Morphology's place in the grammar

Table of Contents

1 Introduction
1.1 Inflection and derivation defined

2 Distributed Morphology
2.1 How is the grammar organized in Distributed Morphology?
2.1.1 A different kind of morpheme
2.2 Different organization in syntax and morphology
2.2.1 Merger
2.2.2 Fusion
2.2.3 Fission
2.3 How are sounds mapped onto morphemes?
2.3.1 When the Vocabulary is not enough

3 Two types of inflection
3.1 Proving that there are two types of inflection
3.1.1 Lacking forms
3.1.2 Inflectional split
3.1.3 Deflection
3.1.4 Language acquisition
3.2 Making new words with inherent inflection
3.2.1 Plural nouns
3.2.2 Infinitives
3.2.3 Participles
3.3 Split inflection instead of split morphology?

4 Word-class-changing inflection
4.1 Inflection and derivation revised
4.2 A formal description

5 Conclusion

6 References

1 Introduction

There is much discussion in morphological theory as to where exactly morphology belongs in the mental representation of grammar. Several grammar models have been developed, each aiming at describing the key concepts of our grammar and the position of morphology in particular. Traditionally, there seems to have been a general consensus that there exists pre-syntactic (lexical) and post-syntactic components, but recently this has become an issue of debate. A key issue in this discussion is the process of word formation. While some linguists argue that word formation takes place in a separate morphological component, some say syntactic rules also play a part and some argue that words actually are formed in the syntax.

Numerous linguists have contributed to this discussion, many proposing new models of morphology and word formation. In this paper, two alternate theories that attempt at describing the position of morphology in the grammar will be outlined. Chapter 2 describes Halle and Marantz’ (1993) model of Distributed Morphology, which presupposes that all word formation takes place in a syntactic module and that there is no such thing as a lexical process. Chapters 3 and 4 give an outline of an alternate view to Distributed Morphology. Chapter 3 describes Booij’s (1993) approach at proving that there are two different types of inflection, and that contrary to former theories, inflection can feed word formation. In chapter 4, Haspelmath (1995) provides much the same view as Booij by showing that inflection also can contribute to changing a word’s part of speech category.

1.1 Inflection and derivation defined

The notion of inflection and derivation, and the distinction between these concepts, is a debated issue in recent morphological theory. In order to understand the discussions outlined in this paper, a definition of the traditionally accepted meanings of the terms inflection and derivation is necessary.

The Concise Oxford Dictionary of Linguistics defines inflection as “any form or change of form which distinguishes different grammatical forms of the same lexical unit”^[1]. Inflection is traditionally assumed not to change a word’s part-of-speech category, but alters its grammatical function. Inflection is required by the particular syntactic context that a word appears in. Inflection provides a complete paradigm – for the vast majority of words, all the forms of the paradigm exist and are used. The category of tense for verbs is an example of inflection – the words walks and walked have different grammatical functions, but belong to the same POS category.

Derivation is a process that creates a new word. As such, in the process of derivation, a new word is produced by adding a morph to the word’s base form. Derivation most often changes a word’s part-of-speech category. In addition, derivation is incomplete in the sense that a particular derivational morph cannot be applied to all words of a class. An example of derivation is the creation of the word countess from count. Clearly this affix can not be added to a substantial number of other words to create more derivated words in the same fashion.

Traditionally, there has been a polarisation between derivational and inflectional attempts at describing morphology. As such, morphemes are seen as being either derivational or inflectional. As already mentioned above, derivational morphemes are those that modify part of speech. They usually do not interact with syntactic rules and have an idiosyncratic meaning. Inflectional morphemes, on the other hand, do not change part of speech, do not have an idiosyncratic meaning, but can interact with syntactic rules. This way of classifying morphemes also entails that the morphemes exhibit properties that ensure their assignment to one of the classes and that the grammar is organized in a manner that draws advantage of the different types of morphemes. Also following this way of classifying morphemes, the process of word formation is one of two types – either derivational and located in the lexicon, or inflectional and located in the syntax.

2 Distributed Morphology

Halle and Marantz offer an alternative approach to morphology. Their theory Distributed Morphology is a radically novel theory of the mental grammar’s architecture. The theory suggests a grammar model where the syntax is separated from the phonological realization of linguistic elements. The phonological realization of an element is governed by lexical entries that relate morphosyntactic features to phonological features. As such, there is not one specific morphological grammar component. While traditional morphology presupposes that the building-blocks of a word are invariant, Distributed Morphology allows them to change during the process of word formation. In the following chapter, the basics of Distributed Morphology will be outlined.

2.1 How is the grammar organized in Distributed Morphology?

Distributed Morphology divides the organization of the grammar into five different levels: the syntactic levels Logical Form (LF), D-Structure (DS) and S-Structure (SS), the morphological level Morphological Structure (MS) and the phonological level Phonological Form (PF). At each level there are hierarchical groupings of terminal elements, represented as tree structures. The terminal elements of the tree structures are groupings of grammatical features, roughly corresponding to pieces of words. In the syntactic levels the terminal nodes do not have a phonological representation. This is first obtained at the level of MS, through the process of Vocabulary insertion. The MS level functions as an interface between syntax and phonology, matching sounds and grammatical features. The mapping of phonological feature complexes onto morphosyntactic feature bundles takes place after terminal nodes have been combined to make words in the syntax. As a result of this, the structure of words must be assumed to be determined by the syntax.

2.1.1 A different kind of morpheme

Within Distributed Morphology, the term morpheme has a quite different meaning than the well-established one of being a configuration of phonological units with a grammatical or lexical meaning. The terminal elements in the grammar hierarchy are termed morphemes in all the levels, also in the syntactic levels where the morphemes still do not have a phonological representation. As such, in Distributed Morphology, a morpheme is a syntactic terminal node and does not have anything to do with the phonological expression of it. All terminal nodes are called morphemes, regardless of their phonological content.

The actual phonological representation of a particular morpheme is found in the Vocabulary. The Vocabulary consists of a set of Vocabulary Items and bears some resemblance to the traditional notion of the lexicon. The Vocabulary items hold information about their phonological realization as well as in which linguistic environments they can exist.

2.2 Different organization in syntax and morphology

Halle and Marantz point at mismatches between “the organization of the morphosyntactic pieces and the organization of the phonological pieces”^[2] as a motivation for organizing the grammar in the levels described above. There is not one uniform organizational structure of the terminal elements, or morphemes, at the different levels in the grammar - the terminal elements are organized differently in the syntactic levels than in MS and PF. In the syntactic levels LF, SS and DS the morphemes are organized hierarchically, but there is no left-to-right order among them. In the phonological level PF, however, sentences display a linear order of morphemes, allowing words to be spelled out in a way that humans understand. As such, this linear order must be created in connection with the interface level MS. The grouping of morphemes within words follows different principles form the grouping of words into phrases and sentences. Another factor that reinforces the differences between PF and the syntactic levels is that when additional morphemes are required to ensure well-formedness of an expression, they can be inserted in MS. This means that the number of morphemes that need a phonological realization may well change between the syntactic level SS and the phonological level PF where the terminal nodes get their actual sound segments assigned. Subject-verb agreement is an example of a grammatical feature that for many languages is implemented by adding an extra morpheme in MS.

Other grammatical processes that contribute to the differences between terminal elements in the syntactic levels and MS are head-to-head movement of morphemes in the hierarchy, merging of nodes, fusion of sister nodes and fission of single nodes into two separate nodes. These processes all take place on the morphological level of MS.

2.2.1 Merger

Merger is the process that joins two terminal nodes by giving them a head category node. Two independent terminal nodes are still kept under this category mother node. When phonological sequence items are assigned to the terminal nodes in the Vocabulary insertion process, one individual Vocabulary item is assigned for each merged terminal node under the category node. As a result, a new word is formed from the heads of two independent phrases and each head is kept as a separate morpheme in the new word. An example of merger is the combining of the main verb and tense in English.

2.2.2 Fusion

Fusion is a slightly more radical process than merger. In the process of fusion, two terminal sister nodes are fused into a single terminal node. Since this fused node now displays a combination of the features from the original sister nodes, the one Vocabulary item that can be inserted must have the features of both the original terminal nodes. Since fusion actually combines two terminal nodes into one, only one Vocabulary item can be inserted, producing only one resulting morpheme. In many Indo-European languages, a single affix representing number and case is the result of two independent nodes (morphemes) joined in fusion.

2.2.3 Fission

In the event of fission, one individual morpheme, or terminal node, corresponds to more than one Vocabulary item. Instead of allowing the Vocabulary insertion process to stop after assigning one item to the node, fission ensures that the process continues until all Vocabulary items that can be inserted have been. This feature is probably most notable in agglutinating languages, such as Georgian. In Georgian, through the existence of fused clitic terminal nodes, a clitic morpheme can incorporate features such as subject, object and indirect object arguments.

2.3 How are sounds mapped onto morphemes?

The fact that the terminal nodes undergo change at MS is not the only difference between the syntactic levels and MS. Another important distinction is the systematic difference between the types of features the terminal nodes in the syntax and morphology levels display. In the syntactic levels LF, DS and SS the terminal nodes have no phonological features, but only morphosyntactic and semantic features. The morphosyntactic features are extracted from the Universal Grammar, which is a language independent module. Language-specific features are not necessary or even available at LF, DS and SS. The only syntactic level that uses some language-particular information is DS. But even there the information is on the category and concept level and the Vocabulary of the language is not available. As such, the semantic and syntactic features that are chosen for a particular terminal node in the syntactic levels are picked without regarding if that set of features actually exists in the Vocabulary. The only restriction that applies is that the feature complexes must satisfy general and language-specific constraints on the combination of features. If a morpheme consists of a feature sequence that does not exist in the Vocabulary, a sound sequence will still be assigned. Insertion only requires that the feature sequence in the Vocabulary is non-distinct from the feature sequence in the terminal node, and the Vocabulary item that matches the most features of the node will be inserted. This basically means that a Vocabulary item can be underspecified for the morphosyntactic feature bundle that it realizes. The only constraint is that it cannot contain features that conflict with the features of the terminal node.

The Vocabulary of a given language consists of entries with two distinct sets of features – phonological and morphosyntactic/semantic features. Through the process of Vocabulary insertion, the phonological features are matched with the morphemes at MS. Distributed Morphology does not operate with a lexicon in the conventional sense. The processes that other theories place in the lexicon are spread between various other components in Distributed Morphology.

In the process of Vocabulary insertion, the vocabulary items in question compete to realize the terminal elements in MS. Generally speaking, the Vocabulary item that most closely matches the feature specification in the terminal node is picked, but in the cases where there is more than one potential match, two types of conditions apply: context free and context dependent insertion. In context free insertion, the entry whose category is compatible with the category of the terminal element and whose feature set is compatible with the feature set of the terminal element, is chosen. In context dependent insertion, or conditioned allomorphy, the alternative Vocabulary items in question only differ with regards to possible insertion contexts and phonological features. Generally speaking, the allomorph with the most specified context will be preferred for insertion. This means that the most specified allomorph that still fits according to the feature specification in the terminal node will be chosen.

[...]

^[1] The Concise Oxford Dictionary of Linguistics (1997)

^[2] Halle & Marantz (1993), p 115