Simonovic arsenijevic - in and out of paradigms - bcn2013
1. In and out of paradigms:
How to do everything with very few affixes
Marko Simonović Boban Arsenijević
Utrecht University University of Niš
1
2. Aim of the talk
• Present data from Serbo-Croatian (S-C) about
phonologically identical or similar suffixes
traditionally analyzed as unrelated.
• Argue for the minimal analysis, treating the sets
of similar suffixes as single suffixes.
• Explain the phonological and semantic
differences in terms of paradigm membership.
• Point to the crucial role paradigm membership
and surface generalizations play in processing,
and in language change.
2
3. Part I: Adjectives and participles
• Five traditional grammar’s endings, related by
sharing an –n, figure in the derivation of passive
participles and adjectives in S-C.
• Participial: -en, -an.
Adjectival: -en, -an, -(a)n (the fleeting /a/).
• We argue that there are only two endings
involved, -n and -a/e (the latter with two variants
selecting complementary phonological contexts).
• Uniform semantics, sensitivity to paradigms.
3
4. Passive participle
• Regularly derived Serbo-Croatian passive
participles end in an /n/.
• Traditional grammar recognizes two different
endings: –an and –en.
• –en goes only to palatal contexts, –an mostly
to the non-palatal ones (to –a stems).
(1) gleda-an -> gledaan voli-en -> voljen
watch-an love-en
watched loved
4
5. Semantics
• Semantically, the two endings are equivalent.
• They derive participles, expressions with hybrid,
verbal and adjectival semantics.
• The ‘adjectival semantics’ of participles is non-
gradable (their gradable dimensions inherited
from the verbal stem, as in well/better educated,
are ignored as irrelevant, cf. Kennedy & McNally).
(2) lepljen #lepljen-iji bolje lepljen
glued glued-Comp better glued
5
6. Adjectives in -n
• A majority of S-C adjectives ending in an /n/.
• In traditional grammars, 3 different endings:
1) the ending –an, limited to palatal contexts
of primitive stems, 2) the ending –en in non-
palatal primitive stems, 3) the unconstrained
ending –(a)n.
(3) pliš-an glin-en mir-(a)n
velvet-an clay-en peace-(a)n
velvetAdj, clayAdj peaceful
6
7. Semantic asymmetries
• –an and –en endings derive non-gradable
adjectives (in most cases the material something
is made of, or other defining property).
• Endings reserved for classificational adjectives?
• –(a)n adjectives are unconstrained in this sense.
(4) a. #led-en-ija figura, b. #košt-an-ija srž
ice-en-Comp figure bone-an-Comp core
c. mir-n-iji kraj
peace-n-Comp neighborhood
‘a/the more peaceful neighborhood’
7
8. Classificatory
• Arguments in favor of treating –en/an as an
ending deriving classificatory adjectives.
• –en/an adjectives do not nominalize.
(5) kopn-en(*-ost/*-stvo/*-ina/*-ota…)
land-en N N N N
• –en/an endings are in complementary
distribution with the classificatory –ski.
8
9. Phonological asymmetries
• -en only in non-palatal contexts
• -an only in palatal contexts
• -(a)n has an epenthetic /a/, inserted when
the ending is word-final.
(5) smeš-an smeš-n-og
laugh-an laugh-n-GenMSg
‘funny’
• /a/ is epenthetic ⇒ the actual ending is –n.
9
12. How many suffixes?
• Ockham’s razor.
• –en/an is one suffix across the board, and –n
is a separate suffix?
• –en/an derives non-gradable adjectives
specifying an absolute, pervading, presence of
a property on/in the argument .
• –n derives possibly gradable adjectives
specifying an underspecified relation between
the argument and the meaning of the stem.
12
13. Or even simpler
• Two suffixes, –n and –e/a?
• –n is semantically the lightest adjectival ending,
specifying an uncostrained relation between the
argument and the meaning of the stem.
• –e/a subspecifies that the relation is of a
particular type: a total, absolute, pervading,
presence of a property on/in the argument
(therefore non-gradable).
• Problem: the inverse phonological distribution of
–en/an between participles and adjectives.
13
14. Argument 1
• Let us first present two arguments in favor of
the simpler analyses presented.
• Argument 1: Whenever there is a pair of
adjectives, one with the ending –n and the
other with the ending –en/an, their semantics
is strictly as predicted.
(6) a. med-n-i b. med-en-i
honey-n-PL honey-en-PL
‘related to honey’ ‘made of honey’
14
15. Even participles
• Additional support that –en/an is the same
across the board comes from the fact that
(adjectival) participles can be members of the
same kind of pairs.
(7) a. od-seč-n-i b. od-seč-en-i
off-cut-n-PL off-cut-en-PL
‘abrupt’ ‘cut off’
15
16. Argument 2
• No –en : –an adjective pairs attested.
• Also phonological reasons: the two suffixes
appear in different phonological contexts.
• But this only strengthens the argument: two
items in complementary phonological
distribution, sharing semantic effects are most
probably two phonological variants of the
same item.
16
17. Argument 3
• Participles aside, –en/an suffixes take only
monomorphemic stems, while –n may also
combine with the more complex ones.
(8) a. vun-en-i c. ne-pri-seb-n-i
wool-en-PL not-by-self-n-PL
‘woolen’ ‘nervous, uncalm’
b. iskr-en-i d. bez-po-treb-n-i
sparcle-en-PL without-over-need-n-PL
‘frank’ ‘unnecessary’
17
18. Explaining away the problem
• –en only in palatal contexts in participles, only in
non-palatal in adjectives; –an almost the oposite.
• Arsenijević & Simonović (2013): different
phonological constraints in and out of paradigms.
• Paradigm: the maximal set of forms productively
derived from a morphosyntactic class, while
preserving semantic transparency.
• Participles are part of the verbal paradigm,
–en/an adjectives are not members of their
stems’ paradigms.
18
19. Paradigm-membership marking
• The inverse distribution: an emergent
regularity between phonological and morpho-
syntactic factors that facilitates the different
effects/processing of the same suffix in
paradigmatic and non-paradigmatic domains.
• This also explains the very fact that there are
two variants, since their inverse distribution
clearly shows that there’s little phonological
reason to have them both.
19
20. Part II: Nominalizations
• Ending –je derives collective nouns from
nouns, NPs , PPs and participial VPs.
• In its deverbal nominalizations, derived nouns
take different prosodic patterns depending on
the aspect of the verb.
• We argue that this split again corresponds to
the split between within paradigm derivations
and those with some idiosyncratic properties.
20
21. The collective -je
• Collective nouns are not productive in S-C any
more, hence also not paradigm members.
(9) a. list + -je > lišće b.crep + -je > *crepje
leaf roof tile
• They have the syntactic and semantic
properties of mass nouns (singularia tantum,
much/*many, cumulative and divisive),
• while still preserving the meaning of an
assembly (e.g. compatible with one by one).
21
22. Prosody
• By a rule, –je nominalizations take a prosodic
pattern different from that of the base.
(10) a. kAmeen : kAmEEn-je
stone stoneColl
b. sa zvEEzdA : sAA-zvEEzd-je
with stars constelation
c. jEzIk slOvo : jEzik-o-sloov-je
language word linguistics
22
23. Deverbal nominalizations
• Regular gerunds in S-C: past participle + –je.
• S-C verbs usually come in aspectual pairs (a
perfective and an imperfective).
• Gerunds are derived from the imperfective
variant.
(11) a. jed-en(-je) b. po-jed-en(*-je)
eatimpf-en-je over-eatperf-en-je
‘eating’ int. ‘eating up’
23
24. Perfective nominalizations
• There is a limited set of –je nominalizations
derived from perfective verbs.
• Resultative semantics (Ignjatović 2013).
• Virtually all lexicalized (idiomatic).
(12) u-stolic-en-je
in-chair-en-je
‘enthronement’ (not generally placing
something or someone in a chair)
24
25. Prosodic asymmetries
• Perfective nominalizations behave like other
–je nominalizations, in taking up a different
prosodic pattern, imperfective nominalizations
preserve the prosody of the stem.
(13) pOstA-ti (infinitive)
a. pOstA-ja-an-je b. postA-An-jE
becomeperf-Impf-an-je becomeperf-n-je
‘becoming’ ‘creation’
25
26. Paradigms
• Explanation: imperfective nominalizations
belong to the verbal paradigm.
• S-C nominalizations preserve lexical prosody
within the paradigm, and drop it otherwise
(Arsenijević & Simonović 2013).
• In this way, gerunds (paradigm members) are
distinguished from resultative (and other)
nominalizations, even though derived by the
same suffix.
26
27. Support
• Even with an imperfective nominalization, if it
is idiomatic, the prosodic pattern may change
with its leaving of the paradigm (by receiving a
count interpretation):
(14) pUtova-ti (infinitive)
a. pUtova-an-je b. putovA-An-jE
travelimpf-an-je travelimpf-an-je
‘traveling’ ‘trip’
27
28. Theoretical consequences
• Arsenijević & Simonović (2013): structurally
complex stems within the paradigm, primitive
ones outside (cf. Marvin 2002, Roy 2010…).
• Phonological correlates of semantic/syntactic
properties: (via) surface generalizations and
selective lexicalization, (to) grammatical
constraints.
• The ontological status of classificatory properties,
in views taking the syntactic and morphological
behavior as evidence?
28
29. Conclusion
• Phonology is sensitive to paradigm membership.
• As paradigms are special interpretation domains
(transparency, inheritance of semantic properties
of the stem etc.), phonology of the derived forms
closely correlates with their semantics.
• Suffixes have an ever-applicable version for
paradigm members and a more demanding one
for word formation.
• Productive suffixes get ‘recycled’ in less
productive domains.
29