Plurals Arabic in L2 Acquisition
1
Introduction
Arabic pluralization is a morphological system strikingly different from English What is the learning path for such a system? To what extent does L2 acquisition of it parallel L1? Not very much What accounts for the difference? initial state? nature of the input? learning mechanism?
Mary Ann Walter
Dept. of Linguistics and Philosophy, MIT
GALANA 2004 December 17-20, 2004 University of Hawai'i at Manoa
7
Summary
L2 learners engage in morphological decomposition Apparent advantage of feminines is due to their hypothesis that final a always indicates feminine and that no feminine stem lacks it Recall that it`s a better cue for feminine in L2 input than in L1 input (75 vs 60%) In its absence, a broken plural is used as a default Broken outputs mostly fall into the language's existing classes, and eventually reflect the input distribution
4
Results - correct
Mean % correct by age and stem type
100 90 80 70 60 50 40 30 20 10 0 2 3 age (years) 4 5 100 90 80 70 60 50 40 30 20 10 0 1 2 group 3 4
Mean % correct by group and stem type
2
% correct
m f b
% correct
m f b f-a f-no-a
8
Whence the L1/L2 Difference?
Nature of the input? Frequency: Broken forms are even more predominant in L2 input than in the rest of the language could explain their choice as default But most common broken type still <80% of feminine count for all groups
100% 90% plural type percentage 80% 70% 60% 50% 40% 30% 20% 10% 0% 1 2 Group 3 4 masc fem broken
Plural morphology
English morphological inflection typically consists of an overwhelmingly frequent regular pattern and a few irregular exceptions (Daugherty and Seidenberg 1992) Arabic nominal plurals fall into three classes: regular w/masculine suffix, regular w/feminine suffix, and irregular or "broken" w/internal stem modification (Wright 1996) Broken plurals are subdivided into 30-odd types (5 most frequent below) Masculine: Feminine: Broken: Singular mutarjim Tabbaax `akl sayyaara jariida jism qalb rajul kitaab Plural mutarjim-iin Tabbaax-iin `akl-aat sayyaar-aat jaraa'id ajsaam quluub rijaal kutub Gloss translator cook food car newspaper body heart man book Template
L1
Feminines rapidly reach ceiling Masculines and brokens still not completely acquired at 5 years Semantic/phonological coherence in a class does not aid learning
L2
No overall effect for group level Better performance on feminines But cf separation between fem w/final a and without (dotted lines)
waraa'id awraad wuruud wiraad wurud
5
Results error analysis
Mean % errors by age and type
100 90 80 70 60 50 40 30 20 10 0 2 3 age (years) 4 5 100 90 80 70 60 50 40 30 20 10 0 1 2 group 3 4
Variability: Is it still greatest among feminines in L2 input? To be determined... Initial state?
m>b f>b b>b m/b>f m/b>f2
Mean % errors by group and type
% errors
f>b b>b m/b>f
Final a can indicate feminine, unithood, or nothing at all ~75% of feminine stems have it (in L2 input; 60% in native input) ~20% of broken stems also have it (constant across input types)
(type counts from textbook glossaries (Thackston 1996, Brustad et al. 1995, 1997) and Kouloughli 1992)
% errors
m>b
Neither cross-linguistic naturalness nor overwhelming suffixation in English leads to its overgeneralization in Arabic, even though this is what Arabic L1 learners do So learners do not seem to straightforwardly transfer expectations about frequencies from L1 to L2 Another test: relativizer use in L2 Arabic. Forthcoming! But frequency might still be driving their behavior indirectly: Rarity of broken forms in English makes the Arabic ones highly salient Learners soon notice their frequency in Arabic relative to English This results in overextension through hypercorrection-like a process
Stems taking masculine plurals are limited to human males not all human male stems take masculine plurals Feminine stems exhibit most phonological variability, and are largest single class if brokens are decomposed (Boudelaa and Gaskell 2002) Regular stems NOT a majority: (B&G 2002)
m f
L1
m>b: masculine stem gets broken plural f>b: feminine stem gets broken plural b>b: broken stem gets incorrect broken pl m/b>f: masc/broken stem gets fem pl m/b>f2: above w/out broken a stems Tabbaax `akl jariida jism
L2
*Tabaa'ix *'ukuul *ajraad *jismaat
Distribution of brokens: (K 1992)
waraa'id awraad wuruud wiraad wurud other
L1 errors overwhelmingly involve overapplication of feminine suffix In L2 feminine overapplication is relatively rare Virtually disappears without broken stems w/final a (dotted line) Feminine stems still advantaged because of majority w/final a Errors with broken outputs predominate
b
3
Method
L1 data: from Ravid and Farah (1999); real-word picture-naming, 12 children per age group L2 data: 43 adult English-speaking subjects, none native speakers of Arabic or other Semitic language, grouped based on course membership: Group 1 (n=7): completed one-month intensive Arabic course Group 2 (n=15): at end of 1-year university Arabic course Group 3 (n=16): at end of 2nd year university Arabic course Group 4 (n=5): at end of 4th year university Arabic course, also some immersion experience Written questionnaire with the singular stem (in Arabic script) and English gloss of 42 Arabic words in random order, all of which were in their textbook glossaries Instructed to write the Arabic plural of each word, and to make a guess even if they were not sure Stimuli were divided into 3 classes: Masculine stems: 14 items Feminine stems: 14 items, 9 w/final a Broken stems: 14 items, 3 w/final -a
6
% e r r o r s % i n p u t
40 35 30 25 20 15 10 5 0
Results L2 broken plurals
Group 1 Group 2 Group 3
40 35 30 25 20 15 10 5 0
40 35 30 25 20 15 10 5 0
Group 4
9
wr on g oth er
Selected references
Boudelaa, S. and M.G. Gaskell. 2002. A re-examination of the default system for Arabic plurals. Language and Cognitive Processes 17, 321-343. Brustad, K., M. al-Batal, and A. al-Tonsi. 1995. A...