323-Morphology

Download Report

Transcript 323-Morphology

323 Morphology
The Structure of Words
4. Inflection and derivation
(This page last updated 25 OC 06)
4.1 Inflectional Categories
Inflectional Categories include such features as tense, person, number, perfective,
progressive, relevance, aspect, voice in verbs, and number, gender, person, case in nouns, and
positive, comparative, and superlative in adjectives and adverbs. The task it to determine whether an
affix or a morphological operation is grammatical (inflectional) or derivational (a lexical property). Two
approaches to the problem are the dichotomy approach, which divides morphemes into distinct
classes (usually 2), and the continuum approach, which states that morphemes range from clearly
inflectional to clearly derivational (or the reverse).
An inflectional Dimension refers to a specific set: a set of inflectional categories. The members
of this must be mutually exclusive. To be mutually exclusive, two such members cannot occur in the
same paradigm. A verb, for example, cannot be both present and past at the same time. The example
that H gives is [+Tense], the more-or-less standard name for the set that includes [+Past] and [-Past].
Another such set is [+Number] which includes [+Plural] and [-Plural]. Some languages have an
intermediate set: [+Dual]. [+Dual] refers to quantities of two: two books, two men, two cars, and so
forth. A noun cannot be singular and plural at the same time. The holds for dual. Besides number,
there is another dimension that nouns are marked for: Case. Case is marked only in English pronouns:
he, him; she, her; I, me; they, them; we, us; who, whom. Other languages have more cases. Adjectives
and adverbs (A) are marked for degree: new (positive), newer (comparative), newest (superlative).
The Bantu languages tend to be marked with class markers. The classes divide semantic properties
into a few grammatical categories which obligatory.
A paradigm is most often listed in a table (or grid). The first dimension is person, the second
dimension is number, and the third dimension is tense:
1
4.1 Inflectional Categories
A one dimensional paradigm can be illustrated in the following way. Number is the only dimension:
[-Pl]
raft-ø
[+Pl]
raft-s
Number
The first two dimensions of the English verb are person and and number:
Number
[ -Pl]
[ +pl]
[ +1 st ] play-ø
play-ø
[ -1 st ]
p lay-ø
play-ø
p lay-s
play-ø
[ +Perso nal]
Person
[ -Personal]
Person has a second layer: [+Personal] which contains [+1st] and [--1st]. [+1st] refers to the first
person and [-1st] refers to the second person. All these feature are binary. Binarism is a theoretical
concept in which everything is paired off — either as plus or minus. The concept of ‘plus’ and ‘minus’
falls within marking theory. In marking theory, the marked pair (plus) is the non-default, and unmarked
(minus) pair is the default. The marked (+) feature is
2
4.1 Inflectional Categories
the more specific feature of the binary pair.
In the verb system, there exists a third dimension in inflection. Many languages show a tense
dimension including English:
Numb er
[ -Pl]
[ +pl]
[ +1 st ] play-ø
play-ø
[ -1 st ]
play-ø
play-ø
play-s
play-ø
[ +Personal]
Person
[ -Perso nal]
[ -Past ]
[ +Past ]
play-ed
The third dimension is not easy to show on a two dimensional plane. The angled lines are the best I
can do for the moment, but it is not ideal. Number and person play no role in the past tense inflectional
paradigm. Number does occur in the past tense, but it occurs only in the suppletive pair was and were.
The grammatical distinction of number is covert except for these two verbs.
In many languages, tense is not the distinctive feature, but perfect. The feature [+Perfect] refers to the
completion of an action. Sometimes the past tense of tensed languages can be
3
4.1 Inflectional Categories
translated into the perfect of a perfect language. This is true of the activity of the verb has been
completed. Actually, whether an action has been completed or not is called aspect, a fourth level of
dimensions. However, aspect is periphrastic in English and does not enter the inflectional paradigm. In
Russian the situation is worse. Aspect is marked by suffixes and prefixes and becomes tangled up
with derivational prefixes, which are often the same form: pisat’ ‘to write, imperfective’, na-pisat’ ‘to
write, perfective’; pod-pisat’ ‘to sign, perfective’, pod-pis-yvat’ ‘to sign, imperfective’.
Another verbal dimension is voice. Usually, it is periphrastic but in some languages such as Swedish,
it is marked by inflection (or the ending could be a clitic). Some languages mark relevantness, where
one action is relevant to another action or state. English is one of them:
E.g. Earlier John saw the movie; *earlier John has seen the movie; John has seen the movie.
Nouns usually show three dimensions: number, person, and Case. English nouns are covertly marked
for person and overtly marked for number:
E.g. book, books; ox, ox-en; goose, geese.
Person becomes????? over in agreement, which is discussed below. Case is not marked in English
nouns, though it is in some pronouns:
E.g. John saw Mary, Mary saw John; He saw her, she saw him.
He and she occur in nominative Case, while him and her occur in the accusative Case. Other such
pronoun pairs include:
E.g. I, me; we, us; they, them; who, whom.
You does not overtly mark Case nor does it. Pronouns in the possessive form may be said to be
overtly marked for the possessive Case. There can be two forms for possessive pronouns: one if the
pronoun precedes the possessed item and one that contains a final ‘-n’ or ‘s- if it does not precede the
possessed form:
E.g. My book, that book is mine; your book, that book is yours, our cat, this cat is ours; her car,
that car is hers.
.
4
4.1 Inflectional Categories
Some pronouns do not make this distinction:
E.g. His house, that house is his; its tail, ?*this tail is its.
The latter construction is dubious, but if it passes, the form is the same.
In many languages, if the verb is marked for person and number, it must agree with the subject of the
clause. If the subject is third person plural, the verbal ending must be the same. Overt agreement in
English occurs only in the third person singular:
E.g. he helps, *he help. The cats meow. *the cats meows.
In some languages, the verb may also agree with the direct object, occasionally with the indirect object
(dative object). H notes that verbs agree with their direct object as well as their subjects, for example in
Yimas:
E.g. Krayŋ
narmaŋ
k-n-tay.
frog
woman
object-subject-saw
Note: another interesting fact here is that the object precedes the subject, a very uncommon word
order. The same kind of agreement occurs in Georgian. If the indirect object is in the subject position,
the verb agrees with it. The same does not occur in Russian, for example. English is restricted to
subject-verb agreement.
Many languages show adjective-noun agreement as in Russian, although English does not:
nov-aja knig-a ‘new book’, nov-yje knig-i ‘new books’.
Demonstratives must agree with the noun for number they are modifying in English:
E.g. this book, these books; that story, those stories.
Many Indo-European languages Bantu languages have this kind of agreement.
Other types of agreement are possessor-possessed agreement in Arabic, postposition-complement
agreement in Classical Nahuatl.
Agreement occurs within the dimensional categories of languages, although the dimension may not be
overt in one of the inflected forms. English is a prime example of this. For example, Russian nouns do
not overtly mark gender, although gender is often associate with various declension class. The word
5
sudja ‘judge’ is masculine though the class marker ‘a’ in the nominative singular form generally is
associated with the feminine gender:
4.1 Inflectional Categories
E.g. nov-yj (masc. singular) sudj-a ‘judge’, *nov-oe (neuter singular) sudj-a.
The lexeme for ‘judge’ must include the feature masculine ([+Masc]) so that the adjective can properly
agree with it. Agreement is matching like features. I use linking theory to account for this, a topic I
hope to introduce later.
Dependent Verb Forms are forms that are restricted to embedded clauses. Relative clauses are
restricted to embedded clauses:
E.g. There is her friend, who is tall. *There is her friend and who is tall.
The complementizer (mood marker) that cannot occur in a matrix clause
E.g. I know that Saigon fell. *That Saigon fell. Saigon fell.
The same holds for gerunds and infinitives:
E.g. I ran out of water washing the car. *I washing the car. I was washing the car.
These examples in English do not involve inflection, though in some languages they do.
4.2 Derivational Meanings
Whereas inflectional categories are obligatory in the grammar, derivation categories are not.
Agreement, tense, number, gender are obligatory in many languages and these forms cannot be
omitted or altered. One may say a drinker, but it is not obligatory. One may say one who drinks. Many
derived forms are preferable, the non-derived form exist even though it may sound stilted or some
such thing: Kindness is preferable. That of being kind is preferred. There are lexical items (words) that
can’t be formed by a transposition of related words as above. A different lexical item would have to be
used: The king beheaded the knight, The king chopped off the knight’s head.
4.2.1 Derived Nouns
Apparently, nominal derivation is more common than verbal derivation. Certain types of affixes
are productive and follow a pattern. For example, the suffix ‘-er’ referring to an agent is common and
follows the pattern V -> N (agent, one who):
E.g. bake+agent -> bak-er; grop+er -> grop-er; taste+agent -> tast-er.
6
4.2 Derivational Nouns, Verbs and Modifiers
A similar pattern is followed by verb+instrument (that which one uses to V)
E.g. slice+instrument -> slic-er; peel+instrument -> peel-er; poke+instrument -> pok-er.
Another pattern is V -> N: verb+theme (ee):
E.g. divorc+ee, invit+ee. govern-ee, employ+ee.
Other derivations:
V -> N: verb+result (ion):
E.g. decide+ion -> decis+ion, revise+ion -> revis-ion, destroy+ion -> destruct-ion.
V -> N: verb+state-of (ment)
E.g. entrap+state -> entrap-ment, enhance+state -> enhance-ment.
A -> V: verb+become -> redd-en, black-en, white-en (subject a theme: his face reddened)
A -> V: verb+causative -> redd-en, black-en, white-en (subject an agent: they reddened his face.
N -> A: noun+pertaining to -> bush+pertaining -> bush-y, juice_pertain -> juic-y.
Adj -> Adv: adjective+in a ___ manner:
E.g. slow+manner -> slow-ly, bright+manner -> bright-ly, rapid+manner -> rapid-ly.
4.3 Properties of Inflection and Derivation
Most of the properties distinguish them are listed:
7
4.3 Properties of Inflection and Derivation
Inflect ion
( grammat ical)
Derivat ion
(lexical)
ob lig at o ry
opt ional
relevant t o synt ax
not relevant t o synt ax
not replaceable by a single word
replaceab le b y a single w ord
same concept ual meaning as base
new concept ual meaning
relat ively abst ract meaning
relat ively concret e meaning
regular semant ic meaning
of t en unpredict ab le meaning
not really relevant t o base meaning relevant t o base meaning
unlimit ed (almost ) applicabilit y
limit ed app licabilit y
usually o ccurs at p eriphery of w ord usually o ccurs close t o base
less base allomorphy
more base allomorphy
may b e cumulat ive
non-cumulat ive ( almost always)
not it erable
it erab le
The first four criteria are the most reliable ones. The first is very reliable, since one cannot choose to
omit an inflectional affix at will. That is, one must say John sleeps, not *John sleep. As I discussed
earlier, it may be always possible to find some way to get around a derived form even if the resulting
construction is very awkward and clumsy and nearly unacceptable though grammatical.
Cumulative refers to a single morpheme that contains more than one feature:
E.g. The verbal inflectional suffix in English ‘-s’ contains [-Pl] and [-Personal]. It is rare to find a
derivational suffix that clearly contains two or more readily identifiable features.
8
4.4 Conceptualizations in Morphological Theory
4.4.1 The Dichotomy Approach
In the dichotomy approach, inflection and derivation are considered separate entities; each are an
independent set. This is sometimes called split morphology, Word building rules are assumed to occur
in the lexicon before syntax. Inflection is assumed to occur in the syntax. That is, derivation is presyntactic and inflection is post-syntactic.
As noted before, an inflectional suffix almost always occurs peripherally in the word, whereas
derivation is non-peripheral. If inflection is post-syntactic, then this claim readily explains why inflectional
affixes are peripheral. Let us see how this would work. In L322 I worked out a theory involving linking,
which I briefly introduce here.
Chomsky originated the idea back in the 80s starting with chains. If a category moves, it leaves a
trace. A link is established between the moved category and the trace position. If the category is moved
a second time, another link is built. Two connected links form a chain. A chain in theory can be
indefinitely long, though they rarely contain more then four links and that is uncommon.
Next I added to the theory that if a lexical item shows agreement, the agreement slot on the noun
is empty:
E.g. [N CAR+[Agr __]] where the underscores represent the agreement position. Next I posited a
quantifier position (Q). Q is a set that contains quantifiers
Q = {one, two, three, four, five, some, several, few, a few, more, many, … }.
This set is not lexical. I consider them operators, a term taken from logic. If ONE locally c-commands a
noun, then a link must be established between it and its complement, which is true for all Lexemes and
Operators that require a complement. This is shown below for one hand:
9
4.4 Conceptualizations in Morphological Theory
NP
QP
N
HAND
[ _ _ Pl]
Q
ONE
[ -Pl]
In the above diagram AP is omitted for the sake of clarity and other nodes dominating NP are also
omitted. The Linking Principle states that only like features may be linked; that the two like features
must share the same feature sign: [+X] <--> [+X]. Unacceptable is [-X] <--> [+X]. One kind of speech
error occurs when features don’t match. Speakers tend to correct themselves so that the features
agree. If a feature has no positive or negative sign, then a sign must be inserted. If an unmatching sign
is inserted, the construction fails because it violates the Linking Principle. Thus a minus sign must be
inserted before ‘Pl’ in the above figure:
NP
QP
N
HAND
[ -Pl]
Q
ONE
[ -Pl]
The link is shown here in red. The colour is for illustrative purpose; it has no theoretic significance.10
4.4 Conceptualizations in Morphological Theory
If the features do not match, we obtain the following ungrammatical phrase:
E.g. *One stones.
Note that one is inherently singular: [-Pl]. The remaining numerals are [+Pl] including zero:
E.g. I have zero books. *I have zero book.
Such sentences with zero are very unusual. Mass nouns are singular: rain, mush, grass, oil, meat.
If a verb agrees with anything, it will agree with the subject. There is one big problem in Chomsky’s
theory. There is no explanation why verbs should agree with the subject. For the moment, let us adopt
a the original idea of a sentence:
E.g. S -> NP + VP, VP -> V + NP
Syntax has advanced well beyond the above, but the version I am assuming is too involved to go into
here. Assume the following structure, where there is a link between he subject and the verb:
S
VP
NP
QP
Q
ONE
[ -Pl]
N
HAND
[ -Pl]
[ -Personal]
V
[ _ _ Pl]
[ _ _ Personal]
The feature [-Personal] is innate in all nouns. Only the first and second pronouns may be
[+Personal]. It is a property of the verb that is must be marked for plural and personal; they are not
innate:
11
4.4 Conceptualizations in Morphological Theory
S
VP
NP
QP
Q
ONE
[ -Pl]
N
HAND
[ -Pl]
[ -Personal]
V
[ -Pl]
[ -Personal]
Note that in addition to Q and N being linked, N and V are also linked. Hence plural agreement is
required in both links, which form a chain. This is as far as I will go here. It introduces how morphology
interacts with syntax. Remember the third goal: Explanatory Goal. The theories of binary features,
unacceptable empty features (neither plus nor minus), linking and feature matching are the
explanatory theories that have the power here. Together they form a set of constraints that limit the
power of grammars. To account for theory of linking that apparently limits links to be built between
heads and complements will force a major change in Chomsky’s theory of phrase structure rules.
However, there are many syntacticians who revere Chomsky as a god. And being a god, everything he
says must be necessarily correct
In the continuum approach, H lists in his table 4.6 a continuum from simple inflection to
derivation. The continuum table is rather interesting. The factor cumulative in features is rather
logical, once we see that some features can be represented in a single morpheme. The English suffix
‘-s’ = [+Pl] is not cumulative since only one feature is marked. The same holds for Eng. Past tense:
12
4.4 Conceptualizations in Morphological Theory
[+Past] refers to only feature: tense. The non-progressive participle suffix [-ed] is cumulative in that it
may refer to the passive voice:
E.g. Max was watched by his boss.
Or it may refer to the perfect relevance:
E.g. Max’s boss has watched him.
This may not be the best example, since I believe that ‘-ed’ should be listed as two morphemes each as
a different dimension. The only good example in English is the verbal ending ‘-s’. It contains two
features, which are cumulative: ‘-s’:
E.g. { [-Pl], [- Personal]}.
This is more or less natural.
H notes the derivational affixes can mark one or of a set of lexical features. There is no evidence
here that supports a continuum approach and fails to support the dichotomy approach.
4.4.3 Contextual, Inherent, and Derivation.
This approach makes good sense. Contextual and inherent refer to inflection. Contextual inflection
includes agreement. Verbs must agree with the subject nouns that they are in construction with, nouns
must agree with quantifiers and determiners which they are in construction with. Another contextual
inflection is case-marking in nouns, adjectives and determiners in the languages that require it.
In the theory of unspecified features that I outlined above, the feature must agree with the
category that locally c-commands it. The link is contextual. Inherent features are those which are part of
the lexical make-up of the category:
E.g. gender in nouns and the third person in nouns. All nouns are third person with a very small
class which may be 2nd person. Gender in English nouns is nearly always based on a semantic feature.
If a noun is [-Human], it is nearly always neuter in the sense that it may be replaced with the pronoun it.
If it is [+Human], The noun is either feminine or masculine in that they may be replaced the pronoun.
The choice of the pronoun he or she is depend on other factors than pure semantics. Pets and domestic
animals with names are usually referred to with he or she appropriate for the name. Other animals are
usually referred to with it unless the sex of the animal is known — this is an optional rule.
13
4.4 Conceptualizations in Morphological Theory
Tense, voice, aspect, relevance and mood are not exactly inherent in the verb, but these categories
always subsume a verb.
E.g. RUN is not inherently present or past, but we must mark them for tense. Tense is an
operator which must adjoined to some verb, whether it is a main verb or an auxiliary verb:
E.g.. Mary bough-t a car, Mary ha-s fallen ill, Mary is falling ill. If tense were inherent in the verb,
one would expect the feature to remain in the verb, but that does not happen. Tense is adjoined to the
first auxiliary verb dominated by VP. The same holds for aspect, voice, relevance, mood and polarity.
Polarity deals truth and negation. Note that these operators are not inflection, but that they are
grammatical as opposed to lexical.Verbs are never marked inherent for these operators. Infinitives,
gerunds, and participles are formal constructions and they are marked [-Tense]. They are not
operators.
14