Applicative voice is a valency-increasing operation that promotes a peripheral thematic role (benefactive, locative, instrumental, malefactive) to object status. Applicative morphemes, frequent in Bantu languages, allow constructions like 'build-APPL a house for him' (benefactive applicative) to function as syntactically transitive or ditransitive structures.
You've studied valency-changing operations — processes that add or remove arguments from a verb's argument structure. You've also worked with thematic roles: the semantic labels (Agent, Patient, Beneficiary, Goal, Instrument, Location) that describe how participants relate to an event. Applicative voice is where those two frameworks intersect: it is a valency-increasing operation that takes a role that is normally expressed as an adjunct (a peripheral, optional phrase) and pulls it into the core argument position of the verb.
To see the contrast, start with a simple transitive clause: "She built a house." The verb *build* has two core arguments — an Agent (she) and a Patient (a house). If you want to add a Beneficiary, you do so with a prepositional phrase: "She built a house for him." That for-phrase is an adjunct — syntactically optional, not required by the verb's valency. In a language with a productive applicative morpheme, the verb itself can be marked to incorporate that Beneficiary directly: "She build-APPL him a house," where *him* is now a direct object, a core argument. The preposition disappears; the morpheme on the verb absorbs it. The result is a ditransitive structure with three core arguments: Agent, Applied Object (the new one), and Patient.
Different thematic roles produce different applicative types. A benefactive applicative promotes a Beneficiary ("she cooked-APPL him food" = she cooked food for him). A locative applicative promotes a Location ("she slept-APPL the mat" = she slept on the mat). An instrumental applicative promotes an Instrument ("she hit-APPL the stick him" = she hit him with the stick). A malefactive applicative promotes someone adversely affected ("she cooked-APPL him food" in a context meaning she cooked his food without permission). Bantu languages like Swahili and Chichewa are the canonical examples because their applicative morphemes are highly productive and their argument-structure effects have been extensively documented.
Why does this matter for typology and syntactic theory? The applied object behaves like a core argument in several ways: it can become the subject of a passive, it controls agreement morphology on the verb, and it has the word-order properties of a direct object. This is evidence that grammatical relations like "object" are not just semantic notions but syntactic positions that can be assigned to different thematic roles by morphological means. Applicatives thus reveal a crucial design feature of clause structure: the number of semantic roles present in an event can exceed the verb's default valency, and morphological operations exist to restructure which roles occupy which grammatical slots. From your earlier work on case systems, you can also see that languages with rich case morphology often don't need applicative morphemes — they can simply case-mark a Beneficiary differently — which is one reason applicatives are more developed in languages with less robust case systems.