Indonesian verbs take two suffixes, -kan and -i, which show properties of both causative and applicative morphemes (Kaswanti 1997 inter alia).

(1)	"Causative" -kan
a.	Tika tewas
	Tika dead
	'Tika is dead'
b.	Ia menewaskan Mary
	3SG meN-dead-CAUSE Mary
	'He killed Mary'
(2)	"Applicative" -kan
a.	saya memanggang biskuit untuk Eric semalam
	1SG meN-bake biscuit for Eric last night
	'I baked a biscuit for Eric last night'
b.	saya memanggangkan Eric biskuit semalam
	1SG meN-bake-BEN Eric biscuit last night
	'I baked Eric a biscuit last night'
c.	saya memanggangkan biskuit untuk Eric semalam
	1SG meN-bake-BEN biscuit for Eric last night
	'I baked a biscuit for Eric last night.'
(3)	"Applicative" -i
a.	saya menanam pohon di halaman belakang kita
	1SG meN-grow/plant tree in back yard 1PL
	'I grew/planted a tree in our backyard'
b.	saya menanami halaman belakang kita dengan pohan
	1SG meN-grow/plant-APPL back yard 1PL with tree
	'I grew/planted a tree in our backyard'
*c.	saya menanami pohon di halaman belakang kita
	1SG meN-grow/plant-APPL tree in back yard 1PL
	'I grew/planted a tree in our backyard'
	(Grammatical only if this sentence means 'plant the tree repeatedly')

While -kan and -i seem to share many properties, they have many differences as well: 1) In ditransitives -kan occurs when the patient is the direct object of the verb:

(4)	saya memberikan/ *memberi surat untuk/kepada Peter
	1SG meN-give-KAN mem-ber(i)-i letter for/to Peter
	'I gave a letter to Peter.'

In contrast, -i is found when the goal is the direct object:

(5)	saya memberi/ *memberikan Peter surat
	1SG meN-give-i meN-give-KAN Peter letter
	'I gave Peter my letter.'

These facts and others were described in detail in Kaswanti (1997).

The main problems presented by -kan and -i are the following:

(A)	Why can -kan be used as both an apparent causative and as an applicative? What determines when a causative or an applicative interpretation is found?
(B)	What determines the different distributions of -kan and -i as applicatives? While -i has a distribution similar to that of prototypical applicatives (e.g. in Bantu) and occurs only in "dative shifted" structures, -kan is compatible with both a "dative shifted" and an "unshifted" structure (compare 2 and 3)

We will argue that -i should be analyzed as an instance of incorporation along the lines of Baker (1988): an abstract preposition is incorporated into the verb, making the object of the preposition the internal object of the derived verb. In contrast, -kan has the effect of adding an internal argument to the argument structure of the verb. This feeds "dative shift", but -kan does not itself cause "dative shift" as does incorporation.

Apparent causatives involve a discrepancy between syntactic subcategorization and theta role assignment. They are created when -kan adds an internal argument to a verb that has only a single, internal argument (e.g. unaccusative verb). In unaccusatives, the argument that is assigned the patient theta role cannot be syntactically internal because the verb does not subcategorize for an internal syntactic argument. -Kan adds a new internal argument position, and the patient is realized syntactically in that position. Burzio's Generalization, which we rephrase as (6), explains the causative interpretation:

(6)

syntactically realized (casemarked) internal argument <-> theta marked subject The creation of a theta marked external argument is a secondary effect of the addition of the internal argument by -kan. The causative interpretation, thus, is a secondary effect of the addition of an internal argument to the argument structure.

Putting aside idiosyncratic, lexicalized uses of -kan, we claim that a unified analysis is possible for all productive uses of the suffix.

References

Baker, Mark (1988). Incorporation. University of Chicago Press. Kaswanti Purwo, Bambang (1997). The direct object in bi-transitive clauses in Indonesian. In: Grammatical Relations. Amsterdam: John Benjamins.