Differential indexing in Kamang: a viewpoint alternation

3.1 Language and data

Kamang is a severely endangered Alor-Pantar (Papuan/non-Austronesian) language spoken by around 6,000 people in Indonesia (Schapper 2014: 286–287). The language possesses up to seven prefix paradigms that index the S or P argument. Table 2 gives three of the most common prefix paradigms – the /a/-series, /o/-series, and /e/-series^[7] – plus the dative and self-benefactive paradigms, which are phonologically similar to the /e/-series.

Around a third of verbal bases obligatorily bear an index from a single prefix series; the rest are either never prefixed, or participate in DAI. Various alternation types are attested – between prefixation and zero-marking, or between prefixation with indexes from different series (see Fedden et al. 2013, 2014; Schapper 2014; Walker et al. 2023). This study focuses on the alternation presented in (1) above and in (7) below; namely, between /e/-series prefixation and zero-marking.

The /e/-series alternation has been described as indicating greater patientivity of S on motion and posture verbs (Schapper 2014). This has two typical interpretations: greater affectedness and/or greater duration. In (1a), zero-marking is used for a less-affected S argument whereas, in (1b), an /e/-series prefix “indicates the dog’s running away was caused by an external event, such as someone kicking at the dog” (more affected; Schapper 2014: 255). In terms of duration, the prefixed form indicates that a motion or posture was held for a long time. In (7b), nih ‘sit’ acquires the “somewhat lexicalized meaning … ‘live, reside’” (Schapper 2014: 326) with an /e/-series prefix.

The present study aims to provide a more detailed picture of the /e/-series alternation through a qualitative analysis of examples attested in a corpus of spoken Kamang. The corpus includes monologic narratives, conversations, and structured (i.e., using stimuli) and unstructured elicitation recorded by Antoinette Schapper and George Saad between 2010 and 2020.^[8] Part of this corpus – annotated for grammatical relations and animacy with the GRAID schema (Haig and Schnell 2014, 2022) – is publicly available (Schapper et al. forthcoming). Target verbal bases were those recorded in the lexicon or the annotated sub-corpus as alternating between zero-marking and /e/-series prefixation. Utterances containing these verbs were then extracted for comparison.

There are two points of note. First, plural forms of the /e/-series are homophonous with /a/-series plurals, which may lead to ambiguity. To confirm alternation with the /e/-series, verbal bases needed to be attested at least once with an unambiguous singular prefix (all of them were). Second, the /e/-series is phonologically similar to the dative and self-benefactive series. The phonetic differences between the three series in natural speech can be minimal if not imperceptible (Al-Gariri 2022). Since the dative prefix occurs only very rarely (Fedden and Brown 2017: 424–426; Schapper 2014: 324), ambiguity with the /e/-series is unlikely. However, the self-benefactive is frequent in discourse. To minimize the chance of including self-benefactives, I selected examples that (i) previous researchers had glossed as /e/-series and (ii) had accessible recordings in which no glottal stop was audible. Some examples found only in elicitation data could be discarded when the speaker explicitly mentioned the glottal stop (by referring to the glottal stop grapheme <′>). Consequently, only the clearest instances of /e/-series prefixation were included in the study, totalling 123 tokens for 21 lexical verbs. Some verbs occur much more frequently with zero: here, 20 examples were selected for comparison, prioritizing good quality recordings and clear surrounding context. Given the modest sample size (a common challenge in research on endangered languages) and the exclusion of unclear cases, future research will likely find that additional verbal bases can alternate.

Several features of Kamang grammar are relevant to this study (for a full description, see Schapper 2014). There are several ways to mark aspect: non-obligatory perfective and imperfective suffixes, stem suppletion, and aspectual serialization constructions. Aspectual serialization involves a sequence of two verbs, the second of which expresses aspect. Completive aspect is marked by lai ‘finish’ and continuous aspect by dii ‘lie’. The latter is illustrated in (8).

Also relevant here is the resultative construction: “one verb encodes the result of the event denoted by another verb. Result verbs are intransitive and occur as the final verb in the serialization” (Schapper 2014: 345). As shown in (9), the resultant state of the S argument is expressed when the previous verb is intransitive (‘fall down’).

If the previous verb is transitive, the final verb expresses the resultant state or position of P (Schapper 2014: 346). In (10), the P argument (semantically a location) of tal ‘make fence’ is kadii ‘house’, which in turn is the S argument of the resultant state beela ‘surround(ed)’.

In other verb sequences, the final verb expresses a property of the A argument of the preceding transitive verb. While the final verb in the resultative construction in (10) expresses a property of P, the final verb in an otherwise identical construction in (11) expresses a property of A. That is, ‘a man’ is standing, not the child he is holding.^[9] There is currently no evidence that prefixation behaviour differs according to whether the final verb expresses the resultant state of the S, A, or P of the previous verb, but this remains a question for future research.

Finally, Kamang does not require any arguments in the clause to be overt (e.g., in (8) the S argument ‘they’ is unexpressed). Independent arguments (pronouns or lexical noun phrases) are permitted to co-occur in a clause with the prefix that indexes them, although co-occurrence is generally dispreferred (see Walker et al. 2023).

3.2 Kamang /e/-series alternation

Verbal bases that participate in the /e/-series alternation are listed in Table 3 along with their frequencies in each construction. The translation is of the unprefixed lexical base; where prefixation leads to an idiosyncratic meaning change – a common feature of elaboration constructions – this is given in the rightmost column. Note that unprefixed posture verbs can have both telic ‘stand up, sit down’ and atelic readings ‘stand, sit’ (cf. Schapper 2014: 326).

The items in parentheses in Table 3 were discounted due to insufficient data in the corpus to fully substantiate the alternation (e.g., due to difficulties in ruling out a transitivity difference). Three of these are emotion/cognition verbs (maringki ‘startle’, biee ‘afraid’, and mai ‘hear’): as mentioned, cross-linguistically such verbs frequently participate in DAI of S, and in elaboration constructions like resultatives and middles. Also relevant, then, is the idiosyncratic meaning shift from mitan ‘complete’ to the emotion/cognition verb ‘understand’ when prefixed.

Aside from emotion/cognition verbs, Table 3 shows the groups of motion and posture verbs, which Schapper (2014) previously identified as participating in /e/-series alternation. These semantic groups are also common in elaboration constructions; nevertheless, there is evidence of lexical restrictions even within a semantic class. For motion verbs, tak ‘run’ alternates, but semantically similar verbal bases like maa ‘walk’, loo ‘walk’, and lila ‘fly’ are not (yet) attested with /e/-series prefixes. For posture verbs, lexicalization is manifested in the idiosyncratic meaning difference between unprefixed nih ‘sit’ and ‘live; reside’ when prefixed. The final group in Table 3 is labelled “other” due to its heterogeneity. It includes verbal bases that seem inherently intransitive and non-volitional (‘sleep’) and inherently transitive and volitional (‘buy’, ‘sell’), as well as bases that appear to be inherently durative (‘work’, ‘be bright; shine’) or inherently punctual (‘bump into’, ‘finish’).

In terms of systematic meaning differences between the alternants, increased duration and affectedness is only part of the story. In (12a), the S argument has been in position for a long time, yet latsi ‘stand’ is zero-marked. Conversely, in the prefixed ge-latsi in (12b), the S argument goes volitionally to stand ‘up on top’ and does not remain there long. It is difficult to impute a “greater affectedness” reading here.

While greater affectedness is often a property of arguments indexed with the /e/-series, it is one of several properties associated with elaboration constructions, including completion, inchoativity, resultativity, and features of middles. As the following examples illustrate, the /e/-series alternation is best analysed as a viewpoint split between zero-marked verbs that express events holistically – either with all three phases, or without internal phasal partitions – and /e/-series-marked verbs that shift focus to the elaboration phase.

Example (13) shows two almost identical resultative serializations, the final verb dii ‘lie’ expressing the position of the P argument of the preceding verb kuh ‘cover’: botil ‘bottle’ in (13a) and sukuu ‘hole’ in (13b).

Unprefixed dii in (13a) is best analysed as a state proper with no internal phases (the utterance was elicited to describe a static picture). In (13b), the utterance is part of a narrative expressing a series of dynamic events. Here, the prefixed verb is associated with a dynamic event with internal phases. This includes initiation (the initiator is unexpressed but clear from the context), but the focus is on the elaboration phase: ge-dii expresses a resultant state, which implies an inchoative phase.

Example (14) demonstrates another resultative serialization. Here the alternation applies not to a motion or posture verb, but to awila′ ‘fill’. In elicitation, the speaker described the meaning difference as a matter of degree; that is, the meaning is neutral ‘fill’ when zero-marked and completive – ‘completely full’ – with an /e/-series prefix. As described above, completive aspect is associated with the end phase of an event, which also implies a greater degree of affectedness of the indexed argument.

In (15), silang ‘descend’ alternates between a zero-marked variant with habitual reading in (15a), reflected in the translation ‘used to go down’, and a single punctual event denoted by the prefixed variant in (15b). In a habitual reading, the whole event from initiation to completion is relevant since it repeats in its entirety. This contrasts with the prefixed variant, which denotes a single punctual event with internal phases, of which the elaboration phase is in focus: what is important is that the woman has descended, resulting in her being in the hole. This pair of utterances – the only clear example involving habituality identified so far in the corpus – might seem contrary to the previous description of the /e/-series as indicating long duration; however, as mentioned, elaboration phases have telic and atelic characteristics due to including both inchoative and stative phases.

As well as intransitive verbs, several alternations involve transitive verbal bases, such as baila ‘buy’, balkei ‘sell’, and beta ‘bump into’. The last-mentioned is a naturally reciprocal event associated with middle semantics (Kemmer 1993: 102–108). The prefix on beta ‘bump into’ indicates greater affectedness of P: in elicitation, a speaker explained that unprefixed beta means ‘bump into’ or ‘barge’, while the prefixed version means ‘push over’. This follows from greater focus on the elaboration phase of the event, which features P. The corpus yields one alternation pair, shown in (16), which describe the same stimulus (Fedden et al. 2010b), in which man A bumps into man B, causing the latter to step backward. The two speakers frame the event differently: in (16a), man A merely bumps an arm, whereas in (16b), beta is prefixed when it indexes man B, who is more affected by the push than an arm.

The difference in meaning between prefixed and zero-marked baila ‘buy’ and balkei ‘sell’ is unclear. There are two possible interpretations, both of which fit an elaboration-focus analysis. First, the prefixed variant in (17b) may indicate greater focus on P than A. Indeed, the A argument is not mentioned when the prefix is present, while in the zero-marked variant in (17a), the speaker emphasizes the agentivity of A by using the agentive pronoun.^[11]

The second analysis of (17) is a more telic interpretation of the prefixed verb: a resultative sense in which (all) the goods have entered into the resultant state of ‘bought’ or a completive sense in which the purchase is complete (compared to ‘being in the process of buying’). The latter interpretation is supported by the co-occurrence of the imperfective suffix -si with unprefixed baila in (17a).