue and offers many corresponding examples from polysynthesis and MG.

3 Polysynthetic traits in Modern Greek

Fortescue (1994: 2601) mentions a number of "traits that tend to cluster in languages displaying polysynthetic morphology." He also says that none of these traits are criterial on their own. In the following, I will cite these traits (in italics) while keeping Fortescue's numbering. Each trait is accompanied with some characteristic examples from polysynthetic languages and MG.

(a) Noun stem incorporation within the verbal complex (also of adjectival stems within nouns).

Polysynthesis: Noun incorporation (NI) is the compounding of a nominal stem with a verb, in order to construct a complex verb. In Cayuga, for example, the nominal root -ahy- 'berry' is compounded with a verbal root -kw- to form the verb stem -ahyakw-, which then can be used as a base for verbs like ka'hya:kwas 'I am berrypicking' (cf. Mithun 1994: 5024; see also sections 5 and 5.1).

MG: A pattern of NI⁹ in MG is found in (13a). The non-incorporated pattern in (13b) has exactly the same meaning. Details about NI in MG and its limited productivity are found in section 5.1.

(b) A large inventory of bound morphemes (suffixes or prefixes) hand in hand with a somewhat limited stock of independent stems.

Polysynthesis: In (14) from Greenlandic Eskimo, traditionally regarded as the polysynthetic language par excellence (cf. Sasse undated: 4), the whole sequence is considered as a verb form, with aliikku being the only root in the verb complex whereas all other morphemes are suffixes (= li is an enclitic), i.e. non-root bound morphemes or the final inflectional suffix (cf. Mattissen 2003: 282).

MG: In MG there is also a large inventory of bound morphemes, cf. the prefixes kata- 'completely', para- 'excessively' kse- 'completely', 'very much', etc., afto- 'self', alilo- 'each other', TAM and inflectional suffixes, together with a limited stock of independent stems, cf. especially foreign loans like tést 'test' and asansér 'elevator' (see also section 6).

(c) Derivational (word-formation) processes productive in the formation of individual sentences, the verbal word being a minimal sentence (this can include shifts back and forth from nominal to verbal and the reverse within a single such word).

Polysynthesis: Eskimo languages have an immense potential for linear concatenation of affixes with "concrete meanings" as Sasse (undated: 4) reports. (15) is a classical example adopted from Finck (1909).

In the following I cite Sasse's description of the derivation patterns in (15): "The noun of instrument aalisaut 'fishing-line' in (15b) is a nominalization derived from the verbal stem aalisar- 'to fish' (15a). From aalisaut a second derived noun, aalisa-ut-issiaq 'something suited for a fishing line' (15c) can be formed by means of the suffix -issiaq-. This, in turn, enters a verbalization process by adding the verbalizing suffix -si- 'get' (15d). The whole complex is further extended by -niar- 'try,' which gives us 'try to get something...' and finally bound by the tense and pronominal suffixes which make the word-form complete (15e)." (Sasse undated: 4)

MG: Productive derivational patterns such as these in (15) are not attested in MG. The derivation process can take place at most twice, e.g. changing a noun or an adjective to a verb and again a verb to a noun, see (16).¹⁰

(d) Pronominal marking of subjects and objects or other core actants on verbal forms (so-called 'polypersonalism') and of possessors on nominal forms.

Polysynthesis: Abkhaz, a North Caucasian language, is a typical example of polyperso-nalism. See, for example, (17) where the elements Ø, ɭə, and y mark the core actants on the verb stem te-yt'.

(18) and (19) – again from Abkhaz – exemplify the pronominal marking of possessors on nominal forms. In (18) the element sə- marks the nominal form y°nə and in (18) the element yə marks the nominal form y°nə̀.

MG: In the extended verbal words in (20), the pronominal markers tu-and to-, and the inflectional ending -i mark the core actants on the verbal head p- (see also section 2). Pronominal marking of possessors on nominal forms is not attested in MG.

(e) Integration of locational, instrumental, and other adverbial elements into the verbal complex as affixes (including auxiliary-like pre- or postverbs).

Polysynthesis: As Mithun (1988: 44) reports, Maidu, a language of north-eastern California, has fifteen verbal suffixes which indicate the direction of motion. They supply such meanings as 'downward, to the end,' 'separation, off from,' 'against, up to, alongside of,' 'upward,' 'down, on ground,' 'away, away from,' 'down into a hole, into a house, into a box,' etc. See some examples in (21).

The same author quotes some examples from Lakhota, a Siouan language of South Dakota. The prefixes in (22), do not refer directly to specific instruments but rather indicate types of motion.

MG: MG adverbials can appear as prefixes in strict order before the head verb, see for example the position of the adverb ksaná in (23). The phenomenon of adverb incorporation (AI) in MG is discussed in detail in section 5.2.

(f) Many potential 'slots,' relatively few of them obligatory, fillable with specific (semantic) types of morpheme.

Polysynthesis: This trait is already discussed in section 2. See Table 1 for Cayuga.

MG: Similarly, this trait is attested in MG and is already discussed in section 2. See Tables 2 and 3 for an overview of the slot patterns. In (24) there is one more example: the valence operator afto- 'self' strictly appears between the incorporated adverb and the verb base.

(g) Productive morphophonemic processes resulting in several allomorphs (phonological shapes) for both lexical stems and bound morphemes.

Polysynthesis: We have already seen an example from Cree in section 2. Cayuga is a more prominent case: eight sets of pronominal prefixes are distinguished according to the initial sound of the verb base, e.g. for the combination 1S/2S one gets the allomorphs gõC- (C set), gõya- (a set), gõyẽ- (ẽ set), gõ- (i set), gõyẽ- (õ set), gõyo- (o set). In the third person singular feminine the i of i-initial bases is swallowed by the feminine prefix e-. Thus we get the forms e-gá:nya's 'she is paying (it)' from a consonantal base and e-dá:gra's 'she keeps falling' from an i-initial base (cf. Sasse 1998).

MG: Some examples are already given in section 2. Two special cases involving the sonority hierarchy are given in (25) and (26).

In (25), from the two vocals u and e survives the first one, which is higher in the sonority hierarchy (sonority hierarchy for MG: a>o>u>e>i), and the accent stays on the vocal which survives. In (26) su‑ looses u in front of the clitic (or in other instances in front of the definite article with the same form) and then a second rule o – e → o gives the complex its final form with the stress on the vocal which survives.

(h) Nonconfigurational syntax (relatively free word order).

Polysynthesis: In a nonconfigurational language, meaning is not (or not largely) determined by word order. As Baker (1996: 10) reports, in a simple transitive sentence of Mohawk the subject, verb, and object can appear in any of the six logically possible orders.¹¹ See (26).

MG: In MG there is a relatively free word-order in some discourse contexts, cf. the word orders SVO and VSO in (28b) associated with approx. the same meaning.

(i) Head- (or double-) marking type of inflection (cf. Nichols 1986).

Polysynthesis: In most polysynthetic languages, syntactic relations are marked on the head of a constituent, cf. the pronominal markers in (17) from Abkhaz above.

MG: The agreement marker on the verb is an obligatory and prominent head-marking element. The relevant phenomenon is discussed in section 4 in detail.

Table 4 summarizes the discussion so far in relation to MG. 1, 2, 3, and 4 are values given for "pervasive", "non-pervasive but recurrent", "marginal", and "not attested", respectively.

As shown in Table 4, the only pervasive trait is the existence of a large inventory of bound morphemes. Adverbials as verb affixes, allomorphs for both lexical stems and bound morphemes, nonconfigurational syntax, and the head-marking inflection are non-pervasive but recurrent. NI is marginal. The rendering of sentences by means of derivation and the pronominal marking on verbs and nomina are not attested. According to this table, there are apparent tendencies towards polysynthesis in MG.

In the next section, I will focus on Nichol's (1986) typological account and her insights as regards the head-marking patterns cross-linguistically. These patterns suggest a strong connection of MG to polysynthesis. On top of this, I will mention some diachronic issues related to the evolution of head-marking elements in polysynthesis and MG.

4 Head-marking and dependent-marking grammar (Nichols 1986)

4.1 Head-marking and dependent-marking

According to Nichols (1986), in the area of languages which are rather morphologically oriented,¹³ two types can be assessed: the 'head-marking languages' and the 'dependent-marking languages.' In the following, I cite her main description of the phenomenon:

MG, like most Indo-European languages, is mainly dependent-marked. The dependent-marking patterns appear in case marking, in the construction of the possessive NP, in the attributive phrase (agreement in adjectives), in clause relations, relativization and other subordination (cf. Nichols 1986: 59–64 for the description of these phenomena; see also Charitonidis 1999).

The head-marking patterns relate MG to polysynthesis. In the latter, the head-marking material is usually concentrated before a verbal head. The corresponding Greek patterns appear esp. in the verbal agreement and the pronominal marking of actants (the so-called 'clitics') before the verb root. See a dependent marking pattern in (29) and its corresponding head-marking pattern in (30).

In line with Nichols's suggestions, voice in MG can be regarded as a head-marking element of an indirect type because it carries information about the verb's valence – although it does not directly mark the occurrence of particular nominals filling the valence in the clause (cf. Nichols 1986: 64). See (31).

4.2 Patterns which favour head-marking cross-linguistically

Patterns which favour head-marking cross-linguistically and in MG (cf. Nichols 1986: 77–79) are (a) person and number agreement on verbal heads, (b) the negation, which is associated with the head of the constituent rather than with the dependent in its scope, cf. (31) from MG in which the negation dhen has floated away from the NP óla ta mejéthi toward the verb in clause level, and (c) the semantically based hierarchy of adverbials, cf. (33) – section 5.2 on AI gives a general pattern for this hierarchy.

VO languages of all types (SVO, VSO, VOS) are among languages with head-marking tendencies. MG is a VO‑language (according to Nichols's categorization; cf. 1986: 79–83) and is expected to have these tendencies. On the other hand, the nonconfigurational patterns and the great freedom of choice as to word order in MG (cf. (h) of section 3) is opposed to non-head-marking languages which "tend to be confined to the verb-final type" (cf. Nichols 1986: 104).

4.3 The appositive nature of dependents in polysynthesis and Modern Greek

In polysynthesis, when nominal dependents appear together with pronominal markers on the verbal head, the role of the dependents is appositive, cf. (17) from Abkhaz, repeated below as (34). The nominal dependents 'man,' 'woman,' and 'book' are in apposition to the coreferential pronominal markers on the verb.

In languages with consistent head-marking such as Abkhaz, "full NP's are included only for emphasis, focus, disambiguation, etc.' (Nichols 1986: 107). Similar patterns are attested in MG, in which the pronominal markers on the verb have the same reference as the external nominal phrases, which are included in the sentence usually for emphasis or disambiguation, cf. (35). It is noteworthy that Greek pronominal clitics are grammaticalizing definiteness as in head-marking languages.

The head-marked dependency is unilateral, in that the dependent requires the head but the head does not require the dependent; the head can occur alone with the same reference (cf. Nichols 1986: 108). This actually happens in (34) and (35) from Abkhaz and MG, respectively, in which the verb constitutes a minimal sentence together with the pronominal markers. The inflectional agreement suffix on the Greek verb suggests another unilateral relation, i.e. in sentences such as (36) the subject is optional, i.e. it can be absent in some situational contexts (see section 7).

Head-marked patterns such as that in (36) are also attested in many Indo-European languages, cf. English, German, a.o.

4.4 Headword migration and cliticization

In both Romance languages and MG, direct and indirect objects alternate or co-occur with clitics on verbs. As mentioned in section 4.3, the pronominal "clitics" before the Greek verb can be regarded as verbal markers (= affixes), in co-reference with case-marked nominal dependents. A related cross-linguistic phenomenon is the combined appearance of headword migration and reduction of whole words to affixes via cliticization. In a diachronic view, these reduced elements are derived from former dependent NPs having migrated, as it were, towards a verbal head as affixes (cf. Nichols 1986: 72). Joseph (2002^b) reports that parallels between the independent pronouns and the weak pronouns are attested in many languages. In MG there are parallel forms such as aftón/ton (3SG.ACC), eména/me (1SG.ACC), etc. In the Algonquian (polysynthetic) language Cree, the free pronouns ni:ya 'I' and ki:ya 'you' are similar to the reduced forms (= prefixes) ni- and ki-, that mark first and second person, respectively (cf. Joseph 2002^b: 95).

4.5 Conclusion

Although mainly a dependent-marking language, MG contains many head-marking patterns. Person/number and agreement markers on the verbal head, voice and negation are the most prominent. The semantically based hierarchy of adverbials and VO-order favor head-marking patterns in both polysynthesis and MG. In both polysynthesis and MG, when nominal dependents appear together with pronominal markers on the verbal head, the role of the dependents is appositive. In both polysynthesis and MG, the pronominal markers are derived from former dependent NPs. I conclude that head-marking is synchronically and diachroni-cally motivated in both polysynthesis and MG and that this fact brings MG closer to polysynthesis.

5 Incorporation

Before I proceed to further evidence which supports a polysynthetic view of the Greek verb, I would like to define the notion of incorporation. As Mithun (1994) argues, this notion is traditionally used with a specific meaning, referring to noun incorporation (NI), i.e. the compounding of a nominal stem with a verb, in order to construct a complex verb (see (a) of section 3).

In the literature, we usually find a more general account of "incorporation," which, except of the traditional definition of NI above, includes the incorporation of adverbials/adverbs, verb roots, nouns with adverbial function, adjectives, sentence adverbials, etc. (cf. Spencer 1995: 455–459 for examples from Chukchi).

Furthermore, as Mithun (1994) notes, the notion of "incorporation" is sometimes used extensively to cover various types of morphologization, whereby markers appear in the verb complex, which correlate to independent words in the same language or other languages. These can be pronominal affixes, verb affixes, etc. which modify grammatical relations (cf. Mithun 1994: 5025 and see note 8).

In this paper, I use the term 'incorporation' in the last two broad senses. The incorporation of pronominal prefixes and other elements into the verbal complex inside slot patterns has been discussed in section 2. In the following, I will proceed with the presentation of the two main incorporation patterns in MG, i.e. (a) NI (see section 5.1) and (b) AI (see section 5.2). At the same time, relevant examples from several polysynthetic languages will be presented.

5.1 Noun Incorporation

In a language like Cayuga, the main type of incorporation is the productive NI, esp. the incorporation of an object (theme) into the verbal complex. In (a) of section 3 I have discussed this phenomenon on the example of the verb stem -ahy-a-kw- 'berry picking.' In that case, the element -a- is the so-called stem-joiner (SJ). As Sasse (1999: 85) notes "it does not have any meaning in and of itself but simply serves in tying the two roots together." This type of incorporation is very common and yields compounds from almost any verb, e.g. the root -kw- can serve as the base of verbs such as -hnyõhs-a-kw- 'pick squash/pumkins', -yẽt-a-kw- 'get wood,' etc. (see Sasse ditto).

Another type of incorporation in Cayuga is the lexicalized incorporationwhich yields NR+VR compounds in a specific sense, e.g. 'mind+drop' + REPETITIVE = 'forget,' 'cold + enter' + BENEFACTIVE = 'get a cold or flu,' 'throat + dry' = 'be thirsty,' etc. (cf. Sasse 1999: 88).

In the literature it is sometimes discussed whether there is productive NI in MG or not (see esp. Rivero 1992 and Smirniotopoulos/Joseph 1998). The issue is whether NI is a product of syntactic rules, i.e. compositional/ad hoc incorporation, or a product of lexical rules, i.e. a process which yields idiosyncratic items in relation to their source.

I would like first to proceed with some data from NI. The relevant patterns show composi-tional and non-compositional semantics, see (37) and (38), respectively.

The main indication pleading for a non-compositional/lexical analysis of the formations in (37) and (38) altogether is the union vowel -o-, which is recognized as the 'union' vowel in Greek compounds (cf. the SJ -a- in Cayuga above). Spur-of-the-moment creations such as ghato-vlépo 'look at cats,' 'be a cat-looker,' pito-trógho 'eat pittas,' 'be a pitta-eater,' and rodho-kiló 'roll-tires,' 'be a tire-roller' confirm the existence of a productive object-verb or NI pattern with an active and intransitive verb. However, the acceptability of these formations is extremely restricted. There is thus good reason to believe that NI in MG is a lexical, non-compositional, i.e. non-syntactic process (cf. Smirniotopoulos/Joseph 1998).¹⁶

Summarizing, NI in Cayuga and MG can be regarded as a compounding process sharing many similarities such as compositional and/or non-compositional semantics, morphological processes like the addition of union vowels, and slot patterns, in this case the appearance of a NR just before the VR.

Another kind of incorporation in polysynthesis is the incorporation of a NR with instrument role, see (39) from Mohawk with the incorporated noun root hióhs 'elbow' denoting a body part.¹⁷

Similar patterns are found in MG, cf. the verb podh-o-pató 'tread on sb/sth,' consisting of the verb base pató 'tread' and the incorporated noun pódhi 'foot' which denotes a body part having the instrument role.

5.2 Adverb Incorporation

AI is not a unique characteristic but only an indication of polysynthesis, cf. Cayuga which shows no AI (cf. Sasse 1999).

As Spencer (1995) notes, it is difficult to find incorporation of temporal or aspectual adverbials in a given language. For example, in Chukchi, a palaeosiberian polysynthetic language, words like jep 'still' and temporal adverbs like igər 'today' and ajwe 'yesterday' fail to incorporate. The results of incorporation usually appear as semantic units, (esp. prototypical predicates consisting of verb + object) and any sequence combination which is difficult to interpret as such a unit, e.g. a verb together with a sentence adverb, is rare and may even be ungrammatical in a given language (cf. Spencer 1995: 461). In line with this insight, perfective markers and aspectuals fail to incorporate both in polysynthetic languages and MG.¹⁸

Let us now look at the regular AI patterns in polysynthesis and MG. In Nahuatl, a polysynthetic Uto-Aztecan language from Central Mexico, manner adverbs and directionals are productive incorporators, cf. (40b) and (41), respectively.

In (40b) and (41), an adverb appears before the verbal root and after the object marker Ki, a slot pattern very similar to the Greek one, in which incorporated adverbs appear before the verb root and after incorporated pronominal objects (see section 2). Two examples from MG with an incorporated manner adverbial (see (42)) and a directional (see (43)) exhibit the same patterning as the Nahuatl examples above.

In (44), I cite some examples with manner incorporators from MG.

Aktionsart²⁰ adverbs incorporate into the verb stem of many polysynthetic languages, e.g. in Chukchi, cf. (45). The same happens with the Aktionsart adverbs in MG, cf. (46).

As to the Greek verbs which allow AI, these are non-stative and refer to actions, processes, and changes of state (cf. Rivero 1992: 300f. and the references therein), cf. the grammatical pattern in (47b) with the ungrammatical one in (48b).

As in the case of NI (see section 5.1), AI cannot be regarded as a regularly productive, compositional (= syntactic) process. On the one hand, the union vowel -o-, which appears equally in MG noun and AI is an indication of a lexical compounding process. On the other hand, there are many semantic and morphological complications, e.g. the free adverb phrase can have a different meaning than the incorporated pattern, cf. stékome kondá 'I stand closely' and kondostékome 'I stop for a little time' or the incorporated adverb can appear in an etymologically and morphologically distinct form than in the free adverb phrase, cf. perpatáo ghríghora and ghorghoperpatáo, both meaning 'walk quickly' (see Smirniotopoulos/Joseph (1998) for further examples and details).

Nonetheless,if we depart from a pure syntactic analysis we have to admit that there are several regular patterns which give AI in MG a character similar to that of polysynthesis proper (see also next section).

6 Modern Greek in relation to affixal and compositional polysynthesis

In this section, I present Mattissen's (2003) distinction of affixal and compositional polysynthesis, and then apply this distinction on MG. I first cite her comprehensive description of these two language types:

The pervasive existence of non-root bound morphemes in MG is already mentioned in (b) of section 3. The question which immediately arises is whether MG consequently follows one of the two strategies mentioned above or a mixed strategy, at least in some of its verb forms. As regards to the affixal strategy (see (i) above), see (14) from Greenlandic, repeated here as (49).

As mentioned in (b) of section 3, in (49) the whole sequence is considered as a verb form, with aliikku being the only root in the verb complex. All other morphemes are suffixes (=li is an enclitic), i.e. non-root bound morphemes or the final inflectional suffix (cf. Mattissen 2003: 282).²² Extreme derivation patterns such as these are not attested in MG, a fact that definitely distances MG from affixal polysynthesis towards compositional polysynthesis.

As regards to compositional polysynthesis, a typical member is Chukchi, allowing several roots within one word form, making use of NI, AI, etc. On top of this, in Chukchi there are non-root bound morphemes expressing, for example, person marking, valence operations (e.g. reflexive, reciprocal), aspectual categories, degree, manner, reversative, direction, etc. (see Mattissen 2003: 282f. and the examples/references therein).

In section 5.1 I have mentioned that NI in MG is likely to follow a compositional strategy only in a limited way, because it shows many lexicalized patterns without a transparent counterpart, i.e. the relevant patterns are not always ad hoc formations (see (ii) above). On the other hand, AI is not particularly productive (cf. Smirniotopoulos/Joseph 1998). However, there are indeed non-root bound morphemes, such as (a) person markers (see section 2), (b) aspectual categories (integrated into TAM suffixes – see sections 1 and 2), (c) valence operators, e.g. the reflexive afto- and the reciprocal alilo- (see (f) of section 3), (d) prefixes which express degree, e.g. para- 'excessively,' kata- 'completely,' 'very much,' kse‑ 'completely,' 'very much,' etc., manner, e.g. ksaná 'again,' reversative, i.e. kse- (equivalent to the English un-),opposition/substitution, i.e. anti- 'opposite,' 'against,' 'instead,' a.o. (see Warburton 1970: 62f).

Mattissen (2003) regards ad hoc verb root serialization as a necessary condition for the assessment of compositional polysynthesis, a pattern which is only marginally attested in MG. I cite two examples from verb root serialization in Chukchi (see (50a) and (51a)) with their synonymous analytic counterparts (see (50b) and (51b), respectively).

Patterns such as those in (50) and (51) are not attested in MG. There are only a handful of verb root serializations, such as anavo-svíno'switch on-switch off, 'anev-o-katevázo 'take up-take down', anev-o-katevéno 'go up-go down' trogho-píno 'eat-drink', anigho-klíno 'open-close', beno-vghéno 'come in-get out', mapped onto a concrete semantic pattern, i.e. action/opposite of the action (trogho-pino may be better considered as a coordinative compound). The verbs tremo-pézo 'blink', 'flare', (literally: 'tremble-play') and tremo-fégo 'coruscate', 'flare' (literally: 'tremble-beam/radiate') are lexicalized formations.

These restricted patterns vis-à-vis verb root serialization force us to characterize MG as a language close to the polysynthetic 'Mixed II' type, i.e. a language with non-root bound morphemes, NI (although restricted, see section 5.1), with one or more roots per verb form (see NI in section 5.1 and AI in section 5.2). Polysynthetic languages of this type are Takelma and Blackfoot (this categorization relies on Mattissen 2003: 287; for the other types of polysynthetic languages see Mattissen ditto).

7 Reference and predication strategies in Modern Greek and polysynthesis

Apart from the fact that NI and AI appear to be processes with restricted productivity in MG, resulting in many cases in word forms with non-compositional semantics (see sections 5.1 and 5.2), the attested slot patterns (see section 2) have no pervasive character and the appearance of the pronominal markers before a verbal head is massively influenced by discourse factors, cf. the following situations:

(i) John asks Mary what the teacher does with a pupil in the classroom. John can't see the scene. The book is mentioned as a noun in John's question (see (52)).

(ii) John asks Mary what the teacher does with a pupil in the classroom. John can't see the scene. The book is not mentioned in John's question at all (see (53)).

(iii) John asks Mary what the teacher does in the classroom. John can't see the scene. The book and the pupil is not mentioned in John's question at all (see (54)).

In (52b) there are two pronominal markers before the verbal head, in (53b) one and in (54b) none. This patterning is not the same with the patterning of pronominal markers in polysynthetic languages. The pronominal markers in Cayuga, for example, are obligatory and in principle denote two core arguments, i.e. "agent" and "patient" or "actor" and "undergoer". The precise interpretation of these arguments is not fixed but conventionalized, according to the lexicalized argument structure of the verb, e.g. the first argument may be agent, experiencer etc., the second argument may be patient, goal, location (in the last case with the addition of an applicative suffix), etc. (see Sasse 1999: 37–39). On the other hand, in MG the first PM denotes the goal (or sometimes the beneficiary) and the second PM the theme (see (52)).

On top of this, the elaboration principle of sentence structure which appears in Cayuga and the other North American languages is not the same as in MG. In particular, the appositive relation of a coreferential participant expression (word) to the pronominal markers of a verb base denoting the fundamental situation, is different. As Sasse (1988: 194) argues, in Cayuga the verbal character of all minimal units (i.e. words) in the sentence, force this appositive relation as a functional necessity. To become less abstract, cf. (55).

All three sentence units in (55) are "verbs" and contain verb bases, i.e. kẽ''see', yahshé: 'be two people', and ksa:'-áh 'be a child'. The pronominal two-place prefixes ka:khe and kae are referential and obligatory. We see thus that the elaboration pattern (approx.) "I saw persons" → "they were two people" → "they were children" is forced by the verbal character of these three units, whereby the basic situation, i.e. (approx.) 'the seeing of persons,' must be mentioned in the first place. This is not the case with a corresponding MG sentence, cf. (56).

In (56), the appositive relation of ta pedhiá to the pronominal marker ta- before the verb ídha is not motivated by the verbal character of the contained units, since in MG there is a clear verb-noun distinction (the verb is predicated of nominal arguments). On top of this, the pronominal marker ta- is optional, i.e. it can be absent in different contexts (cf. (52)–(54)).

For further differences between the systems of polysynthetic languages and MG, cf. Mithun (1988). The grammatical (esp. nominal) categories established in the western literature for the description of European languages, such as number, gender, and case have no immediate correlates in the morphological systems of the North American languages. The inflectional categories of the European noun are mostly derivational in the polysynthetic languages and are marked on the verb (cf. the notion of head-marking in section 4). The conclusion to be drawn is that, notwithstanding all attested analogies, both systems are not structurally isomorphic (cf. Mithun 1988; cf. also Sasse 1988, 1993).

8 Conclusions

The analysis in sections 1–7 has shown that MG is a language with many polysynthetic tendencies/traits but is not a polysynthetic language. As already mentioned in Charitonidis (2007), MG has a moderate, and in several cases strong, index of synthesis (the two extremes being isolating and polysynthetic) and a moderate index of fusion (the two extremes being agglutination – with straightforward segmentability – and fusion – with no segmentability).²³

The slot patterns of the Greek verb described in section 2 yield polymorphemic words and are similar to the patterns in many polysynthetic languages, esp. the North American languages. However, the pronominal markers have no obligatory character (see section 7); there is also no pronominal head-marking on nominals in combination with a co-referential NP (see section 3).

Section 3 has underscored a pervasive phenomenon common in MG and many polysynthetic languages: the existence of a large number of bound morphemes together with a limited stock of independent stems. Fortescue's (1994) eight traits of polysynthesis received four different values. It is interesting that four of these traits have received the value 2, i.e. "non-pervasive but recurrent."

Head-marking is an important trait of polysynthesis but in MG it is not pervasive. The inflectional verb ending is a regular head-marking element but is also attested in many Indo-European languages. However, this trait is synchronically and diachronically motivated in both polysynthesis and MG (see section 4). The development and strengthening of this trait in MG may contribute to the evolution of a new system away from the traditional dependent-marking strategy and simple synthesis towards head-marking and polysynthesis.

NI is marginal in MG. Again, there are many similarities with polysynthesis but the attested patterns do not point to a productive/adhoc process (see section 5.1). On the other hand, AI is a recurrent phenomenon showing many regularities (see section 5.2). The insertion of a noun or an adverb into slot patterns together with NI/AI yields a compound polysynthetic pattern (see section 2).

Mattissen's (2003) account of polysynthesis suggests that MG is definitely dissociated from the affixal strategy, operating for example in Greenlandic Eskimo (see section 6). However, the fact that MG is associated with languages such as Takelma and Blackfoot following the compositional strategy (in Mattissen's terms) does not suggest the isomorphic status of MG with these languages.

The problem of isomorphism is taken up in the last section. Leaving aside that the inflectional agreement-ending denotes the agent in most of the cases, the presence of object markers is massively influenced by discourse factors. On top of this, there is no consistent agent-patient encoding within the verb complex by means of inflection – a regular phenomenon in polysynthetic languages such as Cayuga. This optionality is closely connected to the elaborated noun-verb distinction and the predominant dependent-marking patterns in MG which distance MG from the verb-oriented polysynthetic languages.

The juxtaposition of independent pronouns or noun phrases to the incorporated pronominal markers (as well as to the agreement ending) in MG has a different character in comparison with the polysynthetic juxtaposition. The structure of apposition in the North American languages is functional and has the role of mentioning the presupposed entities in a basic situation throughout the sentence. In MG, apposition is mainly used for selectively emphasising or disambiguating entities embedded in a concrete situational context (see section 7).

Notes

^* I would like to thank Hans-Jürgen Sasse for his inspiring ideas and comments on drafts of this paper. I would also like to thank Irene Philippaki-Warburton for her supporting comments on some early research and Stavros Skopeteas who motivated me into writing on this topic again. back

¹ Comrie (1981) notes that "both agglutinating and fusional languages, as opposed to isolating languages, have inflections, and it is therefore misleading to use a term based on (in)flection to refer to one only of these two types. The availability of the alternative term fusional neatly solves the terminological dilemma" (see Comrie 1981: 42). back

² In MG the verb is inflected for mood (indicative, subjunctive, imperative), aspect (perfective, imperfective), voice (active, passive), tense (present, past), person (1^st, 2^nd, 3^rd), and number (singular, plural). The future is expressed by a particle tha followed by the present or dependent (the verb-form distinction present/dependent relies on the account of Holton et. al. (1997); it is not used in this paper because it impedes the comparison of MG with the polysynthesic languages under examination). There is also an indeclinable active gerund in ‑ondas, an active and passive present participle, and a past passive perfect one in -ménos (kurasménos 'tired'). Formally, the finite forms of the verb (those with personal endings) consist of a stem + (optionally) the perfective aspect maker (-s- in active, -th- in passive) + personal ending (indicating person, number, tense, mood, voice). Past forms are usually prefixed by e- (the "augment"). In the active 1st conjugation, the present endings are -o, -is, -i, -ume, -ete, -un, and the past endings are -a, -es, -e, -ame, -ate,-an. There is also a 2^nd conjugation with two types (for further details cf. Holton et al. 1997). back

³ The usual citation form for the Greek verb is the 1st person singular. back

⁴ A list of the 'object pronouns' or 'weak pronominals' can be found in Joseph (2002^a: 3). back

⁵ Marginally, the pronominal marker in genitive expresses the beneficiary role, cf. (i) back

(i)	Mu-éftiakse for.me-he.repaired	to the	aftokínito. car
	'He repaired the car for me.'

⁶ Baker (1996) names this type of incorporation "robust." Accordingly, noun incorporation in a specific language is robust if (a) it is reasonably productive, (b) the noun root is fully integrated with the verb morphologically, (c) the noun is referentially active in the discourse, and (d) both the noun root and the verb root can, in general, be used independently (cf. Baker 1996: 19). back

⁷ As Joseph (2000^b) notes, mi(n) cannot be used as a general non-indicative negator since imperatival forms cannot be negated per se but only their "surrogate" prohibitives. These always contain a finite verb form, cf. mí(n) fíghis 'don't leave.' For further arguments about the special status of mi(n) see Joseph (2000^b: 107 Fn. 5). back

⁸ Joseph (2000^b) adopts the position made by Zwicky (1994: xiii) that the notion of clitic is not a genuine category in grammatical theory and can be rejected as unnecessary. According to this position, the binary division of the elements of grammar in affixes and words assigns these elements to morphology or syntax, respectively, and is indispensable. Clitics are thus considered as atypical affixes (for further details see Joseph 2000^b: 89–91). For a syntactic approach see Philippaki-Warburton/Spyropoulos (1999), who consider pronominal clitics as phonologically dependent words. back

⁹ In this paper, I use the term 'incorporation' barring strict syntactic considerations which rely only on compositional patterns (cf. Baker 1996 a.o.). As we will see in sections 5.1 and 5.2, MG and polysynthesis exhibit both non-lexicalized (compositional) and lexicalized (non-compositional) patterns, whereby syntactic theory cannot give a homogenous account. Accordingly, a compounding/affixal analysis seems more adequate and is adopted here. back

¹⁰ Cf. Alexiadou (to appear) for details of this process. back

¹¹ Baker (1996: 10) implies that discourse ("pragmatic") factors influence word order in Mohawk but he does not go into detail. back

¹² As Baker (1996: 137: f.26) notes "what ne is remains somewhat mysterious; its core synchronic use seems to be simply to mark a postverbal NP constituent under certain conditions. However, there is a significant amount of variation in its use both within and across speakers". back

¹³ Nichols (1986) does not examine languages of the isolating type. back

¹⁴ According to Nichols (1986: 57), "the head is the word which governs, or is subcategorized for – or otherwise determines the possibility of occurrence of – the other word. It determines the category of its phrase." back

¹⁵ In MG a governed adverbial is ksaná 'again' and an outer adverbial is akómi 'still.' See also section 5 on AI. back

¹⁶ In this paper, syntax and syntactic theory are meant in relation to the American generative tradition (cf. the syntactic framework in Chomsky 1981). back

¹⁷ Sasse (1999: 88) reports similar incorporation patterns in Cayuga. back

¹⁸ However, this behaviour is far away from regarding it as universal, since temporal adverbs and sentence adverbials do incorporate in Chukchi, as Spencer (1995: 461f.) shows. back

¹⁹ The typical ending of the MG adverbs is -a. The typical union vowel in Greek compounds is -o- which deletes the adverbial ending -a. When the verb root begins with the vowel a, the union vowel -o- is usually deleted (see h'). With the adverb ksaná 'again' the union vowel does not appear (see 42), perhaps an indication of incorporation in the syntax (cf. Smirniotopoulos/Joseph 1998). back

²⁰ In Rivero (1992: 307–314) 'Aktionsart' is distinguished from aspect. 'Aktionsart' is considered as 'a characteristic of the inherent meaning of a verb' and may appear as durativity or iterativity. According to the same author, 'aspect' adverbs, such as akómi 'still' and mólis 'just,' fail to incorporate in MG, as opposed to 'Aktionsart' adverbs, such as sixná 'often' and dhiplá 'twice,' which do incorporate. I suggest that this distinction is not always valid, given the fact that adverbs such as sixná, which modify situations irrespective of the perfective/imperfective (= aspect) distinction, fail to incorporate, cf. the grammatical (ia) and (iib) with the ungrammatical (ii). An incorporated pattern with miláo, i.e. sixnomilái, was attested only once in a poem by Ioanis Vilaras (see http://arcadia.ceid.upatras.gr/arkadia/culture/personalities/vilaras.htm, accessed May, 2008) and thus cannot be regarded as a corroborating evidence.

Furthermore, the 'Aktionsart' adverb páli 'again' was not attested in an incorporation pattern, against Rivero's prediction (Rivero 1992: 307). This issue requires further investigation in order to define the exact semantics of the incorporating adverbs. back

²¹ In the Chukchi examples I follow Spencer (1995) who indicates incorporation with '=' before the verb stem. back

²² The derivation of another Eskimo sentence is exemplified in (c) of section 3. back

²³ This categorization relies on Comrie (1981); see section 1. back

Abbreviations

References

Alexiadou, Artemis (to appear): "On the role of syntactic locality in morphological processes. The case of (Greek) derived nominals". In: Giannakidou, Anastasia/Rathert, Monika (eds.): Quantification, definiteness and nominalization. Oxford.

Baker, Mark C. (1996): The polysynthesis parameter. New York.

Beller, Richard/Beller, Patricia (1979): "Huasteca Nahuatl". In: Langacker, Ronald (ed.): Studies in Uto-Aztecan grammar. Vol. 2: Modern Aztec grammatical sketches: 199–306.

Charitonidis, Chariton (1999): Head-marking und Inkorporation im Neugriechischen. Unpublished MA paper. Universität Köln.

Charitonidis, Chariton (2007): "Towards a typological classification of Modern Greek". Studies in Greek Linguistics 27.

Chomsky, Noam (1981): Lectures on government and binding. Dordrecht.

Comrie, Bernard (1981): Language universals and linguistic typology. Chicago.

Dixon, Roland B. (1911): "Maidu". In: Boas Franz (ed.) (1911): Handbook of American Indian languages. Bureau of American Ethnology, Bulletin 40.1. Washington: 679–734.

Dixon, Robert M. W./Aikhenvald, Alexandra Y. (eds.) (2002): Word. A cross-linguistic typology. Cambridge.

Fortescue, Michael (1983): A comparative manual of affixes for the Inuit dialects of Greenland, Canada, and Alaska. Kopenhagen. (= Man and Society 4).

Fortescue, Michael (1994): "Morphology, polysynthetic". In: Asher, Roland E./Simpson, J. M. Y. (eds.): The encyclopedia of language and linguistics. Vol. 5: 2600–2602.

Joseph, Brian D. (1990): "The benefits of morphological classification. On some apparently problematic clitics in Modern Greek". In: Dressler, Wolfgang U. et al. (eds.) (1990): Contemporary morphology. Berlin: 171–181.

Joseph, Brian D. (2002^a): "'Word' in Modern Greek". In: Dixon, Robert M. W./Aikhenvald, Alexandra Y. (eds.) (2002): Word. A cross-linguistic typology. Cambridge: 243–265. Available at http://www.ling.ohio-state.edu/~bjoseph/publications/2001word.pdf (accessed February 2008).

Joseph, Brian D. (2002^b): "Defining 'Word' in Modern Greek. A response to Philippaki-Warburton/Spyropoulos 1999". In: Booij, Geert/van Marle, Jaap (eds.) (2002): Yearbook of morphology 2001: 87–114.

Henry, Reg/Hill, H. Louise (eds.) (1994): Õ:gwéh Goga:'. Vol. 7. Brantford/Ontario.

Hewitt, B. George/Khiba, Zaia K. (1979): Abkhaz. Amsterdam. (= Lingua descriptive studies 2).

Holton, David et al. (1997): Greek. A comprehensive grammar of the modern language. London.

Mithun, Marianne (1988): "System-defining structural properties in polysynthetic languages". Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung (ZSPK) 41/4: 442–452.

Mithun, Marianne (1994): "Word-formation. Incorporation". In: Asher, Roland E./Simpson, J. M. Y. (eds.) (1994): The encyclopedia of language and linguistics. Vol. 9: 5024–5026.

Mithun, Marianne (2004): "The Non-universality of Obliques". Talk at the conference Syntax of the world's languages, Leipzig, August 5, 2004 (Handout).

Nichols, Johanna (1986): "Head-marking and dependent-marking grammar". Language 62: 56–119.

Philippaki-Warburton, Irene (1992): Εισαγωγή στη Θεωρητική Γλωσσολογία ("Introduction to theoretical linguistics"). Athens.

Philippaki-Warburton, Irene/Spyropoulos, Vassilis (1999): "On the boundaries of inflection and syntax. Greek pronominal clitics and particles". In: Booij, Geert/van Marle, Jaap (eds.) (1999): Yearbook of morphology 1998. Dortrecht.

Rivero María-Luisa (1992): "Adverb incorporation and the syntax of adverbs in Modern Greek". Linguistics and Philosophy 15: 289–331.

Sasse, Hans-Jürgen (1988): "Der irokesische Sprachtyp". Zeitschrift für Sprachwissenschaft 7/2: 173–213.

Sasse, Hans-Jürgen/Keye, Alfred (1998): Far more than one thousand verbs of Gayogoho:nǫ' (Cayuga). A handbook of Cayuga morphology. Manuscript. Universität Köln, Germany and Woodland Cultural Centre, Brantford, Ontario.

Sasse, Hans-Jürgen (1999): Cayuga. Unpublished manuscript. Canberra: Research Centre for Linguistic Typology.

Sasse, Hans-Jürgen (undated): Polysynthetic languages. Chapter 1: Introduction. Unpublished manuscript.

Sischo, William R. (1979): "Michoacan Nahuatl". In: Langacker, Ronald (ed.): Studies in Uto-Aztecan grammar. Vol. 2: Modern Aztec grammatical sketches: 307–380.

Skorik, Piotr Ja (1948): Očerki po Sintaksisu Čukotskogo Jazyka. Inkorporatsija (Outline of Chukchee syntax. Incorporation). Leningrad.

Smirniotopoulos, Jane C./Joseph, Brian D. (1998): "Syntax versus the lexicon. Incorporation and compounding in Modern Greek". Journal of Linguistics 34: 447–488.

Spencer, Andrew (1995): "Incorporation in Chukchi". Language 71/3: 439–489.

Zwicky, Arnold M. (1994): "What is a clitic?" In: Nevis, Joel A. et al. (eds.) (1994): Clitics. A comprehensive bibliography 1892–1991. Amsterdam/Philadelphia: xii–xx. (= Amsterdam Studies in the Theory and History of Linguistic Science. Series V: Library & Information Sources in Linguistics 22).