The Indonesian suffixes -i and -kan are often cited as classic examples of "valence increasing" morphology. Chung (1976) describes them as "object-creating" suffixes which promote an oblique argument or adjunct to core argument status (specifically Direct Object), and thereby increase the syntactic transitivity of the verb as illustrated in (1). However, Kaswanti (1995, 1997) points out that the syntactic effects of these suffixes, and -kan in particular, are not uniform. He identifies two basic patterns, which are illustrated in the following contrast (from Kaswanti 1997):

(1)	a.	John membeli-kan buku itu untuk Mary.
		John buy book that for Mary
		'John bought that book for Mary.'
	b.	John membeli-kan Mary buku itu.
		John buy Mary book that
		'John bought Mary that book.'
(2)	a.	John memberi Mary buku itu.
		John give Mary book that
		'John gave Mary that book.'
	b.	John memberi-kan buku itu kepada Mary.
		John give book that to Mary
		'John gave that book to Mary.'

Kaswanti takes these two verbs, 'give' and 'buy', as prototypes and attempts to classify other roots into two lexical classes according to the pattern that they follow most closely. Chung posits three lexical classes, depending on whether her Dative rule is triggered by -kan, -i or -Ø. One problem for this type of analysis is the fact that some verbs allow both of the patterns illustrated above, and thus must belong to two different classes. For these roots, as Kaswanti notes, the form meN-root-kan is ambiguous, as illustrated in (3):

(3)	(Kaswanti 1995)
	a.	Ibu menjahit-kan saya baju.	[Benefactive]
		mother sew me shirt
		'Mother sewed me a shirt.'
	b.	Saya menjahit-kan baju ke tailor.	[Causative]
		I sew shirt to tailor
		'I had my shirt sewn by a tailor.'

A more revealing analysis can be achieved by positing two homophonous suffixes, which I will refer to as -kan₁ (instrumental/ causative/ displaced theme) vs. -kan₂ (benefactive). Some support for this proposal comes from the observation that, when the form meN-root-kan is truly ambiguous, the benefactive use is (almost) always one of the possible senses as in (3). More importantly, the two suffixes are distinguished by their syntactic behaviour. Only -kan₂ (benefactive) can apply "vacuously", i.e. be present where the clause structure does not match the expected output of the rule. Conversely, only -kan₁ can be truly optional, i.e. be absent where the clause structure does match the expected output of rule. The two affixes also differ in terms of input and output requirements. -kan₂ (benefactive) applies only to transitive V roots, whereas -kan₁ occurs with roots of all types: transitive V, intransitive V, ditransitive V, adjective, noun. Furthermore, -kan₂ always creates a ditransitive clause (a double-object construction), whereas -kan₁ creates a mono-transitive clause. Thus -kan₂ always increases the syntactic valence of the verb but in the presence of -kan₁ the syntactic valence of the verb may increase, decrease or remain the same. This suggests that -kan₁ cannot be treated as a normal "valence increasing" affix, but instead must be associated with a template of some kind which determines the argument structure and subcategorization properties of the output form.

Homophony between Causative and Benefactive forms is not necessarily predictable on semantic grounds, but it is not uncommon in serial verb languages (e.g. Thai, Alamblak). In Chinese the prepositional verb gei 'give' is used to mark benefactives and permissive causation, which may suggest the original semantic basis for the homophony in other languages.