[Lingtyp] Response to Dryer

Fri Jan 22 16:56:58 UTC 2016

Much of this discussion is of course philosophy of science 101 - how is that we construct generalizations based on particulars.

James’s statement of this in his work on Radical Empiricism still is one of my favorites: "Just so, I maintain, does a given undivided portion of experience, taken in one context of associates, play the part of a knower, of a state of mind, of 'consciousness'; while in a different context the same undivided bit of experience plays the part of a thing known, of an objective 'content.' In a word, in one group it figures as a thought, in another group as a thing. And, since it can figure in both groups simultaneously we have every right to speak of it as subjective and objective, both at once. (James 1912, Essay I)"

On Jan 22, 2016, at 11:32 AM, Jan Rijkhoff <linjr at dac.au.dk<mailto:linjr at dac.au.dk>> wrote:

This is just a quick response to Matthew about the position of adjectives in the noun phrase (I may give a more elaborate response when I have more time).
Perhaps the more general question to some of the contributors to this discussion is: does Morphosyntactic Typology make sense, if it were true that there is no empirical basis for a comparison, i.e. if actual linguistic forms and constructions are basically incomparable within and across languages?

First (adjective): recall that I only took into account cases that involve un-derived members of the word class Adjective (thus ignoring, for example, classifying/relational adjectives like ‘evolutionary’ in ‘evolutionary biology’).

Second (relative clauses): I do not say that relative clauses cannot occur in between num and N (see, for example, Rijkhoff 2004: 309-311).

My claim is: if a language has a dedicated class of (qualifying) adjectives, a modifying adjective will occur next to the head noun in the NP (‘Principle of Head Proximity’), as in these ‘iconic’ patterns:

   dem num A N     dem A N num     num A N dem     A N num dem
   dem num N A     dem N A num     num N A dem     N A num dem

If a language employs members of these four word classes (which are not regarded as members of semantic categories here) in the same NP, then we should only find these eight word order patterns, because they reflect scopal differences among the various modifiers (‘Scope Principle’) - but only in the case of simple (non-complex), integral (hierarchically organized) NPs (‘Principle of Domain integrity’).* This is true for approximately 50% of the languages in my sample (Rijkhoff 2004: 328; apparent counterexamples to my claim of ‘adjective-noun adjacency’ are discussed on pp. 266-276). The other languages were excluded for various reasons, such as:

- use of verbal or nominal (i.e. complex, phrasal) constructions to express cardinal or adjectival notions (see examples in my first contribution to this discussion);
- co-occurrence restrictions: in some languages a NP can only accommodate one or two modifiers (see e.g. Foley on Yimas);
- constituent has no attributive use (e.g. in Hixkaryana ‘demonstratives are not used adjectivally’);
- the ‘modifier’ is expressed as a bound form (affix, part of compound);
- the ‘modifier’ is an appositional element (rather than a fully integrated constituent of the noun phrase; this is often the case in Australian languages and in languages that have a numeral classifier construction after the noun).

As far as I can remember, these factors do not a major play a role in Matthew’s word order studies, but in my view they are highly relevant in a cross-linguistic investigation of word order patterns. I have discussed them in various places (see especially Rijkhoff 2004: 327-332; also Rijkhoff 2015).

So I had to exclude 50% of the languages in my sample so to be able to make a responsible cross-linguistic comparison regarding the relative order of demonstrative, numeral, adjective, and noun in a simple, integral NP. I think this shows that, if we are careful enough**, we can compare actual forms and constructions across languages. This is possible if we let go of the idea that we can compare everything in all languages. We need to systematically reduce the data we collected for a morphosyntactic investigation until we are left with a database that contains only those units that are similar enough (in terms of function, meaning and form) to allow for a responsible cross-language comparison. Obviously the elimination procedure should be fully transparent, so that others can judge for themselves if the remaining data are similar enough to be compared for the purpose of the investigation.

* In NPs that cannot be characterized as ‘simple (non-complex), integral (hierarchically organized) NPs’, the influence of other factors (e.g. heaviness) may interfere with the Scope Principle.
** If this discussion continues long enough, I may outline what this careful approach involves.

Jan Rijkhoff

Rijkhoff, Jan. 2004. The Noun Phrase. Oxford: Oxford University Press. [Expanded paperback of the 2002 Hb edition]

________________________________
From: Lingtyp [lingtyp-bounces at listserv.linguistlist.org<mailto:lingtyp-bounces at listserv.linguistlist.org>] on behalf of Matthew Dryer [dryer at buffalo.edu<mailto:dryer at buffalo.edu>]
Sent: Friday, January 22, 2016 4:24 AM
To: lingtyp at listserv.linguistlist.org<mailto:lingtyp at listserv.linguistlist.org>
Subject: [Lingtyp] Response to Rijkhoff and LaPolla

I want to return to near the beginning of the recent discussion, when both Jan Rijkhoff and Randy LaPolla raised issues in response to my drawing attention to the fact that Greenberg’s claim that verb-initial languages tend to place the adjective after the noun is misleading (since they do so to the same extent as verb-final languages).

There are two assumptions that Jan and Randy seemed to make that I think are at the heart of the discussion. The first assumption may not in fact be one that either Jan or Randy made, though it would not be an unreasonable assumption to make, given my published work.

The first apparent assumption is that in collecting data on the order of adjective and noun, what I did was to collect this data without paying attention to the grammatical properties of these “adjectives” (what I prefer to call semantic adjectives, though the label is not ideal), particularly the extent to which they behave like verbs and the extent to which they behave like nouns. The second assumption is that the word order behaviour of semantic adjectives varies considerably depending on these other properties. Both of these assumptions are false.

With respect to the first assumption, I have spent considerably more time over the years examining the grammatical properties of semantic adjectives than I have examining the order of adjective and noun, primarily the extent to which they behave like verbs or nouns.  I have definitely not ignored these differences. My reason for spending so much time on this is precisely because I was curious whether there were differences in word order patterns varying with the extent to which semantic adjectives exhibit verbal or nominal properties. The overall conclusion from this is: very little.  The word order patterns of semantic adjectives are very similar, regardless of whether they are verbal, nominal, or neither. Let me discuss briefly two examples of contrary claims in the literature.

In the paper of his that Randy provided a link to, he cites Van Valin (1986) and Lehmann (1986) as providing evidence that ‘adjectives which are highly nominal in nature precede the noun they modify in OV languages, while those which are highly verbal in nature follow the noun they modify’. However, on the basis of a sample of approximately 300 languages in which the semantic adjectives exhibit nominal features, it turns out that this is not true. In my data, adjectives which are nominal in nature more often follow the noun, like adjectives which are verbal in nature. There is one area of the world where we do find a lot of OV languages that are AdjN and in which the adjectives are often nominal, and that is Eurasia excluding Sino-Tibetan and other families of southeast Asia. I suspect that Van Valin’s and Lehmann’s conclusions were influenced by this pattern, which I suspect is simply a reflection of two accidentally coinciding common traits throughout this area of Eurasia, namely OV&AdjN order and a tendency for semantic adjectives to be nominal in nature. It is true that the preference for NAdj order is higher in my data among languages in which the semantic adjectives are verbal, though the difference falls short of statistical significance. It is possible that this difference is real and if so has a ready explanation: relative clauses show a strong tendency to follow nouns among all languages and getting a higher proportion of NAdj order among languages in which semantic adjectives are verbal is hardly surprising. But the difference is much less than the quote from LaPolla would suggest.

The second case involves the position of semantic adjectives relative to the numeral and noun when both occur on the same side of the noun. The usual pattern is for the adjective to occur between the numeral and the noun, in either Num-Adj-Noun or Noun-Adj-Num order. Rijkhoff (2008: 804) claims that exceptions to this, namely Adj-Num-Noun or Noun-Num-Adj, must involve distinct syntactic constructions, such as ones where the adjective is verbal and a type of relative clause. However, in an almost-completed paper of mine reporting the order of demonstrative, numeral, adjective and noun in a sample of over 500 languages, I find little evidence to support this claim of Rijkhoff’s. There are a very small number of cases where his line of explanation is plausible but they are clearly the exceptions. Let me present two kinds of evidence from my paper bearing on this. The data in the following table distinguishes semantic adjectives with verbal properties from semantic adjectives lacking verbal properties, restricting attention to languages where the adjective and numeral occur on the same side of the noun. The numbers are numbers of genera rather than numbers of languages:

Adj closer to noun
than Num

Num closer to noun
than Adj

Percent
closer

Adj is verb

30

4

.88

Adj is not verb

106

12

.90

As the table shows, semantic adjectives which are verbs show the same strong tendency to occur closer to the noun than the numeral that we find with semantic adjectives which are not verbal. And the majority of cases where the adjective is further from the noun are ones where the adjective is nonverbal. In other words, the verbal nature of some semantic adjectives has little effect on their position relative to the numeral and noun.

Let me share one example from my paper that illustrates that it is only semantics which is relevant, not the particular morphosyntactic properties of semantic adjectives.  The following examples come from Nias (Lea Brown, p.c.). Semantic adjectives in Nias are verbal and when used attributively occur in relative clauses. Adnominal numerals also occur in relative clauses. What the data in (1) shows is that the relative clause containing the semantic adjective must occur closer to the noun than the relative clause containing the numeral. The opposite order, shown in (1b), is judged ungrammatical.

1.

a.

No

u-bunu

n-asu

s=afusi

si=dua

rozi.

past

1sg-kill

abs-dog

rel=white

rel=two

clsfr

    N

   Adj

  Num

                  ‘I killed the two white dogs.’

b.

*No

u-bunu

n-asu

si=dua

rozi

s=afusi.

 past

1sg-kill

abs-dog

rel=two

clsfr

rel=white

    N

   Num

   A

                  ‘I killed the two white dogs.’

However if we examine the position of relative clauses involving a verb that is not a semantic adjective, we find the opposite pattern. In (2), the relative clause involving the verb for ‘sleep’ must occur outside the relative clause containing the numeral. The opposite order, in (2b), is judged ungrammatical.

2.

a.

No

u-bunu

n-asu

si=dua

rozi

si=mörö.

past

1sg-kill

abs-dog

rel=two

clsfr

rel=sleep.

    N

  Num

    Rel

                  ‘I killed the two dogs that were sleeping.’

b.

*No

u-bunu

n-asu

si=mörö

si=dua

rozi.

 past

1sg-kill

abs-dog

rel=sleep

rel=two

clsfr

   N

    Rel

  Num

                  ‘I killed the two dogs that were sleeping.’

In other words, despite the fact that all these modifiers involve relative clauses, the one involving a semantic adjective must occur closer to the noun. Even though these semantic adjectives are verbs, they occur closer to the noun, unlike relative clauses involving other verbs. My paper presents data from a number of other languages exhibiting a similar pattern.

In a way this should not be surprising. A common hypothesis is that modifiers representing more inherent properties are more likely to occur closer to the noun. Since ‘white’ is more inherent than ‘sleeping’, it is not surprising that it occurs closer to the noun. In other words, it is the semantics that is relevant, not the particular construction.

In short, it is not the case that I ignored the finer properties of semantic adjectives. Rather, I have collected a large amount of data that shows that these finer properties have little affect on their word order position, contrary to the claims of of LaPolla and Rijkhoff.

Matthew
_______________________________________________
Lingtyp mailing list
Lingtyp at listserv.linguistlist.org<mailto:Lingtyp at listserv.linguistlist.org>
http://listserv.linguistlist.org/mailman/listinfo/lingtyp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20160122/4684bf76/attachment.htm>