This paper discusses ter- vs di- verbs in Indonesian from their argument and semantic structures in order to understand and explain their distributions. It is argued that, while di- is certainly a voice prefix, ter- is not, although in one of its uses it looks like a passive marker. This is due to the fact that both di- and ter- verbs share the property that the most prominent argument in their argument structures (a-str) is an Undergoer. However, evidence is provided to support the idea that ter-, unlike di-, is a morphological head having its own argument and semantic structures. Crucially, ter- imposes a constraint that the structure is syntactically intransitive (i.e. consisting of only one core), and this core Undergoer argument must be shared with an embedded Undergoer argument. This explains why ter- cannot take a base whose core argument is highly agentive (e.g. lari 'run'--> *ter-lari). The a-str based analysis also provides an explanation for volitional contrast such as ter-bakar 'get burnt' and di-bakar '(be) burnt (by someone)', where a volitional agent is implied in the di-verb but absent in the ter-verb. Furthermore, a detailed semantic analysis is proposed showing that 'successful causation' is central/primary to the meaning of ter-. Another important, but optional, component in the semantic structure of ter- is the evaluative meaning of being 'unwanted'. Certain other meanings observed with ter- such as accidental, spontaneous, ability, and inchoative are arguably parts of, or derived from, the primary meaning, possibly in combination with the evaluative meaning. A new integrated analysis proposed in this paper accounts for a range of certain contrasts in acceptability and/or meanings such as terjatuh 'fall off accidentally' vs *dijatuh, terlibat 'be involved' vs *dilibat (vs dilibatkan '(made) involved'), (?*)terpenuh vs terpenuhi 'successfully fulfilled' (vs dipenuhi 'fulfilled'), and tertutup/i 'closed/covered' vs ditutup/i 'closed/covered (with something) by someone'. The analysis is also supported by text statistics, showing that certain ter-verbs have been overused resulting in semantic bleaching and in a change of word category. For example, ternyata 'in fact' (lit. 'get real') and tersebut 'the or that (lit. 'already mentioned') are two most frequently used ter-words in text, which now appear to be used as an adverb and anaphoric determiner respectively.