Pernilla Hallonsten Halling

[attr pred] construction overlap

The overlap of attr and pred is the second most common overlap at the construction level, attested in 12 sample languages. This is a peculiar overlap for a number of reasons. Firstly, it is attested in examples where the interpretation is ambiguous between an NP in attr or and a whole clause in pred. For instance, an individual example may mean either ‘the sad person’ or ‘the person is sad’. Secondly, and following from the first point, in order to have an overlap of attr and pred, the language in question cannot have an obligatory copula. An obligatory copula is expected to appear in pred, but not in attr. The presence of a copula thus clearly separates the encoding in the two functions. Languages with a constructional overlap of attr and pred differ remarkably with regard to whether it appears to be the only strategy, or whether they exhibit other alternatives for attr and pred, respectively. As stated in the previous section, I do not attempt to fully establish this difference, but it is still interesting to notice the tendencies in certain languages. In Maltese (Afro-Asiatic), only context appears to separate attr from pred. In Warekena (Arawakan), only a few Adjectives are used in the construction where attr and pred overlap, whereas other constructions are clearly separable (see examples from these two languages in appendix B.) In this section, I present examples from two languages, one in which the encoding pattern is entirely ambiguous between attr and
pred (Bora), and one where the encoding appears to be ambiguous between the two,
although this is not made explicit in the source (Gooniyandi).
We begin by looking at Bora (Witotoan), where the construction in question is am- biguous in terms of whether it instantiates attr or pred.
(7.2) Bora (Witotoan) (Thiesen 1996: 50) [attr pred]

Ímí
good
cáraca.
chicken
attr/pred

‘the good chicken/the chicken is good’
In the example in (7.2), the descriptive root ímí is used, and the example could be either
an instance of attr or of pred. However, this is not the only option for encoding
the attr and pred functions. In order to avoid this ambiguity, in attr, the property word ímí is combined with a classifier, which turns it into a so-called ‘qualifier’ (Thiesen 1996: 50).

(7.3) Bora (Witotoan) (Thiesen 1996: 50)

Ímí-ibye

good-3sg.m

cáraca
chicken
majchó.
eat
attr

‘The good chicken eats.’

According to Thiesen & Weber (2012), qualifiers can be analyzed as NPs that are in apposition with the NP that they modify, rather than being a modifier to it. Bare property words, as in example (7.2), appear to be more commonly used in pred than in attr. The overlap and the alternative construction with a classifier in attr are captured in constructional-typological terms in (7.4), where the overlap is also highlighted. As it is not possible to distinguish attr from pred, the item that is modified in the former and of which the property is predicated in the latter is labeled as ‘N/S’ for ‘noun or subject’ in both functions. While it does not really make sense to denote an item in attr as ‘S’, or to label the subject with ‘N’, it is necessary to do so here in order to capture the ambiguity, that is, the total overlap.

(7.4) attr and pred in Bora

attr

Function: property modification within referring expression

Form 1: G.MOD N/S

Form 2: G.MOD-clf N Example: (7.2), (7.3)

preD

Function: property predication

Form: G.MOD N/S

Example: (7.2)


In other languages, encoding used in attr and pred appears to be exactly the same, even though the encoding is described separately for the two functions in the language description. This is the case in Gooniyandi (Bunaban), where Adjectives cannot really be distinguished from Nouns (cf. section 6.4.1). In pred, the property word, which may carry a clitic (see 7.5b below), appears to follow the subject, as illustrated in (7.5). In attr, the order of the property modifier and the head is free, illustrated with examples (7.6–7.7) (McGregor 1990: 456-457). This means that in cases where the property word, without a clitic, follows the Noun, the example is ambiguous between attr and pred (Bill McGregor, p.c.).

(7.5) Gooniyandi (Bunaban) (McGregor 1990: 456-457) [attr pred]

1. 
   yaanya
   

other
biddinyiwoorloo
wasp
thiwa
red
yaanya
other
biddinyiwoorloo
wasp

gooroogooroo
black
‘One wasp is red, another is black.’
pred

2. 
   minyawoo
   

cat
wajaddanyi
different
lambardi
little
thadda
dog
nyamani=nyali
big=rep
pred

‘Cats are different, they’re little; dogs are big.’
(7.6) Gooniyandi (Bunaban) (McGregor 1990: 265) [attr pred]

jalandi
belt
gooroogooroo

black

attr

‘a black belt’
(7.7) Gooniyandi (Bunaban) (McGregor 1990: 297)²

ngirndaji
this
labawoo

white

jiga
flower
attr

‘This is a white flower.’
Structurally, attr and pred have exactly the same appearance, in the case where the property word follows the head in attr. This seems to be a constructional overlap, as illustrated in the notation in (7.8). The label ‘N_prop’ is used here since Adjectives cannot be distinguished from Nouns in Gooniyandi. The ‘N/S’ label is used on the same grounds as for Bora, described above.
(7.8) attr and pred in Gooniyandi

attr

Function: property modification within referring expression
Form 1: N/S N_prop Form 2: N_prop N Example: (7.6), (7.7)

preD

Function: property predication Form: N/S N_prop(=cli) Example: (7.5)
As discussed in chapter 6, Bora has root and lexeme level overlaps of all three functions and Gooniyandi has root and lexeme level overlaps of attr and pred. The overlap of attr and pred appears to imply the same or a greater overlap on the root and lexeme levels.
In table 7.1, all languages with an overlap of attr and pred on the construction level are listed. A map with the same languages is found in figure 7.1. The constructional notation for the examples in each respective language is included in the table. The rightmost column indicates whether or not the overlap in question is stated in the source. Where this is the case, the example is explicitly treated in the source as ambiguous
between attr and pred, or the encoding of the two is stated to be identical. The
distinction between stated and non-stated overlaps is only indicated for the overlap of
attr and pred, and not in the following sections, since it is only here that it applies.
² This example is repeated from (6.15b) in section 6.4.1.

Table 7.1. Languages with an [attr pred] overlap on the construction level

Languages Construction Stated in source

Bininj Gun-Wok (Gunwinyguan) pro-N/S-ADJ(-tns.asp) Bora (Witotoan) MOD N/S
Gooniyandi (Bunaban) N/S ADJ ³
Imonda (Border) N/S ADJ-l _nmlz
Kewa (Nuclear Trans New Guinean) ADJ N/S; N/S ADJ X Koyra Chiini (Songhay) N/S ST.V
Lahu (Sino-Tibetan) N/S ST.V ve_nmlz/rel
Maltese (Afro-Asiatic) N/S ADJ.num.gen
Nuu-chah-nulth (Wakashan) ST.V=ind.per.num N/S Slave (Athapaskan-Eyak-Tlingit) N/S ST.V
Warekena (Arawakan) N/S ST.V-mi _DER
Yimas (Lower Sepik-Ramu) N/S ST.V-k _irr-ncl.num

As illustrated in table 7.1, in more than half of the languages with a constructional overlap of attr and pred, examples are explicitly stated to be ambiguous between the two functions. The rest of the languages have encoding that appears to be the same for attr and pred, and which would be ambiguous in meaning between attr and pred if confirmed. This expectation is implied in the constructional notation in table 7.1. The total overlap of attr and pred may seem surprising, since attr encodes a phrase and pred a whole clause. As soon as the NP in attr is expanded with other elements, or the tense is changed in pred, differences are bound to be exposed. The question may be raised of how far the generalization of ambiguity between attr and pred holds, and whether it only applies to one or a few examples in a specific language. Based on the present analysis, it appears that this type of encoding overlap can be assumed to be fairly local. Alternatively, it could be argued that the constructions attested in attr and pred are partly homonymous, and that a wider range of examples, for instance, with pronouns or other tenses, would expose this homonymy. In other words, it may be the case that in languages where certain grammatical markers are absent, or where a certain word order holds, the encoding of attr and pred simply coincides. Regardless of what perspective is taken, the examples attested do instantiate encoding overlaps of attr and pred. These results, and the problems that may follow from them, do not make any essential contribution to the discussion of adverbs, which is why these problems will not be discussed any further here.

³Although this overlap is not stated in McGregor (1990), Bill McGregor (p.c.) confirms it.

overlap no overlap

Figure 7.1. Languages with [attr pred] construction overlap

3.