Polysynthesis: A Diachronic and Typological Perspective
Summary and Keywords
Polysynthesis is informally understood as the packing of a large number of morphemes into single words, as in (1) from Bininj Gun-wok (Evans, in press).
'I cooked the wrong meat for them again.'
Its status as a distinct typological category into which some of the world’s languages fall, on a par with isolating, agglutinating, or fusional languages, has been controversial from the start. Nevertheless, researchers working with these languages are seldom in doubt as to their status as distinct from these other morphological types. This has been complicated by the fact that the speakers of such languages are largely limited to hunter-gatherers—or were so in the not too distant past—so the temptation is to link the phenomenon directly to way of life. This proves to be oversimplified, although it is certainly true that languages qualifying as polysynthetic are almost everywhere spoken in peripheral regions and are on the decline in the modern world—few children are learning them today.
Perhaps the most pervasive of the traits that give these languages the impression of a “special” status is that of holophrasis, which can be defined as the (possible) expression of what in less synthetic languages would be whole sentences in single complex (usually verbal) words. It turns out, however, that there is much greater variety among polysynthetic languages than is generally thought: there are few other traits that they all share, although distinct subtypes can in fact be distinguished, notably the affixing as opposed to the incorporating type.
These languages have considerable importance for the investigation of the diachronic complexification of languages in general and of language acquisition by children, as well as for theories of language universals. The sociolinguistic factors behind their development have only recently begun to be studied in depth. All polysynthetic languages today are to some degree endangered (they are dying off at an alarming rate), and many have been poorly studied if at all, which makes their investigation before it is too late a prime goal for linguistics.
Polysynthetic languages are more than just languages with very long words. Although this is undoubtedly part of the picture (see Greenberg, 1960, for a quantitative approach putting Eskimo at the head of the field), languages like German and Sanskrit also have long words, but no one would want to call them polysynthetic. We need to be clearer at the outset as to what counts as a polysynthetic language apart from sheer length of words.
The dimension of polysynthesis has, since it was first integrated into the framework of morphological typology in the 19th century by Humboldt (1836), proved more difficult to define than the other major morphological “types” among the world’s languages. The actual term “polysynthesis” had earlier been coined by Duponceau (1819) in characterizing the Indian languages of North America—Humboldt himself called it the “incorporating” (einverleibende) type, and this is symptomatic of the problems surrounding the use of the term, since not all highly synthetic languages display (canonical) noun incorporation. For some researchers the term just refers to an extreme elaboration of the dimension of synthesis, one of Sapir’s two orthogonal typological axes, along with morphological “technique” (Sapir, 1921, pp. 120–146). These correspond more or less to Comrie’s “index of synthesis” as opposed to his “index of fusion” (cf. Comrie, 1983, pp. 42–43). For others (within the generative camp) it is an independent category or parameter with far-reaching theoretical ramifications (cf. Baker, 1996).1 It has been argued back and forth whether noun incorporation is morphological or syntactic (see Sadock, 1986). The truth may depend both on definition and on particular language family.
In Fortescue (1994) an attempt was made to enumerate the various traits that tend to cluster together to create the appearance of a distinct polysynthetic type. A simpler “rule-of-thumb” definition of a polysynthetic language can be adopted here, one that combines two essential features: holophrasis (independent “word-sentences” with bound core argument pronominals) and the possibility of more than one lexically “heavy” element (lexical or affixal) in a single such “word-sentence.” This combines an inflectional and a derivational component: bound pronominals on the one hand and more or less open-ended derivational/incorporative productivity on the other. Language families that clearly fall within this definition are listed at the end of this article Appendix 1 (those with question marks following are “polysynthetic with qualifications,” falling short in one or other of the two halves of the definition). Within this definition are both languages that display compounding of more than one lexical stem in a single word and those that do not, i.e., ones that display a form of “quasi-incorporative” construction in which lexically heavy verbalizing affixes “incorporate” noun stems (like Eskimo-Aleut). In general, there is a remarkable diversity of structure among the different languages that fit under this rubric.
Most polysynthetic languages are of the “head-marking” (or “double-marking”) type (Nichols, 1986), though some (such as Chukchi) are to a certain degree “dependent-marking,” at least as regards NPs. By “head-marking” Nichols refers to languages in which syntactic relations such as subjecthood or possession are morphologically marked on the head of a constituent (the verb in the case of a VP, the possessum in the case of an NP) as opposed to on the dependent (its argument(s) or possessor). The reverse is the case in dependent-marking, and the relationship is marked on both head and dependent in double marking.
There is a further distinction that cross-cuts the affixal/compounding distinction, namely that between polysynthetic languages that organize their morphologies in a “templatic” manner (with a fixed order of slots) and those that organize them in a “scopal” manner, with an ad hoc ordering of morphemes by scope according to expressive needs (cf. Mattissen, 2003, for the distinction). The former overcomes the natural rigidity of morphology by organizing it into a large number of “slots,” at least some of which are open to an array of (incorporated) exponents, whereas the latter allows syntax-like flexibility into the morphology itself. However, these are not watertight categories, and many polysynthetic languages combine elements of both types in different segments of the morphology (cf. Rice, 2000).
Polysynthesis primarily concerns verbs and may not involve nouns at all in highly head-marked language families like Na-Dene and Wakashan. Some polysynthetic language (like Eskimo-Aleut) can, however, produce considerable elaboration of nouns as well and may furthermore allow recursive shifts between verbal and nominal base several times within a single word. The development of nominal case or elaborate noun class/gender distinctions is not typical of polysynthetic languages but can certainly be found among them—thus nominal case in Eskimo-Aleut and noun class systems in Yimas and Adyghe, for example, where the nominal class of arguments is drawn into the verbal morphology. Noun incorporation as such is not exclusive to polysynthetic languages, nor are “heavy” verbalizing affixes; neither is the possibility of holistic single verb clauses (with bound pronominals as in Bantu and, arguably, the classical Indo-European languages). Morphological recursivity is not restricted to polysynthetic languages, but where it develops elsewhere—as in English—it is usually characteristic of highly developed written styles, which is also the case with the proliferation of nominal compounding in languages like High German and Classical Sanskrit.
Polysynthesis has come recently to be seen in terms of the more general “complexification” of grammar (Trudgill, 2011) or grammatically “mature phenomena” (Dahl, 2004). Both these authors are concerned with the sociological underpinnings of complexification—the conditions for “more and more of the same” continuing unchecked. They represent a more empirical approach to the sociolinguistic development of polysynthesis than vague suggestions made in the past as to the lifestyles of speakers of such languages (typically as “hunter-gatherers”). Trudgill discerns five principal factors leading to grammatical complexity: low degree of (external) contact, high stability, small population size, dense social networks, and a large amount of community-shared information (Trudgill, 2011, p. 146). The most tangible of these—low degree of contact and small size—will be in focus in the present article. Together these factors presuppose a degree of isolation and low demographic density, i.e., small groups spread over large areas with little contact with other groups.
1. The distribution of polysynthetic languages today
Consider first where polysynthetic languages are concentrated geographically today, namely in areas of low population density (in pre-colonial times at least). The language areas shown in Figure 1 (very crudely) are just extremes on the scale of synthesis, but the fit with areas of respectively high and low population density (represented by the distribution of black dots) is notable. A few smaller pockets of polysynthetic languages are not marked. There is also a diachronic dimension not shown directly: the population in Western Europe, for instance, was of course much lower several centuries ago and the degree of synthesis of the older languages concomitantly higher. It can be seen that areas where high demographic density = relatively analytic languages (dependent- or zero-marking) include the following:
China, Southeast Asia, northern India, West Africa, Western Europe, and central California (Penutian). Areas where low demographic density = polysynthetic languages (head- or double-marking) include most of North America, most of Meso-America, Amazonia, Chile, northeastern Siberia, northern Australia, and the Sepik valley (New Guinea).
There are, however, numerous exceptions (apart from the massive inroads of colonial languages—for instance, in the Americas—obscuring the picture), and these call for explanation. Sometimes an explanation does suggest itself, as in the case of highly polysynthetic Nahuatl, which was brought from the thinly populated north not long before historically attested times and has become less synthetic in recent times in the dense demographic situation of the central valley of Mexico today (see Canger, in press).
Areas that diverge from the expectations above include the following:
(a) areas of high density but a rather high degree of synthesis—Japan and Korea (large island/peninsula isolation), southern India (little mobility)
(b) areas of low density but a rather low degree of synthesis—southern Africa (Khoisan) (high phonological complexity), Tibet (still largely migratory)
Clearly demographic density is not the only factor behind the emergence of polysynthesis—too much special pleading is necessary to account satisfactorily for all cases like those under (a) and (b). What needs to be addressed—given the partial linkage of demographic density and degree of synthesis—is whether there is some inevitability about the tendency to complexification in areas of low demographic density, and in that case what might have stopped this “normal” development in the languages given under (b) above? And what are the factors that allowed it to develop in the languages under (a)? It is possible that a further factor that needs to be taken into account here is mobility (relatable to Trudgill’s factor of stability), with languages under (a) representing communities with relatively low mobility, those of (b) relatively high mobility.
Before returning to these central questions, let us consider a specific case illustrating the interplay between demographic density and degree of synthesis. Sherzer (1976, pp. 159–160) describes the original situation in central California, where the languages are predominantly dependent-marking, as one of intense contact between groups displaying considerable bilingualism.2 This contrasts with the northwest coast of America, where polysynthesis reigns, an area which was almost equally densely populated and mobile, but with much multilingualism. Apparently conservative attitudes to one’s language (and more hostile relations with one’s neighbors) played a role there, in an environmentally rich region of highly competitive small tribal units. Presumably this can be linked to Trudgill’s factors of internally “dense social network” and “large amount of community-shared information.” Such attitudes also play a role in the densely populated pueblo area of the Southwest, where bilingualism is rare despite much cultural diffusion (Sherzer, 1976, pp. 163–164). More expected is the complex polysynthetic morphology of languages of the American sub-Arctic (Athabaskan and Algonquian), spoken by small, highly mobile groups with little awareness of linguistic or tribal identity and meeting almost exclusively speakers of the same or similar languages. Sherzer argues that the greater degree of contact in linguistic areas generally leads to greater uniformity, but this obviously does not wipe out linguistic diversity as such (compare again the northwest coast)—it may depend on which languages one is in contact with and how similar they are to one’s own.
It is necessary to introduce here Sapir’s rather mysterious notion of “drift”—“the unconscious selection on the part of [a language’s] speakers of those individual variations that are cumulative in some specific direction” (Sapir, 1921, p. 155). The direction taken “may be inferred, in the main, from the past history of the language.” For Sapir drift was an inherent tendency toward some “type” as goal (e.g., toward polysynthesis in the case of synthetic languages or toward greater analyticity in the case of modern European languages). It could operate on different “levels,” with different strands “conspiring” or not to nudge it in a specific direction, and thus a stable endpoint may never be reached. “Knowledge of the general drift of a language is insufficient to enable us to see clearly what the drift is heading for. We need to know something of the relative potencies and speeds of the components of the drift” (Sapir, 1921, p. 163). A more up-to-date (and simpler) take on Sapir’s notion—as followed in the present article—is that there is just one global morphosyntactic drift toward greater complexity and synthesis plus one major counterdrift toward greater morphosyntactic simplicity. Clearly the former has been the major force behind the emergence of polysynthetic languages, however this came about. The question of what, in functional terms, might “drive” this drift will be returned to below.
2. Pathways into polysynthesis
Turning now to the question of how a language might become polysynthetic, it would appear that embedding in a larger geographical region where head-marking is already dominant is a prerequisite (Nichols, in press), though there may occur subsequent displacements.3 Note that this does not equate polysynthesis with head-marking: whereas head-marking is essentially a matter of inflection, polysynthesis as defined above contains an essential derivational component. There are plenty of more or less analytic head-marking languages, and, as mentioned, some polysynthetic languages may display dependent-marking features, especially on nominals (these may reflect either secondary developments or relics from an earlier stage). Thus almost the whole of North and South (and Meso-) America is characterized by a notable degree of predominantly head-marking synthesis. This is also true of northeast Siberia and the American Arctic (but with double-marking a significant areal component). The polysynthetic Sepik languages are embedded in an area of considerable head-marking synthesis that falls short of actual polysynthesis, as is also the case in northern Australia (see, respectively, Foley, in press; Evans, in press). Apparent exceptions such as the northwest Caucasus and the Kiranti languages of India and Nepal can probably be explained in terms of location in “residual zones” into which languages of varying degree of synthesis have impinged (cf. Nichols, 1992, p. 273). In the case of Sora, this is part of a wider head-marking Munda presence in eastern India, much reduced by inroads from Indo-European and Dravidian languages.
Head-marking languages and families appear to follow three main pathways into polysynthesis (Fig. 2) according to whether they already at an early stage display: (a) productive verbalizing affixes but little or no compounding; (b) compounding of various kinds but no productive verbalizing affixes; or (c) clause chaining/verb serialization in a fixed order. Mattissen (2003) distinguishes types (a) and (b) but subsumes (c) under (a) (she also has hybrid combinational types). Naturally these are idealized types. The idea is that languages of the (not yet polysynthetic) type indicated in the central boxes will complexify naturally toward polysynthesis as they move into or became absorbed in a developing head-marking area. The target would then be the major types of polysynthesis indicated in the boxes on the right of the figure.
Some of the language families that appear to reflect the three distinct pathways are the following:
Note that this is not to claim that languages of type (a) might not already have had some verbalizing affixes before being “captured” by areal head-marking, or that the verbalizing affixes of type (b) might not at some earlier stage have developed from verb + noun compounding, or that the clauses of type (c) might not already have included incorporates. The claim is only that when these languages became “fixed” on their course toward polysynthesis they were already predominantly of the three types indicated.
Although reconstructing earlier stages of complex languages with little descriptive time depth is fraught with difficulties, there may nevertheless be synchronic symptoms of the paths followed. Thus type (a) (via incorporation) will not usually develop heavy verbalizing affixes reduced from verbs, while type (b) (via heavy verbalizing affixes) will not usually produce reduced nouns as (quasi-) incorporates (and eventually affixes). Type (c) may, however, develop either. Symptomatic of type (c) is holophrastic verb-serializing where one of the verbs will typically be “light” and semantically subordinate to the superordinate one (eventually becoming purely adverbial). There are cases where one of the verb roots (the lighter one) will have been fully grammaticalized but have left behind material originally affixed to it (opaque or still to some degree transparent) between it and the main verb—Givón’s analysis of Na-Dene discussed below is a case in point.
There may also develop hybrid types with secondary developments associated with other types, e.g., by type (b) languages drawing upon light lexical verbs to add to their stock of verbalizing or adverbial affixes. Languages of the “bipartite stem” type (as typical of many languages in southern Oregon/northern California) in which stems may consist of combined bound forms alone (e.g., prefixes for instrument or causal agent and suffixes for location or direction of movement) may represent a further stage of the compounding type (DeLancey & Golla, 1997, p. 193, suggest as much for proto-Penutian). Algonquian languages may represent a midway point in this development, with the minimal verbal predicate consisting of an obligatory initial stem plus obligatory final element (either affix or, secondarily, another stem), which determines the grammatical status of the whole (e.g., its transitivity). Northern Hokan languages of this type appear to be purely affixing today (but lack the heavy verbalizing affixes symptomatic of type (b)).
2.1. Examples of the three main pathways
Let us look at a specific case of each of the three main pathways. First, on the face of it, the most straightforward pathway into polysynthesis is that displayed by Iroquoian languages, all of the “compounding,” noun-incorporating type and situated in the middle of the large North American head-marking area. Chafe (2000) discusses the “florescence” of noun incorporation in that family. He traces the full development in the northern languages from simpler roots reflected in Cherokee, which he sees as lying closer to the proto-language. Mithun (1984) hypothesized successive stages in the development of noun incorporation across the languages of the world, starting with the universal tendency of indefinite objects to cluster tightly with verbs even before full morphological coalescence is achieved and the noun loses its syntactic independence (as in Hungarian and Turkish). In Cherokee only body parts and some classifiers and clothing nouns are incorporated at Mithun’s stages of compounding Type I and “case manipulating” (oblique promoting) Type II. In the northern branch of the family the highest, “classificatory” Type IV is attained, passing via pragmatic “discourse manipulating” Type III. The latter is reached when nouns expressing background or incidental information in the clause are incorporated in the verb. The implication is that the northern languages have developed out of a stage resembling Cherokee. Chafe contrasts simple body part incorporation in Cherokee as in kakvʔskwo.ʔa “I am washing my face” (-kvh- “face,” -hsk- “head,” -wo.- “wash”) with “pragmatic” incorporation in northern Iroquoian as in the following example from Seneca (2000, p. 51).4 This form would only have been used if it was intended to downplay the object “shoes”; if the shoes were “newsworthy” information they would have been expressed by an independent word.
'We went and got shoes'
It is probably not possible to reconstruct any pre-incorporating stage for the family, which undoubtedly lies very far in the past. The choice to follow the compounding head-marking pathway must have been a very early one since the family as a whole is so verb-heavy that compounding is limited to noun incorporation in verbs (no nominal compounding as such). This is widely utilized for forming complex nominal expressions too (descriptive verbal expressions in referential use).
Eskimo-Aleut, earlier stages of which have been reconstructed in some detail, illustrates the second, affixal pathway. It is in some ways a typical affixing-only type of polysynthetic language but is unusual in having a fairly complex case system and displays double- rather than head-marking. This may betray its Uralic(-like) origins—it chose ultimately to develop along a more head-marked way than other northern Siberian languages like the Ugric and Samoyedic ones, which continued by and large along the dependent-marked Uralic one. Northern Samoyedic languages have, however, developed optional double-marking from Uralic dependent-marking origins (cf. Nikolaeva, 2014, p. 143 for Nenets) and are on their way toward (but have not yet reached) affixing type polysynthesis with “heavy” verbalizing suffixes. The “nudge” to Eskimo-Aleut that pushed it the way it went was probably the development of the ergative clause via a focal nominalization construction, as described in Fortescue (1999). This was a “markedness shift” determined by linguistic/functional system pressure.5 It was based on a possessed passive participial (cf. Trask’s 1979 nominalizing “type B” pathway into ergativity). The result was the following type of clause in West Greenlandic, where REL is relative/possessive case:
piniartu-p tuttu tuqup-paa
hunter-REL reindeer.ABS kill-3SG>3SG.IND
'The hunter killed the reindeer.'
The transitive inflection here reflects an earlier passive participial plus (3rd person) possessor ending, as in tuqut-(t)a-a “his killed thing,” with the t replaced by the p of the corresponding intransitive paradigm (an Inuit development). So, roughly, “the hunter’s killed thing (was) the reindeer.” Note crucially that the transitive verbal inflections are taken from the nominal possessive inflections (here 3SG -a). In other Inuit dialects (and Yupik) the participial source of the verbal paradigm is still quite transparent. The development of the transitive ergative clause resulted in the “double-marking” of the family, characterized by verbal polysynthesis and nominal case co-existing and interacting. Nominal morphology was co-opted in this way into clausal function (a common enough phenomenon within Uralic). Note that Eskimo languages (especially the Yupik branch) are still developing new tense/mood paradigms from nominalizations (participial forms).
The earlier nominative/accusative active clause type remained when the object was indefinite and the verb intransitive (or “anti-passivized” if inherently transitive), the subject in the absolutive and the object in an oblique case (instrumental = accusative), as in:
piniartuq tuttu-mik tuqut-si-vuq
hunter.ABS reindeer-INST kill-ANTIP-3SG.IND
'The hunter killed a reindeer.'
But this “choice” of ergativity must have been relatively late and was in all likelihood preceded by a dependent-marking, nonergative stage. In other words, hypothetical “early choices” such as dependent-marking (normally not leading to polysynthesis) can later be annulled or reversed. If we are lucky such historical shifts may be reconstructable (as in the case of Eskimo-Aleut), but this will of course not always be possible. Note that the head-marking tendency increases across Siberia from west to east, the direction the ancestors of Eskimo-Aleut speakers presumably took from the Old World into the New. In Alaska they would have met and mixed with speakers of probably already head-marking, polysynthetic Na-Dene languages.
Eskimo-Aleut must also have increased an earlier more modest stock of derivational affixes, e.g., by producing “verbalizer” affixes from the incorporation of auxiliary verbs. Thus Eskimo *-(ŋ)u- “be” (with epenthetic ŋ after vowel stems) has been retained in Aleut as a-/u-. Compare WG kina-a-vit? (“who are you?”) with Atkan Aleut kiin a-Xt, the former with the affix -u-/-a- corresponding to the Aleut independent auxiliary. The Aleut auxiliary is ingredient in various TAM constructions where Eskimo has affixes. That this particular combination is relatively recent in WG can be seen by the fact that it is the singular form of the interrogative stem that it is combined with—it could have been the plural form kikkut with an appropriate inflection. Incorporated auxiliaries could in theory have drawn their own following (e.g., aspectual or adverbial) affixes with them. The acquisition of “heavy” verbalizing affixes would have opened up the way to morphological recursivity.6 Note that Eskimo-Aleut retains a clear distinction between nominal and verbal stems (and affixes), which can be exploited morphologically, unlike in affixing-only but nonrecursive Wakashan. Its stock of affixes has been increasing ever since through the coalescence of new lexicalized combinations and the splitting off of allomorphs in particular contexts (cf. Fortescue, 1985, pp. 194–197). This process is likely to have been ongoing for a long time and would have continued after the formation of the ergative clause since it is still in progress today, as in the following example from WG. Here -niit- is a lexicalized combination of locative case -ni and semi-independent stem it- “be (in a place)” (the degree of lexicalization varies from dialect to dialect). Note that this “new” affix takes a whole NP as its base here (not a possibility in Uralic languages, for example).
nirrivi-up qaa-va-niip-put baaja-t pingasut
table-REL (on)top-3SG-be.on-3PL.IND pilsner-PL three
'There were three pilsners (beers) on the table.'
Na-Dene languages can be used to illustrate the third, clause-combining pathway into polysynthesis, although they appear on the surface to be of an affixing (prefixing) type today (with the addition of more recent postposition incorporation at the periphery of the verb). These highly conservative languages display diachronic layering that retains traces of very early structures (in the affixes closest to the verbal stem) which have become greatly eroded and reanalyzed. The discontinuous, “entangled” ordering of its prefixes (inflectional and derivational) strongly suggest an earlier stage of the verb-serializing, clause-combining type (c). Givón (1999, pp. 122 ff.) shows how a number of adverbial prefixes in Tolowa (Oregon Athabaskan) reflect earlier verbs, via old postpositions (not those incorporated transparently in left-most position within the verb template), and similarly for perfect(ive) aspect prefixes (1999, pp. 143–149). This development can be schematized as below:
An example of an adverbial prefix (“inceptive, upward, out”) that can be related to a lexical verb (xa- “raise, carry raised”) is seen in (6).
'I start to sing'
(7) is an example of a perfective prefix ghu- that has left traces of its verbal origin in the idiosyncratic effect upon a following 1st person subject prefix and (crucially) in the nasalization of its vowel in the case of a 3rd person singular subject (usually marked by zero). This suggests an earlier nasal perfective suffix on a light verb ghu- (cf. gha- “come”).
It is possible that Na-Dene and Yeniseian, if they are related as argued by Vajda, share this early pathway into polysynthesis, with light verb plus heavy verb clauses combined or serialized (see Vajda, in press, as regard to Yeniseian).
2.2. Examples from other polysynthetic families
Looking at some rather less transparent cases—and always bearing in mind the proviso that the reconstruction of early stages of polysynthetic languages is by necessity rather speculative—consider now the Algonquian family. It appears today to show some features of affixing type (b) as regards its governing “finals” (in meanings like “being,” “having,” “moving,” or “arranging”) and its derivational recursivity, as in the example of “secondary” derivation below (Wolfart, 1996, p. 426) where VTA and VAI are markers of the transitivity and animacy of the verbal stages of the derivation.
'be a habitual wife-watcher' (i.e., one who keeps an eye on his wife)
But quite different from Eskimo-Aleut are its elaborate compounding of stems, in particular its body part and classifier “medials” (the latter in meanings such as “wood,” “stone,” “cloth,” or “rope,” referring either to the shape or consistency of the subject/object or to a locational or instrumental function). These attach to any kind of root (verbal or nominal), often correspond to independent nouns, and must have developed from incorporates. Can we be sure which process came first historically? The governing finals (obligatory, unlike medials) could once have been independent verbs developing into suffixes after the language had moved some way toward fully fledged polysynthesis along the compounding pathway. Evidence as to what might have happened can be seen in transitive finals with meanings like “by hand,” “by vision,” “by sticklike tool,” “by heat,” and “by mental action” (Wolfart, 1996, p. 429). There is obviously an overlap in meaning with the classifier medials. In fact, the transitive finals of this type look like the fusion of a classificatory or body part medial with an abstract transitive final (finals generally determine the transitivity/valency of the complex word). If this is correct it would support the scenario suggested above, i.e., the primary compounding pathway and the secondary development of “heavy” transitive finals, further elaborated by the secondary (recursive) introduction of independent verb stems as “derived finals.”
The situation sketched above can be compared with that of the Wakashan languages, which are clearly of the affixing type today, with no compounding of any kind and a wide range of lexically heavy “governing” suffixes, as in the following from Nuuchahnulth (Nakayama, 1997, p. 169), where PRO is a proverb “peg” element and INV is “inverse.”
'They used to live with us.'
Wakashan languages also have a wealth of locative-directional lexical affixes lacking in both Algonquian and Eskimo-Aleut. These “restrictive” (nongoverning) lexical affixes display typological anomalies that strongly suggest areal influence (if not direct borrowing) from neighboring Salishan, where lexical affixes, all of the restrictive type, do not show such anomalous behavior (see Fortescue, 2009). This supports the contention that the direction of the original pathway into polysynthesis in this family was of the affixing type (b), with later areal developments introducing affixes of the restrictive type. This is not to deny that at least some Wakashan governing affixes ultimately derive from lexical verbs, as Mithun (1984, pp. 887–889) suggests, though the evidence for this has largely been lost through time, much as in the case of Eskimo-Aleut.
The earliest stages of the development of polysynthesis in a family may be obscured by areal developments or by movement into a large area characterized by a different structural type, so the apparent choice of pathway (shared with dominant neighbors) does not necessarily represent the “earliest” one. This may even effect basic marking type—despite this generally being “conservative” and tenacious, as Nichols claims.
The case of Chukotian (Chukchi, Koryak, Alutor, and Kerek), with its “onion-skin” morphology built up of successive circumfixed layers, is particularly revealing in this respect. It displays an extreme form of incorporating polysynthesis that arguably reveals relatively recent florescence. Its derivational affixes are also more lexically transparent and fewer in number than those of “old” polysynthetic Eskimo-Aleut. For diagnostics of “newer” vs. “older” polysynthesis (essentially a matter of transparency), see Fortescue (2013). The circumfixes appear to have developed when original 1st/2nd person subject suffixes had eroded to the point where newer prefixes (from pronouns) were added. At the same time, comitative case circumfixes appeared on case-marked nouns with prefixation of an adverb meaning “together,” which is also ingredient in derived circumfixes meaning “having X” (the source of the past/passive participial below). Prior to the development of the ergative clause, the starting point would have been the purely suffixing nominative-accusative clause type typical of northern Asia. Like them, it would have been purely dependent-marking, as it still largely is today despite the ergativity (egregiously absent in distantly related Itelmen/Kamchadal).
But this ergativity is almost exclusively manifest in its case marking and appears to be due to influence from Eskimo-Aleut on the emerging perfect tense paradigm, based in turn on a past/passive participial, a similar source to the phenomenon in Eskimo-Aleut mentioned above (cf. Fortescue, 1997, for a detailed discussion of this diachronic process in Chukotian). The linguistic consequences and socio-cultural setting of long-term contact between Eskimos and Chukchis have been studied in some detail.7
Derivational extensions of the uninflected verbal stem by circumfixes meaning “want to,” “negative,” and “make” could appear “bracketed” within the circumfixed inflectional material—the “want to” one was further grammaticalized to produce a “future” paradigm joining those already forming the outer “onion skin.” The derivational and incorporative possibilities must have been present before the development of the ergative clause—these languages have reached Mithun’s Type III in the evolution of incorporation, and there are some traces of stages I and II in remotely related Itelmen (Mithun, 1984, p. 883). Nouns must have already been “incorporatable” within verb-like/participial circumfixes like Chukchi ɣa-wopqə-len “having a moose (in it—e.g. a forest),” from wopqə “moose” (and cf. the comitative case prefix). Nominal compounding (modifiers plus head) is also much developed in the family, and complex NPs can in turn be incorporated.
The following example from Koryak (Kurebito, in press) illustrates the end result of the “florescence” of polysynthetic incorporation on this compounding but dependent-marking basis.
'I killed a reindeer as material for a winter coat.'
Such cases must be distinguished from the converse, of head-marking languages being affected by dominant dependent-marking neighbors, which may also occur. Head-marking languages like Ket and Tariana have developed nominal case systems from moving into the proximity of surrounding dependent-marking neighbors. In the case of Ket this was probably a matter of relatively recent complexification via bilingualism with Samoyedic and other dependent-marking neighbors (cf. Vajda, in press). In the case of Arawak language Tariana, it has developed, besides case distinctions, verb compounding patterns borrowed from dominant neighbor family Tuconoan (aktionsart enclitics with a lexically “heavy” adverbial meaning). These contrast with a few grammaticalized aspectual suffixes from native serial constructions (Aikhenvald, 2000). Whether this means that Tariana (and other northern Arawak languages) developed polysynthesis via the clause-combining pathway as opposed to the compounding one (as suggested by the southern Arawak languages displaying noun incorporation) is uncertain.
Transparency of agglutination cuts across degree of synthesis (recall Sapir’s orthogonal axes) and may simply reflect how recent the development from concatenative morphology toward more synthetic fusional type has taken place. Note that polysynthetic morphology (whether of the templatic or hierarchical/scopal type) may remain strictly agglutinative for a long time (with fusional developments in limited regions of the complex word, e.g., Eskimo inflectional endings).8
In general, there is no necessity to suppose that the head-marking “choice” always precedes—or follows—some other early choice such as that to take a compounding as opposed to a single-stem affixing pathway forward. Thus Givón (2009, p. 95) posits an early bifurcation between embedding and verb-serializing languages in the development of complex clauses. The further stage of polysynthetic words with multiple verb stems may derive from either starting point. In the case of Iroquoian languages, the model illustrated in Figure 2 suggests that the proto-language was probably of a compounding type already before it was drawn into a head-marking area (or before one developed around it). On the other hand, both head- and dependent-marking languages may exploit lexical compounding as a secondary development, the latter especially through nominal compounding, the former through “incorporating” and “clause-combining” polysynthetic structures. Since compounding cuts across the marking type distinction, head-marking can hardly always be an “earliest” choice itself, but neither can compounding, or indeed such further “choices” as flectional vs. agglutinative type, or of basic word order type, all of which are largely independent of degree of synthesis (though there may be some weak statistical trends toward correlation). This is not to say that certain further typological consequences and dependencies may not fall out from such choices. It remains to be seen to what degree head-marking and the global drift toward complexification correlate.
3. Halting or reversing the drift toward greater synthesis
In the context of his discussion of the historical development of complex syntax from simple parataxis, Givón (2009, p. 120) suggest that the further compacting of syntax into complex verbal morphology may serve the purpose of increasing automaticity (and speed of processing), a matter of “chunking.” This may well be a major factor behind the drift toward increasing head-marking complexity. From this perspective, polysynthetic languages, far from being holistic in a “primitive” or child-like sense, may represent highly evolved forms of grammar—though hardly the best suited for interlanguage communication. Polysynthetic languages are quintessential exemplars of “esoterogenic” languages in which group identity and orality is central, as opposed to “exoterogenic” ones, which tend to simplify in the service of greater communicability, ultimately as lingua francas (cf. Crowley, 2000). For those languages that do follow the drift toward accumulating complexity, unchecked by the need to communicate with powerful neighbors, one can ask: What early choices determine whether they will follow a polysynthetic as opposed to some other pathway forward? The extreme elaboration of nominal case and/or classification systems is a priori unlikely to lead to polysynthesis, and neither is the packing of complex phonological/semantic distinctions into monosyllables (like Khoisan, but also Nilotic Dinka, and to some degree Sinitic). Phonological complexity in general cuts right across the synthetic/analytic divide—thus in highly polysynthetic North America, phonology is complex in the west, simpler in the east. If there is only one global kind of positive drift as suggested above, that toward increasing complexity and synthesis, does this imply that all languages start from a more analytic/isolating stage by simple incremental steps? Or can there be historical fluctuation, with simplifying counterdrift coming into play at certain historical periods but not at others in the development of a language? If this does occur it is not unreasonable to think that it will be relatable in principle to fluctuations in Trudgill’s five factors, for example, in population density (and mobility).
Figure 3 presents a highly schematic view of what can happen when the drift toward greater synthesis is checked or even reversed by meeting some dominant language(s) of a lesser degree of synthesis. Less compact arrows = isolating, more compact ones = moderately synthetic (agglutinating), and most compact = fusional or polysynthetic, so the left-to-right axis shows drift toward complexification. Contact (involving accommodation by adult speakers) with a less synthetic language (on the right) can lead to halting or reversing that drift—but so can increased demographic density and openness of contact with speakers of other dialects of the same language/family. There is no reason why a particular language family should not encounter such episodes more than once.
Within Australia, for example, the Pama-Nyungan (suffixing) languages have taken the dependent-marking pathway toward elaborate case systems, whereas the non–Pama-Nyungan languages (mostly prefixing) have taken the head-marking one toward polysynthesis. This may have hindered the development of polysynthesis in the Pama-Nyungan languages despite a considerable amount of (agglutinative) synthesis. Whereas most of the non–Pama-Nyungan languages appear to have taken the compounding pathway into polysynthesis, South Daly languages appear to have taken a clause-combining pathway (Evans, in press).
Just to take one apparent glaring exception to the correlation of low demographic density with polysynthesis discussed above, consider the case of the phonologically complex Khoisan languages. It is likely that expanding the information that can be packed into monosyllables (exploiting tone and vowel quality as well as clicks) was an earlier choice than concatenative morphological elaboration, which is fairly synthetic only in the Khoe languages of the area (it is almost polysynthetic in Nama, which has less phonological complexity than the San languages—cf. Güldemann & Vossen, 2000). In fact early phonological complexity could have blocked the development of any great morphological complexity in the San languages, whereas the Khoe languages (which may not even be genetically related) could have developed phonological complexity through areal influence from San after they were already fairly well down the road toward morphological complexity—it seems at all events unlikely that a San(-like) language would spontaneously develop in that direction with so much complexity already packed into monosyllables.
4. The acquisition of polysynthetic languages
It is a commonplace in child language development that children at an early stage may follow either of two main pathways beyond infantile holophrasis as they develop toward adult language: some children concentrate on elaborating names for things, with a rapid increase in vocabulary but little production of whole clauses (and little increase in phonological fluency), while others are able to produce fluent but stereotyped clauses with adult-like intonation from an early stage but with little elaboration of vocabulary (cf. Nelson, 1973, on the difference between the “expressive” and the “referential” style characterizing different groups of children around the age of 24 months). Applying this to adult languages (purely by analogy), one might say that while some opt for a more analytic “nouny” pathway forward into complexity (like San), others opt for a more “verby” one (polysynthetic languages).9 This parallels the distinction between dependent-marking and head-marking languages (those developing elaborate case and noun class systems and those capitalizing on the synthesis of verbal predicates). Note that a language may be highly polysynthetic (“verby”) while having only a very limited number of verb roots but elaborate derivational resources (as with Na-Dene). What seems generally true is that polysynthetic languages have a limited stock of lexical roots/stems, counterbalanced by highly productive derivational/incorporative resources.
As a child learns to master a polysynthetic language, there is no inherent limit to the combinatorial variability of the morphological verb-sentence, which can be quite syntax-like. In the case of “affixing” polysynthetic languages, any lexical item can serve as the base for derivational processes (which may be further extended by recursion). In contrast, children learning more analytic “nouny” languages will reach a point where all the necessary lexical word forms are acquired in order to deploy the adult language in a socially satisfactory manner—i.e., further morphological development is “capped” by the lexicon. Nominal lexicon may also reach that point for learners of polysynthetic languages, of course, but their ability to produce new verb forms will continue throughout life, limited purely by utility and individual verbal ability. It is these skills of imitation, interpretation, and innovation that are passed on from generation to generation and which may nudge the language forward gradually to greater and greater complexity, “more and more of the same.”
It is instructive to compare the ways in which children do in fact learn polysynthetic as opposed to analytic languages, although there is still to date relatively little documentation of the former. There are several studies of one language family that has taken the affixing pathway into polysynthesis, namely Eskimo-Aleut (as in Fortescue & Lennert Olsen, 1992), and there is at least one study of the acquisition of an incorporating-type language (Mithun, 1989). For the clause-combining type see Nordlinger, Wigglesworth, Kelly, and Forshaw (in press). Although the “ways in” differ in these three cases, this only indirectly reflects the diachronic picture since phonological and other factors besides morphological type also play an important role (stressed syllables in Mohawk but not in West Greenlandic, for example). The existing studies do all show a “precocious” attention to morphology prior to syntax, however.
Appendix 1: List of Polysynthetic Language
Northern Hokan (Palaihnihan, Shastan, etc., but not Pomoan)
Chukotian (not Itelmen)
Sora (and other South Munda)
Chintang (and other Kiranti)
Northwest Caucasian (all)
Tukang Besi (Sulawesi)12
South Daly River
Dixon, R. M. W., & Aikhenvald, A. Y. (Eds.). (2002). Word. A cross-linguistic typology. Cambridge, U.K.: Cambridge University Press.Find this resource:
Evans, N., & Sasse, H.-J. (Eds.). (1999). Problems of polysynthesis. Studie Typologica 4. Berlin: Akademie Verlag.Find this resource:
Givón, T. (2009). The genesis of syntactic complexity: Diachrony, ontogeny, neuro-cognition, evolution. Amsterdam: J. Benjamins.Find this resource:
Mahieu, M.-A., & Tersis, N. (Eds.). (2000). Variations on polysynthesis, the Eskaleut languages. Amsterdam: John Benjamins.Find this resource:
Mahieu, E., Chen, C.-H., Geber, D., & Manouilidou, C. (2009). Noun incorporation and its kind. Lingua, 119(2).Find this resource:
Mithun, M. (1999). The languages of Native North America. Cambridge, U.K.: Cambridge University Press.Find this resource:
Nichols, J. (1986). Head-marking and dependent-marking grammar. Language, 62, 56–119.Find this resource:
Rowicka, G. J., & Carlin, E. B. (Eds.). (2006). What’s in a verb? Studies in the verbal morphology of the Americas. Utrecht: LOT (Netherlands Graduate School of Linguistics).Find this resource:
Aikhenvald, A. (2000). Areal typology and grammaticalization. The emergence of new verbal morphology in an obsolescent language. In S. Gildea (Ed.), Reconstructing grammar (pp. 1–37). Amsterdam: John Benjamins.Find this resource:
Andersen, G. (in press). Sora. In M. Fortescue, M. Mithun, & N. Evans (Eds.), The Oxford handbook of polysynthesis. Oxford: Oxford University Press.Find this resource:
Baker, M. C. (1996). The polysynthesis parameter. New York: Oxford University Press.Find this resource:
Canger, U. (in press). Nawatl. In M. Fortescue, M. Mithun, & N. Evans (Eds.), The Oxford handbook of polysynthesis. Oxford: Oxford University Press.Find this resource:
Chafe. W. (2000). Florescence as a force in grammaticalization. In S. Gildea (Ed.), Reconstructing grammar (pp. 39–64). Amsterdam: John Benjamins.Find this resource:
Comrie, B. (1983). Language universals and linguistic typology (2d ed.). Oxford: Basil Blackwell.Find this resource:
Crowley, T. (2000). Simplicity, complexity, emblematicity, and grammatical change. In J. Siegel (Ed.), Processes of language contact: Studies from Australia and the South Pacific (pp. 175–193). Montreal: Les Éditions Fides.Find this resource:
Dahl, Ő. (2004). The growth and maintenance of linguistic complexity. Amsterdam: John Benjamins.Find this resource:
DeLancey, S., & Golla, V. (1997). The Penutian hypothesis: Retrospect and prospect. International Journal of American Linguistics, 63, 171–202.Find this resource:
Duponceau, P. S. (1819). Report of the corresponding secretary to the committee of his progress in the investigation of the language of the American Indians. January 12. Transactions of the Historical and Literary Committee, 1.Find this resource:
Evans, N. (in press). Polysynthesis in Northern Australia. In M. Fortescue, M. Mithun, & N. Evans (Eds.), The Oxford handbook of polysynthesis. Oxford: Oxford University Press.Find this resource:
Foley, W. (in press). Polysynthesis in New Guinea. In M. Fortescue, M. Mithun, & N. Evans (Eds.), The Oxford handbook of polysynthesis. Oxford: Oxford University Press.Find this resource:
Fortescue, M. (1985). The degree of interrelatedness between Inuit dialects as reflected by percentages of shared affixes. International Journal of American Linguistics, 51(2), 188–221.Find this resource:
Fortescue, M. (1994). Polysynthetic morphology. In R. E. Asher et al. (Eds.), The encyclopedia of language and linguistics (Vol. 5, pp. 2600–2602). Oxford: Pergamon Press.Find this resource:
Fortescue, M. (1997). Eskimo influence in the formation of the Chukotkan ergative clause. Studies in Language, 21(2), 369–409.Find this resource:
Fortescue, M. (1998). Language Relations across Bering Strait. London: Cassell.Find this resource:
Fortescue, M. (1999). The rise and fall of polysynthesis in the Eskimo-Aleut family. In N. Evans & H.-J. Sasse (Eds.), Problems of polysynthesis (pp. 257–275). Studie Typologica 4. Berlin: Akademie Verlag.Find this resource:
Fortescue, M. (2009). Lexical suffixes and the position of Proto-Wakashan within the Northwest Coast linguistic area. Canadian Journal of Linguistics, 54(1), 27–42.Find this resource:
Fortescue, M. (2013). Polysynthesis in the Arctic/Sub-Arctic: How recent is it? In B. Bickel, L. A. Grenoble, D. A. Peterson, & A. Timberlake (Eds.), Language typology and historical contingency. Festschrift for Johanna Nichols (pp. 241–264). Amsterdam: J. Benjamins.Find this resource:
Fortescue, M., & Lennert Olsen, L. (1992). The acquisition of West Greenlandic. In D. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 3, pp. 111–219). Mahwah, NJ: Lawrence Erlbaum.Find this resource:
Givón, T. (1999). Internal reconstruction: As method as theory. In S. Gildea (Ed.), Reconstructing grammar (pp. 107–159). Amsterdam: John Benjamins.Find this resource:
Givón, T. (2009). The genesis of syntactic complexity: Diachrony, ontogeny, neuro-cognition, evolution. Amsterdam: J. Benjamins.Find this resource:
Greenberg, J. H. (1960). A quantitative approach to the morphological typology of language. International Journal of Linguistics, 26, 178–194.Find this resource:
Güldemann, T., & Vossen, R. (2000). Khoisan. In B. Heine & D. F. Nurse (Eds.), African languages, an introduction (pp. 99–122). Cambridge, U.K.: Cambridge University Press.Find this resource:
Humboldt, W. von. (1836). Über die Verschiedenheit des menschlichen Sprachbaues und ihren Einfluss auf die geistige Entwickelung des Menschengeschlechts. Berlin: Königlische Akademie der Wissenschaften.Find this resource:
Kurebito, M. (in press). Polysynthesis in Koryak. In M. Fortescue, M. Mithun, & N. Evans (Eds.), The Oxford handbook of polysynthesis. Oxford: Oxford University Press.Find this resource:
Mattissen, J. (2003). Dependent-head synthesis in Nivkh. Amsterdam: John Benjamins.Find this resource:
Mithun, M. (1984). The evolution of noun incorporation. Language, 60(4), 847–894.Find this resource:
Mithun, M. (1989). The acquisition of polysynthesis. Journal of Child Language, 16, 285–312.Find this resource:
Nakayama, T. (1997). Discourse-pragmatic dynamism in Nuu-chah-nulth (Nootka) morphosyntax (Unpublished doctoral dissertation). University of California, Santa Barbara.Find this resource:
Nelson, K. (1973). Structure and strategy in learning to talk. Monographs of the Society for Research in Child Development, 38 (serial no. 149).Find this resource:
Nichols, J. (1986). Head-marking and dependent-marking grammar. Language, 62, 56–119.Find this resource:
Nichols, J. (1992). Linguistic diversity in space and time. Chicago: University of Chicago Press.Find this resource:
Nichols, J. (in press). Polysynthesis and head-marking. In M. Fortescue, M. Mithun, & N. Evans (Eds.), The Oxford handbook of polysynthesis. Oxford: Oxford University Press.Find this resource:
Nikolaeva, I. (2014). A grammar of Tundra Nenets. Berlin: Mouton de Gruyter.Find this resource:
Nordlinger, R., Wigglesworth, G., Kelly, B., & Forshaw, B. (in press). The acquisition of Murrinh-Patha. In M. Fortescue, M. Mithun, & N. Evans (Eds.), The Oxford handbook of polysynthesis. Oxford: Oxford University Press.Find this resource:
Rice, K. (2000). Morpheme order and semantic scope: Word formation in the Athapaskan verb. Cambridge, U.K.: Cambridge University Press.Find this resource:
Sadock, J. (1986). Some notes on noun incorporation. Language, 62, 19–37.Find this resource:
Sapir, E. (1921). Language. New York: Harcourt, Brace & World.Find this resource:
Sherzer, J. (1976). Areal linguistics in North America. In T. A. Sebeok (Ed.), Native languages of the Americas (Vol. 1, pp. 121–173). New York: Plenum Press.Find this resource:
Trask, R. (1979). On the origins of ergativity. In F. Plank (Ed.), Ergativity (pp. 385–404). London: Academic Press.Find this resource:
Trudgill, P. (2011). Sociolinguistic typology. Oxford: Oxford University Press.Find this resource:
Vajda, E. (in press). Ket. In M. Fortescue, M. Mithun, & N. Evans (Eds.), The Oxford handbook of polysynthesis. Oxford: Oxford University Press.Find this resource:
Wolfart, H. C. (1996). Sketch of Cree, an Algonquian language. In W. C. Sturtevant & I. Goddard (Eds.), Handbook of North American Indians (Vol. 17, pp. 390–439). Washington, DC: Smithsonian Institution.Find this resource:
(1.) Baker argues specifically for a head movement analysis of noun incorporation, which he regards as crucial for polysyntheis. Note that his strict definition of polysynthesis excludes affixing-only languages like Eskimo-Aleut and Wakashan.
(2.) Within the central Penutian languages of the Californian linguistic area (DeLancey & Golla, 1997), see the development from a dependent-headed, verb-ablauting proto-language much like Yokuts to the somewhat more complex peripheral Penutian languages today.
(3.) Nichols regards marking type—the morphological marking of grammatical relations either on syntactic heads or dependents within a construction—as a conservative, stable feature of language, perhaps the deepest, most tenacious typological trait of all those distinguishing languages (Nichols, 1986, p. 89).
(4.) AND is “andative” and EX.PL.AGT is exclusive plural agent. Chafe (2000) discusses the apparent divergence from the successive ordering of Mithun’s stages that Cherokee represents, since it skips Type III but displays some classificatory incorporation. He suggests a possible reversal of the order of Types III and IV (p. 50).
(5.) A markedness shift involves the reanalysis of a “marked” (more complex and/or contextually specific) construction as unmarked, e.g., a passive construction becoming the norm for active sentences as the old active type becomes limited in distribution.
(6.) Particularly significant are its ubiquitous combinations of participial nominalizations plus existential verbalizers of the “be” and “have” kind.
(7.) Thus one can differentiate between the situation where Chukchi reindeer-herding fathers took Eskimo wives inland and the children grew up ultimately learning Chukchi as their principal language but influenced by their mother’s native language, and, on the other hand, the case of less successful reindeer hunters who came down to the coast and lived as itinerant traders among sedentary marine-hunting Eskimos. By and large it was adult Eskimos who had to learn Chukchi to communicate with them, and gradually Eskimo receded as more and more Chukchis came down to the coast, learning the Eskimo way of life and forming their own communities interspersed with the Eskimo ones. A great many coastal Chukchis and Koryaks are probably Eskimos who have shifted language (Fortescue, 1998, p. 194).
(8.) “Fusional” in the sense of complex morpheme boundary adjustments and in the sense (as here) of portmanteau morphemes need of course to be distinguished. Some polysynthetic languages do not display either—Ainu is a case in point (possibly through Japanese influence).
(9.) Whether there is some social/cultural correlate of the choice of a “nouny” as opposed to a “verby” route forward is an open question—for instance, if greater value is given to fluent, stereotyped speech patterns in some cultures, as opposed to more weight given to the elaboration of nominal lexicon for naming objects and places in others.
(10.) And perhaps other lowland Mayan languages with noun incorporation. All of these are only mildly polysynthetic.
(11.) Tariana, under the influence of neighboring Tucanoan, does not mark pronominal objects of transitive verbs on the verb (unlike most Arawak; cf. Aikhenvald, 2000). So Tariana probably qualifies as “polysynthetic with qualifications.”
(12.) And possibly some other Austronesian languages of southwest Sulawesi, the northern Philippines, and Taiwan.