View
217
Download
1
Category
Tags:
Preview:
Citation preview
How do Hebrew speakers ‘trick or treat’?An experimental study of verb formation from Hebrew CCVC nouns
Roy Becker-Kristal, UCLA Linguistics
roybecker@humnet.ucla.edu
Introduction• This experimental study tests how Hebrew speakers coin
novel denominative verbs from known CCVC nouns.• Beyond a mere description of a language-specific state
of affairs, this study puts to test various morpho-phonological theories.
• Particular attention is given to the experimental methodology, since comparable studies are rare in the morpho-phonological literature.
BackgroundHebrew verb conjugations (binyanim)• A small close set of CV template paradigms. Any
Hebrew verb must pertain to one such template paradigm or binyan.
• There are five binyanim, three of which, namely katal, hiktil and kitel can have any argument structure (niktal and hitkatel are limited to intransitives). katal and hiktil are skeletally rigid, while kitel is more ‘plastic’.
• The (sub-)binyanim are named after their Past-3rd-Sg-Msc template (e.g. katal=/C1a.’C2aC3/)*.
* The binyanim are more commonly named using the consonants p/f-’-l, e.g. pa’al, pi’el, hif’il etc. The use of k-t-l, which is common in the noun template system, is more transparent, and hence used here.
BackgroundThe binyanim: katal• The default binyan in Ancient Hebrew.• Huge lexical count.• Rather non-productive in Modern Hebrew.• Highly diverse base template paradigm (six-to-nine very
different templates), e.g. /lis.’tom/ to seal:
/sa.’tam/ Past3SgMsc /sat.’mu/ Past3Pl
/so.’tem/ PresSgMsc /sot.’mim/ PresPlMsc
/jis.’tom/ Fut3SgMsc /jis.te.’mu/ Fut3Pl
(Prescriptively also: /stam.’tem/ Past2PlMsc, /sit.’mu/ Imp2Pl)
BackgroundThe binyanim: hiktil• Strongly associated with causation, but can
accommodate any argument structure.• Large lexical count.• Productive mostly in secondary derivation (of causatives
from existing non-causative verbs).• Very stable base template paradigm (only two-three
rather similar templates). e.g. /le.hak.’lit/ to record:/hik.’lit/ Past3SgMsc/hik.’lat.nu/ Past1Pl/mak.’lit/ PresSgMsc /mak.li.’tim/ PresPlMsc/jak.’lit/ Fut3SgMsc /jak.’li.tu/ Fut3Pl(Prescriptively also: /hak.’let/ Imp2Sg)
BackgroundThe binyanim: kitel (canonical C1VC2VC3 form)
• The default binyan in Modern Hebrew.• Huge lexical count.• Highly productive.• Moderately stable base template paradigm (five rather
similar templates), e.g. /le.ka.’bel/ to receive:
/ki.’bel/ Past3SgMsc /kib.’lu/ Past3Pl
/ki.’bal.nu/ Past1Pl
/me.ka.’bel/ PresSgMsc /me.kab.’lim/ PresPlMsc
/je.ka.’bel/ Fut3SgMsc /je.kab.’lu/ Fut3Pl
BackgroundThe binyanim: kitel (non-canonical C0VC0VC0 forms)
• kitel can accommodate clusters in its consonantal slots:CCC-initial: /stip.’tez/ he stripteasedCCCC-medial: /tins.’keb/ he transcribed*CC-final: /sig.’ment/ he segmented*
• Moderate lexical count, but highly productive, especially when accommodating loanwords.
• Stable base template paradigm (three templates: C0iC0eC0, C0iC0aC0, C0aC0eC0).
• CVCCVC kitel verbs (‘kirtel’) are frequent. CCVCCVC kitel verbs (‘kristel’) are also attested. Peculiarly, CCVCVC kitel verbs (‘kritel’) are extremely rare.
*Speech-technology jargon. Linguistic circles use different terms.
BackgroundThe binyanim: kitel sub-binyanim• The plasticity of kitel is exploited to accommodate bases
containing affixes, creating several sub-binyanim. • These sub-binyanim host many denominative verbs from
affix-/reduplication-derived CVCCVC nouns. • Prefixed sub-binyanim: iktel, miktel, tiktel, iktel• Suffixed sub-binyanim: kitlen• Reduplication sub-binyanim: kitlel, ktolel.• All these sub-binyanim have moderate-to-low lexical
count, but kitlel is extremely productive and iktel has also been becoming fashionable.
BackgroundNouns and denominative verbs in Hebrew• Nouns (and adjectives) enjoy much greater structural
flexibility in Hebrew (and other Semitic languages). There are many more CV-templates, and templates are not at all obligatory.
• Therefore, the derivation of a denominative verb involves phonological mapping of free structures onto rigid frames, with inevitable faithfulness violations. Examples:excessive truncation:/kon.fi.gu.’ats.ja/ configuration → /kin.’feg/ he configuredfull reduplication:/’zap/ zap → /zip.’zep/ he zapped
BackgroundDenominative verbs in Hebrew (Cont.)• For some structures it is in fact impossible to derive a
denominative verb, e.g.:CV: /’tsi/ fleet → ???CVCVCC: /ko.’mand/ command (in linguistics) → ???
• Yet, in most cases, not only is it possible to derive a denominative verb, but native Hebrew speakers also have clear intuitions how to derive it. Hypothetical examples:CVCCVC: /’bun.ke/ bunker → /bin.’ke/ he bunkeredCVCC: /’ift/ shift → /if.’tet/ he pressed the ‘shift’ key
• The overwhelming majority of newly coined denominative verbs in Hebrew are in either the ‘canonical’ (C1i.’C2eC3) or in the C1VC2.’C3VC4 variants of kitel.
BackgroundCCVC nouns and denominative verbs• CCVC nouns are tri-consonantal, therefore all possible
binyanim are relevant for verb derivation.• The template paradigms of the most productive sub-
binyanim (canoincal kitel and kitlel) are the least structurally faithful: /C1C2VC3/ rendition of the base never occurs in their paradigm, whereas it does occur in katal, hiktil and various less productive variants of kitel such as iktel or ktolel.
• Thus there is a conflict between the morpho-lexical requirement to use the default operation and the phonological requirement to maximize source-derivative faithfulness.
BackgroundInformal systematic survey of CCVC nouns and
denominative verbs in the Hebrew lexicon:• 300-350 CCVC nouns (and adjectives).• 39 unambiguous cases of denominative verbs derived
from CCVC nouns: 10 in katal, 19 in hiktil, six in kitel, three in kitlel, four in other kitel variants.
• Although these numbers are not reliable for quantitative analysis, the morpho-lexical vs. phonological conflict is apparent: The otherwise preferred strategies kitel and kitlel fare much worse that the template-faithful hiktil.
BackgroundLexical survey (cont.):• Apparent ‘island of reliability’ in CCiC: out of 12 CCiC-
sourced verbs, nine (75%) are in hiktil.• Balanced variation between strategies elsewhere.• Huge variation in CCoC-sourced verbs: 10 verbs, five
different strategies:katal: e.g. /xa.’ap/ he napped, from /’xop/ naphiktil: e.g. /his.’mil/ he turned left, from /’smol/ leftkitel: e.g. /bi.’lef/ he bluffed, from /’blof/ bluffkitlel: /bil.’geg/ he blogged, from /’blog/ blogktolel: /no.’e/ he scrounged, from /’no/ scrounge
Theoretical ApproachesThe ‘Semitologist’ Approach*:• In Semitic languages, lexical entries include the
consonantal root as a component in their representation.• Derivation of the phonological contents of a novel word
requires the extraction of a consonantal root, the choice of a template, and the combination of the two together.
• Since derivation goes through consonantal root extraction, vocalic contents and prosodic structure of the source of the derivation should play only a marginal role.
• Verb derivation strategies should not differ significantly between CVCVC-sourced and CCVC-sourced nouns, nor between CCVC nouns containing different vowels.
*Berent et al. (1997,2001,2002) clearly manifest this approach, but might not necessarily subscribe to the current line of argumentation.
Theoretical ApproachesThe ‘Universalist Phonological’ Approach*:• Semitic languages adhere to the same phonological
constraints and representations as other languages.• Lexical entries do not contain the consonantal root as an
independent representation component.• Derivation of the phonological contents of a novel word is
achieved by direct source-derivative mapping, while adhering to faithfulness and markedness constraints.
• Direct mapping and source faithfulness imply that, in the derivation of denominative verbs, the skeletal structure and vowel quality of the noun source should play a significant role in determining the derivation strategy.
*See e.g. Bat-El (1994a, 1994b).
Theoretical ApproachesThe ‘Lexicalist’ Approach*:• Morphological strategies are determined by the speaker’s
knowledge of paradigm distribution in the existing lexicon.• When choosing a morphological strategy, the speaker
applies arbitrary statistical knowledge on phonological structures, rather than direct phonological mapping.
• The choice of derivation strategies of novel CCVC-sourced denominative verbs should match the distribution of strategies among the existing 39 CCVC-sourced verbs.
• Hypothetical strategies absent from the Hebrew lexicon should be avoided even if they are phonologically optimal.
*This inflection-based approach, advocated in e.g. Albright(2002), Albright&Hayes(2002), might not extend to derivational morphology.
The ExperimentExperiments on novel word derivation• No standard format (unlike inflectional ‘wug’-tests).• The morphological task is somewhat unnatural (most
people do not invent words on a daily basis).• The task tends to be open-ended, and is seldom confined
to a small n-ary choice.• Many times no derivational strategy is appropriate – a
problem for a strict forced-choice task.• High risk of over-creativity on one hand, and strategy drift
on the other.
The Experiment - GuidelinesHow to do an experiment on derivation?• Ensure that the subjects are in the right state of mind!• Obviate the task by using real word stimuli (not ‘wugs’)
and straightforward meanings for the derivatives!• Keep the task simple! Cover most relevant strategies and
have the subjects choose, rather than derive themselves.• Don’t force the subjects to choose – leave a default
alternative with no derivation available!• Instruct subjects to prefer naturalness over creativity!• Keep the experiment short, but don’t forget to use fillers!• If the experiment is phonological, do not rely solely on
written form!
The Experiment - Subjects• To date, 24 subjects have participated (a few others were
recruited but dropped out in the middle).• Native Hebrew speakers, aged 22-36, educated
(undergrads to post-docs).• Homogeneous and representative of General Israeli
Hebrew (mostly second generation native speakers).• Living in Israel at present or have recently relocated from
Israel to Los-Angeles.• No thorough background in theoretical linguistics.• Neither academic nor professional experience in applying
prescriptivist norms.
The Experiment - Stimuli• 63 nouns, ranging from core vocabulary, through slang
and loanwords to English words used in code-mixing.• 48 CCVC stimuli, comprising of 16 CCiC stimuli and 8
stimuli for each of CCeC, CCaC, CCoC and CCuC.• 15 control/filler stimuli: 5 stimuli for each of CV.’CVC,
VC.’CVC and CCV.’CVC, all perfectly balanced for vowel qualities in both stressed and unstressed syllables.
• Stimuli never had existing semantically and morpho-phonologically related verbs (denominative or not).
• Stimuli were phonologically compatible with all of the available strategies (not yielding phonotactic violations).
• Different stimuli never shared the same consonantal root.
The Experiment - Task• Given a noun stimulus, a related predicative meaning,
and a sentence illustrating the use of the predicate, the subject had to choose the most natural denominative verb among eight derivational options (in Past-3rd-Sg-Msc).
• For CCVC stimuli, these options included the seven (sub-) binyanim katal, hiktil, kitel, kitlel, iktel, ktolel (lexical count: 1), ktilel (fictive, lexical count: 0).
• For the control conditions there were also seven options, but these varied (since, for some stimuli, certain tri-consonantal options had to be replaced by dummies).
• The eighth option was always a paraphrastic rendition of the predicate with no derivation. The subjects were instructed to choose this default option only as last resort.
The Experiment - Procedure• For each experiment session, the subject receives by e-
mail a PowerPoint presentation, an answer sheet and detailed installation instructions.
• The subject is instructed to ‘condergo’ the session while sitting alone in front of the computer.
• The presentation begins with a multi-media introduction to denominative verb derivation, illustrating both real verbs and (obvious) intuitions about some hypothetical verbs.
• The introduction states explicitly that the experiment tests how Hebrew speakers derive verbs from nouns. However, the phonological aspects studied are not disclosed.
• The subject is explicitly instructed to prefer naturalness to creativity.
The Experiment – Presentation• The 63 stimuli were divided to three sessions of 21 stimuli
in each (16 CCVC and 5 of the other conditions).• The order of the 5 non-CCVC stimuli was randomized.• The order of the 16 CCVC stimuli was randomized, and
they were divided to four blocks of four stimuli. These four blocks were interlaced between the non-CCVC stimuli.
• For each stimulus, the order of the seven non-default derivational strategies was randomized.
• Graphemic as well as carefully-controlled, naturally-sounding, recorded auditory renditions were created for both the stimuli and all the derivational options.
• Each stimulus was presented, with its derivational options, on a separate slide in a PowerPoint presentation.
[shvil]
שיבל .7{/i.vel/ (kitel)}
שבל .4{/a.val/ (katal)}
שבילל .1{/vi.lel/ (ktilel)}
שבולל .5{/vo.lel/ (ktolel)}
השביל .2{/hi.vil/ (hiktil)}
הוליך בשביל .9{led through the path}
שיבלל .6{/iv.lel/ (kitlel)}
אישבל .3{/i.vel/ (iktel)}
דרךצרהואולאסלולה{latinization} {// path (a narrow and/or unpaved road)}
בשביללהוליךהפועלמשמעות{Verb meaning: “to lead through a path”.}
דוגמת שימוש: המדריך ________ את החניכים.{Usage example: The guide ____ the boy-scouts.}
האזינו לכל אפשרויות הגזירה וסמנו בדף את זו שנראית לכם כמתאימה ביותר{Listen to all derivations and mark the most appropriate one on the questionaire}
The Experiment – ExampleFollowing is an example of a stimulus presentation slide (English glosses in curled brackets are later addition):
The Experiment – Special FocusSpecial focus on hiktil in CCVC stimuli:• The number of CCiC stimuli is twice that of other vowels.• Two binary non-phonological parameters are controlled:
a) transitivity: whether or not the predicate takes a direct object (strict causativity was avoided, because of the risk of a ceiling effect).b) homophony: whether or not the Hebrew lexicon already has a hiktil verb with these three consonants.
• Each combination of {vowel, ±trans, ±hom} is represented by two stimuli (four stimuli for CCiC).
• Homophony in other binyanim: For the 48 stimuli there were 13 potential katal homophones, 11 kitel, 5 kitlel and 2 potential iktel homophones.
CV.’CVC Results• Hegemony of kitel (and
kitlel) as expected.• Somewhat surprising
viability of katal, maybe due to /CV.’CVC/ skeletal faithfulness.
• Negligible use of hiktil (and iktel).
• Results are in line with all three theoretical approaches.
derivation strategies for CV.'CVC (n=110 out of 120 )
iktel4.5%
kitlel19.1%
kitel60.0%
hiktil5.5%
katal10.9%
katal hiktil kitel kitlel ktilel ktolel iktel
CV.’CVC Results• Hegemony of iktel.• Mostly in line with the
phonological approach.• The ‘Semitologist’ claim
that the nouns contain an underlying // cannot be ruled out.
iktel is also in line with the lexical distribution approach, but ktilel is not (yet it’s marginal).
derivation strategies for VC.'CVC(n=101 out of 120)
katal2.0%
kitlel2.0%
kitel2.0%
ktolel2.0%
iktel80.2%
hiktil6.9%
ktilel5.0%
katal hiktil kitel kitlel ktilel ktolel iktel
CCV.’CVC Results• CCV.’CVC stimuli do not
constitute a problem for derivation.
• Equally co-dominated by 4-consonant kitel variants CVC.’CVC (skeletally unfaithful, common in the lexicon) and CCV.’CVC (skeletally faithful but non-existent in the lexicon).
• Reduplication is superfluous and appears only marginally.
derivation strategies for CCV.'CVC (n=115 out of 120)
krotlel1.7%
kirtel41.7%
krotel7.0%
kritlel7.0%
kritel39.1%
kirtlel3.5%
kirtel kirtlel kritel kritlel krotel krotlel
CCVC Results• Pluralism of strategies.• Marginal use of katal.• Substantial increase in
the popularity of hiktil, ktilel and ktolel, mostly at the expense of kitel.
• Both the phonological and lexicalist approaches can explain the popularity of hiktil, but ktilel and ktolel can be explained only by the phonological approach.
derivation strategies for CCVC(n=1059 out of 1152)
katal5.9%
iktel7.1%
ktilel14.6%
kitlel18.1%
kitel14.2%
hiktil28.2%
ktolel11.8%
katal hiktil kitel kitlel ktilel ktolel iktel
CCeC Results
Relative to CV.’CVC:• CCeC is probably the
‘default’ CCVC source.• High popularity of hiktil
and ktilel (/i/-faithful).• Slightly higher popularity
of iktel (/e/-faithful, but root-prefixed). Might be non-significant.
• Significant decrease in popularity of kitel, but remains viable.
• Stability of kitlel.
derivation strategies for CCeC (n=176 out of 192)
katal5.1%
iktel9.1%
ktilel21.0%
kitlel19.9%
kitel18.2%
hiktil22.2%
ktolel4.5%
katal hiktil kitel kitlel ktilel ktolel iktel
CCaC Results
Relative to CCeC:• Higher popularity of katal,
tightly related to the /a/-melody. Surprising given the unfaithful skeletal structure (katal is CCVC-faithful in future tense, where the vowel is /o/), and the non-productive status of katal.
• Decrease in popularity of hiktil and ktilel.
• Like CCeC otherwise.
derivation strategies for CCaC (n=171 out of 192)
iktel9.9%
katal13.5%
ktilel15.2%
kitel15.8%
hiktil15.8%
kitlel23.4%
ktolel6.4%
katal hiktil kitel kitlel ktilel ktolel iktel
CCiC Results
Relative to CCeC: • hiktil rules (/i/-faithful,
Island of Reliability).• hiktil dominance explains
decrease in katal, kitel, kitlel, ktolel and iktel.
• Stability of ktilel (also /i/-faithful). Unclear whether stability is genuine or an artifact of counter-balance between /i/-faithfulness (positive) and dominance of hiktil (negative).
derivation strategies for CCiC (n=368 out of 384)
ktolel1.9%
katal3.0%
iktel4.9%
ktilel20.4%
kitlel14.4%
kitel10.1%
hiktil45.4%
katal hiktil kitel kitlel ktilel ktolel iktel
CCoC Results
Relative to CCeC:• ktolel (/o/-faithful) rules.• ktolel dominance explains
the decrease in kitel, kitlel,iktel and hiktil.
• Significant drop for ktilel, perhaps also for hiktil.
• Probably higher popularity of katal, explainable by the future tense (/Ciktol/) or by lexical statistics. Might be non-significant.
derivation strategies for CCoC (n=177 out of 192)
ktilel5.1%
iktel6.2%
katal6.8%
hiktil13.6%
kitel14.7%
kitlel16.4%
ktolel37.3%
katal hiktil kitel kitlel ktilel ktolel iktel
CCUC Results
Relative to CCeC:• Similar to CCeC, except
for the ‘mirror image’ in ktolel vs. ktilel.
• Back-rounded faithfulness explains ktolel popularity, while height difference underlies its more limited span than in CCoC.
derivation strategies for CCuC (n=167 out of 192)
ktilel4.8%
katal4.8%
iktel7.8%
kitlel21.0%
kitel16.8%
hiktil25.1%
ktolel19.8%
katal hiktil kitel kitlel ktilel ktolel iktel
Transitivity Results• Transitivity facilitates
derivation: 96% vs. 89% non-default responses.
• Transitivity significantly increases hiktil (and also kitel) responses.
• This increase is probably responsible for the decrease in katal, kitlel and ktilel responses (and not vice-versa).
derivations for transitive (n=549) and intransitive (n=510) predicates
0
5
10
15
20
25
30
35
katal hiktil kitel kitlel ktilel ktolel iktel
% o
f n
on
-def
ault
res
po
nse
s
intransitive transitive
Homophony Results• Results are based on
responses for 24 (vs. 24) stimuli for hiktil, 13 (vs. 35) for katal and 11 (vs. 37) for kitel.
• Quite predictably, potential homophony impedes derivation.
The effect of homophony onhiktil, katal and kitel
0
5
10
15
20
25
30
35
hiktil katal kitel
% o
f no
n-d
efau
lt re
spo
nse
s
no homophony homophony
Analysis - Stability• ktolel, hiktil, ktilel and
katal are very sensitive to the source vowel.
• Conversely, ktilel, kitel and iktel are stable across different vowels.
• This suggests (at least) two different derivation mechanisms. Notice in particular the difference between iktel and the four sensitive strategies.
Stability across stimulus vowels
0
10
20
30
40
50
katal hiktil kitel kitlel ktilel ktolel iktel
% o
f non
-def
ault
resp
onse
s pe
r vo
wel
Analysis – hiktil vs. ktilel
hiktil ktilel h/k
CC{eaou}C![+trans-hom]
71/507(14.0%)
63/507(12.4%)
1.13
CC{eaou}C[+trans–hom]
61/184(33.2%)
17/184(9.2%)
3.59
CCiC![+trans–hom]
113/276(40.9%)
67/276(24.3%)
1.69
CCiC[+trans–hom]
54/92(58.7%)
8/92(8.7%)
6.75*
• Phonologically, hiktil and ktilel are equivalent (/CCiC/ faithful). Why is there hiktil bias?
a) morpho-syntactic and lexical knowledge.
b) hiktil as island of reliability for CCiC.
• When both factors are combined hiktil is as popular as kitel for CV.’CVC nouns!!!
* 1.13x3.59x1.69=6.85≈6.75 Where is the catch …??? ;-)
Analysis – ktilel vs. ktolel• For non-low vowel CCVC stimuli, ktilel and ktolel mirror
each other: Strong bias towards the fully faithful vowel in CCiC and CCoC, and slightly weaker bias towards the backness&rounding faithful vowel in CCeC and CCuC.
CCiC CCuC
ktilel ktolel i/(i+o) ktilel ktolel i/(i+o)
20.4% 1.9% 91.4% 4.8% 19.8% 19.5%
CCeC CCoC
ktilel ktolel i/(i+o) ktilel ktolel i/(i+o)
21% 4.5% 82.2% 5.1% 37.3% 12.0%
CCaC
ktilel ktolel i/(i+o)
15.2% 6.4% 70.3%
Analysis – hiktil vs. ktolel• This mirror imaging is only partially replicated for hiktil vs.
ktolel in non-low vowel ![+trans –hom] CCVC stimuli. The CCiC~CCoC asymmetry is predicted by the /i/-hiktil island of reliability, but CCeC~CCuC asymmetry is unexplained.
CCiC CCuC
hiktil ktolel i/(i+o) hiktil ktolel i/(i+o)
40.9% 2.2% 95.0% 12.3% 22.1% 35.7%
CCeC CCoC
hiktil ktolel i/(i+o) hiktil ktolel i/(i+o)
20.9% 6.2% 77.1% 9.2% 39.2% 19.0%
CCaC
hiktil ktolel i/(i+o)
13.5% 7.1% 65.4%
Tentative Conclusions• The experiment elicited derivations successfully: only 8%
of no-derivation answers, equally distributed between all conditions; only marginal over-creativity; the results are always interpretable by at least one theoretical approach.
• The results for CV.’CVC and VC.’CVC stimuli are predictable by all approaches. As such they are not too informative, but reinforce the experimental design.
• The popularity of kritel for CCV.’CVC nouns, of katal for CCaC, of ktolel for CCoC and CCuC, and of ktilel for CCiC and CCeC (and CCaC) can only be explained by the universalist phonological approach: In choosing these strategies the speakers neither extract a consonantal root nor consult the existing lexicon.
Tentative Conclusions• The phonological approach also explains hiktil~ktilel
equivalence for CC{eaou}C ![+trans –hom] stimuli and the ktolel~ktilel mirror image for CC{ieou}C stimuli.
• The differences between the phonologically equivalent hiktil and ktilel elsewhere are best explained by morpho-syntactic and lexical knowledge, and demonstrate the (secondary) role of lexical statistics in derivation.
• The stability of kitel, kitlel and iktel across all CCVC conditions is best explained by the root-extraction and/or the lexical distribution approaches. The account by the phonological approach, relying on /CV…VC/ as the optimal base (single-consonant edges) requires more ‘hand-waving’, in particular for iktel.
Selected ReferencesAlbright, A. (2002). ‘Islands of reliability for regular morphology: evidence from Italian.’
Language 78, 684-709.
Albright, A. and Hayes, B. (2002). ‘Modeling English past tense intuitions with minimal generalizations’. In Maxwell, M. (ed): Proceedings of the 6th Meeting of the ACL Special Interest Group in Computational Phonology. Philadelphia, July 2002. ACL.
Bat-El, O. (1994a). ‘System modification and cluster transfer in Modern Hebrew.’ Natural Language and Linguistic Theory 12, 571-593.
Bat-El, O. (1994b). ‘Resolving prosodic mismatch in Modern Hebrew verb formation.’ In van der Hulst and van de Weijer (eds): Leiden in Last: HIL Phonology Papers I. The Hague: Holland Academic Graphics. 25-40.
Berent, I., and Shimron, J. (1997). ‘The representation of Hebrew words: Evidence from the Obligatory Contour Principle.’ Cognition 64, 39-72.
Berent, I., Everett, D. L., and Shimron, J. (2001). ‘Do phonological representations specify variables? Evidence from the Obligatory Contour Principle.’ Cognitive Psychology 42(1), 1-60.
Berent, I., Marcus, G., Shimron, J. and Gafos, A. (2002). ‘The scope of linguistic generalizations: evidence from Hebrew word formantion.’ Cognition 83, 113-139.
Recommended