Linguistics in Amsterdam 5-1 (January 2012)Vadim Kimmelman: Word order in Russian Sign Language

To refer to this article use this url: http://www.linguisticsinamsterdam.nl/05/nr01/a01

4 Word order in RSL

In this section, the results of my analysis of the corpus and experimental data will be presented. The position of the subject will be discussed first, followed by a discussion of the position of the object; thirdly I address the order in locative clauses, and finally doubling of predicates. Since doubling is presented in a separate section, I will not address this issue in the first three sections. I excluded from analysis all clauses consisting of a verb only. Thus, when I state, for example, that the SV order appears in 95% of the clauses, this means that it appears in 95% of the clauses in which the subject is expressed (and expressed only once).

4.1 Subject position

next section

In the corpus data, the subject preceded the predicate in the absolute majority of the cases (95%, 170 out of 179 clauses). In the experimental data, the subject always preceded the predicate. In the nominal and adjectival clauses, the only argument always appears before the predicate. Therefore, it is possible to immediately conclude that the position of the subject is pre-verbal.

In some languages, the position of the subject depends on the transitivity of the verb (Dryer 2007). For example, in Spanish subjects of intransitive verbs can appear in the postverbal position, but subjects of transitive verbs cannot. Therefore, it might be worth investigating whether in RSL subjects of transitive and intransitive verbs also behave differently. However, there is no research on transitivity in RSL, and there is no reliable methodology to decide which of the verbs in the corpus are transitive and which are intransitive. Nevertheless, it is still possible to assess the hypothesis that transitivity influences the position of the subject. There are some clauses in the corpus in which the object and the subject are overt, so the verbs are obviously transitive. There are also many clauses in which only the subject is overt; in these cases, the object is either covert (in the case of transitive verbs) or there is no object in the argument structure (in the case of intransitive verbs). Thus the latter group of clauses should contain clauses with both transitive and intransitive verbs, while the former group contains only transitive verbs. If transitivity influenced the position of the subject, we would expect these two groups to show different distribution of the subject position, because clauses in the two groups differ with respect to transitivity. In reality, however, in both groups the percentage of the VS order is only 5%. We can therefore conclude that the position of the subject in both transitive and intransitive clauses is pre-verbal.

Prosodic properties of subjects support this analysis. In most cases of the SV order, the subject does not constitute a separate prosodic unit (EDU). As shown in Table 3, in 65% of the cases the subject and the verb are within one prosodic unit.

Table 3: Prosody with SV order

SV

1 EDU

2 EDU

total

116

63

179

%

65

35

100

There are several types of situations in which the subject constitutes a separate EDU. First, at the beginning of a narrative, subjects are quite often (10 cases) introduced in a separate EDU, in fact, following the subject there is often a boundary of a super-discoursive unit (20)[11].

Situations in which the subject constitutes a separate super-discoursive unit occur only at the beginning of narratives (with one exception, [X3-41], which is, however, at the beginning of a new episode in the narrative). Therefore, it is reasonable to assume that this is a discourse-related strategy, specifically used to introduce the main participants of the narrative.

Second, in some cases non-manual marking accompanies a subject which appears in its own EDU, which may be indicative of topicalization (but, of course, I have no proof yet that this is in fact topicalization). The non-manual marking consists of lowered eye brows and a head nod or only raised eye brows, and is used to introduce information known to the addressee (21). For instance, in 21 the subject lady fat is marked with this non-manual marker, which might be a sign of a movement.

If in the examples of this sort subjects are indeed topicalized, then the prosodic boundary between the subject and the rest of the clause is expected, because topics are often intonationally separated in SLs (Aarons 1994).

Thirdly, in some cases the signer hesitated between the subject and the verb, thus creating a prosodic boundary between them (22).

If we discard these three types of situation, then we are left with only 15% of clauses in which the subject is separated from the verb by a prosodic boundary. Therefore, I conclude that in the default case, the subject and the verb constitute one prosodic unit.

In the nine clauses with VS order, the prosodic facts are different. In four cases out of nine, there is a prosodic boundary between the verb and the subject. Therefore, the more marked word order (VS) is also more marked prosodically, which is in line with the hypothesis formulated in Section 3.

Considering my position that prosodic boundaries must not be equated with clause boundaries, the sequence of clauses in (23) is of interest.

These two clauses are almost identical, but in the second one the subject son is separated into its own EDU. In other words: the same syntactic structure is mapped onto two different prosodic structures.

In the following section, the position of the object is discussed, including the factors that can influence it. I have checked whether these factors also influence the position of the subject, but none of them appeared to do so. I will therefore not discuss these factors for the subjects. The subject in RSL is clearly pre-verbal, and most likely the 5% cases with VS order can be attributed to afterthoughts or the like.

4.2 Object position

Determining the position of the object(s) in RSL is more intricate than specifying the position of the subject. Therefore, this issue will be apprached in several steps. First, the quantitative data concerning the position of the object in the corpus is presented, and then the factors that influence the position of the object are discussed.

Before turning to the discussion of the object position, I want to mention that I also observed objects in sentences with three-place (ditransitive) predicates (like give, send etc.). However, such clauses were too infrequent to allow for a systematic analysis. In addition, I faced the problem of not having sufficient data to determine which object is the direct one and which is the indirect one in a three-place predicate. Therefore, in the discussion below, I refer mainly to clauses where one object was expressed and discuss the position of this object.

4.2.1 General picture

In the corpus data, objects are expressed in 105 clauses (the experimental data are discussed below in sections 4.2.2 and 4.2.3). In 74% of the cases, the order is OV, and in 26% of the cases, it is VO[12]. From this one can conclude that in the default situation, the object precedes the verb. However, later in Section 4.2.2, I will argue that it is not the case.

Pre-verbal objects are usually not divided from the verb by a prosodic boundary; see Table 4:

Table 4: Prosody with OV order

OV

1 EDU

2 EDU

total

56

22

78

%

72

28

100

When the word order is OV, the object is separated from the verb by a prosodic boundary in 28% of the cases. Still, there are also several types of situations when there is a boundary between the object and the verb.

Firstly, sometimes the signer hesitates, which naturally results in a prosodic boundary. In example (24), the direct object umbrella is divided from the verb because of the hesitation[13].

Secondly, sometimes the signer clarifies or specifies the meaning of the object by means of some additional description, which can also lead to a prosodic boundary. In (25), for instance, the signer used the second sign umbrella2 to clarify the meaning of the first sign umbrella1, and this resulted in a prosodic boundary between the two nouns.

Thirdly, when there is a role-shift between the object and the verb (Engberg-Pedersen 1993), the object forms its own prosodic unit (26).

In example (26) the signer takes the role of the brother calling the sister after signing the object sister, resulting in a prosodic boundary. Within the framework of Prozorova (2009), this result is straightforward because role shift usually requires a body shift, which for Prozorova defines a prosodic boundary.

Fourthly, there was one instance of a clause with OSV order (27), which suggests that the object, which ias also marked by raised eye brows, is topicalized. The possible topicalization may explain the prosodic boundary between the object and the rest of the clause.

Fifthly, locative clauses behave prosodically different from non-locative clauses. In these clauses, objects are more often prosodically separated from the verbs. Locative clauses will be discussed in detail in Section 4.3.

In sentences with post-verbal objects, it is more common for the object to be contained in its own EDU. Table 5 shows that 50% of the clauses with the VO order (i.e. 15 out of 30) consist of two EDUs..

Table 5: Prosody with VO order

VO

1 EDU

2 EDU

total

15

15

30

%

50

50

100

If the hypothesis formulated in the methodology section is right, then this means that VO order is less basic, and thus also more prosodically marked. Below, however, I will show that this conclusion is not correct.

Turning again to my claim that prosodic boundaries must not be equated with clause boundaries, the sequence of clauses in (17) – repeated here as (28) for convenience – is of interest:

In (28a) the object forms its own EDU, while in (28b), which is manually identical to the (28a), the object and the verb are not prosodically separated.

4.2.2. Verb class and object position

For many SLs, the verb class has been shown to have an influence on the word order (Kegl 2004a, 2004b; Rathmann 2001; Vermeerbergen et al. 2007; de Quadros 1999; Milković et al. 2006; Hendriks 2008). Agreeing verbs may behave differently from plain verbs, and classifier constructions can also display a different syntax.

In order to test whether the same holds in RSL, I had to determine verb classes first. This process is not trivial, and the existence of clear-cut verb classes (such as plain and agreeing verbs) has even been questioned (Schembri & Cormier 2009). Concerning verb classes, the following decisions were taken:

In the corpus data, agreeing verbs were only used in three clauses with overt arguments. However, plain verbs and classifier constructions were used sufficiently often to allow for a comparison. In 78% of the clauses with classifier constructions (63 out of 81), the object was pre-verbal (OV), while in 22% it followed the verb (VO). This distribution is very similar to the distribution in the corpus in general. The picture was different, however, for sentences with plain verbs. Only in 4 out of 11 cases the word order was OV, while in the other 7 cases it was VO. However, given the small number of clauses with plain verbs and an overt object, it would be premature to determine whether the verb class influences word order.

In the experimental data, too, most of the verbs were classifier constructions. The distribution of word order within this group of clauses is approximately the same as in the corpus data: 81% (36 cases) display OV order and 19% (8 cases) have VO order[14]. With plain verbs (be.afraid, sell, build, behave), the (S)VO order appeared eight times in the experimental data, while (S)OV and OSV were observed one time each. With agreeing verbs, the (S)VO order was used eight times while SOV was attested only once. Therefore, it is clear that the experimental data confirm the results obtained from the corpus data: plain and agreeing verbs in RSL are used predominantly with the SVO order, while sentences with classifier constructions show a clear preference for the SOV order.

This is further confirmed when we focus on the pool of SVO examples with plain and agreeing verbs. In the next section, I will show that there are additional semantic and syntactic factors that can favour the SVO order, namely reversibility of the situation and animacy or heaviness of the object. However, among the clauses with SVO order, 11 contain a simple inanimate object, in two cases the object is heavy, in four cases the object is animate, and three out of these four cases are reversible (all others are non-reversible). For example, in example (29) with the agreeing verb look, the object is not heavy, it is inanimate, and the situation is not reversible.

Therefore, most instances of SVO order with plain and agreeing verbs cannot be attributed to other factors; thus this order is really determined by the verb class.

4.2.3 Other factors

Several other factors turned out to have a potential influence on the position of the object in RSL. Before discussing these factors, however, I want to mention two factors that could possibly be thought to be of influence, but that do not appear to be in these data.

First, working with the corpus I got an impression that the use of fingerspelling could influence the position of the object. However, when I considered the fingerspelled objects in the corpus (e.g. g-r-u-š-i in (20)), the predominant object position was still pre-verbal.

Secondly, I also supposed that the use of pronouns vs. full noun phrases could influence word order, as has been described for various signed and spoken languages of the world, including ASL (Wilbur 2002) and HZJ (Milković et al. 2006). However, in my data, the use of pronominal objects did not result in a word order different from that observed with full noun phrases[15]. Maybe a larger corpus would reveal that these factors also play a role, but based on the data analyzed here, no such influence can be assumed.

Still, there is evidence for the influence of several other factors on word order in RSL. In the data I analyzed, all verbs inflected for aspect (habitual or progressive, in other words, aspect types that are phonologically realized by reduplication) appeared in the clause-final position. However, given the small number of examples, this issue requires further research. Interestingly, verbs marked for aspect have been reported to appear clause-finally in some other SLs, too, namely in ASL (Chen Pichler 2001) and LSB (de Quadros 1999).

As in other SLs – for instance, VGT (Vermeerbergen et al. 2007), LIS (Volterra et al. 1984), HZJ (Milković et al. 2006), and LSB (de Quadros 1999) – reversibility influences word order in RSL. In reversible situations the (S)VO order is preferred.

In the corpus data, reversible situations are much rarer than non-reversible ones. In order for a situation to be reversible, the arguments usually should be either both animate, or both inanimate. The latter type of situation did not appear in my data at all, so in all reversible clauses discussed here both arguments are animate.

Table 6 shows that in the corpus data reversible situations displayed the (S)OV order 8 times and the (S)VO order 7 times, while in non-reversible clauses the SOV order was more dominant (74%).

Table 6: Influence of reversibility in the corpus data

Number

%

Reversible

(S)OV

8

53

(S)VO

7

47

total

15

100

Non-reversible

(S)OV

70

74

(S)VO

24

26

total

94

100

The distribution for the experimental data (Table 7) shows a similar picture: approximately half of the reversible clauses contained the SVO order, while in the non-reversible clauses the SOV order was predominant.

Table 7: Influence of reversibility in the experimental data

Number

Reversible

(S)OV

5

(S)VO

5

OSV [16]

1

total

11

Non-reversible

(S)OV

21

(S)VO

8

OSV

1

total

30

Animacy of the object can influence its position in RSL, just as in some other SLs (e.g. LSA (Massone & Curiel 2004), HZJ (Milković et al. 2006), NGT (Coerts 1994), and LIS (Volterra et al. 1984)). Obviously, this factor is related to the previous one, as in my data only animate objects participated in reversible situations. When discussing this factor, we therefore have to keep in mind that it cannot be decided whether one of these two factors is not a direct result of the other.

In the corpus data, animate objects appeared in the post-verbal position more often than inanimate objects. As can be seen in Table 8, the VO order was used in almost half of the clauses with an animate object, while inanimate objects were predominantly (74%) pre-verbal.

Table 8: Influence of animacy in the corpus data

Number

%

Animate

OV

8

57

VO

6

43

total

14

100

Inanimate

OV

70

74

VO

24

26

total

94

100

Again, the experimental data yielded similar results. As shown in Table 9, animate objects are as frequently pre-verbal as post-verbal, while inanimate objects are mostly pre-verbal.

Table 9: Influence of animacy in the experimental data

Number

%

Animate

OV

5

33

VO

5

33

VOV [17]

5

33

total

15

100

Inanimate

OV

33

75

VO

8

18

VOV

3

7

total

44

100

The last factor that I want to discuss is the heaviness of the object. “Heavy” objects, that is, object NPs that contain dependent material and which are therefore phonologically more heavy, tend to appear in clause-final position in, for example, English, as is illustrated by the sentence pair in (30). This phenomenon is referred to as “Heavy NP Shift” (Larson 1988):

I considered all object NPs containing more than one sign heavy. For example, an object can be modified by an adjective, it can be repeated, or it can be first signed and then fingerspelled. I reasoned that even one additional sign might make an object heavy, because in the RSL discourse I analyzed most of the noun phrases contained only a single sign. Also, the duration of a sign in a sign language is generally longer than the duration of a word in a spoken language (Bellugi & Fischer 1972), which means that it is “easier” to make a sign language NP phonologically heavy.

The corpus data did not provide evidence for the claim that the heaviness of an object influenced its position. Rather, as shown in Table 10, the positioning of heavy and non-heavy objects was strikingly similar: both appeared predominantly in the pre-verbal position.

Table 10: Influence of heaviness on the position of the object in the corpus data

Number

%

Heavy

OV

66

74

VO

23

26

total

89

100

Non-heavy

OV

12

75

VO

4

25

total

16

100

However, the experimental data yield a different picture. As is evident from Table 11, heavy objects appear more often in post-verbal than in pre-verbal position, while with non-heavy objects the OV order is predominant.

Table 11: Influence of heaviness on the position of the object in the experimental data

Number

%

Heavy

OV

1

10

VO

6

60

VOV

3

30

total

10

100

Non-heavy

OV

37

74

VO

8

16

VOV

5

5

total

50

100

At this point, I can only speculate about why the experimental data but not the corpus data showed an influence of heaviness on the object position. One possibility is that my definition of heaviness was too weak: maybe objects in the experimental data were in fact heavier than objects in the corpus data, and some of the objects in the corpus which were considered heavy should not have been analyzed as such. Probably a larger data pool is necessary to test this factor.

To sum up, the data suggest that the use of fingerspelling and pronouns does to influence the position of objects in RSL. Aspectual marking on the verb and the heaviness of objects are likely to have an impact on the position of objects, but more data is necessary to verify this claim. Finally, the reversibility of the situation and animacy of the object do influence the object’s position. One should keep in mind, however, that – as was shown in section 4.2.2 – the factor that determines the basic position of the object in the first place is the verb class.

4.3 Locative clauses

As mentioned in section 2, there is good reason to analyze locative clauses in SLs separately since they have been shown to behave similarly across different SLs, and even descriptions rendered in pantomime by non-signers show a similar pattern (Laudanna & Volterra 1991).

The corpus contains 70 locative clauses with one or more arguments expressed. Usually, in these constructions, the argument labeled S is the Figure which is located or moved relative to the Ground labeled O. Sometimes, in case of object manipulation, there are three arguments: the Agent (S) who performs the manipulation, the first object (the Figure) which is being manipulated, and the second object (the Ground) in relation to which the Figure is manipulated. Example (31) is of the latter type, but in this example, the Agent is not expressed, while the Figure is round.object (‘tray’) and the Ground is chair.

Surprisingly, however, the word order in locative clauses in the corpus turned out not to be different from the word order in other clauses (see Table 12).

Table 12: Word order in locative clauses in the corpus

Number

%

SV

46

98

VS

1

2

total

47

100

OV

24

83

VO

5

17

total

29

100

The subject preceded the verb in all clauses but one, and the object was also mostly pre-verbal. This distribution, however, is not very informative, because the difference between locative and non-locative clauses should appear when both the subject and the object are expressed, as this is when the OSV order should surface (that is, the Ground-Figure order found in previous studies).

In the corpus, there were 8 locative clauses containing both the subject and the object. However, only one of these clauses showed the expected OSV order (32); moreover, the object in this clause is non-manually marked (by raised eye-brows) which can be a sign of topicalization.

The other word orders were SOV (5 cases) and SVO (2 cases).

Therefore, on the basis of the corpus data, I was not able to confirm my hypothesis that locative clauses would show a word order different from that of other clauses in RSL. One should mention, however, that even in locative clauses, the SOV order is not unexpected because of the tendency to place animate Figures before the Grounds (Volterra et al. 1984). Therefore, we are left with only two clauses that do not use this locative word order. However, it is also curious that the OSV order which is very prominent in locative sentences in other SL (section 2) was not used.

Closer inspection of the prosody of locative clauses revealed an interesting pattern. As I have shown above, in non-locative clauses most pre-verbal objects (the SOV order) are not separated from the verb by a prosodic boundary in the corpus. Interestingly, in locative clauses with the same word order, almost half of the objects are followed by such a boundary (see Table 13).

Table 13: Prosody with the OV order in locative clauses (corpus data)

OV

1 EDU

2 EDU

total

13

10

23

%

56

44

100

Thus, there is at least a prosodic difference between locative clauses as a group and non-locative clauses.

The experimental data showed a different pattern with respect to word order: the OSV order is the most frequent one when both arguments are expressed (8 clauses out of 23) (33). The other word orders are SOV (34), and, in 8 clauses, OV (35); for the latter cases, it is impossible to deduce whether they are underlyingly SOV or SVO. There is also one case each of the SVO order (36), of the OVS order (37), and of the OVSV order (38).

Thus, the OSV order is the one usually used in locative clauses in the experimental data. In contrast, in non-locative clauses this order is used in only two of the cases (see Table 7), and both times the objects are non-manually marked, which may be a sign of topicalization (39).

Animacy of the subject influences its position relative to the object in locative clauses in RSL (as in other SLs). In 5 cases, animate subjects are placed before the object (the SOV and SVO orders), and in 5 cases after the object (the OSV, OVS, OVSV orders). In contrast, inanimate subjects (5 cases) were consistently placed after the object (the OSV order).

I must conclude that the results of the experiment and of the corpus analysis are different. On the basis of the corpus data, it is impossible to conclude that locative clauses in RSL differ from non-locative clauses syntactically/with respect to word order (prosodically they do), but on the basis of the experimental data, it is clear that locative clauses in RSL are created using the same mechanism (i.e. Ground-figure order) as in other SLs. This discrepancy can be explained, however, once we take into account that RSL (and probably other SLs as well) uses two strategies for creating locative clauses, which I will refer to as the ‘syntactic strategy’ and the ‘spatial strategy’.

According to the syntactic strategy, locative clauses are created by means of the same rules as other clauses. When a signer uses this strategy, s/he neither uses signing space nor a simultaneous construction. As the rules are the same as in other clauses, the word order will most likely be SOV for RSL, because verbs in locative clauses are almost always classifier constructions. Consider example (40), illustrated in Figure 2.

FIG2

Figure 2: The cat is on the chair (syntactic strategy)

In this example, the word order is SOV, as it would be in a non-locative clause with a classifier predicate. Note that the signer does not use the signing space for localization of referents. The sign chair is not localized in a specific location; rather, it is articulated in neutral space, slightly to the right of the signer. However, the classifier construction cl:sit is not directed to the right, that is, it does not spatially agree with the location of the chair. Also, there is no simultaneity in this example.

The spatial strategy is a universal visual strategy, probably determined by the cognitive mechanisms of representing locative situations (Laudanna & Volterra 1991, also see Perniss 2007). According to this strategy, the bigger object (Ground) is articulated first, followed by the Figure (mobility); also the animate object is mentioned first (animacy). Therefore, the word order in a clause created on the basis of this strategy will be OSV, or SOV with animate subjects. A signer using this strategy locates referents in space and uses these locations to express the spatial relation between the referents. Simultaneity is also likely to be used. This strategy is employed in example (41), illustrated in Figure 3.

FIG2

Figure 3: The cat is on the chair (locative strategy)

Example (41) is identical in content to (40). However, the word order here is OSV, as the Ground is mentioned first. The signer located the chair to the right in the signing space, and then the classifier construction is directed towards this location. Another example of the spatial strategy was presented in Figure 1, where a simultaneous construction was used.

It should be emphasized again that, if we look at word order only, it is not always possible to distinguish these two strategies. The SOV order can be used with both strategies if the subject is animate. However, the OSV order unambiguously identifies the spatial strategy, and the SVO order the syntactic strategy. Moreover, the use of space and simultaneity can distinguish these two strategies. A different question is whether there is a sharp boundary between the two strategies. Is it possible for a signer to use space actively, including simultaneity, but still stick to the SVO order, or to use the OSV order without using space? The latter seems extremely unlikely. I did not find any such examples in the experimental data. However, this is a question for future research.

As for the two available strategies, a hearing non-signer describing a locative situation with gesture does not have a choice: only the spatial strategy is available to him. In contrast, a signer can choose between the two strategies, and the reasons why s/he decides to use one strategy over another are probably extra-linguistic, or at least extra-syntactic. It is reasonable to suppose that when a signer is asked to describe a single picture with a locative situation, the spatial strategy is likely to be used, as it is natural to use a dedicated strategy to describe a spatial situation. When a signer is telling a story, however, the situation is different. The story usually does not consist exclusively of locative situations; it consists of a series of events which are signed using the syntactic strategy. When a locative situation appears among other non-locative events, it is also likely to be signed using the syntactic strategy, as switching between strategies is cognitively demanding. This might explain why locative clauses in the corpus data did not appear to differ from non-locative clauses, while in the experimental data they were clearly different.

Table 14: Spatial and syntactic strategies

Spatial strategy

Syntactic strategy

Order

OSV or SOV order

SVO or SOV order

What determines word order

Universal principles (mobility and animacy)

Language-specific syntactic and semantic rules

Space

Active use of space

Less or no use of space

Simultaneity

Active use of simultaneity

No use of simultaneity

Used in what circumstances?

When describing single spatial situation

In narratives

Table 14 describes the distinction between the two strategies in RSL. However, it is likely that these two strategies are available in other SLs, too. The only cell in this table that may require modification to adapt it for other SLs is the word order in the syntactic strategy, as this is a language-specific feature.

4.4 Doubling

In this section, doubling of the verb (or the nominal predicate) is discussed. As I have mentioned in section 3.4, I decided to consider sequences containing two occurrences of one verb to be a single clause, if these occurrences are only separated by the arguments or adjunct of this verb, and if the occurrences are either identical or different only in morphological or non-manual marking. That is, if between the two occurrences of a verb referring to one situation another verb appears, I did not analyze this sequence as one clause containing doubling, but as three separate clauses. In making this decision, I do not deny that the mechanisms governing doubling in discourse are principally different from the mechanisms governing doubling in syntax (in clauses), but this research was only focused on the syntax of RSL.[18]

Doubling was analyzed only on the basis of the corpus data because prosody was an important parameter. In particular, I was interested in whether the occurrences of the doubled element would be separated by a prosodic boundary and whether the placement of this boundary can be used to determine what was the base position of the doubled element.

Before discussing doubling of verbal predicates, I will briefly mention doubling of nominal predicates. The corpus contained two clauses where the nominal predicate was doubled. In both cases, the occurrences of the predicate were identical. In the first case the clause constituted one EDU (42), while in the second case there was a prosodic boundary between the occurrences of the predicate (43)[19].

Verbal predicates are repeated in 21 clauses in the corpus. In 14 cases the occurrences of the verb are identical, so they can be classified as verbal echoes (44).

In the other 7 cases the occurrences were different, and what is important, the second occurrence was always more marked (which is in line with the findings of Fischer & Janis (1990) for ASL). In two clauses the second occurrence was inflected for aspect: once progressive aspect (45), and once distributive aspect (46)[20].

In two cases the second occurrence of the verb was marked with a meaningful (emotional) non-manual expression (47); see Figure 4 for illustration of the two occurrence of the verb look.

FIG2

Figure 4: The difference in the non-manual expression between the first and the second occurrences of the verb look

In three cases the occurrences of the classifier construction which was doubled (X3-30, X2-30, Z3-32) were different in the shape of the movement in that the second occurrence contained a more iconic, detailed movement (48).

In one of the clauses with doubling, the occurrences of the verb were adjacent (G2-16). In 16 cases the object was placed between the occurrences (the VOV sequence; see e.g. (46) and (47)), and in 4 cases an adverb (the VAdvV sequence; e.g. (44)).

If we look at the prosody of the clauses with doubling, we can see that the picture is quite diverse. Consider Table 15 showing the prosodic patterns of the cases in which the object or the adverb intervened between the occurrences of the verb:

Table 15: Prosody in clauses with doubling

Object

Number

Adverb

Number

V O V

2

V Adv V

1

V/ O/ V

6

V Adv/ V

2

V/ OV

4

V/ Adv V

1

VO/ V

3

OV/ OV

1

Total

16

Total

4

In most cases both occurrences of the verb and the object constituted separate EDUs. V/OV and VO/V boundary placement were less common, while the situation with all elements included in one prosodic unit was even less common. When the adverb was placed between the occurrences of the verb, there was usually a prosodic boundary in the clause, too.

To sum up, there are three observations that can be made concerning doubling of the predicates. Firstly, doubling of predicates is a fairly common clause-level phenomenon in our corpus (it appears in 21 out of 773 clauses), and in most cases the occurrences of the predicate are identical. Secondly, when one of the occurrences of the predicate is marked (manually or non-manually), it is always the second one. Thirdly, in most cases the clause with doubling constitutes more than one EDU and the placement of the prosodic boundaries in the clause seems arbitrary. The last two observations will turn out to be relevant for the discussion of the basic word order in the next section.