It's Complicated: The Relationship between Syllabification

Preview:

Citation preview

It’s Complicated:The Relationship between Syllabification &

Morphology in PIE

Andrew Miles ByrdUniversity of Kentuckyandrewbyrd@uky.edu

October 27th, 2012

General Overview

œ Ellen Kaisse : “The relation between word formation andphonology is complex.” (2005:45)

œ Dieter Gunkel: analysis of innovative Greek formations in-ma(t)-, many of which started to be formed to the weak stemallomorph as early as Pindar (2011).

œ Motivated by Trochaic Shortening (HL ! LL), though only ifa weak root allomorph existed elsewhere in the root’sparadigm (cf. kequmai ‘I poured’).qeuma ‘stuff poured’ ! qumageuma ‘meal’ 9 g‘ma

œ Such examples of morphology influencing a phonologicalprocess abound throughout the world’s languages, bothIndo-European and non.

Morphology & Syllabification

œ Today we’ll be investigating a problem of sonorantsyllabification in Proto-Indo-European (PIE), whose“exceptions” I also believe to be also driven by morphology.

œ First things first : what exactly do we know about PIEsyllabification?

History of the Problem : Meillet

œ First significant steps in the understanding of sonorantsyllabification: Meillet 1934.

œ Four easy principles that describe the distribution of sonorantsin PIE.

First Principle

œ “If a sequence of two sonorants follows a vowel or is in the firstsyllable of a word : the first is consonantal, the secondsyllabic.”

VRR ! VRR˚

#RR ! #RR˚

Examples

œ PIE *sru“-tó-s ‘flowed (masc.nom.sg.)’ ! *srutós

> Skt. *srutáh. , Gk. rhutós

Second Principle

œ “If a sequence of two sonorants follows a single consonant andprecedes a vowel : the first is syllabic, the secondconsonantal.”

CRRV ! CRR˚V

Examples

œ PIE *“ku“nés ‘dog (gen.sg.)’ ! *“kunés

⇢ Skt. śúnah. , Gk. kunós

Third Principle

œ “If a sequence of two sonorants follows a vowel and precedeseither a consonant or the end of the word: the first isconsonantal, the second syllabic."

VRRC ! VRR˚C

VRR# ! VRR˚#

Examples

œ PIE *néu“n ‘nine’ ! *néu

“n˚

> Skt. náva, Lat. novem, Gk. enné(w)a;œ PIE *neu

“nti“- ‘90’ ! *neu

“n˚

ti-> Skt. navatí-, Av. navaiti-

Fourth Principle

œ “If a sequence of two sonorants stands between two vowels :the first forms the second half of a diphthong, the second isconsonantal.”

VRRV ! VRRV

Examples

œ PIE *oi“u“os ‘one’ ! *oi

“u“os

> Av. aiva, Cyp. oiwos ‘only’

History of the Problem : Schindler

œ Schindler (1977:56) recognized a broader pattern withinMeillet’s rules of PIE syllabification and very elegantlycollapsed all four observations into one phonological rule.

Schindler’s Rule of Syllabification

œ When given two adjacent sonorants, the rightmost one alwaysbecame the syllable nucleus, if that sonorant was not adjacentto a ‘true’ vowel (*e, *a, *o, etc.).

∑+sonorant°syllabic

∏! [+syllabic] /

Ω°syllabic ,

#

æ Ω°syllabic ,

#

æ

(iterative from right to left)

History of the Problem : Schindler

œ Since publication, Schindler’s ‘right-to-left’ formulation hasbeen widely accepted by nearly all scholars to date and hasbecome the standard view of PIE syllabification.

œ But: five instances where Schindler’s rule does not correctlypredict the syllabification reconstructable for PIE!

Schindler’s Exceptions

(1) Roots of the shape *#R“R“-.

a. *u“i“eth2-, not *ui

“eth2-

> Ved. vyathate ’rolls’b. *u

“reg-, not *ureg-

> Ved. vrájant- ‘going’c. *mneh2-, not *m

˚neh2-

> Gk. Ímnhsa ‘remembered’

Schindler’s Exceptions

(2) Word-Medial *æ[mn- ! *m- or *n- (Asno Gesetz).

a. *“ghei“mno- ! *[“ghei

“]æ[mno]æ ! *[“ghei

“]æ[mo]æ

> Lith. žiemìnis ‘winter’

b. *h2a“kmnés ! *[h2a“k]æ[mnés]æ ! *[h2a“k]æ[nés]æ> Skt. aśnah. ‘stone (gen.sg.)’

Expected: **[“ghe][i“m˚

]æ[no]æ, **[h2a][“km˚

]æ[nés]æ

Schindler’s Exceptions

(3) /CR1R2V/ ! [CR1R˚

2]æ[V]æ,

iff /CR1R2C/ (! [CR1R˚

2]æ[C-]æ)

a. *[tri][(i“)om] ‘three (gen.)’; cf. * [tri][bhis]

b. Skt. -[vyas], -[vi][yas]; cf. -[vı]

Expected: **[tr˚]æ[i

“om]æ, **[u]æ[yas]æ

Schindler’s Exceptions

(4) Accusatives in *-im, *-um(s), *-r˚

m(s).

a. PIE *ménti“m ! *[mén]æ[tim]æ ‘mind (acc.sg.)’

b. PIE *suhx

num(s) ! *[suhx

]æ[num(s)]æ ‘son(s) (acc.)’

c. PIE *p@h2tr˚

ms ! *[p@h2]æ[tr˚ms]æ ‘fathers (acc.)’

Expected: **ménti“m˚

, **s(e)uhx

nu“m˚

(s), **p(@h2)trm˚

s

Schindler’s Exceptions

(5) Nasal-infixed presents.

a. PIE *i“u“ngénti

“!*[i

“un]æ[gén]æ[ti]æ

> Lat. iungunt ‘they join’b. PIE *bhi

“ndénti

“! [bhin]æ[dén]æ[ti]æ

> Lat. findunt ‘they split’

Expected: **[i]æ[u“n˚]æ[gén]æ[ti]æ, *[bhi

“n˚]æ[dén]æ[ti]æ

First Fix : Kobayashi 2004.

œ In his 2004 book, Kobayashi was able to account for nearly allof the exceptions in (1), (2) and (3) by assuming the followingtypologically-common principle of syllabification:

Onset MaximizationSyllabify as many consonants as possible within the onset.

Onset Maximization

œ Thus:

a. PIE *[u“i“et]æ[h2- : PIE *-[u

“i“eh2s]æ

b. PIE *[mneh2-]æ : PIE [*h2a“k]æ[mnés]æ

œ If a consonant cluster could be syllabified as an onset, it was.

*u“i“- and *mn- were licit onsets in PIE.

Second Fix : Keydana 2008

œ In 2008 Keydana suggested handling the exceptions in (4) –exceptions of the type *[mén]æ[tim]æ – in a very innovativefashion.

œ He assumed that there was a general tendency within PIE forcoronal codas to be avoided in PIE, formalized through thereconstruction of a constraint *R/C ‘Coronal sonorants areblocked in coda position’.

œ Through the constraint ranking *R/C ¿ *ComplexOnset, weget:

*[“ku“n˚]æ[bhis]æ

*[mén]æ[tim]æ

Second Fix : Keydana 2008

œ But what about...

a. *[sru]æ[tós]æ ‘flowed’b. *[“kli][æ[tós]æ ‘leaned’c. *[“klu]æ[tós]æ ‘famous’

œ If simple codas were preferred over complex onsets, we’dexpect:

a. *[sr˚u“]æ[tós]æ ‘flowed’

b. *[“kli“][æ[tós]æ ‘leaned’

c. *[“kl˚u“]æ[tós]æ ‘famous’

œ Also, assumption = ad hoc!

Third Fix : Cooper 2011:199ff.

œ To address exceptions of the type *[mén]æ[tim]æ *-m, Coopervery cogently argued that *m was different from all of theother sonorants.

œ New Sonority Hierarchy:

a, e, o ¿ i“, u

“, r, l, n ¿ m ¿ obstruents

œ *m : *menti“m :: *d : *i

“d ‘it’

To Review

œ Exceptions 1 - 4 have good explanations.

œ Despite the efforts of a number of scholars, the nasal-infixpresent is handled poorly by everyone’s rules of syllabification.

œ Question for today:

Why is */i“u“ngénti

“/ syllabified as *[i

“un]æ[gén]æ[ti]æ?

Current Explanation : Syllabification as Analogical

œ Fortson (2010:71) : “(M)orphological analogy surely played arole, since the root that this form comes from is *i

“eug- or

*i“ug-, and *i

“ung- is closer to the root in shape and sound than

*iu“n˚

g-.”

œ But why were alternating syllabifications of the type R˚

ª Rtolerated everywhere else in the proto-language?

*“ku“on- ª *“ku

“n˚

- ª *“kun- ‘dog’

œ Hayes (2009:251): “In most languages, syllabification ispredictable: starting out from the string of segments, one canpredict the syllabification.” ’

Syllabification must be derived through phonological rules.

Hypothesis 1 : Syllabification as Underlying

œ Hayes : “In most languages..."?

œ Can syllabification be underlying within the forms themselves?

œ There is at least one reconstructable instance of lexicalizedsyllabification in PIE: *kur-, not expected **ku

“r

˚- (Forssman

1980)

a. Hitt. kurkaš ‘foal’b. Pahl. /kurrag/, Fars. kurra ‘foal’c. Gk. kurnoc ‘bastard’

Hypothesis 1 : Syllabification as Underlying

œ Barra Gaelic (Borgstrøm 1937)

ar.an ‘bread’ s aL.@g ‘spectacle’a.ram ‘army’ s a.Lak ‘hunt’

œ Bosch & de Jong 1997, Green 1997 : these are differences instress, not differences in syllabification!

PIE *kur- must be explained otherwise!

Hypothesis 2 : the Prosodic Word

œ English -mpl- cluster:

a. camp leader [khæmp][li:][dr"]

b. employ [Em][phloI]

œ Word-initial sub-:

a. sublunar [s@b][lu][nr"]

b. subliminal [s@][blI][mI][nl"].

Hypothesis 2 : the Prosodic Word

œ These differences in syllabification are due to the prosodicword (pword).

œ Pword (!):the prosodic category universally accepted to be the domain ofsyllabification (Booij 1995)

œ The prefix sub- is typically a distinct pword: [s@b][lu][nr"]

non-cohering affix

œ In subliminal it is not: [s@][blI][mI][nl"]

(cohering affix)

Hypothesis 2 : the Prosodic Word

œ Were one to analyze certain affixes in PIE as beingnon-cohering, they would syllabified independently.

œ Perhaps our exceptions could be explained accordingly?

œ So, for example, perhaps the previous cited exceptions mentimand bhindénti were actually parsed as two separate pwords inPIE:

a. (men.ti)! + (m)!

b. (bhi.den.ti)! + (n)!

Two Problems

1. Invariably evolves into a circular argument:

This suffix is non-cohering because it syllabifies strangely, andvice versa.

2. More seriously: under this analysis we actually expect:

a. **[mén]æ[ti]æ[m˚

b. **[bhi]æ[n˚]æ[dén]æ[ti]æ

Hypothesis 3 : Let’s Generativize!

What Would English Do?

A Curious Pair from English

œ lantern & apron

apron /"eIprAn/ ! ["eI]æ[p(h)rn"]æ

lantern /"læntErn/ ! ["læn]æ[t(h)r"n]æ

œ If syllabification is an automatic process and affects allsequences equally, how can the sequence [Crn] be syllabified as[Crn

"] in one instance but [Cr

"n] in the other?

A Curious Pair from English

œ Answer: the position of the underlying vowel of the secondsyllable (/prAn/ vs. /tErn/)!

œ This makes good diachronic sense, as syllabic sonorants usuallyderive from an intermediate stage of reduced vowel + R:

["eIp(h)r@n] > ["eIp(h)rn"]

["lænt(h)@rn] > ["lænt(h)r"n]

Back to PIE

œ Ablaut: a form with an underlying full vowel was reduced tozero when unaccented.

œ Like English, if that vowel was immediately followed by asonorant, that sonorant was subsequently made the nucleus ofthe syllable.

œ Thus,

*/h1éi“men/ ‘way’ ! [h1éi

“]æ[mn

˚]æ, just like apron

*/méntei“m/ ‘mind (acc.)’ ! *[mén]æ[tim]æ, just like lantern

How do we know?

œ How exactly do we know that these are truly the underlyingforms?

œ These are proterokinetic nouns!

*/h1ei“ménes/ ’way (gen.)’ ! *[h1i]æ[méns]æ

*/mentéi“es / ’mind (gen.)’ ! *[mn

˚]æ[téi

“s]æ

The crux of the matter

œ In short, it appears that these unexpected syllabifications – aswell as expected ones – are to be viewed as the direct result ofa preference for nucleification of a segment that was moraicearlier in the derivation before processes of syncope took place.

œ This preference may be formalized for Optimality Theorythrough the constraint Max-µ (’Do not delete a mora (µ).’)

A Principled Approach

Indo-European syllabification should now be re-envisioned asobeying three straightforward and typologically grounded principles.

Principle 1

Syllabification is persistent throughout the phonological derivation.It occurs pre-syncope, as well as post-syncope.

Persistant Syllabification

Underlying Form: */h1éi“men/ */méntei

“m/

Syllabification : *[h1éµi“µ][meµnµ] *[méµnµ][tei

“µmµ]

Syncope: *[h1éµi“µ][mnµ] *[méµnµ][ti

“µmµ]

Resyllabification : *[h1éµi“µ][mn

˚µ] *[méµnµ][tiµmµ]

Surface Form: *[h1éi“]æ[mn

˚]æ *[mén]æ[tim]æ

Note that mora assignment is persistent, too.

Persistant Syllabification

Underlying Form: */h1éi“men/ */méntei

“m/

Syllabification : *[h1éµi“µ][meµnµ] *[méµnµ][tei

“µmµ]

Syncope: *[h1éµi“µ][mnµ] *[méµnµ][ti

“µmµ]

Resyllabification : *[h1éµi“µ][mn

˚µ] *[méµnµ][tiµmµ]

Surface Form: *[h1éi“]æ[mn

˚]æ *[mén]æ[tim]æ

Note that mora assignment is persistent, too.

Principle 2

Sonorants that were moraic earlier in the derivation, remain as suchpost-syncope.

Moraic Sonorants

Underlying Form: */sreu“tós/ */“ku

“onbhi

“s/

Syllabification : *[sreµu“µ][tóµsµ] *[“ku

“oµnµ][bhiµsµ]

Syncope: *[sru“µ][tóµsµ] *[“ku

“nµ][bhiµsµ]

Resyllabification : *[sruµ][tóµsµ] *[“ku“n˚µ][bhiµsµ]

Surface Form: *[sru]æ[tós]æ *[“ku“n˚]æ[bhis]æ

Further Evidence: Compensatory Lengthening

Principle 3

Non-moraic sonorants occupy the syllable nucleus if and only ifmore important constraints are violated.

1. Onset: Syllables must have onsets.

2. MaxST: No syllable may violate the Maximum SyllableTemplate.

Obeying Onset

œ PIE */“ku“onés/ ‘dog (gen.sg.)’ ! *[“ku][nés]

6! *[“ku“n˚][és]

œ PIE */kwetu“ores/ ‘four (nom.pl.)’ ! *[kwe][tu][res]

6! *[kwe][tu“r˚][es]

Obeying the Maximum Syllable Template (MaxST)

œ PIE */“ku“onés/ ‘dog (gen.sg.)’ ! *[“ku][nés]

6! *[“ku“nés]

œ PIE */kwetu“ores/ ‘four (nom.pl.)’ ! *[kwe][tu][res]

6! *[kwe][tu“res]æ

æ[“ku“n- and æ[tu

“r- are not licit onsets in PIE.

Parallels

It’s my prerogative!

Its maI pr"rAg@tIv

Returning to the Final Exception

Underlying Form: */bhei“nedénti

“/ ’they split’

Syllabification: *[bheµi“µ][neµ][déµnµ][tiµ]

Syncope: *[bhi“µ][n][déµnµ][tiµ]

Resyllabification: *[bhiµnµ][déµnµ][tiµ]

Surface Form: *[bhin]æ[dén]æ[ti]æ

Schindler

œ So what does this mean for Schindler’s theory of right-to-leftsyllabification, and every subsequent theory of IE syllabificationthat is based off it?

œ Each of these theories reflects an epiphenomenon, a patterninferred by linguists from the surface reconstructions, and*not* the actual underlying forms.

œ But there is a straightforward reason why Schindler’s theoryworks so well...

Why does Schindler’s rule seem so right?

œ In a survey of athematic suffixes containing two sonorants,each is of the shape *-RVR :

a. *-u“er

b. *-u“en

c. *-ter

d. *-men

e. *-sel

Why does Schindler’s rule seem so right?

œ In a survey of PIE verbal roots containing two sonorants, allbut one is of the shape *(C)RVR(C)- :

a. *h2u“ers- ‘rain’

b. *dhreu“gh- ‘deceive’

c. *tu“er“k- ‘cut’

d. *u“el- ‘roll’

Why did Schindler’s rule seem so right?

œ This entails that in the zero-grade of both types, *-RR and*(C)RR(C)-, respectively, the rightmost sonorant will alwayssyllabify, given that it had been moraic before syncope tookplace

Hence *-RR˚

and *(C)RR˚(C)-!

A New Approach

Note that the additional machinery, which Keydana & Cooper bothassume to explain Schindler’s 4th exception, is rendered unnecessaryby our current analysis of Indo-European syllabification.

œ *-m may or may not be less sonorous than the othersonorants.

œ But “exceptions” like mentim no longer need to explained assuch.

A New Approach

œ To conclude, our new theory makes three well-foundedassumptions:

a. Syllabification is persistant.b. The underlying form is changed via syncope, whose

derived form continues the original moraicity ofsonorants.

c. Syllabification respects phonotactic constraints.

A New Approach

œ This new theory of syllabification is better for two simplereasons:

a. Schindler’s exceptions almost completely disappear.

b. Even better : the regular outcomes and the exceptionsare driven by the same principles.

A New Approach

œ For those paying close attention – what to do with*[tri]æ[om]æ?

œ Perhaps in this isolated instance we may explain theanomalous syllabification by invoking a pword boundary.

Dutch: [ro:d]![Axt@x]! ! [ro:t]æ[Ax]æ[t@x]æ ‘reddish’

English: [2n]![Eksklusıv]! ! [2n]æ[Ek(s)]æ["sklu]æ[sıv]æPIE: [trei

“]![´om]! ! [tri]![´om]!

œ The genitive plural suffix -´om would thus be a non-coheringaffix.

Thanks!

ReferencesBooij, Geert. 1999.The Phonology of Dutch.Oxford University Press: Oxford & New York.

Bosch, Anna and Kenneth de Jong. 1997.The Prosody of Barra Gaelic Epenthetic Vowels.Studies in the Linguistic Sciences 27(1).1–15.

Cooper, Adam. 2011.Syllable Nucleus and Margin in Greek, Vedic, and Proto-

Indo-European.PhD dissertation, Cornell University.

Forssman, Bernhard. 1980.Hethitisch kurka- Comm. ‘fohlen’.Zeitschrift für vergleichende Sprachforschung 94(1./2.).70–74.

Fortson, Benjamin. 2010.Indo-European Language and Culture: An Introduction (2nd ed.).Chichester, U.K. & Malden, Mass.: Wiley-Blackwell.

Green, Antony Dubach. 1997.The Prosodic Structure of Irish, Scots Gaelic, and Manx.PhD dissertation, Cornell University.

Gunkel, Dieter. 2011.The emergence of foot structure as a factor in the formation of greek

verbal nouns in -ma(t)- .MSS 65.77–103.

Kaisse, Ellen M.. 2005.Word-formation and phonology.In P. Štekauer and R. Lieber (eds.), Handbook of Word-Formation,

25–47. Dordrecht: Springer.

Keydana, Götz. 2008 [2010] .Hohe Sonorität in der Koda: Indo-Germanische Lösungen für ein

phonetisches Problem.Historische Sprachforschung 121.54–64.

Kobayashi, Masato. 2004.Historical Phonology of Old Indo-Aryan Consonants.Tokyo: Fujiwara.

Meillet, Antoine. 1934.Introduction á l’étude comparative des langues indo-européennes (7th

ed.).Paris: Hachette.

Schindler, Jochem. 1977.Notizen zum Sieversschen Gesetz.Die Sprache 23.56–65.

Recommended