📄 Ideophones and verbal constructions with the verb ‘say’ in Amharic

Summary

The paper asks whether Amharic ideophones form a distinct word class and how they behave inside complex predicates. These predicates pair an uninflected coverb with an inflected light verb, most often alä 'say'. Earlier descriptions treated all such coverbs as one group. Meyer instead sorts coverbs into distinct types: besides ideophones, he identifies interjections, adverbial elements, nouns or verbs that have undergone reduplication, and quoted material. He then defines ideophones on Amharic-internal morphosyntactic grounds.

The data is a convenience corpus: printed sources (mainly Leslau's 1995 reference grammar), twenty years of observed speech, and targeted online searches. Native speakers checked all of it.

The decisive test is causal alternation. Only ideophones can switch between a noncausal reading with 'say' and a causal reading with adärrägä 'do'. Amharic can also build ideophones from verbal roots with two templates, one intensifying and one weakening the event.

A comparison across Ethiosemitic shows that languages near Amharic share its rich coverb inventory, while distant ones have fewer coverb types. The work clarifies lexical classes, verb classes and contact effects in the region.

Key points

• Amharic has about 57 million speakers in Ethiopia, about 25 million of them second-language speakers.
• Ideophones do not inflect and carry final-syllable stress. They take the clause-final verb slot, can have direct objects, can be modified by converb clauses, and reduplicate to mark pluractionality. Some can be nominalized with -ta.
• 'Say'/'do' complex predicates are frequent enough to count as a separate inflectional verb class. They are also a productive way to absorb loanwords that resist root-and-template morphology.
• Intensive vs. attenuative ideophones are derived from the same root and contrast with a neutral inflected verb. In Tigrinya, gemination alone (sɨbbɨr vs. sɨbɨr 'break') marks the difference.
• This derivation also appears in Tigrinya, Argobba, Kistane and Harari, all in close contact with Amharic, which suggests areal spread.

Interesting ideas

• I1. The construction can be split into two parts: a coverb carrying the lexical meaning and a light verb carrying tense, person and causation. Analysing them separately avoids treating the light verb as the real predicate.
• I2. A word class is defined by a language-internal test (whether it allows the noncausal/causal swap) rather than by meaning such as "sound imitation". This yields a checkable criterion instead of an intuition.
• I3. Degree (intensified vs. attenuated) is treated as a regular derivation from one root, so "break completely" and "break somewhat" share a single event core.
• I4. Light-verb frames serve as an entry route for words that do not fit the native morphology. This explains how unfamiliar lexemes become usable predicates.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (frames): English light-verb constructions like "take a walk" or "give a shout" could get their semantic frame from the noun, with the verb supplying only grammar and agentivity.
• I2 (tokens): Onomatopoeic items in "the balloon went pop" could be tagged by distributional diagnostics rather than by a lexicon of sound words.
• I3 (world): Event states could store a degree attribute linked to one predicate, so "smashed" and "cracked a bit" update the same object state at different intensities.
• I4 (tokens): Unknown verbs or nonce words inside "do a X" or "go X" patterns could get a predicate analysis from the light verb instead of failing at lemmatization.