12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 1 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

Subject

Key-Topics

DOI:

17. Functional Perspectives on Syntactic Change

MARIANNE MITHUN

Linguistics

»

Historical Linguistics

syntax

10.1111/b.9781405127479.2004.00019.x

Functionalist approaches to linguistics rest on the fundamental assumption, underlying a broad spectrum of
work, that language is shaped by its use. Functionalism represents a point of departure rather than a unified
theory or codified model of language, but it does have important theoretical implications. It implies that the
ultimate goal of linguistics goes beyond description (as in structuralism) and even generalization (as in
typology) to explanation of an inclusive kind.

Of course much modern linguistic theory seeks to be explanatory in some sense. Under some approaches,
explanation has been framed chiefly in terms of theory-internal consistency. A model of language is
constructed and described in terms of abstract, inherent structural principles. Individual constructions are
then explained by their conformity with the principles. Functional explanations have tended to be wider
ranging, encompassing both language-internal and language-external considerations. Linguistic structures
are seen to be shaped by a variety of forces, including the many physiological, cognitive, and contextual
factors involved in their acquisition and use. Pertinent physiological factors include, for example, the motor
abilities that constrain articulation. Cognitive factors include general capabilities rather than specific
linguistic structures, such capacities as memory, pattern recognition, abstraction, generalization, and
routinization of repeated tasks. Contextual factors represent perhaps the largest and most varied set,
including text structure, communicative goals, language contact, and the myriad other features of the
extralinguistic context that can affect the way communication and ultimately language are shaped. These
three kinds of factors, physiological, cognitive, and contextual, are intertwined in most communication.
Because communication is effected by all components of the linguistic system working in concert, value has
been placed increasingly on considering linguistic structures within the context of the grammar as a whole,
and within the context of communication, thought, and interaction. As a result, functionally oriented work
has been based, where possible, on spontaneous speech recorded in its natural setting.

1

Few modern functionalists would maintain that there is a synchronic, one-to-one correspondence between
linguistic form and function. Synchronic systems are understood as the historical products of sequences of
individual diachronic events, each motivated in one way or another at the time it occurs. The diachronic
dimension thus plays a key role in explanation. This focus contrasts with the secondary role accorded
diachrony under some theoretical approaches in which primary attention is paid to those aspects of
language hypothesized to be innate and thus immune to change. Under such approaches, language change
has sometimes been viewed more as a phenomenon to be explained in terms of synchronic constraints, or
as evidence for particular universal structures. As a result, the kinds of phenomena investigated have varied.
Functionalist approaches have tended to focus on those aspects of language that do change, that can be
seen to be shaped by processes of acquisition and use. Arbitrariness is recognized as an integral feature of
grammar, but explanations are sought for the development of arbitrariness as well. Perhaps the most
fundamental source of apparent arbitrariness is the process of grammaticization, the cognitive routinization
of recurring structures. When the individual decisions involved in building complex expressions are

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 2 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

of recurring structures. When the individual decisions involved in building complex expressions are
automated, fine judgments need not be made each time a structure is used. The French subjunctive, for
example, could be seen to have a basic irrealis function, but speakers do not evaluate the degree of reality
of the situation at hand before they utter every subjunctive form. Its use is triggered automatically by certain
grammatical and lexical contexts. Such arbitrariness is itself quite functional: the automation of whole
structures frees the mind for attention to more novel aspects of the message (Mithun 1989). Arbitrariness
can also result from processes of change. As is well known, grammatical changes that simplify one area of
the grammar often complicate others. Furthermore, the ongoing process of syntactic change can also create
arbitrariness when the motivation behind one change is obscured by the next.

The explanation of syntactic change in terms of communicative function is not new. In quite early work one
finds an assumption that when the communicative efficacy of language is impaired in one way, speakers
instigate compensatory changes. Harris and Campbell (1995: 21–3) point to early scholars who explained the
rigidification of word order by the loss of inflectional case, beginning with Ibn Khaldûn in the fourteenth
century on Arabic (Owens 1988: 270) and continuing with Bernard Lamy in 1675 on French (Scaglione 1981:
41). Adam Smith (1761) and Johann Herder (1772) held similar views on the motivation of language change
in general. Cognitive abilities involved in the acquisition and use of language have also long been adduced
as forces shaping language change. In 1816, Franz Bopp explained the development of the Indo-European
infinitive in terms of the reanalysis of an original nominal form as a verb (Disterheft 1980). Hermann Paul's
1880 discussion of analogy and restructuring in grammatical change emphasizes the role of pattern
recognition, reanalysis, and extension by both children and adults (Paul 1880). The importance of the
cognitive routinization of repeated tasks, resulting in the grammaticization of frequently used syntactic and
morphological structures, was appreciated by a number of early comparativists and discussed eloquently by
Meillet in 1912.

2

Aspects of the context in which communication takes place have long been noted in

discussions of change. The role of language contact, for example, was discussed as early as the eleventh
century by Ibn Hazm of Cordova (Harris and Campbell 1995: 33). When Adam Smith (1761) attributed the
rigidification of word order to the loss of case inflection, he located the ultimate cause of the change in
language shift: inflectional categories were lost as adults learned a second language imperfectly. All of these
lines of research have continued to the present day with increasing sophistication and rewards, as more has
been learned about individual languages and as general patterns have been compared. The role of functional
considerations such as these in the understanding of syntactic change will be illustrated in the following
sections.

1 Routinization and Reanalysis: The Yup'ik Subordinative

A long-recognized capacity of the human mind is the ability to automate repeated tasks. This process is one
of the most powerful forces shaping grammar on several levels of structure. Over time, frequently recurring
discourse patterns can become routinized in syntactic constructions. Such a process has been hypothesized
to underlie the development of English complement constructions, for example. They are assumed to have
evolved from series of two clauses, the first containing a demonstrative that which points cataphori-cally to
the fact stated in the second (Allen 1980). Independent words that recur frequently in certain constructions
can evolve into grammatical particles, clitics, and affixes. Such a process can be seen in progress in the
evolution of the English word full into the adjective-forming suffix -ful of beauti-ful and grace-ful. As pre-
formed templates, the grammatical structures that result from such processes require less attention from
both speaker and hearer during the production and understanding of speech. A second well-known cognitive
capacity is the ability to abstract patterns. Such processes are easily observable as children acquire their first
language, producing along the way forms that reflect overgeneralizations or alternate analyses of existing
patterns. A third familiar ability is the extension of recognized patterns to new contexts, observable as
speakers exploit the tools at hand for new expressive needs.

These abilities can play important roles in the shaping of syntactic structures. They can be seen, for
example, to underlie a syntactic construction in Central Alaskan Yup'ik, an Eskimo-Aleut language spoken in
southwestern Alaska. In Yup'ik, as in all Eskimoan languages, both nouns and verbs consist of an initial root,
any number of post-bases (primarily derivational suffixes), and a final inflectional ending. For nouns, the
ending marks number and case. Possession can be expressed by a transitive pronominal suffix specifying
the possessor and the possessum. For verbs, the ending consists of a mood marker and a pronominal suffix
complex specifying the core arguments of the clause. A sample noun (‘my grandmother’) and verb ('she told
me about it’) can be seen in (1).

3

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 3 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

me about it’) can be seen in (1).

3

(1)

Among the Yup'ik mood suffixes are some that express common modal distinctions, such as the indicative,
the interrogative, and the optative (for polite commands), and others that function primarily to link clauses
in various ways, namely the subordinative, the participial, and a set of connective moods: several
contemporatives (‘while,’ ‘as,’ ‘when in the past’), the precessive (‘before’), the concessive (‘although, even
if’), the contingent (‘whenever’), the consequential (‘because’), and the conditional (‘if, when in the future’).
Of special interest here is the mood referred to as the subordinative.

Subordinative clauses are frequent in spontaneous Yup'ik speech, often corresponding to what would be
packaged as independent clauses in English. In the passage in (2) the speaker, Mrs Charles, described the
transport of a butchered moose across a portage. The pieces of meat were to be shared by two families:

(2)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 4 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

The subordinative mood -lu- serves to link actions or states that are portrayed as related elements of a
larger event or episode.

The subordinative construction includes a grammatical requirement that the subjects of all subordinative
verbs must be coreferent with the subject of a main clause (though the specification of the higher subject is
not always explicit). The abbreviation R in the glosses of the pronominal suffixes in (2) indicates that the
argument is coreferent with the overarching subject, the two people loading the boat. Sentences like that in
(3) are unambiguous. Gender is not distinguished in Yup'ik, but the subject of the verb ‘leave’ must be
coreferent with that of ‘watch’:

(3)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 5 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

The Yup'ik subordinative could be approached in several ways. Most historical linguists would begin by
seeking the source of the subordinative marker. Because the construction exists essentially as such in all of
the modern Eskimo-Aleut languages, its origins cannot be reconstructed from comparative evidence. There
is no attestation of an earlier stage of development in historical documents. The modern languages do
provide a clue, however. Fortescue et al. (1994: 410) propose a connection between the subordinative suffix
-lu and the enclitic =llu ‘and, and also, too’ that persists in all of the languages. In the standard Yup'ik
orthography, used here, the digraph ll stands for a voiceless lateral ł. The voiced l of the subordinative suffix
-lu is automatically devoiced following a voiceless segment, so it actually has two variants: -lu and -llu. One
approach to the reconstruction of the subordinative construction would be to stop at this point, having
determined that the subordinative suffix may be descended from a conjunction meaning ‘and’ and citing the
requirement of subject coreference as unsurprising evidence of the universality of the subject category.

The construction holds further interest for a functional approach to diachronic syntax, however. It
exemplifies the kind of syntactic construction that results from the routinization of a recurring discourse
pattern. In Yup'ik, as in the other Eskimoan languages, the general pragmatic relation of a sentence to the
preceding discourse may be indicated by the enclitic =llu. Such a link can be seen in (2d): “[They took the
pieces one by one and placed them behind the boat.] And when they were finished.” It appears that
sequences of clauses that were especially closely related pragmatically, sharing the same subject, came to
represent a recognizable complex construction in themselves. Repeated use resulted in the routinization or
grammaticization of the discourse pattern. As the complex construction became routinized, a kind of
reanalysis occurred. The conjunction llu became increasingly fused with the preceding constituent, first to an
enclitic as =llu ‘and,’ and then, in one construction, to a verbal suffix -lu. The suffix was then reanalyzed as
a member of the inflectional mood paradigm, complementary in function to other moods marked by verbal
suffixes, such as the indicative.

The fact that the subordinative construction requires subject coreference is also of interest. In general, the
grammar of Yup'ik shows strong ergative/ absolutive patterning. Both the case suffixes on nouns and the
pronominal suffixes on indicative verbs represent ergative and absolutive categories. If, as in some current
theories, the category of subject is considered a purely structural phenomenon, its unique role in the
subordinative construction makes little sense. Once its function is taken into account, however, its
prominence in just this area of the grammar is easily understood. We know that speakers’ choices of
subjects are not random. Given an array of participants in an event, certain preferences emerge. Semantic
agents tend to be preferred over semantic patients. First persons are preferred over second, and second
over third. Humans are preferred over animals, and animals over inanimate objects. Identifiable (definite)
arguments are preferred over unidentifiable (indefinite) ones (Silverstein 1976; Chafe 1994: chs 7–8; and
others). Thus in spontaneous English speech we find sentences like Sam grabbed the ball more often than
sentences like The ball was grabbed by Sam; sentences like I saw your mother yesterday more often than
Your mother saw me yesterday; sentences like He was hit by a car more often than sentences like A car
hit him; sentences like She ate the last cookie more often than The last cookie was eaten by her; and
sentences like Sally met a man in the produce section more often than A man met Sally in the produce
section (Chafe 1994). None of these preferences determines subject choice on its own. The observed
preferences reflect the general function of subjects: they serve as a point of departure for the clause.
Semantic agents tend to initiate actions. Speakers tend to present information from their own point of view
(thus the person and animacy hierarchies). Speakers typically take common knowledge as a point of
departure then move on to what is new. Of importance here is the fact that closely associated clauses tend
to share a common point of departure. For this reason, clause-combining constructions like the Yup'ik
subordinative that link clauses portrayed as elements of a larger event frequently show subject continuity.

The shaping of the Yup'ik subordinative illustrates the inseparability of cognitive and contextual factors in
the development of syntactic constructions. The cognitive routinization of the construction took place

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 6 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

the development of syntactic constructions. The cognitive routinization of the construction took place
because of the frequent occurrence of a certain discourse pattern. Repeated use of the marker led to an
erosion in form, which ultimately led to a functional and structural reanalysis of the conjunction as an
inflectional mood suffix.

2 The Extension of Patterns for New Communicative Functions: The Yup'ik

Past Contemporative

Most historical linguists recognize the extension of an existing pattern to a new domain as a force in
syntactic change. Different approaches to diachronic syntax have tended to focus on different aspects of the
process, however. For some, extension has been seen primarily as a simple generalization of structure. For
some it has been viewed simply as one entry in a catalog of possible kinds of change, the limits of which are
to be explored. Functional approaches focus on the reasons motivating changes of this kind: the expressive
needs of speakers. Such an approach can be illustrated with the evolution of another Yup'ik mood marker.

One of the Yup'ik contemporative mood suffixes, the past contemporative -ller-, is usually translated ‘when’
(in the past). It is used to situate one event in time with respect to another:

(4)

Comparative evidence indicates that the past contemporative was not present in Proto-Eskimo-Aleut or even
Proto-Eskimo. Its diachronic source is easy to identify, however. It is descended from a past nominalizer, a
derivational suffix that is reconstructed for Proto-Eskimo by Fortescue et al. as *-ł∂- (Yup'ik orthographic ll
is [ł], e is [э], and r is; velar and uvular fricatives regularly alternate with stops in final position):

(5)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 7 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

The past nominalizer persists in modern Yup'ik as well. Its use can be seen in (6):

(6)

The shift from nominalizer to subordinator did not occur in a single leap. Traces of its diachronic source
remain in the past contemporative construction. As noted earlier, nouns carry inflectional endings indicating
number, case, and possession. Possession is shown by a transitive pronominal suffix specifying both the
possessor and the possessum. The noun ‘boat’ in (7) contains a suffix indicating a third person plural
possessor and the third person singular boat, followed by the locative case marker:

(7)

Each of the mood markers in Yup'ik is associated with a particular paradigm of pronominal suffixes. For the
most part, verbs in the connective moods share the same paradigm. Intransitive verbs in the past
contemporative mood are inflected differently, however. They still carry the possessive suffixes used with
nouns, followed by the locative case marker:

(8)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 8 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

It is clear, however, that past contemporatives like ayallratni ‘as they were leaving’ are no longer analyzed as
nominal constructions. They have been reanalyzed as verbs. In Yup'ik, possessive nouns are identified by an
ergative case suffix. The ergative can be seen in its genitive function in (9), where it marks the man as the
possessor:

(9)

In past contemporative constructions like ‘when the man left,’ the noun ‘man’ is no longer classified
grammatically as a possessor. It appears in the absolutive case, as the only core argument of the intransitive
clause ‘(when) the man left’:

(10)

In transitive clauses, the traces of the nominal source of the past contemporative are disappearing. The past
contemporative mood is now usually followed by the transitive pronominal suffixes that appear with verbs in
other connective moods:

(11)

The possessive counterpart of the 3SG/3PL suffix for nouns would be -atni: angya-atni ‘in his boats.’ Yet
even with transitives the reanalysis was not instantaneous. Alternations persisting in the modern language
between nominal and verbal endings show that the past contemporative construction is still in the process of
evolving. Yup'ik speakers accept both tangallr-atni (with nominal ending) and tangallr-aki (with verbal
ending) for ‘when he watched them.’

The extension of the past nominalizer to an inflectional mood marker has been gradual, taking place in
small, tentative steps, the kind of increments that are possible only for changes shaped by use. The source
structure was a possessive construction, consisting of a noun referring to the possessor in the ergative (=

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 9 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

structure was a possessive construction, consisting of a noun referring to the possessor in the ergative (=
genitive) case, and a nominalized verb containing possessive and locative case suffixes (‘at their leaving’). As
the function of the construction began to shift, the nominalizing suffix split from its derivational source and
took its place in the inflectional mood paradigm used with verbs. The noun referring to the possessor,
originally marked with an ergative suffix, was coded as an absolutive, the sole argument of an intransitive
clause. Next, the endings on the nominalized verb began to shift. On transitives, the possessive suffixes
began to be replaced by the pronominal suffixes used with other connective moods, a shift that is still in
progress. On intransitives, the replacement has not yet taken place.

Comparative evidence shows that the evolution of the past contemporative is part of a larger constellation of
similar processes, all involving the extension of derivational suffixes to new functions as inflectional mood
markers. Jacobson (1982) lays out the range of nominalizing constructions in modern Yup'ik along a
continuum, from those which are still primarily nominal at one extreme, to those which are primarily verbal.
The gradual evolution of the system as a whole indicates even more clearly that such extensions become
established in the grammar slowly through use, rather than instantaneously as the result of a single
structural reanalysis.

Also of interest is the fact that extension of this type, in which nominalization evolves into subordination, is
not uncommon cross-linguistically. A functional analysis might go beyond documentation of the frequency
of the shift to a consideration of why this particular development should occur so often, of what might
stimulate speakers to make this particular leap. A possible explanation might lie in the function of the
nominalizer, which can allow speakers to reify an action. The nominalization of whole clauses permits
speakers to treat events as entities rather than independent predications, entities that can be integrated into
larger sentences.

Language as an Integrated Tool of Communication: Aleut Clause Structure

An important consideration in much functional work is the fact that all aspects of the grammatical system
work in concert for purposes of communication.

For this reason, the study of an individual syntactic construction in isolation can fail to yield the same kind of
understanding that might be possible when the language is considered as a whole. An example of the
interaction of different parts of the grammar can be seen in Aleut, the sole representative of the second
branch of the Eskimo-Aleut family. Aleut is spoken on the Aleutian, Pribilof, and Commander Islands off
Alaska and Siberia. The language and its relation to the Eskimoan languages have been described by Knut
Bergsland in a number of articles, especially Bergsland (1986, 1989, 1997), and a magnificent dictionary
(1994). Of special interest is the language's unusual clause structure.

Grammatical relations are indicated in Aleut by pronominal endings on verbs and case suffixes on nouns. In
intransitive clauses, the verb carries a pronominal ending referring to the single core argument. For first and
second persons, the ending is a pronominal enclitic. For third persons, it is a suffix distinguishing number. A
verb alone can function as a clause in itself, with reference to the argument specified by the pronominal
ending (examples cited here are from the Atkan dialect):

(12)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 10 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

Third person arguments may also be identified by a noun containing a suffix that distinguishes number. The
suffixes match those that appear on verbs: -x for singulars, -(i)x for duals, and -s for plurals.

(13)

In clauses with two nominal core arguments, both nominals show the same suffixes. Definiteness is not
distinguished:

(14)

If the semantic agent of a clause with two core arguments is not represented by a separate nominal, it is still
specified pronominally by the suffix on the verb (the tense glossed as present is used both for ongoing
events and immediate pasts, those of current relevance):

(15)

If, however, the semantic patient is not represented by a separate nominal, the structure of the clause
changes. The agent nominal carries the suffix -m, identified by Bergsland as an ergative:

(16)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 11 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

The suffix on the verb changes as well. In intransitive sentences like “The man is working” in (13), and
sentences with two nominals like “The man is eating fish” in (14), the ending is -x, matching that on the
agent noun ‘man.’ In transitive clauses without a separate patient noun, however, the ending changes. A
comparison of the sentences in (17) shows that the pronominal suffix on the verb suffix specifies the patient
or absolutive (in the plural, both the ergative and absolutive case suffixes on nouns appear as -s):

(17)

If the clause contains no independent nouns at all, the pronominal suffix on the verb represents a
combination of the agent and patient, but only one is mentioned overtly. If both arguments are singular, the
singular suffix -a is used. If either or both are dual, the dual suffix -kix is used. If either or both are plural,
the plural suffix -(ng)is is used (there is no gender distinction in the language):

(18)

The system can result in ambiguity, as can be seen from the translation of (19), but since the pronominal
suffixes are only used anaphorically to refer to already established referents, the ambiguity is seldom a
problem:

(19)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 12 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

We thus have an unusual system in which ergative marking appears on the agent noun, and transitive
suffixes appear on the verb, but only when the semantic patient is pronominal (anaphoric). Why should such
a system exist? Comparative evidence shows that it evolved from a system essentially like that of its
Eskimoan relatives, to which we now briefly return. The Eskimoan languages show a straightforward
ergative/absolutive pattern. Grammatical relations are expressed by pronominal suffixes on verbs and case
suffixes on nouns. The pronominal suffixes refer to two participants if the verb is transitive, and to one if it
is intransitive, whether independent nominals are present or not (as in Aleut, absolutives are distinguished
by their lack of case marking, but they do carry number suffixes):

(20)

The verbs may stand alone as full clauses in themselves with no change in form. Pronominal reference comes
from the pronominal suffixes: neraa ‘she ate it,’ iptuq ‘she disappeared.’

Many Yup'ik verb stems are ambitransitive, that is, they may be inflected either as transitives or intransitives.
The verb nere- ‘eat’ is inflected as a transitive in (20a) above, and as an intransitive in (21):

(21)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 13 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

An important feature of Yup'ik syntax is the fact that transitive absolutives must be identifiable (definite). If
the semantic patient of an action is indefinite, it must be expressed as an oblique, in Yup'ik the ablative. The
clause is then grammatically intransitive:

(22)

Such a grammatical requirement is actually not unusual cross-linguistically. Kapampangan, for example, an
ergative language of the Philippines, shows the same restriction on transitive absolutives as Yup'ik (Mithun
1994). In essence, in these languages syntactically transitive constructions require a high degree of
transitivity in the sense of Hopper and Thompson (1984). If a semantic patient is indefinite or incompletely
affected by an event, the event is expressed as a syntactic intransitive.

Aleut shows a surprising departure from the relatively common, stable system found throughout the
Eskimoan languages. The requirement that a semantic patient be definite in transitive clauses has evolved
into a requirement that the semantic patient be pronominal (anaphoric). It is only under these circumstances
that the ergative suffix -m can appear on the agent noun and that the transitive suffixes can appear on the
verb. This typologically strange development was stimulated by a change in another part of the grammar.
Bergsland reports that final syllables in Aleut underwent phonological reduction. As a result, three of the
Proto-Eskimo-Aleut case suffixes, the Proto-Eskimo-Aleut ergative *-m, ablative/instrumental *-mek, and
locative *-mi, merged into a single Aleut form: -m. The merger of the distinction between the original
ergative *-m and oblique *-mek would have caused havoc in the existing system, in which -m marked
ergative agents and -mek oblique patients. The same form would be used to mark contrasting functions. In
sentences corresponding to the Yup'ik “The woman ate the fish” in (20a), the suffix -m would mark the
semantic agent, but in sentences corresponding to “The woman ate a fish” in (22), the same suffix -m would
mark the (formerly oblique) semantic patient. The syncretism of the cases would have seriously impeded
communication.

The remedy is understandable in terms of the structure at the time. In the original system, only identifiable
(definite) nominals could serve as transitive absolutives. The most identifiable arguments are of course those
that can be represented by pronouns. This is just the context in which traces of the original transitive
construction have remained in Aleut, clauses with pronominal transitive absolutives. The original ergative
marker remains, but the context has been narrowed from all clauses containing identifiable absolutives of
any kind to only those containing pronominal (anaphoric) absolutives, without independent noun phrases. In
the parent language, all core arguments were represented by pronominal suffixes on the verbs, whether
additional nominals were present in the clause or not. In modern Aleut, only those transitive arguments not
identified by separate nouns are represented by the pronominal suffixes. As Bergsland points out, the Aleut
transitive suffixes on verbs are derived from a subset of the transitive suffixes on Proto-Eskimo-Aleut verbs:
Aleut singular -a is cognate with Yup'ik -a 3SG/3SG, Aleut dual -kix is cognate with Yup'ik -ke-k 3DU/3DU,
and Aleut plural -(ng)is is cognate with Yup'ik (ng)i-t 3PL/3PL. He notes that the Aleut forms are clearly
innovations, reductions of the earlier, more elaborate system.

The unusual evolution of the original ergative system has led to a number of other innovations within Aleut.
The special treatment of pronominal (anaphoric) arguments in clauses has been extended to possessive
constructions. In Proto-Eskimo-Aleut, as in the modern Eskimoan languages, nouns referring to possessors
carry ergative suffixes, and nouns referring to possessions carry transitive suffixes specifying the possessor
and possessum. These are similar in form to the transitive suffixes that appear on indicative verbs. When
Aleut clause structure became sensitive to the difference between nominal and pronominal arguments,
possessive constructions underwent a similar development. Now the possessor is specified pronominally on

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 14 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

possessive constructions underwent a similar development. Now the possessor is specified pronominally on
the possessum only if it is not represented by an independent nominal. Proto-Eskimo-Aleut also contained a
locative case suffix *-mi, which became indistinguishable from the ergative and ablative due to the
phonological changes mentioned above. As a result, locatives can no longer be expressed inflectionally in
Aleut, but only analytically in a phrase. Finally, the changes in the specification of grammatical relations have
affected Aleut word order. As seen in (14a) above (“The man is eating the/a fish”), if both core arguments
are represented by nouns, neither noun carries case marking. As a result, constituent order has become
quite rigid, nearly invariant SOV, in contrast with the fluid order of the Eskimoan languages.

The syntactic structure of modern Aleut becomes explicable once the diachronic processes involved in its
development are considered. Massive syntactic restructuring was triggered by phonological changes that
affected the shapes of morphological markers. The changes in shape nearly destroyed the original case
paradigm, which in turn compromised the efficacy of existing syntactic constructions. Such changes can only
be explained when the language is considered as a whole, its parts interacting in communication. The
changes would not have taken place if grammar were not shaped by its communicative function.

4 The Communicative Context: Conjunction

Perhaps the greatest variety of factors shaping language come from the context in which communication
takes place. The effects of two of them can be seen in the evolution of one of the most basic syntactic
constructions: coordinating conjunction. Conjunction might be assumed to be among the most universal and
stable of constructions. Yet even among closely related languages we find unexpected variety, both in the
degree to which coordination is grammaticized and in the inventories of coordinating constructions that
exist (Mithun 1988). In many languages there are no grammaticized coordinating constructions whatsoever,
and in many others the constructions that do exist can be seen to have evolved surprisingly recently.

Speakers and writers of European languages might wonder how a language could function without
grammaticized coordination. An examination of spontaneous speech in its natural context quickly provides
an answer: links among constituents can be shown by intonation. Coordination is typically signaled
intonationally whether overt conjunctions are present or not. Very closely related constituents may be
combined with little break in intonation. Somewhat looser bonds may be shown by “comma intonation,”
usually a pause and a non-final pitch contour.

In a surprisingly large number of languages, the diachronic sources of coordinating constructions are easily
traced because the constructions have come into the grammar so recently. Such a situation can be seen
among the Northern Iroquoian languages of northeastern North America. Though not mutually intelligible,
the languages share most of their morphological and syntactic patterns. Yet each shows a different
coordinating construction. The basic coordinating conjunction in Cayuga, for example, is descended from a
Proto-Northern-Iroquoian discourse particle *ohni? ‘also, too,’ whose reflexes remain in all of the modern
languages. In Cayuga, the particle has been reduced to hni?. An example of the particle in its original use
can be seen in (23). After he had selected a hammer, a customer in a hardware store was asked whether he
needed anything else:

(23)

The path along which such a particle could develop into a conjunction is easy to see. The sentence in (24)
was the answer to a question about what a family was going to plant:

(24)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 15 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

With pauses separating the nouns, the potatoes could be interpreted as an afterthought, an extra addition to
the list. The particle now occurs in contexts where it clearly could not mark an afterthought. A guest
watching three children play asked their names. His host’ s reply is in (25):

(25)

The particle does not yet appear with conjoined verbs or clauses, except in its original adverbial role. Verbs
and clauses are usually linked simply by intonation.

The other Northern Iroquoian languages all contain coordinating conjunctions, but the forms are not
cognate. Some developed along a route parallel to that of Cayuga hni?, like the Seneca kho, which evolved
from a different particle meaning ‘too’ and also follows the final conjunct. Others developed along a slightly
different path, arising from a particle that serves to link new sentences to the preceding context. Such a
development can be seen in Mohawk. Historical documents indicate that there was no regular coordinating
construction in Mohawk a century ago (Mithun 1988, 1992). Modern Mohawk, however, contains a fully
grammaticized conjunction tanu? ‘and’ that conjoins clauses, predicates, and nominals. The diachronic
source of the conjunction can be seen both through comparative evidence and in the historical record.

A discourse particle tá: can be reconstructed for Proto-Northern-Iroquoian that functioned to tie new
information to preceding discourse. It has reflexes in most of the modern languages, usually at the
beginning of paragraph-like units, and is translated variously ‘and so,’ ‘so then,’ ‘so now,’ and ‘now then.’ It
appears pervasively in an extensive Mohawk text recorded in 1896–7 (Hewitt 1903) (Hewitt’ s orthography,
which differs slightly from that used in Mohawk communities today, is retained in examples from his work):

(26) Mohawk discourse particle tá:
[I am thinking that, perhaps, I should recover from my illness if ye would uproot the tree standing in
my dooryard and if there beside the place from which ye uproot the tree I should lay myself in a
position recumbent.]
Tá, e'thóne? né rao ñ kwéta? wahatiro ñ totáko?
so, at that time the his people they tree uprooted
“So thereupon his people uprooted the tree” (Hewitt 1903: 282.5)

The same texts show a compound particle tahnu ‘furthermore, moreover,’ apparently formed from the
particle tá: and another particle nú:wa ‘now.’ Like the particle tá:, it often appears at the beginning of a
new paragraph, relating the new information to preceding context, but it also precedes sentences and
clauses within paragraphs:

(27)

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 16 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

The particle tahnú? is no longer used in Mohawk, though one speaker remembers an elderly relative who
used it. A descendent of the particle appears as a regular conjunction in modern Mohawk, the phonologically
reduced, unstressed particle tanu? ‘and.’ It is used to conjoin constituents of any type:

(28)

(29)

It is striking that coordinating constructions in many languages of the world are relatively recent loans from
European languages, languages with which they have been in contact for no more than a century or two.
Bogoras (1922: 881) noted the presence in Kamchadal, a Luoravetlan language of Siberia, of local Russian
conjunctions i, dai ‘and,’ je ‘but,’ potom ‘after that,’ and others. Texts in Tiwi, an Australian aboriginal
language documented by Osborne, show the conjunction and (Osborne 1974). In his survey of
Mesoamerican languages, Suárez notes that “in most of these [Mixe-Zoque] languages coordinating particles
have been borrowed from Spanish, but in spite of that, coordination through mere juxtaposition (with
different meanings according to context) is still very common” (1983: 109). In Tequistlatec-Jicaque
languages, “constituents of the clause and clauses may be linked by coordinating particles; in Coastal
Chontal some of these particles are native, but in Highland Chontal all particles with this function are
borrowings from Spanish” (1983: 115). “Coordination is made largely through juxtaposition in Huixtan
Tzotzil. In Tojolabal, the same mechanism is found, although there are coordinating particles borrowed from
Spanish” (1983: 120). In Huave, “in most cases coordination is marked with particles borrowed from Spanish,

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 17 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

Spanish” (1983: 120). In Huave, “in most cases coordination is marked with particles borrowed from Spanish,
and the constructions with a reduced second clause match the Spanish patterns so closely that these have
probably been imitated too” (1983: 132). South American languages in contact with Spanish, such as Guaraní
and Quechua, exhibit this phenomenon as well (Cole 1982: 78–80). The borrowing of conjunctions may be
facilitated by a structural feature. Since they often occur at the edges of constructions, integrating them into
a new language need cause relatively little syntactic disruption.

The prevalence of conjunctions borrowed from European languages may be due to another factor as well.
The source languages for these borrowed conjunctions often have substantial literary traditions. Literacy
itself may contribute to the development of grammaticized conjunctions in two ways. First, written language
cannot exploit the powerful cue of intonation for indicating links among constituents. Punctuation can
provide only a faint shadow of the fine gradations of pitch and rhythm available in spoken language. Second,
written language has been shown to differ structurally from spoken language in important ways (Chafe
1985, 1987, 1994). Speakers, under constraints of memory and production time, typically produce
syntactically simpler constructions than do writers. They tend to introduce only one important piece of
information per intonation unit. Writers, by contrast, have the luxury of time to produce long, elaborate
sentences, embellishing earlier statements or rephrasing at will. Accordingly, written sentences are typically
longer and packed more tightly with information. While speakers use more sentence-initial conjunctions
than writers (32 versus 0.9 per 1000 words in a sample analyzed by Chafe), writers conjoin significantly
more constituents within clauses (23.8 per 1000 words versus 9.9). Such differences can be seen today
between modern spoken and written Mohawk. Relatively recently, Mohawk speakers have begun to write their
language. The conjunction tanD' ‘and’ appears considerably more frequently in their written Mohawk than in
their speech. The development of systematic overt specification of grammatical relationships in literary
languages is clearly functional, crucial for guiding readers through highly complex structures without the aid
of intonation.

But why should contact with these languages result in the sudden gram-maticization of coordinating
constructions in spoken languages? Intonation in spoken language can indeed signal various degrees of
linkage among constituents, but the precise nature of the links can be vague. Series of noun phrases may
indicate open or closed sets of entities, alternatives, or apposition. The formal grammaticization of
conjunction can provide systematic overt disambiguation. Series of intonationally linked clauses can show
the same vague relationships. They may represent sequential events (‘and then’), simultaneous situations
(‘and,’ ‘while’), a contrast (‘while’), purpose (‘in order to’), or items related in a number of other ways. Prior
to the grammaticization of clause conjunction, relationships between juxtaposed clauses are usually
interpreted from context, or, when necessary, specified by discourse adverbials. The grammaticization of
coordination offers systematic specification of the relationship. The fact that this tool should be so easily
borrowed, even by speakers who are not themselves literate in either the donor or the recipient language,
confirms the power of expressive need in shaping grammar. Speakers constantly exploit and extend the
devices available to them to meet new communicative needs. In language contact situations, there is no
reason that devices available in one language should not be extended to communication in another. The
forms themselves are apparently easily borrowed, as in Kamchadal, Tiwi, and the Mesomamerican languages
cited above. The mere concept of overt expression of coordination appears to be easily borrowed as well.
The Cayuga, Seneca, and Mohawk conjunctions were developed from discourse particles already present
within the languages, but they evolved quite rapidly into markers of grammatical coordination just at the
time that speakers were beginning to be educated in European languages, English for the Cayuga and
Seneca, and French for the Mohawk. The sudden development of syntactic coordinating constructions under
these conditions again illustrates the intertwining of cognitive and contextual factors in shaping syntactic
structure. Stimulated by contact with other languages, speakers have reorganized their grammatical systems
in order to increase their expressive resources for communication.

5 Conclusion

Functional approaches to syntactic change share a general assumption that languages are shaped in
significant ways by the physiological, cognitive, and contextual circumstances surrounding their use.
Functionalism is neither a codified theory nor a set of instructions for mechanical analysis: it does not offer
formulae for discovering explanations in the way that the comparative method might provide procedures for
reconstructing earlier sound systems. This characteristic is not unique to functionalist approaches; it is
shared by all work on diachronic syntax. It is due to the nature of the subject matter. Syntactic change can
be stimulated and facilitated by a wide variety of factors, often working in concert. Their presence in a

12/11/2007 03:39 PM

17. Functional Perspectives on Syntactic Change : The Handbook of Historical Linguistics : Blackwell Reference Online

Page 18 of 18

http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747919

be stimulated and facilitated by a wide variety of factors, often working in concert. Their presence in a
language does not guarantee that a given change will take place, only that it may be rendered more likely.
Examples abound of divergent developments in closely related languages that appear to share all major
relevant structural and contextual properties. What is called for is careful consideration of all of the
circumstances surrounding a change, both language-internal and language-external, that might have
stimulated or facilitated it, and common sense in assessing their interrelationships. As more is learned about
recurring constellations of phenomena, a clearer, more detailed picture should emerge of the kinds of forces
that motivate specific changes and of their interaction.

1 A survey of major synchronic functionalist work through the early 1980s can be found in Nichols (1984).

2 Further discussion of current work on grammaticization can be found in Bybee, Heine, and Traugott, this
volume.

3 Yup'ik examples cited here come from the speech of members of the Charles family of Bethel, Alaska. I am
especially grateful to Elizabeth Charles Ali for her help in transcribing and analyzing the material. Examples are
given here first in their surface forms and then, if appropriate or necessary, with an indication of their underlying
morphemic segmentation.

Cite this article

MITHUN, MARIANNE. "Functional Perspectives on Syntactic Change." The Handbook of Historical Linguistics.
Joseph, Brian D. and Richard D. Janda (eds). Blackwell Publishing, 2004. Blackwell Reference Online. 11 December
2007 <http://www.blackwellreference.com/subscriber/tocnode?
id=g9781405127479_chunk_g978140512747919>

Bibliographic Details

The Handbook of Historical Linguistics

Edited by: Brian D. Joseph And Richard D. Janda
eISBN: 9781405127479
Print publication date: 2004