Transcript 323-Morphology

323 Morphology
The Structure of Words
3. Lexicon and Rules
(This page last updated 26 SE 06)
•3.1 Productivity and the Lexicon
•
The lexicon is in theory infinite, but in practice it is limited. Human beings know only a certain
amount of information at any one time and it is impossible for a human to know an infinite amount of
information. This holds in the lexicon, as well. Comparing a lexicon to a dictionary (the printed
lexemes), a dictionary can hold only so much information at one time. The list can grow and grow, but
it is never infinite.
•
The potentiality for making up new words by means of the rules of word building is potentially
infinite, but this has never been proved. Nevertheless, it possible to create a large number of words,
larger than what most humans could possibly memorize. Thus we must distinguish between actual
words and potential words.
•
A neologism is a new word that has been created. Neologisms that do not catch on except
occasionally are called occasionalisms. Note that this word was probably created recently and I
doubt if it has really caught on. If true, then the word occasionalism is itself an occasionalism.
•Affixes that are readily adjoined to words to create new words (bases and stems) are called
productive.
•
E.g. The English suffix ‘-er’ can be added to most verbs that denote an agent oriented action:
doer, fixer, baker, worker, runner, swimmer, writer, and so forth. The same suffix can also denote an
instrument ‘cooker, pickle slicer, popcorn maker, double-boiler, but it is doubtful that this verb
productive, though it may be productive if the semantic class is known. Other affixes are clearly not
productive:
•
E.g. ‘-ic’, ‘ion’, ‘-ive’, ‘be-’, ‘de-’, ‘a-’, re-’ and so forth.
•Another problem with unproductivity (sic) is that unproductive affixes easily change the meaning of
the word.
3.1 Productivity and the Lexicon
E.g. head, be-head; give, forgive; stand, understand; woman, womanize; and so
forth.
There are affixes that are very productive, rather unproductive, somewhat unproductive, very
unproductive. H lists a finer list of productiveness (p. 42). Another problem are complex words
that are lexical, but underlying base is not lexical. To illustrate this, consider disgruntled. It is
derived from the base *‘gruntle’, which is not a lexeme with the associated meaning of
disgruntled. I take the view that forming bases is productive given the restrictions on the base,
but the base is not always a lexeme. There no way to be absolutely sure whether a given base
will or will not be a lexeme. As a consequence, all lexemes must be enterred in the lexicon. If a
base is created, one must check to see if it is a lexeme, or one may occasionally determine a
lexical meaning for the new base thus creating a new word, as I did with unproductivity above.
H argues that a word-form lexicon is more desirable. A word-form lexicon is one in which
every declined or conjugated form of each word is listed in it. Inflected forms are generally
predictable given the class forms of each lexeme, except the irregular ones such oxen, children,
brethren; is, are, be, was, were (being and been) are regular (except for the pronunciation of
been in the US and in Canada whether the American pronunciation has taken over the earlier
one which is still standard in Britain.
Even so, there is evidence that all the word forms of everyday usage are memorized and
listed in the lexicon. (I read a paper at SFU claiming that the lexicon is divided into two parts: the
list of lexemes and the list of word-forms derived from them. Each set of word forms derived from
a lexeme are linked to that lexeme at little cost to the grammar.) Linking is another research topic
of mine, which I cannot get into here.
3.1 Productivity and the Lexicon
H mentions that a lexicon should be elegant which means the least number of rules that will produce
all the inflected forms. The lexical part of the lexicon contains a list of all lexemes that a speaker has.
The word-form part of the lexicon contains the inflected forms for each inflectable lexeme
(conjunctions, prepositions and other functions are not inflected in English):
play
plays
PLAY
played
playing
The lexeme PLAY is connected to the word-forms play, plays, played, and playing by means of a link.
The links are for information transference from the lexeme to the word-form, which we might call
formation, and from the word-form to the lexeme; the latter is called interpretation. The most
common word-forms are most likely memorized. The word-form component will differ for each speaker
just each speaker probably knows a different set of lexemes, everybody’s experiences are unique to
that individual. The hypothesis is that speakers normally draw from the set of word forms in forming a
sentence. To form an unusual word, he must form the word-form from the lexeme using the rules of his
grammar. The above diagram is incomplete, but it will suffice for now.
3.2 The form of Morphological Rules
A morpheme rule is any kind of regularity that is ‘noticed’ by speakers and is reflected in their
unconscious linguistic knowledge (H p. 44). Though there may be several formal descriptions that can
be conjectured, H will discuss two formalisms: the morpheme based model and the word-based
model.
3.2.1 The morpheme based model
In this model morphemes are combined together to form a new form, expressed by a set if wordbuilding rules. H compares these to syntactic rules forming phrases, clauses and sentences. Consider
the following words as examples:
E.g. fox -> fox-es, school + house -> school+house, build -> re-build, con-trast ->
con-trast-ive-ness, sad -> sadd-est.
Word-structure (word-formation) rules:
word-form <--> stem (+ inflectional suffix)
stem <--> base + lexical meaning (bad format here)
base <--> {{(deriv. prefix +) {root, base} + (deriv. suffix)} , {stem + stem}}
inflectional suffix = -es, -est
derivational prefix = rederivation suffix = -ive, -ness
root = fox, school, house, build, happy, sad, down, never, do, be, and so forth.
Phrase-structure rules (top down and bottom up):
S <--> NP + VP
VP --> V + NP
NP <--> Det + Adj + N or better NP <--> Det + [Adj + N] (an intermediate phrase).
N = car, house, mouse, stupidity, delight, forever, down, …
V = run, sleep, smoke, rise, depend, forage, …
Note: The symbol ‘<-->’ means that a form on the left side of the arrow is mapped into the
structure on the right and the form on the right side is mapped into the left side.
3.2 The form of Morphological Rules
D = {the, this, these, that, those}
Q = {{a, an, one, ø}, some, few, a few, several, … }
A = {happy, red, large, petite, long, deep, fuzzy, …}
Some syntacticians question question whether rules such as the VP expansion rule is really
necessary. For example, the lexical entry for DESTROY should include the fact that it requires a direct
object (a complement):
E.g. [V DESTROY + ____ NP].
They query whether the rule ‘VP -> V + (NP)’ is really necessary.
I don’t like the idea that the VP ‘rule’ is really a rule. Rather, it is a statement of sets:
E.g. VP is a set that contains V and NP; V and NP are members of the set VP.
This is merely a statement of sets. I will go one step further and write it as:
E.g. {VP} <--> <V, NP>.
Note: In set theory notation, the comma enclosed in angled brackets indicates linear
order:
VP is a set that contains the ordered set V then NP. This notation is not normally used in linguistics;
the plus ‘+’ denotes order as shown above.
Note: the curly braces can be omitted once it is understood that VP, V and NP are each
a set.
The lexical expansion above is a statement that in essence says:
If one member of the set V is DESTROY, then the second set is NP, which is the
complement of the verb.
What remains in question is how to account for an optional member. In reality, there are no optional
members. Recall that ø as a phonological sign is permitted in set theory. An optional member actually
exists; it merely has ‘ø’ as its sign:
The S ‘John likes to eat’ implies he likes to eat something. The pronoun may take on a zero
form for certain verbs: [V EAT [NP ø]].
3.2 The form of Morphological Rules
The lexical entry for EAT now should be:
{V EAT, {complement, NP, {-ø, ø}.
By ‘-ø’ I mean it must have a phonetic sign.
Not all verb take a zero complement such as DESTROY.
[V DESTROY, {complement, NP, -ø}].
The logic for the ø complement rests in set theory and the 3-component theory. The complement fills
the function role, and its form is ø, and its sign has to be ø. If it has no form how can it have a sign?
Each component constitutes a set, usually called the complement of the verb, or an argument of the
verb:
E.g. :{COOK, {-ø, +ø}, {-ø,+ø}.
A morphological form has no form, but it has a sign:
E.g.: {[+Pl] (of certain nouns), ø, voiced final obstruent}:
E.g.: calf, calves = /kæf, /kævz/.
The fact that EAT takes theme argument prompts this analysis.
In morphology, the plural suffix ‘-s’ would have the grammatical feature indicating that it requires a
noun as a host:
H: [/z/, N ___, ‘plural’].
D: {[+Plural], N ___, /z/].
The ordering is not crucial, but it should be used consistently. The square brackets are often used to
denote a feature. The ‘+’ (or ‘-’) is a binary value:
E.g. [+Plural] = ‘plural’, [-Plural] - ‘singular’.
This distinction becomes important once the theory of binary oppositions is adopted.
3.2 The form of Morphological Rules
3.2.2 The Word Based Model.
Here, the fundamental significance of the word is significant. Rather than breaking words down, a
word-schema is formulated. For example, the following English verb-forms can expressed in the
following:
a)
hits, sits, types, knows, feels, acts, procrastinates, regurgitates, and so forth
b)
[/Xz/, V, ‘third person singular of V]
Where X is the Lexeme of each verb in a), V and /z/ ‘third person singular of Vs’. /X/ is a phonemic
string such as /plej/. the word-schema a) that there is a list of word forms that end in /z/, and that they
are verbs (V), and that /a/ is the third person singular of V.
There is a closely related schema:
E. g [/X/, V, ‘x’]
Now.the two schemas can be represented in the following mapping correspondence:
E. g. [/Xz/, V, ‘third person singular of X] <--> [/X/, V, ‘x’].
3.2 The form of Morphological Rules
This model eliminates the need for morphemes, bases, or roots. Words are related by mapping one
scheme to another.
E.g. play/played:
[/X/, V, ‘x’] <--> [Xd, V, ‘past tense of ‘xd’]
“there is a string /X/, it is a verb (go, write, play, cough, …), and ‘a function
‘x’”; this
corresponds with the string /Xd/, the same verb, and the past tense of ‘xd’
‘xd’ is the function x
of /X/, and ‘d’ which has the function ‘past tense’.
E.g. [/plej/,V, ‘engage in games’] corresponds with [/plej/. V. ‘engage in games
past
tense’. PLAY <—> PLAYD. PLAY stands for the first bracketed sequence in the above
line
and
PLAYD stands the second sequence.
Now is this perfectly clear? Methinks not. These rule schemas don’t cut the mustard as afar as I am
concerned. Dr. A. told me that these are explanatory, but just descriptive. If they cannot lead to an
explanatory goal, why bother. At least we should try to become familiar with them, just in case I turn out
to be on the wrong track.
Let us go along with set theory, followed by many logicians and mathematicians and possibly others in
other fields. First, the following: “hits, sits, types, knows, feels, acts, procrastinates, regurgitates” form a
set in their own right, just as the uninflected form (infinitive form) is a set and the proposed Lexemes of
them form a set:
E.g. HIT, SIT, TYPE, KNOW, FEEL, ACT, PROCRASTINATE, REGURGITATE.
The third person singular is a set with one member: /z/. Recall that each lexeme and grammeme has
three properties:
E.g.: {HIT, V stem, /hit/}; {3 P. Sg. (3PS), suffix, /z/}.
The form hits is a set that contains two subsets (members):
E.g.: {HIT, 3PS} <--> {/hit/+/z/}.
This tells us that {HIT, 3PS} can be mapped to {/hit/}+{/z/} = {/hitz/}. Note the phoneme symbol ‘/’ is a
set marker for phonemic sets; using both ‘{/’ is totally redundant. I did so here to emphasize this point.
Using one or the other is fine; just that ‘/’ give us more information.
3.2 The form of Morphological Rules
First I’ll do writing as two morphemes:
WRITE
funct ion
V
form
/ rajt /
sign
[+Progressive]
function
[+Suffix]
[Noun Host]
form
{iñ}
sign
The feature [+Suffix, Noun house] accounts for the adjunction of the suffix to the noun stem: writing
(writ-ing). This operation takes place in the syntax in the version of 322 that I taught until 2003.
In the word-based model we obtain:
‘Prog of / rajt /
f unct ion
V-wordf orm f orm
V-wordf orm
f orm
/ rajt /
/ rajt iñ /
sign
‘writ e’
f unct ion
sign
Let’s look at an example of word derivation.
E.g. work+agent <--> worker. (more or less)
[/X/]V, ‘x (= an agent)’] <--> [/X/]V, N, one who ‘xs’]. But is it really desirable to do so? H seems to think
so, but I am being a bit reserved for reasons that I will talk about in the next chapter (I hope).
3.2 The form of Morphological Rul es
WRITE
function
agent
f unct ion
V-wordf orm f orm
[ +Suf f ix, V-host ] f orm
/ rajt /
sign
/ iñ/
' writ e'
function
sign
'one who writes'
f unct ion
V-wordf orm f orm
V-wordf orm
f orm
/ rajt /
/rajtr/
sign
sign
H generalizes the word-based by replacing ‘write’ with ‘X’:
E.g. ‘one who Xs’
In the morpheme based model, one can do similarly:
{X, V-wordform, /x/}, where /x/ is the sign of X.
There does not appear to be a big difference with the exception that the word-based grammar uses
repetition. H uses ‘X’ which really means the set of words of a given class. The section following on
morpheme subtraction seems to support the notion of a morpheme rather than a string of phonemes.
Set theory predicts morphemes, word-based grammar does not. H mentions that bases are not
necessary. Another difference is that in word-based grammar, the entire word-based form corresponds
to another word-based form. In morpheme based grammar, the mapping is from function to sign and
vice versa.
3.2 The form of Morphological Rules
In the morpheme based model we need to be more specific of rule writing. The concatenation rules of
Morphology will join sign of the lexeme and the sign of the grammeme. The feature [+Suffix] tells us
that the grammeme is a suffix to be adjoined to the right end of a lexeme, and the feature [V-host] tells
us that the host, the lexeme to which it must be adjoined, must be a verb. Similarly, [N-host] tells us
that the host is a noun. Although Chomsky uses A for modifiers, we must be careful to exclude degree
words and phrases from A: A = {adjective, adverb}. Therefore, WRITE and [+Progressive ([+Prog]) <-> /rajt-iŋ /. The rules for obtaining a verb and one of its inflectional suffixes is determined in the syntax,
especially where syntax and morphology overlap. I taught this approach in syntax (L322) for several
years. Where as some linguists believe there is no formal division between syntax and morphology
(the distinction is one of convenience rather than formal), there are others who believe that the words
syntax and morphology should not be used in the same sentence. (I suspect that H is leaning in the
direction of the latter.)
The schemas are not a theoretical device, but a descriptive device. Recall the 3 goals of a theory. The
main point of word-based grammars are the schemas. That is where a true comparison will occur. The
schemas have no explanatory value. Recall that our third goal is to find the best explanatory system
that will account for the facts of the corpus.
H will have more to say on this later. Therefore, we should not make any conclusions at this time, but
we should try to understand both approaches.
3.3 The form of Morphological Rules
3.3.1 Pattern Loss
Pattern loss is one (or more) inflectional categories. For example, H mentions Ancient Greek
and its evolution into New Testament Greek. The nominative case forms for Ancient Greek adelphós
‘brother’:
adelphós (singular), adelphó: (dual), adelphoí (plural)
By the time of NT Greek, the dual had disappeared completely without a trace leaving two
grammatical categories: Sg. and Pl.
3.3.2 Coalescence (Merger)
Coalescence is a diachronic change where two syntactically separate word-forms (related
grammatically) coalesce or merge into one complex word-form. In Old Russian, the reflexive of a verb
was formed with a reflexive functional word. By Modern Russian the reflexive had been
phonologically reduced so that it could not bear independent stress. The form had become
inseparably conjoined to the end of any-word form of a particular verb. Not
all verbs form the
reflexive in this way. :
E.g. OR ‘mytɩ sebja’ to wash oneself.
MR ‘mytsja’ to wash oneself.
In MR some verbs still take the modern form of the reflexive to form the reflexive. The reflexive
suffix has other functions and can change the meaning of the stem. This form seems to be more of
weak clitic that must adjoined on to a verb. This is somewhat similar to H’s hypothesis that ‘walk did’
(roughly) -> ‘walked’. In pre-English, the past t. suffix became adjoined to the verb stem.
3.3.3 Analogical Change
Analogical change is a when an inflectional pattern is modified to be another pattern.
There are verbs in English which were regular (weak) at one time but became
irregular (strong) following the pattern of another verb class.
E.g. the past tense of dive is dived in standard English.
3.3 The form of Morphological Rules
Over time the substandard form dove came into existence due to the pattern of strive - strove.
Arrive follows the same pattern. The interesting item here is that the analogical change applies to
the past tense, but not the non-progressive participle (in my dialect, at least; based on for some
verbs; based strive, strove, striven:
E.g. dive, dove, dived (*diven); arrive, arrove, arrived (*arriven),
But for other verbs, the non-progressive participle as well; based on sting, stung, stung:
E.g. dig, dug, dug; drag, drug, drug (this one really gets the purists going). 3.3.4
Reanalysis
Reanalysis is when a morpheme — a root — loses its function and becomes part of
morpheme. H cites the Ancient Greek example of kithára.
kithára
guitar
kithar-îzo to play the guitar: -izo is adjoined to the stem, ‘to use X’
kithar-is-té:s A guitar player (who does something with X, X a noun): -tés ‘one
who uses X.
later:
kithar-isté:s A guitar player (who does something with X, X a noun).
Through reanalysis, the two suffixes ‘-es-’ and ‘-tés’ became merged as ‘-istés’ single
‘one who does something with X, it is directly added to noun stems.
another
suffix:
Secretion is defined (roughly) as the process when a string of phonemes is reanalyzed
so that the string becomes a morpheme with such-and-such a function. H cites the
example
of alcoholic (alcohol-ic) which influences new words ‘X’ to first form a blend
workaholic. A new
morpheme arose ‘aholic’, when ‘work-’ is subtracted from workaholic leaving ‘aholic’ behind as a
suffix, when can then be added to some nouns with the meaning of one who indulges in X to a
relatively high degree.
The form of Morphological Rules
3.3.5 Other Changes
Phonological changes cease to be regular and predictable:
E.g. At one time the Old-English word for ‘foot’ was ‘fo:t’ and ‘feet’ was ‘föti’. The
umlauted vowel gradually became /e:/, and at some point in time, the plural
marker
for Old English, /I/ was gradually lost leaving /fo:t/ (Sg.) and /fe:t/. After
the Great
Vowel Shift (16th to 18th C), /e:/ -> /I/, and inexplicably, /o:/ in foot’
delaxed to ‘/ʊ/’. After
the loss of word-final (plural marker) /I/, the conditioning
factor was lost. The alternation
is now morphological.
Semantic shift of the morpheme is another change:
At one time in the earlier stages of Russian, the suffix /l/ referred to a participle,
but it is unclear what the original function of the participle was. The participle
combined with an auxiliary verb which corresponds with the auxiliary be in English.
In time
the auxiliary was lost in this construction as was the original way of forming the past tense. The lparticiple took on the meaning of the past tense.
Go to Course Outline, Go to Chapter 2, Go to Chapter 4