EFFICIENCY TOOLS IN THE SPEECHES OF MARTIN LUTHER KING, JR. 
M. Cassandra Foster Smith 
Computer Science Program 
Howard University, School of Engineering 
and 
Software Troubleshooters (as of June, 1980) 
Washington, D.C. U.S.A. 
Abstract 
This thesis represents the results of a 
computer-aided analysis of aspects of speeches 
of the Reverend Martin Luther King, Jr. Specif- 
ically, the analysis has investigated the 
occurrence of indicators of the efficiency 
function--tools facilitating the comprehension 
of a discourse by a hearer or reader--in four 
speeches of Dr. King. 
Contrary to the expectations of many who 
anticipate complex grammatical structures in the 
discourse of those who are speechmakers before 
many and diverse audiences, this study has 
demonstrated that the speeches of Dr. King are 
replete with simple structural devices-- 
sequential clauses as opposed to embedded 
clauses, sentences in which there are clear 
linkages between clauses, and clear linkages 
between sentences, to name a few. 
The analysis of the texts of Dr. King was 
accomplished in part by a computer program which 
used as input a surface semantic description of 
a sentence as a basis for predicting the syn- 
tactic function of elements of the sentence. 
Rationales Scope, .and Methodology 
Nida and Taber I propose norms for deter- 
mining the functions of stylistic features. 
They consider the occurrence of features mani- 
festing the efficiency function--for the 
enhancement of comprehension on the part of the 
hearer/reader--and the occurrence of features 
manifesting the special effects function-- 
serving to stimulate hearer/reader interest in 
the discourse. Both formal and lexical features 
might be utilized to effect either of the two 
functions. They further propose that no dis- 
course is restricted to the utilization of a 
single funtion--either efficiency or special 
effects--since a speaker/writer would be con- 
cerned both that his discourse is understood 
and that it evokes interest. This approach to 
style, in terms of its effects on an audience, 
has been discussed by others. See, for example, 
Chapman 2, Helloway 3, and Ohmann~ 
Martin Luther King, Jr. is considered by 
many to have been an effective speaker of our 
times. Certainly, he was effective in attract- 
ing large audiences to his podium, and official 
statistics on the attendance and frequency of 
turnout at his speeches would bear this out. 
Consider, for example, the turnout for the March 
on Washington where the estimated audience was 
200,000. Reston 5 of the New York Times report- 
ed: "It was King who...touched the audience... 
until then the pilgramage...had been a specta- 
cle." Whether this attraction was due to his 
political persuasion, his philosophy of non- 
violence, curiosity, or the sheer beauty of his 
oratory is not a point at issue here. What is 
at issue is that he did have a message, and he 
certainly must have desired his message not only 
to be heard and understood but to be, in addi- 
tion, a catalyst to the furtherance of his point 
of view. 
One might expect, then, that the features 
proposed by Nida and Taber manifesting both the 
efficiency and special effects function would 
be present in the discourse of a speaker of the 
caliber of Dr. King. This study reflects an 
acceptance of these proposals as norms, and in 
so doing endeavors to contribute to the sub- 
stantiation of them as such by measuring the 
extent of reliance by Dr. King on specific fea- 
tures manifesting the efficiency function. 
This is not to say that features, in general 
lexical, manifesting the special effects func- 
tion are not present in the discourse of 
Dr. King, but explicit in this study is a goal 
of specifying a computable process (an algorithm 
/device) for the recognition and identification 
of formal features. 
By statistical accumulation of data evi- 
dencing the presence of certain features 
enhancing the efficiency function, this study 
will endeavor to show that there is indeed a 
high occurrence of formal features for the 
enhancement of the efficiency function in the 
speeches of Dr. King. Specific parameters to 
be addressed herein are: 
i) intersentence markers 
2) markers of relationships between 
clauses 
3) sentences with simple structure 
4) potential terminals in a sentence 
5) clauses in sequence in a sentence 
To accomplish the statistical accumula- 
tions, the research contained herein utilizes a 
sentence parser which provides semantic coding 
as a basis for the recognition of the specific 
units of interest in Dr. King's discourse. The 
parsing algorithm will contain tools for a) the 
recognition of specific kinds of intersentence 
-167-- 
markers; b) the recognition of markers between 
sentences; c) the recognition of simple sen- 
tences; d) the recognition of potential termi- 
nals in a sentence; and e) the recognition of 
sequential vs. embedded clauses in a sentence. 
Clearly, an investigation into the fre- 
quency of occurrence of a limited set of 
stylistic features in the oratory of a single 
person will not permit definitive conclusions 
on the universal set of features present in all 
oratory. Nor will it define an algorithm (or 
device) for the consistent recognition of such 
features. "But it is anticipated that these 
results will serve as substantiation of proposed 
norms (as mentioned above) and as a cue to both 
the minimal information necessitated in the 
processing of natural language as well as a cue 
to a methodology for the recognition of formal 
stylistic features. 
The appropriateness of an examination of 
limited stylistic features, however, has been 
sanctioned by many stylists. Hough 6 states: 
"The choice of this word rather than that, of 
this kind of syntactical construction rather 
than another, is a visible fact, whose nature 
and effects can be examined." He further 
elaborates: "Study a particular stylistic 
device as it is employed in an individual work 
of art .... We are inquiring how a specific 
configuration is used for a special...purpose .... " 
Ullman 7 is also encouraging: "This is the 
way stylistic analysis proceeds: It takes a 
particular device in a language or in a limited 
corpus and examines the different effects which 
emanate from it. But one could also reverse 
the process and investigate, for example, the 
various devices through which irony is expressed 
in the language or corpus." 
It is further anticipated that many lin- 
guists will insist that a speaker such as 
Dr. King will have demonstrated a wide variety 
of structural tools in his oratory. This study 
proposes to show evidence, as has Ohmann 8 in his 
analyses of Faulkner, et al, substantiating 
Milic 9 that "the language of an individual is 
restricted...though...able to interpret the full 
range of syntactic patterns...the language 
affords, for some reason he seems to depend for 
his expressive needs on merely a selection from 
the available total." Further discussion of the 
expectation of restricted use of language tools 
can be found in Chapman 2 and Ohmann. ±0 
Since it is expected that the structural 
patterns present in Dr. King's oratory are very 
limited in number and recurring, one anticipates 
that a parsing algorithm for the recognition of 
these structural units will be simple. A parser 
based on a transformational model, for example, 
would contain many "complications...irrelevant 
to the particular problem...the student of style 
may wish to feel free to use the simplest gram- 
matical model that satisfies the requirements 
of his particular job. ''II Further, it is 
anticipated that the algorithm used in this 
study in identifying the regular recurring 
structural patterns (the analogies) in the 
speeches of Dr. King will also identify the 
anomalies by the rejection of them as analogies. 
Given the above considerations concerning 
the rationale and methodology for the study and 
the delimiting of the scope, the following sec- 
tion will discuss the specific speeches of 
Dr. King comprising the corpus for the study. 
The Corpus 
Four speeches have been selected to com- 
prise the corpus for the study. Text I is 
popularly known as I had a Dream. 12 Text II is 
the Nobel Prize Acceptance Speech. 13 Text III 
is the text of Dr. King's statement on the eve 
of his death, 14 and Text IV is a sermon deliv- 
ered by Dr. King at his home church in Atlanta, 
Georgia. 15 
The texts cited above were selected as 
representative of the gamut of the speech types 
delivered by Dr. King based on the makeup of 
the audiences which were present. Text I was 
delivered to general audience composed of old 
and young, black and white, rich and poor, 
educated and uneducated. Text II was delivered 
to a more elite audience, probably all highly 
educated and distinguished in their fields--the 
Nobel Prize Committee. The third text repre- 
sents an impromptu speech delivered by Dr. King 
on the eve of his death. The final text was 
delivered by Dr. King to his usual audience-- 
his flock at his home church. 
Text I contains 81 sentences, 1,644 words, 
and 156 clauses. The mean (simple average) 
words per sentence is 20.2, and the mean clauses 
per sentence is 1.9. Text II contains 43 sen- 
tences with a total of 1,156 words and 122 
clauses. Here, the mean sentence length is 
26.91 words, and the mean number of clauses per 
sentence is 2.7. Text III consists of 23 sen- 
tences, 34 clauses, and 256 words. The mean 
number of words per sentence is 11.13, and the 
mean number of clauses per sentence is 1.48. 
Text IV contains 136 sentences, 2,316 words, 
and 250 clauses. The mean sentence length is 
17.05 words. The mean number of clauses per 
sentence is 1.8. 
Combining the data from the four texts, 
the corpus consists of 283 sentences, 562 
clauses, 5,391 words. The mean sentence length 
for the corpus is 18.971, and the mean number 
of clauses per sentence is 1.982. The standard 
deviation of the sentence lengths is 12.96 
words. The standard deviation of the clauses 
168 
per sentence is 1.32. Table i summarizes the 
mean sentence lengths in words by text and for 
the corpus. Table 2 contains comparable data 
for clauses per sentence. 
These statistics have been obtained by a 
manual tallying of the sentences, clauses, and 
words in the texts and as such are 
TABLE i 
MEAN WORDS PER SENTENCE BY TEXT AND CORPUS 
Mean 
Text Words/Sentence 
I 20.2 
II 26.9 
III 11.13 
IV 17.02 
Corpus 18.97 
TABLE 2 
MEAN CLAUSES PER SENTENCE BY TEXT AND CORPUS 
Text 
Mean 
Clauses/Sentence 
I 1.93 
II 2.79 
III 1.48 
IV 1.84 
Corpus 1.99 
preliminary and for the purpose of demonstrat- 
ing the volume of information under study. 
Further, the tables presented here are for the 
purpose of describing the raw data worked with 
in this research. Further sections contain 
discussions of the data with respect to the 
parameters under investigation and the signifi- 
cance of them to the style of Martin Luther 
King, Jr. A fuller study of the research dis- 
cussed herein may be found in Smith 16 from which 
this study was extracted. 
The reader should be forewarned that this 
paper is intended to be read by linguists. It 
is assumed that the reader is knowledgeable of 
past and present trends in the branches of 
linguistics. Readers who are uninitiated in 
the field might consider Dinneen 17 which is 
fairly objective in its treatment of the trends 
in linguistics from ancient times to the mid 
1960's. 
Text Analysis Proc@dure 
In this section the grammar and the algo- 
rithm for the analysis of the texts comprising 
the corpus are presented. As stated earlier 
the goals of the analysis are: a) the recogni- 
tion and identification of explicit semantic- 
syntactic markers between sentences, b) the 
recognition and identification of explicit 
semantic-syntactic markers indicating relation- 
ships between clauses, c) the recognition and 
identification of sentences with simple struc- 
tures, d) the recognition and identification of 
potential terminal points in a sentence, and 
e) the recognition of sequential versus embedded 
clauses in a sentence. 
The Grammar 
The grammar underlying the parsing algo- 
rithm is presented below. It is defined as an 
ordered 4-tuple: <V,V~,SC,'S'> where V is the 
vocabulary of terminal symbols (or words from 
sentences or items from the lexicon); V" is the 
vocabulary of nonterminal symbols (the semantic 
codes, etc.); SC is a set of semantic compati- 
bility rules; and "S" is an initial string or 
sentence. The semantic compatibility rules 
are presented below as a numbered set of rules 
in Backus-Naur Form to more efficiently indi- 
cate alternative reductions. 
In Backus-Naur (or Backus-Normal) Form 
(BNF) alternative reductions for an element 
of V or V" are specified in one rule. This 
is in contrast to the method used generally 
by linguists in specifying a separate rule 
for each alternative. For example, many lin- 
guistic systems would specify rule 1 as: 
E÷EE 
E ÷ EAE 
E ÷ E-lexeme 
E ~+ EV O E. 
In the notation utilized here nonterminal sym- 
bols are enclosed in angular brackets (< >), 
and the alternatives are separated by vertical 
bars ( I )" 
The Semantic Compatibility Rules 
i. < E >: = <E> <E> I <E> <At <E> I E-lexeme 1 
<EV> <0> <E> 
2. < S >: = <K> I <K> <RCUTS> <S> I <RCUTS> <S> 1 
3. RCUTS: = WNILEIALTHOUGHIWHEREITHATIWHEN I 
WNOIWHOSEIWHICHIWHATIBECAUSEIWNYIUNTILI 
ASIWHENEVERIBUTIIFIFORISOIEVERYTIMEI, I 
- I " I ' I ANDIORI : 
4. < K >: = <E> I <X> <E> <X> I <X> <E>I<E><X> 
169- 
5. < X >: = <0> \] <A> I<RHT> 
6. < A >: = <A> <A> I <A> <RHM> <A> I A-lexeme 
7. < O >: = <0> <0> I <A> <0> \] <O> <A> I <0> 
<RHM> <O> I O-lexeme 
8. <RHT>: = <RHT> <0> I <RHT> <RHT> I <RHT> 
<P~M> <RHT> I RHT-lexeme 
9. <RHM>: = ANDIORIBUTINORINEITHERIEITHER I 
In the above rules E-lexeme, O-lexeme, 
A-lexeme, etc., refer to a word from the lexicon 
belonging to the semantic domain indicated. 
RCUTS represents a subset of relators (R). 
The grammar described above is based on a 
linguistic statement developed by Dr. Michael 
Zarechnak 18 for the analysis of weather report 
data. The approach discussed here represents 
an expansion and extension of the approach to 
more general literature. The author provided 
the computational approach and methodology for 
the weather report analysis. Using Kholodo- 
vich's 19 theory of subsets it was determined 
that the weather report sentences could be 
represented in a limited number of frames. 
(See Smith20, 21 for a summary.) In the treat- 
ment of the sentences as string formulae this 
grammar is also influenced by the work of 
Zellig Harris. 22 
With respect to the goals listed at the 
beginning of this section, the analyzer (parser) 
described above accomplishes the identification 
of explicit markers between clauses, and rejects 
those sentences in which there are no explicit 
markers between pairs of clauses. Via the 
recognition of sequential clause boundaries, 
it also indicates potential terminal points 
in a sentence and the number of clauses per 
sentence. Since the kernelization algorithm 
relies on explicit markers between sequential 
clauses to effect kernelization and rejects 
sentences with embedded clauses, it indicates 
sequential vs. embedded clauses. In flagging 
sentences with introductory relators, the 
analyzer has also accomplished a portion of the 
identification of explicit markers between 
sentences. 
Some final observations should be added 
at this point concerning the goals and motiva- 
tion for the analyzer described above. The 
analysis is accomplished by the analysis of 
word-level semantic units to obtain information 
on the surface sentence structure. No deep 
semantic structure analysis (ala generative 
semantics) summoning, for example, hypothetical 
underlying propositions or deep syntactic struc- 
ture analysis (ala transformational theory) 
providing, for example, proposed kernel struc- 
tures underlying verbal (infinitive, gerundive, 
participial) phrases is included. 
The clause (or kernel sentence or configu- 
ration) discussed herein is strictly a surface 
structure unit in which a finite verb form 
occurs with or without its subject or object 
noun-phrase. That is to say, a configuration 
with one or more deleted noun-phrases but with 
a finite verb form is considered a clause. 
This type of analysis, then, in terms of 
the surface structure alone is considered by 
some to be a current trend. J. D. Fodor 23 for 
example, states: "The move has been from the 
interpretation of deep structures alone (as in 
Katz's theory) via the interpretation of deep 
and surface structures (as in Jackendoff's 
theory) to the interpretation of surface struc- 
tures alone." For a contrary view one is 
referred to Woods. 24 
Statistical Measures of Style 
In the analysis of selected texts this 
study has accumulated statistics on certain 
features proposed as manifestations of the 
efficiency function in the discourse of Martin 
Luther King, Jr. These specific features were 
chosen following the dictum of stylists such as 
Chatman: 25 "we count features only to demon- 
strate that they are distinctive enough to be 
noticed without recourse to statistics; the 
count is significant only insofar as it corre- 
sponds to a preformed impression in the sensi- 
tive reader's mind." The study contained herein 
may well exemplify what Enkvist 12 alluded to in 
the following: "It is the task of Linguistic 
Stylistics or Stylolinguistics to set up inven- 
tories and descriptions of stylistic stimuli 
with the aid of linguistic concepts." 
The Style of Martin Luther. King, Jr. 
Ullman 7 discussed two general approaches 
to stylistic studies. In the first of these 
the stylist begins with a stylistic device and 
analyzes a corpus for effects from the device. 
In the second approach the stylist begins with 
an effect and analyzes a corpus for devices 
which have contributed to the effect. This 
approach might be graphically depicted as 
follows: 
effect~ ~aevice 
~~de~e e 
-- %~4eo 
- 170 
The approach taken in this stylistic study 
of the oratory of Martin Luther King, Jr. might 
be considered analogous to the second approach 
above. Its graphic depiction is as follows: 
~ c~ o~.9~ 
efficiency~ ~ sequential clauses 
~oe e 
such that the effect proposed is the efficiency 
function (facilitating understanding on the 
part of the hearer/listener), and the devices 
under inspection in the analysis are a) inter- 
sentence markers, b) markers of relationships 
between clauses, e) sequential clauses, d) sim- 
ple sentences, and e) potential terminal points. 
With respect to intersentence markers it 
has been shown that all of the texts are 
strongly linked by repetition, anaphora, and 
introductory relators. An inspection of the 
texts shows that there is indeed a preponder- 
ance of sentence to sentence links plus links 
joining early parts of the texts to latter 
portions. The text coherence is accomplished 
in part in text I by repetitions such as "I 
have a dream", "Let freedom ring.", etc. In 
text II, one can observe the various repetitions 
and references to "I accept the Nobel Prize for 
Peace", "this award", "this prize", "I accept", 
etc. Text III is less linked by anaphora and 
repetition than the other three texts. Dr. King 
relies more on introductory relators with 56.52% 
of the sentences in this impromptutext deliv- 
ered on the eve of his death containing intro- 
ductory relators. Text IV is linked by repe- 
tition and references such as "It's midnight.", 
"the bread of hope", "the bread of faith", "the 
bread of love." Clearly, therefore, this 
device is heavily present in Dr. King's oratory. 
The second device indicative of the effi- 
ciency effect under analysis in the texts is 
the presence of markers between clauses. In 
the four texts 151 of the 283 sentences are 
multi-clausal. Of these 137 (48.41%) contain 
explicit markers between clauses. Fourteen of 
these (4.95% of the corpus) contain clauses in 
which the boundaries between are unmarked. 
Further, with respect to the third device-- 
sequential clauses--141 sentences (49.82% of the 
corpus) are composed of clauses in sequence 
while i0 sentences (3.53% of the corpus) contain 
embedded clauses. Thus with respect to marked 
versus unmarked clause boundaries and sequential 
versus embedded clauses, 127 sentences (44.88% 
of the corpus) contain clauses in which the 
interclause boundaries are marked and in which 
the clauses are in sequence. Twenty-four sen- 
tences (8.48% of the corpus) contain sentences 
in which the interclause boundaries are unmarked 
or in which embedded clauses are present. Fur- 
ther, none of the sentences with unmarked clause 
boundaries contains embedded clauses, and none 
of the sentences containing embedded clauses 
has unmarked interclause boundaries. 
In considering the fourth device--simple 
sentences--this study has shown that 132 sen- 
tences (46.64% of the corpus) are simple with 
respect to the number of clauses (one) per sen- 
tence. Further, since clauses in sequence are 
less complex and easier for the listener to 
decode, one may consider the 127 sentences with 
clauses in sequence and marked interclause 
boundaries simple also. Thus, 259 sentences 
(91.52% of the corpus) are simple. 
Finally, with respect to potential terminal 
points in a sentence, one can see that clauses 
in sequence provide potential terminal points. 
Thus, the 141 multiclause sentences which 
contain clauses in sequence are evidence of the 
potential of terminal points in addition to 
the point at the end of the sentence. 
Clearly, from the above one can see that 
the five devices proposed as contributors to 
the efficiency effect are present in high pro- 
portion in the oratory of Martin Luther King, 
Jr. This study has shown that, consciously or 
not, Dr. King has structured his oratory such 
that there is a preponderance of tools which 
aid the hearer/reader in the comprehension of 
the text. One might conclude by agreeing with 
Nichols 26 who quotes Aristotle as having said 
that to know what to say is not enough, but 
one must also know how it must be said. She 
summarizes by stating that, indeed, the effec- 
tiveness of Adlai Stevenson, John F. Kennedy 
and, Martin Luther King is in part due to the 
"uniqueness of their handling of the language 
patterns." 
General Conclusions 
As was stated in the introduction, this 
study does not purport to capture the total 
style of Martin Luther King, Jr. It has 
endeavored to show evidence of specific struc- 
tural (and lexical to a very small extent) 
devices which enhance the efficiency function. 
Anyone inspecting these texts of Dr. King's 
171 
may observe that there is a wealth of both 
structural and lexical tools which might provide 
inspiration to further study of the oratory of 
Dr. King. 
With respect to structural analysis, one 
might consider the role of infinitive phrases, 
gerunds, and participles in Dr. King's texts. 
One might also be interested in analyzing the 
clause construction of his sentences with re- 
spect to modifiers, the presence or absence of 
the subject and object, etc. 
There are also indications of studies which 
might approach an analysis of Dr. King's style 
on the lexical level. One observes many in- 
stances, especially in "I Had a Dream ''12 of 
metaphor , simile, collocative clash (unusual 
combinations), and dated and outmoded words. 
These are just a few of many indications of 
further direction in the study of the style 
of Dr. Martin~Luther King, Jr. 
This research was supported in part by the 
Faculty Research Program in the Social Sciences, 
Humanities, and Education, Office of the Vice 
President for Academic Affairs, Howard Univer- 
sity, Washington, D.C. 

References 

I. Nida, Eugene A. and Taber, C. R. Com~onen- 
tial Analxsis of Meaning. Mouton, The 
Hague, 1974. 

2. Chapman, Raymond. Linguistics and Litera- 
ture! An Introduction to Literary 
Stylistics. T.&A. Constable, Edinburgh, 
1973. 

Holloway, John. Matthew Arnold. The Victo- 
rian Sage: Studies in Argument. Archon 
Books, Hamden, N.J., 1962. Reprint 
edition, Babb, Howard S. (Ed.). Essa~fs 
in Stylistic Analysis. Harcourt Brace, 
Jovanovich, 1972, 293-320. 

Ohmann, Richard M. Literature as sentences. 
Essa~sin Stylistic Anal~sis. Babb, 
Howard S. (Ed.) Harcourt Brace, 
Jovanovich, 1972, 353-64, 

Reston, James. The March on Washington. 
New York Times. August 29, 1963. 

Hough, Graham. Stxle and Stylistics. Human- 
ities Press, New York, 1969. 

Ullman, Stephen. Two approaches to style. 
Patterns of Lit erar XStyle. Strelka, 
Joseph (Ed.). The Pennsylvania State 
University Press, University Park, 
1971. 

Ohmann, Richard M. Generative grammars and 
the concept of literary style. Word 
20, 1964, 423-39. 

Milic, Louis. Introductory essay. Stylists 
on Styl i . Milic, Louis T. (Ed.). 
Charles Scribener's Sons, New York 
1969, 1-24. 

Ohmann, Richard M. Prolegomena to the 
analysis of prose style. Style in 
Prose Fiction. Martin, H. (Ed.). 
Columbia University Press, New York, 
1959. Reprint Edition, Babb, Howard S. 
(Ed.). Essays i n Stylistic Analysis. 
Harcourt Brace, Jovanovich, New York, 
1972, 36-49. 

Enkvist, Nils E. Linguistic S~listics. 
Mouton, The Hague, 1973. 

King, Martin Luther, Jr. Text of speech 
delivered at Lincoln Memorial, 
Washington, D.C.: I had a dream. 
August 28, 1963. 

King, Martin Luther, Jr. Nobel Prize 
acceptance speech at Oslo, Norway. 
December i0, 1964. 

King, Martin Luther, Jr. Text of Dr. King's 
statement on the eve of his death. 
April 3, 1968. 

King, Martin Luther, Jr. Message to the 
world: It's midnite. (Date unknown). 

Smith, M. Cassandra. The Grammatical Struc- 
ture of Oratory. University Microfilms 
International, Ann Arbor 1980 (Forth- 
coming). (1979 Ph.D. Dissertation, 
Georgetown University, Washington, 
D.C.) 

Dinneen, Francis P. An Introduction to 
General Linguistics. Holt, Rinehart 
and Winston, New York, 1967. 

Zarechnak, Michael and Coyne, E. Semantic 
analysis of natural language state- 
ments. Linguistics 182. Mouton, The 
Hague, 1976. 

Kholodovich, A. A. Theory of word sub- 
classes. Voprosyjazykoznanija i. 1960, 
32-43, 

Smith, M. Cassandra. Status of a tutorial 
reading course and data gathering for 
semantic input for a syntactic analy- 
sis. 1977. (Course paper Georgetown 
University, Washington, D.C.) 

21. Smith, M. CassaNdra. Kernel sentence analy- 
sis based on the theory of subclasses. 
1978. (Course paper Georgetown Univer- 
sity, Washington, D.C.) 

22. Harris, Zellig S. String Analysis of Sen- 
tence Structure. Mouton, The Hague, 
1962, 29-42. 

23. Fodor, Janet Dean. Semantics: Theories of 
Meaning in Generative Grammar. Thomas 
Y. Cromwell, New York, 1977. 

24. Woods, William A. What's in a link: founda- 
tions for semantic networks. Represen- 
tation and Understanding. Bobrow, 
Daniel and Collins, Allan (Eds.). 
Academic Press, New York, 1975, 35-82. 

25. Chatman, Seymour. Linguistics, poetics and 
interpretation: the phonemic dimension. 
The Quarterly Journal of Speech XLIII. 
1957. 

26. Nichols, Marie H. Rhetoric and style. 
Patterns of Literar~Style. Strelka, 
Joseph, (Ed.) Pennsylvania State 
University Press, University Park, 
1973, 130-43. 
