Grammatical Categories From Parts of Speech to Form-Classes PDF

Development Team Principal Investigator: Prof. Pramod Pandey Centre for Linguistics / SLL&CS Jawaharlal Nehru University, New Delhi Email: pkspande...

Development Team Principal Investigator: Prof. Pramod Pandey Centre for Linguistics / SLL&CS Jawaharlal Nehru University, New Delhi Email: [email protected] Paper Coordinator: Prof. Ayesha Kidwai Centre for Linguistics / SLL&CS Jawaharlal Nehru University, New Delhi Email: [email protected] Content Writer: Prof. Ayesha Kidwai Content Reviewer: Prof. S. Imtiaz Hasnain Department of Linguistics, Aligarh Muslim University [email protected] Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes  Description of Module Subject Name Linguistics Paper Name Grammatical Categories Module Title From Parts of Speech to Form-Classes Module ID Lings_P6(a)_M2 Quadrant 1 E-Text Table of Contents Introduction The early 20th century: Otto Jesperson American Structuralism: Bloomfield (1926) and Fries (1956) A Comparative Evaluation Conclusion References Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes  From Parts of Speech to Form-Classes 1. Introduction The modern dissatisfaction of the classical traditions of part of speech identification was based primarily on the ad hoc-ism that characterised the procedure. Modern grammarians asked a number of questions that could receive no principled answer from either classical European tradition or that of the grammarians. What was the reasoning by which such assignment did not distinguish, say, abstract nouns like happiness or quantifiers like everyone by a distinct part of speech label? What was the principled reason why the number of part of speech labels had to be limited to eight? What role did parts of speech have in the analysis of the formal properties of sentences? These questions led to a questioning of the relevance of the notion of parts of speech itself. 2. The early 20th century: Otto Jesperson Otto Jespersen took the radical step of restricting the relevance of the notion ofparts of speech to the syntax.. Although he still assigned part of speech labels to words, Jespersen propounded a view of grammar in which syntactic-functional relations were encoded only at the level of syntax, whereas part of speech assignment made reference to only morphological form and lexical semantics. The rules of grammar map certain parts of speech onto distinct syntactic functions, but these syntactic functions are not included in the definitions of parts of speech. A brief summary of Jespersen's theory of syntax is in order before we proceed. In his conception, phrases (and sentences) are not merely an arrangement of words, but rather they encode semantically hierarchical relationships, where one word is defined or modified by another word, which in turn may be defined or modified by a third word, and so on. This leads to words being ranked as primary, secondary, and tertiary, where a primary word (in his terminology, a principal) is one that makes the most special(ised) contribution to sentence meaning. A secondary word (an adjunct), on the other hand, is one that modifies or defines a principal, and a tertiary word (a subjunct) is one that modifies/defines an adjunct. Thus a phrase like [a slightly unhappy student], slightly is a subjunct, a and unhappy are adjuncts, and student is Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes  a principal. A secondary may be joined to a primary in two ways: junction (attributive) and nexus (predicative). Jespersen's idea was that syntactic ranks are fundamentally insensitive to part of speech information, thus disputing the assumption that occurrence in the same function necessarily entails an identical part of speech assignment. Jespersen’s own classification is accomplished purely on the basis of morphological distribution and lexical meaning—substantives, pronouns, adjectives, verbals, adverbs, prepositions, conjunctions, interjections—and is in effect not so different from Sweet’s, except that his reference to form and lexical meaning alone does not suffer from allowing overlapping classes. For Jespersen, for example, a verbal participle would be classed under verbids (a sub-class of verbals). The fact that this form could, in the syntax, be ranked as a principal (Reading is fun) or an adjunct (the reading boy) can at best be described as a tendency, just as could be the frequency of principals being nouns. However, nothing in Jespersen’s approach could explain why adjectives or prepositions could never be primaries, adverbs and interjections as nexus, and so on. Thus, Jespersen’s strict exclusion of form-related information from the operations of the syntax deprives the grammar of valuable syntactic generalisations. 3. American Structuralism: Bloomfield (1926) and Fries (1956) The methodology advocated by Bloomfield in 1926 marked modern linguistics’permanent rupture with the Greek\Latin classical tradition, at least as far as the value assigned to of parts of speech units of basic grammatical analysis was concerned. Bloomfield’s (1926) appeal to syntax as the primary parameter is more explicit than Sweet’s. Based on a view of the sentence as a sum of minimum free morphologic constructions (=words) concatenated into phrases (his non-minimum syntactic construction) and arranged in a certain order, Bloomfield defines parts of speech through the criterion of positioning (ordering) in a sentence. His reasoning goes as follows (1949: 198-199): (1) Bloomfield’s part of speech algorithm a. Since not every word/phrase can occur in all the ordered positions in a syntactic construction (i.e. Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes  phrase or sentence), it is necessary to associate each syntactic position with a functional meaning. b. Consider every word/phrase that can occur in a particular syntactic position as having the function of encoding the functional meaning associated with that syntactic position. c. Consider all words/phrases having the same function as constituting a single form-class. d. Define the class-meaning as the union set of functional meanings a form-class can have. e. When a form-class contains relatively few forms, the meanings of these forms may be called sub- categories. f. If we keep to the level of the word (“a minimum free form”), then all the morphologic construction members of a form-class constitute a word-class. g. The maximum word-classes of a language are the parts of speech of the language. To understand Bloomfield’s method, let us take the example of adjectives (e.g., angry) and verbal past participles (e.g., chastised), both of which may modify the noun head of a noun phrase. The algorithm in (1) will ignore the fact that there are clear meaning differences between these two modifiers—the pure adjective is purely [descriptive], but the participial always denotes a [result] in that we presuppose that a prior event has resulted in the state denoted by the participle (e.g., chastised necessarily presupposes an event of scolding having taken place). Instead, Bloomfield’s algorithm will initially focus only on the fact that both these classes can occur between the article and the noun, i.e. in the frame in (2). We can work out the algorithm as in (3): (2) [The ___ man] (3)a. Assign functional meaning to ___: MODIFICATION b. Identify adjective and verbal participles as having the function: MODIFIER c. Group adjective and verbal participles as one [FORM-CLASS: MODIFIER] d. Two class-meanings in the FORM-CLASS: MODIFIER: ‘descriptive’ and ‘result’ e. Two SUB-CATEGORIES in the FORM-CLASS: MODIFIER: [descriptive] and v [result] f. One WORD-CLASS in the FORM-CLASS: MODIFIER = [ADJECTIVE] It should be apparent that Bloomfield’s algorithm fails to distinguish verbal participles and adjectives as Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes  distinct parts of speech; and this fact continues to hold even though adjectives and verbal participles occur in other syntactic constructions. For example, adjectives may also occur in the comparative-superlative frame, one from which verbal participles are barred. In addition, Bloomfield also considers articles and demonstratives as a special sub-category of adjectives, thereby invoking a third sub-category [delimiting], but the expansion of subcategories leaves the conclusion in (3f) unaffected. This is because of the way that form-class is defined in (1c), by which there is no requirement that all members of the form -class must all individually exhibit each one of the set of functions; rather, the definition here only requires that members of a form-class share just one function —MODIFIER, in this case. As a consequence, Bloomfield’s approach in effect creates overlapping form-classes, and therefore parts of speech—a verbal participle may be therefore a member of the part of speech [VERB] as well as [ADJECTIVE]. Fries (1956) represents a rather extreme structuralist implementation of the methodology of distributional analysis and substitution advocated by Bloomfield. Taking the distribution of a word as the position of a word in the sentence, and eschewing all reference to lexical meaning, Fries sets out three basic frames for the determination of form classes: Frame A: The concert was good (always). Frame B: The clerk remembered the tax (suddenly). Frame C: The team went there. Fries’ method identifies nineteen form-classes, of which the first four may be termed major word-classes, as they contain 67% of total instances of vocabulary: Class 1: Words that can substitute for concert (e.g. food, coffee, taste,etc.) and words that can substitute for clerk, tax and team Class 2: Words that can substitute for was, remembered and went. Class 3: Words that can substitute for good. Class 4: Words that can fill the position of there. Other classes that may be identified include determiners; modal verbs; the negative particle “not”; adverbs of degree; coordinating conjunctions; prepositions; the auxiliary verb “to”; the introductory “there”; interrogative pronouns and adverbs; subordinating conjunctions; interjections; the words “yes” Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes  and “no”; the so-called attention-giving signals: look, say, listen; the word “please”; the forms “let us”, “lets” in request sentences. Objections to Fries’ methods abound, not the least of which is the fact that his frames did not test all the relevant positions, given that only intransitives and transitive verbs were used. It results in classes (of varying sizes) that had no characterising principle of class membership other than syntactic position\function, i.e. “faceless” classes, in that no unique form or function could be isolated as their unique characterising property (except for the many single-membered classes). Moreover, as Hockett (1958) points out neither the inclusion nor exclusion from the form-class is arrived at on a principled basis: “[The pattern of interchangeability] defines a form-class which includes she, he, it, John, Mary, the man at the corner, my friend Bill, and so on endlessly, but which by no means includes all forms, since we can name many which are excluded: her, him, them, me, yes, no, ripe, find her, go with us tomorrow.” This strategy would therefore predict that her and she, him and he, are never together members of the same form-/word-class, an analysis that cannot be justified, either based on structuralist morphology (where these forms would be grammatically conditioned allomorphs) or native speaker intuitions. At a more general level, implicit in the structuralist conception of language was the conception that syntax does not care if the objects it ordersare sentences, phrases, or words—they aremerely different ‘sizes’ of items. Jespersen’s complete separation of the language of syntax from the language of morphology becomes significant in this regard, because his claim is that syntax is about relations between forms, and not the forms themselves—occurring in the same function makes no predictions about the form-classes to which a particular morpheme\word belongs. 4. A Comparative Evaluation A fact not emphasized enough in the linguistics literature is that both the descriptivist and structuralist methodologies ended up characterising the identification of parts of speech or word-classes as a procedurally latergeneralisation, i.e. closer to the ‘surface’ of grammatical theorising. Thus for example in Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes  Jespersen and Sweet’s methods, forms were first divided into whether they could bear inflection or not, and only then were parts of speech identified. In Bloomfield’s system, the identification of form-class and then word-class was a necessary prerequisite for part of speech identification, and the whole procedure starts by the identification of syntacticfunction. It is therefore possible to consider whether parts of speech, or even form- and word-class, can be seen as the basic units of grammatical analysis. Figure 1. Pāṇini’s model of grammar (from Kiparsky 2009) By asking this question, we would move closer to the ‘other’ classical tradition—the Pāṇinian one. As Kiparsky (2009) shows, Pāṇini’s negative answer to the question was based on a model of grammar (Figure 1), in which the start point is given by a level of representation that encodes abstract (sentential) semantic relationsand abstract morphological features, and then proceeds derivationally through subsequent levels of representation to build an ultimate phonological representation. In such a model, searching in the outputs of the last stage of derivation for the basic units of grammatical analysis is bound to be fruitless. The extent to which Bloomfield’s ideas are similar to those of Pāṇini—he too takes the level of functional (sentential) meaning as the starting point of his algorithm—will be apparent; what is different is the former’s reluctance to abstraction, because to admit a feature like [abstract tense] into Bloomfield’s system would be a gross violation of the positivist tenets of structuralism. In the absence of a sufficient degree of abstraction however, linguistic theory cannot be built; a purely positivistic methodology such as that instantiated by Fries and Bloomfield above will lead to no significant generalizations at all. In all such abstraction, however, as our brief critique of Jespersen’s separation of syntax from morphology has shown, a principled mapping of morphological and syntactic information must be found. The lesson to take away from the bird’s eye-view of the traditions of grammatical inquiry that we have examined in this unit is that while morphological and syntactic form and function should both be Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes  necessary conditions for defining grammatical categories, stating grammatical categories in terms of the word is not the optimal strategy. Neither is an approach that imposes the structuralist dogma of what can be seen, heard and measured too literallythe best way to reach the proper unit of description and analysis. 5. Conclusion In the unit, we have traced the how the word-based conceptions of parts of speech were displaced by the spread of American structuralism, which advocated a strictly distributional approach to the identification of form-classes. The unit concludes with a comparative evaluation of the approaches discussed thus far, where it is suggested that an adequate theory of the basic units of linguistic analysis must move away from word-based approaches and admit some degree of principled abstractness in its postulations. Paper : Grammatical Categories Linguistics Module : From Parts of Speech to Form-Classes