The Pahoturi River language family, with special reference to its verbal puzzles

4.1.1 Analytic stem and synthetic stems

To discuss the topic of verbal stems, we use the superordinate term lemma to refer to the lexicon entry of a single verb, or a set of related verb stems. For instance, the Kawam lemma ‘to baptize’ consists of the analytic stems kumbog (npl) and kumbumeny (pl) and the synthetic stems ngkubog (npl) and ngkubumeny (pl). In contrast, the Kawam lemma ‘to see’ only contains the analytic stem ikop. Analytic stems differ from synthetic stems in that no inflectional material attaches to them directly, but must occur instead on a following auxiliary element.^[7] Consider the contrast between the realization of the lemma ‘to baptize’ with a plural analytic stem kumbumeny preceding an inflected auxiliary verb (19) and with a directly inflected singular synthetic stem ngkubog (20).

4.1.2 Restricted and extended stems

In both Idi and Ende, verb stems can be divided into two main subclasses: restricted and extended. They correspond to a difference in lexical aspect: punctual/telic versus durative/atelic. In Idi, restricted and extended stems form different conjugation classes. Restricted stems do not take a past tense prefix, and they take a different set of agreement suffixes. There is often a formal similarity between restricted and extended stems from the same lemma, but derivational processes are often opaque and no longer productive. Below, the difference between inflected forms for the remote past is shown by restricted -gädzn- (21) and extended -gädz- (22), both meaning ‘take out’. The different forms reflect different event structures: a python can be removed from a hole all at once (punctual), while honeycomb is removed bit-by-bit (durative).

This differential inflection marking for restricted and extended stems is not observed in Ende.

4.1.3 Valency of stems

The majority of PR verb stems are either intransitive (subcategorized for a single S^[8] argument) or transitive (subcategorized for an A and O argument). There are also many ambitransitive stems that can be intransitive or transitive without any overt derivational process. These are mostly patientive, where the S in the intransitive use corresponds with the O in the transitive use: examples include Kawam zän ‘enter (intransitive; 4.1.3); put into (transitive; 4.1.3)’ or Idi nglbn ‘stand up, grow up (intransitive); lift up (transitive)’. Valency cross-cuts the lexical aspectual distinction; thus, both transitive and intransitive verbs are found in restricted and extended classes.

The intransitive template is also used in passive and anti-passive constructions with verb roots that typically take the transitive template.

4.2.1 Multiple and distributed exponence

As in other languages in the area (see for example, Carroll et al. 2016; Döhler 2018), PR languages exhibit distributed exponence in their verbs, meaning morphemes are specified for multiple grammatical categories and those categories are distributed across morphemes (Carroll 2022; Carroll et al. 2016; Caballero and Harris 2012; Harris 2017). Under Caballero and Harris’ (2012) typology, multiple exponence in PR is affixal and formally distinct, with no licensing dependency on each other. In other words, to determine the values for verbal categories, the entire verb must be considered.

The Ende form in (25) illustrates this complexity. Each morpheme in (25) has as many as five potential interpretations across tense (remote past, recent past, future/irrealis), arguments (A, O), person (1, 2, 3), and number (singular, dual, plural). For example, the first morpheme in (25), n-, provides five possible interpretations (rec.tr.sgO; rec.tr.plO; fut.3>2sg; fut.2>3sg; fut .2>3 pl ), while the final morpheme -eyo has three possible interpretations (rem.3nsgA; fut.3nsgA; fut .2 nsg A).^[9] The co-occurrence of these two morphemes disambiguates the verb as being future tensed since that is the only interpretation compatible with both morphemes. Accordingly, the verb in (25) has an unambiguous tense, despite the ambiguity at the morpheme level.

The glossing line in (25) illustrates how indexing of the O argument is manifest across four morphemes: the tense prefix (n-), the O-prefix (ə-), the stem (ɽəntmeɲ), and the pluractional suffix (-ajm). In transitive verbs with a single O argument, these morphemes all agree with the person and number features of that O argument and by association also accord with one another.

4.2.2 Tense-Aspect-Mood marking

All varieties have three morphological perfective tenses: remote past, typically used for events that happened before sunset yesterday, recent past for events earlier today and last night, and future for upcoming events, imperatives, and other modal extensions. A fourth inflectional paradigm, called imperfective, primarily has past habitual and irrealis uses. Examples (26) and (27) show the two past tenses in Agob, distinguished by differing tense prefixes for the verb oter ‘to sleep’.

Formal marking of tense-aspect-mood (TAM) is distributed across the verb, appearing primarily in the first prefix slot (which also marks valency), the first suffix slot (which also marks verbal number), and the final suffix slot (which also indexes S and A arguments).

Several verbal particles are identified which add further modal meanings to a clause, like the future particle ményi in Agob (28).

The present progressive can only be expressed in an analytic construction, see (29). This type of periphrastic construction is also used for the other TAM values for non-inflecting verbs and is discussed in Sections 4.1 and 6.1.

4.2.3 Participant number

Participant number marking (Corbett 2000) is particularly interesting in PR, as verbs exhibit different types of number marking in multiple locations. Suffixes typically agree in number with S and A arguments, while prefixes agree with O arguments. Many verb stems also alternate according to the plurality of the absolutive participant: if the S or O argument has a plural referent (greater than two), the use of a plural stem is obligatory: compare for instance Idi wadzan ‘enter, put into (nonplural S/O)’ and dzardzar ‘enter, put into (plural S/O)’. Plural stems can also be used to mark event number, for example, with actions that are performed several times or at several places.

Interestingly, affixes and stems do not distinguish the same number categories. In Idi, the O-prefix distinguishes singular and non-singular (2+) objects, while the stem distinguishes non-plural (1–2) and plural (3+) participants. Thus, a dual object is indexed with a nonsingular prefix and a nonplural stem, see (30). In this example, paucal number (a small number greater than 2) for the A can only be deduced from the use of the suffix -m, referring to plurality of an S or A, on the second verb. See Schokkin (2022b) for an in-depth discussion of verbal number in Idi.

Both Idi and Ende have a dedicated slot between the stem and the S/A suffix slot, which takes a form marking the (non-)plurality of both A and O in transitive verbs. In Ende, if the A is singular and the O is plural (3+), the -neg suffix is used, see (31). If the A is nonsingular (2+) and the O is plural (3+), -aeb is used, see (32).

Thus, we see multiple ways in which number marking is distributed across verbs. Sometimes, a number value does not have a dedicated form, but is constructed by combining different elements with other values, see (30). Elsewhere, number is marked simultaneously in several places on the verb without serving a disambiguation function, as in (31) where the plural (3+) quantity of the O is indexed in the prefix, the stem, and the suffix immediately following it.

6.1.1 The uninflecting element: analytic stems

As mentioned, most lexical verbs have two stems: an analytic form used in a periphrastic construction for the present tense, and a synthetic form used for the other tenses. For example, the verb ‘to bite, eat (meat)’ has the analytic stem dhndhg (47), and the synthetic stem -ndhg- (48).^[12]

In addition to the verb lemmas with both types of stem, we also find many lemmas containing only analytic stems. These are never inflected directly but always followed by a separate form, not only in the present, but also the other TAM categories. Frequently occurring examples are yeka ‘to tell, speak’ (also a noun meaning ‘language’) and yéndhpä ‘to see’ (related to yéndhép ‘eye’). Below, the use of yéndhpä is illustrated for the present, remote past, and recent past tenses, respectively.

Analytic stems subcategorize for transitivity of the accompanying form. Yeka, for instance, combines with either transitive or intransitive inflected forms, whereas yéndhpä occurs exclusively with a transitive inflected form. Many analytic stems refer to bodily processes, emotions and physical sensations, or to states, such as si ‘to urinate’, ngndä ‘to cry’, or yéndu ‘to sleep’.

Borrowed verb forms only occur in analytic constructions and are never inflected directly. Obvious loans are, for instance, the English verbs miks ‘to mix’ and lus ‘to lose’. Also English nouns, such as sespin (‘to boil in water’, lit. ‘saucepan’) or paia (‘to shoot with a gun’, lit. ‘fire’) are encountered in this construction. Facilitating the use of loanwords as verbs is a function of complex predicates observed across the Southern New Guinea region and elsewhere. For example, in Lavukaleve, a Papuan language of the Solomon Islands, there is a special verb-adjunct construction used just for loanwords (Terrill 2003: 399–400). For discussion of Komnzo (Yam), see Döhler (2018: 115), for Urdu, see Butt (2010), and for Bardi, see Bowern (2010).

Within the category of analytic stems, it is difficult to make a principled distinction between primarily verbal and primarily nominal stems. As mentioned in Section 3.2, analytic stems can function in non-finite subordinate clauses, taking nominal case marking. They can also function as core arguments to verbs (bearing a core case clitic, see Section 3.2), and be possessed. Below, (52) shows allative case on käkäly ‘to load’ (synthetic stem -käly-) to express purposive meaning, and (53) shows locative case on wélwél ‘to wait’ (synthetic stem -wl-) to express simultaneity of events. These suffixes are also found on nouns (see (54) and (55)) with the same semantics, as well as in their original spatial sense. Crucially, there is no formal difference in flagging case on a stem with uncontroversial verbal origins like ‘to load’, (52), and an uncontroversial nominal stem like ‘basket’, (54).

When a verbal lemma has both an analytic and a synthetic form, the analytic stem always contains either the same amount of or more morphological material than the synthetic stem. Thus, the former are analyzed as derived from the latter.

6.1.2 The inflecting element: auxiliary verbs

We will now turn our discussion to the elements bearing inflection in analytic constructions, and show that they are best analyzed as auxiliaries. We consider Idi analytic constructions a type of complex predicate: they involve two predicational elements that predicate as a single unit, and whose core arguments map onto a mono-clausal syntactic structure (Butt 2010: 2). Mono-clausal structure follows from the fact that the analytic stem always appears unmarked in the construction under discussion here. By contrast, when an analytic stem is used in a subordinate clause, and thus the construction as a whole is multi-clausal, the analytic stem is marked by a nominal case suffix, see (52–53). Neither is there any indication that the analytic stem functions as a core argument to the inflected element, as it is never marked with a core case clitic.^[13]

Although the issue has received much attention in the typological and theoretical linguistic literature (both formalist and functionalist), it has proven difficult to establish good cross-linguistic criteria to distinguish between the various types of complex predicates. One such attempt puts forward the following criteria to tell apart light verb constructions from other complex predicates, most notably auxiliary verb constructions (Seiss 2009: 509; also cf. Butt 2010; Butt and Lahiri 2002):^[14]

1. Light verbs are always identical in form to the corresponding main verb, whereas auxiliaries are usually form identical only at the initial stage of reanalysis from verb to auxiliary.

2. Light verbs always span the entire verbal paradigm, and are not restricted to appear with just one tense or aspect form.

3. Light verbs do not display defective paradigms.

4. Light verbs exhibit subtle lexical semantic differences in terms of combinatorial possibilities with main verbs, and are thus restricted in their combinations. Auxiliaries, on the other hand, are not restricted in their combinatorial possibilities (although they do not have to combine with every main verb).

The first criterion implies that in languages exhibiting light verbs, there will be a fully productive lexical counterpart to each of these elements. What is more, in order for the elements to count as light verbs “proper,” they have to be identical in form to these lexical counterparts. It is here that Idi analytic constructions depart most radically from more prototypical light verb constructions as found in many languages of Australia and the Indian subcontinent. Only a very limited number of Idi stems are encountered fulfilling the function of an inflecting element in a complex predicate, and moreover, these stems are encountered in only two additional functions: in copula and experiencer-object constructions, discussed below in Section 6.1.4–6.1.5. Also in these latter functions, they can be considered semantically empty forms.

Present tense is only expressed with an analytic construction, with a dedicated auxiliary -r- ∼ -l- (plural stem -nga-); see Appendix E, and, for example, (49). Other TAM values exhibit both analytic and synthetic constructions, and here we find perfective -g- (suppletive 3sgO -gä-, plural -gädzi-), and imperfective -nd- (plural -ndr-). None of these are found outside the three construction types under discussion in this paper, which indicates that they cannot serve the same function as lexical verbs. Instead, they are highly grammaticalized items serving as a vehicle for verbal inflectional categories like TAM, directionality, and person/number of core arguments, when combining with forms that are not allowed to bear verbal inflection themselves.

A brief contemplation of potential lexical sources points to the same conclusion. The present progressive stem is homophonous with the basic motion verb ‘to go ∼ come’,^[15] whereas the perfective and imperfective stems are formally similar to the lexical verb ḡg ‘to make’ and the 1|3sg present copula =nd, respectively. However, their inflectional paradigms do not show much similarity (but note that both the present progressive stem and the motion verb show the same type of person suppletion, a rare feature cross-linguistically). The inflecting forms in analytic constructions may have grammaticalized from lexical verbs at some point, but this historical relationship is now shrouded in obscurity due to the complexity of Idi inflectional patterns.

Thus, it is clear that the inflecting forms in the Idi analytic construction do not meet the first criterion as stated above, to be considered light verbs. Additionally, as the present tense has a separate form, they do not adhere to criterion (2) very well either. The forms do not show defective paradigms compared to lexical verbs, as, for instance, the copulas do, and so they do meet criterion (3). As for criterion (4), there appear to be few to no combinatorial restrictions on non-inflecting and inflecting forms, which indicates that Idi also does not meet this criterion.

Based on this, it appears that the inflected forms in Idi analytic constructions would be better analyzed as auxiliary verbs, as opposed to light verbs. This raises issues because the term auxiliary verb is reserved for elements that consitute a complex predicate exclusively with unambiguously verbal forms. Anderson (2006, 2011 provides a thorough cross-linguistic study of auxiliary verb constructions and other complex predicate types, taking a functional-constructional approach, and includes discussion of cases which may not be considered auxiliary verb constructions under stricter approaches. Nevertheless, Anderson (2011: 796) defines auxiliary verb constructions as “mono-clausal verb phrases that minimally consist of an auxiliary verb component that contributes some grammatical content to the expression and a lexical verb component that contributes lexical content to the expression” [emphasis ours]. When discussing light verb constructions, the author is more agnostic as to what the non-inflecting element can pertain to (2011: 812): “one element is predicative and lexically rich, but the other element is lexically empty and merely instantiates the argument structure of the construction and allows for the predicative element to be expressible morphosyntactically.”

Anderson states in a following footnote that in some languages the boundary between light verbs and auxiliaries may be near impossible to draw, citing data from the Papuan language Mek, and it is likely not the author’s intention to propose a principled distinction, per se, between auxiliary verb and light verb constructions. The above examples have shown that Idi too poses difficulties for classifying uninflected elements in analytic constructions as either primarily verbal or non-verbal. This is problematic for analyzing these constructions as auxiliary verb constructions under a strict definition. Moreover, in the other construction types to be discussed (in Section 6.1.3), they do combine with forms that are unambiguously nominal, although they do not form a complex (verbal) predicate with these forms. In terms of its function, the Idi analytic construction fits Anderson’s definition of light verb construction well.

Another recurring issue in the analysis of complex predicates is which of the participating forms, if any, can be considered the head of the construction. From the literature, it appears that certain characteristics associated with the predicate head can be shared by two or more forms in both light verb and auxiliary verb constructions. Thus, this criterion appears not to be a good diagnostic in distinguishing between the two types of constructions. Both the Idi non-inflecting form and inflecting form in the analytic construction can be said to have characteristics of (verbal) predicate heads. The non-inflecting form determines the valency of the construction as a whole, since it subcategorizes for arguments, and can be marked for verbal number, whereas the inflecting form (in addition to bearing all TAM, directional, and agreement marking) is needed to form a grammatical sentence and can be considered the syntactic head of the construction.

The above discussion has shown that Idi (and by extension, the PR family) poses a challenge to the current typology of complex predicates. It appears that the definition of either the light verb construction or the auxiliary verb construction has to be broadened to accommodate the Idi phenomena. Either a definition of light verb constructions has to allow for the fact that the light verbs in question could entail highly grammaticalized elements not occurring also as lexical verbs, or a definition of auxiliary verb constructions has to allow for a non-inflecting element with a structurally ambivalent status in terms of its part-of-speech.

Despite the difficulties presented in analyzing the construction as a whole, the inflecting stems that can occur in Idi analytic constructions clearly form a separate verbal category based on language-specific criteria, and henceforth will be referred to by the term auxiliaries.

6.1.3 Other constructions that show similarities to analytic constructions

Two further construction types in Idi resemble the analytic construction, but do not form a complex predicate: constructions in which the auxiliaries function as copulas (Section 6.1.4), and experiencer-object constructions (Section 6.1.5). Due to their similarities, these construction types may be connected to the analytic construction, and could explain its origin. The sections below discuss the constructions in turn, followed by a discussion of a possible historical relationship with the analytic construction.

6.1.4 Copula constructions with auxiliaries

In addition to the copular enclitics, which show a reduced inflectional paradigm and can attach to different parts-of-speech (see Section 4.3), Idi also exhibits copular use of the same auxiliaries encountered in analytic constructions. Each of the three is attested in such constructions. The present progressive auxiliary is encountered both with the sense ‘be’ and ‘become’, while the perfective and imperfective auxiliaries mostly have the sense ‘become’. Because the predicative expression (PE) in a copula construction is not marked by a core case clitic, many of the constructions in which one of the auxiliaries is used as a copula are formally identical to analytic constructions. Below, examples are given: (56) and (57) with the present progressive auxiliary, and (58) and (59) with the perfective auxiliary.

These examples show that the auxiliaries can fulfill both the copular function of expressing a relation of identity or class membership (see (56, 58)), and expressing an attributive relation (57, 59), in addition to other copular functions such as expressing locatives. When it expresses a relation of identity, the PE is a noun such as la mg ‘old man’, and when it expresses an attributive relation, the PE is an attributive adjective such as mer ‘good’ or miäng ‘with flowers’ (derived from a noun with the attributive suffix).^[16] We also find auxiliaries with transitive inflection, particularly when the predicative expression refers to a property, as in (59). In essence, this yields a causative of a copula construction: ‘they made it (be) good.’

The copular enclitics are not attested with the ‘become’ sense, and neither do they occur with transitive inflection. But (56–57), for instance, do have counterparts with the 1|3sg present copula =nd, cliticized to a demonstrative form. At present, the exact factors that determine the use of the auxiliary-as-copula construction instead of the use of the enclitic form are not well understood. The permanency of the reference may play a role, in addition to whether the referent is available in the immediate physical or discourse context or not. This remains an area for further research.

6.1.5 Experiencer-object constructions

The construction termed experiencer-object construction is transitive, and contains two argument NPs, one referring to an affected entity and one to an affecting entity, in addition to an inflected auxiliary. The main difference between experiencer-object constructions and analytic constructions is that the former are not complex predicates. For them, the auxiliary is unambiguously the sole predicate head, and the other constituents are unambiguously nominal, serving as core arguments to the verbal predicate head, marked by a core case clitic when expressed as full NPs. The two construction types have in common that in both cases, the inflected verb is semantically empty, and primarily functions to license the morphosyntactic expression of a particular event type, by serving as a vehicle for verbal inflectional categories. The main lexical contribution comes from the NP referring to the affecting entity, and this can be said to be the “semantic head” of the construction.

Experiencer-object constructions typically convey states or changes of states, in which a human or highly animate participant (the experiencer) is affected by an entity over which they have no control. The experiencer is the O argument of the construction, whereas the affecting entity is the A argument. Typically, the O is highly animate but low in volition, whereas the A is inanimate or abstract. Experiencer-object constructions tend to have OAV constituent order, whereas in general, AOV is more standard. This construction type is widely attested in Papuan languages; see for example, Foley (1986: 121–124, 2018: 925–926) and Pawley and Hammarström (2018: 113–115).

Not many lexical verbs refer to bodily sensations and emotions, and experiencer-object constructions often convey these meanings. They employ the same auxiliaries as mentioned above: -r- ∼ -l- for present tense, and -g- or -nd- for the other tenses. Some examples are given below; (60) has the typical OAV order shown by experiencer-object constructions, while (61) shows AOV order.

From the examples, it is clear that the meaning of the construction as a whole is determined by the noun heading the NP referring to the affecting entity. Substituting dhndhg ‘meat’ for wot ‘garden food’ has the effect that now the construction specifies that the affected human participants are hungry for meat specifically.

The noun referring to the affecting entity is usually abstract, like ‘sleep’, ‘hunger’ or ‘thirst’, but can also refer to an animate being. Example (62) comes from a story in which the hunting protagonist encountered a Papuan black snake (Pseudechis papuanus), a venomous snake species notorious for its aggression. The storyteller relates how he was paralyzed with fear after he narrowly escaped, and how his companions described his condition:

OAV word order is also a minority pattern encountered in analytic constructions, as exemplified below. Reported speech is introduced by an analytic construction combining the semantically empty form ada ‘like this; thus’ with an inflected auxiliary. When both the A and O arguments, the speaker and recipient of the speech act, respectively, are humans, both OAV and AOV orders are found: example (63) shows OAV order and (64) shows AOV. Across the board, AOV appears to be more common in analytic constructions.

Nevertheless, rather than a clear-cut divide between experiencer-object constructions and “regular” transitive constructions, there is a gradient scale on which constructions can be placed with respect to the affectedness and degree of volitionality of a human O argument relative to that of the A argument, which has a bearing on constituent order. This is furthermore governed by information structure: topical O arguments tend to be fronted, even when inanimate.

6.1.6 Grammaticalization of analytic constructions

The previous sections have shown that there are two major ways of forming a verbal predicate in Idi: by a synthetic construction, in which a lexical verb stem is inflected directly, and by an analytic construction, a complex predicate in which a lexical stem is uninflected and an auxiliary form takes all the inflection. It was argued that the Idi analytic construction challenges the current typology of complex predicates, both because the uninflected element is not unambiguously verbal, and because the inflecting elements are only encountered as facilitating predication of forms that cannot do so otherwise, and not as full lexical verbs. We then went on to show that there are two further construction types with the same auxiliary forms that are not complex predicates: a copula construction in which a noun or adjective is predicated with the addition of an auxiliary form, and an experiencer-object construction in which a (change of) state of a human experiencer is expressed by a noun referring to the affecting entity, again predicated with the addition of an auxiliary. While in both cases the auxiliary is the syntactic head (without which predication would not be possible), the non-verbal element nearly completely determines the meaning of the construction. Because the same auxiliary forms are present in all three construction types, we question whether they are related and if they could have developed out of each other historically. All have in common a clear division of labor between the element that bears verbal inflectional marking, and the one that does not: the latter contributes lexical semantic meaning, while the former enables the morphosyntactic expression by a verbal predicate.

The constructions also differ in various ways. In the experiencer-object construction, the relevant NPs are marked as core arguments, and so the construction is not a complex predicate, but syntactically headed by the sole (auxiliary) verb. Auxiliary-as-copula constructions are a more difficult matter. Because analytic and copula constructions look the same, there are many borderline cases where it is unclear whether the non-inflecting element is primarily nominal or verbal. A case could be made quite easily for an utterance like (57) that it is a complex verbal predicate, and that miäng ‘with flowers; flowering’ functions as a non-inflecting verb, similar to e.g., mk ‘to chase’ in (53): while miäng is not attested with a synthetic (inflecting) stem counterpart, neither is mk. Analyzing (56) in the same way is a further stretch, but not impossible.

Apparently, the structural ambiguity with respect to the uninflected forms in Idi analytic constructions occurs not only because word class membership is difficult to determine across the board, but also because these constructions appear to be closely related to, and may have grammaticalized from, constructions that are more clearly bipartite noun-verb collocations, like the experiencer-object construction. Here, due to the preferred OAV constituent order, the NP in the A grammatical function usually occurs directly preceding the auxiliary verb. It seems a reasonable assumption that in some cases, this form may have been reanalyzed as part of a complex predicate, rather than as an argument to the auxiliary, particularly if it is used in this function more often than as a noun with clearly referential function.

This may have happened, for instance, with the form yéndhpä ‘to see’ (related to yéndhép ‘eye’), which even appears to have retained a fossilized core case marker. In contemporary Idi, yéndhpä plus an auxiliary can occur accompanied by two overt core arguments A and O, as was shown in (49) and (50). Thus, yéndhpa is not analyzed as a core argument to the auxiliary, but rather as a non-inflecting form that combines with the auxiliary into a complex predicate. It may have once been the affecting entity in an experiencer-object construction, however. While the exact grammaticalization process is unclear, another factor that makes such a process plausible is the fact that core arguments are often left unexpressed when retrievable from the discourse context. Thus, the following reanalysis could have happened:

When the NP with A grammatical function is dropped in (66) – which often happens in connected speech – the analytic construction is identical to a presumed experiencer-object construction that could be its source. This potential example is illustrative, because the form in question continued to exist independently as a noun with a slightly different meaning, which, if our inferences are correct, does not seem to have happened very often.

While we cannot be certain about the exact historical developments, the suggested grammaticalization route would pave the way for more forms with nominal origins, plus derived infinitives from synthetic verbal stems, to be recruited into the formation of complex predicates together with auxiliaries. This may have led to the current situation where this type of predicate is now a firmly established feature of Idi, and PR, morphosyntax.

6.2.1 Ditransitive indexing and animacy

Two of the morphemes, the tense prefix and the pluractional suffix, agree with whichever of the T and R arguments is more animate, according to the animacy hierarchy in (71).

In this hierarchy, first and second person arguments are equally animate and are more animate than third-person human referents, which are more animate than third-person non-human or inanimate referents. If the T and R arguments are equal in animacy, these affixes will index the recipient argument by default.

Example (72) illustrates how object-marking affixes in ditransitive verbs preferentially agree with R arguments as opposed to T arguments when the animacy and number are equal. In (72a), the tense prefix indexes second-person, corresponding with a second-person R argument, and in (72b), the tense prefix indexes first-person, corresponding with the first-person R argument.

However, if the T argument of the ditransitive verb ranks higher on the animacy scale than the R argument, the tense prefix will agree with whichever argument is more animate. In (73), the verb indexes the more animate second-person T as opposed to the third-person R argument.

In (73), all three O-indexing morphemes have a non-singular or plural value, indicating that they all index the first-person plural T argument.

6.2.2 Ditransitive indexing and number

The other two O-indexing morphemes—the O-prefix and the stem—do not index person, only number. Thus, these morphemes are not sensitive to the above animacy hierarchy. The O-prefix indexes the dual argument if either or both the patient and recipient are dual. The stem indexes the plural argument if either or both the patient and recipient are plural. Because the O-prefix and the stem are not sensitive to the animacy hierarchy, the grammar may generate forms with concordant agreement where all affixes agree, that is, when the plural or dual argument is also the most animate. However, this system also allows forms with discordant agreement, with some morphemes indexing the theme and others the recipient. This discordance may arise if the less animate argument is plural or dual.

The examples below show multiple giving scenarios, which show concordance or discordance among the object-indexing markers in the verb. The forms in (74a) and (74b) show concordant agreement with the T and R argument, respectively. In both examples, the most animate argument is also plural. In (a), the most animate argument is the first-person nonsingular theme, and all three object-marking affixes agree with the theme. This pattern is repeated in (b) for the more animate first-person nonsingular recipient.

Conversely, the forms in (75a) and (75b) show discordant agreement. In (75a), the tense prefix is zero-marked, indicating indexation with the R argument (second-person singular) while the O-prefix indexes the T argument (first-person dual). As stated earlier, the tense prefix will agree with the R argument by default, unless the T argument is more animate. The O-prefix, however, must agree with whichever object is dual, which in this case is the T argument.

Example (75b) is similar in that the theme and recipient arguments are equal in animacy. Again, the tense prefix and the unrealized pluractional suffix indicate agreement with the recipient by default. The dual-sensitive O-prefix is unrealized, indicating default agreement with the recipient as neither the theme nor the recipient is dual. Finally, the plural-sensitive stem agrees with the theme argument because the theme argument is plural, while the recipient is singular.

The examples in (76) differ from those of (75) in that the arguments now differ in animacy. In (76), the theme argument is inanimate plural while the recipient argument is second-person dual. The unrealized tense prefix and the unrealized pluractional suffix indicate indexation of the most animate argument: the recipient. The dual-sensitive O-prefix also agrees with the recipient because it is dual. On the other hand, the plural-sensitive stem demonstrates agreement with the inanimate theme argument because it is plural.

The discordant agreement in (76b) follows the same pattern. The animacy-sensitive pluractional suffix agrees with the third-person plural recipient argument as it is more animate than the inanimate theme argument. The plural-sensitive stem agrees with the only plural argument: again the recipient. However, the dual-sensitive O-prefix deviates from the rest of the form, indexing the dual theme argument.

This pattern whereby the stem is sensitive to the distinction between nonplural and plural while the prefix is sensitive to the distinction between nondual and dual is opposite to the pattern in Yam languages like Nen, where stems are sensitive to dual and prefixes distinguish singular and nonsingular (Evans 2014, 2017, 2019b). It is premature to conjecture on the source of this discordant agreement, but it is clear that languages in this region feature inflectional patterns by which differing categories of number are distributed across multiple exponents in the verbal form.