Untitled Document

Morphology

John Goldsmith
Spring 2004

Text: Understanding Morphology, by Martin Haspelmath.

Chapter One

1. Morphology is the study of word structure. It interacts with syntax and phonology.

With syntax: can syntax be analyzed with no access to the internal morphological structure to the word? or better put, just what information regarding word structure is it optimal for syntax to have access to?

With phonology: to what extent are phonological alternations or generalizations influenced by morphological structure?

2. Morphology as the decomposition of words into morphemes. This is a big step: how is analysis of words into morphemes motivated? When is it desirable to have morphological analysis without dividing a word into morphemes?

3. More basically: what is a word? This will be one of the major think-questions of the course. What is a word, and how do we recognize one when we see it? Can the speech stream be decomposed into words, or is it possible that there is stuff out there - morphemes? - that do not form parts of words? Lexical items.

4. De Marcken algorithm for finding words in a language.

5. Variation across languages: analytic/synthetic distinction: isolating - analytic - synthetic - polysynthetic. Average number of morphemes/word: Vietnames: 1.06; German: 1.92; Old English: 2.12; Swahili: 2.55; Sanskrit: 2.59; Greenlandic Eskimo: 3.72.

6. Haspelmath's "goals of linguistic theory". 1. Elegant description; 2. Cognitively realistic description. 3. System-external explanation. 4. A restrictive architecture for description. His figure 1.1 on p. 8. Functionalist versus formalist modes of explanation.

7. Swahili example?

Chapter Two

1. Word forms, lexemes, and lemmas (lemma = H's word-family). A lexeme is an inflectional family, i.e. a paradigm, of word-forms (e.g., boy, boys; jump, jumps, jumped, jumping).

2. A paradigm: a set of related forms, a choice of which can be determined or accessed by syntax. A template of related forms; for a given (fixed) stem, the meaning of the stem is fixed, and the choice of the position in the paradigm specifies grammatical information and in most cases a grammaticized meaning as well. Generally exceptionless, in the sense that all position are filled (available) in a paradigm, with semantic regularity. Generally the items in a paradigm are all from the same major lexical category.

Related words not from the same paradigm are related derivationally. A family of derivationally related words is a lemma (word family).

3. Examples. Clear examples, less than clear examples.

4. Compounds are words composed of pieces from two distinct lexemes (also, two distinct lemmata).

5. Morphemes. Traditional definition: minimal meaningful unit of linear analysis. Automatic segmentation of text. Greenberg square.

6. When do morphemes have identifiable meanings? Clear cases when they do; clear cases when they don't. Intermediate cases: stems; affixes. Can an affix have a grammatical function and no identifiable semantics? German umlaut. Why are affixes short? related to their frequency.

7. Affixes, bases (or stems), and roots. When the entries of a paradigm share a contiguous amount of phonological material, we identify that as the stem of the paradigm, and the other material as the affix that identifies that part of the paradigm. Example. Inflectional affixes.

8. Derivational affixes defined in parallel fashion. When two lexemes are derivationally related, typically one is more morphologically complex than the other, i.e., formed from the other by the addition of an affix. When this is not the case: conversion; or ? A stem which is not related to a morphologically simpler lexeme is a root.

9. Suffixes, prefixes, infixes, circumfixes? (ge-X-t).

10. A stem can itself sometimes be morphologically analyzed: it may be derivationally related to another lexeme.

11. Example from text (20) where the affixes denote body parts, in Bella Coola (Salish).
Example from text (20): bound roots in English, which are difficult to identify as roots or affixes: bio- and -crat.

12. Formal operations that relate word-forms and lexemes: affixation; compounding. Templatic association (Afro-Asiatic, Penutian). Base modification (voicing of final fricatives in English to form verbs, or plurals). Illustrates historical morphologization of phonological effects. Palatlization (example in text). English strong verbs. Reduplication. Subtraction: French unmarked (masculine) adjective stems. Acronyms, clippings, alphabetisms. "The base of a morphologically complex word is the element to which a morphological operaiton applies."

13. Morphemes and allomorphs. Example: Turkish vowel harmony (2.23, p. 26). How much of this is taken care of by the phonology -- and is therefore not the business of the morphology? Automatic phonology; automatic morphophonology; non-automatic morphophonology; morphological suppletion.Suppletion in buy/bought, catch/caught, teach/taught, seek/sought, how many others? Italian stem choice for andare: vád-o 'I go'; vá-i 'you go', va 's/he goes' and-iámo 'we go', and-áte ýou pl. go', vá-nno 'they go'. Crucial point: allomorphs must have the same meaning, and their choice must be predictable from phono-morpho-syntactic information (unless they are in free variation, which is rare).

14. Section 2.6: some difficulties in morpheme analysis. So many difficulties! and right from the start.
1. How different can morphemes be and still be called allomorphs? How many examples of the pattern are necessary -- is one enough? good/bett: are these allomorphs of a single morpheme?

Chapter Three: The Lexicon and Rules

1. Productivity. New words: neologisms. Morphological processes that can create new words are called productive. A suffix such as English's -al suffix is not productive, but speakers are still aware of the relationship arrive/arrival. Haspelmath: "The reason why languages may have unproductive rules is that complex words, like simple words, may be listed in the lexicon. The lexicon is the linguists' term [sic] for the mental dictionary that language-users must be equipped with, in addition to the grammatical rules of their language." So morphological rules describe relationships between words ("in the lexicon"), and they may or may not be productive in addition. "Only when a rule is productive and is observed to be extended to new bases can we be sure not only that the rule exists in linguists' descriptions, but that it is also cognitively real." Haspelmath sees a gradation of productivity: -ness/-ize/mis-/-ee/-eer/-al/-th/-ter.
View 1 of lexicon: Never includes regular complex formations: inclusion of the principle or generalization is sufficient, and including its effects in addition is to be avoided. Associated with the morpheme lexicon view. This view is comfortable with marking diacritic features on morphemes to show which ones undergo non-regular processes (positive exceptions). E.g.: warm: can be combined with -th.
View 2 of lexicon: contains all of the (possible or observed) words of the language. "uneconomical, and thus inelegant: One important goal of linguistics is to provide an elegant description of language structure, and lists are inherently inelegant" (Haspelmath, p 43). Boy do I disagree with that statement! Lists are, to the contrary, inherently elegant. They hae a well-understood structure. H continues: "A general methodological principle of linguistics (as in any other science) is that as many facts as possible should be subsumed under general rules and principles rather than merely stated in the form of a list." H points out that stems of high frequency (have, make, say) are more likely to have irregular inflected form than stems of medium frequency.

2. The form of morphological rules
2.1 The morpheme-based model, in which morphemes are concatenated (like words): the syntax of words (title of book by E. Selkirk). If we decide to assign morphemes to category, then this can be expressed either as phrase-structure rules (e.g., Word -> Stem + Suffix_1), or the categories can express the combinatorial possibilities directly: e.g., ness is of a category "Adj_stem \ Noun_stem", which means if you left-concatenate an adjectival stem, the result is a noun-stem. This is sometimes written: [ Adj __ ] (as in H.).

2.2 Word-based model. Unfortunately, it's hard to specify just what this model says: what kinds of patterns between strings can it countenance? The simple examples are cases like [ X ] <--> [ Xz ], but what other patterns can be found? In the cases given in the book, the element ("X") found on one side is also found on the other; but this is not always assumed to be the case ("all kinds of non-concatenative processes can be described with it quite naturally"). The example given is conversion, which is as concatenative a process as any: you concatenate nothing. Another example: Hindi/Urdu long/short vowel [ maar 'causative' ] <--> [ mar 'activity' ]; schema [ XVVC ] <--> [ XVC ]

Back formation, a historical process: editor -> edit; baby-sitter -> baby-sit. I think Haspelmath's account of baby-sitter is wrong; he suggests that "baby-sitter" originally was formed by the familiar N1-N2 compound rule, as if from the two words baby and sitter. It is, rather, an example of a particular kind of compound, often called a kind of compound, often called a synthetic compound, involving a verb, preceded by an argument (often a theme of the verb [truckdriver], but not always: theater-goer ). See http://www.linguistlist.org/issues/6/6-758.htmlfor a discussion. What has happened, on the word-based and schema-based model, baby-sitter is reanalyzed along the lines of the [ X ] <--> [ Xer ] schema.

3. Morphological change
3.1 Pattern loss: Loss of elements in an inflectional paradigm (e.g., dual number; past subjunctive in French).

3.2 Coalescence of words and patterns: grammaticalization, grammaticization: when a syntactic pattern is reanalyzed as a morphological pattern. Example: future in French, Spanish: from an earlier encliticized form of the verb habere.

3.3 Analogical change: extension of patterns, and leveling (loss of allomorphy or morphophonology). Neither of these is properly thought of as analogy in most cases: the pattern is larger than any single instance of its form. Example: weak verbs become strong verbs (English: sticked > stuck; dived > dove ). Here the "strong" pattern wins over the concatenative ("weak") pattern.
Leveling: loss of allomorphy or morphophonology. Example: earlier je treuve/nous trouvons, modern je trouve/nous trouvons (although French retains je peux/nous pouvons, je veux/nous voulons).

3.4 Reanalysis.

3.5 Morphophonology: formerly triggered by the phonological content of certain morphemes, later triggered by the "features" -- the context -- instead. Example: umlaut in German.

Chapter Four: Inflection and Derivation

1. Rules of derivational morphology serve to "create" a type of word that exists independently of that rule, while rules of inflectional morphology are the only way to create the word-forms that they create (e.g., 3rd person singular forms of verbs). (Problem: inherent plurals, like police).

2. Inflectional morphology must be understood as specifiying a position in a multi-dimensional chart (table, grid). Each dimension is a feature; each position along that dimension is a value (what Haspelmath calls an inflectional category) -- the terms feature and value are more general than their use here in morphology. Values are (by definition) mutually exclusive. From a formal point of view, we can think of features as mappings (functions) from words to labels.

Example of a grid: Spanish indicative verbal inflection: 3 dimensions: person (1,2,3), number (sg, pl), and tense (present, future, imperfective). How do we argue that each of these is an independent dimension, rather than these forms being cells in a 1-dimensional paradigm of size 18?

Notation: feature/value matrices.

Common inflectional categories: Nouns: number, case. Verbs: tense, aspect (perfective, imperfective, habitual, ...), mood (indicative, subjunctive, imperative...). Nouns, Verbs, Adjectives and adpositions: agreement in number, case, person, gender.

3. Asymmetries: of two sorts. The general notion of asymmetry refers to the following situation: When feature F has value f1, then feature G has two or more distinct values g1 and g2; but when F has a different value, f2, then G has only one feature value g1 (or more generally, G has feature values when F has value f2 than when it has value f1). Example: 1. German adjectives distinguish between three genders in the singular, but not in the plural. 2. Spanish has an indicative/subjunctive contrast in the present and past, but not in the future.

These two examples illustrate two different kinds of asymmetry: in the first, a contrast (or distinction) is lost or ignored; in the second, a value in the paradigm (chart, table) is missing and morphological form steps in to fill it.

4. Agreement markers: agreeing in person, number, gender, and case. Important to distinguish two cases: what Haspelmath calls NP agreement and noun agreement: NP agreement is agreement among arguments of a proposition (typically subject NP and object NP of a verb), while noun agreement occurs inside of an NP, typically agreement of determiners, adjectives and participles with the head noun (e.g., Swahili wa-le wa-tu wa-refu "those tall people" [Haspelmath wrongly translates this as having a "two" in the gloss: the only 2 is the class 2 marking, which is the Bantuist's way of saying human-plural]; Italian quest-a nuov-a casa "this new house").

Haspelmath points out that the dimensions are defined by what morphology realizes anywhere; the marking on the agree-ing form may be morphologically distinct, but the information in the agreement-head may not be "morphologically marker": French: un-e [fem] femme [fem.] intelligent-e [fem.]: femme is not morphologically analyzable into two pieces, one of which marks the feminine gender. Hence we refer generally to morphosyntactic features: expressed by the morphology, transmitted by the syntax.

5. Other inflectional categories: 1. adjectives: comparative and superlative. 2. verbs: active/passive voice. 3. verbs: polarity (affirmative/negative). 4. verbs: independent/dependent clause type.

6. Derivational morphology. Meanings more diverse: e.g., -in: sit-in, be-in, love-in; -nik: peacenik, beatnik, nudnik; French -ier: forms names of trees.

7. V -> Noun. Agent, patient, instrument, action.

Week	Reading for this week	Assignment after this week
1	Chapters 1 and 2	Write a one to two page (single-spaced) paper in which you attempt to explain what a word is, using characteristics that are language-independent (that is, without appealing to specific facts about a language except in order to illustrate your point). This is an extremely difficult task, and the point is for you to come to appreciate how difficult it is. You will want to think about phonology and syntax. Can you give ideas that will be helpful in deciding what a word is in a language for which that decision has not yet been made? What do your criteria tell you about the number of words in the English sentence, "Isn't it cold?" You may assume that your reader is familiar with the reading for this week.
2	Chapter 3	Assume for purposes of discussion that you believe that there are words, and that each language has a grammar that determines their possibilities of combination; you recognize that there are difficulties in defining words, but you believe that these difficulties will someday be overcome. Now take a stand on the question as to whether there are morphemes, and whether words are properly understood as being combinations of morphemes. Acknowledge that there are difficult cases for any view on this point; rather than trying to establish your point with one knock-down example, be nuanced and sensitive to linguistic fact. The alternative to the there-are-morphemes view is most likely the word-based model, though you are free to design your own alternative if you choose. Whether you believe there are morphemes or not, you must define what a morpheme is or how it is identified, and you should give clear cases that suit that definition or procedure. One to two page paper, like last week. This is very hard, too, and the assignment is designed help you understand what the issues are that are at stake in this week's reading.
3	Chapter 4	Select a language other than English, and describe three aspects of the morphology in this language. Explain the type of morphology (inflectional or derivational, or perhaps you prefer a more refined classification) of each type. If you encounter problems in classifying them as inflectional or derivational, explain the problems.
4	Chapter 5 On productivity: Probabilistic Phonotactics and Morphological Productivity by Hay and Baayen	Exercises (pp 83-84) 2, 4, 5. Submit name of language for project.
5	Chapter 6, plus up to p. 130	Exercises (pp 113-114) 1, 2.
6	Chapters 7 and 8	Exercises (pp. 146) 3, 5.
7	Chapter 9	First draft of first section of project due.
8	Chapter 10	Exercise 3, p. 207.
9	Chapter 11	First draft of project due
10	Chapter 12	Project due.

Course project: I would like each of you to construct a morphology problem on a language of your choice. It should be self-contained, in the sense that it could be given to a linguistics student to solve with just the data that you present in it. It should be organized into sections. In the first section, you present data that allows the student to decompose words into component morphemes. It should have at last three sections. Past the first section, the sections can deal with any topic that we deal with in the course (or any that is relevant to morphology): for example, the difference between derivational and inflectional morphology; the feature structure of the morphemes; the constituent structure of words; morphophonology; and so on. I would like a brief statement by email from you by Week 4 regarding what language you will use, and some initial thoughts on what is interesting about it. By week 7, I would like a first draft including at least the first section, and by week 9 I would like a first draft of the whole project. The final version will be due on the last day of class.