Giving voice to space: the grammar of Northern Alta spatial roots

2.1 The study of the grammar of space

One of the starting points of research on the way languages express spatial relations is related to the observation that all languages have ‘where is X’ questions. Levinson (2003: 64) notes that these questions “tend to elicit answers in which the location of a figure is specified as in some relation to a ground”, where the ‘figure’ is the entity to be located, and the ‘ground’ is the entity with respect to which it is located.^[1] Levinson and Wilkins (2006: 3) distinguish a domain in which the figure is static (‘stasis’) from a domain in which the figure is moving (‘kinesis’). The stasis domain is further subdivided depending on whether the strategy utilized to locate the figure relies on the selection of one of the facets of the ground object to specify an angle between the two elements, in which case it is considered to belong to the ‘angular domain’, also known as ‘frames of reference’ (FoR), or ‘coordinate systems’. Strategies not involving such specifications belong to the ‘non-angular’ domain, which includes the subdomain ‘topology’. These two partitions (stasis vs. kinesis, angular vs. non-angular) result in a number of specific subdomains that, taken together with the variety of linguistic strategies used to express them, represent the backbone of the ‘grammar of space’ of a given language (Figure 1).

Figure 1:

Conceptual subdivisions of the spatial domain (Levinson and Wilkins 2006: 3).

The comparative research presented by Levinson and Wilkins (2006) shows that there is considerable variation in the strategies used in spatial expression. On the basis of this variation, the authors question earlier generalizations in this field, arguing that they originate from studies of familiar languages such as English. Levinson and Wilkins observe that “the majority of languages do not code topological distinctions in adpositions, do not have satellite-framed motion systems, and do not have an intrinsic plus relative frame-of-reference inventory” (2006: 550). In the particular case of topological relations (i.e. relations expressed by the English prepositions in, at, near, between, etc.), the results of the study demonstrate that such notions are not universally encoded in specific parts of speech but are distributed throughout the clause. For example, languages with small or minimal sets of adpositions may express spatial relations with other resources within their grammar and lexicon. A well described example is the case of the Mayan language Tzeltal, in which the derivational morphology and small set of roots allow the production of nouns and verbs expressing topological relations (Brown 2006: 234).

More recent studies on the world’s languages have contributed to developing the field of the grammar of space in different directions. For example, the investigation of Mayan languages has contributed to positing new partitions of the spatial domain, as reflected in Bohnemeyer’s (2017) classification of spatial concepts. The classification builds on the stasis versus kinesis division, and distinguishes between static “inherent and essential” properties (i.e. shape, size) and static “variable” properties (2017: 327).^[2] The different types of variable static properties form three subdivisions that will be explored in Section 8: ‘location’, which specifies the location of a figure in relation to a ground; ‘disposition’, which is concerned with features of the figure’s position, configuration, or shape;^[3] and ‘orientation’, which refers to the alignment of one of the figure’s facets with a cue.^[4] Bohnemeyer’s classification also presents subtypes of kinesis, which include one of the subdomains explored here: spatial descriptions of ‘routes’.

2.2 The morphology of Philippine-type languages

Philippine languages are known to exhibit elaborate systems of verbal morphology. For example, Ilocano is a Northern Philippine language for which Rubino (2000: xviii–xxi) lists more than 450 different verbal affixes. Rubino (2005: 329) notes that all roots derive verbal and nominal forms, and that most verbs can be decomposed into a root and an affix. As a result, roots from unexpected domains, such as onomatopoeic roots, can be “thoroughly integrated into the highly productive derivational and inflectional morphology of the language” (Rubino 2001: 303). Another set of roots for which these verbal derivations are possible is the set of roots with spatial meanings, as evidenced by dictionaries of Ilocano (Rubino 2000) and Tagalog (English 1986). This derivational pattern is observed for Tagalog in Fortis (2003), which describes it as a mechanism that allows the “incorporation of spatial relations in verbal semantics”, but it has not been explored in subsequent studies on spatial expression in Philippine languages.^[5] The aim of this paper is to contribute to previous spatial research by investigating this pattern of expression in the Philippine language Northern Alta.

The productivity of such morphological systems appears to be a feature, not only of Philippine languages, but also of Philippine-type languages,^[6] for which Blust (2013: 453) notes that it is not unusual to have 200–300 distinct affixes or affix combinations. One of the most relevant features of the verbal morphology of these languages is their system of verbal alternations, also known as the Philippine-type voice system (Himmelmann 2005: 113). Such systems exhibit a subset of affixes that appear on verbs and determine whether the subject is an actor or a patient. One of the characteristics of such systems is that they distinguish at least two and sometimes three different non-actor voices, the so-called ‘undergoer voices’. Note that not all voices are available to any particular root (Reid 1992: 67–68), and that some derivations produce idiosyncratic meanings (Latrouite 2011: 198). For these reasons, voice affixes are most often treated as derivational. The question arises which voice forms are available to spatial roots and how these combinations are used in the Northern Alta spatial language.

6.1 Actor voice stative mə-

Stative AV verbs are formed with the prefix mə-. Clauses headed by stative verbs in actor voice express states of affairs (e.g. bilaj ‘life’ – məbilaj ‘alive’) or properties (e.g. biʔət ‘laziness’ – məbiʔət ‘lazy’), and have an experiencer as their subject, as shown in (11) and (12) respectively.

In spatial contexts, AV stative verbs may express the location of the subject (e.g. bəŋli ‘middle’ – məbəŋli ‘to be in the middle’), its orientation (e.g. saʔlaŋ ‘front’ – məsalaŋ ‘to face’), disposition (e.g. saŋil ‘act of leaning’^[14] – məsaŋil ‘to lean’), or manner of motion, as in (13).

6.2 Actor voice 1 <um>

With roots that denote properties or states, the AV1 infix <um> produces verbs expressing the acquisition of the property/state denoted by the root (e.g. baʔik ‘smallness’ – bumaʔik ‘to become small’, ʔiŋəl ‘anger’ – ʔumiŋəl ‘to get angry’), as in (14). Verbs derived with <um> are also used in meteorological expressions, such as ʔuddən ‘rain’ – ʔumuddən ‘it rains’. With roots denoting actions, AV1 verbs express the initiation of an event or action (e.g. lbut ‘act of boiling’ – lumbut ‘to boil’) or the completion of an action (e.g. datəŋ ‘arrival’ – dumatəŋ ‘to arrive’).

In their spatial uses, AV <um> verbs express that the subject assumes a location, as (15), an orientation (e.g. saʔlaŋ ‘front’ – sumaʔlaŋ ‘to face’), or a disposition (e.g. ʔitnud ‘act of sitting’ – ʔumitnud ‘to sit’). Such verbs also appear in the domain of routes, in which they express the initiation or completion of a movement.

6.3 Actor voice 2 men-

AV2 mən- verbs express an action associated with the referent of the root, e.g. laʔad ‘walk’ – mənlaʔad ‘to walk’, bilaj ‘life’ – mənbilaj ‘to live’, latih ‘rattan’ – mənlatih ‘to collect rattan’, saʔluŋ ‘hut’ – mənsaʔluŋ ‘to live in a hut’, butag ‘betelnut’ – mənbutag ‘to chew betelnut’. AV2 men- verbs contrast with AV1 <um> verbs (Section 6.2) in that the latter are more punctual, whereas the former are more durative. AV2 mən- verbs may describe two kinds of situations: middle situations, i.e. situations in which “the development of an action is confined within the agent’s personal sphere so that the action’s effect accrues on the agent itself” (Shibatani 2006: 234), as shown in (16). AV2 verbs may also describe active situations, i.e. situations in which the action extends beyond the personal sphere, in which case they may take an object marked in the oblique or locative case, as in (17).

AV2 mən- middle verbs are used in the expression of routes (ʔusul ‘act of going downstream’ – menʔusul ‘to go downstream’) whereas those describing active situations are used in the expression of orientation (e.g. saʔlaŋ ‘front’ – mensaʔlaŋ ‘to face’).

6.4 Patient voice -ən

Verbs derived with the PV suffix -ən form clauses in which the subject is affected in some way. PV verbs are the most unmarked subtype of UV verbs in that their perfective form only carries the perfective infix <in>, as shown in (19), whereas in LV and CV the perfective is formed by combining their voice affix with the perfective infix <in>, as shown in (20) and (40) respectively. PV verbs may denote a change of state (e.g. ʔətul ‘act of cutting’ – ʔətulən ‘to cut something’), communication (e.g. ʔujug ‘act of mocking’ – ʔujugən ‘to mock someone’), cognition (e.g. ʔisip ‘thought’ – ʔisipən ‘to think about something’) or perception (e.g. ʔalub ‘smell’ – ʔalubən ‘to smell something’). (18) shows a PV clause in which the subject undergoes a change of state.

PV verbs are used in the expression of orientation (saʔlaŋ ‘front’ – saʔlaŋən ‘to face something’) and in the domain of routes (laʔad ‘walk’ – laʔadən ‘to walk a certain distance/to walk to a certain place’).

6.5 Locative voice -an

Verbs formed with the LV suffix -an express an action that is oriented towards a location. The subject of LV clauses expresses a superficially affected patient, as in (9). It may also express a goal (e.g. suli ‘act of returning’ – sulian ‘to return somewhere’), a recipient (e.g. ʔatəd ‘act of giving’ – ʔatədan ‘to give someone’), a stimulus (e.g. ʔinta ‘act of seeing’- ʔitan ^[15] ‘to see something’), or an addressee (e.g. biddə ‘act of saying’ – biddan ‘to tell someone’). (20) below shows an LV verb in which the subject expresses a goal.

LV forms are used in the expression of orientation (liʔud ‘back’ – liʔudan ‘to turn your back on something’) and routes (dman ‘act of passing’ – dmanan ‘to pass somewhere’).

6.6 Conveyance voice ʔi-

The subject of a CV verb is an entity that is moved across space by an agent marked in the genitive case. Thus, these verbs can be understood as expressing caused motion. The motion may be concrete (e.g. sumə ‘act of hiding’ – ʔisumə ‘to hide something’) or abstract (e.g. puluŋ ‘word’ – ʔipuluŋ ‘to say something’). In some verbs, motion results in a change of state (e.g. gisa ‘act of stewing’ – ʔigisa ‘to stew something’^[16]). In the case of verbs expressing concrete motion, the movement can be translational, i.e. shifting from one point in space to another, as in (21), or non-translational, as in (22), in which the verb describes the action of rotating the open part of a hut towards the shade, so that the thatched roof can protect it from the sun.^[17]

As shown in the preceding examples, the result of the caused motion can be a new location, as in (21), or a new orientation, as in (22). The result of caused motion may also be a new disposition, as will be shown in Section 8.2.

6.7 Directional pa-

The directional affix pa- is not part of the voice system, but it is a derivational affix that also appears in a number of spatial descriptions. It may combine with spatial roots (e.g. taʔpu ‘top’ – pataʔpu ‘towards the top’), place names (e.g. ditiʔi ‘Diteki village’ – paditiʔi ‘towards Diteki’), or roots expressing motion (e.g. suli ‘act of returning’ – pasuli ‘returning, on the way back’). Directionals formed with pa- can function as adverbial modifiers or heads of predicates, as in the first and second parts of (23) respectively, and also as nominals, as in (24).

Directional pa- forms are used in orientation descriptions, as in (48), and in routes, as in (49).

7.1 Location questions

In Northern Alta, ‘where is X’ questions can be verbless, simply formed with the interrogative pronoun ʔadinu ‘where’ and the subject, as in (25). The interrogative pronoun ʔadinu may also be followed by the locative copula ʔisaj ‘to be at’, as in (26).

7.2 Basic locative construction (BLC)

Northern Alta uses the construction headed by the locative copula ʔisaj ‘to be at’ as its BLC.^[19] In these constructions, the subject represents the figure and the constituent marked in oblique or locative case, as in (27) and (28) respectively, expresses the ground.^[20] Oblique ground phrases typically have an indefinite reading, whereas locative phrases are definite and can be headed by place names.

Although oblique and locative phrases may express locations, none of them specify the nature of the spatial relationship between the figure and the ground object. Speakers instead resort to the set of spatial roots presented in the following section.

7.3 Set of spatial roots

Spatial roots are a class of roots that express spatial relations such as coincidence, propinquity, contact or containment between figure and ground, and angular relations, i.e. relations involving the use of a coordinate system. Spatial roots may occur in their bare form as nouns and may also take many of the affixes presented in Section 6 (see Appendix). As shown in Table 1, most spatial roots form binary oppositions, i.e. pairs of terms that are opposite in meaning.

‘Front-Back-Side’ (FBS) roots express location when occurring as nouns, but differ from the other subcategories in that their verbal derivations have an orientational reading. Such FBS-derived verbs are crucial in the expression of orientation (see Section 8.3).

Left-right terms most often refer to inherent facets of figure objects, as in the expression ‘to the man’s left’, but they are sometimes used from the perspective of the viewer, as in ‘to the man’s left (from our perspective)’. Contiguity is the label for a subcategory of spatial terms expressing relations of contact and coincidence. It includes the pairs ‘inside-outside’ and ‘middle-extremity’. Note that the term disalad evokes both notions of ‘inside’ and ‘under’, and thus it would also match the vertical category. The term digdig refers to landscape features such as a seashore or riverbank but also to regions projected from the side of an object (i.e. digdig ni bəli ‘beside a house’). The term digdig differs from the FBS term tagidlid ‘side’ in that its stative derivation has a locational (non-orientational) reading ‘to be beside’, as in (32).

Vertical terms come in two subsets with different uses. The term taʔpu ‘top’ is most often used in its nominal form and usually refers to mountain peaks. It is sometimes used to refer to the highest point in the sky, and to the upper part of the man’s body, in the Man and Tree task. As to the term taŋʔal, in its nominal form, it refers to the property of height and often derives stative verbs that can be used to describe both concrete things (e.g. ‘tall trees’) and more abstract concepts (e.g. ‘high prices’ or ‘high blood pressure’). As for təpdəʔ and siduŋ, the former expresses the property of lowness and can be used to refer to lower regions of objects (i.e. the lower part of a door), whereas siduŋ can only refer to areas underneath objects but not to low regions.

The category ‘stream terms’ includes the pair diŋatu ‘upstream’ – dilud ‘downstream’, which play an important role in several domains of Northern Alta spatial expression. The term dipaniŋ ‘across’ is mainly used to refer to the area across a river, but can also refer to a river located across a mountain. It can also be used in smaller scale settings and refer to ‘the other hand of a person’ or ‘the other side of a person’. As to dəbbəliw ‘opposite side’, it is only used in large-scale settings and may refer to a place on the other side of a sea, a river, or a village. Finally, sunset-sunrise terms are used in spatial descriptions but are not included in Table 1 because they are not roots but rather derived terms: ʔalaŋan ni sinag ‘sunrise place, the coming place of the sun’ and əlbugan ‘sunset place, sinking place’.

7.4 Spatial roots in ground phrases

The lexical head of the BLC ground phrase can be a noun or a place name, as in (27) and (28), respectively. Ground phrases may thus be headed by a spatial root occurring as a noun, as with gəsaʔ ‘outside’ in (29). The spatial noun may also be possessed by a genitive phrase indicating the ground. In such cases, the spatial noun refers to a part of the ground or to a region projected from it. (30) shows a ground phrase with the spatial noun disalad ‘inside’ possessed by the noun bəli ‘house’.^[21]

8.1 Location

Location can be of two types, depending on whether the strategy employed relies on the specification of an angle or not. The locative construction, the Northern Alta BLC (see Section 7.2), is the preferred construction in both angular and non-angular descriptions.

8.1.2 Frames of reference

The data collected for the domain of frames of reference (in which descriptions specify an angle) comes from the M&T task (see Section 4). The dataset shows that, of the three different frames of reference (FoR) distinguished in Levinson (2003), (i.e. intrinsic, relative, and absolute), the intrinsic frame is used in 63.3% of the descriptions, the absolute frame in 12.6% of the cases, and the relative frame in 8.5%.^[23] Formally, locative expressions using any kind of FoR are expressed with the BLC.

Levinson and Wilkins (2006) describe the intrinsic FoR as a strategy that involves the partition of a ground object or landmark into facets, on which search domains can be projected (e.g. the expression ‘the man is in front of the church’ relies on the front facet of the church to project a search domain). In the case of the data collected with M&T, the most common strategy treats the man as the ground for which it distinguishes a front facet, a back facet, and two sides. Such a strategy locates the tree in a region determined by one of the man’s facets. Thus, the resulting intrinsic expressions have ground phrases in which the search domain is defined by FBS or left-right terms, as in (34)–(35) respectively.

(34)

ʔisaj=sila

ti

saʔlaŋ =na=i

ʔin

pun

nən

ʔajuh

be.at=also

obl

front=3s.gen=sp

abs

trunk

gen

tree

‘the tree is in front of him’ (142.137, M&T R33)

(35)

ʔisaj	ti	diwanan =na	ʔin	ʔajuh
be.at	obl	right=3s.gen	abs	tree
‘the tree is on his right’ (142.185, M&T R12)

The absolute frame of reference uses arbitrary fixed bearings as the anchor of the coordinate system (Levinson 2003: 48). Such bearings may be defined by elements such as the sun’s path, wind directions, mountain slopes, river drainages, etc. Speakers of Northern Alta sometimes use the local river as the anchor of the coordinate system to disambiguate M&T cards, and thus rely on the set of stream terms to define a search domain. (36) shows two such absolute descriptions, with the man and the tree respectively treated as figures, and the tree and the man as respective implicit grounds.

(36)

ʔisaj

ti

diŋatu

ʔin

taʔu

ʔisaj

ti

dilud

be.at

obl

upstream

abs

man

be.at

obl

downstream

ʔin

ʔajuh

abs

tree’

‘the man is upstream (of the tree), the tree is downstream (of the man)’

(147.281, M&T R32)

Speakers of Northern Alta may also rely on the bearings defined by the sun’s path, even though such a system is not found in M&T descriptions of location. A number of orientational descriptions use the sunset-sunrise locations as fixed bearings; these could arguably be described as using an absolute FoR, as in (48) in Section 8.3.2 below.

As to the relative frame of reference, it involves a mapping of the axes of the observer’s body onto the ground, as in the expression ‘the tree is to the right of the church (from my perspective)’. Unlike the intrinsic description presented in (35), such a relative expression would not remain true if the observer walked around the scene (Levinson and Wilkins 2006). Most descriptions of M&T locations rely on the observer’s lateral axis, thus their ground phrases have regions defined by left-right terms. As shown in (37), relative expressions using left-right terms are formally identical to their intrinsic counterparts presented in (35): in both cases, the ground can be expressed as a possessor. Both (35) and (37) describe the scene presented in Figure 2.

Figure 2:

M&T card R12.

(37)

ʔisaj

ti

ʔawili =na=i

ʔin

pun

ʔa

ʔajuh

be.at

obl

left=3s.gen=sp

abs

trunk

lk

tree

‘the tree is to his left (of the man’) (137.30, M&T R12)

8.2 Disposition

Brown (2006: 231) defines dispositional terms as terms that encode features of space, shape, configuration, or position.^[24] Northern Alta exhibits a set of least 30 roots to express such dispositional features. The data collected with the PSPV task (see Section 4) clearly shows that most scenes are described with AV stative verbs formed with such dispositional roots. (38) and (39) show stative derivations of the roots ʔedsaŋ ‘lie’, pajuŋ ‘stand’ and saŋil ‘lean’.

CV derivations of dispositional roots are also used in certain scenes of the PSPV task. Clauses headed by CV dispositional verbs involve an actor that caused the object to be in a certain disposition. Such constructions appear in pictures in which the figure is perceived as being in an atypical position, as with the pot in Figure 3, which is described in (40).

Figure 3:

PSPV picture 12.

Table 2 provides a sample of dispositional roots, together with their stative AV and dynamic CV derivations, all of which were collected with the PSPV task.

8.3 Orientation

The data for the domain of orientation was collected using the M&T task and involves two different strategies: one that describes the man in relation to the tree and the other in relation to a cue located outside the picture (picture-internal vs. picture-external cue).^[25] The two strategies differ in the grammatical treatment assigned to two types of cues.

8.4 Routes

Routes contrast with location, disposition, and orientation in that the latter three are static. The data for this domain was collected using a route description task consisting of a series of 13 questions of the type ‘how to go from point A to point B’, with the constraint that the path description must avoid local roads. The purpose of this constraint was to obtain verbs describing actions like ‘going up a hill’ or ‘crossing a river’ and to verify whether verbs derived from the set of spatial roots are used in this context. The dataset reveals that Northern Alta consistently favors the use of an additional set of roots, complementary^[26] to the set of spatial roots presented in Table 1: the set of path roots. Table 3 presents the set in relation to their spatial counterparts.

Path roots involve a figure moving in relation to a ground, e.g. ʔuddiʔ ‘go upstream’ presupposes a figure that moves along a particular path determined by the river, and dudul ‘to go beside’ presupposes a figure that reaches a particular goal, i.e. the side of a ground object. Verbs derived from these roots can be referred to as path verbs in the sense of Rau et al. (2012: 5), i.e. “verbs that express a trajectory of the figure with respect to the ground”.

In addition to the predicate, path and goal can also be expressed elsewhere in the clause. In AV clauses, path and goal can be expressed in locative or oblique DPs, whereas in PV or LV both path and goal can be treated as the subject and thus expressed in absolutive DPs. Manner of motion is expressed by stative verbs that function as either modifiers of the predicate, as in (13) above, or as heads of oblique DPs. As for the figure, it is expressed by absolutive DPs in clauses with AV path verbs or by genitive DPs in UV clauses, as in (50) and (51) respectively (both examples are further discussed below). Due to space limitations, it is unfortunately not possible to provide a more complete depiction of the available strategies for expressing the components of motion.

The data collected with the route description task shows that verbs derived from path roots appear in three forms: AV <um> (42%), directional pa- (30%), and, to a lesser extent, AV2 men- verbs (13%), with only two instances of PV or LV verbs. In many of the route descriptions, we find a directional pa- verb (see Section 6.7) at the beginning or at the end of the description. In most descriptions, the directional form is followed or preceded by sequences of clauses headed by AV2 and AV1 verbs, with AV2 expressing durative movement and AV1 expressing the completion of a movement. The alternation of directionals, of AV2 and AV1 verbs forms the structure of most route descriptions. (49) is the last part of a route description from a village called Kalapnit to a coastal village named Dibut, which is only accessible by boat or mountain hike. In order to reach Dibut, the traveller has to follow a stream towards the source, climb a hill, and finally descend towards Dibut village:

As to the limited use of UV verbs, (50) and (51) show an AV-PV alternation of the verb ʔəlwas ‘cross a river’. (50) has a predicate headed by the AV1 derivation, the figure is the subject, and the ground entity defining the path is marked in the locative case (the ground phrase is delimited by square brackets). In (51), we find a predicate with a PV form of the same root. The subject of the clause is the ground entity, and the figure is marked in the genitive case.

9.1 The role of spatial roots in spatial expression

In relation to research question 4, which is concerned with spatial roots and their relation to other lexical resources, the preceding subsections show that the spatial domains explored rely on three different sets of roots: spatial roots (Table 1), dispositional roots (Table 2), and path roots (Table 3). The data also shows that each set of roots is predominantly used in one particular subdomain. Spatial roots differ from the two other sets in that they may occur in both nominal and verbal forms. In their nominal form, spatial roots play a central role in the expression of location, appearing as heads of the BLC’s ground phrases in both topological and coordinate systems. As to their occurrence as verbs in the expression of location, it is attested but marginal. On the other hand, verbs derived from FBS roots (a subtype of spatial roots) form predicates that are fundamental in the expression of orientation.

Disposition is expressed with verbs derived from dispositional roots, whereas routes are expressed with verbs derived from path roots, a set of roots that is complementary to spatial roots. Verbs derived from the three sets of roots are represented in the central part of Figure 4, and they are linked with an arrow to the spatial domain in which they operate. On the other hand, the nominal occurrence of spatial roots in the BLC is represented by a circle located on the right side of Figure 4. The circle is linked to the subdomain of location, in which the BLC and spatial roots (in ground phrases) play a central role.

9.2 Figure and ground information

In relation to the issue of the distribution of spatial information in the clause (research question 2), we observe that Northern Alta is a language that expresses ground information in spatial nominals, in path-verb predicates, and less importantly, in stative locational predicates. The language may also express figure information in predicates headed by FBS and dispositional verbs.

Starting with ground information, it is clear that the locative verb ʔisaj, which is the head of the BLC predicate, can be used with any kind of figure and thus does not specify any kind of figure-related information. The BLC ground phrase, on the other hand, is headed by spatial nominals that convey information on the ground. For example, the term saʔlaŋ ‘front’ presupposes a faceted ground, whereas the term siduŋ ‘underneath’ entails an elevated volume that generates a sheltered space underneath (i.e. the crown of a tree or the seat of a chair). In addition, due to the derivational pattern discussed in Section 2 and further explored in the Appendix, ground information, such as volumetric properties, axial properties or topographic features, may also be expressed in the predicate when it is headed by one of the path verbs presented in Section 8.4 or by a locational stative verb. For example, the path verb mən-ʔuddiʔ ‘go upstream’ implies a body of water such as a river or a stream which flows, and the stative verb mə-disalad ‘to be inside’ involves a ground with a hollow space.

As for information on the figure’s shape or geometry, it is encoded in both stative dispositional and orientational predicates (Sections 8.2 and 8.3 respectively). For example, the stative orientational verb mə-saʔlaŋ ‘to be facing’ presupposes a figure (not a ground) with a front facet, whereas the dispositional verb mə-ʔədsaŋ ‘to be lying’ entails a figure with an axis longer than the other one, with the long axis displayed horizontally.

Thus, Northern Alta is a language in which both figure and ground information can be expressed in the predicate. This particular distribution of spatial information is illustrated in Figure 4, in which the figure and ground information are represented with horizontal lines placed under the different types of spatial predicates.^[27]

9.3 Voice alternations in spatial expression

Concerning the relationship between voice (voice-marking form and voice alternations) and the different spatial subdomains (research question 3), the data suggests that each subdomain favors a particular subset of voice affixes and that voice-marked constructions play two different roles in subdomains of spatial expression: they express pragmatic markedness of locations or dispositions, and the centrality of cues/grounds in orientation and route descriptions.

In the domain of location, there are two competing constructions. The locative clause (Section 7.2) is the most common device in the expression of location and, like the BLC, it is the most unmarked construction.^[28] Location can also be expressed by voice-marked clauses in stative AV, as shown in (32) and (33). Our elicitation sessions confirm the locational semantics of such forms and their morphological productivity, which can be observed in the second column of the table in the Appendix. As a non-default type of locational clause, such voice-marked locative clauses are pragmatically more marked than the BLC. A possible reading of such statives is that the location of the figure is highlighted because it is perceived as nontypical, similarly to what happens with disposition (see Section 8.2), but for the domain of location, such a possibility still needs to be investigated in future work. As to the domain of disposition, the data shows that most scenes are expressed with voice-marked constructions. The preferred construction in this domain is a clause with the predicate headed by an AV stative derivation of a dispositional root. As shown in Section 8.2, when a particular setting is perceived as atypical (see Figure 3), speakers use a construction headed by a CV dynamic derivation of a dispositional root. Both dispositional constructions treat the figure as the subject and mark the ground in oblique or locative case, but differ in the fact that the CV dynamic derivation introduces an additional participant: the causer.

Moving on now to orientation, it is the subdomain expressed with the widest variety of constructions, several of which are voice-marked (Section 8.3). Orientational voice-marked expressions include clauses with predicates headed by AV verbs, either stative or dynamic, and dynamic verbs in undergoer voices, including PV and LV. The alternation between stative and dynamic verbs in the M&T task may be explained in terms of whether the spatial array is construed as a state of affairs or as a situation involving an animate agent that willfully orients his body towards a tree (a situation that is also described with the perception verbs ‘see’, ‘look’ presented in Section 8.3). Among the instances of dynamic verbs, we find two undergoer voices that sometimes alternate with actor voice: PV and LV. We have seen that such UV forms only appear in picture-internal orientation descriptions (i.e. those in which the tree is the cue) but not in picture-external descriptions. A possible explanation for this difference is that in picture-internal descriptions, the same cue appears in every card. This recurrence increases the possibility of treating the cue as a central element and selecting it as the subject of the orientational predicate. On the other hand, the cue in picture-external descriptions is variable and thus is perceived as less salient and is more likely to be marked in a case that has a non-definite reading, i.e. the oblique case in an AV clause. We have seen that the subject argument in LV constructions can take different semantic roles, such as stimulus or goal, as in (20) above, whereas in PV the subject is a more affected patient. Thus, considering the semantics of these affixes, LV constructions are more to be expected than PV in orientation descriptions, given that the cue/tree is only superficially affected, if affected at all. A difference between these two types of predicates is that PV forms only rely on the man’s front facet so that only the cards in which the man faces the tree are described in PV. A possible reading of these PV forms is that the tree is perceived as an obstacle that the man has to overcome or as something that is about to undergo a change of state (i.e. cut or hit in some way).

Finally, we have seen that in the domain of routes, most constructions are headed by verbs formed with a path root and the affixes AV1 <um>, AV2 mən-, and directional pa- (Section 8.4), and that the alternation between AV2 and AV1 verbs structures route descriptions. Examples of undergoer voice are scarce in this domain, but (50) shows a ground (the river) that is expressed in a locative phrase at the first mention and becomes more central at the second mention, as in (51), in which it is treated as the subject of the PV clause.

Figure 4 lists the types of affixes used in each spatial subdomain and uses stars to represent voice alternations in the domains in which they are most relevant: disposition and orientation.