Honaida Ahyad Michael Becker Stony Brook University Annual

Honaida Ahyad & Michael Becker Stony Brook University Annual Symposium on Arabic Linguistics 2016 6/11/2021

Verb type Doubled verbs (short vowel) Hollow verbs (long vowel) Perfective sadd tamm faaz saal naam Imperfective ji-sudd ji-timm ji-fuuz ji-siil ji-naam # 109 24 53 48 3 “to block” “to agree” “to win” “to melt” “to sleep” 6/11/2021

The derivational approach (Brame 1970) The Max-Feature account (Rosenthall 2006) The bilateral templatic analysis (Chekayri & Scheer 2005) Perfective-Imperfective paradigms (Gafos 2003) 6/11/2021

Today, we will show that speakers learn the distribution of vowels in verbs and extend this vowel distribution to nonce words (e. g. Berent 2007) 6/11/2021 4

133 short vowel verbs (doubled) 101 long vowel verbs (hollow) A quarter contains an emphatic (pharyngealized). [t d s z] vs. [tˤ dˤ sˤ zˤ] Emphatics with short vowel verbs [u] Emphatics with long vowel verbs [ii] 5

Verb type emphatic %[u, uu] (pharyngealized) short vowel dˤal ~ ji-dˤil “to get lost” tˤaar ~ ji-tˤiir “to fly” sˤaam ~ ji-sˤuum “to fast” %[u, uu] (plain) 97% [u] tˤaχ ~ ji- tˤuχ “to shoot” long vowel non-emphatic 32% [uu] ʕal ~ ji-ʕil “to sicken” baχ ~ ji-buχ “to spray” ʕaad ~ ji-ʕiid “to repeat” 77% [u] 59% [uu] ʃaaf ~ ji-ʃuuf “to see” 6/11/2021

Hypothesis: The presence of emphatics and vowel length will determine the backness of the imperfective vowel. 6/11/2021

104 participants 86 females + 18 males Age: M=32, range 18 -50 Recruited through social media Native speakers of Urban Hijazi Arabic (Makkah, Jeddah, Madinah, Taif) 6/11/2021

A word list consisting of 60 nonce verbs, each in three forms: Perfective, always with [a, aa], e. g. [zˤaadˤ] Imperfective with a front vowel [i, ii], e. g. [ji-zˤiidˤ] Imperfective with a back vowel [u, uu], e. g. [ji-zˤuudˤ] 6/11/2021

Experigen (Becker & Levine 2014) Forced choice task Perfective Imperfective 1 Imperfective 2 6/11/2021

mixed effects logistic regression model in lme 4 with maximal random slopes for participants. positive β = more [u/uu] β (Intercept) 0. 29 0. 11 2. 65 <0. 001 Emphatic 0. 02 0. 19 0. 11 < 1 Vowel length 0. 21 0. 19 1. 12 <1 Emphatic: Vowel length SE – 1. 53 0. 38 Z value Pr(>|z|) – 4. 08 <0. 001 6/11/2021

6/11/2021

The Minimal Generalization Learner (Albright & Hayes 2002, 2003, 2006) was trained on the real words of Urban Hijazi Arabic, and tested it on the same nonce perfective verbs that were given to the participants. The model generated imperfective forms and confidence scores that mimicked the choices of the native speakers. 6/11/2021

For the short vowels, the MLG prediction is almost entirely [u], but still correlated with actual choices. 6/11/2021

Good match between the MGL predictions and the participant choices for the long vowels. 6/11/2021

The Sublexical Learner (Becker & Gouskova to appear, Allen & Becker 2015). Like the MGL, the model is trained on real words, tested on the same perfectives as the participants. The model generates imperfective forms with their predicted probability. Two sublexicons: V → [+high +front] V → [+high +back +round] Both sublexicons include short and long vowel verbs Constraint set: we used the environments that the MGL discovered. 6/11/2021

Good match to the participants' choices for short and long vowel verbs. 6/11/2021

6/11/2021

Speakers have implicit knowledge about the choice of vowel in the imperfective based on vowel length and the presence of an emphatic. Phonetic motivation: it's hard to see how emphatics can select more front vowels and more back vowels. We claim that speakers extend the distribution of vowels from their lexicon, as predicted by the Minimal Generalization Learner. 6/11/2021

Our analysis unifies the treatment of long and short vowel verbs (cf. separate treatment in Brame 1970). Our results lend support for the view that speakers of Semitic languages can learn trends in their lexicon at a level of representation in which both vowels and consonants are present, not purely consonantal roots (Berent et al. 2007). 6/11/2021

Christina Bethin, Ellen Broselow, Robert Hoberman 6/11/2021

Allen, Blake & Becker, Michael. 2015. Learning alternations from surface forms with sublexical phonology. lingbuzz/002503. Albright, Adam & Hayes, Bruce. 2002. Modeling English past tense intuitions with minimal generalization. In M. Maxwell (Ed. ) Proceedings of the Sixth Meeting of ACL Special Interest Group in Computational Phonology (pp. 58 -69). Philadelphia: ACL. Albright, Adam & Hayes, Bruce. 2003. Rules vs. analogy in English past tenses: A computational/ experimental study. Cognition 90: pp. 119 -161 Albright, Adam, & Hayes, Bruce. 2006. Modeling productivity with the gradual learning algorithm: The problem of accidentally exceptionless generalizations. In Gradience in grammar: Generative perspectives, eds. Gisbert Fanselow, Caroline Fery, Matthias Schlesewsky, and Ralf Vogel, 185– 204. Oxford: Oxford University Press. Becker, Michael & Gouskova, Maria. (To appear). Source-oriented generalizations as grammar inference in Russian vowel deletion. Linguistic Inquiry 47: 3. Becker, Michael & Levine, Jonathan. 2014 Experigen – an online experiment platform. Available at http: //becker. phonologist. org/experigen. Berent, Iris, Vaknin, Vered, & Marcus, Gary F. 2007. Roots, stems, and the universality of lexical representations: Evidence from Hebrew. Cognition 104: 254– 286. Berko, Jean. 1958. The child’s learning of English morphology. Word 14: 150– 177. Brame, Michael. 1970. Arabic phonology: Implications for phonological theory and general Semitic. Doctoral Dissertation, MIT. 6/11/2021

Chekayri, Abdellah & Scheer, Tobias. 2004. The non-appearance of glides in the hollow verbs of Classical Arabic. Paper presented at the first conference on Arabic language and linguistics, Oxford 30 -31 July. Chekayri, Abdellah & Scheer, Tobias. 2005. Biliteral approach to weak verbs in Arabic. Paper presented at the 19 th Arabic Linguistics Symposium, University of Illinois 1 -3 April. Coetzee et al. 2007. Post-nasal devoicing in Tswana. In Jürgen Trouvain and William J. Barry, eds. ICPh. S XVI. p. 861 -864. (pdf) (With Susan Lin and Rigardt Pretorius. ) Frost, R. , Deutsch, A. , & Forster, K. I. 2000. Decomposing morphologically complex words in a nonlinear morphology. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26, 751– 765. doi: 10. 1037/02787393. 26. 3. 751 Gafos, Adamantios. 2003. Greenberg’s asymmetry in Arabic: A consequence of stems in paradigms. Language 79 (2), pp. 317 -355. Rosenthall, Sam 2006. Glide distribution in Classical Arabic verb stems. Linguistic Inquiry 37, pp. 405 -440. Ussishkin, Adam, Reimer, Dawson, Colin, Wedel, Andrew & Schluter, Kevin. 2015 Auditory masked priming in Maltese spoken word recognition, Language, Cognition and. Neuroscience, 30: 9, 1096 -1115, DOI: 10. 1080/23273798. 2015. 1005635 Zsiga, Elizabeth, Gouskova, Maria & Tlale, One. 2006. On the status of voiced stops in Tswana: Against *ND. Proc. NELS 36. Amherst, 721 -734. 6/11/2021

Pharyngealized consonants have low F 2, like [u/u: ] and unlike [i/i: ] More [u] with pharyngealized consonants – could be assimilation * C V [+low] [–back] More [i: ] with pharyngealized consonants – dissimilation? * C V [+low] [+back] Cf. *NT vs. *ND (Zsiga et al. 2006, Coetzee et al. 2007) 6/11/2021

Honaida Ahyad Honaidah. [email protected]. edu Michael Becker Michael. be[email protected]. edu Michael. [email protected]. org 6/11/2021