📄 The Verbs with Hamza as First or Third Radical in Modern Arabic Dialects

Summary

Czapkiewicz examines how Arabic verbs whose first or third root consonant is a glottal stop (hamza) changed in modern spoken dialects once hamza dropped out of their sound systems. He also asks which of these changes date back to the Middle Ages or earlier and which are more recent. He collects verb forms from roughly thirty dialects, ranging from Mauritania and Morocco through Egypt, the Levant, Iraq and the Gulf to Yemen, Oman–Zanzibar and Bukhara. He arranges them by derived stem class and lines each one up with its Classical Arabic counterpart. He then sorts the observed differences into a small set of change types and explains them with phonetic loss, classical notions of hamza "lightening" (takhfif) and analogy. His main claim is that every change began in phonology. One innovation is comparatively late: an initial hamza vanishes and the stem gains a final y. In several dialects this caused initial-hamza and final-hamza verbs to fall together as y-final verbs. The study matters because it shows a single sound loss reshaping morphology and even shrinking the verbal lexicon.

Key points

• The data cover verbs across stem classes I–X plus some combined classes, each tagged with dialect codes (e.g., M, T, E, KA, IB, KD, O-Z, YS) and the Classical form.
• The change types include:
  • hamza lost with no replacement;
  • initial hamza replaced by w;
  • a lost initial hamza compensated by a final y;
  • a vowel at the start of the stem in place of hamza;
  • class VIII formed as if the verb were w- or y-initial;
  • a final hamza replaced by y or by a glottal-type consonant.
• After initial hamza is lost, a stem is left with only two consonants (e.g., kla "eat" < 'akala). It is then filled out with y to fit the triconsonantal verb template, parallel to how biconsonantal nouns take w/y in derived forms.
• All changes except initial hamza → final y were already underway in medieval times.
• Hamza loss reduced the number of hamzated verbs. Some were replaced by unrelated verbs, e.g., in Tunis for "take" and "ask".

Interesting ideas

• I1. Some dialect verbs have two stem shapes, chosen by whether the following suffix starts with a consonant or a vowel (e.g., KA 'ekel vs kel), so stems should be modelled as allomorphs with explicit selection rules. — For Mova (tokens): The en morphology could store competing stem shapes with an explicit selection condition rather than listing them as unrelated forms, keeping lemma choice explainable.
• I2. A form that falls below the minimal template is "repaired" by adding a default segment, and this one principle accounts for many surface forms at once. — For Mova (learn): Self-induced repair rules could likewise be stated as "restore the required template with the default filler", then checked against negative controls before being adopted.
• I3. Each variant form is aligned to one shared reference form (Classical Arabic), and the differences are classified into a closed set of change types rather than described one by one. — For Mova (tree): UD dialect conversion could align each variety-specific analysis to a canonical reference and label each mismatch with a type from a fixed list, giving auditable conversion logs.
• I4. Analogy is invoked explicitly to explain why forms move into another inflectional class once their distinctive trait disappears. — For Mova (world): The compiled verb-class knowledge could record which classes merge under which conditions, so lookups for irregular or nonstandard forms fall back to the class they have joined.