On-going Projects

Adjective predicates can take two negators: the circumfix [ma-Adj-ʃ] and [maʃi].

Main research questions:

  1. Is this a case of free variation?
  2. Is [ma-Adj-ʃ] or [maʃi] more preffered in some adjectives more than others? what factors play a role here? and on what's the nature of these factors (syntactic, phonological, etc)?
  3. How can we formally account for this variation?

The temporal stability of medial consonant sequences in Moroccan Arabic

Main research questions:

  1. Are medial consonant sequences in Moroccan Arabic part of the same syllable or are parsed into separate syllables?
  2. How can we use acoustic measurements to analyze the temporal stability of medial consonant sequences?
  3. Do geminate clusters in medial positions show the same timing properties?

Augmentation in the Moroccan Arabic broken plural: a minimal size requirement (with Michael Becker)

Main research questions:

  1. Can the augmentation of the Moroccan Arabic CCuC plurals to CCuCa be analyzed as a minimal size requirement?
  2. What is the role of the final [-a]? and how does it differ from similar morphemes in Moroccan Arabic and in other Arabic varieties?
  3. What evidence exists to support the claim that the final [-a] in CCuCa is part of a recent diachronic change?

Completed Projects

Variable and exceptional assimilation of the definite article [l–] in Moroccan Arabic

This paper investigates the morpheme-specific phonological behavior of the Moroccan Arabic definite article [l-] and its variable assimilation process with nouns that begin with the palatal fricative [ʒ]. By analyzing of a corpus of these nouns, I identify one major factor that influences this variability: the context following [ʒ].


The results of a nonce word experiment I conducted show that speakers follow the lexical tendencies when generalizing to [ʒ]-initial nonce words,which does not align with the assumptions proposed in the literature (Harrell, 1962; Heath, 1987, 1989; Freeman, 2016).


I show that a learnability approach that uses Maximum Entropy Grammar (Goldwater and Johnson, 2003) with Lexically-indexed Constraints (Pater, 2000, 2009) successfully accounts for both the categorical behavior of known [ʒ]-initial words as well as the variability predicted for nonce words.

Distance-based sibilant harmony in Moroccan Arabic

In this paper, I investigate the variable patterns of regressive sibilant harmony in Moroccan Arabic, focusing on the influence of distance between harmonizing segments. Specifically, the palatal fricative [ʒ] triggers harmony in the alveolar fricatives [s] and [z], transforming them into [ʃ] and [ʒ], respectively. Through experimental analysis, I identify one major factor that affects the likelihood of harmonization: the number of intervening segments between the sibilants.


The results of an experimental study I conducted demonstrate that harmony occurs more likely when there is only one intervening segment between the harmonizing fricatives, with harmonization occurring 52% of the time, compared to 26% when there are two or more intervening segments. However, factors like the voicing of the target sound and morphological complexity had minimal impact on harmony rates.


I propose an analysis using Maximum Entropy Grammar (Goldwater and Johnson, 2003) alongside Agreement-by-Correspondence (Rose and Walker, 2000, 2004), which successfully accounts for both the observed variation and distance effect, respectively. Additionally, lexically-indexed constraints (Pater, 2000, 2009) help capture the exceptional behavior of certain lexical items.

Developing a Shiny App for MaxEnt learner with Hidden Structure (Advisor: Joe Pater)

In this project, I developed an interactive Shiny application, MaxEnt with Hidden Structure in R, designed to assist linguists in generating phonological grammars (weights) using a Maximum Entropy model. The app was built based on the existing HGR model developed by Staubs (2014). HGR finds solutions for learning problems (with or without hidden structure), generate distributions over forms, and performs online learning simulations.


In addition to creating the shiny app interface for the HGR model, I added two main features:

  1. In cases where the training data come with raw frequency counts, instead of probabilities, I added the option of normalizing globally, not only within tableau as in the original HGR model.
  2. Second item
  3. I added the option to edit the generated weights and updating an existing grammar.

The source code is available on Github.

MA Thesis: Broken plural formation in Moroccan Arabic: a stratal optimality theory account

The previous accounts of broken plural formation are successful in accounting for some broken plural patterns in Classical Arabic, but I show that they are insufficient to account for broken plural formation in Moroccan Arabic. I show that a Stratal OT (Kiparsky, 2000; Bermúdez-Otero, 2003; Rubach, 2003; among others) analysis of this phenomenon is needed as an alternative account. I adopt the representational assumptions made by Al Ghadi (1990) regarding the syllable template of Moroccan Arabic. The analysis proposed assumes that there are two main levels to broken plural formation: the stem level and the word level. At the stem level, the infixation of the broken plural morpheme takes place, while syllabification and epenthesis occur at the word level. The analysis I propose accounts for the major broken plural patterns in Moroccan Arabic, i.e. CCVC, CCVCa, and CCVCeC.


The thesis is available here.