I. What is a Ghost Segment? A. Segment-Zero Alternations II

Outline 

I.

What is a Ghost Segment? 

A. Segment-Zero Alternations 

II.

Historical Jers in Old Church Slavonic 

A. invention of ь and ъ in orthography, high lax vowels, front and back 

1. note that Lunt describes ь as being similar to /

​ɪ​/ and ъ as similar to /ʊ/ 

judging by the comparisons he makes to english words (pit and put),  

a. but ъ was only rounded in a few regions, and could also be described 

as lax /

​ɯ​/ 

b. BUT it was originally lowered to o so /u/? 

2. reduced vowels rather than full vowels 

B. Jer-Shift 

1. in some places “weak,” disappear 

a. where not followed by a syllable with another yer 

2. in other places “strong,” lower articulation 

a. where the next syllable contains a weak yer 

i.

the pattern of strong yers on alternating syllables is Havlík’s Law 

b. originally lowered to e and o (in orthography), became other things in 

other dialects 

C. How do we know this is happening? 

1. stem vъzьm- becomes vozьmi in imper. 2p.sg, but vъzemъ in a 

participle 

III.

Rule Based 

A. Lightner 1965, Russian 

1. Originally posits tense/lax distinction, which is not relevant otherwise in 

Polish, and lax vowels never surface as such 

B. Scatton 1975, Bulgarian, SPE 

1. all information on Scatton taken from Jetchev 1997 

2. posits underlying high lax vowels /ǐ/ and /ŭ/ 

a. /ŭ/ → [ə] and /ĭ/ → [e] 

b. absolute neutralization 

3. Rules DEL and LOW

 

a. or more simply

 

b. LOW comes after DEL 

4. so every ∅-inflected word must actually end in /ŭ/ 

C. Gussmann 1980, Generative, SPE, Polish 

1. note that I do not have a copy, and all information is pulled from other 

sources 

2. Gussmann argues for underlying segments like /i/ and /ɨ/ but [-tense] in 

underlying inventory which never surface (ĭ and  ̆)  

a. absolute neutralization 

3. “two kinds of /e/” i.e. that which alternates with zero and that which does 

not, these also alternate with /i/ and /ɨ/, within this there is an /e/ that 

triggers palatalization and one that does not for both the alternating and 

the nonalternating /e/’s 

4. establishes Lower as a rule  

a.   [+syll, 

{ [-high

    [+high 

+high,    →

   -back]

   /    _ C

​

0

​

  -tense] 

 -tense] 

{ ∅ 

b. so /ĭ/ and / ̆/ become /e/ and /ɘ/  

i.

note here that Gussmann writes /ɘ/ as Baby Gamma, the obsolete 

version of /

​ɤ​/, which would be correct if Gussmann was using the 

unrounded high back vowel /

​ɯ​/, as would follow from the 

historic yer, rather than the unrounded high mid vowel as his 

basis 

ii.

are /e/ and /ɘ/ not tense though 

5. also Derived Imperfect Yer Tensing 

a. [+syll, +high, -tense]  →  [+tense]  /  _  C

​

0

​

 aj]

​

DI 

D. Rubach 1984, Lexical and Cyclical, SPE, Polish 

1. assumes Gussmann’s rule of Lower and Yer Deletion, with Lower as Cyclic 

and Deletion as Post-Cyclic  

a. also Labio-velar /j/ Insertion as Cyclic accounting for yers that trigger 

palatalization, with palatalized nonalternating /e/’s being palatalized 

underlyingly 

∅   →   j    /   [-coronal] _ [+syll, -high, -back] 

2. Gussmann’s abstract high lax vowels represented by Rubach as //ĭ// and 

// ̆// 

3. modifies Lower so //ĭ// and // ̆// become /e/ and /ɘ/ (which Rubach 

writes /γ/, presumably because he, like I, cannot write Baby Gamma) and 

separates out Yer Deletion 

a. modified: [+syll, +high, -tense]  →  [-high]   /    _ C

​

0

​

  [+high, -tense]  

b. and:

[+syll, +high, -tense]    →

 ∅

 

c. then Vowel Spell-Out turns /ɘ/ to /e/ 

4. Havlík’s Law explained by Lower’s Cyclicity 

5. Zec 1988 runs parallel for Bulgarian 

IV.

Autosegmental 

A. Spencer 1986, Lexical and Autosegmental, with Rule-Based, Polish 

1. Non-linear Analysis: CV Tier/prosodic template 

a. rule of e-association, ignoring rightmost vowel slot 

i.

rightmost slot marked extrametrical i.e. bar on association 

➥

“historical metric behavior of [yers]” 

ii.

cyclic, as in Rubach 1984 

b. post-cyclic rule deletes rightmost vowel slot 

c. j underlying, deleted in post-cyclic process when interconsonant 

2. Gussmann’s DI Tensing as lexical redundancy rule 

a. pXv~pɨx]

​

DI

​

 and dVr~dzier]

​

DI

  

b. but following the theory of McCarthy 1981:

 

3. Underspecification and explaining where /e/ comes from 

a. using the Archangeli (1984) take on underspecification in epenthesis 

rather than underlying abstract vowels 

b. find maximally unspecified matrices where /e/ is underspecified 

compared to other vowels in inventory 

 

c. Redundancy Rules 

i.

Default Rules (according to Archangeli, part of UG) 

ii.

Complement Rules, of the form [ ] → [⍺Feature] fill in the 

unspecified features for the opposite value specified in matrix 

4. Morpholexical Rules in the nonlinear approach 

B. Kenstowitz and Rubach 1987, Lexical and Autosegmental, Slovak 

1. yers are segments unassociated on skeletal tier

 

2. justification through lengthening rules in Slovak 

3. instead of Yer Deletion, Stray Erasure as in Steriade 1982, deletes all 

unassociated segments 

4. “floating vocalic matrices” 

C. Bethin 1989, Autosegmental, Polish 

1. rejects Rubach 1986 because cannot predict phonetic quality of vocalized 

yers 

2. treat yers as floating feature [-cons], unassociated on skeletal tier 

a. don’t need to specify melody

 

3. like Rubach 1986, trying to account for syllabification, which they argue 

takes place on skeleton 

D. Szpyra 1992, Autosegmental, Polish 

1. Yers are invisible to syllabification and in fact block it 

a.  final consonant behaves as though extrametrical even though 

preceded by yer 

b. rejects Spencer 1986 due to inability to explain syllabification, but 

agrees with underspecification approach 

i.

mention of epenthetic /e/ not related to yers 

c. rejects Bethin 1989 and Rubach 1986 because missing timing slot would 

make yers completely invisible to syllabification and incapable of 

blocking it 

i.

consonants on either side would be considered adjacent, which 

they are not 

ii.

ALSO issue with floating melodies is that it assumes unrelation 

between skeletal and melody tiers, that they are underivable, 

unpredictable mismatches between them 

➥

but in languages where length doesn’t matter, skeletal tier 

is melody driven 

➥

if lexical representations have no redundant properties, no 

skeletal info at all in language, because timing tier can be 

derived from melody 

➥

Polish, being without true geminates, long vowels, no 

empty vowel slots (as we have rejected Spencer), then 

skeletal is derivable always, and floating melodies can 

project a skeletal unit 

➥

but if they could project timing slots, not invisible to 

syllabification 

d. but yers do not ALWAYS block syllabification, there are exceptions 

2. Feature Geometry 

a. assuming skeletal tier not primitive, derived from phonemic tier 

b. yers as empty root nodes, no melodic features 

i.

underlyingly neither vowels nor consonants 

➥

no [±cons], can’t participate in syllabification 

c. assuming also that unsyllabified units can’t be in a syllable, can only be 

extrametrical on the periphery 

i.

so empty root node blocks syllabification 

d. Vocalizing Yers

 

i.

rest of features filled in by universal/language specific rules, ala 

Spencer or rather Archangeli’s underspecification 

➥

WHAT ARE REDUNDANCY STATEMENTS I DON’T GET IT 

ii.

why do they turn into vowels? 

➥

syllabic wellformedness 

➥

argument that the syllable structure of Polish is CCVC 

3. Lower 

a. not necessary to explain: 

i.

occurence of palatal glide (syllable position dependant), so no 

deletion rule necessary 

ii.

blocking of nasalization (more evidence suggests nasal vowels are 

underlying) 

➥

see Iwan 2015 probably 

iii.

Nasal Backing and Raising (better to use syllable structure) 

b. we’re just adding yers to justify Lower 

c. new! improved! 

 

d. vocalization makes new syllable nuclei, unsyllabified consonants attach 

to them  

e. Yer Deletion/Stray Erasure 

i.

unsyllabified segments can’t be removed in certain languages, 

maybe including Polish 

ii.

therefore

 

f. not cyclic 

V.

Optimality Theory 

A. Jarosz 2005, Output-Output, Polish 

1. yers are /

​ɪ​/ and “back” /ɘ/, alternate with /ɛ/ rather than /e/ 

a. note that /ɛ/ is actually lax so why weren’t we using it the whole time 

HUH 

b. richness of the base means ALL languages undergo absolute 

neutralization so that’s not an issue 

2. *[

​⍺high​][​𝛽tense​], ​dep-v​ ≫ *​ComplexCoda​ ≫ ​Ident​[​high​] ≫ ​Max-V 

3. Contextual Correspondence

 

B. Gouskova 2012, Lexically Indexed Constraints, Russian 

1. exceptional morphemes rather than segments 

2. deletion of Mid vowels 

C. Rubach 2013, Lexically Indexed Constraints, Polish 

1. response to Gouskova 2012 

2. points out that all mid vowels in morpheme should exhibit yer behavior 

and delete, but don’t, and the cost of trying to mitigate this issue is very 

high 

VI.

Appendix: Actual Polish Vowel Inventory 

A.  

 

i 

ɨ 

ɛ 

a 

u 

ɔ 

round 

- 

- 

- 

- 

+ 

+ 

high 

+ 

+ 

- 

- 

+ 

- 

low 

- 

- 

- 

+ 

- 

- 

back 

- 

- 

- 

- 

+ 

+ 

front 

+ 

- 

+ 

+ 

- 

- 

tense 

+ 

+ 

- 

- 

+ 

-