Reconstruction, Cultural 389
future, there is a direct intuition that future-tense sen-
tences are not now either true or false. Antirealists may
deem some class of sentences to be nondescriptive, or
descriptive but not involving a realist notion of truth.
Among those who accept realist truth, there are those
who analyze truth in terms of the existence of objects
and universals, and others who take an antirealist
attitude toward either objects or universals.
See also: Descriptions, Definite and Indefinite: Philosophi-
cal Aspects; Empty Names; Nominalism; Objectivity in
Moral Discourse; Reference: Philosophical Theories; Se-
mantic Value; Verificationism.
Bibliography
Aristotle (1984). ‘De interpretatione.’ In Barnes J (ed.) The
works of Aristotle. Princeton: Princeton University Press.
1, 25–27.
Ayer A J (1946). Language, truth and logic. London: Victor
Gollancz.
Blackburn S (1993). Essays in quasi-realism. New York:
Oxford University Press.
Reconstruction, Cultural
J T Katz, Princeton University, Princeton, NJ, USA
ß 2006 Elsevier Ltd. All rights reserved.
One of the main tasks of historical and comparative
linguistics is reconstruction – the recreation, to the
best of our abilities, of a linguistic situation at some
point in the past. Using broad knowledge of tenden-
cies in language change and employing above all the
justly famous and powerful technique known as the
Comparative Method (the classic account remains
Meillet, 1925), scholars are able to explain with
some precision the closely linked matters of linguistic
evolution and relationship, specifically how a given
language has evolved from its linear predecessor and
how the language in question is related to other ton-
gues that are not part of the same line but whose own
histories show that they once intersected with it.
Most work concentrates on phonology and morphol-
ogy since sounds and smaller combinations of sounds
are in a certain obvious sense the easiest units of
language to deal with; the reconstruction of syntax
is, however, playing an increasingly large role in
linguistic research.
But it is important not to forget semantics and
pragmatics: since speech exists in the first place for
communication, coming to terms with the meaning of
Devitt M & Sterelny K (1999). Language and reality: an
introduction to the philosophy of language. Oxford: Basil
Blackwell.
Dummett M (1973). Frege: philosophy of language.
London: Duckworth.
Dummett M (1978). ‘Realism.’ In Truth and other enigmas.
London: Duckworth. 145–165.
Dummett M (1991). The logical basis of metaphysics.
Cambridge, MA: Harvard University Press.
Dummett M (1992). ‘The metaphysics of verificationism.’
In Hahn L E (ed.) The philosophy of A. J. Ayer. La Salle,
IL: Open Court. 129–148.
Dummett M (1993). ‘Realism and anti-realism.’ In The seas
of language. Oxford: Oxford University Press. 462–478.
Field H (1980). Science without numbers. Princeton:
Princeton University Press.
Mackie J L (1977). Ethics: inventing right and wrong.
Harmondsworth: Penguin.
Quine W V O (1970). Philosophy of logic. Englewood
Cliffs: Prentice Hall.
Russell B (1956). ‘On denoting.’ In Marsh R (ed.) Logic and
knowledge. London: George Allen and Unwin. 41–56.
Wright C (1993). Realism, meaning and truth. Oxford:
Blackwell.
the reconstructed morphemes and words is an evident
desideratum. While this may sound straightforward,
interpreting the lexical evidence that emerges from
historical and comparative linguists’ occupation
with formal matters of phonology and morphology
is often not at all easy: semantic change is very much a
matter of (human) tendencies rather than (scientific)
rules (see, e.g., Fortson, 2003), relying on cultural
situations and norms that may be hard for outsiders
to fathom, especially outsiders who are looking back
on a bygone civilization. It is true that even some
phonological changes can seem non-evident unless
taken apart and examined very closely, but in histori-
cal semantics, the ‘surprise factor’ is always there.
Without an understanding of culture-specific religious
and legal rituals, for example, it is not possible to
explain why there should be a morpho-phonological
relationship between the English words blood and
bless (the verb originally meant ‘consecrate with
blood’) or exactly how the Latin word testis (
*trit(j)o-sth2-o-, literally ‘standing as third’) could
have come to mean in the first place ‘witness’ and
then also ‘testicle’ (see Katz, 1998).
Whatever one may think of the Sapir-Whorf
hypothesis, there is clearly some sort of relationship
between worldview and linguistic expression, and the
goal of cultural reconstruction (sometimes known as
390 Reconstruction, Cultural
linguistic paleontology) is to use language and ety-
mology to gain access to the way people talked about
their environment. Forms in related languages may
lead scholars to reconstruct proto-words for ‘fish’ and
‘loom’, but it requires more than a reckoning of
sounds and morphemes to determine with any degree
of plausibility just what species of fish speakers were
referring to or exactly what kind of instrument they
actually used for weaving. There is of course an im-
portant difference between fish and looms, namely
that the one is natural, the other man-made. When
scholars speak of cultural reconstruction, they fre-
quently mean to encompass both. Perhaps largely
because of the inordinately great role that the evi-
dence of words for kinds of fish and trees – especially
the salmon (Diebold, 1985 gives a salutary account)
and the beech (see, e.g., Friedrich, 1970: 106–115) –
has played in scholars’ highly controversial attempts
to specify the homeland (‘Urheimat’) and early migra-
tions of the Indo-Europeans (see, e.g., Mallory, 1989),
the reconstructed vocabulary of the natural world has
historically figured more prominently than words for
items of material and intellectual culture. My empha-
sis in what follows is on the latter, as well as (in part
on account of my own limited competence) on Indo-
European peoples. (Most textbooks on historical
and comparative linguistics rely heavily on Indo-
European evidence, but two good ones that do not
and that do discuss cultural reconstruction are Crow-
ley, 1997, whose examples come mostly from Austra-
lia and the Pacific, and Campbell, 2004, who stresses
the languages of the Americas.)
In the earliest days of scientific linguistics, the rela-
tionship between proto-language and proto-culture
was taken for granted (Jacob Grimm, for example,
was interested as much in what it meant to be German
as in what it meant to speak it), but the rise of Neo-
grammarianism in the late 19th century led to a heavy
emphasis – still not overcome – on formal methodol-
ogy and the view that the concomitant investigation
of cultural questions (the so-called ‘Wörter und
Sachen’-approach, championed most of all at that
time by Otto Schrader: see especially Schrader, 1907
and 1917–1929) was at best the poor stepsister. (It is
telling that there is no chapter on cultural reconstruc-
tion in the latest compendium of research on historical
linguistics, Joseph and Janda, 2003.) The situation is
slowly changing, though the adherence of certain
scholars to repugnant ideological (racist or national-
istic) interpretations of linguistic data has not helped
raise the status of reconstructive work in academia
(see, e.g., Lincoln, 1999). As for the relative status
of linguistic techniques vis-à-vis others in the
reconstruction of prehistory (Day, 2001 provides a
fine synthesis), it remains to be seen what will come
of the widely publicized and aggressive battles betw-
een linguists and archaeologists (Blench and Spriggs,
1997–1999 is a very useful collection, especially in
this context volume 2 [1998], subtitled Correlating
archaeological and linguistic hypotheses) and now
also geneticists and evolutionary biologists (see, e.g.,
Gray and Atkinson, 2003, whose time-scale nearly all
linguists consider impossible).
Cultural reconstruction is above all part of the
history of ideas, a way into the metaphors people
live and lived by: that we surf the web, for example,
and that the Navajo call corn , literally
‘enemy-food’, because they learned of it from the
Pueblo and that early Indo-Europeans personified
evening as ‘the shrouded one’ because in their mythol-
ogy, the sky got dressed at the end of the day (see
Janda, 2000: 200–211 and Katz, 2000). Coming to
grips with the nuances of words is often essential for
determining the pragmatic fulcrum of semantic
change and, more generally, provides a good window
onto a culture’s organization and values: what it does
with the natural materials found in the environment;
how it functions sociopolitically and economically at
both the micro (family) and the macro level (society
as a whole); what its legal and religious systems are;
and how it expresses itself artistically. (The most
remarkable works that cover all these topics are
Benveniste, 1969 and the papers collected in Watkins,
1994: volume 2; see also Buck, 1949, Gamkrelidze
and Ivanov, 1984: volume 2, and Mallory and
Adams, 1997 [plus Zimmer, 1999], as well as the
good summary accounts in Watkins, 2000: vii–xxxv
and passim and Fortson, 2004: 16–47.) Let us briefly
consider each of these areas in turn.
It is largely uncontroversial that the Proto-Indo-
Europeans were an agricultural people who plowed
(*h2erh3-: Latin arāre, whence English arable) their
fields (*h2eĝ-ro-: Latin ager, as well as English acre; a
derivative of the root *h2eĝ- ‘drive’), cultivated grains
(probably at least wheat or barley, though precise
linguistic reconstructions are difficult), bred cattle
(*gwow-: Latin bōs and English cow) and other live-
stock (*pek̂u-: Latin pecū and, interestingly, English
fee [< Old English feoh, in effect ‘valuable four-
legged property’]), made ample use of dairy products
(a variety of words can be reconstructed more or less
precisely, including the predecessor(s) of Latin lac and
English milk), and spun (*(s)neh1-: Latin nēre; cf. also
English needle) wool (*h2/3wlh1-neh2-). Questions
about the use of native animals,˚ vegetables, and min-
erals that generate somewhat more excitement con-
cern two related matters: the domestication of the
horse and the invention of the wheel and evolution
of the chariot. The distribution of the preform
*kwe-kwl(h1)-o-across the Indo-European languages

(e.g., English wheel and Tocharian B kokale ‘chari-
ot’), coupled with some archaic features that appear
to be part of the paradigm’s accentual pattern, makes
it virtually certain on linguistic grounds that the
Proto-Indo-Europeans had at least this one word for
‘wheel’, a derivative of the root *kwel(h1)- ‘turn’
(a second widespread word is based on the root
*ret-‘run’). This is interesting for a number of rea-
sons, not least because the description of the Proto-
Indo-Europeans’ world sketched in the first sentence
of this paragraph suggests a late Neolithic society of
the 4th millennium B.C., which is precisely when we
find the first archaeological evidence for wheels (ca.
3300 B.C.), an invention that would, not very surpris-
ingly, have been referred to in the first instance with a
wholly transparent neologism (see Fortson, 2004:
38). The place and date of the domestication
(*demh2-: Latin domāre and English tame) of the
horse (*h1ek̂wo-: Latin equus, whence English
equine) is more controversial still, and it must suffice
here to state that there is possible evidence for tamed
horses in Eurasia around 3500 B.C.
Indo-European society was certainly patriarchal
and patrilineal (its kinship system is generally thought
to have originally been more or less of the Omaha
type, but see Hettrich, 1985 for an interesting dis-
sent); accordingly, we know more about men’s
exploits than women’s, though E J W Barber (see,
e.g., Barber, 1991) has deftly reconstructed the lin-
guistic and material world of ‘women’s work’, the
production of textiles. The self-designation of the
Proto-Indo-Europeans is unfortunately not wholly
clear and neither are the oldest words for ‘clan’,
though it is good to remember that civilizations
the world over frequently call themselves simply
‘The People’ (e.g., Navajo Diné), which is roughly
the meaning of the reasonably widespread preform
*teuteh2- of, for example, Dutch and Deutsch (the
German word for ‘German’). The leader of the tribe
was the *h3rēĝ- ‘king’ (built to the root *h3reĝ-
‘straighten; rule’), a word preserved in Latin rēx and
found in such Germanic names as Richard (< Old
High German Rı̄co-hard ‘strong in rule’), not to men-
tion Theodoric and Dietrich, which, indeed, both go
back precisely to a compound of *teuteh2- plus
*h3rēĝ-, ‘people’s king’. An issue that has exercised
scholars for the past couple of generations is the
utility of Georges Dumézil’s ‘trifunctionalism’ as a
model of Proto-Indo-European society. According to
Dumézil (see, e.g., Dumézil, 1958), free males
belonged to one of three classes: kings and priests
(first function), warriors (second), and farmers and
other commoners (third). The linguistic and philolog-
ical evidence for this – and for its wider application to
divinities – remains disputed.
Reconstruction, Cultural 391
Because legal and religious texts tend cross-cultur-
ally to be repositories of linguistic archaism, it is a
priori likely that investigating the language of a num-
ber of such documents will tell us a great deal about
the proto-culture in question. The vocabulary of
Indo-European law (‘that which is laid down’
< *legh- ‘lie, lay’, a root that some believe also lies
behind Latin lēx) has not yet received as much atten-
tion as it deserves, but it seems likely that the princi-
pal words for ‘law’ were *jows (cf. Lat. iūs and the
compound iū-dex ‘judge’, literally ‘law-pronouncer’)
and one or more derivatives of the root *dheh1- ‘put’,
as in Classical Greek thémis, Vedic Sanskrit dhā´ma-,
and English deed and doom (which originally meant
‘judgment’, witness doomsday). It is noteworthy (see
Watkins, 1995: 434, with n. 5) that the very first
attested Indo-European word that has come down
to us is a 19th-century B.C. borrowing into Old
Assyrian of (pre-)Hittite išhiul ‘contract’ (a derivative
˘
of the verb for ‘bind’). Recent work on Greek and
Indic marriage by Stephanie W Jamison (see, e.g.,
Jamison, 1999) has added considerably to our under-
standing of Indo-European legal rights and responsi-
bilities, including – remarkably – those of women. As
for religion (Latin religiō is probably etymologically
‘that which binds (-lig-)’, a derivative of *leiĝ-), the
Proto-Indo-Europeans were polytheistic, with the
head, or ‘father’ (*ph2ter-), of the gods – addressed
in the vocative as Vedic Sanskrit dyàus. pı́tar, Greek
Zeû páter, and Latin Iū-piter (i.e., Jupiter) – the deifi-
cation of the bright sky, *djew-. It is worth noting
that the Latin word deus (as in ‘dei-fication’) is a
derivative of this same root, whereas Greek theós (as
in ‘poly-the-istic’), like also Latin fēstus (whence our
word festival), goes back to *dheh1s-, which may well
be an extension of the root for ‘put’ noted above.
The English word god, incidentally, probably derives
from *ĝ hew- ‘pour’ and refers to the spirit immanent
in a tumulus, perhaps a kurgan, the Eurasian burial
mound that some archeologists, notably Marija
Gimbutas, associate with the Proto-Indo-Europeans
(see Watkins, 2000: 31, resuming an argument made
in 1974; Kupfer, 2004 presents a different and to my
mind less convincing picture of the details). Our
knowledge of Proto-Indo-European religious culture
has recently been greatly expanded in two surpassing-
ly brilliant books by Michael Janda (Janda, 2000,
2005).
Finally, there is verbal art, a cover term that sub-
sumes a range of genres in which stunning archaisms
and delightful new turns of phrase are woven together
in a special way. In Proto-Indo-European society, the
poet (for whom there is no single reconstructible
designation) was one of the most highly regarded
figures, and thanks to two profoundly influential
392 Reconstruction, Cultural
works, Schmitt, 1967 and Watkins, 1995, and the
further research they have inspired (e.g., Matasović,
1996), the nuances of what is usually called ‘Dichter-
sprache’ are becoming ever more apparent. It is clear
that at least some metrical and strophic forms found
in the individual daughter languages go back to Proto-
Indo-European times and that certain artful tech-
niques were part of the repertoire of bards who told
tales, which then became the myths of the separate
Indo-European peoples (see, e.g., Puhvel, 1987) – of
the beginning of the world and of glory in battle and
of fights against monsters. For comments on the ad-
vances in and pitfalls of attempting to reconstruct the
details of the genre of epic in Proto-Indo-European,
see Katz, 2005; the inherited tradition of riddles is the
subject of Katz, forthcoming.
Cultural reconstruction in effect puts history back
into historical linguistics, and this is a difficult busi-
ness, for it involves expanding formal linguistic meth-
ods into the larger and messier world of realia, of real
life as it is and was once lived. Still, once we acknowl-
edge the simple truth that real human beings spoke
the languages that we are doing our best to recon-
struct, it becomes part of our job as linguists to take
account of, and reconstruct as best as we can, the
circumstances in which these languages were used
and under which they developed.
Bibliography
Barber E J W (1991). Prehistoric textiles: the development of
cloth in the Neolithic and Bronze Ages with special refer-
ence to the Aegean. Princeton: Princeton University Press.
Benveniste É (1969). Le Vocabulaire des institutions
indo-européennes (2 vols). Paris: Éditions de Minuit.
[English translation: Indo-European language and soci-
ety. Palmer E (trans.). Coral Gables: University of Miami
Press, 1973.]
Blench R & Spriggs M (eds.) (1997–1999). Archaeology
and language (4 vols). [I: Theoretical and methodological
orientations (1997); II: Correlating archaeological and
linguistic hypotheses (1998); III: Artefacts, languages
and texts (1999); and IV: Language change and cultural
transformation (1999)]. London: Routledge.
Buck C D (1949). A dictionary of selected synonyms in the
principal Indo-European languages: a contribution to the
history of ideas. Chicago: University of Chicago Press.
Campbell L (2004). Historical linguistics: an introduction
(2nd edn.). Cambridge, MA: MIT Press.
Crowley T (1997). An introduction to historical linguistics
(3rd edn.). Auckland: Oxford University Press.
Day J V (2001). Indo-European origins: the anthropological
evidence. Washington, DC: Institute for the Study of Man.
Diebold A R Jr (1985). The evolution of Indo-European
nomenclature for salmonid fish: the case of ‘huchen’
(Hucho spp.). Washington, DC: Institute for the Study
of Man.
Dumézil G (1958). L’Idéologie tripartie des indo-européens.
Brussels: Latomus.
Fortson B W IV (2003). ‘An approach to semantic change.’
In Joseph & Janda (eds.). 648–666.
Fortson B W IV (2004). Indo-European language and cul-
ture: an introduction. Malden, MA: Blackwell.
Friedrich P (1970). Proto-Indo-European trees: the arboreal
system of a prehistoric people. Chicago: University of
Chicago Press.
Gamkrelidze T V & Ivanov V V (1984). Indoevropejskij
jazyk i indoevropejtsy: rekonstruktsija i istoriko-tipologi-
cheskij analiz prajayzka i protokul’tury (2 vols). Tbilisi:
Izdatel’stvo Tbilisskogo Universiteta. [English translation:
Indo-European and the Indo-Europeans: a reconstruction
and historical analysis of a proto-language and a proto-
culture. Nichols J (trans.). Berlin: de Gruyter, 1995.]
Gray R D & Atkinson Q D (2003). ‘Language-tree
divergence times support the Anatolian theory of Indo-
European origin.’ Nature 426(6965), 435–439.
Hettrich H (1985). ‘Indo-European kinship terminology
in linguistics and anthropology.’ Anthropological Lin-
guistics 27, 453–480.
Jamison S W (1999). ‘Penelope and the pigs: Indic perspec-
tives on the Odyssey.’ Classical Antiquity 18, 227–272.
Janda M (2000). Eleusis: Das indogermanische Erbe der
Mysterien. Innsbruck: Institut für Sprachwissenschaft der
Universität Innsbruck.
Janda M (2005). Elysion: Entstehung und Entwicklung der
griechischen Religion. Innsbruck: Institut für Sprachen
und Literaturen der Universität Innsbruck.
Joseph B D & Janda R D (eds.) (2003). The handbook of
historical linguistics. Malden, MA: Blackwell.
Katz J T (1998). ‘Testimonia ritus italici: male genitalia,
solemn declarations, and a new Latin sound law.’
Harvard Studies in Classical Philology 98, 183–217.
Katz J T (2000). ‘Evening dress: the metaphorical back-
ground of Latin uesper and Greek hésperos.’ In Jones-
Bley K, Huld M E & Della Volpe A (eds.) Proceedings of
the Eleventh Annual UCLA Indo-European Conference,
Los Angeles, June 4–5, 1999. Washington, DC: Institute
for the Study of Man. 69–93.
Katz J T (2005). ‘The Indo-European context.’ In Foley
J M (ed.) A companion to ancient epic. Malden, MA:
Blackwell. 20–30.
Katz J T (forthcoming). ‘The riddle of the sp(h)ij-: the Greek
Sphinx and her Indic and Indo-European background.’
In Pinault G-J & Petit D (eds.) Langue poétique indo-
européenne. Louvain: Peeters.
Kupfer K (2004). ‘Zur Etymologie von nhd. ‘‘Gott’’ und
deren Implikationen für die nominalen Kategorien des
Indogermanischen.’ In Fritz M & Wischer I (eds.) Histor-
isch-vergleichende Sprachwissenschaft und germanische
Sprachen: Akten der 4. Neulandtagung der Historisch-
vergleichenden Sprachwissenschaft in Potsdam 2001.
Innsbruck: Institut für Sprachen und Literaturen der
Universität Innsbruck. 119–128.
Lincoln B (1999). Theorizing myth: narrative, ideology, and
scholarship. Chicago: University of Chicago Press.
Mallory J P (1989). In search of the Indo-Europeans: lan-
guage, archaeology and myth. London: Thames & Hudson.

Mallory J P & Adams D Q (eds.) (1997). Encyclopedia of
Indo-European culture. London: Fitzroy Dearborn.
Matasović R (1996). A theory of textual reconstruction in
Indo-European linguistics. Frankfurt am Main: Lang.
Meillet A (1925). La Méthode comparative en linguis-
tique historique. Oslo: Aschehoug. [English translation:
The comparative method in historical linguistics.
Ford G B Jr (trans.). Paris: Champion, 1967.]
Puhvel J (1987). Comparative mythology. Baltimore: Johns
Hopkins University Press.
Schmitt R (1967). Dichtung und Dichtersprache in indo-
germanischer Zeit. Wiesbaden: Harrassowitz.
Schrader O (1907). Sprachvergleichung und Urgeschichte:
Linguistisch-historische Beiträge zur Erforschung des
indogermanischen Altertums (3rd edn., 2 vols). Jena:
Costenoble. [First edition: 1883.]
Reconstruction, Morphological
B Fortson, University of Michigan, Ann Arbor, MI, USA
ß 2006 Elsevier Ltd. All rights reserved.
Introduction
The reconstruction of the morphology of a proto-
language is beset with certain difficulties that do not
attend the reconstruction of phonology. Peeling back
the layers of historical change and undoing them – the
essence of reconstruction – is in principle easy when
the changes are regular and exceptionless, as is the
case with sound change (at least under the Neogram-
marian model). Morphological change, as is well
known, is quite a different kettle of fish; however,
the usual characterization of it as entirely irregular
is perhaps too strong, and depends on where one
seeks to localize a given change and how one defines
‘morphological change’ in the first place.
Certain changes called morphological change are in
fact lexical changes, that is, they result in a change to
individual items in the lexicon and do not affect the
morphological system of the language. (Even a lexical
change that removes from the language the last
remaining example of a once-productive morphologi-
cal process has merely erased all evidence that such a
process once existed; since the process was no longer
productive anyway, the morphological system was
unaffected by the word’s disappearance. An example
of such a case is the dialectal English generalization
of was throughout the preterite paradigm of the
verb be that removed the last example of the once-
common distinction between the singular and plural
in the preterite.) The ‘morphological system’ will
be, for convenience, here defined as the sum total
of elements and rules needed for derivation and
Reconstruction, Morphological 393
Schrader O (1917–1929). Reallexikon der indogerma-
nischen Altertumskunde: Grundzüge einer Kultur-und
Völkergeschichte Alteuropas (2 vols). (2nd edn., ed. by
Nehring A.) Berlin: de Gruyter. [First edition: 1901.]
Watkins C (1994). Selected writings (2 vols). Oliver L (ed.).
Innsbruck: Institut für Sprachwissenschaft der Universität
Innsbruck.
Watkins C (1995). How to kill a dragon: aspects of Indo-
European poetics. New York: Oxford University Press.
Watkins C (2000). The American Heritage dictionary
of Indo-European roots (2nd edn.). Boston: Houghton
Mifflin.
Zimmer S (1999). ‘Comments on a great book.’ Journal of
Indo-European Studies 27, 105–163. [Review of Mallory
& Adams (eds.).]
inflection. Examples of changes that are often called
morphological changes but whose effects are typical-
ly lexical in scope include analogy, paradigm leveling
and split, folk etymology, false segmentation (of the
type [a] nadder > [an] adder), double-marking, back-
formation, contamination, and taboo deformation.
As lexical changes, they are by definition unpredict-
able and irregular, and their effects can greatly ob-
scure morphological history (as well as phonological
history).
When the effects of these changes cause a learner
to reanalyze part of the data as reflecting a morpheme
or rule that was not present in the source grammar,
then the learner’s morphology will come to be dif-
ferent from the morphology of the source grammar
and true morphological change will have occurred.
A true morphological change, involving (for example)
a new morphological rule, is in fact ‘regular’ in that it
affects in the same way all the members of whatever
class or category it applies to. (The class may be small
and limited in number, or it may be open, as is the
case with the nouns to which the s-plural rule applies
in English.)
Once a new productive rule enters the language, it
will usually erase traces of earlier rules that applied
to the same forms. But it is to be expected that the
morphology of the most frequent lexemes in a lan-
guage will have the least opportunity of being over-
looked by a learner, and that these items will stand
the best chance of surviving in the face of morpholog-
ical change. For this reason, irregular forms – those
that are not generable by the productive morphologi-
cal rules of a language – are highly represented among
a language’s most frequent items. These exceptions
are therefore a valuable source of information about

Mallory J P & Adams D Q (eds.) (1997). Encyclopedia of
Indo-European culture. London: Fitzroy Dearborn.
Matasović R (1996). A theory of textual reconstruction in
Indo-European linguistics. Frankfurt am Main: Lang.
Meillet A (1925). La Méthode comparative en linguis-
tique historique. Oslo: Aschehoug. [English translation:
The comparative method in historical linguistics.
Ford G B Jr (trans.). Paris: Champion, 1967.]
Puhvel J (1987). Comparative mythology. Baltimore: Johns
Hopkins University Press.
Schmitt R (1967). Dichtung und Dichtersprache in indo-
germanischer Zeit. Wiesbaden: Harrassowitz.
Schrader O (1907). Sprachvergleichung und Urgeschichte:
Linguistisch-historische Beiträge zur Erforschung des
indogermanischen Altertums (3rd edn., 2 vols). Jena:
Costenoble. [First edition: 1883.]
Reconstruction, Morphological
B Fortson, University of Michigan, Ann Arbor, MI, USA
ß 2006 Elsevier Ltd. All rights reserved.
Introduction
The reconstruction of the morphology of a proto-
language is beset with certain difficulties that do not
attend the reconstruction of phonology. Peeling back
the layers of historical change and undoing them – the
essence of reconstruction – is in principle easy when
the changes are regular and exceptionless, as is the
case with sound change (at least under the Neogram-
marian model). Morphological change, as is well
known, is quite a different kettle of fish; however,
the usual characterization of it as entirely irregular
is perhaps too strong, and depends on where one
seeks to localize a given change and how one defines
‘morphological change’ in the first place.
Certain changes called morphological change are in
fact lexical changes, that is, they result in a change to
individual items in the lexicon and do not affect the
morphological system of the language. (Even a lexical
change that removes from the language the last
remaining example of a once-productive morphologi-
cal process has merely erased all evidence that such a
process once existed; since the process was no longer
productive anyway, the morphological system was
unaffected by the word’s disappearance. An example
of such a case is the dialectal English generalization
of was throughout the preterite paradigm of the
verb be that removed the last example of the once-
common distinction between the singular and plural
in the preterite.) The ‘morphological system’ will
be, for convenience, here defined as the sum total
of elements and rules needed for derivation and
Reconstruction, Morphological 393
Schrader O (1917–1929). Reallexikon der indogerma-
nischen Altertumskunde: Grundzüge einer Kultur-und
Völkergeschichte Alteuropas (2 vols). (2nd edn., ed. by
Nehring A.) Berlin: de Gruyter. [First edition: 1901.]
Watkins C (1994). Selected writings (2 vols). Oliver L (ed.).
Innsbruck: Institut für Sprachwissenschaft der Universität
Innsbruck.
Watkins C (1995). How to kill a dragon: aspects of Indo-
European poetics. New York: Oxford University Press.
Watkins C (2000). The American Heritage dictionary
of Indo-European roots (2nd edn.). Boston: Houghton
Mifflin.
Zimmer S (1999). ‘Comments on a great book.’ Journal of
Indo-European Studies 27, 105–163. [Review of Mallory
& Adams (eds.).]
inflection. Examples of changes that are often called
morphological changes but whose effects are typical-
ly lexical in scope include analogy, paradigm leveling
and split, folk etymology, false segmentation (of the
type [a] nadder > [an] adder), double-marking, back-
formation, contamination, and taboo deformation.
As lexical changes, they are by definition unpredict-
able and irregular, and their effects can greatly ob-
scure morphological history (as well as phonological
history).
When the effects of these changes cause a learner
to reanalyze part of the data as reflecting a morpheme
or rule that was not present in the source grammar,
then the learner’s morphology will come to be dif-
ferent from the morphology of the source grammar
and true morphological change will have occurred.
A true morphological change, involving (for example)
a new morphological rule, is in fact ‘regular’ in that it
affects in the same way all the members of whatever
class or category it applies to. (The class may be small
and limited in number, or it may be open, as is the
case with the nouns to which the s-plural rule applies
in English.)
Once a new productive rule enters the language, it
will usually erase traces of earlier rules that applied
to the same forms. But it is to be expected that the
morphology of the most frequent lexemes in a lan-
guage will have the least opportunity of being over-
looked by a learner, and that these items will stand
the best chance of surviving in the face of morpholog-
ical change. For this reason, irregular forms – those
that are not generable by the productive morphologi-
cal rules of a language – are highly represented among
a language’s most frequent items. These exceptions
are therefore a valuable source of information about
394 Reconstruction, Morphological
earlier morphological rules in a language, and gave
rise to the oft-repeated maxim that morphological
reconstruction proceeds on the basis of exceptions.
Following this maxim does not immunize the recon-
structivist from pitfalls, however, and it is to these that
we now turn.
Reconstructing on the Basis of
Exceptions
Synchronic innovatory morphological processes with
diachronic implications include primarily analogy, an
imprecise (and sometimes over-invoked) term that
will be used here for convenience to cover any exten-
sion of a morphological process to a lexeme to which
it previously did not apply. Forms are innovated daily
in this way; some of them get picked up by other
speakers, make it into the historical record, or both.
Since analogy involves the extension of synchroni-
cally productive morphological processes, any and all
forms produced by productive processes can be sus-
pected of being historically recent creations, gener-
ated on the fly by individual speakers. (‘Productive’ is
used here in its broadest sense, to describe any process
that can be analogically extended.) If two related lan-
guages have a particular productive process in com-
mon (say, the -s plural of English and German), the
presence of cognate or identical forms exhibiting
that process (e.g., English pl. sofas, German Sofas)
in no way guarantees that the languages’ ancestor had
those forms too (we would not want to reconstruct
a Common West Germanic paradigm *sofa, *sofas).
It is only when cognate forms are found exhibiting
the same synchronically unproductive process that
cognacy of both the process and the forms can be
entertained. Thus, the equation of is/were with
German ist/waren, evincing identical and unpredict-
able suppletion, strongly suggests common inheri-
tance of the two paradigms.
Even here, though, one can be led astray by parallel
independent developments. The equation of English
mouse/mice with German Maus/Mäuse is good prima
facie evidence for a Common West Germanic word
for ‘mouse’ that formed its plural by umlauting, ex-
cept that we know on independent grounds that um-
laut happened well after the breakup of Common
West Germanic, and additionally spread to German
only some time after it had run its course in English.
Without such knowledge, however, the comparative
method would dictate the reconstruction of an um-
lauting paradigm for proto-West Germanic, since in
both English and German, umlaut is not a regular and
predictable process of forming plurals.
Similarly, one of the ways of forming a perfect in
Latin involved changing a root vowel a or ā to ē or e,
as in the 1st person singular perfects ēgı̄ cēpı̄ fēcı̄ stetı̄
dedı̄ from (infin.) age-re cap-ere fac-ere stā-re da-re.
Latin’s relative Oscan had this process too, as exhib-
ited by hipid (historically *hēbēt), the 3rd person
singular perfect subjunctive of the verb corresponding
to Latin habeō ‘have.’ A perfect *hebuı̄ or *hēbı̄, not
attested for Classical Latin, was analogically created
at some point in the pre-Romance period, because
certain forms of the Italian simple past (1st sing.
hebbi, 3rd sing. hebbe, 3rd pl. hebbero) continue a
perfect stem with an ēˇ. If all we had at our disposal
was Oscan hipid and Italian hebbi, hebbe (plus a
knowledge of the sound changes), we would probably
be tempted to reconstruct *hēˇb- as this verb’s perfect
in the common ancestor of Italian and Oscan, namely
Proto-Italic. (Students of these languages will know
that the data are being fudged a bit here for illustrative
purposes; the long-vowel perfects like ēgı̄ are histori-
cally quite different from the short-vowel perfects like
stetı̄.) Conservatively, we are licensed to project only
the a/e pattern onto Proto-Italic, rather than (without
external evidence) the exhibition of that pattern by a
particular lexeme. And the pattern itself can only be
projected onto the protolanguage if it cannot be
shown to be independently innovated in its daughters.
When cognates are compared across several lan-
guages and some of the cognates are morphologically
peculiar in their traditions and others are not, if the
morphological peculiarities are comparable, then the
simplest hypothesis is that they are inherited and
the other languages have regularized the peculiari-
ties. A well-known example of this concerns the re-
construction of the Proto–Indo–European word for
‘daughter-in-law,’ a feminine noun that has the ex-
pected feminine paradigm in *-ā- in some languages
(e.g., Sanskrit snus. ā´, Old English snoru, Old Church
Slavonic snŭkha) but an unexpected o-stem paradigm
in others (o-stems are usually masculine or neuter;
e.g., Greek nuós, Latin nurus, Armenian now [genit.
nowoy]). Since there is no reason for speakers to
have remodeled an inherited ā-stem feminine as an
exceptional o-stem feminine, the o-stem forms are
surely inherited, and the ā-stem forms independent
regularizations.
Less clear-cut, of course, are cases where determin-
ing what is regular and what is irregular is a whole
task unto itself. Perfects of the Latin 3rd conjugation
were built with a multitude of different formations,
essentially all of which underwent some analogical
spread at one time or another. The class was closed –
practically no new verbs during the Classical period
entered the 3rd conjugation – meaning there is little
evidence for which, if any, of these formations
was a default for the class. Comparisons between
3rd-conjugation perfects and cognate forms in other

Indo-European languages must therefore be made care-
fully, with as detailed an understanding as possible of
the histories of the different classes.
The dictum to reconstruct on the basis of excep-
tions also invites caution for another reason, namely
that exceptional forms themselves are not necessarily
more archaic, and can lead the linguist astray in the
absence of more detailed data. The history of the
perfect plural stem in Sanskrit is a good case in
point. Perfects of roots of the shape CaC were formed
in Sanskrit by reduplicating the root-initial conso-
nant; thus to par ‘fill’ was formed pa-par-, the singular
perfect stem. The dual and plural perfect stem was
formed in the same way but with absence of the
root vowel, hence pa-pr-, with an internal consonant
cluster. As a result of sound changes, the prehisto-
ric dual/plural perfect stems *sa-sd- (root sad ‘sit’),
*ya-ym- (root yam ‘extend’), and *ya-yt- (root yat
‘stretch’) became sed-, yem-, yet-. The resulting
pattern singular sasad-: plural sed-, yayam-: yem-,
yayat-: yet- was gradually extended to other roots.
A linguist looking at this stage of Sanskrit (a bit
anterior to the earliest attested stage), simply going by
the raw numbers where perfect plural stems of the
type papr- outnumbered those of the type sed-, might
decide that the sed- forms, being exceptional, were
older. In this case, though, they are the rising upstarts.
A look at a later stage would reveal a different situa-
tion. By the time of the Rigveda (the earliest Sanskrit
text), ten CaC-roots that originally formed their per-
fects with reduplication in the plural stem have taken
over the sed- pattern (e.g., pec- to the root pac
‘cook’), in most cases completely ousting reduplica-
tion, and an equal number preserve reduplication; by
the later Classical stage, the number of roots preserv-
ing reduplication had been reduced to five (Whitney,
1889: 284–285). The linguist looking at late prehis-
toric Sanskrit would obviously need to consider other
factors than raw numbers in order to come to the
correct conclusion.
Morphotactic Comparison
Cognate languages sometimes agree in having iden-
tical rules for the concatenation of derivational or
inflectional morphemes (morphotactic rules), but
without cognacy of the actual morphemes (as in
Siouan languages). In such instances, some linguists
have been tempted to apply the comparative method
to the morphological templates (the sequences of
fixed positions taken by the morphemes), and claim
that the templates are inherited even if there has
been innovation in the morphemes themselves. (An
analogy in the lexical domain would be to claim that
a protoparadigm was suppletive for a particular word
Reconstruction, Morphological 395
if the daughter languages show suppletion in that
word but with different suppletive forms, such
as Russian let ‘years’ [genit. pl.], god [sing.] compared
to Polish lata [pl.], rok [sing.].)
How valid the results of such comparison are is a
somewhat vexed question. Under traditional under-
standing, morphological reconstruction cannot oper-
ate without cognate morphemes to compare; and
certainly if the goal is to reconstruct the morphologi-
cal inventory of a parent language, comparison of
morphotactics alone is bootless. Additionally, since
inflectional morphemes are positioned by syntactic
rules, what is really being compared in this scenario
is the syntactic systems of two cognate languages. In
either case, little information about the morphology
of the protolanguage can be gleaned.
Similar situations occur outside the realm of con-
catenative rules, as when two cognate languages
share the same case-system but different case-endings.
The two Tocharian languages have, between them,
seven innovated nominal cases, the endings of three of
which (comitative, allative, and ablative) are not cog-
nate. Does one then reconstruct for Proto-Tocharian
the existence of those three desinential categories for
Proto-Tocharian in the absence of cognacy of the
morphemes themselves? Clearly the answer is no,
since if one posits this system, the development from
it to the attested Tocharian languages would entail
the loss and replacement of all three original endings
(either all of them in one language, or some of them in
one language and the rest in the other). More likely,
Proto-Tocharian had a collection of postpositions,
some of which underwent independent but parallel
grammaticalization in its two daughters.
Convergent evolution must not be overlooked as a
source of morphological similarity. The two unrelated
Caucasian languages Georgian and Udi (the former is
Kartvelian, the latter Lezgian [Northeast Caucasian])
have the same preverb–plus–verb template, in which
an agreement marker intervenes between preverb
and verb, but the two systems have entirely different
sources within their respective families (Harris, 2003).
From within a single family, the Slavic languages have
a category of animacy in nominal paradigms, whereby
accusatives of animate nouns have the same endings
as the genitive, rather then the ‘regular’ accusative
(of inanimates). Each Slavic language today has a
different system. In Russian, all noun classes partici-
pate in it; in Novi Serbo-Croatian, only o-stems do.
In Old Church Slavonic, which was fairly close to
reconstructed Common Slavic, this morphosemantic
feature was extremely limited – to the accusative
singular in o-stems. The system has therefore expand-
ed in each Slavic language independently and partly
along parallel lines; if Old Church Slavonic were
396 Reconstruction, Morphological
not attested, we would likely reconstruct too much of
it for Common Slavic (S. Thomason, personal com-
munication).
Internal Morphological Reconstruction
Based on Allomorphy
The synchronic establishment of underlying mor-
phemes on the basis of allomorphic alternations is
essentially the same thing as the internal reconstruc-
tion of a language’s morphological history. To take
a simple example, Latin has various present active
infinitive formations, exemplified by amāre ‘to
love,’ movēre ‘to move,’ capere ‘to take,’ audı̄re ‘to
hear,’ esse ‘to be,’ velle ‘to want,’ ferre ‘to carry.’ We
can isolate a suffix -re in the first four forms, but
the last three forms exhibit a doubling of a conso-
nant followed by -e. Examination of other forms in
the paradigms of these three verbs reveals the verb
stems to be es-, vel-, and fer-, indicating that the
infinitive suffix began with a consonant that assimi-
lated to the preceding consonant. Other evidence
from the language reveals that the consonant for
all these forms was once s, which was rhotacized
to r between vowels (hence amāre, movēre, etc.);
remained following another s (es-se); and assimilated
to a preceding liquid (whence vel-le, fer-re). From this
evidence, we can reconstruct *-se as the pre-Latin
present infinitive active suffix. (Additional evidence
would allow us to refine the reconstruction to *-si.)
Some forms of allomorphy are not so simple, and
involve what has often been called a subtraction mor-
pheme. In French, for example, there is no way to
predict that the feminine of /mi/ ‘sent’ (passive parti-
ciple) requires the addition of /-z/ while the feminine
of /fri/ ‘fried’ requires /-t/, and the feminine of /So/
‘cold’ requires /-d/. One can analyze this alternation
with the following rule: derive the masculine from the
feminine by deleting the final consonant of the fe-
minine. But many linguists doubt that such rules ac-
tually exist; they have proposed an alternate rule
for the French data, namely, posit the final consonant
as underlyingly present in the masculine forms, and
then posit a morphophonological rule that deletes the
final consonant. This does not involve actual subtrac-
tion of an element from one category to form another
category, but allows both forms to be independently
derived from a single underlying form. However,
under either of these scenarios, there is a serious
problem: a French speaker would be unable to form
the feminine of an adjective to which he or she only
knew the masculine. Various varieties of French, in
fact, have reanalyzed one of the possible consonants
coming at the end of feminines as a default feminine
marker (for example, [-z] in the case of Northern
French). Speakers thus, clearly analyzed the feminines
as derived from the masculines and, from the pool of
distinctive differences between the two, latched onto
one as the default.
Cases like this are quite common, and they typical-
ly come about in the same way: sound changes elimi-
nate the final sound or syllable from one set of forms
but not from another set where it was protected by
additional derivational or inflectional material. The
history behind the French facts is well known: what
synchronically appears as a derivational suffix (in this
case, /-z/, /-t/, or /-d/) is in reality the old stem-final
consonant. At an earlier stage, the masculines were
/miz/, /Sod/, and /frit/, while the feminines were
/mize/, /Sode/, and /frite/; /-e/ was the feminine suf-
fix. In time, deletion of word-final stops produced the
masculines /mi/, /So/, and /fri/, and subsequently
word-final schwa deletion produced the feminines
/miz/, /Sod/, and /frit/. For a more detailed discussion
of the process with other examples, see the important
study by Hale (1973).
The synchronic French data allow one to recon-
struct the whole history except for the phonetic shape
of the original feminine suffix (all that can be said for
sure, based on the evidence presented above, is that
there was one that was added to the stem-final conso-
nant). A more difficult situation exists in Hawaiian,
where the verb form known as the passive is formed
with the suffix [- ia] in almost all verbs, with a small
handful of exceptions where the consonant is some-
thing other than a glottal stop. Faced with these facts,
a lucky soul might hit upon the idea that -ia (which is
what the variant suffixes have in common) was the
original suffix, and the consonant preceding it was
an old stem-final consonant that had been deleted in
other forms where it would have been word-final;
but other analyses are also possible without addition-
al evidence (such as [- ia] is innovatory and one of
the exceptional variants was the original suffix). The
truth is that [- ia] was generalized in the same way
that [-z] was generalized as a feminine marker in
Northern French (only [- ia] has spread much, much
more widely than [-z].
Syncretism
All reconstruction is limited by the available evi-
dence. Syncretism, the collapse of inflectional systems
(such as case-systems) due usually to a combination
of phonological and analogical factors, can obliterate
evidence of original inflections. The modern Ro-
mance languages, for example, have lost all the
Latin case distinctions in nouns, and the few fossi-
lized remnants would not suffice to recreate any full
Latin paradigm (e.g., from French, the place-name

Aix that continues the Latin locative plural Aquı̄s
‘at the waters’; the pronoun rien ‘nothing’ < Latin
accusative singular rem ‘thing’; encore ‘again’ dimly
preserving the ablative singular ending -ā in (*)hinc
hā (h)ōrā, where the -o- continues *-au- < -ao- < *-ā
ō- [ō alone would have come out eu, as in heure <
hōrā]; and the noun chandeleur ‘Candlemas’ < Latin
genitive plural *candēlōrum).
The preservation of relic forms through seman-
tic specialization (essentially the process known as
Kuryłowicz’s Fourth Law of Analogy), leading to
their being learned as separate lexical items not
part of the paradigms of the words that they were
originally inflections of, is necessary for any kind of
internal reconstruction of inflectional changes such as
gain or loss of case-endings, paradigm leveling, and
paradigm split. In Latin, a v (also written u) was lost
between a consonant and u, leading to allomorphy
in some nominal and adjectival paradigms; but this
v/u was restored, so that what should have been, e.g.,
nominative singular ecus ‘horse,’ genitive equı̄, was
restored to equus equı̄, and what should have been
parus ‘small,’ genitive parvı̄ was restored to parvus
parvı̄. The restoration erased the effects of the sound
change, but luckily for scholars, forms like parum
‘too little,’ originally the accusative singular (from
parvum) which was specialized as an adverb, re-
mained because, due to their separation from their
original paradigm, they escaped the restoration of the
Reconstruction, Syntactic
S Thomason, University of Michigan, Ann Arbor,
MI, USA
ß 2006 Elsevier Ltd. All rights reserved.
Syntactic reconstruction has not figured prominent-
ly in historical linguistic investigations, as can be sur-
mised from the fact that the index of the recent 881-
page Handbook of historical linguistics (Joseph and
Janda, 2003) lists just seven pages, all in the same
article, where it is discussed. As Fox observes,
‘‘Syntactic reconstruction is a controversial area . . .
scholars working within the framework of the classi-
cal Comparative Method have been far less successful
in applying their methods here than in the case of
phonology or even morphology’’ (1995: 104; see also
Jeffers, 1976). And in discussing this topic elsewhere,
Fox does not point to any methods other than the
Comparative Method that have offered promising
results (1995: 104–109, 190–194, 250–253, 261–270).
Reconstruction, Syntactic 397
-v- and provide evidence for the leveling that must
have taken place elsewhere.
The above survey has touched on some of the prob-
lems that attend morphological reconstruction. Read-
ers interested in a listing of the different available
procedures should consult the carefully laid-out
material in Koch (1996).
Bibliography
Fisiak J (ed.) (1980). Historical morphology. The Hague:
Mouton.
Hale K (1973). ‘Deep-surface canonical disparities in rela-
tion to analysis and change: an Australian example.’ In
Sebeok T A (ed.) Current trends in linguistics 11: Dia-
chronic, areal, and typological linguistics. The Hague:
Mouton. 401–458.
Hammond M & Noonan M (eds.) (1988). Theoretical
morphology. San Diego: Academic Press.
Harris A C (2003). ‘Preverbs and their origins in Georgian
and Udi.’ In Booij G & van Marle J (eds.) Yearbook of
morphology 2003. Dordrecht: Kluwer. 61–78.
Hock H H (1991). Principles of historical linguistics
(2nd edn.). Berlin: de Gruyter.
Joseph B D & Janda R D (2003). The handbook of historical
linguistics. Oxford: Blackwell.
Koch H (1996). ‘Reconstruction in morphology.’ In Durie
M & Ross M (eds.) The comparative method reviewed.
Oxford: Oxford University Press. 218–263.
Whitney W D (1889). Sanskrit grammar (2nd edn.).
Cambridge: Harvard University Press.
Efforts to reconstruct syntax can be traced at least
as far back as 1893–1900, when Delbrück (as cited in
Lehmann, 1992: 32) reconstructed OV word order
for Proto-Indo-European. By far the most ambitious
early effort at reconstructing syntax is Schleicher’s
‘Proto-Indo-European fable’, which was considered
a rash enterprise even in those pre-Neogrammarian
times (1868; as cited and translated in Jeffers and
Lehiste, 1979: 107):
Avis akvāsas ka
‘A sheep and horses’
avis, jasmin varnā na ā ast, dadarka akvams, tam, vāg-
ham garum vaghantam, tam, bhāram magham, tam,
manum āku bharantam.
avis akvabhjams ā vavakat: kard aghnutai mai vidanti
manum akvams agantam. Akvāsas ā vavakant: krudhi
avai, kard aghnutai vividvant-svas: manus patis varnām
avisāms karnauti svabhjam gharman vastram avibhjams
ka varnā na asti.
Tat kukruvants avis agram ā bhugat.

Aix that continues the Latin locative plural Aquı̄s
‘at the waters’; the pronoun rien ‘nothing’ < Latin
accusative singular rem ‘thing’; encore ‘again’ dimly
preserving the ablative singular ending -ā in (*)hinc
hā (h)ōrā, where the -o- continues *-au- < -ao- < *-ā
ō- [ō alone would have come out eu, as in heure <
hōrā]; and the noun chandeleur ‘Candlemas’ < Latin
genitive plural *candēlōrum).
The preservation of relic forms through seman-
tic specialization (essentially the process known as
Kuryłowicz’s Fourth Law of Analogy), leading to
their being learned as separate lexical items not
part of the paradigms of the words that they were
originally inflections of, is necessary for any kind of
internal reconstruction of inflectional changes such as
gain or loss of case-endings, paradigm leveling, and
paradigm split. In Latin, a v (also written u) was lost
between a consonant and u, leading to allomorphy
in some nominal and adjectival paradigms; but this
v/u was restored, so that what should have been, e.g.,
nominative singular ecus ‘horse,’ genitive equı̄, was
restored to equus equı̄, and what should have been
parus ‘small,’ genitive parvı̄ was restored to parvus
parvı̄. The restoration erased the effects of the sound
change, but luckily for scholars, forms like parum
‘too little,’ originally the accusative singular (from
parvum) which was specialized as an adverb, re-
mained because, due to their separation from their
original paradigm, they escaped the restoration of the
Reconstruction, Syntactic
S Thomason, University of Michigan, Ann Arbor,
MI, USA
ß 2006 Elsevier Ltd. All rights reserved.
Syntactic reconstruction has not figured prominent-
ly in historical linguistic investigations, as can be sur-
mised from the fact that the index of the recent 881-
page Handbook of historical linguistics (Joseph and
Janda, 2003) lists just seven pages, all in the same
article, where it is discussed. As Fox observes,
‘‘Syntactic reconstruction is a controversial area . . .
scholars working within the framework of the classi-
cal Comparative Method have been far less successful
in applying their methods here than in the case of
phonology or even morphology’’ (1995: 104; see also
Jeffers, 1976). And in discussing this topic elsewhere,
Fox does not point to any methods other than the
Comparative Method that have offered promising
results (1995: 104–109, 190–194, 250–253, 261–270).
Reconstruction, Syntactic 397
-v- and provide evidence for the leveling that must
have taken place elsewhere.
The above survey has touched on some of the prob-
lems that attend morphological reconstruction. Read-
ers interested in a listing of the different available
procedures should consult the carefully laid-out
material in Koch (1996).
Bibliography
Fisiak J (ed.) (1980). Historical morphology. The Hague:
Mouton.
Hale K (1973). ‘Deep-surface canonical disparities in rela-
tion to analysis and change: an Australian example.’ In
Sebeok T A (ed.) Current trends in linguistics 11: Dia-
chronic, areal, and typological linguistics. The Hague:
Mouton. 401–458.
Hammond M & Noonan M (eds.) (1988). Theoretical
morphology. San Diego: Academic Press.
Harris A C (2003). ‘Preverbs and their origins in Georgian
and Udi.’ In Booij G & van Marle J (eds.) Yearbook of
morphology 2003. Dordrecht: Kluwer. 61–78.
Hock H H (1991). Principles of historical linguistics
(2nd edn.). Berlin: de Gruyter.
Joseph B D & Janda R D (2003). The handbook of historical
linguistics. Oxford: Blackwell.
Koch H (1996). ‘Reconstruction in morphology.’ In Durie
M & Ross M (eds.) The comparative method reviewed.
Oxford: Oxford University Press. 218–263.
Whitney W D (1889). Sanskrit grammar (2nd edn.).
Cambridge: Harvard University Press.
Efforts to reconstruct syntax can be traced at least
as far back as 1893–1900, when Delbrück (as cited in
Lehmann, 1992: 32) reconstructed OV word order
for Proto-Indo-European. By far the most ambitious
early effort at reconstructing syntax is Schleicher’s
‘Proto-Indo-European fable’, which was considered
a rash enterprise even in those pre-Neogrammarian
times (1868; as cited and translated in Jeffers and
Lehiste, 1979: 107):
Avis akvāsas ka
‘A sheep and horses’
avis, jasmin varnā na ā ast, dadarka akvams, tam, vāg-
ham garum vaghantam, tam, bhāram magham, tam,
manum āku bharantam.
avis akvabhjams ā vavakat: kard aghnutai mai vidanti
manum akvams agantam. Akvāsas ā vavakant: krudhi
avai, kard aghnutai vividvant-svas: manus patis varnām
avisāms karnauti svabhjam gharman vastram avibhjams
ka varnā na asti.
Tat kukruvants avis agram ā bhugat.
398 Reconstruction, Syntactic
‘A sheep on which there was no wool saw horses, one
drawing a heavy wagon, another a great burden, and
another carrying a human quickly.
The sheep said to the horses: The heart feels anguish
seeing the man driving the horses.
The horses said: Listen, sheep, the heart feels anguish
having seen the man, the master, make the wool of sheep
into warm clothes and there is no wool for the sheep.
Having heard that, the sheep turned away into the field.’
Even aside from the fatal flaws instantly evident to
any modern Indo-Europeanist (or, for that matter, to
any of the Neogrammarians of the 1870s) – most
notably the Sanskrit-like phonology, especially the
vowel system, which no one now believes to be
possible for Proto-Indo-European (PIE) – Schleicher’s
fable struck Indo-Europeanists as so unpromising
that hardly anybody since then has felt confident
enough to reconstruct even a short discourse for
Proto-Indo-European (in spite of a few proposed
‘updates’ of Schleicher’s fable, e.g., one by Hermann
Hirt in 1938, as cited in Jeffers and Lehiste, 1979:
108). The problem, after the phonology is cor-
rected, is that PIE syntax is too poorly understood
to justify constructing PIE sentences. But the issue of
whether syntax can be reconstructed at all recurs in
the literature of historical linguistics.
Two waves of interest in the topic, one of them
widely considered abortive and the other current and
growing, account for most of the research. The first
wave, which crested in the 1970s, was based on the
famous typological morphosyntactic generalizations
first presented in Greenberg, 1963. These generaliza-
tions focused in particular on clusters of morpheme
orderings that tended to appear together as a bundle –
for instance, OV word order with postpositions,
Adjective-Noun, Determiner-Noun, Relative-Noun,
and other word-order patterns. A language in which
all the statistically predominant ordering patterns ap-
pear together is said (in this view) to be typologically
consistent. The fundamental argument relevant to
typologically-based syntactic reconstruction is that
inconsistencies in ordering patterns, especially as
reflected in minor patterns in restricted environments,
reflect older consistent patterns, providing a means of
reconstructing the earlier morpheme-order patterns.
When evidence for certain patterns is lacking (for
instance because of the absence of texts in an attested
ancient language), the existing orderings, e.g., Noun-
Adjective, are used to infer other orderings, e.g., VO
word order. In other words, the typological clustering
of the various ordering patterns provides directionali-
ty of syntactic change and thus permits the reconstruc-
tion of earlier syntactic orders.
The most prominent proponent of this approach
to syntactic reconstruction is Winfred Lehmann (e.g.,
1973, 1974, and see also 1992: 238–240); Lehmann
views typologically ‘inconsistent’ languages – that
is, languages that violate Greenberg’s ordering gener-
alizations – as predictably transitional, en route to
consistency. The entire typological approach to syn-
tactic reconstruction is sharply criticized in Watkins,
1976, with a focus on Lehmann, 1974. Watkins
laments the fact that (as he sees it) ‘‘the rebirth of
Indo-European syntax has taken place in the bed
of Procrustes’’ (p. 305), that ‘‘[s]yntax is now viewed
as coterminous with word order’’ (ibid.), and that
the approach ‘‘elevates some of Greenberg’s extreme-
ly interesting quasi-universals to the dubious status
of an intellectual strait-jacket . . .’’ (p. 306). He also
presents arguments against specific predictions of the
typological approach, such as the fact that some ty-
pologically ‘inconsistent’ languages have apparently
remained ‘inconsistent’ for many centuries.
A partly comparable search for universal principles
that might lead to valid syntactic reconstruction is
Givón’s much-cited dictum, ‘‘Today’s morphology is
yesterday’s syntax’’ (1971: 413). Like Lehmann,
Givón claims that clues to earlier word-order patterns
are to be found in relics, specifically, for Givón, in
the order of morphemes in words – the crucial
assumption being that processes of cliticization and
then affixation of formerly independent words must
have fossilized the original relative ordering of the
words. This approach too has been sharply criticized
(see e.g., Comrie, 1980 and Harris and Campbell,
1995: 199ff.). One empirical problem for the claim
is that affixes sometimes change positions in the word
relative to other affixes and even relative to the stem,
and clitics sometimes change position when they turn
into affixes. A major reason for the widespread skep-
ticism about all approaches to syntactic reconstruc-
tion that focus on reconstructing word-order patterns
is hinted at in one of Watkins’ objections: not only is
there much more to syntax than word order, but it
is also impossible to know, from surface word order
alone, the nature of the grammar that produced that
word order. So, for instance, a dominant SVO sentential
word-order pattern in one language might result from a
grammar that specifies SVO word order, but in another
language it might result from a verb-second rule.
The search for universals that would permit the
establishment of a firm directionality in syntactic
change, and thus a sure means of syntactic recon-
struction, is continued in some of the research
on grammaticalization, and this research goes well
beyond word-order studies. Still, proposals of univer-
sality have not met with general acceptance. See, for
instance, Harris and Campbell (1995: 336–338) for a
criticism of claims that certain morphosyntactic
changes are predictably unidirectional. Many

linguists, including Harris and Campbell themselves,
would agree that a weaker sense of directionality
might prove useful in syntactic reconstruction, in
the same way that probabilities are used in phonolog-
ical reconstruction: all things being equal, one would
reconstruct *p rather than *b for an intervocalic p : b
correspondence, because intervocalic voicing is more
common, and more likely on phonetic grounds, than
intervocalic devoicing. Some (not all) of the proposed
universals of morphosyntactic change would meet
this ‘more probable’ criterion.
The recent upsurge of interest in syntactic
reconstruction is much less homogeneous than the
typological approach of the 1970s. Perhaps the most
cited recent contribution to the debate, both in its
own right and as a stimulus to further research
(much of which offers opposing views), is Harris
and Campbell’s Chapter 12, ‘Reconstruction of syn-
tax’ (1995: 344–376). Harris and Campbell’s ap-
proach is criticized in Lightfoot, 2002a, and the
discussion is continued in a follow-up exchange
(Campbell and Harris, 2002; Lightfoot, 2002b). The
two most difficult issues, for syntactic reconstruction
in general as well as in these articles, concern the
problem of identifying syntactic correspondences in
related languages and the problem of directionality.
Neither issue is new, but the former in particular has
received more attention recently than in (for instance)
the 1970s (but see Watkins, 1976).
The crux of the correspondence problem is articu-
lated by Werner Winter (1984: 622–623, and see
also the discussions in Harris and Campbell, 1995:
345–353 and Rankin, 2003: 201–206): ‘‘Sentences
are formed, not learned; morphemes and simple lex-
emes are learned, not formed.’’ The difficulty is that
application of the Comparative Method requires ele-
ments that show recurring correspondences in both
form and meaning, and in practice this has always
meant reconstructing proto-language features on the
basis of cognate morphemes – that is, morphemes in
two or more daughter languages that are descended
from the same morpheme in their shared parent lan-
guage. The method has been spectacularly successful
in the reconstruction of individual morphemes, both
words and affixes, and of phonological features
abstracted from the reconstructed morphemes; recon-
struction of partial morphological patterns has also
been carried out successfully, especially where ana-
logic changes that have transformed some daughter-
language paradigms are fairly transparent. But purely
syntactic structures – those which, unlike construc-
tions that are anchored by such morphemes as
the question particle used in the formation of yes/
no questions in some language families – offer no
cognate morphemes to base a reconstruction on.
Reconstruction, Syntactic 399
As a means of overcoming this difficulty, Harris
and Campbell propose the concept of syntactic corre-
spondence. Although they acknowledge that ‘‘syntax
has nothing quite like the duality of patterning (dou-
ble articulation) in phonology’’ (1995: 349), they
argue that sentences in related languages can be con-
sidered to correspond if they mean the same thing and
if the grammatical morphemes that are crucially rele-
vant to the structure point under consideration are
cognate (p. 349), and if they are ‘‘responses to identi-
cal or essentially identical stimuli in two stages of
a language’’ (p. 350). This last notion – that identical
stimuli can be predicted to produce linguistically
equivalent utterances – seems at best overoptimistic;
it is surely not supported by psycholinguistic evi-
dence. Harris and Campbell also emphasize that
their goal ‘‘is to establish correspondence patterns,
not corresponding sentences’’ (p. 350), so in fact
their proposal is that there are cognate sentence
patterns, not cognate sentences per se.
Crucially, they argue for the existence of regularity
in syntactic change, equivalent to the regularity
hypothesis for phonological change (pp. 326 ff.) –
but with a sense of regularity, namely ‘‘rule-governed
and non-random’’ (p. 326), that differs sharply from
the sense in which phonological change is hypoth-
esized to be regular. An analogic change that levels a
morphophonemic stem alternation in an inflectional
paradigm is regular in Harris and Campbell’s sense,
but it is not directly comparable to the concept of
regular sound change that has guided applications
of the Comparative Method for over a century: regu-
lar sound change is form-based only, while analogic
change is both form- and meaning-based (as seen,
for instance, in certain analogic leveling processes in
Serbo-Croatian that affect noun declension but
not adjective declension – Thomason, 1976). The
morphosyntactic changes discussed by Harris and
Campbell in this context are presumably driven by
analogy, not by sound change. Another major prob-
lem that arises for Harris and Campbell’s notion of
cognate sentence patterns is the same as a problem
noted above for the typological approach to syntactic
reconstruction: the same syntactic output in two dif-
ferent languages might well be produced by quite
different grammars.
A different line of research aimed at reconstructing
syntax focuses on the implications of changes in
parameter settings (e.g., Roberts, 1998; Longobardi,
2001; compare the critical discussion in Lightfoot,
2002a: 127–130). But until an approach based on
parameter settings is worked out in more detail, in-
cluding specific examples, the idea is difficult to assess.
Can any safe general conclusions be drawn about
the feasibility of syntactic reconstruction? Perhaps.
400 Reconstruction, Syntactic
There is fairly widespread agreement that syntactic
identity in all the daughter languages – for instance
SOV word order – permits the reconstruction of
that pattern in the parent language, and nobody (as
far as I know) has recommended that all efforts
to reconstruct syntax should be abandoned because
they are hopeless. On the contrary, a cautious opti-
mism is widespread and seems justified. Watkins, for
instance, argues for the possibility of reconstructing
poetic formulas of two and three words – and there-
fore bits of proto-language syntax – where the for-
mulas are cognate morpheme by morpheme in the
daughter languages (1995). Pires and Thomason
(forthcoming) argue that, at least at relatively shallow
time depths, the Comparative Method can sometimes
be used successfully to reconstruct morphosyntactic
constructions, for instance the Proto-Romance future
(using a deliberately restricted data set for testing
purposes); here the analysis rests on shared cog-
nate morphemes, on directional probabilities (clitic-
to-affix is more likely than affix-to-clitic), and on
drift (independent clitic-to-affix changes in several
languages). The prospects seem to be dimmer for
successful syntactic reconstruction where the con-
structions under comparison are not expressed at
least in part by cognate morphemes and where there
is no identity in all the daughter languages. Certainly
proposals that don’t meet at least one of these criteria
are all highly controversial at present.
See also: Comparative Constructions; Morphological
Change, Paradigm Leveling and Analogy; Syntactic Con-
structions.
Bibliography
Campbell L & Harris A C (2002). ‘Syntactic reconstruction
and demythologizing ‘‘Myths and the prehistory of gram-
mars’’.’ Journal of Linguistics 38, 599–618.
Comrie B (1980). ‘Morphology and word order reconstruc-
tion: problems and prospects.’ In Fisiak J (ed.) Historical
morphology. The Hague: Mouton. 83–96.
Delbrück B (1893–1900). Vergleichende Syntax der indo-
germanische Sprachen. Strassburg: Trübner.
Fox A (1995). Linguistic reconstruction: an introduction to
theory and method. Oxford: Oxford University Press.
Givón T (1971). ‘Historical syntax and synchronic mor-
phology: an archaeologist’s field trip.’ Chicago Linguistic
Society 7, 394–415.
Greenberg J H (1963). ‘Some universals of grammar with
particular reference to the order of meaningful elements.’
In Greenberg J H (ed.) Universals of language. Cam-
bridge, MA: MIT Press. 73–113.
Harris A C & Campbell L (1995). Historical syntax
in cross-linguistic perspective. Cambridge: Cambridge
University Press.
Jeffers R J (1976). ‘Syntactic change and syntactic recon-
struction.’ In Christie W (ed.) Current progress in his-
torical linguistics (ICHL 2). Amsterdam: North Holland.
1–16.
Jeffers R J & Lehiste I (1979). Principles and methods for
historical linguistics. Cambridge, MA: MIT Press.
Joseph B D & Janda R D (eds.) (2003). The handbook of
historical linguistics. Oxford: Blackwell.
Lehmann W P (1973). ‘A structural principle of language
and its implications.’ Language 49, 47–66.
Lehmann W P (1974). Proto-Indo-European syntax.
Austin: University of Texas Press.
Lehmann W P (1992). Historical linguistics (3rd edn.).
London: Routledge.
Lightfoot D W (2002a). ‘Myths and the prehistory of
grammar.’ Journal of Linguistics 38, 113–136.
Lightfoot D W (2002b). ‘More myths.’ Journal of Linguistics
38, 619–626.
Longobardi G (2001). ‘Formal syntax, diachronic minimal-
ism, and etymology: the history of French chez.’ Linguistic
Inquiry 32, 275–302.
Pires A & Thomason S G (Forthcoming). ‘How much
syntactic reconstruction is possible?’ In Ferraresi G &
Goldbach M (eds.) Principles of syntactic reconstruction.
Rankin R L (2003). ‘The comparative method.’ In Joseph &
Janda (eds.). 183–212.
Roberts I G (1998). ‘Review of Harris & Campbell.’
Romance Philology 51, 363–370.
Schleicher A (1868). ‘Eine fabel in indogermanischer
ursprache.’ Beiträge zur vergleichenden Sprachforschung
5, 206–208.
Thomason S G (1976). ‘What else happens to opaque
rules?’ Language 52, 370–381.
Watkins C (1976). ‘Towards Proto-Indo-European syntax:
problems and pseudo-problems.’ In Steever S B, Walker
C A & Mufwené S S (eds.) Papers from the Parasession
on Diachronic Syntax. Chicago: Chicago Linguistic
Society. 306–326.
Watkins C (1995). How to kill a dragon: aspects of Indo-
European poetics. Oxford: Oxford University Press.
Winter W (1984). ‘Reconstructional comparative linguis-
tics and the reconstruction of the syntax of undocument-
ed stages in the development of languages and language
families.’ In Fisiak J (ed.) Historical syntax. The Hague:
Mouton. 613–625.