Our knowledge about tense, aspect and modality (TMA) in the Oceanic languages of Melanesia has so far been severely limited by the lack of available data. Habituality in particular, as one of the less described TMA categories, has not yet been widely discussed for this group of languages. Based on corpus data and elicitations, we give a detailed overview of four languages, identifying common trends and addressing specific questions of general concern. These include the relation of habituality to (im)perfectivity and the relation between habituality and irrealis.
The Oceanic language family roughly includes between 450 (Lynch et al. 2002) and 520 (Hammarström et al. 2017) individual languages. Most of them are spoken in Melanesia, specifically in Papua New Guinea, the Admiralty Islands, the Solomon Islands, Vanuatu, and New Caledonia. In contrast to some of the bigger Oceanic languages of Polynesia such as Samoan or Maori, most of the Oceanic languages of Melanesia are spoken by relatively small communities of speakers, often do not have a standardized variety, a written tradition, or official status. Accordingly, they are often comparatively under-documented.
It should therefore not come as a surprise that our knowledge about tense and aspect in this group of languages is rather fragmentary. At the same time, habitual aspect is one of the less described aspectual categories and many of the existing grammatical descriptions do not address this category explicitly. There are of course exceptions to this generalization. To name but two, Bril (2016) reports a marker kua for Nêlêmwa (New Caledonia), which appears to be restricted to habitual contexts, mostly of the past:
There are a few other Oceanic languages which have been reported to have markers that exclusively express habituality. For example, Bali-Vitu has a system of portmanteau subject proclitics that express TMA distinctions along with person-number features of the subject; this system is said to distinguish between the categories of realis, realis perfect, and realis habitual (Ross 2002a). The auxiliary or preverbal particle rere in Siar is said to specifically express habitual aspect (Ross 2002b), similar to the auxiliary fani in Kokota (Palmer 2002). Banoni and Port Sandwich have both been reported to have a post-verbal marker expressing habituality (Lynch and Ross 2002a, b). This picture seems to confirm the claim by Filip (2015) contra Dahl (1985) that dedicated habitual markers are not exceptional cross-linguistically, but rather more widespread than previously acknowledged. On the other hand, none of these languages have so far undergone sufficiently detailed investigations to confirm that the markers in question are in fact exclusively used for the expression of habituality.
In any case, in the languages under investigation, we did not find markers that expressed habituality to the exclusion of other perfective or modal meanings. We attempted to identify the main ways of expressing habitual aspect in our four subject languages, based on existing accounts, general and targeted elicitations, and especially corpus data. We found that the languages do show significant similarities in how they express habitual aspect, some of which may be characteristic of the related languages in this region more broadly. At the same time, we also found significant variation, which speaks to the diversity of Oceanic languages of Melanesia. The main means of expressing habituality in the four subject languages are:
Auxiliaries deriving from a verb with the meaning stay;
In Mav̋ea: an imperfective affix.
Crucially, in all four languages, habituality does not have to be expressed explicitly.
While a full theoretical exploration of our findings is outside the scope of this paper, our results offer an interesting window on some of the questions that have long accompanied the study of habituality. In the next section, we will briefly give some background on these theoretical questions and explain how our findings relate to them. We will then give an overview of the subject languages and our methodology. In Sections 4 to 7, we will discuss each of the subject languages in detail. We summarize our results and draw conclusions in Section 8.
2 Theoretical background
In this section, we want to clarify some of the terms we use and give some back- ground of the theoretical questions to which our findings are potentially relevant. In particular we will comment on the following points:
the definition of habituality and genericity;
the relation between habituality and the distinction between perfective and imperfective aspect;
the relation between habituality and the distinction between realis and irrealis mood;
the relation between habituality and tense.
First of all, we would like to comment on the term habituality. Comrie (1976: 27) suggests the following definition:
The feature that is common to all habituals […] is that they describe a situation which is characteristic of an extended period of time.
Krifka et al. (1995) define characterizing sentences, which comprise habituals, as
propositions which do not express specific episodes or isolated facts, but instead report a kind of general property, that is, report a regularity which summarizes groups of particular episodes or facts.
Many authors distinguish between genericity and habituality. However, the distinction between the two terms is by no means trivial and varies between authors and linguistic subfields. According to Krifka et al. (1995), there are two types of generic statements: (1) those referring to kinds as The potato was first cultivated in South America and (2) generalizations over events as John smokes a cigar after dinner. This latter type of generic statement is also often described as a habitual. In this sense, habituals may be understood as a subclass of generics.
By contrast, in Dahl (1985), the distinction between generics and habituals is the distinction between what is generally the case and what is usually the case (also compare Dahl 1995). In a similar vein, Boneh and Doron (2012) distinguish between a habitual operator Hab, which is an existential quantifier and a generic operator Gen, which is a universal quantifier.
For this article, we took all these distinctions into consideration, but did not find them to play a major role in our subject languages. As far as we could find, statements about kinds can be expressed by the same means as statements about habits of an individual; and whether or not a characterizing statement may have exceptions does not appear to play a major role in how these meanings are expressed. On the other hand, we did not systematically elicit data for all our subject languages to exhaustively test the impact of these distinctions. So far, we can only conclude that they do not figure prominently in the expressions of genericity and habituality in our subject languages. As a result, we do not systematically distinguish between kind-referring statements such as snakes eat small birds and mammals and characterizing expressions such as Mary works in her field every day. Moreover, while we considered both stative and non-stative event descriptions, the focus of this article is on non-stative ones.
The main phenomenon under investigation may be defined as ‘descriptions of regularly recurring events that characterize the behavior of individuals or kinds over a certain period of time’. This is not meant to be a theoretically strict definition, but only as a working characterization that helped us identify relevant contexts.
In this article, we will frequently use the term habitual aspect to refer to these types of statements. By doing so, we do not intend to take a specific stand on the exact relation between habituality, aspect and mood. Starting with the relation between habituality and aspect, there is considerable disagreement in the literature. While Comrie (1976) classifies habituality as a special case of imperfective aspect, most later authors have leaned towards to a more complex assessment of the situation. Thus, Dahl (1985) treats habituals and generics as their own aspectual category, not subordinate to imperfective. Citing Mønnesland (1984: 54), Dahl (1985: 79) stresses the hybrid nature of habitual aspect in terms of its perfectivity:
one can use a pf. verb, thus stressing each individual total event, or use an ipf. verb, which means that the stativeness of unlimited repetition takes precedence’. The first solution is normally chosen in Russian, Polish, and Bulgarian, whereas the second is preferred (even if it is not always the only possible alternative) in Czech, Slovak, Sorbian (a West Slavic language spoken on the territory of the German Democratic Republic) and Slovene. In Serbo-Croatian both aspects are possible.
A similar view is taken by Hengeveld and Mackenzie (2008). Arche (2014) analyzes habituals as imperfective descriptions of series of perfective, or bounded, events (also compare Ferreira 2016 for a related view). Filip (2015) proposes that habituality, or characterizing genericity is not a subcategory of tense or aspect at all, but constitutes a separate and independent category.
In our study, we found that habituality is often expressed by the same means as other types of imperfective aspect, in particular auxiliaries that have developed from verbs meaning stay, reduplication and an imperfective aspectual affix in the case of Mav̋ea. This strengthens the position that habituality is a special case of imperfective aspect. At the same time, we also found some intriguing support for the view that habitual aspect is semantically more complex than imperfective or perfective aspect alone: In particular in Daakaka and Mav̋ea, reduplication is frequently combined with an additional marker of imperfectivity to express habituality.
Another question that is frequently and controversially discussed concerns the reality status of habitual statements.
As Givón (1994: 270) puts it:
The status of the habitual, a swing modal category par excellence, is murky for good reasons. From a communicative perspective, habitual-marked clauses tend to be strongly asserted, i.e. pragmatically like realis. Semantically, however, they resemble irrealis in some fundamental ways. To begin with, unlike realis, which typically signals that an event has occurred (or state persisted) at some specific time, a habitual-marked assertion does not refer to any particular event that occurred at any specific time. Further, the reference properties of NPs under the scope of habitual resemble those of NPs under the scope of irrealis.
The last of these observations is responsible for the classification of habituals as non-veridical by Giannakidou (1995). Cross-linguistically, these considerations are reflected by the fact that many languages use markers associated with irrealis to express habituality (see Cristofaro 2012 and references therein for a comprehensive overview). In our subject languages, the distinction between realis and irrealis statements is relatively prominent. We find the murkiness diagnosed by Givón (1994) reflected by the fact that, in all four subject languages, habitual statements can occur both in realis (or unmarked) environments and in utterances specified for irrealis.
Lastly, various authors have discussed the relation between habituality and tense. Dahl (1985) and others have observed that quite a number of languages exhibit specific markers of habituality that are restricted to past contexts such as English would or used to (even though Binnick 2006 has argued that used to does not necessarily express habitual aspect at all). We could not find markers that were specific to expressing habituality in the past. While few publications comment on habitual aspect in future contexts, we tried to find such contexts too and have included our findings in the article. We did however not find anything unexpected in these contexts.
3 Data and methodology
The languages in this study all belong to Oceanic group within the vast family of Austronesian languages. They are all spoken in Melanesia, either in New Guinea (Saliba-Logea) or in Vanuatu. The rough locations of their main speaker communities are indicated in Figure 1. Their family relations are sketched in Figure 2.
All the subject languages have certain structural properties in common, some of which are listed below (also compare Dunn et al. 2008):
serial verb constructions;
reduplication of verbs (and sometimes nouns);
a distinction between inclusive and exclusive person features;
a rather tight-knit predicate structure obligatorily including a prefixed or pro-clitic subject agreement marker and the verb root, which may optionally be reduplicated; Depending on the language, this structure may be augmented by different types of mostly preverbal tense, aspect and mood (TMA) markers;
grammatical differentiation between alienable and inalienable possession. Daakaka, Mav̋ea and Saliba-Logea have an additional system of possessive classifiers, but Nafsan does not.
However, there are also significant differences between the subject languages. Saliba-Logea in particular differs from the Vanuatu languages in two important ways.
Firstly, the basic word order of Saliba-Logea is SOV, while the Vanuatu languages share the order of SVO. Secondly, in Saliba-Logea, TMA marking is optional, while in Daakaka and Mav̋ea, every assertive sentence needs to contain a marker conveying some TMA information. For Nafsan, our research indicates that the subject proclitics that have been described as simultaneously encoding realis are in fact neutral with respect to TMA information. In this sense, it can also be said to have optional TMA marking, like Saliba-Logea.
Moreover, while all the languages in this study share a clusivity distinction, their pronominal systems and in particular their paradigms of subject agreement markers still differ quite significantly: Daakaka and Mav̋ea have four number distinctions (singular, dual, paucal, plural), whereas Nafsan has three and Saliba-Logea has two. In Mav̋ea and Nafsan, at least some subject agreement markers are portmanteau morphemes which also encode TMA information, but not in Daakaka and Saliba-Logea.
We will also see that all the languages in this study differ significantly with respect to their paradigms of TMA markers.
For this study, we explore corpora that have been created from documentary fieldwork for each language. The corpora we use are the following:
Daakaka: von Prince (2013)
Nafsan: Thieberger (2006a)
Mav̋ea: Guérin (2006)
Saliba-Logea: Margetts et al. (2017)
There are several stories, themes and genres that are shared widely across this region. This allowed us to identify comparable contexts across corpora that we looked at in detail. They include accounts of how life used to be in the past as compared to today; traditional stories about how certain animals came to exhibit the characteristics we see today; and instructions about how to perform certain techniques and crafts.
We have imported these corpora from their native SIL Toolbox format to the ANNIS platform hosted by Humboldt-Universität zu Berlin for improved facilities for searching and analysis (Zipser and Romary 2010; Druskat 2018). We rely on corpus data, along with elicitation data from generic questionnaires such as the one in Dahl (1985) and existing descriptions.
Last but not least, some of us have already completed elicitations based on storyboards that target specific TMA-categories. We will report on the results from one particularly relevant context for Mav̋ea, Nafsan and Daakaka. This context comes from Vander Klok (2013), which is a storyboard about Bill who always forgets his stuff on his way to work. The corresponding picture is shown in Figure 3.
Since more fieldwork targeted specifically at habitual aspect has not been carried out so far, and since all four subject languages are still relatively under-documented, it is possible that our findings do not exhaust the possibilities of expressing habitual aspect in each language. They do however offer a reliable view on frequent expressions and general trends.
Daakaka is spoken by roughly one thousand speakers most of whom live in the West of the island of Ambrym, Vanuatu. The facts reported in this section were first described in von Prince (2015). The core verbal complex consists of a subject agreement marker, which encode four number distinctions and four person distinctions; a clitic encoding TMA and polarity values of the clause; optionally an imperfective auxiliary; the verb root, which may be reduplicated; optionally a resultative suffix and/or a transitivizing enclitic. The main verb can also be followed by one or more serial verbs. Table 1 gives an overview of the basic structure of the verbal complex.
|na, …||=m, …||du,pwer||…||…||…||=ne|
The following sentence illustrates a verb complex with an auxiliary and resultative suffix in addition to the obligatory elements, which are printed in bold:
Table 2 shows the system of subject agreement markers. The paradigm of TMA markers is given in Table 3. Their exact shape depends on their environment. They can be enclitic to a preceding subject agreement marker or proclitic to the proceeding verb if it starts with a vowel. In the case of a third person singular or inanimate subject, there is no subject agreement marker to cliticize to. In this case, the TMA markers are realized as monosyllabic words, where the vowel is determined loosely by the subsequent verb. Table 3 shows the paradigm of TMA and polarity markers.
|Change of State||bwet|
In Daakaka, a simple assertion about the episodic past or present is marked only by the realis marker:
The same form can also be used for habitual and generic clauses, as illustrated by the following example:
This is however not the most common way to relate habitual meanings. Far more frequently, habitual meanings are expressed by an imperfective auxiliary in combination with a reduplicated verb form, or with only one of these two forms.
At the same time, no single expression in Daakaka can be claimed to be canonical for habitual aspect. When von Prince ran the Vander Klok (2013) storyboard, five different speakers produced the habitual context in five different ways, ranging from an unmarked structure with simple realis as in (6), over the auxiliary mas, over a subordinate structure with kuowilye that usually describes abilities, to the imperfective marker bwe and finally a reduplicated verb:
All these different expressions are explored in more detail below.
4.3 Aspectual auxiliaries
Daakaka has two imperfective auxiliaries, pwer and du, which also have main verb counterparts with the meaning stay, be at. The difference between pwer and du is that pwer is specific to singular subjects, while du can only be used with plural subjects. 
The auxiliary pwer is often shortened to pwe and often contracts with a preceding realis marker m- to the form bwe(r) [ᵐbʷɛr]. These auxiliaries are not specific to habitual contexts but can express other types of imperfective aspect as well. They frequently express progressive aspect:
Example (12) shows a non-generic habitual relating to the working routine of the subject.
In the following example, the imperfective plural auxiliary du is used to express a generic property of a bird species called eya, ‘white-eye’.
In addition to those aspectual auxiliaries, Daakaka also sometimes uses the modal auxiliary mas to mark habitual contexts, which is described in more detail in Section 4.5.
As in many Oceanic languages of the region, reduplication in Daakaka is very productive and can have a wide range of functions. It can derive attributes from verbs and express various kinds of pluractionality – this can imply that an action is performed repeatedly by the same actors, or that it is performed on various patients or by various agents (von Prince 2015).
Reduplicated verbs without further aspectual marking can occur in habitual contexts, but typically only in contexts where the plurality of the subject alone is enough to license its use:
But there is reason to believe that the reduplication process does contribute to the habitual interpretation. Crucially, in negative habitual contexts, reduplication also occurs quite often, even with generic singular subjects such as in (15):
However, the most frequent structure to be found in habitual contexts features a combination of the imperfective auxiliaries pwer and du described in Section 4.3 and a reduplicated verb form. Also in reverse, such combinations of the imperfective auxiliaries and reduplicated verbs typically express a habitual meaning. To the extent that there is any structure specific to habitual aspect in Daakaka, this would be it.
There is no structural distinction between kind-referring generic contexts such as (16) and more narrowly habitual ones such as (17). In (16), we see a context that describes general features of a species (hawks) and we find a combination of an imperfective auxiliary and a reduplicated verb form. In (17), we find the same structure expressing a regularly recurrent action by concrete individuals.
Also, the same structure can be used for habitual assertions about the future as well as habitual descriptions of the past or present. Both of the following sentences come from a story which explains how the megapode and the chicken, who used to be very good friends, came to look so different and lead such different lives. The first of the following two examples is a stretch of direct speech by the megapode to the chicken, talking about the future. The second sentence then describes the current state of affairs.
Another example for a habitual description, with auxiliary and reduplication, referring to the future is given below. Like other future habitual contexts, this one is non-factive.
The same combination of auxiliary and reduplicated verb form is typically used in contexts of a past state of affairs that no longer holds. In these cases, the distal marker t indicates that what used to be the case is no longer the case (von Prince 2017).
4.5 Other means of expressing habitual aspect
Apart from the expressions explored so far, Daakaka has an inventory of adverbials such as webung kevene ‘every day’ and the (partial) loans from Bislama taem kevene ‘every time, always’ and oltaem ‘always’ to explicitly mark the regular re-occurrence of an event.
It also borrows the auxiliary mas ‘must’ from Bislama to express law-likeness: 
Another option that we find in Daakaka is the use of the subordinating verb kuowilye ‘know’, which can generally be used to express possibility or ability. This is described in the following section.
4.6 Habituality and irrealis
Most habitual contexts in Daakaka are marked as realis. We have seen in (21) above that the distal marker is frequently used in contexts of habitual states of affairs in the past that have ceased to continue into the present – these contexts are frequently translated into English by would or used to. The distal marker also expresses counterfactuality in conditionals and complement clauses. In this sense, one might say that past habituals are associated with irrealis mood in Daakaka. However, von Prince (2017) has argued that the distal can refer to the actual past as well as to counterfactual developments and that the discontinuity reading is a result of its contrast to the realis marker. In this sense, it would be wrong to take examples such as (21) as evidence for a connection between habituality and non-reality, since the distal marker in these cases refers to actual events.
The main non-realis marker in Daakaka is the positive potential marker, along with its negative counterpart. These can also be found in habitual contexts quite frequently. In most cases, this has to do with the indicative conditionals and generic temporal clauses that are often to be found in these environments. However, in some cases the use of the potential mood marker is not conditioned by these structures but appears to contribute to the habitual reading. In the following example, the potential mood marker alternates with the realis marker in the description of a ritual in which a piece of wood is made to move:
Another quite common situation is illustrated in the next example: Here, the habitual use of the treefern leaves as kindling is described as a generic possibility. This possibility is explicitly referred to as such by the verb kuowilye ‘know, be possible’. When kuowilye expresses possibilities rather than knowledge, its complement clause is always headed by the potential mood marker. In the following example, the potential marker encodes a sequence of clauses as complements to the possibility expressed by kuowilye, before the speaker switches back to realis mood.
In sum, the default mood for habitual contexts in Daakaka is realis. Past habituals are often marked by the distal mood marker. The potential mood occurs in stretches of discourse, typically in the context of conditionals and complement clauses, but sometimes also in unembedded clauses, where it alternates with realis mood.
Nafsan is also known as South-Efate, after the region of Efate island in Vanuatu where it is spoken by about six thousand speakers. The data used in this article were collected in Erakor village, situated on the outskirts of Port Vila. Grammatical elements with TMA values in Nafsan can occupy different morphosyntactic positions, typically preceding the verb. The set of these morphosyntactic slots is frequently referred to as preverbal complex in Oceanic languages. Thieberger (2006b) offers the composition of the preverbal complex as shown in (25). ‘Sub’ refers to subject agreement markers usually called subject proclitics and they are the only obligatory marking of the verb. The subject proclitics are also portmanteau morphemes that carry TMA values, at least in the case of the irrealis and perfect proclitics. They cliticize to the next available element, and this can be a TMA marker, an auxiliary verb, a benefactive phrase, or the verb.
The subject proclitics are divided into three paradigms given in Table 4. Each paradigm can combine with specific TMA markers, as shown in Table 5.  The auxiliary verbs, on the other hand, do not seem to pose any restrictions on the choice of subject proclitics.
|1pl (incl)||tu=||tuk=||tu=, tui=, tukoi=|
|1pl (excl)||u=||ko=||ui=, koi=|
As in the other subject languages, a sentence in Nafsan need not have specific aspectual marking to express habituality. The same form that is typically used for episodic past and present states and events can also be used for habitual meanings.
In Nafsan, the general form of subject proclitics functions as an agreement marker that gets its TMA interpretation from the context of the story. It has also been called ‘realis’ by Thieberger (2006b), because of its frequent usage in contexts with past and present reference. In (26), we can see that the event of ‘walking to prayers’ receives a habitual interpretation from the context, but there is no specific grammatical marker that would be the source of the habitual interpretation.
The general subject proclitic without any further marking can also be used for habitual present contexts:
5.3 Aspectual auxiliaries
In Nafsan the verb to ‘stay’ can function as an auxiliary verb with progressive, habitual, and generic meaning. Example (28) shows to with a progressive interpretation.
Past habituals can be set in a context of a narrative taking place in the past, like (29). The past reference can also be more explicitly marked by different adverbials, like malnen ‘then’ in (30). In (30), we can also see that the predicates marked by to establish a habitual reference, while the last clause remains grammatically unmarked for habituality. The adverbial sernrak ‘always’ and the general form of the subject proclitic on the verb are sufficient in this context for the habitual interpretation.
Besides past habituals, we can also find to in present habituals as in (31), (32) and (33).
The auxiliary to is also prominently used in generic, kind-referring predicates. In her recent fieldwork, Ana Krajinović had the opportunity to create a small riddle for children which characterizes different animals. Each page describes an animal without saying which animal it is and at the bottom of the page there is the question ‘Who am I?’ The book contains riddles about 6 animals. The auxiliary to was consistently used to express generic statements with non-stative verbs. The following examples come from the description of a gecko.
In Nafsan, reduplication does not appear to be as productive as in the other subject languages. According to Thieberger (2006b), reduplication can be used to express pluractionality or intensification, but many of the corresponding examples, the relation between the base form and the reduplicated form is not as transparent as one might expect, as in mar ‘breathe’ vs. marmar ‘rest’. In other cases, reduplication can be used for detransitivizing a verb root or for nominalizing it. In many cases, however, the difference between the bare verb root and the reduplicated form is either entirely obscure or not perceivable at all.
In our study of Nafsan, reduplication could not be found to play any significant role in the expression of habituality.
5.5 Other means of expressing habitual aspect
Like Daakaka, Nafsan occasionally uses the auxiliary mas from the homophonous Bislama word which roughly translates as ‘must, always, absolutely’. In the corpus data, it is quite rare and not attested in habitual contexts. But in the elicitation based on Vander Klok (2013), two out of seven speakers used mas to express the regular recurrence of Bill’s forgetfulness, as illustrated by (38):
Nafsan can also use the verb tae ‘know, can’ to express habituals. In the above mentioned riddle ‘Who am I?’ the verb tae was used to express a generic statement about a girl from Vanuatu. This is a part of the description:
5.6 Habituality and irrealis
In Nafsan, we often find irrealis proclitics in habitual contexts. Kind-referring generics seem to prefer the general proclitic and the auxiliary to.
Another important part of the irrealis marking of habituals is that besides the irrealis proclitic they seem to require the posterior irrealis marker fo. The main function of fo is marking the predicate as posterior to a given reference time.
The usage of irrealis in past habituals seems to be related to the basic irrealis function of expressing possibilities or predictions in the relative future in Nafsan. One reason to draw this conclusion comes from comparing the irrealis-marked habituals with potential-marked habituals. In the same story from which example (40) was taken, we find the potential marker fla expressing a possibility in a habitual context, where it is clear that those possibilities had to be recurrent in the temporal frame that was set in the earlier example (40).
In example (41) we can see a sequence of habitual events with the first two clauses marked by the marker fla and the last one with irrealis and the posterior fo.
Although habitual statements in irrealis mood mostly refer to past contexts in the corpus data, we can find cases with present or kind-referring meaning:
The same combination of irrealis and posterior marking is also used for future habitual contexts:
One out of seven speakers with whom Ana Krajinović ran the storyboard by Vander Klok (2013) produced irrealis forms to express a habitual context:
Mav̋ea, also known as Mavea or Mafea is spoken by about 30 people on the eponymous island of Mav̋ea off the island of Santo, Vanuatu. Assertive clauses generally contain a close-knit verb phrase that can contain a variety of TMA and polarity markers, in addition to subject agreement markers. The basic structure as described by Guérin (2011) is shown in Table 6.
The following example shows a verbal complex that incorporates a fairly large number of the above categories:
For singular subjects, there are two lists of subject agreement markers, depending on whether a sentence expresses realis or irrealis. The full paradigm of subject agreement markers is shown in Table 7.
Among the verbal prefixes, we will find the imperfective prefix l(o) to be particularly relevant for the expression of habituality. In fact, l(o) is the most important means of expressing habituality in the language, while auxiliaries and reduplication play a subordinate role at best.
The prefix m̋e- has been described as expressing iterative aspect, which one might suspect to also play a role in related aspectual contexts. However, it appears both from the corpus data and from the description in Guérin (2011) that m̋e- literally means ‘again, back’ or ‘not anymore’ in combination with negation, rather than expressing iterativity more generally.
For the examples here, we cite the morpheme-level of annotations from the corpus data, which does however not always correspond to the underlying forms.
In Mav̋ea, assertions about bounded past events typically contain no other verbal prefixes than the (realis) subject agreement marker. This is illustrated in (46):
The same form can also be used for habitual descriptions, even though this form is less frequent in habitual contexts than a verbal complex containing an imperfective marker, which is described in more detail below. The following sentence reports habitual activities of the people of Mav̋ea, where the first two predicates are prefixed with the imperfective marker, while the last one remains unmarked for aspect.
Similarly, in the following sentence, the first predicate contains the imperfective marker, but the second one does not:
6.3 Aspectual auxiliaries
Like the other Vanuatu languages in this study, Mav̋ea also has a verb to ‘stay, be at’. This verb is not among the three auxiliaries described by Guérin (2011: 80), which are adi ‘can’, leng ‘cannot’ and ria ‘must’.
Even so, it appears that to can also serve as an auxiliary in that it can occur before a verb without interfering subject agreement markers:
These cases are however too rare to draw definite conclusions from. They might be related to structures involving the verb tur(u) ‘stand up’, that are described as serial verb constructions expressing duration in Guérin (2011: 274).
In a much more common pattern, to is followed by a fully inflected second predicate. This second predicate can have a wide range of different meanings such as ‘go’, ‘speak’, ‘look’, ‘cry’ and ‘take’. Even though this structure does not correspond to a canonical auxiliary structure, the function of to in these environments may well be described as that of an imperfective auxiliary. In many such cases, to expresses a continuous or progressive aspect. This can be seen from the following examples:
It also occurs in habitual contexts. This can be seen in the following example:
In addition to these aspectual structures, there is also the aforementioned modal auxiliary adi ‘can’, which can occur in habitual contexts. This is described in more detail in Section 6.5.
In sum, Mav̋ea can be said to have an auxiliary to which has developed diachronically from a verb meaning stay, be at. This auxiliary is apparently cognate with Nafsan to and Daakaka du. It also expresses imperfective aspect and can occur in habitual contexts, although its role in expressing habitual aspect is probably marginal in comparison with the imperfective prefix lo described in Section 6.5.
intensification (with stative property-denoting verbs)
continuous aspect of stative verbs
derivation of nouns and adjectives
Reduplicated verb forms occur in habitual contexts, at least in kind-referring generic ones. Since kind-referring generic statements always range over individuals, it is not entirely clear that the reduplicated form in these cases is not a way to express the plurality of the subject, even though in many of these cases, the subject agreement marker encodes a singular rather than a plural subject. At the same time, you can see in both of the following examples that reduplication in these contexts is not obligatory, at least not for stative verbs such as to ‘stay’ and suruvu ‘sleep’.
6.5 Other means of expressing habituality
As described in Guérin (2011: 228), the expression that is most strongly associated with habitual aspect in the language is the imperfective prefix lo-. This morpheme can also express other imperfective aspects than habituality. Thus, the following example shows the use of this prefix in a progressive context:
Also, as we have seen above, not all habitual contexts require this prefix. But in the vast majority of habitual contexts, we find the prefix lo-. Guérin (2011) already describes it as a marker of habituality, citing elicited kind-referring examples. In the corpus, we also find the imperfective prefix lo- in many kind-referring contexts. The following two example sentences come from a story that is similar to the one about the chicken and the megapode mentioned before in Section 4. This story is about Parrot and Flying Fox, who used to be good friends. Then one day, they painted each other and Parrot got wonderful bright colors, but Flying Fox only got black. At the end of the story, the persistent result of this event is described in generic terms. We find the imperfective prefix lo- with all non-stative predicates here, as well as reduplication in some of the same predicates. As mentioned in Section 2, this combination of two aspectual expressions might be taken as an indication that habituality is more complex than either imperfectivity or pluractionality.
The same means of expressing habituality are also used for non-generic habituals and irrespective of temporal reference to the past or future. Thus the following sentences are taken from the same text. The first sentence expresses a directive speech act directed toward the future. The following two sentences describe that the command was obeyed, so we see a past reference. In both cases, the imperfective prefix lo- is used, sometimes in combination with to ‘stay’. Note that we are dealing with a non-generic habitual context in each case.
Cases of discontinuous past are also not grammatically differentiated from habitual or generic contexts of the present. The following sentences are from the familiar story about Megapode and Chicken. Megapode complains about Chicken, who has painted him only in dull black and reflects on the good old times which are now therefore over. Again, we see the imperfective prefix as a marker of habituality here:
Finally, in addition to the aspectual auxiliary to described in Section 6.3, there is a modal auxiliary that can be used to express habitual aspect. This is adi ‘can’, which mostly expresses abilities and circumstantial possibilities. In the corpus, it is not very frequent and does not occur in habitual contexts. However, in recent elicitations by Valérie Guérin based on Vander Klok (2013), three independent speakers used essentially the same structure to describe the forgetfulness of the story’s protagonist Bill, as illustrated below:
6.6 Irrealis and habituality
Mav̋ea has a distinction between realis and irrealis subject agreement markers, but only in the first and third person singular. In most habitual contexts, the realis set of subject agreement markers are used. To the extent that irrealis markers also appear, this is mostly conditioned by a reference to the future or similar independent factors, as illustrated in (64):
Saliba-Logea is a Suauic language of the Papuan Tip Cluster in Milne Bay Province, Papua New Guinea. Logea is a closely related dialect. Both varieties are named for the islands on which they are spoken, by a rough total of 2500 speakers (Margetts 1999). For the purposes of this paper, we do not differentiate between the two varieties, but refer to both collectively as Saliba-Logea. The verb root is always preceded by a subject agreement marker and potentially by a number of prefixes. The most complex preverbal structures we find in the corpus contain a combination of two prefixes. A sketch of the attested structures is given (65):
The system of subject agreement markers in Saliba-Logea is given in Table 8. 
Table 9 shows the optional verbal prefixes in the language.
|he-||caus-||a causative marker|
|kai2-||body.weight-||a classificatory prefix specifying that an action involves the body|
The verb complex can contain several suffixes, including suffixes marking person and number features of the object. Some of the most frequent suffixes excluding object suffixes are given in Table 10.
|-ma||-to.sp||directional, towards speaker|
|-wa||-to.ad||directional, towards addressee|
An example of a complex verb structure is given in (66):
The prefixes in Saliba-Logea primarily manipulate argument structure rather than aspect, by detransitivizing the verb root or by transitivizing it via a causative prefix. Serial verbs such as gehe ‘finish’ and suffixes like the perfective marker -ko can add aspectual information to the verbal predicate. There are however no suffixes or serial verbs that would generally express imperfective aspect.
In Saliba-Logea, TMA marking is generally optional. Assertions about the episodic past do typically not receive any particular TMA marking:
Likewise, the absence of any TMA marking appears to be the default for habituals in Saliba-Logea. The following sentence is taken from a description of how to make and use baskets. Like most of the text, it is unmarked for specific TMA values:
Another example comes from an explanation of the properties and uses of the sebulu pandanus:
7.3 Aspectual auxiliaries
Saliba-Logea has two verbs with the meaning ‘stay’, miya and bawa. Neither of them is used to mark habitual aspect or other imperfective aspects. Nor are there other imperfective auxiliaries in the language. There are sequences of verbs, and especially directional serial verb constructions are very common. As mentioned in the introduction to Saliba-Logea, some serial verbs such as gehe ‘be finished’ can add aspectual information to the verbal predicate, but they encode perfective rather than imperfective meanings. Other verbs that are likely candidates for diachronic sources for the expression of habitual aspect (Bybee et al. 1994) such as tuli ‘sit’ or naya ‘wait’ are never used with aspectual meanings in the corpus, even when they occur as parts of complex predicates. The closest thing to a serial verb construction with an imperfective meaning are structures where a verb is followed by the inflected verb lau ‘go’. This type of structure is very common in all the subject languages and typically expresses the continued duration of a certain state of affairs and the passage of time in a narrative:
The same effect is achieved by the discourse particle ee which indicates duration of the event expressed by the preceding verb. It can be drawn out to iconically express varying length of time. It does not express imperfectivity as such, but rather continued duration of a certain state of affairs and typically has a telic reading which is translated as ‘on and on until’. The following example illustrates both expressions in combination:
In short, auxiliaries cannot be used in Saliba-Logea to encode habituality, and neither can serial verbs.
In Margetts (1999), reduplication is described as expressing progressive aspect and deriving nouns, attributes or adverbs from verbs.
Mosel (1994: 28) describes reduplication as expressing habitual and progressive aspect, citing the example in (72) to illustrate the former:
More specifically, reduplication in Saliba-Logea has different aspectual meanings for different verbs. There is no clear-cut stative-active distinction in Saliba-Logea but verbs from the different ends of the spectrum tend to behave differently in terms of their semantics of the reduplicated forms. Some stative verbs simply do not allow stem reduplication and for others the reduplicated stem has a reading of a temporary state of being or inchoative. With some stative verbs reduplication has a habitual reading while the unreduplicated form is understood to express a temporary state, such as for gwauyala ‘happy’ in (73) and (74).
The root pitali ‘dry’ allows a habitual reading of the reduplicated stem, as in (76), which was suggested as a statement about a type of synthetic cloth (e.g. ‘this synthetic shirt is always dry, even when it rains’). Alternatively the reduplicated form can have a lexicalized meaning of ‘dryish’, i.e. ‘damp’.
The simple stem yababa ‘bad’ in (77) expresses that the engine is broken. While this may be a permanent state, it is clearly not intended to be, and the implication is that the engine may be fixed or replaced. In contrast, the reduplicated form of yaba-yababa in (78) has a reading of being generally unreliable.
In the context of a human subject, yababa ‘bad’ typically receives a stative reading instead of an episodic reading as in (77). In these cases, the reduplicated form was rejected in elicitation.
The verb bawa ‘stay’ occurs much more often in its reduplicated form babawa or bawabawa than it its simple form. This reduplicated form can be used in episodic contexts, as shown in (81):
But many of the contexts where the reduplicated form is found are habitual:
The unreduplicated form refers more frequently to episodic contexts:
In sum, reduplication in Saliba-Logea can have a variety of functions, depending in part on the lexical meaning of the verb. Habituality is one of those functions.
7.5 Other expressions of habituality
Saliba-Logea is known for using nouns as predicates (Mosel 1994; Margetts 1999). These nominal clauses are usually stative and can also be used to express habitual and generic meanings. Two examples are given below:
7.6 Habituality and irrealis
As mentioned above, finite sentences in Saliba-Logea do not contain any obligatory TMA marking at all. The default for a sentence without TMA expressions is to refer to the actual past or present, but a reference to the future or to possible alternatives to the actual world are also readily available in a corresponding context. In this sense, Saliba-Logea does not strictly implement a distinction between realis and irrealis contexts.
There are some markers that express modal meanings typically associated with irrealis, such as taba, which is used in conditionals and in talking about past possibilities, and bena, which is used to talk about the future, possibilities and obligations. These expressions can often be found in habitual contexts.
In most cases, they occur in (semantically) subordinate environments that require their presence irrespective of the habitual interpretation. Thus, generic conditionals and expressions of dispositions and possibilities are bound to play a significant role in habitual contexts. But the irrealis TMA markings we find in these environments would have to be there even if the context was not a habitual one. This is illustrated by the occurrence of bena below:
However, in some cases, they may be seen as contributing directly to a generic or non-generic habitual reading, as illustrated by the following generic description of the properties of a certain plant:
All four subject languages have certain features in common concerning the expression of habituality:
Habitual contexts are not obligatorily distinguished from assertions about the episodic past or present. In all the languages under investigation, marking of habitual assertions is optional. This means that there is no obligatory differentiation between sentences that are about the episodic past or present – which is the default interpretation of sentences otherwise unspecified for TMA (cf. Smith et al. 2007).
There is no differentiation between generic and non-generic habituals.
Aspectual marking of habitual contexts, if any, is not distinguished from other pluractional and imperfective contexts. The same markers that are used to signal that a sentence is not about a single bounded event in habituals can also be used for related meanings such as progressives or iterative.
The primary means to express habitual and related meanings are aspectual auxiliaries, TMA affixes, and reduplication. The languages differ however in terms of which expressions are available for which meaning.
The expression of habitual aspect is independent from temporal reference. Habitual contexts of the past are encoded in the same way as habituals of the present and future.
Table 11 gives an overview of the main ways to express habitual aspect in each language. Note that more marginal expressions of habituality are not included here.
What is particularly interesting in the light of the ongoing debates about the nature of habituality is the observation that some of the languages in our sample combine two different means of aspectual expressions to yield a habitual interpretation.
As we have seen, in Mav̋ea, the default way to express habituality involves the verbal prefix lo-, which simply expresses imperfectivity, including progressive aspect. In habitual contexts, this prefix is sometimes combined with a reduplicated verb root. The example below is repeated from (56):
Similarly, the most frequent way to express habitual aspect in Daakaka involves reduplication of the verb in combination with one of the imperfective auxiliaries pwer and du. The example below is repeated from (16):
These observations might cautiously be taken to support the view that habitual aspect might be semantically more complex than some other aspects.
Our findings also support Givón (1994)’s initial assessment that habitual aspect is ambiguous in terms of its modal associations – we find it expressed both in environments marked as realis; and in contexts featuring modal auxiliaries or adverbs and irrealis TMA markers or portmanteau subject proclitics.
We would like to thank the editors for prompting us to write this article. We would like to thank two anonymous reviewers for their valuable feedback and suggestions. And we would like to thank the DFG (German Research Foundation) for partially funding this research (MelaTAMP project, 273640553).
first person dual
first person exclusive
first person inclusive
first person plural
first person singular
second person singular
third person dual
third person paucal
third person plural
third person singular
final negation in a complex negative structure
possessive class 1
possessive class 2
tense, mood, aspect
Boneh, Nora & Edit Doron. 2012. Hab and gen in the expression of habituality. In Alda Mari, Claire Beyssade & Fabio Del Prete (eds.), Genericity, 176–191. Oxford: Oxford University Press.10.1093/acprof:oso/9780199691807.003.0006Search in Google Scholar
Bril, Isabelle. 2016. Tense, aspect and mood in Nêlêmwa (New Caledonia). In Zlatka Guentchéva (ed.), Aspectuality and temporality, 63–108. Amsterdam & Philadelphia: John Benjamins.10.1075/slcs.172.03briSearch in Google Scholar
Bybee, Joan L., Revere Perkins & William Pagliuca. 1994. The evolution of grammar: Tense, aspect, and modality in the languages of the world. Chicago: The University of Chicago Press.Search in Google Scholar
Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.Search in Google Scholar
Cristofaro, Sonia. 2004. Past habituals and irrealis. In Yuri A. Lander, Vladimir A. Plungian & Anna Y. Urmanchieva (eds.), Irrealis and irreality, 256–272. Moscow: Gnosis.Search in Google Scholar
Cristofaro, Sonia. 2012. Descriptive notions vs. grammatical categories: Unrealized states of affairs and ‘irrealis’. Language Sciences 34(2). 131–146. http://www.sciencedirect.com/science/article/pii/S0388000111001239. Papers selected from the ’What do languages encode when they encode reality status?’ workshop at the 41st Annual Meeting of the Societas Linguistica Europaea, Forlì, Italy, 17–20 September 2008.10.1016/j.langsci.2011.08.001Search in Google Scholar
Dahl, Östen. 1985. Tense and aspect systems. Oxford: Blackwell.Search in Google Scholar
Dahl, Östen. 1995. The marking of the episodic/generic distinction in tense-aspect systems. The generic book, 412–425.Search in Google Scholar
Druskat, Stephan. 2018. ToolboxTextModules (Version 1.1.0).Search in Google Scholar
Dunn, Michael, Stephen C. Levinson & Eva Lindström. 2008. Structural phylogeny in historical linguistics: Methodological explorations applied in Island Melanesia. Language 84(4). 710–759.10.1353/lan.0.0069Search in Google Scholar
Ferreira, Marcelo. 2016. The semantic ingredients of imperfectivity in progressives, habituals, and counterfactuals. Natural Language Semantics 24. 353–397.10.1007/s11050-016-9127-2Search in Google Scholar
Filip, Hana. 2015. Grammaticalization of generic markers. Talk given at the Workshop Habituality and Genericity in Flux, ICHL22.Search in Google Scholar
Giannakidou, Anastasia. 1995. Subjunctive, habituality and negative polarity items. In Mandy Simons & Teresa Galloway (eds.), Proceedings of SALT V, 94–111. Ithaca, NY: Cornell University.10.3765/salt.v5i0.2703Search in Google Scholar
Guérin, Valérie. 2011. A grammar of Mav̋ea. Honolulu: University of Hawai’i Press.Search in Google Scholar
Hammarström, Harald, Robert Forkel & Martin Haspelmath (eds.). 2017. Glottolog 3.0. Jena: Max Planck Institute for the Science of Human History. [Available online at http://glottolog.org, accessed on 2017-03-31]Search in Google Scholar
Hengeveld, Kees & Lachlan J. Mackenzie. 2008. Functional discourse grammar: A typologically-based theory of language structure. Oxford: Oxford University Press.10.1093/acprof:oso/9780199278107.001.0001Search in Google Scholar
Krifka, Manfred, Francis J. Pelletier, Gregory N. Carlson, Gennaro Chierchia, Godehard Link & Alice Ter Meulen. 1995. Introduction to genericity. In Gregory N. Carlson & Francis J. Pelletier (eds.), The generic book, 1–124. Chicago & London: University of Chicago Press.Search in Google Scholar
Lynch John, Malcolm Ross & Terry Crowley (eds.). 2002. The oceanic languages. Richmond, Surrey: Curzon.Search in Google Scholar
Lynch, John & Malcolm Ross. 2002a. Banoni. In John Lynch, Malcolm Ross & Terry Crowley (eds.), The oceanic languages, 440–455. Richmond, Surrey: Curzon.Search in Google Scholar
Lynch, John & Malcolm Ross. 2002b. Port Sandwich. In John Lynch, Malcolm Ross & Terry Crowley (eds.), The oceanic languages, 650–659. Richmond, Surrey: Curzon.Search in Google Scholar
Margetts, Anna. 1999. Valence and transitivity in Saliba, an Oceanic language of Papua New Guinea. Nijmegen: MPI for Psycholinguistics.Search in Google Scholar
Mønnesland, Svein. 1984. The Slavonic frequentative habitual. In Casper de Groot & Hannu Tommola (eds.), Aspect bound. A voyage into the realm of Germanic, Slavonic and Finno-Ugrian aspectology, 53–76. Dordrecht: Walter de Gruyter.Search in Google Scholar
Mosel, Ulrike. 1994. Saliba. München: Lincom.Search in Google Scholar
Palmer, Bill. 2002. Kokota. In John Lynch, Malcolm Ross & Terry Crowley (eds.), The oceanic languages, 498–524. Richmond, Surrey: Curzon.Search in Google Scholar
Prince, Kilu von. 2013. Daakaka, the language archive. Nijmegen: MPI for Psycholinguistics.Search in Google Scholar
Prince, Kilu von. 2015. A grammar of Daakaka. Berlin & Boston: De Gruyter Mouton.Search in Google Scholar
Prince, Kilu von. 2017. Paradigm-induced implicatures in TAM-expression: A view from the Daakaka distal. Proceedings of Sinn und Bedeutung 21.Search in Google Scholar
Ross, Malcolm. 2002a. Bali-Vitu. In John Lynch, Malcolm Ross & Terry Crowley (eds.), The oceanic languages, 362–386. Richmond, Surrey: Curzon.Search in Google Scholar
Ross, Malcolm. 2002b. Siar. In John Lynch, Malcolm Ross & Terry Crowley (eds.), The oceanic languages, 410–425. Richmond, Surrey: Curzon.Search in Google Scholar
Smith, Carolta, Ellavina Perkins & Theodore Fernald. 2007. Time in Navajo: Direct and indirect interpretation. International Journal of American Linguistics 73(1). 40–71.10.1086/518334Search in Google Scholar
Vander Klok, Jozina. 2013. Bill vs. the weather. Totem field storyboards.Search in Google Scholar
Zipser, Florian & Laurent Romary. 2010. A model oriented approach to the mapping of annotation formats using standards. In Proceedings of the Workshop on Language Resource and Language Technology Standards.Search in Google Scholar
Kilu von Prince is responsible for the overall structure and most of the comparative analysis as well as most of the background and conclusion sections and the section on Daakaka. Ana Krajinović contributed significantly to the theoretical background section and wrote most of the section on Nafsan. Anna Margetts provided substantial data and analysis for the section on Saliba-Logea, in addition to the corpus data cited. Nick Thieberger contributed mostly to the Nafsan section through analyses in the South Efate corpus and additional comments and suggestions. Valérie Guérin contributed mostly to the section on Mav̋ea, through the analyses of corpus data as well as additional insights and comments.
© 2019 Walter de Gruyter GmbH, Berlin/Boston