PROSODIC PATTERNS AND THE MINIMAL WORD IN THE DOMAIN OF G REEK
TRUNCATED NICKNAMEi
Nina Topintzi
UCL
.
:(
), (
), ( ),
,
,
.
,
.
.
Keywords
truncation, prosodic patterns, hierarchical alignment, minimal word
1.
Introduction
Tru ncated item s in Greek m ostly have the shape of a bisyllabic trochee (Malikou ti-Drachman
1999), bu t an interesting subset d oes not conform to this pattern. The aim of the cu rrent stud y is
to provide a unifying analysis for all the attested patterns and derive variation through minimal
changes in the ranking. The general proposal w ill u se the notion of hierarchical alignm ent (HIER
AL)
and the proper alignm ent of feet to the right ed ge of the prosod ic w ord (ALL-Ft-R) to
generate the patterns. Other prosod ic constraints, requirem ents of m aterial preservation and
lexical stress yield the final output.
The stru ctu re of the paper is as follow s. Section 2 presents the d ata. Section 3 provid es an
analysis of bisyllabic troichaic nicknames and trisyllabic nam es, w hile section 4 examines sid eissu es related to inpu t consid erations and anchoring. Section 5 ad d resses the topic of
monosyllabic feet in truncation. Finally, section 6 offers some concluding remarks.
2.
(1)
The data
a. (
): truncation with bisyllabic trochee (normal pattern)
Base
(
) Nickname
alkivia is
alkis
epaminodas
nodas
ko(n)stadinos
kostas, dinos
melpomeni
melpo
1
mar arita
rita
evagelia
evi, lia
loxias
loxas
b. (
(
) nickname and (
): trisyllabic [u nparsed
+ bisyl. trochee] and occasionally
) nickname too.
Base
(
) Nickname
(
) Nickname
aleksan ros
aleksis, alekos
aristotelis
aristos
aris, telis
panajotis
pana os
panos, jotis
ageliki
agelaii
pana inaikos
pana as
an ipaspistis
an ipas
c. stylistic variant (
Base
evagelos
) and (
(
) Variant
(
vagelis
a anasios
anasis
nikolaos
nikolas
d. ( ): bisyllabic [unparsed
Base
): same as (1b)
) Nickname
vagos
anos, nasos
nikos
+ monosyllabic trochee] nickname
( ) Nickname
xariklia
xara
ko(n)stadinos
kostis
The first column presents the base name which is used to address someone (when using the full
nam e instead of a nickname). Exceptional is (1c), w here norm ally the trisyllabic variant is used
instead . Althou gh the trisyllabic form s of (1b) and (1c) are su perficially id entical, they present
som e su btle d ifferences. The form of (1b) is perceived m ore like a nicknam e, w hile m ore like a
stylistic variant in (1c). Moreover, (1b) preserves m arked stru ctures w hile (1c) d ispenses w ith
them (i.e. lack of onsets, hiatus). N onetheless, it is the case that the trisyllabic form in (1b) is
evid ently truncated . In fact, som e base nam es only have a trisyllabic nickname and lack the
norm al bisyllabic one, e.g. aleksan ros > alekos, av erinos > av eris. My claim is that since the
prosod ic patterns of (1b) and (1c) are id entical, the stylistic variant in (1c) shou ld be treated as
truncation too.
Before going on it is w orth noting that the focus of this stud y is on the prosod ic patterns
of tru ncation. The circu m stances u nd er w hich stylistic variants and segm ental alternations
appear will not be discussediii.
2
3.
Word Minimization
3.1. Hierarchical alignment
The analysis initiates by first posing a qu estion, that is, w hat is com m on to (
), (
) and ( )
patterns? All of them have a foot, w hose right ed ge coincid es w ith the right ed ge of the
prosodic word. This indicates that the constraint in (2) is significant.
(2)
ALL-FT-R:
Align (Ft, Right, PrWd , Right): Align the right ed ge of the foot w ith the right ed ge of the
prosodic word.
Moreover, all of them conform to the notion of hierarchical alignment stated below:
(3)
Hierarchical Alignment (HIER AL) [Ito, Kitagawa and Mester 1996]:
Every prosodic constituent is aligned with some prosodic constituent that contains it.
(4)
a.
b.
c.
X
[ [
] ]
[ [
] ]
[ [
[x x]
] ]
To understand
how HIER AL
are all
-----works, let us
---have a look
--- at (4), where
--- , , * *and x--prosod ic constitu ents and
with
and
d om inates all others. In the binary (4a)
is right aligned (R-aligned ) w ith
is left-aligned (L-aligned)
. In the u nary (4b),
is both L- and R- aligned
with . H ierarchical m isalignm ent appears in the ternary (4c), w here althou gh
and
behave
like in (4a), there is also the constitu ent x in the m id d le w hich fails to be aligned in any w ay.
Therefore, ternary structures like (4c) are banned under HIER ALiv.
If w e now test som e hypothetical cand id ates for (2) and (3), w e w ill get the attested
patterns shown in (5d)-(5f). Section 5.2 shows why (5g) does not emerge.
(5)
HIER AL, ALL-FT-R
HIER AL
a.
b.
c.
(
)(
(
)
**
)( )
*
)
*
(
d.
(
)
e.
(
)
f.
( )
g.
( )
h.
ALL-FT-R
(
)
*
3
i.
(
j.
3.2.
)
(
*
)
N icknam es of the (
**
*
) type
The u nm arked prosod ic w ord is a binary foot (McCarthy and Prince 1994 for red u plication,
Benua 1995 and Pineros 2000 for tru ncation, am ong others). This obeys all of the constraints in
(6) [i.e. prosodic well-formedness constraints).
(6)
FTBIN:
Feet are binary on a syllabic or moraic analysis.
ALL-FT-R: Every
PARSE-
foot stands in final position in the PrWd.
: All syllables are parsed into feet.
Particu larly for tru ncation, a constraint that refers to the preservation of base m aterial to the
truncated output is needed too.
(7)
MAX-BT:
Every segm ent in the base has a correspond ent in the tru ncated form. (Benu a
1995)
Evidently
MAX-BT
m u st be low -ranked , since there is a preference to create an optim al prosod ic
form than preserve m ore m aterial in the tru ncated form , thu s a five-syllable base form like
alkivia is will truncate to the bisyllabic trochaic alkis. Therefore, (5e) supports the ranking:
(8)
(
): FTBIN, PARSE- >> MAX-BT
Moreover, the ranking in (9) m ust hold too. This is because cand id ates (5d ) and (5f) are u nd er
certain circum stances to be d iscu ssed later on optimal. The first thou gh violates
PARSE-
, w hile
the latter violates both PARSE- and FTBIN. Hence, these constraints must be low-ranked.
(9)
HIER AL, ALL-FT-R >> FTBIN, PARSE-
Combining (8) and (9) yields (10), an illustration of which follows in (11)v :
(10)
((11)
(
): HIER AL, ALL-FT-R >> FTBIN, PARSE- >> MAX-BT
HIER AL, ALL-FT-R >> FTBIN, PARSE-
Base: [(mar. a.)(ri.t-a)]
a.
b.
(mar. a.)(ri.t-a)
( a.ri.t-a)
c.
a.(ri.t-a)
d.
(ri.t-a)
>> MAX-BT
ALL-FT-R
HIERAL
FTBIN
PARSE-
MAX-BT
*!*
*!
*
m,a,r
*!
m,a,r
m,a,r, ,a
4
(11d ) is the sole w inner, because althou gh it incurs m any
MAX-BT
violations, it satisfies all
higher-ranked constraints perfectly.
3.3.
(
) vs. (
)
Moving on to the trisyllabic form s, I w ou ld like to propose that (
) and (
id entical rankings of constraints. What d istingu ishes them is that (
from the sou rce form , w hile (
) spring from near-
) preserves m ore m aterial
) achieves the optim al prosod ic shape of a bisyllabic trochee. To
get the trisyllabic result w e only need to re-rank the constraints in (8) so that
MAX-BT
now
dominates FTBIN and PARSE- . This is also intuitively straightforw ard , since it sim ply expresses
the fact that more base material is maintained.
(12)
(
): MAX-BT >> FTBIN, PARSE-
Combining (9) and (12) leads to (13). The tableaux below show how (13) is established.
(13)
(
): HIER AL, ALL-FT-R >> MAX-BT >> FTBIN, PARSE-
((14)
MAX-BT >> PARSE-
Base: [ni.(k .la).-os]vi
a.
b.
MAX-BT
ni.(k .l-as)
(ni.k-os)
(14a) w ins because it satisfies
PARSE-
a
o,l!,a
MAX-BT
*
to a higher d egree than its rival (14b). A natu ral qu estion
would be why MAX-BT does not force the full copying of the base, yielding the almost trivial (for
tru ncation) ou tpu t nikolaos. The answ er is that
HIER AL
MAX-BT
is still d om inated by the high-ranked
and ALL-FT-R. Hence (15b) cannot survive. This yields the expected outcome (15a).
((15)
HIER AL, ALL-FT-R
>>
MAX-BT
Base: [ni.(k .la).-os]
a.
b.
3.4.
ALL-FT-R
HIER AL
ni.(k .l-as)
ni.(k .la).-os
MAX-BT
a
*!
*
Summary of rankings
Up to now we have established the following rankings:
(16)
(17)
(
):
(
HIER AL, ALL-FT-R >> FTBIN, PARSE-
):
>> MAX-BT
HIER AL, ALL-FT-R >> MAX-BT >> FTBIN, PARSE-
As it can be easily observed , variation centres on
another.
5
PARSE-
,
MAX-BT
and the relationship to one
4.
Side issues
4.1.
Input assumptions
Up to now it has been im plicitly assu m ed that the choice of the suffix in the truncated form is
lexical. Som e evid ence for su ch an assertion follow s. In nam es like eleni > lena, frosini > froso, the
last vow el in the tru ncated form cannot be recovered from the base, w hile near m inimal pairs
efstrati-os > strat-os, efsta i-os > sta -is that have com m on base stem end ings and suffixes
nonetheless exhibit d ifferent suffixes in the nicknam e. This su ggests that the suffix is
unpredictable, thu s lexical. For this reason I assu m e that there is an inpu t for truncation = / T +
suffix/ . Contra Benua (1995) then, d istinct inpu ts for full and truncated form s are postu lated .
This assu m ption how ever is consistent w ith the m od el of Ito, Kitagaw a and Mester (1996) as
shown below:
Lexical
Structure:
Stem
Stem Trunc/Argot. etc
Surface
Structure:
Base
Trunc/Argot, etc.
The au thors argu e that this m od el potentially solves the problem w hen the tru ncated form
seem s to be m ore faithful to the inpu t than the full form ; This happens in English nicknames,
w here u nd erlying vow el contrasts are m aintained in the tru ncatu m , w hile neu tralized in the
fu ll forms, e.g. G[ ]r from G[ ]rard. H ow ever, I contend that su ch a problem can be
accom m od ated even in Benu a s m od el. Let s take for instance the nam e P[ ]tricia > P[ ]t,
w hich in Benu a s m od el w ill have the representation below . There is a single inpu t for both base
and tru ncatum and relationships Faith-IO and Faith-BT are used in the w ay Benua su ggests.
H ow ever, a new relationship has been ad d ed , i.e. Faith-IT w hich relates the inpu t and the
truncatumvii. Thu s, it is possible to relate the inpu t and the tru ncatu m w ithou t introd u cing an
input for truncation.
Lexical
Structure:
P[ ]tricia
Faith-IT
Faith-IO
Surface
Structure:
Faith-BT
P[ ]tricia
6
P[ ]t
Bu t even this am end m ent cannot accou nt for langu ages like Greek or Germ an (cf. Ito and
Mester (1997) w here it is argu ed that the inpu t for the Germ an tru ncatu ms is: / Tru nc + i/
yield ing form s like [gab-i], [and -i], [gorb-i], etc). In these langu ages, suffixal material in the
tru ncatu m s often cannot be recovered from neither the base (throu gh Faith-BT) nor the inpu t
(throu gh Faith-IT) as in e.g. eleni > lena. If both input and base inclu d e the su ffix i, how is it
possible to generate the tru ncatu m s suffix a in the absence of a d istinct inpu t for tru ncation
that inclu d es the su ffix a? Since this seem s im possible, Greek presents com pelling evid ence for
the distinct inputs approach advocated here [after Ito, Kitagawa and Mester (1996)].
4.2.
Anchoring
In this section w e exam ine w here the copying from the base to the tru ncatu m starts. Crosslingu istically, anchoring to the left ed ge of the PrWd (cf. Benu a 1995 for Japanese, Ito and
Mester 1997 for Germ an, van d e Weijer 1989 for H u ngarian, am ong others) or anchoring to the
head of the foot (cf. Pineros 2000 for Spanish, Cabre and Kenstow icz 1995 for Catalan) is
preferred . Som e languages allow both types, e.g. Spanish profesor > profe (Colina 1996) and
[doro(tea)]PrWd > [(tea)]PrWd (Pineros 2000), or Greek e.g. spiri on > spiros , xarilaos > xaris,
melpomeni > melpo (anchor to the left of PrWd ) or aristotelis > telis, evagelia > lia (anchor to the
foot head). But some Greek names seem not to fit in this picture.
(18)
(e)o
ros > oros
erasimos > simos
aristoksenos > ksenos
The tru ncated form here d oes not anchor to either the left ed ge of the foot or to the left ed ge of
the PrWd . Is this then problem atic? Should w e em ploy a d ifferent kind of anchoring? A
potential alternative is to u se constraints that refer to stem -ed ges. In particu lar, a constraint like
ANCHOR-R-STEM
w ou ld be able to select the correct point as the beginning of copying in the
nam es above. Below , I w ill attem pt to show that althou gh this type of constraints are at first
glance useful, they produce a different set of problems and thus should not be employed.
First, note that the nam es in (18) are tru ly exceptional. There is only a hand fu l of them ,
perhaps su ggesting that they are lexicalised . Moreover, in the d iscu ssion of su perheavy
syllables in Germ an, H all (2002) states that these are only allow ed at the right ed ge of the
prosod ic w ord . N onetheless, they can also be fou nd w ord -medially in som e Germ an proper
nam es. The argu m ent goes that in fact these w ord s form prosod ic com pou nd s w ithou t being
morphologically compounded, e.g. [zalts]PrWd [b
k]PrWd.
One could make a similar claim for Greek, so that a name like aristoksenos would have the
follow ing representation: [aristo]PrWd [ksenos]PrWdviii. What happens then is that anchoring again
applies at the left ed ge of the PrWd , bu t this tim e it refers to the second PrWd , i.e. [ksenos].
Su ch an approach is entirely feasible, since the nam es of (18) w ere at least at som e stage of the
history of the language also morphological compoundsix, but even if they had not been, German
clearly presents a case w here m orphological sim plex form s can be nevertheless prosod ically
compounded.
7
Thu s, the argu m ent for anchoring to the stem seem s rather w eak. Let u s see how ever how the
tw o approaches, nam ely
ANCHOR-L-P r W d / Ft
vs.
ANCHOR-L/ R STEM
fare in term s of other
nicknam es. In (19) a set of nicknam es is presented , one I call d iminu tive nicknam es , since these
are somehow related to the full diminutives.
(19)
Source name
Full Diminutive
Diminutive Nickname
vasil-is
vasil-ak-is
lak-is
erasim-os
erasim-ak-is
mak-is
jonis-is
jonis-ak-is
sak-is
panaj t-is
panajot-ak-is
tak-is
imitr-a
imitr-ul-a
t(r)ul-a
elen-i
(e)len-ul-a
nul-a
paraskev-i
paraskev-ul-a
vul-a
xarikli-a or xar-a
xar-ul-a
rul-a
The qu estion is how these nicknam es are form ed . There are at least tw o possibilities. The first
hypothesis or Foot approach argues that the base is the fu ll d im inu tive. What is copied to the
tru ncate then is the foot (w holly) as in vasi(l-akis) > (lakis) (cf. Topintzi 2002) d ue to
ANCHOR-L-Ft
. On the contrary, accord ing to the second hypothesis or Stem approach, the base is the source
nam e. The inpu t of tru ncation is: / T + -akis or
-u la / and the constraint in (20) accou nts for
the anchoring attested.
(20)
ANCHOR-R-STEM:
Any element at the right periphery of the stem in the base has a correspondent at the right
periphery of the stem in the truncated form.
When this constraint is employed , the initiation of copying varies a bit d epend ing on the size of
the su ffix. Given that the ou tpu t of truncation m ust be a bisyllabic trochee, w hen the suffix is
m onosyllabic, e.g. -os, a w hole syllable m ay be ad d ed , e.g. nas- yield ing nas-os for an inpu t / T
+ -os / ; bu t w hen the suffix is d isyllabic - as is the case in d im inu tive nicknam es - then the
optim al ou tpu t size has alread y been achieved and only an onset is ad d ed accord ing to
ANCHOR-R-STEM
i.e. the last consonant of the stem to attain syllabic w ell-form ed ness, e.g. s-akis
for an input /T + -akis/.
H ow ever, the stem approach faces a problem that d oes not arise in the foot approach.
More concretely, according to the former, it is anticipated that a diminutive nickname may arise
in the absence of the full d im inu tive, bu t this is never the case. On the contrary, it is possible
that there exists a fu ll d im inu tive w ithou t the d im inu tive nicknam e, bu t it is not possible to
have a d iminu tive nicknam e w ithou t any full d im inu tive. In ad d ition, if a native speaker is
asked w here e.g. the nam e sakis com es from , the answ er w ill be that it com es from this
sequence: jonis-is > jonis-ak-is > sak-is, ind icating that an interm ed iate stage, nam ely jonis-akis, is recognised for the generation of the d im inu tive nicknam e. This fact also su ggests that
8
m u ltiple bases for tru ncation m ay exist, as in the schem a suggested by d im inu tive nicknam es,
i.e. source name
full d im inu tive
d im inu tive nicknam e.
Fu rther evid ence for the foot approach com es from red u plicated nicknam es.
(21)
Source name
Reduplicated nickname
paraskev-i
viv-i
a in-a
nan-a
agelik-i
kik-i
mari-a
rir-i
evageli-a
lil-i
imitr-is
mim -is
Here the stressed vowel of the source name which may be within e.g. mari-a or outside the stem,
e.g. a in-a is targeted in red u plication.
ANCHOR-L-Ft
neatly accounts for that, w hile a far less
straightforward account would be possible through ANCHOR-R-STEM.
H aving presented argu m ents against the u se of
ANCHOR-L/ R-STEM,
the anchoring
constraints employed for Greek are shown below:
(22)
ANCHOR-L-PrWd:
Anchor the left edge of the source form (Pineros 2000)
Any elem ent at the left periphery of the source form has a correspond ent at the left
periphery of the truncated form.
(23) ANCHOR-L-Ft:
Anchor the left edge of a foot (adapted from Benua 1995)
Every correspondent of a foot-initial segment is foot-initial.
These constraints present som e variation w ith respect to each other as it is illustrated below ,
w here the base nam e kostadinos can either surface as kostas (w hen
Ft)
ANCHOR-L P r W d
>>
ANCHOR-L
or as dinos (when ANCHOR-L Ft >> ANCHOR-L Pr W d).
((24)
ANCHOR-L Pr W d, ANCHOR-L Ft
Base: [(kos.ta).(di.n-os)]
a.
b.
ANCHOR-L Pr W d
(k s.t-as)
(di.n-os)
*
*
Althou gh there is a lot of anchoring variation in (
) nicknam es, e.g. kostadinos > kostas or dinos,
evagelia > evi or lia, fixed rankings shou ld probably be postu lated in
Pr Wd
>>
ANCHOR-L Ft,
ANCHOR-L Ft
>>
ANCHOR-L Ft
(
) w here
ANCHOR-L
e.g. aristotelis > aristos bu t *totelis and in d im inu tive nicknam es w here
ANCHOR-L P r W d ,
e.g. erasimakis > makis bu t * akis. Below are the rankings
formed when anchoring is also considered.
9
(25) i) (
):
HIER AL, ALL-FT-R >> AN CH OR-L P r W d
ii) d im inu tive (
>>
ANCHOR-L Ft
>> MAX-BT >> FTBIN ,
PARSE-
): H IER AL, ALL-FT-R >> AN CH OR-L Ft >> AN CH OR-L P r W d >> FTBIN , PARSE-
>> MAX-BT
iii) rest of (
):
H IER AL, ALL-FT-R
>>
AN CH OR-L Ft, AN CH OR-L P r W d
>>
FTBIN , PARSE-
>>
MAX-BT
5. Monosyllabic feet
5.1.
Nicknames of the ( ) type
For these nam es I am assu m ing that there is an und erlyingly accented suffix / -is/ or / -a/.
Moreover, stress m u st be realized in w ord -final position. This can be ensu red if the constraint
below is high rankedx.
(26)
MAX (LA):
(after Revithiadou 1999)
A lexical accent of S1 (input) has a correspondent in S2 (output).
(27)
MAX LA, HIER AL, ALL-FT-R >> PARSE-
Base: [(kos.ta).(di.nos)]
Input:
MAX LA
HIER AL
ALL-FT-R
PARSE-
*
/ T
+
a.
b.
-is /
kos.(tis)
(kos.tis)
*
*!
Adding MAX LA in our constraint ranking suffices to produce the correct result.
5.2.
Lack of monosyllabic nicknames
Recall that earlier w e have seen that the cand id ate (5g)
form ed given
H IER AL
and
ALL-FT-R,
the m onosyllabic cand id ate
is w ell
bu t nonetheless this is u nattested . This lack is u nexpected
because Greek ad m its m onosyllabic w ord s e.g. les
m oreover a com parison betw een ( ) and
say-2sg.PRES , mis m u scle
NOM.sg
and
( ) w ou ld favou r the first, since all else equal the
latter also violates PARSE- while the former does not.
To state the problem in term s of em pirical d ata, the qu estion is w hy w e d o not get (kis)
rather than kos.(tis). To accou nt for this m atter, I w ill ad opt the Weak Layering H ypothesis (Ito
and Mester 1992) w hich d istingu ishes betw een the strict m inimal w ord , i.e. a single foot, and
the loose minim al w ord , i.e. a foot accom panied by m inimal u nparsed material, that is a
syllable. In previou s w ork (Topintzi 2002), the notion of Word Binarity (28) had also been u sed .
Its effects are illustrated in (29).
10
(28) Word Binarity:
(Ito and Mester 1992)
P-derived words must be prosodically binary.
P-d erived refers to w ord s that are related to m ore basic w ord s by m eans of prosod icmorphological operations (e.g. truncation).
(29)
a.
Wd
F
b.
c.
Wd
F
Wd
* d.
F
F
Wd
F
In the su pram inim al w ord of (29a) and in the loose m inimal w ord of (29b), there is binary
branching at the w ord level. In the strict m inim al w ord (29c), binary branching occu rs at the
foot level. Finally, the subminimal (29d) presents no binary branching at any level producing an
ill-form ed stru ctu re. If w e w ere now to translate Word Binarity in term s of a constraint, it
w ou ld be the case that it is top-ranked banning m onosyllabic stru ctu res from the d om ain of
truncationxi.
H ow ever, such an approach introd u ces a m echanism
Word Binarity
w hich on closer
inspection proves u nnecessary. The cu rrent proposal argu es that no new m achinery need s to be
involved.
MAX-BT
s ranking sim ply su ffices. Recall the ranking for trisyllabic nicknam es
repeated below:
(30)
(
):
HIER AL, ALL-FT-R >> MAX-BT >> FTBIN, PARSE-
The only ingred ient m issing from this ranking to get the
( ) patterns and exclu d e ( ) is the
effects of MAX LA. If this is added, the ban on monosyllabic nicknames falls out.
((31)
MAX LA , HIER AL, ALL-FT-R >> MAX-BT >> PARSE-
Base: [(kos.ta).(di.nos)]
Input:
*
/ T
a.
b.
+
MAX
HIER
ALL-
LA
AL
FT-R
MAX-BT
PARSE-
-is /
kos.(tis)
(kis)
a,d,i,n,o,s
o,s,t,a,d,i!,n,o,s
*
The ad vantage of this approach is that it d oes not need to introd u ce new machinery in the form
of Word Binarity. This is plainly a side-effect of the gram m ar s structu re. Furtherm ore, the loose
m inim al w ord , i.e. F w hich in Greek can take either the form
11
(
) as in aristos or the form
( ) as in xara can be prod u ced by a single ranking only: (MAX LA ,) H IER AL, ALL-FT-R >> MAX-BT
>> PARSE- .
6.
Conclusion
In this paper, it has been show n that in fact there are three patterns in Greek tru ncation (
(
),
) and ( ) rather than a single d isyllabic trochaic one. These patterns have been prod u ced
throu gh the interaction of H IER AL and
ALL-FT-R
w ith prosod ic and lexical stress requirem ents as
w ell as the am ou nt of copying requ ired each tim e. Specifically, the attested form s are generated
through the following rankings:
Strict MinWord:
(
): HIER AL, ALL-FT-R >> PARSE- >> MAX-BT
Loose MinWord:
i) (
ii)
): HIER AL, ALL-FT-R >> MAX-BT >> PARSE- [to preserve more material]
( ):
MAX LA , H IER AL, ALL-FT-R
>>
MAX-BT
>>
PARSE-
[im posed by lexical stress
requirements]
As a generalisation then, Greek nicknam e form ation m anifests all types of the m inim al w ord ,
both strict and loose.
i
I am grateful to Moira Yip, Angeliki Malikou ti-Drachm an, Ioanna Kappa, Shigeto Kaw ahara
and the participants at the 6th ICGL for fruitful d iscussion and com m ents. All rem aining errors
are my own.
ii ageliki has an even sm aller nicknam e kiki. This belongs to the class of red u plicated nicknam es
(see section 4.2.).
iii A w ord of cau tion; variation in nicknam es is anticipated and in fact attested (Benua 1995, fn.
34, Ito and Mester 1997, fn.5, am ong others). Ku risu (2001) argu es that tru e tru ncation appears
in the form of su btractive m orphology, w here there is also change of m orpho-syntactic
category, e.g. Koasati plural formation involves deletion of the final coda or rime in the singular
form . Prosod ic tru ncation of the type d iscu ssed here d oes not involve any gramm atical change
of this type. It is thus consid ered m ore of a socio-linguistic phenom enon and as su ch, it is
expected to present a fair amount of variation.
iv Angeliki Malikou ti-Drachm an and Marina Tzakosta (p.c.) w ond er w hether H IER AL is really
needed in the constraint ranking. I contend that it is indispensable, because otherwise forms like
(5h) and (5j) w ou ld be possible. For instance, for the base name aristotelis, w e have the ou tpu t
aristos prod uced by the ranking of trisyllabic nam es in (17), nam ely H IER AL, ALL-FT-R >> MAX-BT
>> FTBIN , PARSE- . All else equal, but without H IER AL we would expect a surface truncated form
aristo(telis) since this w ou ld satisfy ALL-FT-R and w ould respect MAX-BT perfectly. Evid ently such
a truncated form is not possible. Hence, the role of HIER AL is decisive.
v In the tableau x that follow I w ill mention no inpu t for the tru ncated form , althou gh I w ill still
assu m e it (see section 4.1), e.g. for Nikolas / T + -as / , for Nikos / T + -os / , etc. To avoid fu rther
complications, MAX-BT violations only cou nt the segm ents of the root w ithin the base that are
copied to the truncatum and not the suffixes.
vi Anthi Revithiad ou (p.c.) observes that it is possible to assu me that the a of nikolas is in fact
part of the base rather than of the su ffix. As a resu lt MAX-BT w ou ld not be violated on that
occasion. Althou gh, this is ind eed true, there are other sim ilar cases that lead us to assu m e that
12
a hypothesis w here the vow el is part of the su ffix and not of the base is more viable, e.g.
pana inaik-os > pana -as, *pana -is, agelik-i > agel-a, *agel-i. H ere the vow el follow ing the final
stem consonant is evid ently part of the su ffix. In the case of nikolas, it is not possible to
d eterm ine w hether it shou ld be nikol-as or nikola-s and in fact this d oes not change anything in
ou r analysis. H ow ever, given the exam ples above, w here su ch a d ecision can be m ad e, reasons
of consistency suggest that the vowel a in nikolas should be considered part of the suffix.
vii Such a relationship presents no problem whatsoever. In fact it makes Benu a s tru ncation
model more symmetrical to the reduplication model of McCarthy and Prince (1995) where
Faith-IR is allowed.
viii The first o in the w ord is the linking vow el of com pou nd s. A m ore precise representation
should show that. Simplifying a bit however, does not affect the argument in any way.
ix My intuition is that at least som e of these nam es, e.g. (e)o oros and aristoksenos are still
morphological compounds. Other speakers agree with this impression too.
x This is a slightly sim plified version of Revithiad ou s (1999) original formu lation. I take here
MAX LA to require correspondence in both the skeletal and melodic tiers.
xi Since Word Binarity is now treated as a constraint, there m ay be languages w here it is low ranked ad m itting m onosyllabic tru ncates. This pred iction is borne ou t. Glow acka (2002) reports
som e m onosyllabic tru ncates for m asculine nam es, e.g. stanis aw > stas, ryszard > rys ind icating
that m onosyllabic tru ncation is possible although by far less com m on than norm al bisyllabic
truncation.
References:
Benua, Lau ra. 1995. Id entity effects in morphological tru ncation. In Papers in Optimality Theory,
ed . Jill Beckman, Lau ra Walsh Dickey and Suzanne Urbanczyk, pp. 77-136. Am herst, MA:
GLSA. [Available on Rutgers Optimality Archive].
Cabre, Teresa and Michael Kenstow icz. 1995. Prosod ic Trapping in Catalan. Linguistic Inquiry
26, 694-705.
Colina, Sonia. 1996. Spanish Tru ncation processes: the em ergence of the u nmarked . Linguistics
34, 1199-1218.
Drachm an Gaberell and Angeliki Malikou ti-Drachm an. 1996. Greek accentu ation. In Word
Prosodic Systems in the Languages of Europe, ed . H . van d er H u lst. Berlin and N ew York:
Mouton de Gruyter.
Drachm an, G., A. Malikouti-Drachm an , E. Georgiou, G. Stavrou, R. Sim eou and E. Trifonos.
1999.
[Cypriot N icknam e Form ation]. In
Proceedings of the 4th International Conference on Greek Linguistics, N icosia: University Stu d io
Press, pp. 487-494.
Glowacka, Dorota. 2002. Truncation in Polish, ms, UCL.
H all, T. A. 2002. The d istribu tion of su perheavy syllables in Stand ard German. The Linguistic
Review 19, pp. 377-420
Ito, Junko and Armin Mester. 1992. Weak Layering and Word Binarity. MS, UC Santa Cruz.
Ito, Ju nko and Arm in Mester. 1997. Sym pathy Theory and German Tru ncations. In University of
M aryland W orking Papers in Linguistics 5. Selected Phonology Papers from Hopkins Optimality
Theory W orkshop 1997/ University of M aryland M ayfest 1997, ed . Viola Miglio and Bru ce
Moren, pp. 117-39. [Available on Rutgers Optimality Archive].
13
Ito, Ju nko, Arm in Mester and Yoshihisa Kitagaw a. 1996. Prosod ic Faithfu lness and
correspond ence: evid ence from a Japanese argot. Journal of East A sian Linguistics 5, 217294.
Ku risu , Kazutaka. 2001. The phonology of morpheme realization. Doctoral Dissertation. University
of California, Santa Cruz
Malikouti-Drachm an, Angeliki. 1999. Constraints on Greek H ypocoristics. Greek Linguistics 97.
Proceedings of the 3rd International Conference on the Greek Language. A thens 25-27 September
1997, ed. A. Mozer, pp. 391-400. Athens: Ellinika Grammata.
Malikouti-Drachm an, Angeliki and Gaberell Drachm an. 1989. Stress in Greek [Tonizm os sta
Ellinika]. Studies in Greek Linguistics 1989, 127-143. University of Thessaloniki.
McCarthy, John. 2002. A Thematic Guide to Optimality Theory. Cambridge: Cambridge University
Press.
McCarthy, John and Alan Prince. 1995. Faithfulness and red u plicative id entity. In University of
Massachusetts Occasional Papers in Linguistics 18, ed. Jill Beckman, Laura Walsh Dickey and
Suzanne Urbanczyk, pp. 77-136. Am herst, MA: GLSA. [Available on Ru tgers Optim ality
Archive].
Pineros, Carlos-Ed u ard o. 2000. Foot-sensitive Word Minimization in Spanish. Probus 12(2), 291324. [Available on Rutgers Optimality Archive].
Ralli, Aggeliki and Loudovikos Touradzidis. 1992. A computational treatment of stress in Greek
inflected forms. Language and Speech 35(4), 435-453.
Revithiad ou , Anthi. 1999. Headmost A ccent W ins: Head Dominance and Ideal Prosodic Form in
Lexical A ccent Systems. The H agu e: H olland Acad em ic Graphics. [Doctoral Dissertation,
University of Leiden].
Topintzi, N ina. 2002. Prosod ic Patterns in
Greek Tru ncation and
Red u plication: a
Correspondence Approach. MA dissertation, UCL.
van d e Weijer, Jeroen. 1989. The formation of d imu nitive nam es in H u ngarian. A cta Linguistica
Hungarica 39, 353-371.
14
This document was created with Win2PDF available at http://www.daneprairie.com.
The unregistered version of Win2PDF is for evaluation or non-commercial use only.