Folio, Linguistica I/l pp.

1940 Societas Llnguistica Europaea, 1980

THE PEIORITIES AND PITPALLS

OF SYNTACTIC RECONSTRUCTION

LYLE CAMPBELL AND MARIANNE MITHUN

O. Introduction. Recently there has been a surge of interest in

diachronic syntax. As some put it, "diaehronic syntax is where it's
at." This recent flurry is most welcome, but the enthusiasm for such
studies has not been matched by equal success, especially in syn-
tactic reconstruction. The purposes of this paper are to examine the
underlying assumptions of several approaches to syntactic recon-
struction, to attempt to account for the lack of success, and to sug-
gest some priorities for future research. Specifically, we consider the
potential transfer from principles of phonologieal reconstruction to
syntactic reconstruction, pointing out the consequences of the ab-
sence of regularity and directionality for historical syntax. We
point out the limitations of the comparative method for syntax,
but indicate its contribution in the reconstruction of morphology.
We criticize heavily reconstruction via word-order typologies.
Finally, we discuss the causes of syntactic change and the utility of
such explanations for reconstruction. Throughout the paper we
emph&size the importance of a holistic approach, viewing language
s a functionally efficient system of interrelated and interdependent
elements and processes, utilizing knowledge of the interrelationships
in reconstruction.
1. The comparative method. Since it has often been hoped and
frequently assumed that principles of diachronic phonology could
be transferred to diachronic syntax, we begin by considering why
phonologieal and lexical reconstruction have been more successful
than the reconstruction of grammar. Phonologieal reconstruction
owes much of its success to the neogrammarian regularity hypothe-
sis, that sound changes are exceptionless, and to directionality, the
predictable direction of many kinds of sound change. Syntactic
change has no direct analogue of either. We consider this lack to be

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
20

the greatest reasori for the limited success of syntactic reconstruc-

tion. Here we consider the benefits accrued to diachronic phonology
from these two factors, demonstrate why they have no counterparts
in syntax, and consider the implications of this for syntactic re-
construction.
1.1. Regularity. That sounds do not change in randm or inex-
plicable ways is important in the application of the comparative
method in several ways. This fact provides a basis for determining
cognate material, for selecting comparable entities from a set of
languages. It also serves s a means of distinguishing similarities
among languages due to common ancestry from those which may
be due to chance, borrowing, or universale. Furthermore, many
sound changes typically ocur in one direction. For example, the
intervocalic voicing of voiceless stops is natural and very frequent
while a change in the opposite direction (voiced to voiceless between
vowels) is hardly known. Such directionality provides a principled
reason for choosing one reconstruction over another. Thus, if a
language, "Lv has intervocalic voiced stops correspndiiig to inter-
vocalic voiceless stops of another language, L2, it is highly probable
that L! changed in accordance with the mpre natural and frequent
direction to voicing in this position, and that L2 did not change,
thereby reflecting the older state. Given this correspondence and
the probable direetion of change, it is incumbent upon us to re-
construct voiceless stops.
Syntactic change is not characterizd by demonstrable regularity
or directionlity. There is n direct analogue of regularity, and
directionality is rarely discussed. The absene of a regularity ana-
logue has to do with the nature of language itself andits duality of
patterning ^double articulation). Sounds reeur in conventionally
learned combinations with other sounds in association with semantic
or grammatical meaning to form morphemes and words. The sound
itself can be seen s a "type" with its various occurrences within
the different morphemes s multiple "tokens" of a single 4itype."
Regularity means that the same sound under equivalent conditions
ehanges in the same way in the different morphemes. When the
type changes, each of its multiple tokens exemplifies the change in
a regulr way. But grammatical rules are not conventionally linked
with form, with phonological ^ubstanee, When a grammatical
pattern changes, it changes all its instances by definition; it isnot

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 21

embedded in any larger conventional associations like sounds in

morphemes and words.
It might seem at first that regularity should be manifested in
syntactic change s well, assuming that syntax merely governs
larger chunks of language than phonology. For example, the dele-
tion of the relative pronoun could be observable in many different
sentences somewhat s intervocalic voicing of stops might be
exemplified by many different words of a language. There is a
crucial difference, however, in the acquisition of the two Sounds
are acquired not independently, but s parts of words in which form
and meaning are linked. Unlike words, sentences are not learned s
wholes. The form and meaning (or function) are acquired separately.
A syntactic pattern is not learned anew for each sentence which
exemplifies it. A Speaker may utter many passive sentences, but
need not have acquired each sentence independently s he must
learn morphemes. Learning the passive rule once is sufficient for
the production of an infinite number of passive sentences. Because
of this one-time-only, single-shot acquisition of syntactic rules there
can be no equivalent of the regularity hypothesis in diachronic
syntax. The syntactic "type" has but a single "token", the general
rule; when this "type" changes, all its "tokens" (the one single to-
ken) change by definition, making regularity in syntactic change
trivial.
The regularity of phonological change helps one to establish
cognates, but if sentences cannot really be cognate, how is compar-
able syntactic material to be identified? Functional or semantic
equivalence in formal description is not enough, since chance,
borrowing, and universale, in addition to common ancestry, may
account for formal similarities. To assume that formal similarity
among rules of different languages implies that they are cognate is
like assuming words to be cognate merely because they share a
common gloss. Furtherinore, the relative role of chance similarity
in syntax is much greater than in phonology. For the form and
meaning of some word in two languages to be similar by chance is
relatively unlikely. But the rnge of possible Variation among func-
tionally equivalent syntactic patterns is so narrow that the probabi-
lity of chance similarity is great indeed. For example, if there are
only two possible Orders for the constituents of comparative con-
struction, Standard-Pivot-Adjective and Adjective-Pivot-Standard,

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
22

then any two languages selected at random have a high probability

of sharing the same order.
1.2. Directionality. Some sound changes are known to be irre-
versible or at least highly unlikely to happen in other than one
predictable direction. This directionality is useful in phonologicai
reconstruction. We can illustrate this with rhotacism. The change of
5 to r between vowels is quite common, but a change of r to s inter-
vocalically is almost unknown. If we are confronted with related
languages in which s of one, L1? corresponds to r of another, L2,
we are obliged to posit *s for the proto language. Knowledge of the
typical direction of change in this instance gives us a principled
reason for choosing *s over a possible *r, since we know that *s may
easily become r in this position, thus accounting for the reflex in
L2, but we have no reason to suspect that *r could become s, which
is what would be required to account for ~LV were we to postulate
*r. The directionality of rhotacism actually obliges us to reconstruct
*s, all eise being equal. Should we have reason to believe that *r
niore accurately reflects the older state, then our brden of proof is
very strong; we must show exactly why this particular instance goes
against expectations.
Our limited understanding of directionality in syntactic changes
is a severe handicap in reconstruction. For example, given the
correspondence set s/r we know the most likely proto sound was *s
because of the preferred direction of change. If however, we observe
that one daughter language exhibits SVO (Subject-Verb-Object)
order, another SOV, and a third VSO, in the absence of directionality
of syntactic change we have no basis for postulating the word order
of the parent.
The directionality of many sound changes is inherently connected
to implicational universale and to markedness relations. These uni-
versals prove most helpful in phonologicai reconstruction, and would
be of extreme value in syntactic reconstruction if only we under-
stood better the implicational relationships holding in syntax.
To illustrate the value in phonologicai reconstruction, consider
the relationship between glottalized and plain obstruents. The
presence of glottalized obstruents (C') in a language implies the
presence of the plain, non-glottalized counterparts (C). The marking
relationship between C' and C guarantees that under normal con-
ditions, if a change takes place involving both these, the glottalized
one may become plain, but the plain one cannot become glottalized.

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 23

This directionality in the implicational universal is useful in re-

construction, just s in the rhotacisni example above: given a cor-
respondence of C'IC, the directionality requires the reconstruction
of *C". We take the establishment of parallel implicational univer-
sale in syntax to be of very high priority for future research.
Some implicational relationships involve mutual dependency
rather than the directionality of markedness. These, too, are helpful
in reconstruction, but in a different way. Given, for example, the
aspiration of ph in some language, we may expect th also; but equally,
given ^, we would expect ph. This is a relationship of symmetry
(pattern congruity) and the presence of a part of the pattern leads
us to expect other parts. The relationship does not permit inferences
about the direction of change. We do not expect ph>th nor th > ph.
If, however, we find evidence suggesting that some proto language
had *ph, then we may suspect that it also had *tfl and seek supporting
evidence. The relationship thus provides clues for reconstruction.
Syntactic universale can provide similar clues valuable for re-
construction. For example, if we know that all languages which can
relativize off of direct objects can also relativize off of subjects, and
if we find relative clauses modifying higher direct objects in the
daughter languages, then we would be prompted to search for
relatives modifying subjects. Similar examples can be drawn from
word-order universale. We have observed that VOS (Verb-Object-
Subject) languages seem always to have an alternative order (VSO)
whenever the object is "heavy", i.e. contains a conjoined noun-
phrase or a relative clause. If this putative universal holds true
under subsequent investigation, then whenever *VOS order can
be reconstructed for a parent, evidence of *VSO with heavy objects
should be sought. To take another example, ergative languages
characteristically possess rules of antipassive (see Johnson 1976).
Evidence of ergativity in a proto language would prompt us to seek
evidence for reconstructing a rule of antipassive s well. The more
we know about implicational universals and the interrelationships
among parts of grammatical Systems, the better our chances for
reconstruction. For this reason we emphasize a holistic approach
which gives serious attention to the systematic interrelationships
which hold in grammatical Systems.
1.3. Morphological reconstruction. Though there is no direct
analogue in syntactic change of regularity and directionality, there
are partial compensations, which we take up now. The first we con-

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
24

sider is morphological reconstruction by the comparatiVe method.

It is often pssible to reconstruct inflectional and derivational
morphology using the Standard techniques of the comparative meth-
od. To the extent that morphologically complex cognates bear
regularly corresponding sounds and that the cognate morphemes
have not undergone substaiitial functional or positional shifts, we
can reconstruct morphoiogically complex words with the normal
techniques f lexical reconstruction. Morphological analysis of
these polymorphemic proto words provides us with much of the
proto morphology. We can reconstruct morphology by the compara-
tive method precisely because morphology is associated with the
phonlogical form of words containing affixes. Reconstructing the
phonological shape of the morphologically complex words essentially
gives us the proto morphology free.
Furthermore, because there is no clear boundary between mor-
phology and syntax and because many syntactic rules are signalled
(at least in part) on the surface by affixes, the reconstruction of
morphology in this way can in some cases take us a goodly distance
toward an understanding of the syntax of the proto language. The
reconstruction of affixes signalling, for example, passive, causative
Infinitive, or nominalizing morphemes suggests that the proto
language had the related syntactic rules. Examples of this techniqe
can be found in Norman and Campbell (1976), where the reltinship
between ergative morphology, Word order, and rules of antipassive
are discussed; in Jeffers (1976), where the coincidence of VSO order
and embedding by nominalization is explored; and others.
Morphological reconstruction by the comparative method is li-
mited, however. Such things s morphological loss, analogical
changes in morpheme shapes, and shifts in their functions are not
easily recovered by the comparative method alone. A particularly
revealing question to consider is how much of the inflectional mor-
phology of Latin could be recovered from the modern Romance
languages with only the techniques of the comparative method.
1.4. Other constraints. There is additional compensation for the
absence of regularity and directionality, though admittedly limited.
We consider two very well-known principles which are nevertheless
worth reconsidering.
One is the uniformitarian principle: one does not postulate for
diachronic states things that are synchronically impossible. That is,
it is not valid to assume things were pssible in the past that are

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 25

impossible at present (also vice versa, that present possibilities were

impossible in the past).
The other is the intettigibility constraint: there must be a conti-
nuity of intelligibility from one generation to the next. These two
constraints together serve to make syntactic changes relatively
more plausible and learnable.
2. Word-order typology. Word-order studies enjoy a large current
popularity and varied practice. For some, the assumption that
languages tend to change toward greater typological consistency in
word order has become almost a direct analogue of the regularity
hypothesisif languages were always to change toward typological
consistency, then syntactic change would be regulr in a certain
sense. We consider here the limitations of word-order typology for
syntactic reconstruction.
In word-order studies, a cohesiveness is noted among certain sets
of syntactic entities. For example, Lehmann (1973, 1974, etc.)
following Greenberg (1966) notes a tendency for languages with
principal constituent order VO (Verb Object) to have prepositions,
to exhibit the order Adjective-Pivot-Standard in comparative con-
strctions, and to place modifiers (relative clauses, adjectives, gen-
itives, etc.) after nominale; etc. OV languages tend to have post-
positions, the order Standard-Pivot-Adjective, and modifiers pre-
ceding the nouns they modify; etc.
Syntactic reconstruction for a single language by word-order
typology is considered by many of its proponents to be akin to
internal reconstruction. Just s in internal reconstruction attested
alternations provide the basis for postulating non-alternating,
unique forms in the past, so deviation from typological consistency
(i.e. Variation in word-order type) is considered an indication of an
earlier consistency of a different type. Scattered OV characteristics
in a primarily VO type language, for example, are taken s possible
indications of an earlier OV type. Inherent in the method is the
assumption that the history of the language can be unraveled back
to some earlier consistency. This assumption is not unlike assump-
tions made in the internal reconstruction of phonological and allo-
morphic variants, but bears closer examination when extended to
reconstruction by the comparative method.
The telng question for typological reconstruction is, does it
require the proto language to be consistent? Some practitioners
seem to answer yes, others no. If the answer is taken to be yes, then

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
the method suffers from two faults. First the method is circular.
As Jeffers (1976b: 987) puts it in bis review of Lehmann:
. . . the method is . . . circular. For L [Lehmann], a con-
struction in Latin or Germanic (be it productive or a relic),
is archaic, non-innovative, and demands no synchronic ex-
planation, if only it is consistent with V typology. At the
same time, these very constructions are said to argue for the
OVcharacterofPIE.
That is, since the archaicness of any construction in a daughter
language is established on the basis of consistency with a parent
type, and since the parent type is reconstructed on the basis of
.archaic daughter types, the method is circular.
More importantly, the assumption that the proto language must
be consistent is a direct violation of the uniformitarian principle.
Since contemporary languages are frequently inconsistent, the op-
tion of an inconsistent proto language in the past cannot be ruled out.
If the answer is no, then reconstruction by this method leads to
several contradictions, since presumably syntactic cohesiveness of
the interrelated elements of the types is the main thing in favor of
this method.
This weaker version postulates a natural tendency inherent in all
language Systems to develop toward typological consistency, though
it admits the existence of conflicting factors moving languages
a,way from consistency. Consider the value of this method for each
of the following cases, simplified to a comparison of two languages
only for illustration's sake.
a. Two daughter languages are typologically consistent in the
isame way (C^). There are two plausible origins for this Situation:
1. *Cr The consistency of the daughters could be a common
inheritance from a parent with the same consistency.
2. *I. The parent language could have been inconsistent but
both daughters became consistent under the natural ten-
dency to develop consistency.
In this case, the method does not provide strong support for choos-
ing either alternative over the other.
b. Both daughter languages are consistent but in different way s
40^2). In the absence of knowledge of the direction of typological

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 27

change, there is no clear basis for choosing either s reflecting the

older state. Reasonable proto states could be the following:
1. *Cr A separate force could have brought the Cx language
out of alignment, then the natural force toward consistency
moved it along to its current C2 consistency.
2. *C2. Some event may have disrupted the consistency of
the C2 language and the natural tendency inoved it along
toC r
3. *I. The parent language was originally midway between
GJ and C2 and a natural tendency toward consistency pushed
each daughter language in a different direction.
The method rules out none of these possibilities but does prompt the
search for the disruptive force required for l or 2. If we could rule
out the existence of such a force, the method would indicate 3.
Unfortunately, we have no procedure for doing so.
c. One daughter language is consistent and one inconsistent
(Cjl^. Possible antecedents are the following:
1. *I. The parent language may have been inconsistent in
the same way s its inconsistent daughter. The other daugh-
ter could have become consistent under the natural tendency
in this direction.
2. *Cr The parent could have been consistent in the same
way s its consistent daughter; other factors interfered with
the consistency of the other daughter.
The method again prompte us to search for a disruptive force. As
above, the method would indicate *I only if we could rule out the
existence of other factors.
d. Both daughters are inconsistent in different ways (I^) Again
in this case, the method alone does not provide a sufficient basis for
reconstructing the typological state of the parent. Possible recon-
structions are the following:
1. *I12. If II and I2 are different points along the same path,
say between Cx and C2, the parent could have been inconsist-
ent in a way equidistant from both Cx and C2. The two lan-
guages could both be developing naturally toward different
consistencies.

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
28

2. *IX or *I2. Some factor could have radically altered the

original inconsistent state of one daughter to a new inconsist-
ent state. The tendency toward eonsistency does not enter
the picture at all here.
3. *C12. If the proto language represents the consistent state
between the two states of the daughter languages, both
languages could ha.ve undergone change away from consist-
ency due to outside factors counteracting the natural tend-
ency toward consistency.
e. Both daughters are inconsistent in the same way (1^). To say
that the parent was consistent and that both languages sponta-
neously developed inconsistency goes against the assumption of a
natural tendency toward consistency. The most probable explana-
tion is that they both reflect the inconsistent state of their ancestor.
This is the only case, then, in which typological considerations do
provide a basis for choosing one proto System over the others.
This brings us to a paradox in the typological approach tp syn-
tactic reconstruction: the method proves useful only when its
underlying assumption, that languages change toward consistency,
fails. That is, only in the case where both languages are inconsistent
in the same way (1^) are strong inferences about the proto lan-
guage's word order (*) possible. This is so because in all other in-
stances the daughters' Orders are either incommensurate (G^,
Ijlg) or potentially explicable by the tendency to develop consist-
ency. Only when neither l nguage's order evidences the tendency
toward typological consistency s predicted by the method does
the method offer aid in reconstruction. But even worse, notice that
it is the normal application of the comparative method which directs
us to postulate * in this case and the reconstruction actually has
nothing to do with the method of word-order typology and the
assumption that languages change toward consistency. In fact, it
is precisely this assumption which prevents us from making stronger
inferences about the possible proto states in the other cases.
These are fairly strong condemnations of reconstruction via word-
order typology, but there are many other serious questions that
word-order typologies must deal with, s well. Why do typologists
fail to show exactly how typological generalizations are involved in
actual documented changes ? Does word order mean different things
for different languages? (For some it functions to distinguish

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 29

primary roles of subject, object and verb; for others, for example,
where cases perform this role, it serves stylistic and topicalization
roles. Are these separate functions commensurble?) Why do some
languages persist in maintaining typologically inconsistent states
over centuries and even millennia without evidencing a tendency to
change towards consistency? (This is the question of lag.) If a
language was consistent (s some assume proto languages must have
been), if a language was in typological equilibrium, then why should
it ever change at all ? If borrowing (a frequently invoked wild-card
to explain the beginnings of inconsistency) can be involved, then
how does it affect typological cohesion or effect typological shifts?
Unanswered, these questions constitute strong criticisms of word-
order typology considerations for reconstruction. We consider some
of these questions in greater detail.
K the only natural tendency in syntactic change were toward
typological consisteney, then all languages should have developed
consistency by now. Unfortunately, difficulties have too often been
attributed to the borrowing of inconsistent features, but without
the necessary supporting evidence. If typological considerations are
to contribute to syntactic reconstruction, certain questions about
syntactic borrowing must be faced. The relative borrowability of
typological features must be more carefully investigated. Perhaps
some characteristics are easily borrowed (because they are peripheral
to the type), while others are virtually unborrowable by languages
in certain states, because aequiring them would throw the language
intolerably. out of typological kilter. It is important to study docu-
mented (documentable) cases of syntactic borrowing to constrain
indiscriminate recourse to borrowing s a wild-card explanation for
otherwise unexplained typological inconsistency. If borrowing is to
be used s an explanation of typological inconsistency, criteria for
demonstrating actual borrowing need to be developed. It is impor-
tant to consider the subsequent consequences of borrowed syntactic
patterns for the word-order type. (See Tai's (1976) interesting
example of the Altaic influence on Chinese.) Since typological
characteristics may vary in their borrowability, so might they vary
in their capacity to set typological shifts in motion. Also, perhaps
certain inconsistencies are tolerated more easily than others. All this
needs serious investigation.
Another important question which requires further research is
whether typological characteristics change in a specific order in the

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
30

course of a typological shift. If there is a consistent order in such

changes, it would be possible to identify the direction of change in
progress and so to infer characteristics of particular proto languages.
Other factors triggering typological shifbs or undermining consist-
ency must also be identified. The degree to which phonological
decay and analogical restructuring, for example, can effect typologi-
cal shifbs by altering crucial typological characteristics must be
examined.
These research priorities point toward a specific goal: the explana-
tion of typological cohesion, and why and how such cohesion may
be broken, altered, or shifted to a new cohesive type. Mere descrip-
tion is not enough. At present word-order typology does not provide
a rigorous technique for syntactic reconstruction. However, word-
order typology may become useful when it is rigorously constrained
and when the questions raised here have been resolved.
We have strongly criticized the use of word-order typology in
syntactic reconstruction. This raises the question, can typology play
ny valuable role at all in syntactic reconstruction ? We believe that
the answer is an emphatic yes. It can play a role in two ways. First,
there does indeed appear to be something substantial about the
tendency for languages to evidence typological cohesion. This offers
some hope for determining the direction of some kinds of syntactic
changes, and the increased potential this offers for distinguishing
between innovative versus archaic patterns can be helpful in re-
construction. The criticisms presented above are largely leveled at
fche deterministic assumptions attributed to word-order change by
many of its practitioners. We need not force proto languages to be
consistent nor do we need to assume that languages must change
toward consistent types over time. If we accept only the implication-
al relationships and universal aspects of the typologies without the
determinism, then we cannot be vexed by such problems s lag,
departure from states of equilibrium, etc. Without the determinism
it becomes fruitful to employ the insights from typology in recon-
struction, believing not that languages must per force develop con-
sistency, but that if they do change, all eise being equal, it will
probably be towards greater cohesion in the interrelationships of
the integrated elements of their type. This is akin to the use of
implicational relationships and markedness in phonological re--
construction. If consistency (typological cohesion) is less marked,
more natural, more expected, more tolerated, than inconsistency,

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 3l

that is, if there is pressure for the elements of the type to be inter-
related in specific and predictable way s, then generally we can
expect changes to be in the direction of less marking, less inconsist-
ency.
The second way typology can contribute to reconstruction is
related to the first. We require a holistic approach and typology can
contribute to the holistic perspective of language. In language there
are systematic interrelationships among the elements of grammati-
cal processes, working in concert, some naturally co-occurring with
others. Typology can elucidate these interrelationships. When the
normal or expected way in which language welds its parts into a
functional unit is determined, we then have useful clues for re-
construction. When we discover that some proto language had
certain components of an interrelated System, we can then search
for other elements of the System known to be often interrelated or
even required. Word-order typologies need to be made more rigorous
to be fully useful, however any systematic relationship known to
hold among parts of grammatical Systems will prove useful. Consider
the following example for ergative languages.
Ergative languages are characterized by "split ergativity," where
noun phrases and pronouns take either ergative or accusative case
marking. This marking is controlled by implicational universale in
a hierarchy such that, for example, ergative marking of second
person pronouns in a language implies that third person pronouns
will also take ergative marking (though ergative third persons imply
nothing for second person, which may then be either ergative or
accusative). Similarly, the presence of ergative marking for human
nouns implies that animate and inanimate nouns will also be
ergative, but ergative inanimates imply nothing for human
nouns. These interrelationships are valuable in reconstruction. If
evidence of second person or human ergatives is found for some
proto language, then we can be almost certain that third per-
son and inanimate noun phrases were also marked ergatively in
that proto language.
Norman and Campbell (1977) illustrate the holistic approach to
comparative reconstruction of Proto-Mayan syntax. They show, for
example, that not only are the case-marking relationships interre-
lated for various kinds of noun phrases in an ergative language, but
that case-marking is interlocked with morphology, syntactic rules,
grammatical relations, word order, prepositions and oblique noun

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
32

phrases, etc. That is, most of the devices of the grammar are inter-
related in such a way that none of them can be fully understood
without understanding the System s a whole and how each inter-
acts with the others. Knowledge of this interaction and interrela-
tionships makes the reconstruction of Proto-Mayan syntax exciting
and reasonably detailed in ways impossible without knowledge of
the universals of ergative languages s holistic Systems.
Finally, the relative probability of competing reconstructions may
be evaluated by considering the degree to which the individual re-
construction fits into the framework of a more comprehensive
phenomenon and the extent to which the reconstruction is compat-
ible with language universale (Dressler 1971). This point is parti-
cularly important: if we could aseertain how well a given feature
fits into a reconstructed grammatical System, then we would have
compensation for the lack of an analogue to the regularity principle
in sound change. Knowing how well a feature fits teils us its relative
inerit compared with competing reconstructions and also how satis-
factory a solution we have found to the problems we are dealing
with. Because of the nature of grammatical Systems, individual
processes have to work in harmony with other processes in the same
system. Language typology offers the possibility of elucidating the
nature of intemal cohesion in grammatical Systems by comparing
the distribution of different types of grammatical processes in the
languages of the world. In this way, grammatical reconstruction
can avoid arbitrariness by using the typological mpdel which best
fits a specific language family s the basis for comparison and re-
construction, s a gauge against which to measure the probabiiity
of the reconstructed system. Principled reconstructions are possible
if we adopt a holistic approach to reconstruction which utilizes
available knowledge about the structure of grammatical Systems in
general. Atomistic approaches which treat individual components
of grammar outside the context of the entire system are likely to
prove disappointing.
The holistic approach we require is not new. It is Meillet's i4 o
tout se tient" (X937: ix) and Brugmann's "Language., . . is a highly
complicated human activity, in which the most diverse factors work
together in mutual interdependence to the whole, a,nd in which
everything is essentially determined by everything eise" (Lehniann's
(1976: 16970) translation). When we couple the interdependencies
of a specific language with the expected or required interrelation-

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 33
ships of universal grammar, syntactic universale, then comparative
syntax has a chance for increased success.
3. Explanation of syntactic change. The more we know about
language generally, the more successful we will be at reconstruction.
If we understood more fully why languages change, we could use
that knowledge in syntactic reconstruction. If we know what causes
a language to undergo a certain change, then we have a better basis
for postulating what it was like before the change. If, for example,
some physical factor such s muscle control accounts for the ten-
dency toward the continued Vibration of the vocal cords during the
production of intervocalic stops, then this in a sense 'explains' the
direction of the frequent change of stops to voicing between vowels.
If we understood why grammars change, we could use the Informa-
tion of observed effects from known causes for syntactic reconstruc-
tions. Such urderstanding would yield motivation for choosing cer-
tain reconstruction over others. In this section, several proposed
explanations of syntactic change will be evaluated in terms of their
potential usefulness in syntactic reconstruction.
It is convenient to classify causes of change into internal and
external factors. By internal we mean cause of change having to do
primarily with the structure of language or the internal workings
of a grammar. Some internal factors that have been proposed in-
volve functional, perceptual, and typological explanations. By ex-
ternal causes we mean those with a focus on factors largely outside
the structure of language (outside the grammar per se). Some exter-
nal factors discussed have been borrowing, stylistic Variation, and
other sociolinguistic factors, etc. (See also Andersen 1974; Campbell
1974, 1976; Ohala 1974a, 1974b.)
3.1. Internal Causes. Languages seem to strive for communicative
efficiency, toward a balance in the requirements for efficient pro-
duction, perception, and learnability (see Kiparsky 1972, Kisseberth
1973). This tendency toward systemic equilibrium necessitates a
hostic approach to language reconstruction. The parts of language
are interrelated with different processes working in concert. Some
processes naturally co-occur with others within syntactic sub-
systems. When it becomes possible to determine the normal or
expected way in which a language welds these into a functional
unit, we can then search for the pieces and relationships in syntactic
reconstruction.

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
34

Many changes re broght aboiit by disrptions in other parts of

the grammar, putting the functional eqilibrium of the language s
a commnicative system under strss. Compensatory changes may
later bring the system into a riew balance. Oiie clear exmple found
in several modern Indo-European laiiguages, inclding English,
Spanish, and others, is conipensaiion for lost case endings by stricter
word Order to Signal grammatical relatiohships among constituents,
especially subjects, predicates, and objects. Another example is
from Lpp: because sund change eliminated certain final conso-
nnts which were the only formal means of signaling certain case
endings, consonnt-gradation in stems was extended to consonants
to which it formerly had not applied; thus the differeince between
gradated and non-gradated stems came to signal what had formerly
been signalled by case endings, endings whose phonological hape
had triggered the gradation before their loss (Anttila 1972). Mithun
(1976) illustrates a similar therapeutic change in Iroquoian, where
various nminalization markers were introduced t identify syn-
tactic functions of constituents obscured by optional focus fronting
rules. Kno (1974) has proposed several universals relating various
kinds of word order, based on the difficulties of decoding which
they present.
A prticularly instructive example was presented by Bever and
Langendoen (1972) which illustrates the interplay between percep-
tual and piOduction needs. They show that in the history of English,
s a iesult of the disappearance of nominal inflectioiis between the
eleventh and fifteenth oenturies, constructions with relative clauses
became perceptually complex. This perceptual complexity was
counteracted by changes in the restrictions on relative clause mar-
kers, which complicated the grammar per se, but removed many of
the perceptually difficult cases (1972: 77). A perceptually complex
sentence such s: 'The secretary discouragedthe man wanted to see the
boss' was grammatical in this earlier period. A complication in the
grammar requiring an overt relative niarker eliminated the diffi-
culty in Interpretation.
Another internal fctor is the so-called analogy based on mbi-
guity. These analogical changes may be seemingly purposeful modi-
fications to increase functional efficiency or accidental mfeinter-
pretations. Misinterpretations or incrrdct abductions in cases of
surface structure ambiguity may account for changes in related
parts of the grammar. Anttila (1972: 1024) presents several

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 35

examples. To consider just one, Greek had mfinitive phrases such s

tliel graphein "I want to write" and ihalei graphein "he wants to
write", but when final -n Was los$, the surface form of the third
person singular and the Infinitive became identical. At some later
time the infinitives were reiriterpreted s third person singular
forms (thelei grapliei 'he wants he writes') and extended to other
persons (tMl graph want I write'). For other examples of the
appearance of a new construction s a reanalysis of another surface
structure construction with which it is ambiguous, see Anttila 1972:
1024, Breckenridge and Hakulinen 1976, Campbell 1977: 126,
Ebert 1976: viixi, Jamison 1976, etc.
Examples of purposeful modifications are very common. We have
already touched upon several in the examples of the extension of
word order to compensate for lost case endings, the Lpp extension
of consonant-gradation to carry the function of lost case endings,
and Bever and Langedoen's (1972) increased complication in English
relative constructions to accommodate perceptual needs. Gases in
Mayan languages are preented by Normari and Campbell (1977)
in which languages obligatorily change certain sentences to anti-
passive voice with oblique rioun phrases in order to resolve ambi-
guous interpretations of the subject and object roles of noun phrases
in regulr active sentences. Additional examples are easily com-
pounded. <
We do not have s our goal in this section to srvey e^haustively
the internal causes of syntactic change. Rather we intend to dllus-
trate the poteritial contribution of such considerations and to call for
a holistic approach to the investigation of causes of language change.
Understanding the interrelationships among the parts of grammar
is necessary for fnctional explanations to be successful.
Also, it is important that we point out additional cautions. It is
ften too easy to find what one is looking for. Even given a non-
controversial cause and a language which has undergone a change
which could have been the result of the cause, thereis still no guar-
aiitee tht the cause (of our interest) is precisely the factor that
brought about the change in the language. Often other causes
working individually or in union with others can bring about results
which strikingly suggest the agency of a different cause which may
have had nothing at all to do with the case. Criteria must be estab-
lished for demonstrating that proposed causes were indeed the
determininig factors in specific changes.

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
36

Investigation of functional factors in linguistic change are parti-

cularly important, given the limitations of the comparative method
for syntactic reconstruction and given the current restricted contri-
bution of typology. An understanding of the internal causes of
change can point the way to valid reconstruction and provide a
means of distinguishing innovations from archaisms.
3.2. External causes. External causes are often less theory-bound
than internal causes and are often determined by language-specific
factors. We will consider several external causes here.
Syntactic borrowing is an important external cause of syntactic
change. Unfortunately it has been both underused and abused s
an explanatory factor. TJntil recently, many denied even the possi-
bility of syntactic borrowing and still many are reluctant to accept
cases of probable syntactic borrowing. On the other hand, many
others have indiscriminately used borrowing (substrata) to explain
the origin of any otherwise unexplained data, especially typological
inconsistency. To achieve the necessary balance between these two
extremes, it is mandatory to discover which aspects of syntax are
borrowable and under what conditions they may be borrowed.
Adequate criteria must also be established for demonstrating actual
borrowings.
One way pf constraining indiscriminate appeal to borrowing is to
amass s much data s possible on clear cases of syntactic borrowing,
along the lines of Thomason and Kaufman (1976). In this way the
probability for various kinds of syntactic borrowing to take place
might be established and favorable or necessary conditions specified.
Some convincing cases of syntactic borrowing are Gumperz 1964,
1969, Gumperz and Williams 1971, Hyman 1975; Nadkarni 1975,
Sapir 1907: 53344, Silverstein 1974, ndTai 1976. A valid type of
argument involves syntactic Variation in the dialects of a single
language which comes from language contact. Silverstein (1974), for
example, demonstrates the existence of grammatical borrowing in
Chinookan by tracing a dialect continuum in which the simple
tense/aspect system of Lower Chinook (on the Pacific Coast) cor-
responds exactly to neighboring Salishan languages, while the multi-
tense system of Kiksht Chinook (in the southern Plateau) corres-
ponds to that of neighboring Sahaptian languages. In this case there
can be little doubt that the very different tense and aspect Systems
are the result of borrowing.
Sociolinguistic factors often account for syntactic changes aixd it

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 37

is important to investigate these. Some are code-switching, stylistic

variants, conversational implicature, honorifics andreverentials, etc.
To take just one example, consider the development of reverentials,
(honorifics) and their impact on grammar. They have such
varied results s turning third person verb forms into second
person (e.g. Spanish, German, Quiche, polite pronouns) turning
plural forms into Singulars (Mam, English, Finnish), and converting
reflexives and applicatives into plain third person singular actives
(Nahuatl) (see Robertson 1976). Ritual speech generally can in-
fluence syntactic patterns to change. Finally, rules for oral litera-
ture, poetry, translation of sacred texts, and many other things can
all influence syntactic patterns to change. Both internal and exter-
nal causes of syntactic change are important areas of investigation
and deserve high priority.
4. Conclusions. Syntactic reconstruction can have a promising
future, but must be confronted with a sober realization of both its
potentials and limitations. In our considerations of various ap-
proaches that have been taken to syntactic reconstruction we have
pointed out that all are limited. The comparative method applied
directly to syntax is largely unproductive because there is no
analogue to the regularity and directionality of phonological change.
This lack is partially compensated for by our ability to reconstruct
morphology by the Standard techniques of lexical reconstruction.
It would be more fully compensated for if we had greater knowledge
of the implicational relationships which hold among the parts and
processes of grammatical Systems. As presently practiced, word-
order typology offers little help because of untenable assumptions
attributed to it by many when used for reconstruction. Typological
considerations generally, however, are of considerable aid in re-
construction. Most progress is likely when a holistic approach is
taken. Further progress will depend on expanded understanding of
the internal and external causes of grammatical change. Though we
have taken a somewhat negative view on syntactic reconstruction
s usually practiced, we hope we have pointed to promising direc-
tions for future success.
LYLE CAMPBELL
M ABIANNE MITHUN
SU N at Albany
ALBANY, N.Y. 12222
USA

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 38

^EFERENCES,;,
ANBESEN, H. 1974. Toward a -typology of change: bifurcatihg''changes
and .binary relations. Histprical linguislbics: jjroceedings : of, the> first
international Conference on historical linguistics, ed. by J. Anderson
and C. Jones,'18 60. Amsterdam: North Holland Publishing Ca. >
ANTTILA, R.. 1972. An introduction to historicaland comparaibive Jinguistics.
New ybrk: MacMlan.
BEVEB, T: and T. LANGENDOEN. 1972. The interaction of speech <perceptin
and grarnmatical structiire in the evolu.tio?} of language. Linguistie change
and generative grammar, ed. by R. Stockwell and R. MacCaulay, 3295.
Blooniington: Indian riiversity ' Press. : ; : .
BBECKENBIDGE,,;-,J. and A. HAKTTLINEN. 1976. Cycle and afber. Papers frpni
the parasession on diachronic syntax;, ed. by S. Steever et aJ., !SO6?.
Chicago Liiiguistic Society. { '
CAMPBELL, L. 1974. On conditions .on ,sound change. Historical li^iguisticsc
proceedings of the first international Conference on histbrical linguistics,
ed. by J. Anderson and C.: Jenes, 8996. Amsterdam: Nbrth Holland
Publishing Co^ ,; , . . . ' .-:
1976. Language contact and sound change. Proceedings of the second
international Conference on historical linguistics, ed. by W. Christie,
181 194. Amsterdam: Nprtji Holland Publishing Co. . (.
i 1977. Quichean linguistic prehistory.! Uniyersity of California Publi-
cations in Linguistics, 81. Los Angeles. ' : ;
DBESSLER, W. 1971. ber die Rekonstruktion der indogermanischen Syntax.
Zeitschrift fr vergleichende Sprachforschung 85. 522. [not seen].
EBERT, A. 1976. Introduction: Papers frohi the parasession n diachronic
syntasq, ed. by S. Steever et al.,;viixviiL Chicago Linguistic Society.
GBEENBEBG,. J. 1966. Some universale pf grammar with particular reference
to the order of meaningful elements. Universals of languaige, ;ed. by
J. Greenbergi 73?-113, Seond edition. 1VOT Press. ,..i >.
GUMPERZ, J. 1964. Hindi-Punjabi code-switching in Delhi. Proceedings
of the ninth international congress of linguistics, ed. by H. Lunt, 1115 24.
T h e Hague: ., : . . . . . . . ... ;: -v.;
1969. Communication in multilingual societies. Cognitive ant^roplogy,
ed. by S. Tyler, 43548. NY: Holi;; Kinehart and Wihston. ;
GMPEBZ, J. and WIl'SO SF. 19 71., onvergence and creolization: $, case from
the Indo-Aryan/Dr^vidian border in India. Pidginization and Creoliz-
ation, ed. by D. Hymes, 15167. Lottdori: Cambridge University Press.
HYMAN, L; 1975. On the dhange froni SOV tp SVO: evicjence fi;om Mger-
. Congo. Word order and word order; change, ed. by C. Li, 11347. Austin:
University of Texas Press. ; :
JAMISON;, S. 1976. Functional ainbiguity ;and syntactic change: the ganskrit
accusative. Papers from the parasession on diachronic syntax, ed. by
S. Steever et-'al., 126 35. Chicago Linguistic Society.
JEFFERS, .B. ;1976a, Typqlogieal shift and change in _complex sent^nce
structure. Papers from the parasessipn on diachronic syntax, ed. by ,S.
Steever et aL;1 136 49: CHiogo'Lihguistic Society. . ?'
1976b. Review of PrQto-ljiiio-^uropean syntax by W. Lehmann.
Lg 52: 982-8.
JOBCNSOK; D. 1976. Ergativity in universal ^ranimai. Unpublished mn^
SCript. . . . . ,_. , j . .;
KIPABSKY, P. 49.72. .Explanation in phonology. Goals of linguistic theory,
ed. by S. Peters, 189227. Englewood Cliffs, N. J.: Prentice-Hall.
KISSEBTH, C. 197.3> Th.e^interaction of phonological rules and the polarity
of language. Indian Unjversity Linguistics Club, Xerox.
KTJNO, S. 1974. The poitiri of relative clauses and conjunctions. Linguistic
Inquixy ^5: 11736,' /.K
LEHMANN, W. 1973. A structural principle of language and its implications.
Lg 49: 47-66. *

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
 1974. Proto-Indo-European syntax. Austin: University of Texas Press.
1976. A preface to diachronic syntactic investigation. Papers from the
parasession on diachronic syntax, ed. by S. Steever et al., 16978. Chicago
Linguistic Society.
MEILLET, A. 1937. Introduction Petude comparative des langues indo-
europeennes. 8th edition. Paris: Hachette.
MITHTJN, M. 1976. Topic, focus, and syntactic drift in Northern Iroquoian.
Paper read at the 1976 annual meeting of the American Anthropological
Association, Washington, D. C.
NADKARNI, M. 1975. Bilingualism and syntactic change in Konkani. Lg 51:
672-83.
NORMAN, W. and L. CAMPBELL. 1977. Toward a Proto-Mayan syntax:
a comparative perspective on grammar. Mayan Linguistics 2, ed. by
N. England. (In press). University of Missouri Press.
OHALA, J. 1974a. Experimental historical phonology. Historical linguistics:
proceedings of the first international Conference on historical linguistics,
ed. by J. Anderson and C. Jones, 35389. Amsterdam: North Holland
Publishing Co.
OHALA, J. 1974b. Phonetic explanation in phonology. Papers from the para-
session on Natural Phonology, ed. by A. Brck, R. Fox, M. LaGaly,
251 74. Chicago Linguistic Society.
ROBERTSON, J. 1976. The structure of pronoun incorporation in the Mayan
verbal complex. Unpublished Harvard University Ph.D. dissertation.
SAPIB, E. 1907. Preliminary report on the language and mythology of the
Upper Chinook. American Aiithropologist 9: 53344.
SILVERSTEIN, M. 1974. Dialectal developments in Chinook an tense-aspect
Systems: an areal-historical analysis. UAL memoir 29.
STEEVEB, S., C. WALKER, and S. MUFWENE. 1976. Papers from the para-
session on diachronic syntax. Chicago Linguistic Society.
TAI, J. 1976. On the change from SVO to SOV in Chinese. Papers from the
parasession on diachronic syntax, ed. by S.. Steever. et al., 291 304.
Chicago Linguistic Society.
THOMASON, S. and T. KATJFMAN. 1976. Toward an adequate definition of
creolization. Unpublished manuscript.

Brought to you by | Nanyang Technological University

Authenticated
Download Date | 6/11/15 11:17 AM
Brought to you by | Nanyang Technological University
Authenticated
Download Date | 6/11/15 11:17 AM