IIIIIIIIIII III

         

III

    

IIIII IIIII L IfIBRARY

305

 

 

 

Michigan State
University

 

This is to certify that the
thesis entitled
STRATIFICATIONAL APPROACHES TO AMHARIC PHONOLOGY

presented by

Michael Eric Bennett

has been accepted towards fulfillment
of the requirements for

m degree in Limujitm

Xéitflk/g&{-/(2K§QL¢EVZI

Major professor

Date Max I7, 1978

0-7 639

 

 

WM 3 m

 

@ Copyright by
MICHAEL ERIC BENNETT
I978

 

STRATIFICATIONAL APPROACHES TO AMHARIC PHONOLOGY

By

Michael Eric Bennett

A THESIS
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Linguistics and Oriental and African Languages

l978

 

 

ABSTRACT
STRATIFICATIONAL APPROACHES TO AMHARIC PHONOLOGY
By

Michael Eric Bennett

This thesis examines how a stratificational model of language can be
applied to Amharic. Chapter I provides overviews of both Amharic and
stratificational theory; it is noted that an adequate phonological
analysis requires at least a partial analysis of the morphology. Chap-
ter II presents the data: the infinitive, instrumental, and participle
forms of basic stems of bi-, tri-, and quadriradical verbs. A morpho-
logical analysis involving prefixation, suffixation, and morphotactic
simultaneity is given, with some alternations conditioned by the tactics
and others accounted for with Special morphons. Chapter III discusses
the realization of special morphons, and provides a phonotactics ac-
counting for the distribution of phonemes within the phonological word
and for the interdigitation of consonants and vowels necessary for the
assumed morphological analysis. A componential analysis of the Amharic
phonemes is given, and a series of tables provides detailed illustration

of the tactic and realizational phenomena which account for the data.

 

Till
min alskade morfar
ELRICK KNUT SUDERQUIST
som sdkte ett nytt liv i det nya landet

och fann det.

iii

 

 

ACKNOWLEDGEMENTS

Many individuals have contributed to this thesis, often in ways
having little or nothing to do with linguistics. To mention all such
persons would require a tome half again as large as this one; a few
are deserving of special recognition, however, for having rendered
service above and beyond the call of duty.

1 must first thank my teacher and friend Tesfaye Abagaz, without
whom it is likely that this work would never have been undertaken.

I would next like to thank the members of my committee, Dr. Seok
C. Song, Dr. Carol Myers Scotton, and Dr. David G. Lockwood. Their
thoughtful suggestions and cogent criticisms have continually shown
them to be able scholars whose footsteps are well worth following in.
Special thanks are due to Professor Lockwood, the chairman of my com-

mittee, and a man for whom there is no more fitting word than mentor.

To my dear and valuable friend Victoria Maddox Kingsbury I can say

only "Thank you,’I hoping that my expression of gratitude is all the more

eloquent for its simplicity.
Finally, my most heartfealt thanks are offered to my parents,

Harold Dean Bennett and Jean Elizabeth deerquist Bennett.

Chapter
I

II

 

TABLE of CONTENTS

List of Figures

List of Tables

Preliminaries

T An Overview of Amharic

2 The Topic

3 The Scape of the Thesis

4 The Theoretical Framework

.4.l The Graphic Notation

.4.2 The Nature and Form of a Stratum
.4.3 Phonology in Stratificational Theory
.4.4 Results of the Analysis

otes to Chapter I

Some Aspects of Amharic Morphotactic Structure
Traditional Analyses of Amharic Morphology
Modern Approaches to Amharic Morphology

l Transformational Approaches to Semitic Word-

Formation

2.2.2 Additional Comments on Semitic Word-Formation

in Transformational-Generative Theory

2.2.3 Stratificational Approaches to Semitic Word-

Formation

2.3 A Stratificational Treatment of Some Amharic
Nominals ‘

3.1 The General Structure of Nominals in Amharic

.3.2 The Morphotactic Structure of Nv-Stems

.3.2.l The Infinitive

.3.2.2 The Instrumental
3
3
t

NN—l

. .2.3 The Participle
. .2.4 The Integrated Description
0 es to Chapter II

vii

ix

—l-—l-—J —-l-—l._l
(DNN ChoomoomwwN-d-J

N—l
OCD

 

 

III

 

Aspects of the Phonology of Amharic
l Introduction
2 Previous Analyses of Amharic Phonology
.2.l Traditional Approaches to the Phonology of
Amharic
.2 Leslau's Phonemic Analysis of Amharic
The Phonemic Status of the Labiovelars
The Status of §_and 3_in Amharic
The Phonotactics of Amharic
The Phonotactics of Segments
The Onset
The Pre-Nucleus
The Nucleus
The Final
The Integrated Phonotactics
A Componential Analysis of the Amharic Phonemes
The Phonemic Sign Pattern
.l The Phonemic Sign Pattern for Consonants
. .2 The Realization of Vowels
The Realization of the Infinitive, Instrumental,
and Participle Forms
.1 Vowel Morphons with Alternate Realizations
.2 The Realization of the Data in Table 5 from
the Level of Morphotactic Structure to the
Level of Phononic Representation
Class I Roots
Class I(A) Roots
Class I(B) Roots
Class I(C) Roots
Class I(D) Roots
Class I(E) Roots
Class II Roots
.1 Class II(A) Roots
. .2 Class II(B) Roots
. .3 Class II(C) Roots
Class III Roots

0301-5me

3
3.
3.
3
3
3
3
3
3
3.
3
3
3
3
3
3
3
3
3

23000 bowuummmmmmmm-bwm

00w

U'I-wa—J

l

l

l

l

l

l

2

2

2

2

3

3.l Class III(A) Roots
.3.2 Class III(B) Roots
4 Class IV Roots

4.l Class IV(A) Roots
.4.2 Class IV(B) Roots
5 Class V Roots

.5.l Class V(A) Roots
.5.2 Class V(B) Roots
6 Class VI Roots

7 Class VII Roots

8 Class VIII Roots

to Chapter III

croooooooooooooooooooooooooooooooooooooooooooo
U) NNNNNNNNNNNNNNNNNNNNNN

zwwwwwwwwwwwwwwwwwwwwww
m.

0..

List of References

Appendix
vi

TOO
103

N

NNNN

.lO

.ll
.l2

 

LIST of FIGURES

The organization of language in stratificational
theory.

The DIAMOND node and its possible connections
Part of a morphemic stratal system.
A fragment of the Amharic morphotactics.

The morphemic sign pattern for the Amharic data in
Table 3.

Phonotactics for Amharic interfixation.
The general morphotactic structure of nominals.
The structure of a complex and a noncomplex Na-stem.

The lexo- and morphotactic structure of the infin-
itive.

The realization of M/Inf2/ for roots in classes I,
III, and IV.

The realization of M/Inf2/.

The lexo- and morphotactic structure of the instru-
mental.

Revised analysis of M/Inf2/ with conditioning by
M/Inst/.

The realization of M/Part/.

The integrated morphotactic description of the infin-
itive, instrumental, and participle forms.

The realizations of MN/U/, MN/0/, and MN/E/.

The conditioning of 2_in loanwords by style.

vii

Page

TO
22

24
24
27
3o

33

34
36

38

39
41

42
SI
54

 

(A)

0.)

wwwwwww

wwwww

#0.)

.l0

.ll
.12
.l3
.l4
.l5
.l6
.l7

 

The general structure of the phonotactics.
The onset position.

The pre-nucleus in the phonotactics.

Revision of Figure 3.5 to account for reduplication.

The nucleus, including zero realization of normal
vowels.

The final position.
The integrated phonotactics.

The alternation of PN/Ap/ and PN/Fr/ as conditioned
by P/ip/ and P/ep/.

The integration of components into the tactics.
The phonemic sign pattern for staps.

The sign pattern for all spirants except P/s_/.
The phonemic sign pattern for P/s/.

The phonemic sign pattern for sonants.

The realization of vowels in terms of components.

Vowel morphons exhibiting alternations.

viii

55
60
60
62

63
65
66

69
7O
7l
72
73
73
75
77

 

Table

01-wa

TO

ll

l2

l3

l4

l5

l6

 

 

 

LIST of TABLES

Words based on the root ML§_'answer.'

AND and OR nodes and their meanings.

Words based on the root ML§_'answer.'

Examples of nominal constructions in Amharic

Data on Amharic Nv-Stems

The
The

phonemes of Amharic according to Leslau (T968).

use of MN/#/ in tactic processing.

Reduplicated verbs in Amharic

A componential analysis of the Amharic phonemes.

The
and

The
and

The
and

The
and

The
and

The
and

The
and

realization of the infinitive, instrumental,
participle forms of a Class I(A) root.

realization of the infinitive, instrumental,
participle forms of a Class I(B) root.

realization of the infinitive, instrumental,
participle forms of a Class I(C) root.

realization of the infinitive, instrumental,
participle forms of a Class I(D) root.

realization of the infinitive, instrumental,
participle forms of a Class I(E) root.

realization of the infinitive, instrumental,
participle forms of a Class II(A) root.

realization of the infinitive, instrumental,
participle forms of a Class II(B) root.

ix

Page

22
26
3T
48
57
6T
67

80

80

8T

8T

83

83

85

 

 

17

18

l9

20

2l

22

23

24

25

26

The
and

The
and

The
and

The
and

The
and

The
and

The
and

The
and

The
and

The
and

 

realization of the infinitive, instrumental,
participle forms of a Class II(C) root.

realization of the infinitive, instrumental,
participle forms of a Class III(A) root.

realization of the infinitive, instrumental,
participle forms of a Class III(B) root.

realization of the infinitive, instrumental,
participle forms of a Class IV(A) root.

realization of the infinitive, instrumental,
participle forms of a Class IV(B) root.

realization of the infinitive, instrumental,
participle forms of a Class V(A) root.

realization of the infinitive, instrumental,
participle forms of a Class V(B) root.

realization of the infinitive, instrumental,
participle forms of a Class VI root.

realization of the infinitive, instrumental,
participle forms of a Class VII root.

realization of the infinitive, instrumental,
participle forms of a Class VIII root.

 

 

85

86

86

88

88

90

90

9T

91

93

 

 

 

CHAPTER I

l Preliminaries
l.l An Overview of Amharic

The Amharic language of Ethiopia is commonly assumed to have
evolved from a language very similar to Ge'ez, the language of the old
Aksumite Empire; Ge'ez is still in use today as the liturgical language
of the Ethiopian Church, but Amharic has replaced it as the official
spoken language of the country. It is estimated that there are present-
ly between seven and eight million native speakers of Amharic.1

Amharic is classified as a Semitic language, and shares a common
historical source with such better-known languages as Hebrew and Arabic.
These languages have a derivational morphology based on a system of con—
sonantal roots which combine with different vocalic patterns to form
words. Table l illustrates this phenomenon with a number of Amharic
words, all of which are based on the root MES 'answer.I Some of the
words have additional affixes as well; these are separated from the
derivational stem by a hyphen.

Amharic has a number of dialectal forms, but the form spoken in the
province of Shoa, which includes the capital city of Addis Ababa, is
generally considered as the standard. This thesis will describe the

speech characteristic of an educated native speaker of the Shoa dialect.

 

Table 1: Words based on the root ML§_'answer.'

Amharic §flg§§_

MaLLaS-a 'he answered'

MaLS 'answer' (n.)

MELS-at ‘(she) having answered'
al-MaLS-am 'I do/will not answer'
MaLES 'answer!’ (59. masc. imp.)

N.B.: In this table, the consonantal radicals of the root have been
capitalized. The transcription used in this thesis is that
commonly used in Amharic studies; see the Appendix for a com-
plete description of it.

l.2 The Topic

This thesis is a preliminary description of the phonological struc-
ture of Amharic. "Phonological structure“ in the sense used here implies
morphophonemic, phonemic, and phonotactic phenomena. One of the primary
objectives of this study is to provide a basis for further work toward
an exhaustive treatment of Amharic phonology.

Because of the interrelatedness of the phonological and morpholo-
gical systems of a language, even a preliminary description of the pho-
nology of Amharic must take into account the radical/interfix pattern of
the morphology, as illustrated above. The phonotactics must account for
the proper interdigitation of consonants and vowels; one way in which
this may be done was pr0posed in a treatment of Iraqi Arabic by Al-
Azzawi (T973), a treatment which could in theory be extended to a typo-
logically similar language such as Amharic. VHer preposal and its short-

comings will be discussed further in Section 3.5.

 

 

1.3 The Scope of the Thesis

In a work of this size, it is impossible to describe the entire
morphological system of Amharic. For this reason, it was decided to
analyze only a relatively small portion of the nominal morphology of
the language. The participle, infinitive, and instrumental nouns of
the basic stems of bi-, tri-, and quadriradical verbs have been selected
to illustrate the treatment of derived nominals. This sample is large
enough to be representative of the phenomena involved in derivation, yet
small enough to be manageable.

The treatment of phonotactic phenomena is primarily concerned with
the distribution of segmental phonemes and the structure of the phono-
logical word; in addition, a componential analysis of the Amharic pho-
nemes will be given. Among the morphophonemic phenomena to be treated
are (l) the alternation between apical consonants and their palatal
counterparts (t/é, g/fi, etc.) and (2) the alternation between labialized
and non-labialized consonants and its relationship to arguments concern-
ing the phonemic status of labiovelar stops. The problem of reduplica-

tion is also briefly discussed.
l.4 The Theoretical Framework

The theoretical approach used in this thesis is commonly known as
stratificational theory. Any theory of language which recognizes two
or more levels of structure within a language may be said to be strati—
fied, but in actual practice this term is usually reserved to describe
that theory of language which stems from the work of Sydney M. Lamb and

H. A. Gleason, Jr. and which has been subsequently modified and extended

by other scholars.

In general, all forms of stratificational theory characterize lan-
guage as a static network of relationships which is capable of trans-
mitting impulses in two directions, to account for both encoding and
decoding.2 This bidirectional network of abstract relations relates
meaning and sound by means of a realizational plane intersected at var-
ious intermediate points by stratal systems, which have as their pri-
mary distinguishing mark a syntax or tactics of "elements" at a cer-
tain level of description. Although we must use words such as "element"
and "unit" to discuss the linguistic system, this should not imply that
the network of a language contains substantive items. It may be said
that the guiding tenet of the stratificational school is that language
is a completely abstract system which relates to substance only at its
peripheral connections to other systems, such as the auditory or arti-
culatory apparatus of the individual.

At the present time, the theory allows for up to six different
stratal systems in a language, although the status of some of these
strata is still open to debate. In any event, there is no theory-
specific universal which requires all languages to have the same number
of strata.

Following the convention that "up” is toward meaning and "down" is
toward sound, the six stratal systems allowed by the theory are, from
top to bottom: gnostology, semology, lexology, morphology, phonology,
and phonetics (see Figure 1.1). Gnostology is presumed to be responsi-
ble for defining the well-formedness of texts in a language, while the

primary domain of semology is thought of as being restricted to the

 

 

r—>ZOH—IJ>Nv-or—>m:c

mz>r"U

 

Figure l.l:

(Toward Meaning)

é—TACTIC PLANES ———>

 

Gnostology

 

Semology
Lexology

L

Morphology

 

 

Phonology

Phonetics

 

\ \ \ \ \

 

 

(Toward Sound)

The organization of language in stratificational theory.

 

smaller pattern of "proposition.“ The lexology specifies the syntactic
patterns of the language in terms of clauses and phrases; this stratum
corresponds most closely to the transformational-generative level of
surface structure. Morphology handles phenomena related to word forma-
tion (although some phrasal structures may also be handled at this
level), while the phonology deals with phonemic distribution, syllable
structure, and so forth. Finally, the phonetics is presumed to account
for non-distinctive variation in the realization of phonemes. Much of
what is relegated to this level has traditionally been described as
"allophonic variation," although some of what has been treated under this
rubric is dealt with at the level of stratificational phonology.

In addition to its tactic pattern, a stratal system also entails

one or more alternation patterns; that is, areas where various alterna-

 

tions take place on the basis of conditioning from neighboring tactics.
The form of the theory presented here allows for two alternation pat-
terns, one above a specific tactics and one below it. This should be
viewed as an option within the theory, however, and not a descriptive

obligation. There may also be some alternation in sign patterns, where

 

simple units of one level are related to combinations of units on the

next.
l.4.l The Graphic Notation

Stratificational theory uses a graphic notation consisting of var-
ious nodes connected to each other by lines. The two basic types of
node are the ANQ_and QR_nodes, both of which may be either ordered or

unordered and oriented either upward or downward. Table 2 gives all the

 

 

Am \ o + av o op omngocpo
.opnpmmoo pp o op mopopog o
”mo omeoogo ocozo:

A... \ 0.3
o op mopopog u go o gosppo
"mo oopoogocz ogozo:

Am \ o av o
op mopopos o An oozo—pop a
”oz< oogoogo ogozo:

Am \ u.ov o op .goo
-go Lopzoppgoo o: :p so .xp
-maoocoppsepm opmpog u new a

”oz< oogoogoc: ogozo:

 

.0

32?:

o op omngogpo .opo
-pmmoo pp o op mopopog o
”mo oogoogo ogmzczoo

3; \ 3
o go o op mopmpog a
“mo oopoogocs oLchzoo

Ao o \ my o >3
oozoppom a op mopopog m
”oz< oopoogo ogozczoo

Ao.o \ my Loogo Lop

-zoppgoo o: op so .zpmzomcop
apzspm u com o op mopopog m
”oz< oopoogocs oLchzoo

 

 

..—<l<

.mmcpcooe epocp oco mono: mo new oz< "N mpnop

 

logical possibilities for these nodes, along with their "translations”
and equivalents in the algebraic notation which can also be used.

The AND and OR nodes occur both within the realizational and the
tactic portions of a linguistic description; another type of node, the
DIAMOND, is used to show the point of intersection of the tactics with
the realizational plane. The DIAMOND in Figure l.2 has all four of its
possible connections, but this need not always be the case; examples of
DIAMONDS having less than four connections will be given in the next
section. Other types of nodes and configurations of nodes will be ex-

plained at such time as they occur.3

connection to a higher stratum via realization
connection to a higher rank in the tactics
connection to a lower rank in the tactics
connection to a lower stratum via realization

hum—-

Figure l.2: The DIAMOND node and its possible connections.

l.4.2 The Nature and Form of a Stratum

Speech is essentially a linear phenomenon, whereas thought is not.
Part of the function of language is to enable human beings to relate

their thoughts to the linear system of actual speech; in stratificational

 

 

theory, this is accomplished in several steps, through the use of tac-
tics of varying configurations.

The higher strata, gnostology and semology, are usually thought of
as having tactics which allow closed loops, thus making the tactics a
reticulum. At the level of lexology and morphology, the tactics tend to
be more tree-like in form. At the level of phonology and phonetics, the
phonological components are arranged into linear sequences and also si-
multaneously occurring bunches. The output of the phonetic system is
still thought to be essentially linear, however. It is this gradual
progression from nonlinear to linear patterning that allows language to
do what it does: both encode and decode linguistic expression.

Further justification for the formal stratification of language
comes from the fact that native speakers can produce "systematic non-
sense" at various points of the linguistic system. For example, slurbz.
is a well-formed output of the phonotactics of English, even though it
presently has no meaning. Similarly, the famous sentence, "Colorless
green ideas sleep furiously" is well-formed at all but the semantic
level of English, at least without an appropriate context. For further
discussion of these points, see Lamb (1966a, l966b) and Lockwood (l972:
6-8).

Figure l.3 shows part of a typical morphemic stratal system and the
phenomena which may occur there; since most of these phenomena occur in
other stratal systems as well, a short discussion of each one will be
given here to help the reader understand our analysis of Amharic.

Composite realization occurs here in the morphemic sign pattern;

 

composite realization occurs whenever a unit at one level of description

 

Lexons Morphotactics
D] D]

  
 

a
I

@I

 

 

 

A

\
l

" E
"J N Morphemic
Alternation
Pattern
C
Morphemic
Sign
Pattern

Lexonic Alternation Pattern
__.~_d

 

K a

Morphemes

 

 

MorphemicI_,
Signs
A

Morphons--+

 

C: Composite Realization D]: Lexonic Diversification 2: Zero Realization
Dz: Morphemic Diversification N: Neutralization E: Empty Realization

D3: Morphemic-Sign Diversification S: Specificational Realization

P: Portmanteau Realization Tactic conditioning shown with dotted lines

Figure 1.3: Part of a morphemic stratal system. (After materials prepared by Dr.
David G. Lockwood.)

 

ll

is related to a combination of units at a lower level. The nodes used
to describe this situation are the ordered and unordered ANDs. In the
examples shown, the morphemes are realized as combinations of morphons,
which are roughly equivalent to classical morphophonemes.

Portmanteau realization, shown with an upward ordered AND, is essen-

 

tially the opposite of composite realization, in that it represents a
situation where two (or more) units are realized as one. The classic
example occurs in French, where the sequence of morphemes §_lg_is re-
alized phonemically as /o/.

Neutralization and diversification form another pair of opposite

 

 

relationships. In the former, one unit on a lower level is related to
two or more alternates on a higher level, while in the latter, one unit
of a higher level has two or more alternate realizations. Diversifica-

tion can be of two types: unconditioned alternation, which corresponds

 

to free variation, and conditioned alternation. Conditioned alternation

 

occurs when the choice of realization is determined by the tactic envi-
ronment. In the figure, the conditioning from the tactics is represent~
ed by dotted lines. The figure also shows the three possible areas of

alternation: the upper (-onic) alternation pattern, the lower (-emic)

 

 

alternation pattern, and the sign pattern.

 

 

Empty realization occurs when a tactic construction requires a cer-

 

tain realization which has no direct connection to meaning, it being
predictable from the tactic structure. An example of empty realization
occurs in French, where the /t/ in a;t;il_'Does he have?‘ is required by
the construction.

Zero realization occurs when a unit from a higher stratum fills a

 

 

 

12

position in the tactics of a lower stratum, but has no further realiza-
tion; zero realization also occurs when the tactics cannot accept an
element from a higher stratum. This situation is indicated with an
ordered OR, the rightmost line of which is a zero element, as shown in
the upper portion of Figure l.3.

Specificational realization occurs when the realization of a.

 

higher-level element is the selection of a specific tactic pattern on a
lower level, as in one of the realizations of interrogation in French

as the word order verb subject as opposed to the declarative order s39;

 

Ject verb.4

l.4.3 Phonology in Stratificational Theory

The study of phonemics within stratificational theory has received
a fair amount of attention, and there are a number of works which pre-
sent a detailed analysis of the phonemic system of a particular lan-
guage: Sullivan (l969), Rogers (1967), and Al-Azzawi (l973), for exam-
ple. In addition, there are many works which deal completely or in
part with stratificational phonology in general, or certain aspects of
it, among them Lamb (l966a, l966b), Lamb and Vanderslice (1976), Lockwood
(1972), Reich (1975, l976), and Sullivan (1975, 1976).

At this point, it is important to remember that stratificational
theory utilizes a unary component theory in phonology, and presumably in
phonetics as well.5 A component is roughly equivalent to the presence

(or absence) of a value in a binarygfeature theory, where the features

 

serve as parameters which must be given values wherever they occur.

Thus in Amharic, the phoneme /b/ will have as one of its components

 

T3

the component /Labial/, and the phoneme /t/ will have /Apical/. The
/b/ is assumed to be non-apical by virtue of the fact that its compo-
nential description includes no component /Apical/, just as the /t/ is
assumed to be non-labial by virtue of the absence of the component
/Labial/; the fact that these two components cannot occur together in
the same segment (in Amharic) is stated implicitly in the phonotactics.

The phonetic tactic pattern conditions alternate realizations of
certain phonemic components. In Amharic, the phonemes /b/ and /o/ are
both marked with the phonemic component /Labial/. The phonetic realiza-
tion of this component is different for these two phonemes, however, be-
ing /lower lip/ in the case of /b/ and /lips rounded/ in the case of
/o/. The phonetic tactics can also automatically supply certain compo-
nents. Thus /b/ is marked phonemically with the component /Voiced/, but
/t/ need not be marked /Voiceless/--the lower tactics insures that all
phonemes not overtly marked for voice will be voiceless.

The analysis presented in this thesis is based on the framework for
phonology described in Lockwood (l972), and a general familiarity with

that work on the part of the reader is presumed.6

l.4.4 Results of the Analysis

As well as providing a preliminary description of the phonology of
Amharic, this thesis demonstrates that the stratificational theory of
language can adequately account for the consonant/vowel patterning found
in Semitic languages such as Amharic; moreover, it is maintained that
stratificational theory is superior to transformational-generative

theory in this area.

 

l4

There have been several studies of Semitic languages from a genera-
tive point of view (Brame T970, Bar-Lev MS), including one on Amharic
(Bender l968).7 Unfortunately, they all have shortcomings.

Brame (1970), for example, requires vowels in the underlying forms
of verb stems in Arabic, ignoring the nearly universal agreement that it
is only the consonantal root which carries the primary semantic meaning
of a word and which serves as the base fer subsequent extension of that
meaning through the use of vocalic patterns required by other semantic
and grammatical considerations. Furthermore, Brame has a whole series
of rules which are needed to delete, rearrange, and change the under-
lying vowels in certain environments; if there were no vowels in the
underlying stem and if the correct surface vowel could be inserted
directly into its proper position, many of these rules could be elimi-
nated.

In his analysis of Modern Hebrew, Bar-Lev accepts the traditional
separation of root and vowel pattern, stating that "The consonantal root
demonstrates an integrity across intervening vowels that testifies to
its separateness as a discontinuous morpheme (MS:2)." To account for
vowel interfixation in Hebrew, he posits two rules, HOP and INSERT,

which operate as shown below (based on Bar-Lev's (lO)):

/mi-a=nzr/ /kdm=i-a/ UNDERLYING FORM
/mi=nzr=a/ /i=kdm=a/ by HOP
/mi=nzar/ /kidm=a/ by INSERT

One problem with Bar-Lev's approach can be seen here in the fact

that the final /a/ in /mi=nzar/ is inserted by the application of INSERT,

 

 

 

l5

but this rule does not apply to /i=kdm=a/ to produce */kidam/.8 How
INSERT is blocked in this instance is not stated, although there is a
possible explanation in Bar-Lev's comment that “Our rules...would hop

the single internal vowel over the root, leaving any affix where it is,
before inserting the vowel (MS:9)." The final /a/ in /kidma/ is evident-
Ty an affix, and INSERT cannot apply. Note, however, that the boundary
between the root and the final /a/ is the same in both forms before the
application of INSERT. Bar-Lev completely avoids the issue of how a
vowel is to be marked as either an affix or as an internal vowel.

It is felt that this problem stems from the obligatory linearity of
Bar-Lev's description, a trait which is also evident in Brame's work.
As yet, transformational-generative notation is unable to indicate the
morphological simultaneity of roots and interfixes. As we shall see in
more detail in Chapter II, in a stratificational treatment, the "inter-
nal vowels" of Semitic languages are related to roots with an unordered
AND, indicating morphological simultaneity. Prefixes and suffixes are
then related to the root/interfix structure by means of ordered ANDs,
indicating that they do not participate in interfixation. If this pro-
cedure is followed, a statement of the phonotactic structure of words
can then correctly account for the proper positioning of all consonants
and vowels; how this principle applies to Amharic is demonstrated in
Chapter III.

In summary, although the approaches used in Brame's and Bar-Lev's
treatments of other Semitic languages could probably be adapted to an
analysis of Amharic, it is felt that the stratificational description
presented in this thesis provides a more natural and psychologically

plausible solution to the problem of interfixation.

 

NOTES to CHAPTER I

For a detailed description of the linguistic classification of Am-
haric, its structure, and its dialects, see Bender et al. (T976).

For discussions of possible activation conventions to be used in
performance models, see Christie (T973, T976), Reich (T970), and
Dell and Reich (T977).

For a complete introduction to the notation system, see Lockwood
(T972), especially Chapter 3.

This phenomenon is discussed more fully in Lockwood (T977) and
(in press).

For an excellent discussion of the differences between unary compo-
nent phonology and binary feature phonology, see Sullivan (T969:
5-12). For a discussion of the use of components in phonetics,

see Lockwood (l973a, 1975).

There have been some recent developments in stratificational pho-
nology which pertain to the use of features which are activated
in sequence and which may stay activated for varying lengths of
time. For details, see Reich (1975, 1976).

This work has only recently been available to me. A cursory in-
spection indicates that although Bender lists roots of a purely
consonantal nature in the lexicon, there is a phonological rule
which automatically inserts vowels between all consonants before
the application of any other rules. Insofar as this is the case,
it opens his analysis up to the criticism of rule superfluity as
applied to Brame's work.

Bar-Lev's INSERT is a mirror-image rule; witness its application
in the following example (MS:6):

/chv=a-a-t/
/a=chv=a-t/ by HOP
/cahav=t/ by INSERT.

l6

 

CHAPTER II

2 Some Aspects of Amharic Morphotactic Structure
2.1 Traditional Analyses of Amharic Morphology

There are a number of older works which treat the morphology of
Amharic at length, among them such works as Praetorius (1879), Isenberg
(1842), Cohen (1936), and Leslau (1968). Although Leslau's text is
fairly recent, his approach to morphology is not substantially different
from that of the other cited authors, and thus it may be included among
the traditional analyses.

These works do not explicitly treat the roots and interfixes as
separate entities, even though recurrence of the radical consonants of
the roots in various related words was noticed and commented upon. Much
of the description is philological in nature, and the following excerpt
from Cohen may be taken as representative of the tone of these tradition-

al studies:

D'autre part la constitution des themes ou radicaux, au
départ dl une forme fondamentale ou theme simple pour chaque
verbe, comporte des modifications diverses a valeur morpholo-
gique: alternances vocaliques, géminations et repetitions des
consonnes radicales, adjonctions de prefixes. Le theme simple
lui- -meme se présente sous des formes variées qui relevent sur-
tout de 1' etude lexicale et ne correspondent pas a des caté-
gories grammaticales actuelles; cependant certains aSpects des
radicaux sont dus a des categories morphologi ues historique-
ment connues et se survivant partiellement. ICohen T936: 157)

17

 

 

18

What usually follows such comments are lists of words illustrating the

particular point under discussion by the author.
2.2 Modern Approaches to Amharic Morphology

Unfortunately, this writer has not had access to any treatment of
Amharic morphology within the framework of a modern linguistic theory,
with the exception of a short stratificational analysis of some verbal
data given in Lockwood (1972:96, 97; 219, 220).1 There were available
two Ph.D. dissertations dealing with Arabic, a typologically similar
language. The first of these, Brame (1970), is a transformational ap-
proach, while the approach of Al-Azzawi (1973) is stratificational. In
addition, a transformational analysis of some aspects of Modern Hebrew

has been provided in the form of a paper by Bar-Lev (MS).

2.2.1 Transformational Approaches to Semitic Word—Formation

Brame's work suffers from a dearth of information concerning word-
formation in Arabic, a lack unfortunately characteristic of transforma-
tional-generative descriptions, at least until recently. In all of his
examples of phonological derivations, he begins with an underlying form
which contains both consonantal and vocalic elements. A typical example
of his approach is the following:

Suppose further that the passive morphology is created by the

following rule:

2) Passive Formation: V + i / ___C] / [+passive]

We assume that the feature [+passive] is assigned to active

stems by the passive, syntactic transformation. This rule

takes active CaCVC, which we assume is specified [+passive],
into CaCiC. We now add the additional rule of Quasi-Assimilation

 

 

l9

to account for the g_of form I perfectives.
3) Quasi-Assimilation: a -+ u / ___CiC / [+passive]

This rule takes CaCiC on to CuCiC as desired. Thus, by
these two rules katab may be correctly turned to passive

kutib, accounting for class I perfective passives. (Brame
1970:438)

 

At no point in his discussion does he propose the separation of the ra-
dicals from possible vocalic interfixes.2 He does, however, propose an

infixation rule, his 7):
7) wa-Infixation: wa [CVX + [CVwaX (Brame 1970:419)

This rule imposes an obligatory, arbitrary linearization concerning wa_
and the rest of the word, an assumption that seems to stem from no other

source than the nature of the formal rules required by transformational

theory.3

Bar-Lev (MS) uses syntactic evidence to justify the linearization
of vowel patterns with respect to the root. He states, “For intramor-
phemes [vowel patterns] with an affix, there is obviously motivation for
assuming that the internal vowels originate as part of the affix (MS:5)."
If there are words with internal vowels which always appear with both
a prefix and a suffix, however, this tack would seem to allow for the
arbitrary decision to order such internal vowels either in front with
the prefix or behind with the suffix. Consider the form /yi=§mor=u/
'they will guard,I which Bar-Lev analyzes as /yi-u=§mor/; could it not
just as easily be /§mor=yi—u/? In either case, the application of his

mirror-image HOP rule will produce the proper result.4

 

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIuuu.u..........-::————e————
20

2.2.2 Additional Comments on Semitic Word-Formation in Transformational-
Generative Theory

In a recent work, Aronoff (1976) makes some remarks relevant to the
problem of interfixation, cryptically mentioning the Semitic languages

in passing:

4.3.1.3. Consequences. We have found two classes of rules
which are Best viewed as NFRs [Word Formation Rules] in a
particular manner, namely as transformations. This is dif-
ferent from the system using labeled frames mainly in that
it forces us to divide the rule into two parts, a structural
description and a structural change. The first part speci-
fies only the base. The second part contains the base and
the result of the operation of the MFR, amalgamated into
one unit. The formal nature of the bifurcation has an in-
tuitive counterpart: the base is an independent entity,
which we know already, for in order to qualify as a base

it must be an independently occurring word and a member of
a major lexical category: the affix (which in most cases
is equivalent to the affixing Operation) cannot be sepa-
rated from the rule, because it is nowhere given any repre-
sentation of its own. This intuitive counterpart is very
different from the view which pe0ple normally have of af-
fixes: namely that they are independently existing en—
tities; that they are morphemes, just like stems, and have
all the properties stems have.

This view has led to many problems, of which I will
mention only the two most commonly discussed. The first
problem is that of the discontinuous morphemes, like the
Semitic vowel patterns. Though I cannot claim to have
solved all the mysteries of Semitic morphology, it is clear
that once we stop thinking of these vowel patterns as items
of the same sort as the stems, we can stop worrying about the
metaphysical import of these discontinuous patterns and be—

in to develop a framework within which they can be studied.
IAronoff 1976:70)

 

Even six years after Brame's dissertation, the problem of Semitic inter-
fixation seems to have received no satisfactory treatment within trans-
formational-generative theory.

As Aronoff points out, the factor that prevents the extraction of

a completely consonantal root which is then modified with various vowel

 

 

21

patterns is the contention that the base of a WFR must be an indepen-
dently occurring word. The fact that Amharic has no word of the form
[mls] would evidently rule out the recognition of a base /mls/ ‘answer'
despite the similarities in meaning in such words as m311asa ‘he an-
swered,l m§l§§_'answeri' (sg. masc. imp.), malallasfi 'he answered re-
peatedly,‘ and many more.

The failure to provide for the representation of interfixes as si-
multaneous with the root at the level of morphology and the requirement
that stems have underlying vowels (at least in some treatments) are seen
as the two main problems in transformational-generative theory with
respect to Semitic word-formation. In the next section we will see how
a stratificational analysis which avoids these problems can be applied

to Amharic.
2.2.3 Stratificational Approaches to Semitic Word-Formation

Both Lockwood's treatment of Amharic and Al-Azzawi's of Arabic em-
ploy a similar strategy at the morphotactic level, one which involves
the use of both ordered and unordered ANDs in the tactics, the unordered
ANDs being used to provide for the morphological simultaneity of roots
and interfixes. Figure 2.1 illustrates the application of this tech-
nique to the data presented in Table l, which is here presented again
for convenience as Table 3.

To help the reader familiarize himself with the network notation,
this figure will now be explained in detail. The OR node at T represents
a class of four different morphotactic constructions. If the path lead-

ing to 2 is taken, the root MLS 'answer' occurs simultaneously with the

 

 

22

 

MLS

Figure 2.1: A fragment of the Amharic morphotactics.

Table 3: Words based on the root ML§_'answer.‘

Amharic Gloss

MELLaS-a 'he answered'

MHLS 'answer' (n.)

MaLS-at '(she) having answered'
al-MaLS-am 'I do/will not answer'

MeLaS 'answerl' (sg. masc. imp.)

23

interfix §§_'second person singular masculine imperative.‘ The ordered
AND at 3 shows the occurrence of the suffix §_'third person singular
masculine' with the construction relating the root to an interfix §g§_
'perfect tense.‘ This construction is represented at 7. The AND
labeled 4 accounts for the negative construction, and 5 accounts for two
possibilities; the small circle at 6 indicates Optionality, allowing the
construction at 9 to occur either with or without the suffix at_'third
person singular féminine.‘ The upward OR at 8 shows that the structure
of 9 can be part of two different constructions at a higher rank in the
tactics. Finally, the fact that the root_ML§ may occur with three dif-
ferent interfixes is captured by the upward OR at 10.

It should be mentioned that for this particular analysis, the inter-
fix §£_has connections to three meanings: 'gerundive,‘ 'nominal,‘ and
'imperfect tense.I Disambiguation takes place at a higher level of the
grammar, where only 'gerundive' can co-occur with 'third person singular
feminine' at, 'imperfect tense' with 'negative' a13..§m, and so on.

Information pertaining to the relative order of the consonants of
a root and the vowels of an interfix is provided by the morphemic sign
pattern; the sign pattern for the Amharic data treated here is shown in
Figure 2.2. Note that the radicals of the root and the vowels of an in—
terfix always occur in the same order. The phonotactics can then be
constructed in such a way that it can use this information concerning
order to correctly position the consonants and vowels of a word. This
is done by arranging the phonotactics to generate syllables of the form
CV, with an Optional consonant at the end of a word. Such a tactics is

depicted in Figure 2.3.

 

 

 

24
MLS as a# a 3:3 a1 am at
A A A A
v \ \ ' I
\‘ ’
S2!
"' \‘T/
s m l # a e a t

 

Figure 2.2: The morphemic sign pattern for the Amharic data in Table
3.

 

 

 

 

 

 

 

Note: Details concerning length (:) have been omitted for simplicity.

Figure 2.3: Phonotactics for Amharic interfixation.

 

25

Nhen the phonotactics receives simultaneous input from the morpho-
tactics, as in the case of a root occurring with an interfix, it must
process a consonant first, then a vowel. The first consonant will be
the first radical of the verbal root, followed by the first vowel of the
interfix, then the second radical of the root, followed by the second
vowel of the interfix, and so on. In certain cases where a cluster of
consonants occurs on the surface, a special morphon must be postulated;
this morphon, f3 fills the vowel position of the syllable, but has no
further realization. A more detailed phonotactics of Amharic will be
presented in Chapter III.

Now that this brief introduction to Amharic morphology has been
presented, we can see how the structure of nominals, more specifically,

derived nouns, can be handled by stratificational theory.
2.3 A Stratificational Treatment of Some Amharic Nominals
2.3.1 The General Structure of Nominals in Amharic

In Amharic, as in English and many other languages, reference to
the syntactic environment of a word may be necessary to categorize it
as either a "noun" or a ”verb." For example, the English word seat

functions as a noun in (1a) but as a verb in (1b):

(la) Please take a seat over there while you wait.

(lb) Please seat yourself over there while you wait.

Nominal constructions in Amharic may be formed from stems which are
essentially nominal, verbal, or adjectival in character; examples of

such constructions are given in Table 4. In this table, hyphens indi-

 

26

.Aocov 3o: >E.e
.moepxeeee_t

.Ammcov 3o: ocp.
.Aocov 3o: o;p_
.Aocov 3o: mcp.

.Aocov zoo.

.mcpxomgo as.
.mcpxooen as.
.mepxeeen esp.
.oepxeeee esp.

.mepxeeee e\eeeee Op.

.mmmso; x5.
.momao; as.
.mmzo; >5.
.omzo; x5.
.momzo; asp.
.momzo; o;p_
.momsos.
.mmzoc 9.3..
.meO: wsv.
.omoon.

mmopo

m>Hmmmmmoo
4<m24m

m>HH<m3uu< mpuphm< 4<m34o
w>HH<mzuu< mpu~pm<
mpuHem<

m>Hh<mzuo< m>Hmmmmmoa
m>~mmmmmoa

m>~p<m=ou< mpuHhm<
mpuHPm<

m>HH<mzuu< m>Hmmmmmom 4<m24m
m>Hmmmmmoo 4<m24o

m>-<m=uo< m>~mmmmmoo
m>Hmmmmmoo

m>~k<msou< mpuHhm< 4<m24o
upoHHm< 4<m24a

4<m24m

m>HH<m2uu< mpoHHm<

mpqum<

Zum
ZmHm

Suhm
Empm
Enhm
Suhm

ZmHm
ZmHm
ZmFm
ZMHm
ZmHm

Suhm
Zupm
Zum
Empm
Suhm
Zum
ZmHm
ZMHm
Emem
Suhm

 

ogzpozgpm oppooposmuoz

e-mpeeee
neo-ememwe.

e-=-ooo-mpeea
cuzumpoom
=-mpeee

weave

cuoueMomwe
ougwomms
c-31emomws
:1Lmomws
gnomme

ziouwwoupma
o-wwo-poo
:-o-pon
oupoo
cuauwwoupoo
=-ooo-pee
”woupoo
cnzupoo
:upmo

poo

opsoge<

.opgoss< :p mcopposgpmcoo poopsoc po moposoxm no opnmp

27

cate morpheme boundaries, and the morpheme classes to which these mor-
phemes belong are shown in the middle column.
It is important to note that not all stems can occur with all suf-

fixes, as the forms *masbar-ott and *addis-e illustrate. To specifical-

 

ly exclude these forms from the morphotactics of Amharic would entail a
great deal of complication, however, as well as obscure the fact that in

other respects the stems masbar and addis behave precisely as the "reg-

 

ular" stem bet, The non-occurrence of the starred forms, and others
based on them, is better treated at other levels of the grammar, speci-
fically the semology and lexology. As mentioned above, this will lead
to a greater generalization of the morphotactics and a simpler statement
of the relationships involved between stems and suffixes.5

The general morphotactic structure of nominals is presented in
Figure 2.4, which shows that a nominal consists of a STEM, followed by
an optional plural marker (PL), an optional determiner (DET), and an op-
tional accusative marker (ACC), in that order. Here, the determiner
class subsumes both possessive suffixes and the definite article, which

never CO-OCCUY‘.

(nominal)

 

Figure 2.4: The general morphotactic structure of nominals.

 

28

The most usual realization of the plural morpheme is géé; the 9.
of this suffix is actually the realization of two different morphons,
MN/o/ and MN/o+/, the latter being a preemptive.6 Justification for it
is found in the plural of words ending in vowels, where the plural form

may occur with or without the final vowel, as in lebaott/leboté 'thieves'

 

from lgbg_'thief.‘

There are also a number of irregular plurals, most of which were
borrowed from Ge'ez. In some cases, both the Ge‘ez and the Amharic
plural are found in alternation, as in the plural of w§r_‘month,‘ which
is either w§rg§§_or w§rat. In other cases, the Ge'ez plural has a

slightly different meaning from the Amharic plural, as in wandemmotc

 

'brothers‘ but wandammamatt ‘brothers to each other,‘ both of which are
from the singular wandemmo 'brother.‘7

As mentioned above, the determiner class includes both possessives
and the definite article. This class conditions the realization of the
accusative ending as n, it otherwise being realized as zero. Compare

(2a) with (2b) and (2c):

(2a) bet ayya
house he-saw
'He saw a house.‘
-(2b) betun ayya
the-house he-saw
‘He saw the house.‘
-(2c) beten ayya
my-house he-saw

'He saw my house.‘

 

29

Having briefly discussed the general structure of nominal con-
structions, we may now continue to a more detailed account of stem-

formation in Amharic.
2.3.2 The Morphotactic Structure of Nv-Stems

As discussed previously, nominal stems in Amharic may have a verbal,
nominal, or adjectival character. The characterization is based pri-
marily on our intuitions concerning the semantic properties of the
various root morphemes. Thus we view bete_‘my house‘ from the root bet_
'house' as being semantically noun-like, but masbare 'my breaking‘ from
the root §§R_‘break' as being semantically verb-like. At the morpho-
tactic level, however, such distinctions become blurred and the dif-
ferent stems function alike, whatever their semantic properties.

To avoid confusion, I will make a distinction between nominal
stems based on the semantic character of their roots; we then have
nominal stems based on roots whose semantic properties are noun—like,

adjective-like, and verb-like, designated Nn-stems, N -stems, and Nv'

a
stems, respectively.

In some cases, there can be more than one "layer" of structure for
a given stem. For example, the Na-stemIQgrd 'cold (one)‘ is related to
an adjectival infix and to a root BBQ 'cold/be cold.‘ This root can
also be used to form morphological verbs, as in yabardal ‘it is/will be
cold.‘ This example is shown in network form in Figure 2.5, which also
shows a noncomplex morpheme ggng9_‘beautiful,‘ for which the stem is

the root as well. The same situation occurs with the Nn-stems.

In this thesis we are concerned with the morphotactic structure of

 

30

L/cold (one)/ L/cold/ L/beautiful/

A

  
  
  

(Na-stems)

l

' (to verbal

M
qon O o
/ . Q / morphol QY)

 

M/e/ . M/BRD/ .

Figure 2.5: The structure of a complex and a noncomplex Na-stem.

nominal stems whose roots are also used to form verbal stems. For pur-
poses of convenience, such stems have been labeled NV-stems, and the
particular stems which we will treat here have traditionally been called

the infinitive (or verbal noun), the instrumental, and the participle.

 

 

 

 

Table 5 presents these forms for verbs from eight different classes of

roots in Amharic, a representative but not exhaustive sampling.8
2.3.2.1 The Infinitive

The Amharic infinitive is glossed in English as either an infinitive
or gerund. Syntactically, the infinitive construction can function as
the subject or object of a clause. The lexeme L/Infinitive/ is simulta-

neous with the root at the lexotactic level, but is realized in two

 

31

 

.mopzp poocomcoo asp mopppez
mo ooppoopo meLo: ogp mcpzoppop we omopmcp pcmcomcoo moop m opoopocp op mew; oom: mp n capo mew .m.z

 

 

 

 

 

 

ppmm ewpmmwe pmopwmms .oeseeee on. mom HHH>
memN ozmemNmE pomLmNms .p:o sopoepm. mmN HH>
”mace emmxeee amazes .eEap we. m¥z< H>
peepeeme espempeememe ampeememe .x_e. os<mo m
peeemmme expemxmmeme emxmmsme .peppmep_ mxmz < >
pnsmp e »:_:e pm smpme .epeeeeem. ”>3 m
Tame espeaze pmeeme .pmem. mo < >H
pnewp map ewpwe pa empme .eezmeee. ”on m
p mm espEmme peEmms .eeee. zm < pop
map; eneeme eeeme .om. mm: o
psozo oApEoomE Eoowe .ocopm. zoo m
psom expEommE Somme .mmpx. z<m < Hp
peepeo espempemme empemme .eo__em. mo<o u
monoo ommnoos mmnoms .3ocog. muo< a
peepe espempee empee .»3 when. ao< o
pom .me espmmnpmeme mm pmeme .eeez. muse m
peeemm expememme ememme .eeeee_ mmm < H
opapopppoo popcmssgpmcp o>pppcpecH mmpcooz upmom poom mmxh mmo—u

.msopmu>z opgogs< so open um mpnoh

32

different morphotactic positions. The morpheme M/Inf]/ occurs as a pre-
fix and the morpheme M/Inf2/ as a simulfix.9 A bistratal description
of the infinitive is given in Figure 2.6.

Examining the infinitive forms given in Table 5, we see that for
the paired groups IA and IB, IIIA and IIIB, and IVA and IVB, the
first radical consonant of the type A roots is immediately followed
by the second radical. For type B roots, however, the first and second
radicals are separated by the vowel §, To account for this, we may pos-
tulate a morphon MN/A/ which is realized as fi_everywhere except before
a long consonant--remember that this fi_fills the vowel position in the
phonotactics but has no further realization, thus allowing a cluster of
consonants to appear.

This solution is satisfying for two reasons: first, there has been
a traditional distinction between type A verbs and type B verbs; type B
verbs always have a long second radical, whereas the second radical of
type A verbs is sometimes long, sometimes not (Leslau 1968:104). Second-
ly, by treating the fi/fi_alternation as phonologically conditioned, we do
not have to set up separate classes in the morphology to condition it.
This would be very costly in terms of the stratificational simplicity
metric since it would mean adding more nodes to the description.10 The
treatment involving a phonologically-conditioned alternation will require
the addition of nodes, too, but not as many as would be required in the
case of morphologically—conditioned alternation. Thus is appears that
the use of a morphon MN/A/ provides the simplest solution.

Next, if we observe the roots in classes III and IV, we see that

the infinitive form ends in a t_which is not present for roots in class

 

33

“Lexotactics

A

 

L . . .
L/Rootl. /Inf1n1.t1ve/ .

A

 

Morphotactics

A

 

 

A
M/Root/ . M/Ich/ .

Figure 2.6: The lexo- and morphotactic structure of the infinitive.

34

’ 'MOY‘phOtaCthS

 

M/Inf

 

 

 

 

 

Figure 2.7: The realization of M/Inf2/ for roots in classes I, III, and
IV.

I infinitives. Compare a class I infinitivemggpgp ‘to break' with a
class IV infinitive p§g§§t"to stay.‘ Both of these have A_and §_ as
the first and second vowels of the interfix, but the root QR_‘stay' con—
ditions a §_after the second vowel of the interfix. Now both class III
roots and class IV roots condition this 3, but class III roots condition
the realization of the second vowel of the interfix as p, not §, The
network which accounts for the realization of M/Inf2/ for roots in
classes I, III, and IV is presented in Figure 2.7.H

A few words of explanation concerning the figure will be given.
First, for all the classes under consideration here, the first vowel of

the interfix is always 5, The second vowel is either p_or g, with §_

 

35

being the most common--or unmarked--choice. The final position of the
interfix is usually empty, but the occurrence of a root in either class
III or IV will condition the realization of a §_in that position.12 The
fact that the conditioning AND for this §_is higher in the tactics than
the the OR node which leads down to classes III and IV obviates the need
to use separate conditioning ANDs for these two classes. Thus the gen-
eralization that both classes III and IV condition the 3 can be stated
with one node in the tactics. The use of one node instead of two also
effects a simpler description according to the simplicity metric.

Type C and D roots of class I are different from the other class I
roots in that they have an initial p_which conditions the zero realiza-
tion of the §_in the infinitive prefix p§_(M/Inf1/). Examining the in-
finitive forms of roots of types C, D, and E, we see that an 3 occurs
where we would expect either §_or a consonant cluster. The participle
of the root EAL§.IS gplgpi, which looks very similar to the participle

of the root SBR, sabari. For the type E root, however, the first vowel

 

is p_and not §, In other respects, though, it seems to be a triradical
verb, so the infinitive form ought to be m§g1§p; this is incorrect, so
we must somehow account for the occurrence of p_in the correct form,
magalfib.

The best solution to the problems posed by roots of types C, D, and
E hinges upon the use of a preemptive morphon MN/a+/. In the phonology,
this vowel will condition the zero realization of an §_which precedes
it and will also condition the zero realization of any vowel which
occurs simultaneously with it. The details of this will be presented

in Sections 3.5.4 and 3.8.2 of Chapter III. If we treat type C, D, and E

 

36

roots as having the preemptive morphon as an inherent part of the root,
we need not make any alterations in the realization of M/Inf2/ as shown
in Figure 2.7.

We do need to make a slight revision in our analysis to deal with
roots in classes 11, V, VI, VII, and VIII, however. For class II roots,
we need to provide for the realization of the infinitive interfix as
zero, and for roots in classes V, VI, VII, and VIII, we need to allow
them to condition an extra §_at the beginning of the morphemic sign for
M/Inf2/. The complete realization of the infinitive interfix is given

in Figure 2.8.13

2.3.2.2 The Instrumental

The instrumental form of a root has the basic meaning of the instru-

ment or means by which the action of the root is accomplished, or the

V,VI, ___
VII, VIII

3 A a a t

 

 

Figure 2.8: The realization of M/Inf2/.

37

time or place the action of the root happens (Leslau:366, 367). So we

have mfitrfigiya 'broom' from IRG 'sweep,"m3dr3§a ‘time or place of ar-

rival' from QRS-‘arrive,"m3§3m:3rty ,‘beginning' from'GMzR 'begin,‘

 

and so on. Syntactically, the instrumental form behaves as a noun.

The instrumental form is related fairly straight-forwardly to the
infinitive; Figure 2.9 shows the basic patterns involved. Here, the
lexeme L/Instrumental/ is realized by the same morphemes as L/Infinitive/
plus a suffix, M/Inst/. The use of the upward OR in the realizational
portion allows us to capture the generalization that the instrumental
form is related to the infinitive form without having to repeat the
same information with more nodes; it would be possible to use upward ORs
directly above M/Inf]/ and M/Inf2/ and relate them to the downward AND
of the instrumental, but that would increase the complexity of the gram-
mar. It is simplest to exploit the information contained in the downward
AND of the infinitive through the use of the upward OR as shown. The
use of the optionality element in the morphotactics allows us to use the
same basic tactic construction to account for either the infinitive or
the instrumental by making the instrumental suffix M/Inst/ optional.

If we examine the instrumental forms of classes III, IV, VII, and
VIII, we see that the final §£_or gt_of the infinitive does not occur:

p§§pp£_‘to hear‘ but masmiya ‘instrument for hearing, mazfirgat 'to
stretch out' but mazargiya ‘instrument for stretching out,‘ and so forth.
The easiest way of handling this is to have M/Inst/ condition the zero
realization of part of the morphemic sign for M/Inf2/; this is shown in
Figure 2.10.

The final thing we need to treat in conjunction with the instru-

38

Lexotactics

 

  
 
 

I-/Root/ . 9

¢ l'/Instrumental/

A

A

Morphotactics

 

 

A
/..
M/Inf1/ . f . M/Inst/

M/Root/ . . M/Infz/

Figure 2.9: The lexo- and morphotactic structure of the instrumental.

39

III, IV,
1' p ,’ VII, VIII

0 ..
V, VI, "‘ M/Inst/
VII, VIII I O

 

 

Figure 2.10: Revised analysis of M/Inf2/ with conditioning by M/Inst/.

mental is the consonant alternation which occurs in the last radical of
a root which is in the instrumental form. We postulate a special vowel
morphon MN/ip/ which conditions a number of morphophonemic alternations
involving “plain" consonants and their palatal counterparts. The "regu-
lar" morphon MN/i/ does not condition these alternations. Compare
tiyatar ‘play‘ with makffita ‘key' (from the root 551 'open‘); the mor-
phonic representation of the first is MN/tiater/, of the second
MN/m'aikAfa‘tipaL The details of this phenomenon are given in Sections
3.6 and 3.7 in Chapter III. Also, note that the y_which occurs in the

transcription can be treated as a purely phonetic phenomenon; thus the

resulting morphemic sign for M/Instr/ is MS/ipa/.

40
2.3.2.3 The Participle

The participle has syntactic properties which are both noun-like
and adjective-like. For example, the participle of the root LM:N 'beg'

has an adjectival character in (3a) but a nominal character in (3b):

(3a) and lfimppfi. lag ayya
a begging child he-saw
'He saw a begging child.I
(3b) legu M flaw
the-child beggar he-is

'The child is a beggar.I

The Amharic use of the participle is very much like the English use of
one noun to modify another, as in ‘a beggar child,‘ where pggggry-a noun
in its own right--is used to modify Epilg,

In all cases, the participle form ends with the morphon MN/ip/,
which palatalizes the preceding radical if possible. In most cases, the
vowel after the next-to—the-last radical is 3, but there are exceptions.
Note particularly the forms of the participle for type B and C roots in
class II: gflpmi_'standing/one who stands‘ from Q9M_'stand' and pip§_
'going/one who goes' from HEQ_'go.' How these forms are accounted for
through the use of special morphons and a complex syllable nucleus will
be examined in detail in Section 3.3 of Chapter III. Figure 2.11 shows

the complete realization of the participle morpheme M/Part/.

 

41

M/Part/ O

A

A

‘

®
.493

‘

Figure 2.11: The realization of M/Part/.

2.3.2.4 The Integrated Description

Figure 2.12 integrates the descriptions of the infinitive, instru-
mental, and participle forms into a complete network showing both the
realizational and morphotactic relationships involved.

Inspection of the diagram reveals that there are two principal
morphotactic constructions, one of which accounts for the participle
forms and the other of which accounts for both the infinitive and the
instrumental forms. Both of these tactic constructions have a root/
simulfix relation, but the structure for the infinitive and instrumental
forms involves a prefix and a suffix as well.

As mentioned in Section 2.3.2.2, the lexeme L/Instrumental/ shares

112'

.meLoe epepoppeee oee .pepee53epmep .espppepeep esp co coppepeomee oppeepoeeeos oepeeoepep esp HN_.N eeempa

 

o 4 I...-..--:.-..®
0 -- - .

 

I

 

I

. _ ® .

‘ @ l..l“.rll.®
4 4 ...© A- @ u:©

9,, Lo 4...@ 4

4 :o

4

 

 

7—4

Ampoomv

O

9 3.5%: a: 5:5,. 0 .5

\emep\z
_
,\ .222

\opopoppeme\o \popcosaepmcp\4

 

 

AmEmpmu>zv

 

moppoopogmeoz

 

 

\e>popepeep\s

 

 

43

as part of its realization the same structure that realizes the lexeme
L/Infinitive/; in Figure 2.12 this is indicated by the upward OR in
the realizational portion immediately below the ordered AND of the in-
strumental.

Another instance of shared structure is illustrated by the inte-
gration of part of the realization of M/Part/ with that of M/Inf2/.

This is possible because for both of these morphemes the conditioning
factors and the realizations caused by them are identical: class II
roots condition a zero realization. The use of the upward OR allows us
to state this conditioning once instead of repeating it separately for
both morphemes.

The diagram also indicates that regardless of what root occurs,
the morphon MN/A/ is always part of the overt realization of M/Inf2/ and the
morphon MN/ip/ is always part of the realization of M/Part/. This is
indicated by the lack of any conditioned alternation of the lines that
lead from the ordered ANDs to these morphons.

The use of unordered ANDs to state conditioning environments in
the tactics allows us to state these environments once instead of having
to separately mark each class which conditions a certain realization;
all root classes which occur at a tactically Tower rank than such an AND
condition the same alternations.

Finally, the use of ordered ORs allows us to utilize the concept of
markedness in our description. The right-hand line always leads to the
unmarked choice in any given instance.14

The treatment of morphological data is now complete, and we may

proceed to the description of the Amharic phonotactics.

NOTES to CHAPTER II

Regretably,, Bender‘s dissertation on Amharic, done within a trans-
formational framework, became available too late for a discussion
of it to be included in this thesis.

The inclusion of vowels in the underlying form of the root runs
counter to most analyses of the phenomenon done within other
frameworks, including those of "traditional" nature. See Bender,
Bowen, Cooper, and Ferguson (1976:73, 82-84), Matthews (1974:131,
132), and Gleason (1961:72, 73) for example.

It might be claimed that transformational-generative theory (or

at least Brame) is not making a claim of literal order here. As
Schane writes in a section of his book appropriately titled "No-
tational Conventions Express Linguistic Generalizations": "The
notational conventions employed in phonology are not to be re-
garded as abbreviatory tricks for saving space when writing rules.
They are intended to capture relevant aspects of phonological pro-
cesses (1973:73)." Either Brame is making a claim that in some
significant sense the wa is ordered before the stem at some level--
in which case his use 6? the convention of linear order is correct--
or he is not making such a c1aim--in which case the notation sys-
tem is faulty in the sense that it has no way of indicating the
simultaneity of the infix and the stem.

If the /o/ in this example is analyzed as an "internal vowel" in
Bar-Lev‘s approach, we would evidently be forced to choose between
/yi-o-u=§mr/ and /§mr=yi-o-u/ as underlying forms of /yi=§mor=u/.
Bar-Lev does not indicate how such a choice would be made or how
the proper rules would then apply to produce the surface form.

Compare Sullivan's (1977) discussion of English morphOTOQY; in
brief, in setting up a morphotactics to account for such words

as reduce, reform, transduce, transform, addUCe, and so on, the
morphotactics accounts for the form *adform, even though it is
not presently in use. To specifically exclude it would create a
more complex tactics. This is essentially the reasoning employed
in describing the Amharic morphotactics here.

 

 

For a more complete discussion of the preemptive within stratifica-
tional theory, see Lockwood (1973b).

Occasionally double plurals occur, such aS‘wandemmamattott; such
forms can be handled relatively straightforwardly,'but are not

 

44

10.

ll.

12.

13.

14.

45
germane to any points made here. Their description is therefore
left unformalized.

In this classification, a class indicates a group of roots which
condition certain alternations in the realizations of various mor-

[phemes which occur with them; type refers to the particular phono-

logical shapes that occur within each class.

Simulfix is a generic term for infixes, interfixes, and superfixes.
For a more detailed discuSsion, see Lockwood (1972:95, 96; 218, 219).

See Lockwood (1972:58, 59) for a detailed description of the sim-
plicity metric now used by stratificational linguists.

In this and subsequent figures, the following notational equiva-
lencies hold:

7X 78 i” :4;

a b c

(Where 2 is a class label and a, b, c, and so on are members of Z.)

It might be hypothesized that the final 3t and at which occur in
the infinitive form of class III and IV roots are really suffixes;
this approach was tried and rejected on the grounds that it pro-
duced a more complex description. Treating the t_and the preceding
vowel as part of the interfix does not necessitate any change in
the phonotactics, moreover, so this analysis will be accepted here.

Notice that roots in classes VII and VIII are triradical roots,
even though they condition a t in the realization of M/Inf2 / in
the same manner as biradical roots in classes III and IV. 2In a
transformational- -generative treatment of this data, it is likely
that such roots would have to be marked in the lexicon as condi-
tioning this 3, since its presence or absence cannot be predicted
by the number of radicals in the root. That would mean, however,
that each root would have to be separately marked; in a stratifi-
cational treatment we may mark the conditioning in the tactics,
which serves the same purpose as marking each "exceptional" root
separately, but which eliminates the redundancies involved in such
fin approach. For further discussion of the concept, see Sullivan
1977 .

A preliminary investigation of Leslau (1968) and Cohen (1936) in-
dicates that the more highly marked a root class is according to
this analysis, the fewer members it has. Clearly more work in this
area needs to be done, but this initial discovery is suggestive of
a pattern in the language.

 

CHAPTER III

3 Aspects of the Phonology of Amharic
3.1 Introduction

This chapter will present an overview of the phonology of Amharic;
among the t0pics treated are phonotactic phenomena such as the structure
of the syllable and phonological word, the realization of phonemes in
terms of components, and severalmorphophonemic alternations, including
the palatalization phenomenon mentioned in Sections 2.3.2.2 and 2.3.2.3
of Chapter II and the realization of the participles of type B and C

roots in class II through the use of a complex syllable nucleus.
3.2 Previous Analyses of Amharic Phonology
3.2.1 Traditional Approaches to the Phonology of Amharic

Most of the early grammars of Amharic, such as Praetorius (1879)
and Isenberg (1842), approached the phonology of the language in terms
of orthography, and spoke of phonological alternations in terms of
changes in spelling. It was not until Cohen (1936) that anything like
a modern phonemic description was given. Cohen talks of a "systeme con-
sonantique" (p. 30) and gives a more or less traditional presentation
in terms of "sounds" and their a110phones. The following example is

typical:

46

 

47

Le p, occulusive sonore, est articulé (ou fait l‘effet
d‘étre articulé) comme un p_francais, uand il est solide.
Il est toujours solide lorsqu‘il est 9 miné. En dehors de
ce cas, il est articulé le plus souvent d‘une maniére léche;
souvent, surtout entre deux voyelles, Te reléchement est tel
qu‘il ne se produit pas d‘occlusion; on entend alors la spi-
rante bilabiale sonore p, (p. 31)
Leslau (1968) presents a learner‘s introduction to the phonemic system
of the language, a description which can serve as our starting point

here.
3.2.2 Leslau‘s Phonemic Analysis of Amharic

Table 6 lists the phonemes in Leslau‘s analysis according to place
and manner of articulation. Leslau augments the discussion of the pho-
nemes with a few notes on allophonic variation, including the labializa-

tion and palatalization of consonants before rounded and front vowels,
respectively.l In addition, he mentions the elision and epenthesis of
§_in certain positions.

It is inmortant to note that Leslau does not include the glottal
stOp in his phonemic inventory, although he mentions it in his discussion
of the Amharic sound system. He does, however, treat a labiovelar
series--kfi, g!) and g!;-as phonemically distinct from the plain velars.
He also accepts the phonemic status of the vowel 3, The status of 3, 3,
and the labiovelars has been disputed, and in the following sections we
will examine this controversy in more detail. Let us begin with the is-

sue of the labiovelars.

 

48

33
3o
3x

poppopm Lopo>opoop

050'

empm>

3op
o ope

O
(56

3

(D
‘l-
.C

U)
'1—
.:

xomo poepcoo pcoep

A 3 Aooopo>v
L Aooopo>

v
_ Aooopo>v
v

>C
C
E

Aooopo>

eeNppeppopm
ooopo>
mmopoopo>

X.) >03”.

ee~p_eppe_m
ooopo>
mmopoopo>

M>N
mNm'
q.

ee~p_eepo_m
ooopo>
mmopoopo>

4410+»
anon.-

pepepee pepeee pepeep

.Awompv :mpmmp op mcpogoooo upgozs< wo mmamcoso ash

H ”spasm;

mmpmmmm
mpo3o>psom
gape
pagopmp

mpmmoc

mopmopemem

mo>ppmopep

moopm

 

”mpcwcomcoo

no epoch

49
.3.3 The Phonemic Status of the Labiovelars

In Amharic, it is possible to find minimal pairs which give evi-
dence of a contrast between plain velar stops and labiovelar stops.2
Leslau (1968:4) cites g3dd3l3 'he killed' versus gw3dd3l3 'he dimin-
ished,‘ k3mm3r3 ‘he piled up' versus kW3mm3r3 'he sold mead,‘ and
q3tt3r3 ‘he hired' versus qW3tt3r3 ‘he counted.‘ These examples seem
to constitute valid evidence in favor of the labiovelar phonemes.

As Leslau himself states, however, "Occasionally the other rounded
consonants contrast with the unrounded ones (p. 3).” Consider, for
example, the participle of the verb ‘die,‘ mfip§i_(from the root M91)
and the word for ‘evening,’ Epip, Clearly there exist potential minimal
pairs involving rounded and unrounded consonants other than the velars.

Ullendorf, noticing that the labiovelars never contrast with the
plain velars before rounded vowels, suggests an archiphonemic approach
such that a labiovelar archiphoneme is realized as a labiovelar
before unrounded vowels but as the corresponding non-labialized velar
before rounded vowels (1955:86, 87). Thus we could re-analyze the root
QQM_‘stand' as QEQM; in the infinitive form p§gpm, where the radical QH_
appears before a rounded vowel, we obtain the expected suspension of con-
trast and the plain velar g_occurs. In the participle form of’QEQM,
however, the radical occurs before an unrounded vowel and the labiovelar
g!_appears: gflppi, In the face of pairs such as ppt§_'he died'/pflp§i_

'one who dies' and rot3 ‘he ran'/rWa§i 'one who runs,‘ Ullendorf‘s

 

 

pr0posal is still seen to be unsatisfactory in that it still does not
eliminate the necessity of having to set up phonemic pairs of rounded/

unrounded consonants for all the consonantal phonemes in the language.

 

50

In this thesis, a special morphon MN/0/ is postulated to account
for the data. It is realized as the phoneme P/o/ when followed by a
consonant, but as P/w/ when followed by a vowel. A similar morphon
MN/U/ is also postulated; it is realized as P/w/ before a vowel but as

P/u/ before a consonant. It is MN/U/ which occurs in the root QUIR

 

‘count' and is realized as P/u/ in quter ‘number‘ but as P/w/ in

 

qw3tt3r3 ‘he counted,‘ where it is followed by the §_of the interfix.

Because of the similarity in the realization of the infinitive and
instrumental forms between type A roots and type B and C roots in class
II, it might be expected that the realization of the participle of app
‘go‘ be parallel to that of QQM_concerning the appearance of a semivowel.
Yet the correct form of the participle of H§Q_is pigg, not *pypg, The
simplest manner of treating this phenomenon is to use a special morphon
MN/E/ which is realized as P/e/ before consonants but as P/i/ before
vowels. Figure 3.1 illustrates the realizations of these three special
morphons and their neutralizations with the regular morphons MN/u/,
MN/o/. MN/e/. MN/1'/, and MN/w/.

Thus we see how the use of these special morphons allows us to
account for the data without having to postulate labiovelar phonemes;
the use of a pre-nucleus position in the syllable also allows us to

account for alternations involving consonants other than velars.

3.4 The Status of §_and g_in Amharic

Hetzron (1964) presents an argument against the phonemic status of

3, and the summary of his article is worth quoting in full:

 

51

 

  
    

Phonotactics

 

 

 

L__,,

 

 

P/w/ P/u/ P/o/ P/1/ P/e/
(O): onset (PN): pre-nucleus (N): nucleus . (S): syllable

Figure 3.1: The realizations of MN/U/. MN/0/, and MN/E/.

En résumé: la voyelle du sixieme ordre [gr-MEB] en
amharique n'est pas un phoneme parce que son alternance
avec absence de voyelle peut étre définie en termes de
phonétique articulatoire: (A) la tendance de ne pas
avoir de groupes de consonnes 3 l‘initial; (B) ne pas
avoir un groupe de plus de deux consonnes 3 l‘intérieur;
(C) en final: la méme chose que 8, avec restrictions.
Ces regles ont--certes--une formulation imparfaite, d‘une
part pour des difficultés d'audition, d‘autre part a
cause de flottements et l‘incertitude qu‘on constate 3
l‘intérieur-méme de l'amharique, mais 1e systeme n‘en
est pas affecté. Lorsque §_apparait quand—méme dans
certains contextes morphologiques en contradiction
apparente a ces regles, afin de faciliter la descrip-
tion, on a recours a un artifice: on introduit la notion
de jointure /-/, consonne fictive qui rétablit la valeur
générale des régles A-B-C et qui nous épargne 1'intro-
duction d'un 3_phonologique, distinct du [91 non-
phonologique, automatique. (p. 190)

 

52

Hetzron‘s analysis, while an admirable attempt to deal with the
problem, is open to question on several counts. First, it seems in-
consistent to say in the same paragraph that one's proposed analysis is
imperfectly formulated due to variation within the linguistic system
under investigation but that this variation does not affect the
validity of the rules meant to describe the system. Granted, one has
recourse to a theoretical construct that can aid the description of
some of the variation, but what is the cost in terms of the description
as a whole?

In an acoustic study of Amharic, Sumner (1957) found instances of
the "epenthetic vowel" §_where it should not occur according to
Hetzron‘s analysis. For example, Sumner records the occurrence of
three-consonant clusters (p. 76, [1.4]; p. 77, [3.6] and [4.61) and at
least one instance each of §_immediately preceded by a vowel within a
word and of §_occurring after two consonants but before another vowel.
He also finds cases where a syllabic nasal occurs where the sequence
g_+ nasal is expected (p. 77, [4.41). Cohen (1936) also discusses the
behavior of this troublesome vowel, especially in Chapter 11, Sections
0 and E, and notes much fluctuation in its occurrence.

Bound up with the status Of-g is theappearance or non-appearance
of a glottal stop; this sound may occur optionally before a word-
initial vowel or word-medially in some loanwords from Ge'ez. Its oc—
currence before a word-initial vowel is especially common in slow or
careful speech, or when a word is spoken in isolation. If words
beginning with g can appear with the glottal st0p, it might suggest

that g_may condition the stop just as the "normal" vowels do, so why

 

53

not phonemicize it? On the other hand, Hetzron mentions that 3_may
be a realization of his "consonne fictive" (1964:189, 190).

When the glottal stop occurs in loanwords, there is certainly a
potential contrast between it and other consonants: s§3p£_‘hour/clock'
(the pronunciation sgpg is also heard) is in near contrast with pgpptp_
'seven.‘ The occurrence or non-occurrence of the glottal stop in loan-
Iuords from Ge‘ez is probably an alternation based on non-linguistic
factors which determine style. This is a type of alternation which can
be handled in a stratificational framework, as shown in Herrick (1977).3

The glottal stop can be viewed as a realization of the phoneme
IP/+/, which marks the beginning of a phonological word; loanwords such
as s3at~s3?at can then be analyzed as having the morphonic representation
MN/s3+at/. Figure 3.2 shows how the occurrence of the glottal stop can
be conditioned by the use of a style which is here called "Ge‘ez." In
the presence of this style, P/+/ occurs in a certain position in some
words and is obligatorily realized phonetically as [?J. The circular
GATE nodes control these alternations, and the ordered 0R indicates that
"Ge‘ez“ is a marked style.

At the present time, it is not possible to fully resolve the issue
of the phonemic status of 3; more detailed phonetic studies may find
significant differences in quality or length between those pfs judged to
be phonemic and those judged to be purely phonetic. It may even be that
aSpects of Hetzron's proposal may be profitably integrated into a future
stratificational analysis of Amharic. In order to more fruitfully focus
on other matters at this time, however, the phonemic status of §_and the

non-phonemic status of g_will be assumed in this thesis.

 

 

54

s3?at~s3at 'hour/clock'

 

 

s a a t
.1
- - Ge'ez . P/+/
=;'D
['2]

.Note: The broken line indicates omitted structure.

Figure 3.2: The conditioning of 3_in loanwords by style.

3.5 The Phonotactics of Amharic
3.5.1 The Phonotactics of Segments

The general structure of the Amharic phonotactics is presented in
Figure 3.3. This figure differs in several ways from that given in
Figure 2.3, but it will still correctly process words consisting of a
root and an interfix, as will be shown in Section 3.8.4.

Line 1 leads down to the word boundary P/+/, which originates in
the morphology as a determined element that signals the beginning of a
phonological word (P-word). Line 2 leads down to a structure which we
may for convenience call the syllable (S); it generates phonemic materi-

al which precedes the final position in the word, indicated by line 7.

 

 

55

(final)

 

 

(onset) (pre-nucleus) (nucleus)

Figure 3.3: The general structure of the phonotactics.

Line 3 leads to the onset position, and line 4 to the pre-nucleus,
which accounts for the pr0per realizations of MN/U/, MN/0/, and MN/E/,
as was seen in Section 3.3. Next, line 5 leads to the nucleus posi-
tion. Line 6 allows for recursion at this level of the tactics, and
line 8 for recursion at the level of the P-word. The Optionality ele-
ments on lines 3, 4, 5, and 6 allow for both the case where recursion
on line 6 stops and where the final position alone stands as a phonolo-
gical word.

A word of explanation with regard to the nucleus position at this
point: As stated previously, the phonotactics must allow for CV se-
quences in the processing of words based on the root/interfix system in
the morphology. In cases where a consonant cluster actually occurs in

the phonetic realization of such words, a special vowel phoneme P/#/

 

56

(realizing MN/#/) ensures that the nucleus position is filled, thus
maintaining the pr0per sequence of consonants and vowels, even though
this phoneme itself has no further realization.

In her analysis of Iraqi Arabic, Al-Azzawi (1973) assumed that all
words displaying surface clusters had to be analyzed as involving an
instance of the "zero vowel.“ Thus an Arabic word such as gpppp_'you‘
(masc. sg.) would have the morphonic representation MN/?an#ta/ and be
processed as

P/?a n# ta/
CV CV CV

despite the fact that there is no reason to believe that this word in-
volves a root and an interfix.

Sullivan (1976:172, 173) has proposed a general principle of network
operation such that an element originating at a higher level is realized
in the first available tactic position at a lower level (providing all
other tactic requirements are met). In relation to Figure 3.3, the
optionality elements on the pre-nucleus and nucleus lines allow the
tactics to accommodate two contiguous consonants (or more) if there is
no vowel that can be processed between them. In the case of Amharic
gpt§_‘you‘ (masc. sg.)--parallel to the Arabic example cited above--the

morphonic representation can be simply MN/ant3/, the tactics processing

in the order

0 onset

O pre-nucleus
nucleus a
onset n
pre-nucleus
onset t

pre-nucleus
nucleus
final 3

O
O nucleus
9
fl

 

57

If we wish to realize the infinitive form of the root §B3_'break,'
however, we must employ the morphon MN/#/ to account for the surface

cluster which occurs. Compare columns 1 and 2 in Table 7.

Table 7: The use of MN/#/ in tactic processing.

Column 1 Column 2
Morphonic input: MN/m3 (SBR-3)/ MN/m'a’ (SBR-#3)/
Tactic processing:
onset m m
0 pre-nucleus
nucleus 3 3
onset S S
O pre-nucleus
nucleus 3 #
onset B B
O pre-nucleus
nucleus 3
final R R
Tactic output: *m3S3BR ' m3SB3R

Clearly the morphon MN/#/ must be used when the correct sequencing
of vowels and consonants IS necessary to realize certain forms. As we
have seen with these examples, however, the tactics presented in Figure
3.3 can economically account both for words based on the root/interfix
system and those which are not.

It is felt that the use of a realizational portion with an accom-
panying phonotactics provides a natural and straightforward description
of the behavior of words based on the root/interfix pattern. The inter-
digitation of the consonants of the root and the vowels of the interfix
is the direct result of the phonotactic patterns of Amharic.

In a stratificational format, one is able to eliminate the super-

 

58

fluous rules necessitated by having vowels in the underlying representa-
tions of roots; compare Brame‘s two-step derivation of the passive stem
in Arabic which was first discussed in Section 2.2.1 of Chapter II and

is here repeated for convenience.

2) Passive Formation: V +~i / ___C] / [+passive]

3) Quasi-Assimilation: a +-u / ___CiC / [+passive]

In order to produce the passive stem pppjp, he must first derive the i_
of the passive stem which is then used to condition the change of the
first vowel of the stem from p_to p, It is likely that a stratification-
al analysis of this phenomenon would result in the morphonic representa-
tion MN/CCC°ui/--that is, root simultaneous with passive interfix--which
would then be processed by the phonotactics to result in P/CuCiC/.

Note that it ought to be possible to formulate a transformational

rule such as the following to account for Passive Formation in Arabic:
Passive Formation: CCC ui +~CuCiC

This rule accounts for the data, but it provides no explanation why
natural languages should have such rules. The same criticism can be
leveled at Bar-Lev's use of such rules as HOP and INSERT, discussed in
Section 1.4.4 of Chapter I. Why should Modern Hebrew have rules that
move elements from one side of the stem to the other and then reverse
the position of these elements with the consonants on either side? To
this writer such rules appear at best counter-intuitive and at worst
h0pelessly pg_ppp,

In the stratificational treatment of Amharic presented here, the

 

 

59

use of a phonotactics to account for the proper positioning of radicals
and vowels in words based on the root/interfix system exploits the uni-
versal fact that all human beings--and hence all natural languages--
process the phonological consituents of words in sequence.

Let us now examine the phonotactic structure of Amharic in greater

detail.
3.5.2 The Onset

Leslau states (l968:5) that long consonants do not occur initially
in Amharic. Phonetically speaking, he is correct, but it considerably
simplifies the tactics if we allow both long and short consonants to oc-
cur in the onset position, with the proviso that MN/+C:/ is realized as
P/+aC:/. Figure 3.4 presents the structure of the onset in accordance
with this requirement. The figure also shows that P/h/ does not occur

with the phoneme of length P/:/.

3.5.3 The Pre-Nucleus

The pre-nucleus occupies the position after the onset; when reali-
zations of MN/Ul, MN/O/, and MN/E/ occur, they pattern as shown in
Figure 3.5. Notice that P/w/ need not occur both in the onset and in
the pre-nucleus. If P/w/ occurs as the first consonant of a word, the
onset position will be unfilled; if a consonant other than P/w/ is to
be first, it will fill the onset position and then may be Optionally
followed by P/w/ in the pre-nucleus. A similar tack holds for P/i/.
When it occurs alone, that is, without an immediately following vowel,

it will be fit into the pre-nucleus and the following nucleus position

 

 

6O

 

 

other
C's

Figure 3.4: The onset position.

(syllable)

’.(recursion)

(onset) . C?

I

 

\(nucleus)
(pre-nucleus) w
1.

Figure 3.5: The pre-nucleus in the phonotactics.

 

 

61

will be empty.

As it turns out, however, a slight alteration must be made in
Figure 3.5 to allow for the prOper realization of a reduplicative pho-
neme P/Cr/. It can be seen from the data in Table 8 that for forms
such as gwagwam3 'he stood up repeatedly,‘ both the initial consonant
of the root and the pre-nucleus are reduplicated, whereas the expected
*pigppg_does not occur, ppppg§_‘he went repeatedly' being the correct
form. Figure 3.6 shows a revision of Figure 3.5 to reflect this situ-

ation.4

Table 8: Reduplicated verbs in Amharic.

 

 

 

Basic Stem Gloss

s3bb3r3 ‘he broke'

m3n3zz3r3 'he made change (money)'
g3dd3l3 ‘he killed‘

sam3 'he kissed‘

qom3 'he stood up'

hed3 'he went'

Reduplicated Stem Gloss

s3babb3r3 'he broke repeatedly‘
m3n3zazz3r3 ‘he made change repeatedly'
g3dadd3l3 ‘he killed repeatedly'
sasam3 'he kissed repeatedly‘
ananm3 ‘he stood up repeatedly'
hahed3 'he went repeatedly‘

Note: Glosses are only approximate.

3.5.4 The Nucleus

Figure 3.7 on page 63 presents the general structure of the nucleus
position in the phonotactics. Here we see the preemptive vowels related

to an ordered OR in the tactics, which ensures that they will take pre-

 

62

(syllable)

A

Q

0
(onset)
(pre-nucleus)

—lo

 

 

Figure 3.6: Revision of Figure 3.5 to account for reduplication.

 

(nucleus)

3

    

(preemptives) (normal vowels)

 

 

Figure 3.7: The nucleus, including zero realization of normal vowels.

cedence over the normal vowels. The normal vowels must be realized as
zero when they co—occur with preemptives, and this is accomplished
through the use of ordered ORs in the realizational portion--the normal
vowels are realized if the tactics will allow them, otherwise they are
realized as zero. The realization of the “place holding" morphon is
the phoneme P/#/, which has connections to the phonotactics but no
further realization. More details concerning the phonotactic arrange-

ment of the vowels will be presented in Section 3.7.3.
3.5.5 The Final

In a great many cases, it is virtually impossible to tell whether
or not a given word-final sequence of two consonants is "broken up" by

an occurrence of the vowel g; for example, in the words q3§(a)n 'thin,‘

 

 

——i——i
64

sad(6)b ‘insult,‘ and n3m(a)r 'leOpard,‘ it is difficult to tell whether
there is an instance of the phoneme P/a/Ior whether the transition from
one consonant to the other alone accounts for the vocalic effect. As
Ullendorf observes, "In most of these cases, it would be wrong to be
dogmatic...(l955:206)."

There are, of course, instances where there is no doubt that the
final cluster occurs without an intervening vowel, and in such cases,
the morphonic sequence MN/...CC+C.../ is fairly consistently realized as
P/...CCe+C.../ (although see Section 3.4). The phonotactics presented

here accounts for this by means of an obligatory final position which

 

must be filled by either (1) a voWel which is neither a preemptive nor
P/#/, (2) a single consonant, (3) a long consonant, or (4) a sequence of
two consonants. The tactics then provides for a determined occurrence
of P/a/ when a long consonant or a cluster is followed by a single con-
sonant. The structure of the final position is given in Figure 3.8. As
mentioned above, the exact inventory of two-consonant sequences is open
to debate and cannot be dealt with in detail at this time. We will use

the cover-term CC clusters to designate the tactic structure where such

 

clusters occur.

3.5.6 The Integrated Phonotactics

Figure 3.9 on page 66 illustrates the phonotactics which results
from the integration of all the tactic phenomena discussed so far. A
few words of explanation concerning this figure will now be given.

First, this tactics treats phonemes in terms of segments; after a

componential analysis of the Amharic phonemes is presented in the next

65

(nucleus) (final)

(preemptives)

 

I

 

CC clusters 9

 

other
C's

Figure 3.8: The final position.

section, we will see that some of the components of the vowels may be
included in the tactics as determined elements. Second, it can be seen
that final long consonants are also related to a determined occurrence

of P/e/ which occurs before a long consonant immediately preceded by
P/+/. This allows a word such as Egg 'hand‘ to have the morphonic re-
presentation MN/+g:/ while also accounting for the determined occurrence
of P/a/ when such words are followed by words beginning with a single
consonant. For example, the phrase eggon3w 'it's a hand‘ is morphonical-
ly MN/+g:+n3w/ but phonemically P/+eg:a+n3w/.5

Let us now examine the phonemes of Amharic in terms of components.

66

 

(final)

(nucleus)

 

I m

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.9: The integrated phonotactics.

67

3.6 A Componential Analysis of the Amharic Phonemes

A knowledge of the phonemic components--phonons--of each of the
Amharic phonemes is needed to allow the economical treatment of certain
morphophonemic alternations; it also allows us to specify determined
elements in the taCtics, thus simplifying the description. Table 9
shows the componential analysis of Amharic which is assumed here. The 1

table is fairly self—explanatory, but a few comments will be given.

Table 9: A componential analysis of the Amharic phonemes.

Lb Ap Fr
p t t k
(H <:t> d g g Vd
p. t. I? 9 GI
f s s h
fh1<: z 2 Vd
5 G1
Ns m n h
Vb r
w 1 y
u 1, 1p 9 Hi
Vo<o e, ep 3
a L0
L9
Lb: labial Ap: apical Fr: frontal Cl: closed
Vd: voiced G1: glottalized Sp: spirant Ns: nasal
Vb: vibrant Vo: vocalic Lg: long

There are three series of stOps and spirants: voiceless (treated
as unmarked), voiced, and glottalized. The nasals and other sonants

will receive a determined specification for voicing at the phonetic

68

level, so they need not be marked as voiced at the phonemic level. The
phoneme P/h/ is marked only for Spirancy, not for glottal articulation

as might be expected. This is because in certain positions there is a

velar allophone of P/h/ and in others a glottal; not marking this pho-

neme for place of articulation will simplify the phonetic treatment of

this alternation by allowing the place of articulation to be completely
determined by the environment.

Two additional phonemes in Table 9 need a word of explanation. The
phonemes P/ip/ and P/ep/ are phonemes which condition the realization of
apical consonants as their frontal counterparts: §_as g, §_as g, l_as
y, and so on. For example, the singular masculine imperative of the
root LB§_‘get dressed' is 1§Q§§, but the singular feminine imperative is
l§p§§, Compare these forms with the singular masculine and feminine
imperatives of the root SRQ_'steal‘: §§p§g_and'§§p§gig The phoneme
P/i/ does not condition this alternation as evidenced by such words as
sipimp_'movie‘ and gigppg ‘cigarette.'

A network diagram of the behavior of P/ip/ and P/ep/ is given in
Figure 3.10. This diagram also shows the zero realization of P/ip/
after frontal consonants, as in the participle form §§l§_from the root
§L§_'be annoyed.‘6 One fact that we have omitted here is the neutrali-
zation of P/¢/, P/t/, and P/§/ before P/1p/ and P/epl; this will be
discussed in Section 3.7.1.

Note that the alternation between apical and frontal consonants can
be treated in terms of components. This diagram accounts for the fact
that P/ip/ has no further realization when it conditions the phonon

PN/Fr/; the AND used for the realization of P/ep/, however, accounts for

 

69

P P

/1/ P/ep/ /e/

 

 

 

 

PN/Ap/ PN/Fr/ (other phonons) (other phonons)
CFr: E, 9, E3 5, 2; fi.

Figure 3.10: The alternation of PN/Ap/ and PN/Fr/ as conditioned by
P/ip/ and P/ep/.

the fact that this phoneme does have a realization as a phonetic vowel,
as in l3be§:e '(I) having dressed,‘ which is the realization of the

phonemic string P/+l3bes:ep/.
3.7 The Phonemic Sign Pattern

The phonemic sign patterns specifies the realization of each pho-
neme in terms of components, usually through the use of downward un—
ordered ANDs just below the phonotactics; in some cases, however, part
of the sign patterns can be integrated into the tactics through the use
of DIAMONDS, a process which results in a simpler expression of informa-
tion according to the stratificational simplicity metric. Figure 3.11

illustrates two networks which have the same effective information, but

 

which differ in surface information; the simplicity count for each is

 

 

7O
9 ‘~
I'II) At} t A" k
Lb Ap DO Cl
Simplicity: 9 + 2 = 11 Simplicity: 6 + l = 7

Figure 3.11: The integration of components into the tactics. (Adapted
from Sullivan 1978.)

given below it.7

3.7.1 The Phonemic Sign Pattern for Consonants

Because the tactic restrictions on two-consonant clusters have been
left unspecified (see Section 3.5.5), it is not clear at this point how
much of the consonantal sign pattern will be integratable into the tac-
tics; for this reason, the consonants are treated as being realized ex-
clusively through the phonemic sign pattern. Figure 3.12 shows the pho-
nemic sign pattern for the stops, including the neutralization of apical
stOps with frontal stOps conditioned by P/ip/ and P/ep/.

Figure 3.13 on page 72 specifies the sign pattern for all the spi-
rants except P/s/; when P/s/ is followed by either P/ip/ or P/ep/, it

/SI /2/

 

 

 

 

IPN/Ap/ PN/Fr/ PN/Sp/ PN/Vd/ PN/Lb/

Figure 3.13: The sign pattern for all spirants except P/§/.

has the same realization as P/§/. For example, compare g3ll3s3 'he un-

covered‘—-phonemically P/+g3l:3s3/--with g3la§ 'one who uncovers'--

 

phonemically P/+g3lasip/. The alternation between PN/Ap/ and PN/Fr/ has
already been described, so all that is needed to ensure the proper rea-
lization of Pls/ in this environment is to provide for a determined oc-
currence of the phonon of closure PN/Cl/. The resulting configuration
is shown in Figure 3.14.
Figure 3.15 specifies the realization of the sonants: nasals,

semivowels, the vibrant, and the lateral. This completes the presenta-
tion of the sign pattern for consonants, and we may now turn our atten-

tion to the vowels.

73

 

 

 

 

P/1p/ P/ep/ P/s,/
@-— ~. A A
\ ‘ 'l ' ' I,
Ali n r
:1
A
1" ‘U’
s‘vl \\ I“ I,
PN/Fr/ PN/Ap/ PN/Cl/ IPN/Sp/ PN/G]/

Figure 3.14: The phonemic sign pattern for P/§/-

 

 

 

 

PM PM P/m/ P/n/ PM/ PM P/y/
H ep
1'
PN/Vb/ PN/Lb/ PN/Ns/ PN/Ap/ PN/Fr/

 

 

Figure 3.15: The phonemic sign pattern for sonants.

74
3.7.2 The Realization of Vowels

Because the treatment of the phonotactic arrangement of vowels is
fairly complete, it is probably safe to integrate as much of the sign
pattern as possible into the tactics, where a phonon can be automatical-
ly supplied once for a whole class of phonemes instead of having to be
repeatedly specified for each one in the sign pattern.

If all possible components are "factored out" of the sign pattern
and into the tactics, Figure 3.16 is the result; it is believed that
this figure represents the simplest tactics that can be constructed and
still account for the realization of the vowels prOperly.

Note that P/#/ has no connections whatever to any phonon; it fills
a tactic position, but that is all. It might be thought that P/3/,
P/o/, and P/e/ also behave in this manner, but inspection of the network
will show that determined phonons are supplied for each of these vowels
at a point higher in the tactics. The phonons PN/Fr/ and PN/Vo/ are
provided for P/e/, PN/Lb/ and PN/Vo/ for P/o/, and PN/Vo/ for P/3/.

Also, the details of the apical/frontal alternation are not shown;
in connection with this, however, the AND that appeared in Figure 3.10
to account for the realization of P/ep/ has been eliminated through the
use of DIAMONDS in the tactics. The preemptive vowels occur higher in
the tactics than the normal vowels, which necessitates using a separate
DIAMOND to relate them to the phonon PN/Vo/.

Lastly, it is simpler to treat the realizations of P/i/ and the
determined occurrences of P/e/ with unordered ANDS in the sign pattern.

This concludes our examination of the Amharic phonotactics. We will

now proceed with a discussion of how it handles the data from Chapter II.

 

75

 

.mpcocoosoo po mseop :p mpo30> po coppmNppoog och ”o_.m oeompd

\o>\ze \pz\ze

Apoopoo sppz copp
e -oceoppo popcoep
\ a\za ea oepeopppeeoov

 

   

\o._\zQ \34\

Zn. BF, \3\..

 

 

>50 3...

WNW... 090$
APo oocpseopoov . \{a
9::

1\ I

Apocp: Amzmpozcv Amzopo::-oeov

 

 

 

 

 

 

 

 

76

3.8 The Realization of the Infinitive, Instrumental, and Participle

FormS' ' '

In this final part of Chapter III, we will examine in finer detail
the realization of the verbal forms presented in Table 5, focusing on
the realization of the special vowel morphons MN/U/, MN/O/, MN/E/, and
MN/A/. We will also focus on the manner in which the phonotactics
processes the morphonic input of the simultaneous root/interfix struc-

ture specified by the morphotactics.
3.8.1 Vowel Morphons with Alternate Realizations

Figure 3.17 presents the network representation of all vowel mor-
phons with alternations that have been postulated to deal with the data.
The behavior of MN/U/, MN/O/, and MN/E/ has been discussed in Section
3.3, so we need only say a few words about MN/3/ and MN/A/.

As mentioned in Section 2.3.2.1 in Chapter II, the morphon MN/A/
is realized in the same manner as MN/3/ when it is followed by a long
consonant but as P/#/ or zero everywhere else. The realization of MN/3/
is somewhat more complicated. Its realization as zero (conditioned by
a following P/a+/) seems to take place only within phonological words;
compare ppp§g_'to know'--morphonically MN/+m3 (a+wq-A:3)/--with gflt§_
aww3geh ‘you (sg. masc.) knew‘-—morphonically MN/+a+nt3 +(a+wq-A:3) ehl.

To account for the behavior of some particles, among them y§_'pos-
session,‘ p3 'with/by,‘ and k§_‘from,‘ the morphotactics must allow them
to occur either as separate phonological words or as a Single phonolo-
gical word consisting of a nominal with a prefix; compare k3 addis and
Igpyiyg, both of which mean ‘from Addis [Ababa]. The morphonic represen-

tation of the first is MN/+k3 +a+dzis/ and of the second, MN/+k3a+d:iS/.

77

.mcoppocgoppo mopppopgxo mcoeoeos Fo3o> ”m_.m mezmpd

\e\ \.\e \6\e \.\e \0\e \=\a \3\e

 

 

 

 

 

 

.L .L@.

-- .6 1

\M\ZZ \w\zz \O\ZZ \:\ZZ \O\zz \D\ZZ

 

   

 

 

78

In other environments, MN/3/ is realized as zero when it occurs
simultaneously with MN/a+/ and as P/3/ elsewhere.
3.8.2 The Realization of the Data in Table 5 from the Level of Mor-

photactic Structure to the Level of Phononic Representation

The following tables will explicitly show how the verbal data from
Chapter II is handled in the analysis of Amharic phonology presented in
this thesis. In each table, the line labeled Mopphotactic Structure
presents the structure of each form in terms of the morpheme classes
which comprise it. In the Morphonic Reppesentation, the morpheme
boundaries are indicated with a space, with the exception of the root/
interfix structure, which is given in parentheses with a raised dot
between the morphemes to indicate their simultaneity. The Morphonic
Alternation Pattern (MAP) illustrates the structure which results from
the conditioning of various morphonic alternations by the phonology,
and the Phonemic Representation is given in terms of segmental phonemes

occurring in the order determined by the phonotactics. The Phonotactic

 

Structure shows the structure of a particular form in terms of pp§p£s_
(_0_), pre-nuclei (P), ppplei (_N_), and the flpfl (_F_). The Phononic
ngresentation shows the realization of a form in terms of the compo-
nents which result after all conditioned alternations have occurred;
components written vertically are simultaneous with each other. In all
rows except the Phononic Representation, the radicals of roots are both
capitalized and underlined for easy reference; the occurrence of the

word boundary is ignored.

79
3.8.2.1 Class I Roots

Class I roots, of which we may recognize five types according to
their surface forms, all condition the same realizations of the in-

finitive, instrumental, and participle morphemes.
3.8.2.1.1 Class I(A) Roots

Table 10 indicates the realization of the infinitive, instrument-
al, and participle forms of a type A root from class I. As can be seen
from the table, the root SBB_'break' has no gemination, and MN/A/ is

therefore realized as P/#/.
3.8.2.1.2 Class I(B) Roots

Table 11 presents the realizations for a type B root in class I;
type B roots always have a long second radical which conditions the re-
alization of MN/A/ as P/3/. In other respects, the realization of

forms for type B roots is the same as that for type A roots.

3.8.2.1.3 Class I(C) Roots

Verbal forms for roots of type C are given in Table 12. Their
main feature is the fact that the first radical is a preemptive, which
conditions the realization of MN/A/ as zero in the morphonic alterna-
tion pattern. In the participle, the preemptive conditions the zero

realization Of MN/3/ in the same manner.

 

8O

pmoupmw

ea
pze>eomo em
o>poo>a<6>eo

mozozo

peee.ppeom
epmpeppeaa

peonmm

ea e>
p: oopo em
e>e>e>eoo>e<

mozozo

peoa.ppoom
epmpoppeee

 

opmmnpwwms

 

ooflmo> mo em mz
o>o>poe>e<o>eoo>eo
uzozozozo
.Q.I.......m..e

map

.0
(5 L5
_J
LL
v
'E

A
p A: . ”Adv we

emep Ameep.ppoomv peep

 

popco53mecp

mpemnmme
Ln v>

04w: Foam mz

o>o>n>o>npo<o>op

ezozozozo
e pmmmmee
map Ameummwv we

map penummwv we

pmeH Ameep.ppoemv peep

 

popcoszepmcp

oznpmeze

e> as am a:
poe>e<o>eoo>eo
uzozozo
mwmlpMHe_

 

o
A: .oupdv :2
Ameep.ppoomv peep

 

e>pppepeeu

commas

e>
_oem mz
e>o>ese<e>es

mzozozo

mmmemme
Ameummmv .5

Ame mmwv we

Ameep.ppeomv peep

 

e>_ppepeep

coppopeomcoep

coppopcomoeoom upcocoga

weapozepm oppoopocogo

coppopcomoeoom opEozoza

oqz

coppmpcomoeoom upcogoeoz

weaposepm oppompogoeoz

.pooe Amvp moo—o o 40 mELoe opopoppcmo oco .popcmssepmcp .o>ppp:p¥:p mcp wo coppoNppmoL oce

.peez. owed ”Amvp mmepo
”pp opomp
coppopeomcmep

coppopcomoeoom upcococa

oezpozepm oppompocogo

coppopcomoeoom opsocoga

a<z

coppopcomoeoom upcocoeoz

weapozepm oppoopogoeoz

.xaeee_.mmm

.pooe A<VH mmopu o mo mseow opopoppcoo oco .popcoE:me:p .o>ppp:pwcp ozp mo coppoNppooe oz»

A<VH mmm_u
”op mpnoh

81

”snow

op

u>
omop~004
Lmo>a<o>

mozoz

 

eWene+<
e

a

pew.wue+<
www.mne+<

 

peee.ppeoe
epmpeepeee

.pooe onp mmopu o we mseom opopoppeoo one .Fmpco53epmep .o>wpp:wwcp oep wo coppo~ppooe mew

peopm

ea
exemes es
e>ese>e<o>

mozoz

a 1.111.

pa64+<

 

epew.ws+<

 

epew.as+<
peee.ppoow
epmpeppeea

.pooe Auvp mmmpu a mo maeom o—opoppeoo one .popco52epmew .o>ppwepwcp ozp wo coppoNPPowg we»

 

 

MMwUmE

up

u>
opom poopmz
o>gmo>a<o>np

uzozozo

 

e emene+<e

 

wee www.mne+<v e
ea. Aww.mne.<v we

 

pmep Aweep.ppeewv peep

 

popcosaepmcp

epewpee

ea
eopzem oomz
e>e>eoe>e<e>eu
azozozo

a emwpume

 

we? Aww.es+<v e

new Awm.ao+<v we

pmep Aweep.ppoowv peep

 

popco52epmcp

manoos
as
e>
em poeomz
e<e>e<e>eo
azozo

.w6ne+<e

 

Aww.muo+<v E
AwK.wue+<v we
“weep.ppeomv peep

 

 

e>pepe_eep

empoe

om opmz
o4o>o<o>op
dzozo
.mwmume

 

Aww.ao+<v e
Awm.m4+<v we
Aweep.ppoewv _eep

 

e>_ppeweew

coppoweomcmep

coppmpcomoeoom upcocoea

weapozepm owpoopoeoeo

coppopcomoeoom opEmeosa

a<z

:owpwpeomoeoom upcoeoeoz

weaposepm oppoopoeoeoz

 

.3oeoe. muo+<

"onp mmo_u
”mp opoop

coppopeomemee

coppopcomoeoom upcocoea

weapooepm oppoopoeoee

coppopcomoeoom upsocoea

a<z

coppopcomoeoom upcoeoeoz

weapozepm uppompozoeoz

 

.we meme. as+<

"onp mmepo
"up e_eep

82
3.8.2.1.4 Class I(D) Roots

The realizational behavior of type D roots in class I is found in
Table 13; again, roots of this type have the morphon MN/a+/ as their
first radical. For roots in this category, the long consonant of the
second radical conditions the realization of MNlA/ as P/3/, which is in
turn realized as zero because it occurs with the preemptive. In the
case of the instrumental and participle of the root A:2i§2 we see the

palatalization of P/s/ by P/ip/.
3.8.2.1.5 Class I(E) Roots

Type E rootS in class I have MN/a+/ as their second radical, which
again conditions the realization of MN/#/ as zero. The details of their

realizational behavior are illustrated in Table 14.

3.8.2.2 Class II Roots

Roots in class II all condition the same realizations of the in-
finitive, instrumental, and participle morphemes; three types may be

distinguished according to the nature of the second radical or the root.
3.8.2.2.1 Class II(A) Roots

Roots in this category have the morphon MN/a+/ as their second
radical. They condition the realization of M/Inf2/ as zero, as do
all class II roots, but otherwise their realizational behavior is un-

complicated, as Shown in Table 15.

 

83

psmm

ea
pzmzesem
o>eoo>e<

uozo

epz+<m

epe.z+<m

ewe.z+<m
peee.pppeoe

epmwoppeee

 

 

 

 

 

 

 

mesomme
ea
Quezmzoeem mz
o>o>eoo>e<e>eo
azozozo
mewz+<mwe
map =+<m we
map z+<m we
pmep Aweep.pppeemv peep
PMHCOEDLHMCH

Somme

mzopom mz
noo>o<o>op

mzozo

 

z+<mwe

 

z+<m we

 

z+<m we
Ameew.pppeowv peep

 

e>pepepeep

cowpopeomcoep

coppMpcmmoeoom upcoeoze
weapooepm oppompocoee
:oppopemmoeoom owsoeoea
o<z

coppmpcomoeoom oweoeoeoz

weaposepm oppoopoeoeoz

 

.mmpx. z+<m "A<VHH mmo—u

.pooe A<VHH mmopu o wo mseow opopoppemo tem .popco53mecp .o>pppepwep oep wo coppoNppmoe mew “mp mpomp

weapon
eae>
_:eses ese>
o>poe>e<e>po

uozozo

 

emeo+<o

 

wawom4+<w

 

epew.mo+<o

peoa.ppoom

epmpeepeea

opnmpmmms

Luo>
cpwznp opv> mz
o>o>Poo>o<o>Puo>np
uzozozozo

 

a emozowe

 

wee Awe.wo+<ov we

 

wee Awm.ws+<ov we
pmep “weep.ppeoev peep

 

popeoszepmcp

 

ewpewwe

e>
no ose> mz
poo>e<e>poo>es
azozozo
.mws+<owe

 

Aww.ws+<ov we

 

Awm.ws+<ov we
Ameep.ppoemv peep

 

e>ppwepeep

coppopeomeoep

coppmpemmogoom upcocoea
weapozepm oppoopoeosq
coppmpcommeomm opsmcosa
e<z

coppopcommgoom upcoeogoz

weapoaepm oppompoeoeoz

 

.eoppeo. ms+<o "hoop mmepo

.pooe Amvp mmopu o wo mseom o—opoppeoo tem .Fopco23epmew .o>pppcwwep oep wo coppo~wpmoe one ”op upon»

84
3.8.2.2.2 Class II(B) Roots

Roots of type B in class II have the special morphon MN/O/ as
their second radical, by which the participle is realized as Shown in
Table 16. In the case of the participle form, MN/O/ is realized as
P/w/ in the pre-nucleus position of the phonotactics; elsewhere it is

realized as P/o/.
3.8.2.2.3 Class II(C) Roots

The realization of the infinitive, instrumental, and participle
forms of type C roots in class II is given in Table 17. The realiza-
tion of the participle involves the morphon MN/E/ and the use of the
pre-nucleus, as for type B roots. For the root shown here, we see the

palatalization of the final radical conditioned by P/ip/.
3.8.2.3 Class III Roots

Roots in this class are characterized by conditioning the realiza-
tion of M/Inf2/ as MN/Aat/. As discussed in Section 2.3.2.1 of Chapter
II, the endings with L are more simply left in the Sign pattern of
M/Inf2/. Note that this does not entail any complication of the pho-
notactics, which has no trouble in placing the final §_in the proper
position. Two types occur in class III, type A with no gemination of

the second radical and type B with gemination of the second radical.
3.8.2.3.1 Class III(A) Roots

The realizational behavior of the root SM"hear‘ is presented in

Table 18; here the morphon MN/A/ is realized as P/#/, which allows for

85

new;

we
e>esew
eao>o>em
wozao
epmemm
_ewmmm
{.9
peee.pppoee

epmwewpeee

.pooe Auvpp mmmpu m yo mELom opopoppeoo wee .popcw52epmcp .o>pppcpwcp oep wo copponPooe mew

wsmzo
em
pzmzop Pm
o>npo>nppu

uozmo

eweummm
ewaummm
page.pppooe
.mflmflmmmumm

.pooL Amvpp mmopo o eo msgow opowprLeo oco .popcoszepmep .o>ppp:pmcp ozp mo coppeNppooL one

mooews
o4o>ed mz
o>ewo>omo>np
dzozozo

e emmmwe

e e.mmm we

eep.mmm we

peep Awwep.pppoeev peep

 

pepeoE:epmeH

opsooos

Lu

oppxmznppw mz
o>o>npo>puo>np

peep Aweep.

azozozo

aewmmmwe
new.mmm we
ee_.mmm we

H$85 peep

 

popem23epmep

eeewe

e>ea mz
e<e>eme>eo

uzozo

mmmwe

Haw...

flmwe
Aweee._ppeoev peep

 

e>epeepeep

Eooms
mznppw mz
opo>puo>np

uzozo

Aweep.pppeoev peep

 

e>peeepeep

coppapeomeeee

coppopcwmmeoom upeoeoee

weapoaepm oppoopocoea

coppmpcomoeoom opsoeoee

a<z

coppmpcommeoom opeogoeoz

weapozepm oppoepoeoeoz

.66..mmm

HAUVHH mmmpu
"mp mpnmp

coppopeomcoeh

coppmpcomoeomm upcocoeo

weapozepm oppoopoeoga

coppmpcommeomm opsoeoee

a<z

coppcpcOmoeoom upcozoeoz

weapooepm oppoopoeoeoz

.Aeev eeeem..mmm

“AmVHH mmmpu
"mp epoch

86

pnxwp

peee.ppppeoe
epmpoepeee

ea
oewzee mz
o>e>poe>e<o>ee

ezozozo
eeewielwe

a J11.

e e Aw..¥wv we

mop Annwmmv we

peep Aweep.ppppeeev peep

 

pepcoE=epwep

penxwpwe
pooped mz
e<e>poe>e<e>eo

ezozozo

Apewnewmv :e
ApoKHHIvJV we.

Amwee.ppppeeev .weH

 

6>we_epwe_

coppopeumcoee

coppepeomoeoom upcoeoee
weaposepm oppoopocoze

coppopcommeoox opsoeoea

m<z

coppopeomoeoom opeocoeoz
weaposepm oppoopoeoeoz

.eeemeee..nmp "vappp mmepo

.pooL Amvppp mmopu o wo mEeow opowoppeoo one .popeossepmep .o>_ppepwe_ mep mo coppeNp—ooe mew "mp opaop

"—
E
U?

ed
pwmz em
o>eso>ee
eozo

spew

ewwemw
eewemm
peee.pp_eeee

epmweppeee

wammE

ed
oeezmzew m:
o>o>eee<e>eo
wzozozo
e emwmwe
eep Awnmwv we

eee Awemmv we

peep Awwep.epepeoev peep

 

popcoszgpmep

memmE
poeomzew m2
e<e>eee<o>ee
wzozozo
pemwwwe
Apewumwv .e

Apomumwv we
Aweep.pppeoeev peep

 

e>eepe_wep

coppopeomeoew

coppopcomoeoom opeocoea
weapooepm uppoopocoea

coppepeomoeoom opeoeoea

a<z

coppmpeomoeoom oweoeoeoz
weapooepm oppoepoeoeoz

.eeee..mw "Aevepp mmepo

.pooL A<VHHH mmopu e wo maeow m—opowpeeo oeo .popeo53epmcp .o>wpwepwcp oep wo coppon—eoe one "mp o—ook

87

a cluster of consonants in the surface forms of the infinitive and the

instrumental.

3.8.2.3.2 Class III(B) Roots

In type B roots of class III, the long second radical conditions
the realization of MN/A/ as P/3/. In other respects their realization-
al behavior is uncomplicated and does not require extended comment; the

details of their behavior are illustrated in Table 19.
3.8.2.4 Class IV Roots

Roots in this class are structurally very similar to those in
class III, the only difference being that where class III roots condi-
tion the realization of M/Inf2/ as MN/Aat/, class IV roots condition
the realization of this morpheme as MN/A'a't/. Again we find two types--
A and B--depending on the presence or absence of gemination of the

second radical of the root.
3.8.2.4.1 Class IV(A) Roots

Type A roots of class IV have no gemination of the second radical;
MN/A/ is therefore realized as P/#/ and a consonant cluster occurs on
the surface in the infinitive and instrumental forms. The realization-

al behavior of such roots is given in Table 20.
3.8.2.4.2 Class IV(B) Roots

Class IV roots of type B have a long second radical which condi-
tions the realization of MN/A/ as P/3/. Table 21 details their reali-

zation.

 

pwxwp euzmpme pmwzmpme noppopnomnenw
nu no

ween onezen a: po we mz
o>nno>o< o>o>nno>n<o>np o<o>nwo>o<o>nn noppepnomonoom opnononn
wozo wzozozo wzozozo onopoonpm oppoepononn
eenwem eeewwwmwe pennen:e coppepeemeeeee eweeeeee
epwnwwm eep Awuwwmv we now wwwuv :e eez
ow.nwmm eow Amnwwmv we Apmmuwmmv ms noppepnomonnom opnononoz
pnen.>ppoom Amwnp.>ppoomv pwnp Amwnp.>ppoomv pwnp unaponnpm oppoepononoz

mpmpuwpnen

 

pepnosonpmnp

 

e>eepepeen

.openenom..wwm "Amv>p mmepu

.poon AmV>H mmepu e wo menow opopoppneo one .pepno22npmnp .o>pppnpwnp onp mo noppeNTPeon one "pm ownee

88

pn:o epnome pmnoms noppopnomnenk
nu Ln

_: pw one: Pu mz pu Pu mz
o>n>o>pu o>o>n>puo>np o<o>n>puo>np noppepnomonnom ownononn
nozo uzozozo wzozozo onzpoznpm oppoepononn
eemwm eepmw we pwmw we eoepepeemeeeee epeeeeee
epwumm eep nwummv we newwummv we eez
newnmm eop Anummv we Apmmnmmv we nowpepnomonomm ownononoz
pnen.>ppooa pmnp Amwn~.>ppooxv Fwnp Amwnp.>ppoomv pwnp mnopoznpm oppoepononoz

memeewpeee

 

pepno53npmnp

 

6>wpwepeep

.xepm..mm

.poon A<V>H mme—o e wo manow opopoppneo one .pepnosanpmnw .o>pppnwwnp onp wo noppeNppeon one

HA<V>H mmepu
"om apneh

89
3.8.2.5 Class V Roots

Class V roots condition the occurrence of extra vowels in the re-
alization of the morphemes M/Inf2/ and M/Part/. A type A and a type B
can be identified, but in this case the determining factor is not gemi-
nation—-type A roots have four radicals whereas type B roots are ana-

lyzed as having five radicals, the third of which is MN/a+/.
3.8.2.5.1 Class V(A) Roots

As can be seen in Table 22, the absence of gemination in type A
roots in class V conditions the realization of the morphon MN/A/ as
P/#/, which accounts for the consonants cluster which occurs in the in-
finitive and instrumental forms. In other respects, their behavior is

straightforward.
3.8.2.5.2 Class V(B) Roots

Roots in this category have the preemptive MN/a+/ as the third
radical of the root. Aside from this, they behave exactly as do roots

of type A in class V, as can be seen in Table 23.
3.8.2.6 Class VI Roots

Roots in this class have the preemptive morphon MN/a+/ as their
first radical. This class conditions the realization of M/Part/ as
MN/3Aaip/; the second vowel of this interfix must be MN/A/, its sub-
sequent realization as P/#/ accounting for the consonant cluster that
occurs in the surface form of the participle. The palatalization of

P/S/ by P/ip/ is also apparent. Table 24 presents the details.

 

90

 

 

 

 

 

 

wcepmnmo ewvmpenmowe wcmpenmoms nowpawnumcenp
Lu o> U> Lu o> o> U> U>
expwoq a; Pu 04wzpw 24 Pu mz Fm n4 pu mz
o>poo>e<e>poe>e< e>o>poe>e<e>poe>eee>en poe>e<e>poo>e<o>ee eoeweeeemeeeee eeeeeeee
nozozozo nzozozozozo wzozozozo eeeeoeepm oppoepeeoee
nemen+<ewm e eemw :+<w .mw :+<e eeeeepeemeeeee epeeeoee
e..eww .amm a: Awwwfiwflmv: we www.mflmmae. we .2:
opemw.mmummm eow Awmwummummmv we Ammmnmmummmv we nowpepnmmonoom ownonnnoz
pnen.>poom pmnp Amwnp.>poomv Awnp Amwnp.>poozv Awnp onoposnpm oppoeponnnoz
.mwmflwmmumm pepeeeeepmen eswpeepwen .xpe. n+<me "Amv> mmepo

.poon Amv> mmepu e wo manow opnwoppneo one .Aepnosznpmnp .o>pppnpwnp onp wo noppeNpAeon one "mN opnew

 

 

 

 

 

pnexmmme epnmxmmeme nmxmmems noppopnomnenp
en nu
w: o; am a: opp: om m2 m2 om m2 m2
o>n>o>puo>o<o>np o>o>n>o>Pun<o>npo>op n>o>poo<o>npo>n4 nowpepnomonomm o_nononn
nozozozo nzozozozozo nzozozozo weaponnpm oppoeponone
e Memeiwm e Amwmwwwmwe .memwmwzwe coppepeemeeeee eweeeoee
owemwummmm enw Amow.x¥mzv we Am em.mxmzv we n<z
owewwummwa eop Awnwummmmv we Aw. <e. .mxmzv ms nowpepnomonoox opnononoz
pnen.>poom pmnp AmwnH.>poomv AwnH Amwnp.>poomv Awnp unsposnpm oppoepononoz
.mpmmwflwumm pepeeeeepmen esepweewen .wpppmee. mew: HA<V> mmepo

.poon A<v> mwepu e wo manow epnwowpnen one .pepn053npmnp .o>pppnpwnp onp wo coppenppeon one "mm apnep

 

91

wman

nu n<
p:o> o>
o>pun>o>nm

nozozo

newewp
epwwummm
enmwnmmm
pnee.nn>peee

epmeewpnee

mexne

nmop n<op
nno>AUmzo>

nozozoz

o.l.l.111

ewexwz+<

 

epeww.w¥z+<
e

 

nemw.w¥z+<

p>
pnen. poom

epmneepnee

 

epmnmNmE

Ln ee
oppzo> o> mz
o>o>Aun>o>omo>np
wzozozozo
e emwieiee

eee Amw.mme we

omen Awwen.nn>eoeev peep

 

Aepnosonpmnp

emmxnes

opom o<oomz
o>nmo>—umzo>np

nzozozozo
..e new?

 

eee Awww.w¥z+ev e
enp Ammm.m¥z+<v ms

 

peep Awnen.n>poeev peep

 

Aepno53npmnp

pemnWNmE
an

poene> e> m2
e<e>poe>o>emo>eo
nzozozozo
pemwmemwe
Apewwummmv we
Apemwemmmv we

Awwen.pn>peoev peep

 

esnwnepnen

mmxnee
om n<opmz
n<o>Aumzo>np
nzozozo
.wwmwmume

 

Awom.mxz+<v E

 

Awww.m¥z+ev we
Awnen.n>eeeev peep

 

eseeneneep

noppnpnomnen»

noppepnomonoom opnononn

onnpoznpm oppoepononn

noppepnomonoom opsononn

o<z

noppepnomonoom opnononoz

onepoznpm oppoeponnnoz

.poon HH> mmepu e no manom opnwoppneo one .pepnosznpmnp .o>wppnpwnp onp wo noppeNAPeon on»

.p36 eepenpm..mmm unp> maepo
new epeee
noppopnomnene

noppepnomonnom omnononn

mnzpoanpm oppoepononn

noppepnomonnom opsononn

a<z

noppepnmmonomm ownonnnoz

onzponnpm oppoepononoz

 

.eee_ we. w¥z+<

.poon H> mme—o e wo manow opopoppneo one .Aepnossnpmnw .o>ppwnpwnp mnp wo noppe~p_eon one

wH> mmer
new opnee

92

3.8.2.7 Class VII Roots

Although the roots in this class have three radicals, they also
condition the t_ending characteristic of the biradical roots of class
III. Roots in this class also condition the realization of the mor-
pheme M/Part/ as MN/3Aip/. It is the realization of this morphon as
P/#/ which accounts for the cluster of consonants which occurs in the
participle. Table 25 presents the realization of the infinitive, in-

strumental, and participle forms for a class VII root.
3.8.2.8 Class VIII Roots

Roots in this class parallel the behavior of roots in class VII
in all ways except that they condition MN/3/ as part of the realization
of M/Inf2/ whereas class VII roots condition MN/a/. The "absorption"
of P/ip/ by the preceding frontal consonant P/<‘E/ is also apparent.
The realizational details for roots in this class are Shown in Table

26 on the following page.

93

upwm

no em
nno<o>o<

nozozo

HHH>

pnee. poem

 

opnpowpnen.

ewpwmwe

enpo em a:
o>nne<o>e<e>en
nzozozozo

eeemwmemwe

Aewummmv we

 

e
p Anw.wnwv we

peep Aweep.ppn>peoev peep

 

FaucszmeCH

ewwpwmwe
_o _o ee mz
e<o>ewe<o>e<o>en
nzozozozo
pwmwmewwe

 

Aewww.wsmv we

 

prmw.unwv we

Awwen.npn>eeeev peep

 

asepnepeep

.poon HHH> mmepo e wo manow opnpowpnen one .Aepnosnnpmnw .o>wppnpwnw onp wo noppeNAAeon one

noppopnomnenp

nowpepnomonoox opnononn

onzpoonpm oppoepononn

noppepnomonoom owsononn

n<z

nowpepnomonoom opnonnnoz

onepoznpm oppoeponnnoz
.ooxonne on..mmw “HHH> mmepu
"om apne»

NOTES to CHAPTER III

The palatalization mentioned by Leslau is a "secondary" or
"phonetic" palatalization, distinct from the morphophonemic
palatalization discussed in Chapter II.

The absence of word-initial consonant clusters in Amharic has
fostered the interpretation of the sequence'velar‘+ w as a unit
phoneme rather than as a cluster.

The reader is referred to this work for a detailed treatment of
certain types of "stylistic" alternation within stratificational
theory.

Figure 3.6 is presumed to be equivalent to the diagram below; re-
search into the use of the reduplication element in a model of
performance may ultimately provide reasons to prefer one to the
other.

 

The tactics presented here allows the sequence MN/...CC+C:.../ to
be realized as P/...CC+eC:.../, but I do not know if the realiza-
tion of P/+/ as the glottal stop ever occurs in this environment.
I suspect that if it does, the resulting phonetic representation

would be PT/...CCe?aC:.../, but I have not seen this recorded in

the literature, nor have I yet had the opportunity to investigate
it with an informant.

94

 

95

There is actually some variation in the appearance or non-
appearance of the i_of the participle; one hears both S316
and s3lti for the participle form of the root §L§913 : and
13 :1 for the participle of the root LY:, and so on. T is
var1ation could very easily be accounTEd for through a minor
modification of the diagram presented here, but it does not
affect the analysis in any substantial way.

For an explanation of the terms effective information and surface
information, see Lamb (l966az4l-56).

 

CHAPTER IV

In this thesis, we have examined how a stratificational model of
language can be applied to Amharic, a language with a morphology based
in part on a system of consonantal roots which combine with vocalic
interfixes to form words.

In Chapter I, an overview of both the language under consideration
and the theoretical framework involved was provided; the main objective
of the thesis was stated as being the preliminary description of the
phonology of Amharic within a stratificational model basically the same
as that presented in Lockwood (1972). TO this end, it was noted that
to adequately analyze the phonology, at least a partial analysis of the
morphology is necessary. The infinitive, instrumental, and participle
forms of the basic stems of bi-, tri-, and quadriradical verbs were
chosen as data to fill this need.

It was also claimed that the stratificational treatment to be pre-
sented has several advantages over possible transformational-generative
approaches to the problem of interfixation. Transformational-generative
treatments of languages typologically similar to Amharic--Arabic and
Hebrew-—were examined and their Shortcomings discussed. Among these
shortcomings are the use of vowels in the underlying forms of roots,
which results in the proliferation of rules which can be avoided in the
stratificational approach, and the obligatory linear representation of

roots and their accompanying vocalic patterns, which can lead to rules

96

97

involving as-yet-unspecified mechanisms for marking vowels as internal
(part of the interfix) or not. In the stratificational treatment pro-
posed here, the representation of the root and its interfix as morpho-
logically simultaneous accomplishes the equivalent of such marking
without the use of'pg_ppp_boundaries or features.

Chapter 11 examined in more detail the problem of linearity with
respect to transformational-generative approaches to Semitic word—
formation. Two stratificational analyses, one of Arabic and one of a
small selection of data from Amharic were then discussed, and it was
observed that the two employed similar strategies in dealing with inter-
fixation. Both represent the root and interfixes as occurring simul-
taneously at the morphotactic level, with the proper interdigitation of
consonants and vowels being accounted for by the phonotactics.

The Amharic data was presented, followed by an analysis of the
morphological phenomena exhibited by the data. The analysis presented
involved both prefixation and suffixation as well as morphotactic si-
multaneity. The realizations of various morphemes were provided for,
alternations being conditioned by verb classes in the tactics. In
addition, a number of Special morphons were postulated in connection
with the phonological effects involved in the realization of the data.

Chapter III presented a detailed account of the behavior of these
Special morphons, examining their relationship to the phonotactics and
to neighboring elements. The bulk of Chapter III was devoted to deter-
mining the form which the phonotactics must take to both properly
account for the distribution of phonemes within the phonological word

and provide for the interdigitation of consonants and vowels which is

98

necessary to deal with the morphological analysis presented in Chapter
II. How the tactics processes words which are not based on the root/
interfix pattern was also discussed.

The problem of transformational-generative approaches to inter-
fixation was returned to, the observation being made that while the
types of rules used in that framework may be able to correctly account
for the data, they do not in themselves provide any explanation of why
natural languages should have such rules. In the stratificational
framework, the correct positioning of consonants and vowels in Amharic
follows as a direct consequence of the fact that the phonotactics must
process phonemes in sequence, just as human beings must sequentially
activate certain parts of the articulatory apparatus when they speak.

Chapter III also provided a componential analysis of the phonemes
of Amharic and discussed certain componential alternations as well.
Finally, through a series of tables which provided detailed illustra-
tion of the tactic and realizational phenomena involved, it was shown
that the analysis proposed in this thesis can indeed account for the
data concerning the infinitive, instrumental, and participle forms of
the root classes discussed in Chapter II.

Space limitations and a paucity of reliable data precluded a
satisfactory treatment of several phenomena, among them the phonemic
status of the vowel §_and the restrictions on word-final consonant
clusters; the problem of non-distinctive (allophonic) variation in
the realization of phonemes has regrettably been left untouched.

Nevertheless, in the sense that the objective of a preliminary

description of the phonology has been accomplished, and that certain

 

99

problems and their possible resolutions have been identified-~however
cursorily-—it is felt that this thesis provides the foundation for a

more complete stratificational description of Amharic.

 

 

 

 

LIST of REFERENCES

 

LIST of REFERENCES

Al-Azzawi, Theresa McGlaughlin
1973 Broken Plurals in Modern Iraqi Arabic. Unpublished Ph.D.
dissertation, Michigan State University.

 

Aronoff, Mark
1976 Word Formation in Generative Grammar. Cambridge: MIT.

 

Bar-Lev, Zev
MS "Hebrew Intramorphemics.”

Bender, M. L.
1968 Amharic Verb Morphology: A Generative Approach. Unpub-
liShed_Ph.D. dissertation, UnTVersity of Texas.

 

Bender, M. L. et al.

 

 

 

1976 Language in EthiOpia. London: Oxford University Press.
Brame, Michael K.
1970 Arabic Phonology, Unpublished Ph.D. dissertation, MIT.
Christie, William M., Jr.
1973 A Stratificational View of Linguistic Changg. Unpublished
Ph.D. dissertation, Yale University.
1976 ”Evidence Concerning Limits on Central Embeddings in Eng-

lish," Forum Linguisticum 1.25-37.

 

Cohen, Marcel
1936 Traité de langue amharique. Paris: Institut d'Ethnologie.

 

Dell, Gary S. and Peter A. Reich
1977 I'A Model of Slips of the Tongue," in The Third LACUS Forum
(R. J. Di Pietro and E. L. Blansitt, Jr., eds.). Columbia,
S.C.: Hornbeam Press.

Gleason, H. A.
1961 An Introduction to Descriptjve Linguistics (revised edition).
New York: Holt, Rinehart, and Winston.

 

Herrick, Earl Myron
1977 Extra-linguistic, Non-semological Conditioning of Catego-
rical Alternation. Unpublished Ph.D. dissertation, Michi-
gan State University.

 

100

 

 

 

lOl

Hetzron, Robert
1964 "La Voyelle du Sixieme Ordre en Amharique," Journal of
African Languages 3.179-190.

 

 

Isenberg, Rev. Charles William
1842 Grammar of the Amharic Language. London: Church Mis-
sionary Society.

Lamb, Sydney M.

 

 

l966a Outline of Stratificational Grammar. Washington, D.C.:
Georgetown University Press.

l966b ggrolegomena to a Theory of Phonology," anguage 42.536-

3.
Lamb, Sydney M. and Ralph Vanderslice

1976 "On Thrashing Classical Phonemics," in The Second LACUS
Forum (Peter A. Reich, ed.). Columbia, S.C.: Hornbeam
Press.

Leslau, Wolf
1968 Amharic Textbook. Berkeley: University of California
Press.

 

Lockwood, David G.

 

1972 Introduction to Stratificational Linguistics. New York;
Harcourt Brace Jovanovich.

1973a "0n the Nature of the Phonetic Stratal System." MS.

l973b "‘Replacives‘ Without Process," in Readings in Stratifi-

 

cational Lingpistics (Adam Makkai and David G. Lockwood,
eds.): University, Alabama: University of Alabama Press.

 

 

1975 "Universal Phonetics and Phonetic Contrast," in M.S.U.
Working Papers in Language and Linguistics 1.29-41.
1977 "Anatactic Relations in Grammar and Phonology." MS.

in press "Grammatical Conditioning in Stratificational Phonology,"
to appear in Forum Linguisticum.

 

Matthews, P. H

 

 

 

1974 Morphology. London: Cambridge University Press.
Praetorius, Franz
1879 Die Amharische Sprache. Leipzig: Halle.
Reich, Peter A.
1970 A Relational Network Model of Language Behavior. Un-
published Ph.D. dTSsertatTOn, University of Michigan.
1975 “Visible Distinctive Features," in The First LACUS Forum

 

(Adam and Valerie Becker Makkai, eds.). Columbia, S.C.:
Hornbeam Press.

1976 "Toward a Sequential Onset Feature Phonology of English,”
in The Second LACUS Forum (Peter A. Reich, ed.). Columbia,
S.C.: Hornbeam Press.

 

 

 

102

Rogers, Henry E.
1967 The Phonology and Morphology of Sherbro. Unpublished Ph.D.
dissertation, Yale University.

Schane, Sanford A.
1973 Generative Phonology. Englewood Cliffs: Prentice-Hall, Inc.

 

Sullivan, William J.

1969 A Stratificational View of the Phonology and Inflectional
Morphology of Russian. Unpublished Ph.D. dissertation,
Yale University.

1975 "Alternation, Transformation, Realization, and Stratification
Revisited," in The First LACUS Forum (Adam and Valerie Becker
Makkai, eds.). Columbia, S.C.: Hornbeam Press.

1976 "Abstractness, the Syllable, and the Fleeting Vowel in Rus-
sian," in The Second LACUS Forum (Peter A. Reich, ed.).
Columbia, S.C.: Hornbeam Press.

 

1977 "A Stratificational View of the Lexicon," Language Sciences
46.11-22.
l978 "Raising: A Stratificational Description and Some Meta-

theoretical Considerations," in Proceedings of the Second
Annual Linguistic Metatheory Conference (1977). Depart-
ment of Linguistics, Michigan State University, East Lansing,
Michigan 48824.

 

Sumner, Claude

1957 Etude expérimentale de l‘Amharique moderne. Addis Ababa:
The University College Press.

Ullendorf, Edward
1955 The Semitic Languages of Ethiopia, a Comparative Phonology.
London: Taylor‘s (Foreign) Press.

 

APPENDIX

APPENDIX

In this thesis, the following system is used to transcribe Amharic:

l. Vowels have their standard IPA pronunciation, with the excep-
tion of 3, which represents the schwa [a], and p, which repre-
sents a high, back unrounded vowel [w].

2. Glottalized consonants are indicated with a subscript dot, with
the exception of the glottalized velar [k?], which is tran-
scribed g,

3. The alveopalatal sounds are marked with a wedge, so that §_is
IPA [I], g_is IPA [d3], §_is IPA [if], fi_is IPA [n], and so on.

4. Consonantal length is indicated by either writing the consonant
twice or by using the Sign 33 as in wanna/wan:a 'principal.‘

 

103

IIIIIIIIIIIIIIIIIIIIIIIII|II|IIIIIIIIIIIIIIIIIIIIIIIIIIII