Linguistics 051
Proto-Indo-European Language and Society
Rolf Noyer
• The comparative method involves demonstrating that two languages are
related through systematic sound correspondences.
• Forms which (on the basis of their meanings) are presumed to be descended
from a common form in an ancestral language are collected into
correspondence sets.
English German Spanish French Latin Greek Sanskrit Russian
father Vater padre père pāter patē′r pitár- (otéʦ)
foot Fuss pie pied ped- pod- pad- (nogá)
for für para pour per para pra pered
fish Fisch pescado poisson piscis (ikʰtʰu-) — (rýba)
salt Salz sal sel sal- hal-s sal-ilā solʲ
six sechs seis six sex hex ṣáṣ šestʲ
seven sieben siete sept septem heptá saptá semʲ
The above data can be used to establish the following correspondences (at
least for initial segments of the forms):
NE NHG Sp Fr L Gk Skt R
f f p p p p p p
s z s s s h s/ṣ s/š
Note that NHG initial S- is pronounced /z/.
Although we often represent linguistic expressions using conventionalized
spelling (orthography), this is a convenience and can be misleading,
because what is really being compared is the sounds and not the letters
used to represent them.
To circumvent this problem and also to be able to represent pronunciations
in languages with no standard orthography, phoneticians have developed
standard uses of letters and symbols to unambiguously represent
pronunciations.
This is known as phonetic (or phonological) transcription and employs a
speical set of symbols.
There are several variant transcriptional types in use. One common one is
that of the International Phonetics Association or IPA.
Indo-Europeanists often do not use the IPA, preferring to use either:
(a) standard orthographies of languages (this entails, though, that the reader
knows how to interpret the spelling systems of many languages), or
(b) transcriptional systems that predate the IPA and are still in use because
they are ‘traditional’ in their own way.
Often, because the languages in question are known only from written
records, the ways in which they conventionally written using Roman letters
merely substitutes a unique Roman letter for each symbol of the original
writing system.
Since this is not a direct representation of pronunciation (a transcription), it
is called a transliteration.
To understand sound changes you need to have some awareness of how
transliteration works and how it differs from transcription.
All of the following ancient languages are written in non-Roman alphabets
but are normally transliterated into a modified form of the Roman alphabet
(but not in IPA). In most cases the descendants of these languages continue
to use a non-Roman alphabet either identical to the one used in the ancient
language or a variant of it.
Sanskrit (written in devanagari)
Avestan (written in the Avestan alphabet)
Hittite and other Anatolian languages, and Old Persian (written using various
types of cuneiform writing)
Old Church Slavonic (written in cyrillic characters)
Armenian (written in the Armenian alphabet)
Runic, a dialect of early Germanic (written in runes)
Old Italic (Oscan, Umbian, Faliscan etc.) were written in a early variant of the
Roman alphabet
Greek (written in the Greek alphabet)
Recostructing the proto-language
If [f] in NE and NHG corresponds systematically to [p] in cognates in the
other languages, then we can consider three scenarios:
proto-form NE, NHG the rest
I *p f p
(implies *p > f in NE, NHG, so they are the innovating lgg.
/p/ then is an archaism while /f/ is an innovation)
II *f f p
(implies *f > p in the other languages, so they are innovating.
NE, NHG are archaic (wrt this property, at least), and the other lgg. are
innovating.
III *# (something else) f p
(implies that both /f/ and /p/ are innovations, derived from some third
and different proto-form)
Simplicity disfavors scenario III since it requires postulating two changes
instead of one.
Scenarios I and II are equal with respect to simplicity, provided certain
assumptions are made.
Scenario I: {all}:*p
{X}: f {all} – {Y}: p
If NE and NHG both descend from some common intermediate language ‘X’
then we need to postulate only one change: *p > *f (in X). The remaining
languages are archaic and don’t have any change.
• Could NE and NHG have individually changed *p > f without having
descended from a common ancestor?
In principle, yes, although this would be a more complicated hypothesis. But
such cases of accidentally parallel development are not impossible, and
when they go unrecognized they can create serious problems in
reconstruction.
Thus identical traits do not guarantee a common ancestor.
This is well-known in biology: the complex image-forming eye developed
independently in a number of clades (e.g. the octopus eye developed
independently of the vertebrate eye although they currently have very similar
appearances).
Scenario II: {all}:*f
{all}–{Y} f Y= {Sp, Fr, L, Skt, Gk, R}: p
Scenario II is equally simple, provided that all the language we have
examined besides NE and NHG descend from an ancestor they share but NE
and NHG do not have.
Reconstruction entails hypotheses about subgroupings based on shared
archaisms and shared innovations.
Ideally, a complete reconstruction of a proto-language will involve sound
changes that are consistent with a branching ‘tree’, where each branching is
defined by a set of sound changes.
In practice, though, language contact and borrowing often obscure this
‘regular’ tree to some extent.
Systematic sound correspondences hold between cognate words (or parts of
words).
Two forms in different languages A and B are cognate if they ‘descend’ from
the same form in a language ancestral to A and B.
• What does this really mean?
In applying the compartive method we must be careful in choosing the
things to be compared — the comparanda (singular = comparandum). The
comparanda need to be cognate, but cognates don’t always have exactly the
same meaning.
The reason for this is because meanings change over time: semantic drift.
Old English NE OE NE
cniht ‘boy’ > knight mōdig ‘courageous’ > moody
gār ‘spear’ > gore rēc ‘smoke’ > reek
brond ‘fire’ > brand sacu ‘conflict, strife’ > sake
burh ‘stronghold’ > borough scacan ‘to hasten’ > shake
cēap ‘purchase’ > cheap swāt ‘blood’ > sweat
drēam ‘joy, revelry’ > dream winnan ‘to strive’ > win
gilp ‘boast’ > yelp
lār ‘advice’ > lore
mǣl ‘time, occasion’ > meal
In order to determine what forms are comparanda one must study the
history of meanings.
Linguists are typically interested in finding the oldest attested meanings
available, since these will be:
a. more reliable as indicators of what forms can be cognate
b. better approximations of the meaning of the proto-form
• Knowledge of language consists of two parts:
a. a rule system expressing systematic facts and relationships.
This is also known as the grammar.
b. a list of arbitrary information that must be memorized,
commonly called the lexicon.
• Change in language can consist of change in the grammar or
change in the lexicon.
Parts of the grammar of a human language:
a. Phonology — permitted sounds and their distributions; ways
in which sounds change depending on their environment of
occurrence
b. Morphology — ways in which elements in the lexicon
(morphemes: roots and affixes) can combine to produce
words
c. Syntax — ways in which words can combine to form
sentences.
d. Semantics — ways in which the meaning of a sentence is
derived from the meanings of its parts (morphemes and
words)
e. Pragmatics — how expressions of language are used to
achieve various communicative functions
• Why do languages change?
• How does your language differ from that of your
parents?
• Why would you change your way of speaking?
• What happens when speakers of different language
come into contact with each other?
— what kind of contact?
— bilingualism
— pidgin, trade language, creole, lingua franca
— borrowing
• What aspects of language are typically borrowed?
Proto-Indo-European Society
• domesticated cattle, pigs, goats and sheep and so had an
advanced notion of property, particularly movable
property (*pek ̑u)
• hiearchical and patriarchal
• patrilocal extended family
• kings (L rēx), priests and poets distinct from common people
• king ruled over a ‘people’ (*teutā)
• warrior band (Männerbund or L comitātus) under a chief
self-designation *h₂ar(i)-ós Skt ā ́rya ARYAN, EIRE
OIr aire ‘freeman’
cf. H āra ‘what is fitting’,
Gk áristos ‘best’ ARISTOCRACY
(free) people *h₁leudʰ-os OE lēod , NHG Leute,
OCS ljudĭje ‘people’
< *h₁leudʰ- ‘to grow’
servants, slaves *h₂antbʰi-kʷol-os L anculus, Gk ampʰípolos
Skt abhi-cara-ḥ ‘servant’
*h₂upo-sth₂-i/o-s Skt úpasti-ḥ ‘servant’
Varieties of leaders
1. *h₃rē′g̑-s L rēx, Skt rāj-aḥ ‘king’
< *h₃reg̑- ‘stretch out the arm, direct’
hence ‘king’ < ‘director’
kings could have priestly functions
L rēx sacrōrum
2. *u̯nákt-s Gk ánaks ‘leader, lord’, Toch A nātäk
3. *u̯ik ̑-pot-s Skt viś-páti-s ‘master of the clan’
clan: *u̯eik -̑ /uoik -̑ o-s ECONOMY
4. *tag-ós Gk tāg-ós ‘leader’, Toch A tāśśi ‘leaders’
< *tag- ‘put in order, arrange’ TACTICS
Proto-Indo-European Sound System
1 2 3 4 5
p t k ̑ k kʷ voicless stop 1 labial
b d g̑ g gʷ voiced stop 2 dental
bʰ dʰ g̑ʰ gʰ gʰʷ voiced aspirated stop 3 palatal
s h₁ h₂ h₃ fricative 4 velar
m n nasal 5 labialized
l lateral velar
r rhotic
6 7
i u high vocoid 6 front
e o mid vowel 7 back
a low vowel {nasal, lateral, rhotic} = sonorant
• The ‘laryngeals’ are a category of segments whose precise
phonetic properties are quite uncertain
• The laryngeals were first reconstructed by Ferdinand de
Saussure, who called them ‘coéfficents sonantiques’ and wrote
them *ə₁ *ə₂ *ə₃
• In terms of their behavior in syllables and because of the
possible ‘gaps’ in the segment inventory, they seem to fit into
the system as fricatives.
• Moreover, when Hittite was deciphered in the 1930s some of
the hypothesized laryngeals appeared in Hittite words, spelled
with a symbol corresponding to a /x/ like sound in Semitic
languages
• This has suggested to some a phonetic realization of /ç x xʷ/;
others believe some may have been pharyngeals like /ʕ ħ/
• Some evidence suggests that *h₃ was voiced
• The laryngeals have 4 principal effects
a. Coloring: h₂ causes an adjacent /e/ to become /a/
h₃ causes and adjacent /e/ to become /o/
b. Lengthening: *VH > *Vː
c. Vocalization: If a laryngeal becomes syllabic it has a vowel
as a reflex in many languages.
Famously, only Greek distinguishes the
vocalized laryngeals: *h₁ > e, *h₂ > a, *h₃ > o
d. Tone: A laryngeal after a vowel can affect the tone
or ‘accent’ of words in Balto-Slavic languages
Syllables are subgroupings of segments; generally the groups that are
formed can be predicted based on the properties the segments have.
The key concept for syllable structure is sonority, an approximate measure
of the amount of energy required in the production of a type of segment.
In PIE the sonority scale is:
most sonorous least sonorous
non-high vowel > high vocoid > sonorant > fricative > stop
The Sonority Sequencing Principle requires that syllables contain a
single sonority peak — the ‘nucleus’, and that they do not fall in
sonority before the peak — the ‘onset’ and do not rise in sonority
after the peak — the ‘rhyme’.
An unusual property of PIE syllables is that all segments except stops
(and possibly /s/) can be syllable sonority peaks.
* r | sēn ‘male’ > Gk ársēn, Skt r̥sabhá-
*ph₂ | tēr ‘father’ > Gk patēr, L pater
*g̑e | nh₁ | tōr ‘father, creator’ > Gk genétōr, L genitor, Skt janitár-
*g̑l ̥h₃ | u̯ō s ‘husband’s sister’ > L glōs, Grk gálōs, Skt girí-