1
International Morphology Meeting, Budapest, May 9-12, 2002
SUFFIX COMBINATIONS, GRAMMATICAL
RESTRICTIONS, AND PARSING
JENNIFER HAY & INGO PLAG
University of Canterbury/Universität Siegen
1. INTRODUCTION
(1)
Which principles or mechanisms constrain the combinatorial properties of affixes?
(2)
Base-affix, affix-affix restrictions
a.
computer-ize, *computer-ate, *computer-en
art-ify, *art-ate, *art-∅
Naz-ify, *en-nazi, *Nazi-ate
passiv-ate, passiv-ize
b.
átom, atóm-ic, àtom-íc-ity
átom, átom-less, átom-less-ness
*atom-less-ity
(3)
- lexical strata
- selectional restrictions
- processing constraints
Main claim:
Both selectional restrictions and parsing constraints are instrumental in determining
possible and impossible suffix combinations. Only well parsable combinations are possible
combinations, and this range of possible combinations is further curtailed by selectional
restrictions.
Hay & Plag: Suffix combinations, grammatical restrictions, and parsing
2
2. EXISTING MODELS
2.1.
Stratum-oriented models (Siegel 1974, Allen 1978, Selkirk 1982, Kiparsky 1982,
Mohanan 1986, Giegerich 1999)
(4)
Class I suffixes: +ion, +ity, +y, +al, +ic, +ate, +ous, +ive, +able, +ize
Class I prefixes: re+, con+, de+, sub+, pre+, in+, en+, be+
Class II suffixes: #ness, #less, #hood, #ful, #ly, #y, #like, #ist, #able, #ize
Class II prefixes: re#, sub#, un#, non#, de#, semi#, anti#
(5)
Problems: numerous counterexamples, dual membership, suffix order within
levels, bracketing paradoxes, no predictions on numerous other restrictions.
2.2.
Against strata, for individual restrictions:
(6)
Fabb (1988): 4 classes of suffixes
a. Group 1: suffixes that do not attach to already suffixed words (28 out of 43)
b. Group 2: suffixes that attach outside one other suffix (6 out of 43)
c. Group 3: suffixes that attach freely (3 out of 43)
d. Group 4: problematic suffixes (6 out of 43)
Problems: Numerous counterexamples, arbitrary classes, no predictions on many
other restrictions.
(7)
Plag (1996, 1999): suffix-particular, base-driven restrictions
Problem: Is everything arbitrary?
3. A NEW ALTERNATIVE: COMPLEXITY BASED ORDERING (HAY 2000, in press)
(8)
COMPLEXITY BASED ORDERING (CBO)
“While some affixes basically tolerate no internal structure, others will tolerate
structure to some minimum degree. The degree of internal structure tolerated by
an affix is not determined by selectional restrictions,
however.
Rather, it is
3
International Morphology Meeting, Budapest, May 9-12, 2002
determined by how much structure that affix, itself, creates. Phrased in terms of
processing, an affix which can be easily parsed out should not occur inside an affix which
can not” (Hay, in press)
(9)
Dual routes in morphological processing:
whole-word access vs. decomposed access
(10)
Relative frequency and parsability (Hay 2001)
The higher the frequency of the derived word in relation to the base word, the less
likely is decompososition. Alternatively, the lower the frequency of the derived
word in relation to the base word, the more likely is decomposition.
(11)
a. insane is much more frequent than sane ⇒ whole word bias
b. infirm is much less frequent than firm ⇒ parsing bias
(12)
Graded decomposability of complex words
a. The same suffix will be differently separable in different words depending on the
respective frequencies of base and derivative. Individual words containing the same
suffix will tend to be more decomposable if they are less frequent than their base
than if they are more frequent than their base (e.g. discernment is more
decomposable than government).
b. Suffixes represented by many words which are less frequent than their bases will
tend to be more separable than suffixes represented by few words which are less
frequent than their bases (e.g. -ish tends to be more separable than -ic).
c. More separable affixes will occur outside less separable affixes. (cf. also Burzio
1994:354)
(13)
HYPOTHESIS 1
Suffixes can be ordered in a hierarchy of juncture strength, such that affixes
following an affix A on the hierarchy can be added to words containing A, but
affixes preceding A on the hierarchy cannot freely attach to words containing A.
Hay & Plag: Suffix combinations, grammatical restrictions, and parsing
(14)
(15)
4
a.
Hierarchy of suffixes: X-Y-Z-A-B-C-D
b.
Possible combinations: BASE-A-B, BASE-X-A-C, BASE-Y-Z-A
c.
Impossible combinations: *BASE-A-Z, *BASE-Y-A-Z, *BASE-X-A-Y
HYPOTHESIS 2 (cf. Plag in press)
The order of affixes is determined by selectional restrictions
4. TESTING COMPETING HYPOTHESES
(16)
Purpose:
Investigate the relationship between processing constraints and grammatical
constraints
(17)
Data set: 15 suffixes from Aronoff & Fuhrhop (in press)
suffix
examples
derived
category
base category
-dom
freedom
stardom
employee
deepen
N
-ee
-en
-er
-ess
-ful
-ful
-hood
-ish
-less
-ling
-ly
-ness
-ship
-th
semantic category of
derivative
N/(ADJ)
semantic
restriction on
base
?
N
V
V/N
ADJ/N
sentient
?
baker
Londoner
princess
N
V/N
?
- monosyllable
- obstruentfinal
person/instrument/etc. -
N
N
female person/animal
-
careful
cupful
childhood
falsehood
Jewish
schoolboyish
greenish
fortyish
careless
duckling
ADJ
N
N
N
N
N/(ADJ)
(male ?)
person/animal
abstract noun
concrete noun
person noun
qualitative adjective
partitive noun
‘state of being X’
-
ADJ/Num
N/ADJ/Num
N=person
noun
similative
?
ADJ
N
N
N
?
animate noun
?
fatherly
deadly
kindness
friendship
depth
growth
ADJ
N/(ADJ)
N
N
N
any
N
ADJ/V
N=person and
time nouns
?
person noun
?
‘without X’
young animal, (young)
human being
similative
quality noun
status, collectivity
quality noun
- monosyllable
‘status, realm,
collectivity’
person (non-agent)
change-of-state verb
phonological
restriction on
base
?
?
5
International Morphology Meeting, Budapest, May 9-12, 2002
(18)
procedure: checking of all possible combinations (N=210) for attestations in BNC,
CELEX, OED, internet
(19)
Attested combinations (BNC, CELEX, OED, internet; see appendix for examples)
Attested two-suffix combinations are indicated by ‘yes’, the first suffix of each
combination is given in the first column, the second suffix in the first row. Blank cells are
unattested combinations.
th
th
en
er
ling
ee
ess
ly
dom
hood
ship
ish
less
ful (adj)
ness
ful (n)
-
en
er
ling
ee
ess
ly dom hood ship ish less ful (a) ness ful(n)
yes
-
yes yes
yes
-
yes
yes
yes yes yes yes
-
yes
yes
yes
yes
yes yes yes
yes
yes
-
yes yes yes
-
yes
yes
yes
yes
yes
yes
-
yes
-
yes yes
-
yes
-
yes
-
yes
-
Hay & Plag: Suffix combinations, grammatical restrictions, and parsing
6
(20)
th
Bidimensional hierarchy of suffixes
Suffixes at the beginning of arrows immediately
precede suffixes at the end of arrows in attested
combinations (BNC, OED, CELEX, Internet)
en
er
ful (N)
ee
ling
ly
ish
ess
dom
hood
less
ness
ship
ful (adj)
International Morphology Meeting, Budapest, May 9-12, 2002
7
(21)
Conclusions from the above:
- Only 24 combinations out of 210 logically possible ones are attested.
- There is a clear hierarchy of suffixes, hypothesis 1 is never violated.
- It is unclear, why such a hierarchy should exist in the first place (if not for the
reasons that underlie hypothesis 1).
- The position of certain suffixes with respect to each other is not always
determinable, since not all suffixes interact.
(22)
RE
What rules out the other combinations (processing or grammar)?
HYPOTHESIS 1
(23)
If processing is responsible, we would expect the hierarchy to reflect the order of
suffixes established on the basis of parsing ratios and productivity (for the latter see
Hay and Baayen, in press)
NB: The following graphs rank the affixes as far as possible on the basis of the monodimensional hierarchy (19), so some are equally ranked (e.g. -ling and -ee, -hood and -ship),
because they are indeterminate with respect to one another. Affixes which are
indeterminate with respect to more than one other affix are omitted (i.e. ful (N)).
Hay & Plag: Suffix combinations, grammatical restrictions, and parsing
(24)
Correlation of token-parsing ratio and hierarchical rank
8
9
(25)
International Morphology Meeting, Budapest, May 9-12, 2002
Correlation of type-parsing ratio and hierarchical rank
Hay & Plag: Suffix combinations, grammatical restrictions, and parsing
(26)
Correlation of productivity and hierarchical rank
10
International Morphology Meeting, Budapest, May 9-12, 2002
11
(27)
Correlations between hierarchy and boundary strength, based on the multidimensional hierarchy (19), chains with at least 5 suffixes:
(as measured in average ranks of parsing ratios and hapax-conditioned
productivity, Spearmans rank correlation)
suffix chain
coefficient
significance
th-en-ee-ess-hood-less-ness
rho = .79
p<.05
th-en-ee-ess-ship-ful-ness
rho = .75
p<.07
th-en-ee-ess-ship-less-ness
rho = .79
p<.05
th-en-ee-ess-dom
rho = 1
p<.02
th-en-ee-ess-ly-ish-ness
rho=.60
p<.16
th-en-ee-ess-ly-hood-less-ness
rho=.67
p<.07
th-en-er-ess-dom
rho=.82
p<.08
th-en-er-ly-ish-ness
rho=.83
p<.05
th-en-er-ess-ly-hood-less-ness
rho = .69
p<.06
th-en-er-ess-ly-ish-ness
rho = .61
p<.16
th-en-er-ling-ship-less-ness
rho = .79
p<.05
th-en-er-ling-ship-ful-ness
rho = .75
p<.07
th-en-er-ling-hood-less-ness
rho= .79
p<.05
th-en-er-ling-hood-less-ness
rho = .79
p<.05
⇒ the hierarchy is largely constrained by processing
RE HYPOTHESIS 2
(28)
If grammatical restrictions are responsible, we would predict that most or all blank
cells are the consequence of a grammatical restriction.
Hay & Plag: Suffix combinations, grammatical restrictions, and parsing
(29)
12
Table of structurally possible combinations (caps in bold print indicate possible, but
unattested combinations, question marks indicate unclear restrictions)
th
th
en
er
-
en
er
ling
-
yes
-
yes
yes
ee
?
ess
?
hood
ly dom hood ship ish less ful (a) ness ful(n)
yes yes
-
dom
ess
yes
ling
ly
ee
-
yes yes yes yes
yes yes yes
YES YES yes
yes YES yes
yes ?
-
YES yes
yes ?
YES
yes yes yes
yes ?
yes
-
yes
yes
yes
-
yes
yes
-
ship
ish
yes
yes
-
yes yes
-
less
ful (adj)
ness
ful (n)
yes
-
yes
-
YES ?
yes
-
- Selectional restrictions rule out the vast majority of combinations
⇒ the hierarchy is largely constrained by selectional restrictions
But:
Why should selectional restrictions result in an implicational hierarchy in the first place?
⇒ processing constraints and grammatical constraints work hand in hand, the grammar
is shaped in such a way that it tends to create structures that are better processable (cf.
Hawkins 1994 for a similar finding on word order)
International Morphology Meeting, Budapest, May 9-12, 2002
13
5. REMAINING PROBLEMS
(30)
Problematic combinations in the data set: e. g. -ness and -less
(31)
Beyond the data set (see also Plag, in press)
Some suffixes (e.g. -ize, -ation, -al) seem to systematically violate hypothesis 1 and
occur inside and outside of each other.
5. CONCLUSION
- The empirical investigation of possible combinations of 15 suffixes has shown that the
predictions of CBO are largely born out by the facts.
- Suffixes can be ordered on a hierarchy of boundary strength.
- Selectional restrictions largely coincide with parsing restrictions.
- Both selectional restrictions and parsing constraints are instrumental in determining
possible and impossible suffix combinations.
- Only combinations that are well processable are possible combinations, and this range of
possible combinations is further curtailed by selectional restrictions.
- Grammar is shaped in such a way that it tends to create structures that are better
processable.
6. REFERENCES
Allen, Margaret. 1979. Morphological investigations. Ph. D. dissertation. University of Connecticut. Ann Arbor:
University Microfilms.
Aronoff, Mark, and Nanna Fuhrhop. In press. Restricting suffix combinations in English. Closing suffixes
and the monosuffix constraint. Natural Language and Linguistic Theory.
Baayen, R. Harald, Piepenbrock, Richard and Gulikers, Leon (1995). The CELEX lexical database (release 2) cdrom. Philadelphia, PA: Linguistic Data Consortium, University of Pennsylvania (Distributor).
Burzio, Luigi (1994). Principles of English stress. Cambridge University Press, Cambridge.
Fabb, Nigel. 1988. English suffixation is constrained only by selectional restrictions. Natural Language and
Linguistic Theory 6, 527-539.
Giegerich, Heinz J. 1999. Lexical strata in English. Morphological causes, phonological effects. Cambridge: CUP.
Hawkins, John. 1994. A performance theory of order and constituency. Cambridge: CUP.
Hay & Plag: Suffix combinations, grammatical restrictions, and parsing
14
Hay, Jennifer. 2000. Causes and consequences of word structure. Ph.D. dissertation. Northwestern University
dissertation.
Hay, Jennifer. 2001. Lexical Frequency in Morphology: Is Everything Relative? Linguistics 39.6, 1041–1070.
Hay, Jennifer. in press. From speech perception to morphology: Affix-ordering revisited. Language.
Hay, Jennifer and Baayen, R. Harald (in press). Parsing and productivity, in Booij, Geert, and van Marle,
Jaap, Yearbook of Morphology 2001. Kluwer Academic Publishers.
Kiparsky, Paul. 1982. Lexical morphology and phonology, in The Linguistic Society of Korea (ed.) Linguistics
in the Morning Calm, 1-91. Seoul: Hanshin Publishing Co.
Mohanan, Karuvannur P. 1986. The theory of lexical phonology. Dordrecht: Reidel.
OED2 on CD. 1994. The Oxford English dictionary. 2nd Edition, on Compact Disc. Oxford: Oxford University
Press.
Plag, Ingo. 1996. Selectional restrictions in English suffixation revisited. A reply to Fabb (1988). Linguistics
34.4, 769-798.
Plag, Ingo. 1999. Morphological productivity. Structural constraints in English derivation. Berlin/New York:
Mouton de Gruyter.
Plag, Ingo. in press. The role of selectional restrictions, phonotactics and parsing in constraining suffix
ordering in English, in Geert Booij and Jaap van Marle (eds.), Yearbook of Morphology 2001. Dordrecht:
Foris.
Selkirk, Elisabeth. 1982. The syntax of words. Cambridge: MIT Press.
Siegel, Dorothy. 1974. Topics in English morphology. Ph. D. thesis. MIT.
7. APPENDIX
Attested two-affix combinations (from OED, if not indicated otherwise)
lenghthen
depthless
flattener (internet)
flattenee (internet)
preacherling
breweress
loverly
printerdom
loverhood
controllership
robberish
leaderless
tumblerful (BNC)
saplinghood
ducklingship
seedlingless
refugee-ess (internet)
employeehood
assigneeship
princessly (internet)
princessdom
priestesshood
governessship
governessless
knightlyhood
wollyish
kingdomless
kingdomful
courtliness
childhoodless (internet)
censorshipless (internet)
kinshipful
amateurishness
aimlessness
carefulness
AUTHORS’ ADDRESSES:
Jennifer Hay, Department of Linguistics, University of Canterbury, Private Bag 4800 Christchurch, New
Zealand, [email protected]
Ingo Plag, English Linguistics, Fachbereich 3, Universität Siegen, Adolf-Reichwein-Str. 2, D-57068 Siegen,
[email protected]