8

Click here to load reader

On "Relatedness Paradoxes" and Related Paradoxes

Embed Size (px)

Citation preview

Page 1: On "Relatedness Paradoxes" and Related Paradoxes

On "Relatedness Paradoxes" and Related ParadoxesAuthor(s): Steven L. StraussSource: Linguistic Inquiry, Vol. 13, No. 4 (Autumn, 1982), pp. 694-700Published by: The MIT PressStable URL: http://www.jstor.org/stable/4178306 .

Accessed: 15/06/2014 18:27

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry.

http://www.jstor.org

This content downloaded from 194.29.185.230 on Sun, 15 Jun 2014 18:27:27 PMAll use subject to JSTOR Terms and Conditions

Page 2: On "Relatedness Paradoxes" and Related Paradoxes

694 SQUlBS AND DISCUSSION

all unbounded feature propagation phenomena are nondirec- tional, and that all apparent directionality of spreading is ex- pressed by local rules which can provide left and right contexts. This would mean that notions like "spread left" would be bro- ken down into a local rule with a right-hand context (like (7) in Capanahua) and a rule (like 14)) that says "[ + X] is general in the Domain [aF]". It is not obvious that tone phenomena can be treated in this way, but the success of this approach for Capanahua suggests an interesting project for future research.

References

Clements, G. N. (1977) "The Autosegmental Treatment of Vowel Harmony," in W. Dressler and 0. Pfeiffer, eds., Phonologica 1976, Innsbruck.

Clements, G. N. (1980) "Vowel Harmony in Nonlinear Gen- erative Phonology: An Autosegmental Model," Indiana University Linguistics Club, Bloomington, Indiana.

Dell, F. (1973) Les regles et les sons, Hermann, Paris. Goldsmith, J. (1976) Autosegmental Phonology, Doctoral dis-

sertation, MIT, Cambridge, Massachusetts. Distributed by the Indiana University Linguistics Club, Blooming- ton, Indiana.

Halle, M. and J.-R. Vergnaud (1980) "Harmony Processes," in W. Klein and W. Levelt, eds., Crossing the Bound- aries in Linguistics, 1-22.

Liberman, M. and A. Prince (1977) "On Stress and Linguistic Rhythm," Linguistic Inquiry 8, 249-336.

Loos, E. (1969) The Phonology of Capanahua and Its Gram- matical Basis, Summer Institute of Linguistics of the University of Oklahoma.

Safir, K. (1979) "Metrical Structure in Capanahua," in K. Safir, ed., Papers on Syllable Structure, Metrical Structure and Harmony Processes, MIT Working Papers in Linguis- tics, Vol. 1, Cambridge, Massachusetts.

Vergnaud, J.-R. and M. Halle (1978) "Metrical Phonology-A Fragment of a Draft," unpublished paper, Department of Linguistics, MIT, Cambridge, Massachusetts.

ON "RELATEDNESS Williams (1981) discusses a number of "relatedness paradoxes" PARADOXES" AND RELATED in English word formation and resolves them by proposing a PARADOXES "new" principle of relatedness, one of whose consequences is

Steven L. Strauss, the abandonment of semantic compositionality. I argue in this University of New Mexico squib (i) that Williams's solution abandons not only semantic

compositionality, but structural (morphosyntactic) well- formedness as well; and (ii) that the correct solution of the "relatedness paradoxes", based on a principle advocated in

This content downloaded from 194.29.185.230 on Sun, 15 Jun 2014 18:27:27 PMAll use subject to JSTOR Terms and Conditions

Page 3: On "Relatedness Paradoxes" and Related Paradoxes

SQUIBS AND DISCUSSION 695

Strauss (1982), allows us to retain both semantic composition- ality and structural well-formedness in word formation.

1. What Is a "Relatedness Paradox"?

According to Williams, a paradox may result when the ordinary principle of morphological relatedness conflicts with ordering restrictions in word formation. The ordinary principle of mor- phological relatedness is this: X is related to Y if X = af Y or Y af, and X and Y share an element of meaning. According to Williams, the ordering restrictions operating on English word- formation rules are as follows:

(1) a. + -affixes must be "inside" #-affixes, inflections, and compounds.

b. #-affixes must be "inside" inflections and com- pounds.

c. Inflections must be "inside" compounds.

(1) is an expanded version of Siegel's (1974) original ordering hypothesis (+ inside #).

As Williams notes, strict adherence to the ordering restric- tions in (1) requires that words like transformational gram- marian, hardheartedness, ungrammaticality, and whitewashed have the structures of (2a), not (2b):

(2) a.

transformational grammar ian

hard hearted ness

un grammatical ity

white wash ed

This content downloaded from 194.29.185.230 on Sun, 15 Jun 2014 18:27:27 PMAll use subject to JSTOR Terms and Conditions

Page 4: On "Relatedness Paradoxes" and Related Paradoxes

696 SQU IBS AND DISCUSSION

b.

+ outside compound

transformational grammar ian

# outside compound

hard hearted ness

+ outside #

un grammatical ity

inflection outside compound

white wash ed

In order to relate these words to transformational gram- mar, hardhearted, ungrammatical, and whitewash, however, the ordinary definition of relatedness requires the structures in (2b), not (2a). Hence the paradox.

Williams's solution to the paradox retains the ordering restrictions of (1), but revises the definition of relatedness in such a way that the words in (2a), despite failing the principle af?X/X af, will still be related to transformational grammar, hardhearted, ungr-ammatical, and whitewash. (See Williams (1981, 261) for the precise formulation of this new definition.) He therefore accepts (2a) over (2b) and notes the concomitant loss of semantic compositionality.

2. Arguments against Williams's Principle of Relatedness

2.1. Relatedness vs. Ordering Restrictions

On a priori grounds, the resolution of such relatedness para- doxes should be sought in a revision of the ordering restrictions, not in the notion of relatedness. This is because the ordering restrictions are absolutely language particular, while the ordi- nary notion of relatedness is based on the principle of semantic compositionality, which is obviously a strong candidate for being a language universal. Regarding the particularity of (1),

This content downloaded from 194.29.185.230 on Sun, 15 Jun 2014 18:27:27 PMAll use subject to JSTOR Terms and Conditions

Page 5: On "Relatedness Paradoxes" and Related Paradoxes

SQUIBS AND DISCUSSION 697

Williams himself notes that Russian and Walbiri inflections may appear inside derivational affixes. Moreover, it is certain that the + /# distinction in English is historically determined, a re- flection of both the Latinate (+) and Germanic (#) origins of the language.

However, Williams does not consider the alternative of relaxing the ordering restrictions (la-c) as the way to solve the paradoxes. This is unfortunate, since a straightforward revision is highly motivated. Notice that all of Williams's ordering vi- olations (2b) have the same structure: a right-adjoining element (+ -suffix, #-suffix, or inflectional suffix) lies outside a left- adjoining element (#-prefix or compound specifier).' In fact, there are no ordering violations involving #-prefixes inside +-prefixes, or #-suffixes inside +-suffixes (e.g. re # en + code, but *en + re # code; satan + ic # like, but *satan # like + ic), etc. Thus, the kinds of violations that occur are systematic, not random. They suggest that the ordering restric- tions do not hold between left- and right-adjoining elements, i.e. that prefixation and suffixation are independent of each other.2 This revision in the ordering restrictions will now permit the structures in (2b) and consequently retain semantic com- positionality.

2.2. In Defense of Strict Subcategorization

Williams's theory leads to more than just a loss of semantic compositionality in word formation. As it turns out, require- ments that affixes attach only to bases of specific syntactic categories (N, V, A, etc.) must also be abandoned. Striking examples of this sort are found among words of the form #-X- +, i.e. bases X with a #-prefix and a +-suffix.

Consider a word such as reanalyzable. Since re- is a #-prefix and -able a + -suffix (cf. reanalyzability), Williams's theory would require that it have the structure shown in (3a), not (3b):

(3)a.

re analyze able

b.

re analyze able

I See Strauss (1982) for arguments that English compounds are formed by left-adjoining a word or compound specifier to a base.

2 In general, it is left adjunction (including compounding) and right adjunction that are independent of each other, with respect to the or- dering restrictions.

This content downloaded from 194.29.185.230 on Sun, 15 Jun 2014 18:27:27 PMAll use subject to JSTOR Terms and Conditions

Page 6: On "Relatedness Paradoxes" and Related Paradoxes

698 SQUlBS AND DISCUSSION

The formative re-, however, does not attach to adjectives (*re- good, *re-long, *re-subtle), yet the only way to interpret (3a) is as a violation of this particular subcategorization restriction:

(4) Adj

Ad

V

re analyze able

Furthermore, this is not an isolated problematic example for Williams's theory. It will obtain whenever a word of the form #i-X- +1 is such that [X- +J] 'is a' category C, to which #i may not attach. Consider the following:

(5) a. mis# V misread, misbehave N misfortune, misbelief

*____ A *misgood, *mislong but mis#[read+ ablelA

b. non# N nonperson, nonword A nonrigid, nonsubtle

*____ V *nonbelieve, *nonspeak but non#[rigid + ify]v

Thus, as with example (4), Williams's insistence on (1) results in word structures which violate strict subcategoriza- tion:

(6)a. Adj b. V

V Ad

ms read able non rigid ify

In order to retain (4) and (6), i.e. structures which do not violate (1), Williams must abandon a uniform interpretation of strict subcategorization in morphology. In particular, his analysis must permit structures some of whose sister nodes are other- wise not permitted to cooccur, a move for which there is no precedent.

This content downloaded from 194.29.185.230 on Sun, 15 Jun 2014 18:27:27 PMAll use subject to JSTOR Terms and Conditions

Page 7: On "Relatedness Paradoxes" and Related Paradoxes

SQUlBS AND DISCUSSION 699

Alternatives to (4) and (6) which do not violate strict sub- categorization are shown in (7):

(7)a. Adj

V

re analyze able

b. Adj

V

V

mis read able'

C. V

Adj

Adj

non rigid ify

In (7a), re# attaches to a verb; in (7b), mis# attaches to a verb; and in (7c), non# attaches to an adjective. All of these examples adhere to strict subcategorization requirements. Furthermore, in none of the examples of (7) is the revised ordering restriction violated (+ can appear outside # provided that one is a prefix and the other a suffix); nor is the ordinary notion of relatedness violated (hence, semantic compositionality is maintained).

Finally, we might ask how the alternative analysis handles (2a). The answer is simple: all of the structures in (2a), with the exception of ungrammaticality, are also permitted, since they violate neither subcategorization restrictions nor (revised)

This content downloaded from 194.29.185.230 on Sun, 15 Jun 2014 18:27:27 PMAll use subject to JSTOR Terms and Conditions

Page 8: On "Relatedness Paradoxes" and Related Paradoxes

700 SQU lBS AND DISCUSSION

ordering restrictions. The representation of ungrammaticality in (2a) is excluded since it violates the restriction on negative un- that it attach only to adjectives.3

Support for this comes from the observation that structural distinctions between (2a) and (2b) correlate with distinctions in semantic interpretation. For example, transformational grammarian can refer either to someone whose vocation in- volves transformational grammar (the (2b) interpretation) or to a grammarian who is transformational (the (2a) interpretation). Similarly, the (2a) interpretation for hardheartedness is a state of heartedness which is hard, while the interpretation for white- washed is a state of being washed white. The fact that these interpretations are odd is irrelevant. The crucial point is that they differ from (2b) interpretations in ways suggested by the distinct structures. In this regard, notice that angrammaticality is unambiguous and that nonrigidify, for example, means (make not rigid); it does not mean (not make rigid). These interpre- tations accord with the proposed structures.

In short, Williams's theory saves strictly language-partic- ular rules (cf. (1)) by sacrificing the universal principles of se- mantic compositionality and strict subcategorization. The pro- posed alternative saves the universal principles by revising the language-particular ones. The revision itself is highly princi- pled. The choice between the two approaches is obvious.

References

Allen, M. (1978) Morphological Investigations, Doctoral dis- sertation, University of Connecticut, Storrs, Connecti- cut.

Siegel, D. (1974) Topics in English Morphology, Doctoral dis- sertation, MIT, Cambridge, Massachusetts.

Strauss, S. (1982) Lexicalist Phonology of English and German, Foris, Dordrecht.

Williams, E. (1981) "On the Notions 'Lexically Related' and 'Head of a Word'," Linguistic Inquiry 12, 245-274.

3Siegel (1974) argues that un- attaches only to adjectives, while Allen (1978) challenges this and permits un-attachment to some noun bases. For arguments against Allen's position, see Strauss (1982).

This content downloaded from 194.29.185.230 on Sun, 15 Jun 2014 18:27:27 PMAll use subject to JSTOR Terms and Conditions