This talk presents a theory of the interfaces of semantics, morphology, and syntax, on the basis of the Voice system of Akkadian. I suggest that features can be charged for positive, negative, or neutral feature values, and that depending on their charge, a feature may attract, repel, or simply pass through the syntactically-defined field of another feature.
Akkadian is an extinct East-Semitic language, attested in Mesopotamia from about 2600 BCE to 75 CE. It derives its lexical words by combining triradical IDIOSYNCRATIC LEXICAL ROOTS with templates. While the ROOT supplies the lexical base of a given word, a template provides further specification on the word class and its relevant grammatical categories. Verbal templates primarily encode TAM-features; Akkadian marks three aspects (perfective, imperfective, resultative) and one mood (imperative). Causatives and anticausatives are formed through template patterns. A pattern features one characterising morpheme and a set vocalic pattern, both of which are recurring throughout the conjugations. The paradigm is given in Table 1 for the 3SG.M, where X-Y-Z serve as stand-ins for the ROOT consonants.
Table 1: The Akkadian template pattern and conjugations
The projection of a ROOT, i.e., its simple combination with a categorising head (Halle & Marantz 1993), is commonly termed a G pattern (for Grundstamm, Ger. ‘base stem’). The two causatives are termed D (for Dopplungsstamm, Ger. ‘doubling stem’) for the geminated second radical featuring throughout the conjugations, and Š (Š Stamm Ger. ‘stem’) for the characterising š- prefix appearing before the ROOT consonants. The two anticausatives are similarly termed after their respective morphemes t- and n-.
Studies thus far (e.g., Kouwenberg 1997, 2010) have provided philological descriptions of causative and anticausative semantics, and rough trends as to which ROOT may derive which patterns. In Kamil (2025), I define the semantic content of all template patterns and morphemes, as well as that of ROOTS, showing that (in)compatibilities between ROOT and template features determine the constraints for derivation. The key feature shared by ROOTS and causatives, which drives the mechanism is telicity.
The two causatives D and Š both add a causation event to the derivation but differ in the aspectual specification of their causation. D causatives mark an atelic causative, introducing a causation event, which intersects with the change-of-state (CoS) event undergone by the Patient, and which does not entail culmination. Š causatives mark a telic causative, introducing a causation event that entails culmination before the onset of the CoS event. ROOTS specified as [+telic] may only derive the [+telic] Š stem, ROOTS specified for [-telic] derive almost only the [-telic] D stem. Through their telicity features, D and Š also substantially differ in the primitive predicates they semantically add to a derivation, i.e., DO for D and CAUSE for Š. I argue that, in Akkadian, the two predicates represent the two different causal subevents of causative events, which ultimately also derive different causer subjects at different heights of attachment. I furthermore suggest that agentive ROOTS also specify a DO predicate, which affects which causative they may derive and which anticausative meaning arises in the derivation of t-stems.
The analysis of the rules of Akkadian verb derivation builds on a patchwork of Distributed Morphology, lexical semantics, and Magnetic Grammar (D’Alessandro & van Oostendorp 2020). I focus in particular on the LEXICAL ROOT and its semantic and syntactic restrictions in the derivation of grammatical features, and, in effect, grammatical forms.