Historical Linguistics:
Questions of reconstruction
and relatedness
Ian Roberts
Downing College
[email protected]
The Indo-European family tree
The Indo-European Language Family
(More) Correspondences
 English: mouse, father, three, fish
 German: Maus, Vater, drei, Fisch
 Latin:
mūs, pater, trēs, piscis
 Kannada: ili, appa, muru, minu
The Comparative Method
 If a similarity between forms in two languages is
observed this can in principle be attributed to:
 Necessity (BUT: linguistic signs are arbitrary, and cf. Kannada etc
etc)
 Chance (this is always the most boring account of anything, but cf.
English dog, Mbambaram dog)
 Borrowing (e.g. Japanese kompyutaa)
 A historical connection: common origin:
 So we can conclude that English and German are quite closely related,
and that Latin is more distantly related to both, while Kannada is
unrelated to either.
Grimm’s Law (or the First Germanic
Consonant Shift) (oder die erste Lautverschiebung)
Verner’s Law (the second Germanic consonant
shift; die zweite Lautverschiebung)
 PIE:
 Gothic:
 German:
bhrātēr- pətērbrōþar
fadar
Bruder
Vater
“brother” “father”
 Why the different medial consonants?
Verner: voiceless intervocalic stops become voiced when the
preceding vowel is unaccented.
Sanskrit, Greek show “father” originally had an unstressed first
syllable.
 (cf. http://mr-verb.blogspot.com/2009/10/verners-law-
movie.html)
The Neogrammarian Thesis
 Sound laws are exceptionless!! (Osthoff &
Brugmann 1878)
 Ausnahmslosigkeit!
 Hence phonological reconstruction can be
relied on.
 But what about syntax?
What we know about PIE I: Phonology
Proto-Indo-European consonant segments
Velar
Labi
a
l
Coron
al
*m
*n
voiceless
*p
*t
*ḱ
*k
*kʷ
voiced
(*b)
*d
*ǵ
*g
*gʷ
aspirate
d
*bʰ
*dʰ
*ǵʰ
*gʰ
Nasal
Plosiv
e
Fricative
*s
Liquid
*r, *l
Semivowel
pala
pla labi
Laryngeal
t
i
a
a
n
l
l
*gʷ
ʰ
*h₁, *h₂,
*h₃
*y
*w
What we know about PIE II:
Morphology










Singular
Anim Neut
Nominative *-s, *-Ø*-m,
Accusative *-m *-m,*-Ø
Vocative
*-Ø *-m, *-Ø
Genitive
*-(o)s
Dative
*-(e)i
Instrumental *-(e)h₁
Ablative
*-(o)s
Locative
*-i, *-Ø
Dual
Anim
Neut Anim.
*-Ø *-h₁(e) *-ih₁ *-es
*-ih₁
*-ih₁ *-ns
*-h₁(e)
*-ih₁ *-es
*-h₁e
*-om
*-me
*-mus
*-bʰih₁
*-bʰi
*-ios
*-ios
*-h₁ou
*-su
Plural
Neut.
*-h₂, *-Ø
*-h₂, *-Ø
*-h₂, *-Ø
 (from Beekes (1995) Comparative Indo-European Linguistics:
An Introduction, John Benjamins).
Syntax?
 Lightfoot (1998:257): “the kind of reanalyses that occur in
catastrophic change constitute cutoff points to reconstruction”.
Proto-languages are no more amenable to reconstruction than
proto-weather.
 Harris & Campbell (1995:353): syntactic reconstruction may
be possible provided we can solve the correspondence
problem. In phonology, this problem is straightforward:
yesterday’s segments correspond in some fairly systematic way
to today’s (e.g. Gmc /f/ is the inherited reflex of PIE /p/). But
what was the Latin parent of L’état, c’est moi?
 Watkins (1976:306): “the confirmation by Hittite of virtually
every assertion about Indo-European word order patterns made
by Berthold Delbrück .. [is] .. as dramatic as the surfacing of the
laryngeals in that language”.
The IE clause
 Hale’s (1995) structure for the Vedic Sanskrit clause:
 [ Topic [CP C [ Focus IP ]]]
 A tendency for the verb to be second
 (cf. also Garrett (1990) on Anatolian, Kiparsky (1995)
on the prehistory of Germanic, Newton (2006) on
Celtic, and Fortson (2004) for a summary)
Older Germanic
 Fuß (2008) on Old High German:
[ Topic [ wh-phrase [ V IP ]]]
 Roberts (1996) on Old English:
[ Topic [ Focus (Verb) [ weak pronoun ..
Latin
 a.


Si bovem ..
serpens momorderit.
If cow-Acc.sg. snake-Nom.sg. has-bitten
“If a snake has bitten a cow”
(DS 116; Cato De Agri Cultura 102.1)
 b.

in adulterio uxorem tuam
si prehendisses
in adultery wife-Acc. your-Acc if you-havecaught

“If you have caught your wife in adultery”
(DS 119; Cato Orat 222.1)

 “probably represents a more archaic typology”
 (Devine & Stephens (2006) Latin Word Order OUP).
PIE Syntax
 Null subjects
(like Modern Italian, Greek)
 SOV word order
 Wh-movement
(like Modern Indic)
(like Modern English)
 Productive topicalisation/focalisation to the
“left periphery” (like Modern Slavonic)
 Second-position effects (pronouns, adverbs,
verbs) (like Modern Germanic, South Slavonic)
Nostratic Syntax
 Definitely head-final (Dolgopolsky, Bomhard
2008).
 Uh …
A related question: quantifying
distances between languages



while languages differ from one another in all aspects
of their structure, some pairs of languages differ from
each other more than others do: Spanish and
Portuguese are very similar to each other indeed,
English is quite similar to German and Japanese is
significantly unlike almost all other languages.
many, if not all, of these degrees of structural and
lexical difference can be correlated to historical
relationships. The central activity of historical linguists
for two centuries has been the establishment and
organisation of these relationships.
Recent developments are changing this picture though
…
Indo-European cladistics
Ringe et al (2000) used techniques from
evolutionary biology to try to identify the firstorder
subgrouping
of
Indo-European
languages.
More recent work on the same idea by Nakleh
et al (2005), Warnow et al (2005) and
http://www.cs.rice.edu/~nakhleh/CPHL
Character
"an identifiable point of grammar or lexical
meaning which evolves formally over the
course of the language family's development,
.. each state of the character ought to
represent an identifiable unique historical
stage of development - a true homology
[shared trait inherited from a common
ancestor]" (71).
Examples of characters
 a lexical character: Eng hand (=1), Ger Hand
(=1), Fr main (=2), It mano (=2), Rus ruká (=3).
  a phonological character: sequence of changes
Grimm's Law, Verner's Law, initial-syllable stress,
merger of unstressed *e with *i except before *r.
Absent = 1, present = 2 (singles out Germanic).
  a morphological character: most archaic
superlative suffix 1. *-isto-, 2. *-ismo-, 3 etc. absent (2
singles out Italo-Celtic).
Database
 
24 languages representing all 10 IE subgroups,
and 322 characters (22 phonological, 15
morphological and the rest lexical).
 
Result of running tree-optimisation software: 18
characters were incompatible with the best tree, "in
computational terms our result is a total failure” (86).
 
Italo-Celtic, Balto-Slavonic, the satem group
and Graeco-Armenian emerge as IE subgroups.
Possible Indo-European tree
(Ringe, Warnow and Taylor 2000)
Controversies for Indo-European
history
 Subgrouping: Other than the 10 major
subgroups, what is likely to be true? In
particular, what about





Indo-Hittite
Italo-Celtic,
Greco-Armenian,
Anatolian + Tocharian,
Satem Core?
Our best PPN (Language, 2005)
Modularised Global Parametrisation
 Developed by Chiara Gianollo, Christina
Guardiano and Giuseppe Longobardi, U of
Trieste
 Uses Universal Grammar syntactic
parameters to measure distances among
languages
Universal Grammar (UG)
 the set of grammatical principles which makes human
language possible (and defines a possible human
language)
 determined by the human genome
 physically exists (in res extensa) in the brain
 otherwise known as the “language faculty”, the
language acquisition device (LAD), the initial state of
language acquisition and the language bioprogram.
Some assumptions:
a. There exists a rich, innate language faculty (UG)
which is a species characteristic.
b. There are no racial or cultural biases towards
particular languages or language types.
c. There is clear evidence that a sentence which is wellformed in one language L may be ill-formed in some
other language L’:
i.
ii.
*John Mary hit.
John-ga Mary-o butta.
UG contains:
 invariant principles
 associated parameters of variation
 OV (e.g. Japanese, German) vs. VO
(e.g. English, Italian).
 UG principles define V, O and how they
go together (VP); a parameter
determines their order.
Parameters tell us what is variant (and by
implication what is invariant) in grammars, and
as such they 


predict the dimensions of language typology
predict aspects of language acquisition
predict what can change in the diachronic
dimension.
“A particular language L is an instantiation of the initial state of the
cognitive system of the language faculty with options specified”
(Chomsky (1995:219)).
Things that can vary inside a simple nominal expression (a
DP):

is number marked?
(English: YES; Japanese: NO)

is there a system of articles?
(English: YES; Japanese: NO)

is there a system of classifiers?
(English: NO; Japanese: YES)
The expression of possession:
 Type A: Possessor > Possessee
John’s sister
John-no imooto-ga
 Type B: Possessee > Possessor
la soeur de Jean (French)
chwaer Siôn (Welsh)
Method
 Parameters limited to the extended nominal
phrase (DP)
 27 languages from 4 families (IE, Semitic,
Uralic and Niger-Kordofanian)
 57 binary parameters
TABLE A
Optimisation by Kitch
Optimisation by UPGMA57
An approach to measuring relatedness which relies on
parametric syntax has certain advantages over an approach
based on lexical similarities:
o discreteness: the values of a parameter do not form
a continuum or cline of any kind
o binarity: a maximally simple range of possibilities
o finiteness: the number of parameters is finite, and in
fact rather small, usually thought to be more than 20
but less than 100
o no uncertainty of comparanda: we are in principle
always sure when we are comparing like with like
(Guardiano & Longobardi (2003:4))
A possible synthesis
 treat parameter values as characters,
thereby adding syntax to the cladistic
comparison.
 The parametric grid can be taken to indicate
the syntactic characters.
Backmutation (or homeoplasy)
"either improbable or vanishly rare" (70), i.e. "we simply
do not find cases in which the contrast between two
elements A and B in a structured system is
eliminated from the language, then .. reintroduced in
precisely the same distribution that it originally
exhibited" (ibid). Clearly true of phonemic
split/merger, loss/gain of inflection, changes in wordmeaning, etc. But is it true of parameters?
 The case of French and the null-subject parameter.
Homoplasy-free evolution
 When a character changes
state, it changes to a new
state not in the tree
 In other words, there is no
homoplasy (character
reversal or parallel evolution)
 First inferred for weird
innovations in phonological
characters and
morphological characters in
the 19th century.
0
0
1
0
0
0
0
1
1
Two issues
 Parallel
development
(analogy
rather
than
homology): such cases must simply be set aside.
Alternatively one can take sets of changes rather
than individual changes as evidence for clades. In
any case, it seems doubtful that syntactic change
poses any problems not already encountered in the
area of phonology.
 much less is known about the syntax of a number of
older IE languages compared to their phonology,
lexicon and morphology (see above).
Conclusion
 Syntax has played a relatively minor role in
establishing relations among languages, but
this can change.
 Parametric
comparison can quantify
grammatical differences and thus play a
major role in developing our theories of
typology, acquisition and change.
 Cladistic methods combined with parametric
comparison may shed light on major
questions in historical linguistics.
THANK YOU!!
Descargar

Historical Linguistics: Questions of reconstruction and