, . . . . , . ..V .V. .V ,
., . . A A V .. .sz
, V , V . V,.. V .A
‘ V , V

 
 

 
 
 

 

 

 

 

 
 

 
 
 
 

 

 
 

 

 

 
 

 
 
 

 

 
 

 

 

 

 
 
 
 
 
 
 

H . . . . . I
. V , , V .V V
N .. 2 V V V V .
., . . I V , V. . V \. V .. V. A . .V . , V. V V V

. . Z. . ... .. V. :.. .V. . . .V I
. V ,A , , . V . . , . . V, ._ . q! . _ V . . . . ...V V V I : .. 1 . , . . . ,., ., . .. ... V V . V . . V V. V

WlESVE

1

MI HIG ISTATE WE SITY LIBRARIES

Hi Ilillllifllwlllulll/Ill/ll!!!Ill/ll

31 93 00906 4456

 

 

 

 

M11

 

I'.

This is to certify that the
thesis entitled

FINAL DEVOICING AND ORTHOGRAPHICAL
INTERFERENCE IN GERMAN

presented by

Jonathan Blaine Abuhl

has been accepted towards fulfillment
of the requirements for

 

M.A. degree in Linguistics

gm ML

Major professor

Date June 5, 1992

 

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

LIBRARY T’
Michigan State
nlversity

 

%

__

PLACE IN RETURN BOX to ram

ovo this checkout from your record.
TO AVOID FINES

Mum on or baton duo duo.

DATE DUE DATE DUE DATE DUE

     
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU is An Affirmative Action/Equal Opportunity institution
china-m 7

 

 

 

 

 

 

 

 

 

 

FINAL DEVOICING AND
ORTHOGRAPHICAL INTERFERENCE
IN GERMAN

BY
Jonathan Blaine Abuhl

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS
Department of Linguistics and Germanic,
Slavic, Asian & African Languages

1992

4W7r'
’,V>/

’)
:7?

7}?

ABSTRACT
FINAL DEVOICING AND

ORTHOGRAPHICAL INTERFERENCE
IN GERMAN

BY
Jonathan Blaine Abuhl

The final devoicing rule of German is somewhat difficult for
American students to learn. This thesis reviews the theory
of markedness and how it relates to final devoicing and
reports two experiments which examine the relationship.

The participants in both experiments were American students
of German. The second study had more participants and
produced significant results. Two main factors were
analyzed: task and class level of subjects, who were
presented with pictures, words and a text to recognize or
read accordingly. The subjects were recorded and their
responses were checked for final devoicing. The responses
in the picture phase were more likely to be devoiced than
those in the word-reading phase. The responses in the word-
reading phase were also more likely to be devoiced than

those in the text-reading phase.

T0 LAURA

iii

ACKNOWLEDGMENTS

Many people have been instrumental in the completion of this
thesis, without whose help I never could have finished.
First, I would like to thank my committee members: Dr.
Hudson for his guidance concerning a topic and for last
minute comments before the final copy, Dr. Falk for her
initial steering of the proposal and direction in writing
and Dr. Lovik for his direction in German linguistics and
his avant garde attitude. I would also like to thank two
individuals instrumental in making possible the experiment
of this thesis: Dr. Dennis "Stats wizard" Preston for his
help with the statistical analysis and Dennis Huffman for
his assistance in the language lab. My family has also
played a large role in the accomplishment of this work.
Thank you Mom and Dad for all your investments in my life,
thanks to Jeff and Janet and Karel for all the encouragement
they've given me and for believing in me. Above all, I wish
to thank my wife Laura for her patience and advice and
encouragement. Her support and love have kept me from

quitting. Thank you Laura.

iv

Table of Contents

List of Tables ........................................... v

IntrOduction.00....000000000000000.00.0.0.000. 0000000000 .1

Chapter 1. Final Devoicing in German

1. O IntrOdUCtiono 0 0 0 0 . . 0 0 0 0 0 0 0 0 0 . 0 0 0 0 0 0 0 . 0 0 0 . 0 . 0 0 0 0 0 3
1 0 1 Final DeVOiCing 0 0 0 0 . 0 . 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0000000000 3
1.2 Theories Concerning Final Devoicing.............5

Chapter 2. Markedness

2.0 IntrOduCtionoooooooooooooooooooooooooooo000000007
201MarkedneSSoooooooooooooooooooooooooo ..... 00000.08

2.2 Con

2.1.1 Child Language Acquisition...............9
Adult Learners of English................9
Japanese Writing System.... ...... .......10
Universals..............................11
Voice Contrast Hierarchy.... ...... ......12
Physiological Support....... ......... ...13

2. .

2. .

2. .

2. .

2. .
lusion.............................. ..... ..13

()h‘Hi‘P‘H
minasuaw

Chapter 3. Language Learnability

UUUUUUUUUU

UGQO‘U‘I-fiUNI-‘O

Introduction........... ................... ...15
Contrastive Analysis Hypothesis (CAH)..........15
Language Similarity............ ..... . ...... ....16
Hierarchy of Difficulty........................18
Orthographical Interference....................19
The Honitor....................................20
The Voice Warp.................................21
Direction of Difficulty........................22
Harkedness Differential Hypothesis (MDH).......23
Conclusions....................................23

Chapter 4. The Pilot Study

4.0 Introduction...................................27
4.1 Hethodology....................................27
4.2 Results........................................29
4.3 Hypotheses. ................................... .30

Chapter 5. The Final Experiment

5.0 Introduction...................... ..... ........32
5.1 Subjects ......................................32
5.2 Tasks..........................................33
5.3 Place of Articulation...... ..... ..... ....... ...36

Chapter 6. Results

6.0 Introduction................... ............. ...37
6.1 Data...........................................37
6.2 Statistical Interpretation.....................40
6.1.0 Introduction............................40
6.1.1 STEPWISE VARIABLE RULE ANALYSIS:
Three Factor Groups.................41
6.1.2 STEPWISE VARIABLE RULE ANALYSIS:
Two Factor Groups...................42

Chapter 7. Conclusions
700 IntrOduction000000000.0000.0000000.00.00000000046
7.1 Suggestions for Future Research................46
7.2 Additional Observations........................48
7.3 conC1uSion00.0.000.000000000000000...00.00.000.50

List of References...... ............................. ...52

vi

LIST OF TABLES

2-1. Examples of Final Obstruents......................4
3-1. Stockwell and Bowen's Hierarchy of Difficulty....18
4-1. Pilot Test Results........... ......... ...........29
6-1. Beginning Level Results......... ..... ............37
6-2. Intermediate Level Results.......................38
6-3. Beginning Level Devoicing Frequency..............39
6-4. Intermediate Level Devoicing Frequency...........40
6-5. First VARBRULE Data... .......... . ..... . .......... 41

6-6. Second VARBRULE Data........... ............... ...42

vii

Introduction

To aid in the understanding of foreign language
acquisition, it is useful to examine certain recurring
phenomena which students typically encounter in learning a
foreign language. Each foreign language has its own unique
obstacles which present themselves to the native English
speaker. The German language is no exception. One
troublesome phenomenon in German is what is commonly known
as final devoicing. Acquisition of the rule of final
devoicing of obstruents by American students is relatively
problematic.

There are two main foci in this thesis: First is the
markedness of the rule of final devoicing. Second is the
presentation of an experiment that will give support for the
hypothesis of orthographical interference in a native
English speaker's acquired final devoicing of German. The
hypothesis is that the final devoicing rule, as acquired by
native English speaking students of German, is influenced by
German orthography. This influence is manifested in
different degrees depending on the task which is used to
elicit the students' verbal responses. Each task used to
get a response focuses on a different variable to influence
devoicing: markedness, orthographical interference and
monitoring. There are three predictions that can be made
based on these variables. When recognizing a picture,

students will be most likely to devoice. When reading a

2
word-list students will be less likely than with the picture
presentation to devoice properly. When reading a text
students will be the least likeliest to devoice properly.
The reasoning for these predictions can be found in chapters
two and three. Hypotheses are foralized in chapter four.

A brief explanation of the German language concerning
the problem of devoicing is presented in Chapter 1. Chapter
2 introduces the theory of markedness and the relationship
between devoicing and markedness. Language learnability is
discussed in Chapter 3, where it is argued that final
devoicing is relatively easier to learn by native English
speakers than final voicing is by native German speakers.
Orthographical interference and response time are also dealt
with in this chapter. Chapter 4 introduces a pilot study,
examining the effects of orthography on those native English
speakers who have acquired the final devoicing rule.

Because of the unrefined nature of the pilot study, a more
in depth look at this subject was warranted, Thus, Chapter
5 describes the modified experiment, and Chapter 6 its
results and the statistical interpretation of those results.
Chapter 7 is reserved for the author's post-experiment
comments, and additional remarks, including the realization
that the final experiment was inconclusive and that the

hypothesis is worthy of further research.

Chapter 1
Final devoicing in German
1.0 German, as well as other languages including Russian
and Turkish, has devoicing of final obstruents. All voiced
obstruents are devoiced in final position. The obstruents
are the fricatives and the stops. Final position is before
a word or syllable boundary. The controversy concerning the
syllable boundary, while briefly mentioned, will not be
examined in this thesis.
1.1 Final devoicing in German
The relationship between German spelling and
pronunciation is quite straightforward, leaving the student
with few exceptions to learn. However, one exception
concerns final devoicing. In final position, although
obstruents are pronounced without voice, the orthography may
show a voiced grapheme. That devoicing is word-final is
uncontroversial. Whether devoicing is syllable-final
(Benware 1986:66) is, however, the focus of some debate.
Vennemann (1972:13) shows how some German words can be
"resyllabified"; the final phoneme of one syllable can re-
attach itself to the beginning of the following syllable.
This would result in the voicing of the otherwise devoiced
obstruent. The author has also observed this phenomenon
with a pausal schwa: a native German speaker said the
normally devoiced und /unt/, but followed by a pausal schwa.
This created the bisyllabic phonological word lundal.

4

Only word-final devoicing is considered here. The
controversy surrounding syllable-final devoicing and the
definition of the syllable are not dealt with in this study.

In the first of the examples below, the German word for
gay, /tak/, is pronounced /tagd/, with the plural
inflection. These words are just a few examples that point
to the fact that German has no final voiced obstruents, and
that some final voiceless obstruents correspond to voiced
obstruents followed by suffixes, e.g. the plural inflection.
Attention must be given to the final obstruent phoneme with
and without inflection in order to recognize the final
devoicing phenomenon. The last three examples show that the

rule is of final devoicing, not prevocalic voicing.

Tabla 2-1 Examples of Final Obstruonts
WWW Wu

day Tag [tak/ Tage [taga/
half halb lhalp/ halbe [hale/
child Kind [kint/ Kinder [kinda/
forest Wald /va1t/ walder Ivelda/
house Haus [haus/ Hauser [hoyza/
nerve Nerv Inerf/ Nervds /nerv6:s/
bed Bett [bet] Betten [betan/
casual salopp /zalop/ saloppe /zalopd/

cupboard Schrank /érank/ Schranke [Srenkdl

5
1.2 Theories concerning final devoicing

The following is a short introduction of theoretical
aspects of final devoicing, including formalizations and
other considerations. Brakel (1983:4) gives his version of
an acceptable representation of the final devoicing
phenomenon. He uses two different words as examples:bund
'colorful' [bunt/ and Berg 'mountain' /berk/.

Using distinctive features and the generative

approach, a rule such as

[+occlusion]-»[+surd] / ___#

could be written and interpreted as either a

phonotactic constraint of certain Germanic and

Slavic languages or as a rule converting voiced

stops to voiceless ones in word final position.
Employing more main-stream terminology, Eckman (1981:197)
presents the rule:

[-sonorant]-[-voice] / ___#

Many introductory linguistic textbooks use German final
devoicing as the prime example of neutralization.
Neutralization refers to the suspension of a phonemic
contrast. ”If a phonemic contrast generally observed is not
found in a given enviroment, it is invariably a two-way
opposition that is involved and normally it is the unmarked
member of such an oposition that occurs in the PLACE OF
NEUTRALIZATION" (Katamba 1989:39). Neutralization is
influenced by markedness. "The concept of markedness which
thus appears for the first time in Prague school phonology

is tied directly to the nature of neutralization" (Anderson,

1985:110). Markedness will be discussed further in Chapter

two.

Recently, Dinnsen (1985:267) has disputed the claim
that German devoicing is neutralization, and he notes
research by Port, Mitleb & O'Dell (1981) and O'Dell & Port
(1983) to support his claim

that the underlying distinctions are preserved

phonetically. In particular, it was found that a

comparison of vowel durations before underlying

voiced and voiceless obstruents word-finally

showed longer vowel durations (by approximately

10%) before underlying voiced obstruents.

Moulton (1962:50) implies that orthography, i.e. voiced
graphemes, might be the cause of certain phonetic
distinctions, the voicing of a devoiced grapheme. However,
Dinnsen and Charles-Luce (1984) have shown by their study
that this is not the case in at least one language. Catalan
exhibits final devoicing of obstruents without the
representation of the voiced grapheme, e.g. <t> is written
instead of <d> for devoiced /d/. Conversely, German words
with underlying voice do exhibit the voiced grapheme. The
study by Dinnsen and Charles-Luce in no way draws into
question the existence of the phenomenon studied in this
thesis. The auditory difference of voice in the word-final
stop in German is measured in this thesis. Exact acoustic
measurements and their interpretation are left for those who

have the necessary apparatus. The equipment used in these

experiments consists of a tape recorder and the human ear.

Chapter 2
Markedness

2.0 Introduction

Trubetzkoy postulated the construct of markedness which
has survived almost 50 years of testing and examination.
Many areas of Grammar have been considered that give support
to his original construct which can be seen in even the most
recent publications. The way a child learns his/her first
language as well as the way an adult learns a second
language gives evidence that final obstruent devoicing is
unmarked. The Katakana orthographical system of Japanese
points to the voiced obstruent character as being more
marked than its voiceless counterpart. Statistical tenden-
cies and implicational universals in languages of the world
also show support that final devoicing is less marked and
the "voice contrast hierarchy" gives evidence that voice is
more marked. Finally, the physiological simplicity of
devoicing points to the fact that voiced stops are more
marked than voiceless stops. The study of markedness and
its relationship with final devoicing are vital to this
thesis. Since final devoicing is less marked, without the
introduction of any other intervening variables, its
acquisition could be assumed to be relatively easy. This
chapter defines markedness and gives many examples that

support this assumption.

2.1 Markedness

Trubetzkoy first adopted the term Markedness.

The question whether the 'strong' or the 'weak'

opposition member of a correlation based on the

manner of overcoming an obstruction of the second

degree is unmarked can, in the final analysis, be

determined objectively only from the functioning

of the particular phonemic system. However, in

any correlation based on the manner of overcoming

an obstruction a 'natural' absence of marking is

attributable to that opposition member whose

production requires the least deviation from

normal breathing. The opposing member is then of

course the marked member. (Trubetzkoy 1969:146)
There are two different forms of markedness: logical and
natural. A logically marked phoneme receives its "status
for reasons internal to the phonological system of the
language, and cannot be determined by phonetic criteria"
(Anderson 1985:110). Anderson uses the final devoicing of
German and Russian as an example of logical markedness. The
devoicing is not obligatory because of any phonetic
characteristic, e.g. the preceding phoneme or consonant
harmony, but because of a boundary. A naturally marked
phoneme is that phoneme of a pair that possesses a phonetic
quality that distinguishes it from its partner, e.g. has a
further deviation from normal breathing. For example, the
voiced phoneme of a voice/voiceless pair is the marked
phoneme. Normal breathing takes place without voice. The
phoneme with the most phonetic characteristics, i.e.
distinctive features, is the marked phoneme.

In 1962 Roman Jakobson, in an article on Ancient Greek

Prosody, gave his definition of markedness: "a marked

9

category tends to be interpreted in relation to the unmarked
one as a compound, complex category opposed to a simple one"
(p. 266). The similarity in their definitions and the fact
that Jakobson and Trubetzkoy were colleagues is just more
evidence that Trubetzkoy positively influenced Jakobson and
many other linguists with his time-tested theory.

Trubetzkoy's influence can be seen in even the most
recent publications. For example, Katamba (1989:89) employs
a definition of markedness as follows: "what is NATURAL can
be said to be UNMARKED and what is not natural can be said
to be MARKED, i.e. in some sense unusual." He also says
that markedness is relative, i.e. dependent on
circumstances.
2.1.1 Child language acquisition

More support for the markedness of final voiced
obstruents can be found in the field of child language
acquisition. Ingram observed the speech of children from
1;6-4;0 years of age. Although he discovered a wide range
of variation among the children, he was able to conclude
that the obstruents tended "to be...devoiced at the end of a
syllable." (1986:227) For example, a child might say /bit/
instead of /bid/ for head. We are not sure, however,
whether this is for physiological or other reasons.
2.1.2 Adult learners of English

Altenburg and Vago (1987) also support final devoicing

(as a less marked phenomenon by their observation of the

10

speech of two adult Hungarians who were learning English.

Subject A occasionally devoiced a word-final

obstruent even if the next word began with a

voiced sound. Thus, we find examples like band

increases ($16) and end of (S6), where the last

consonant of the first word is phonetically

voiceless. This process is not obligatory... The

converse, i.e., word-final voicing before a

following voiceless sound, is not found. (p.153)
The important factor in this example is that neither English
nor Hungarian has final devoicing. Nevertheless, these
speakers often devoiced. This lends credence to the theory
that final voicing is more marked.
2.1.3 Japanese writing system

In the Japanese writing system Katakana, the
voice/voiceless distinction is made apparent in the
orthographical system by the addition of a diacritic,
i.e. ". The /V/ represents all possible vowels that can
follow that consonant, e.g. /ka/, /ki/, /ku/, /ke/, /ko/,
etc. The voiced counterpart of /kV/, i.e. /gV/, is achieved
by the addition of ". Likewise /tV/ becomes /dV/ and lsV/
becomes /zV/. This makes perfect sense in light of
Jakobson's definition of a markedness as a "compound,
complex category". The voiced category is the one with the
additional element, making it orthographically more complex.
This distinction is not made with the spelling of /bV/ and
/pV/. Both of these bilabial characters receive different
diacritics, which makes them both more complex than the

underlying unmarked character of /hV/. The /bV/ receives "

and the /pV/ receives ‘.

11

2.1.4 Universals

Greenberg (1963:xix) noted that there are implicational
universals, of which:

It is asserted universally that if a language has

a certain characteristic, (0), it also has some

other particular characteristic (u), but not vice

versa. That is, the presence of the second (u)

does not imply the presence of the first (a).
Concerning voicing of obstruents, if a language has voiced
obstruents, it has voiceless obstruents, but not vice versa
(Eckman 1984:79). Eckman goes on to give one example of a
non-implicational universal: "In all languages, there are
least two color terms." There is no implication here. This
is simply a fact about the languages of the world.

An example from syntax concerning agents in passive
constructions: some languages, such as Persian and Arabic,
have no expressed agents in their passive voice
construction. Other languages, such as English and German,
have the option of expressing the agent. If a language has
expressed agents in the passive, it will also have passive

constructions without an expressed agent. A [+agent]

construction implies the [~agent] construction, but not vice

versa.
more marked -------------------------- less marked
expressed agent -------------- no expressed agent

Regarding phonology, a language such as Korean has no voiced

12
stop phonemes; other languages, like English, do. If a
language has voiced stops, it will have voiceless ones. The
reverse is, however, not true. A [+voiced] obstruent

implies the presence of a [-voiced] obstruent.

more marked ----------------------- less marked

[+voiced] obstruent ------- [-voiced] obstruent
2.1.5 Voice contrast hierarchy

In 1977 Eckman conveyed an additional universal that
points to the voiceless obstruent of a voice/voiceless pair
as being the more marked, the "voice contrast hierarchy."
If a language has word-initial voiced obstruents, it will
have word-initial voiceless obstruents, but not necessarily
voiced obstruents in the medial or final positions.
Similarly, if a language has word-medial voiced obstruents,
it will also have a word-initial voiced obstruents as well
as word-medial and word-initial voiceless obstruents.
Finally, if there are word-final voiced obstruents, there
are word-initial and word-medial voiced obstruents as well
as voiceless obstruents in all three positions. He
clarifies this markedness phenomenon with the following

diagram (p.322).

13

 

Initially less marked
Med ally
Finally more marked

This shows that the word-final position is more marked
because the existence of a voiced consonant in this position
predicts the existence of voiced and voiceless consonants in
all the other positions.
2.1.6 Physiological support
Support for the markedness of final voicing can also be
found in the explanation of what happens to a person
physically during voicing as opposed to not voicing. Flege
(1982:118) claims that English speakers have to learn to
suppress their (learned) final voicing, if they wish to
accurately acquire a language such as German. He explains
devoicing from a physiological perspective.
Speech researchers generally agree that devoicing
is a "natural" consequence of aerodynamic factors
tending to extinguish the transglottal pressure
difference necessary for glottal pulsing during
the closure interval of final stops. To produce
fully voiced stops an English speaker may need to
learn some means of diminishing the rate at which
pressure increases during stop closure intervals.
lFrom this description we can say that devoicing is more
natural, easier and therefore less marked.
2.2 Conclusion

The concept that final devoicing is less marked than

fidnal voicing is overwhelmingly supported by the previous

14
sections. Each section illustrates the fundamental
definition of markedness. Since final devoicing is less
marked, the prediction is that English-speaking students of
German will have little problem learning to devoice
properly. This prediction is based on the assumption that
if something that is simpler and less complex than an
already existing ability, it will be quite simple to learn.
This phenomenon is tested in the first task of both
experiments, i.e. the students are asked to name a picture.
Some of the pictures employ the final devoicing rule.
Although this is not observed here, the voluntary German
speech of the students could be predicted to be
appropriately devoiced because of the main influence of

markedness, when no other variables are introduced.

Chapter 3
Language Learnability

3.0 Introduction

The contrastive analysis hypothesis (CAH), although
both logical and controversial at the same time, is very
helpful when making predictions about the difficulties that
second language learners will experience. Many theorists
have attempted to predict the relative degree of difficulty
of a certain language to be learned. The relative
markedness of one language in comparison with the other
would reasonably play an important role in its learnability
by speakers of the other language. The markedness
differential hypothesis (MDH) (Eckman 1977) combines both
the concepts of the contrastive analysis hypothesis and
markedness to produce a highly productive, insightful
construct. Stockwell and Bowen's (1965) hierarchy of
difficulty can give conflicting predictions concerning the
ease of final devoicing acquisition. Orthographical
interference and monitoring are also major factors in final
devoicing acquisition. These theories and phenomena are
discussed in this chapter.
3.1 The Contrastive Analysis Hypothesis (CAM)

Lado (1957:1) said, "in the comparison between native
and foreign language lies the key to ease or difficulty in
foreign language learning." Acceptance of the CAH is quite

easy, since it appears very logical. Since German has no

15

16

final voiced obstruents and English does, according to the
CAH, a native German would have difficulty learning to voice
the final obstruents of English and a native American would
have difficulty learning to devoice the final obstruents of
German. However, the CAH makes no prediction about the
relative degree of difficulty from one certain language
being learned to the other; neither language learner would
be said to have a more difficult task. As will shortly be
discussed, the final voicing that must be learned by the
former group is theoretically more problematic than the
final devoicing that must be learned by the latter. This
does not suggest, however, that final devoicing is easy,
rather it is easier when compared to the task of final
voicing encountered by native Germans learning English.
3.2 Language similarity

"An 'interlanguage' may be linguistically described
using as data the observable output resulting from a
speaker's attempt to produce a foreign norm, i.e., both his
[sic] errors and nonerrors" (Selinker 1969:69). Corder
(1978:86) sees the interlanguage process as one that is
affected by the relative difference or similarities of
target and native languages. Depending on the languages'
nearness, the interlanguage will be formed by either a

”restructuring" or a "recreating" strategy.

17

If we believe that another language is only

distantly related structurally to our own (e.g.,

Chinese-English) we may opt for a recreative

strategy, starting from scratch, on the grounds

that the path to the target will be shorter,

while if we believe that the target is closely

related, (e.g., Danish-English) we may prefer a

restructuring strategy as being more economical.
Regardless of the strategy, he believes that the language
learner acquires simplified forms of the target language
first. The two pairs of German phonemes listed below
exemplify the simple and difficult distinction. The
phonemes on the right are simpler than on the left.

lfl/ ------ /1/
Id/ ------ It/

The /0/ has every feature that the /i/ has plus lip-
rounding. Similarly, the /d/ has every feature that the /t/
has plus voice. This is based on the assumption that lip-
rounding and voicing are always more "marked". In fact, the
more complex items are learned later than those that are
simpler. This assumption might be argued, depending on the
languages in question. In an imaginary language with all
rounded vowels except one, the unrounded vowel could be
argued in one sense as marked. This shows that the concept
of markedness is often "relative" (Katamba 1989:89). Since
I will be dealing with two Germanic languages, it is safe to

assume that in the voice/voiceless distinction the voiced

phoneme is marked.

18

3.3 Hierarchy of difficulty

Stockwell and Bowen (1965:10) have noted a "Hierarchy
of Difficulty” based upon "the conditions of negative,
positive, and zero transfer". They compare English and
Spanish but their system can easily be adapted to apply to
English and German. They have set up a three-way
correspondence using the conditions of On optional, 92
obligatory, and 9 zero. These conditions refer to the
occurrence of certain phonological phenomena in the
languages under contrast. The same phenomena are considered
to predict whether there is negative, positive or zero

transfer. Their hierarchy is written as follows:

Table 3-1: Stockwell and Bowen's Hierarchy of Difficulty

 

 

 

 

 

 

 

 

 

 

 

 

 

DIEFICULTY COMPARISON
Magnitude Order L1 L2
I 1 0 Oh
2 0 Op
3 Op Ob
II 4 Ob Op H
II 5 Ob
i 6 0p
III I 7 0p Op
‘ I 8 Oh Oh

 

An example of the most difficult correspondence, English 9
and German Qh, would be the phoneme /x/. English does not

employ this velar fricative while German does. This would

19
cause negative transfer at its extreme and would
theoretically be the most difficult to learn.

This hierarchy of difficulty lacks the ability to make
unambiguous predictions concerning magnitude of difficulty
relating to one phenomenon. For example, if our comparison
looked at the graphemic presentation of both languages, we
might say that English (L1) is obligatory in its
representation of final stops by their graphemic
counterparts, e.g. the voiceless velar stop is always
represented by <k>, however this same stop is represented by
either <g> or <k> in German (L2). For English this is
obligatory, but for German this is optional. This analysis
would put the final devoicing rule in the II magnitude, 4th
order of difficulty. However, a manipulation of the formula
can put the results in magnitude I, lst order, e.g. English
has no final devoicing = 0, German has obligatory final
devoicing = ob.

3.4 Orthographical interference

A major problem studied in this thesis is the fact that
native English speakers have to devoice a final stop, even
though the final grapheme continues to show voice. In
English the final graphemes <b>, <d> and <g> are all
pronounced with voice. In German they are devoiced. Lado

(1957) pointed out this problem as follows:

20

the student transfers the habits from his [sic]

native language to the foreign one, he [sic] will

have difficulty both in reading and in writing the

foreign language whenever a symbol the same as one

in his [sic] own language represents a different

sound in the foreign language.
More recently, Brown and Haynes (1985:20) say:

the skills of L1 reading may interact with the

acquisition of L2 reading skills, with the

possibility that transfer or interference may

occur as a function of the pattern of similarities

and differences between the two languages.
3.5 The Monitor

The Monitor hypothesis (Krashen and Terrell, 1983) was
postulated to account for some variation in performance in
the area of second language acquisition. The monitor is
used as an "editor" of a second language learner's output
when this output is not spontaneous or initiated by the
speaker. A student is able to monitor his/her output when
three criteria are met. First, the performer has to have
enough time. Second, the performer has to be thinking about
correctness, or be focused on form. Finally, the performer
has to know the rule. The Monitor hypothesis has elicited
both support as well as opposition. Support can be found in
Variation in Interlanguage (Tarone 1988:70).

Still other studies have followed from the

Labovian approach (cf. pp. 39-44, in which

'attention paid to speech' plays a crucial role in

causing the individual to styleshift from a casual

to a more careful speech style).
Opposition can be found in Hulstijn and Hulstijn (1984).
They found that the Monitor played no role in reducing the

amount of grammatical errors of second language learners.

21
Their findings do not, however, affect the hypotheses or
results of this thesis, since their focus was grammatical
errors and not pronunciation errors. I assume that the
Monitor functions, or fails to function, differently
depending on which area of SLA is being observed.
3.6 The voice warp
Merritt Ruhlen (1986:18) points to an interesting
phenomenon that he calls the "Voice Warp". He says that the
contrasts between voiceless and voiced stops is one of the
most common among the languages of the world. The following
shows the usual pattern:
p t k
b d g
Gamkrelidze (1975), as quoted by Ruhlen, "observed, however,
that if there is a gap, or hole, in the system it is far
more likely to be p or g than any of the other four". Place
of articulation may have an almost obvious role. The stop
phoneme with the largest area between the point of
articulation and the vocal cords may be more readily voiced,
i.e. the bilabial_§tgp. Conversely, the phoneme with the
smallest area, i.e. the yela; stop, may be more readily
voiceless. A stop can only be voiced prior to its release.
After its release, it takes on the acoustic characteristics
of the following phoneme. The bilabial stop has a greater
tendency to be voiced because it has a larger cavity into

‘which this prereleased-voiced air can be released than the

22
velar stop which has a smaller cavity. This is not contrary
to Eckman's "voice contrast hierarchy”. Eckman makes no
reference to place of articulation.
3.7 Direction of difficulty
Moulton (1962:50) made some interesting observations
regarding direction of difficulty. Moulton refers to the
final devoicing phenomenon among the German speakers as a
"habit”.
This habit is so firmly fixed that when [a speaker
of German] learns a foreign language, he [sic]
finds it very difficult to pronounce voiced
obstruents in one of these positions.
Later Moulton expresses the major hinderance of the native
English speaker learning German,i.e. orthographical
differences. This problem area also turns out to be one
focus of my experiment.
Fortunately, if the student refuses to let himself
[sic] be misled by the spelling, the German
alternation of voiced and voiceless obstruents is
not hard to learn.
His hypothesis of direction of difficulty is based on the

fact that German has no final obstruents with voice, while

English does. We can see this in the following examples:

German- No final voicing ----------

English-4 Final voicing mph [mab/
19.32 Idid/
league llig/

23
Without stating it directly, he laid a basis for what was
later to be called the Markedness Differential Hypothesis
(MDH) .
3.8 Markedness Differential Hypothesis (MDH)

Now that we are familiar with the devoicing phenomenon
in German along with the role of markedness, we can accept
an amendment to the contrastive analysis hypothesis which is
the markedness differential hypothesis (MDH).

The areas of difficulty that a language learner

will have can be predicted on the basis of a

systematic comparison of the grammars of the

native language, the target language and the

markedness relations stated in universal

grammar... (Eckman 1977:321)

With this construct, the prediction can be made that the
native German speaker would have more difficulty learning to
voice final obstruents than the native English speaker would
have learning to devoice the final obstruents. The more
marked the phenomenon is, the more difficult it is to learn
(all other things being equal). Such a claim is supported
by a study by Dinnsen and Eckman (1975) as cited by Eckman
(1977:321).

3.9 Conclusion

The difficulty that native German speakers have when
trying to learn to voice the final obstruents of English has
already been established (Eckman 1977). As a teacher of
German to native English speakers, I have also noticed that

the final devoicing rule of German remains somewhat

problematic. In Table 2-1 the obvious difference between

24
German orthography and German pronunciation concerning final
obstruents was pointed out. In much the same way, the
plural <s> might seem difficult to a native speaker of
German when Americans pronounce it as /s/ or /z/.

The voice warp might predict that some stops are more
difficult to devoice than others. The /g/ might lend itself
to easier devoicing because of the small area into which the
prereleased, postglottal air can flow. Conversly the /b/
might be more difficult to devoice because of its large
cavity. The assumption here is that the /b/ is easier to
make than the /g/. Spelling, however, offers a
complication. A native speaker of English has learned to
always read the final voiced grapheme with voice. However
in German all final voiced obstruent graphemes are to be
read without voice.

After this chapter, some conflicting predictions might
be made concerning the acquisition of the final devoicing
rule of German by American students. The CAH and its
modification, the MDH, would predict its ease of acquisition
by American students because final devoicing is less marked
than final voicing. Those theories supporting the
orthographical interference explaination would predict a
disruption in the theoretically easy acquisition of final
devoicing. The hierarchy of difficulty could predict the
ease of difficulty depending on the wording of the formula.

The monitor model also plays a part in whether a student

25

will correctly devoice the obstruents in the final position.

I believe that all theories represented above can be
combined to make clear predictions about possible responses
concerning final devoicing. When a student is asked to give
a verbal response without any written stimulus, he/she will
more than likely devoice properly. This can be predicted on
the basis of the markedness differential hypothesis and the
availability of time to focus on form. When a student is
asked to read a word that employs the final devoicing rule,
he/she will not devoice as often as in the earlier phase,
because of the orthographical interference, but will
continue to devoice some. The student will voice some of
his/her previously devoiced phonemes because of the presence
of the final voiced graphemes. The reason some final
devoicing continues could be explained by the monitor model.
With the single word the students have the time to focus on
form and to remember the rule of final devoicing. When a
student reads a text with words that employ the final
devoicing rule, he/she will decrease the number of devoiced
phonemes further yet. This can be explained by the presence
of the orthographical interference and the absence of the
time to monitor. When the students read a text, the flowing
speed that is used when reading sentences out loud precludes
any chance to focus on form, i.e. to monitor.

For this thesis, there was no experiment performed on

native German speakers learning English, It is therefore

26
impossible at this time to define what is "relatively easy"
and "relatively difficult". The following two chapters will
show, however that after two years of instruction,
approximately 88% of the English-speaking students observed

may have acquired some ability to devoice final obstruents

in German.

Chapter 4
The pilot study

4.0 Introduction

In order to show the difficulty of final devoicing
acquisition and to give a brief glimpse of the situation
concerning orthographical interference a pilot study was
conducted in the spring quarter of 1991. It also provided
an impetus for further study. The hypothesis was as
follows: In the beginning stages of language development
native English speaking American students learning German
will begin to devoice some final obstruents. However, when
presented with a printed text, they voice some of the
previously devoiced obstruents.

The null hypothesis is: German orthography has no
effect on the "Final Devoicing Rule” that has been acquired
as part of the interlanguage of an American learner of
German. The results suggest that Americans' ability to
devoice the final obstruent is interfered with or hindered
when they read text as opposed to picture recognition.

4.1 Methodology

The subjects consisted of eight American students in
the third quarter of the first year of German at Michigan
State University. There were two parts to the experiment:
picture recognition and the reading of a text. Each student
was first presented with eight unlabelled pictures of simple

vocabulary items. The student was required to recognize the

27

28
picture and give the German name for the pictured item. This
task was implemented to see if any acquisition of final
devoicing had occurred. Only three pictures had items that
employed the final devoicing rule. The other five
pictures were used as distractors. The following is a list
of the German words for the pictures used along with the
corresponding English glosses. The starred words are those

which require final devoicing.

Zug* train
Katze cat
Mann man
Hund* dog
Ball ball
Frau woman
gelb* yellow
Buch book

After the recognition exercise, each student was given
a short text to read aloud. The text contained simple
vocabulary, interspersed with words employing the "Final
Devoicing Rule". This task was given to observe any
possible orthographical interference. The following is a
copy of that text and its translation:

Ich kenne ein Verb,* das 'kaufen' heiBt. Ich

kenne auch eine Frau, die etwas kaufen will. Sie

mochte ein Fahrrad* kaufen, weil sie auf dem Berg*

wohnt. Sie mbchte auch einen Apfel kaufen, weil

sie Hunger hat.

I know of a verb that is called 'buy'. I also

know a woman who wants to buy something. She

would like to buy a bike because she lives on the

mountain. She would also like to buy an apple

because she's hungry.

Each student was tape recorded to avoid experimenter

bias. Only one-syllable words were focused on, so only word

29

final devoicing was tested as opposed to syllable final.
Only words with voice/voiceless distinction of the stops
/b/-/d/, /d/-/k/ and /g/-/k/ were used. The occurance of
the underlying /v/ and /z/ in German is relatively rare when
compared with the stops. I did not feel that the students'
vocabularies were advanced enough to expect the recognition
of pictures of items that end in either underlying voiced
fricative /v/ or /z/.
4.2 The results

The following table shows the frequency of word-final
devoicing (signified by the Y for yes) and the failure to
devoice (signified by the N for no). The judgements
concerning the success or failure to devoice were made by
Jonathan Abuhl. The dashes mean that the student did not
know this word or needed prompting.

Table 4-1: Pilot Test Results

Subject Recognition Reading

b/P d/t glk b/P d/t g/k
1. -- Y Y N N N
2. Y N N N N N
3. N N N N N N
4. -- Y N N N N
5. -- N N N N N
6. Y N N N N N
7. -- Y N N N N
8. N N N N N N

30

It is apparent in the recognition phase, that many of
the students had not acquired/learned the final devoicing
rule. Those that did show some devoicing tendency failed to
perform correctly for every voice/voiceless distinction.
However the surprising data appeared during the reading
phase. Not one student was able to hurdle the
orthographical discrepancy. Every student voiced the stop
in the final position.

The data of this pilot study are too few to do any in-
depth statistical analysis. However, the tendency to lose
any devoicing during the reading of a text is apparent
enough to launch another slightly modified test with more
subjects, to gain some significant data.

4.3 Hypotheses

The considerations of markedness predict that final
devoicing would be relatively easy. The considerations of
orthographical interference predict difficulty in final
devoicing. Considerations of monitoring predict variation
in devoicing depending on the task. I hypothesize, then,
that:

Final devoicing will occur when there is no

graphemic presentation (because devoicing is

unmarked and acquired). With the presentation of

words, the occurrence of devoicing decreases

(because of orthographical interference), less so

with single words (because monitoring is

31
possible).
Three tasks were developed to observe the above
mentioned factors, i.e. markedness, interference and

monitoring. These tasks are outlined in the next chapter.

Chapter 5
The Final Experiment
5.0 Introduction

The pilot study was just a springboard for a more
complex study that would provide more convincing evidence.
The mistakes and shortcomings in the pilot study were
revised in order to reduce the number of variables that
might have an effect on the data. The second study (spring
1992) took into account three different factors: subject's
level of study, task, and place of articulation.

5.1 Subjects

There were 42 subjects in two groups. One group
consisted of 25 students from two different German 103
classes. These students were in the final quarter of the
first year of German. The second group consisted of 17
students from three different German 203 classes. These
students were in the final quarter of the second year of
German.

Each subject was given the same instructions and test
items in the same order. Each subject was recorded in the
recording booth at the Wells Hall language lab. The
cassette recorder was a Yamaha K-1020 and the microphone was
an Electric voice, model 676. The judgements as to whether
final devoicing had taken place were made by Jonathan Abuhl.
After the student sat down, he/she was asked to read the

following directions out loud:

32

33

-No grade will be given for your performance.

-For the first task you will be shown some

sketches. The first two cards are
colors. Try to recognize the pictures
and give the German word which
represents them. If you don't know a
word, I will try to make the item in
question clearer, but don't worry if you
don't know some.

-For the second task you will have to read some

words that I have printed on some
cards.

-For the third task you will read a short passage.
There were two reasons for such a formal presentation of the
instructions. First, each subject received the same
instructions to avoid possible experimenter influence.
Second, while the student read, the lab technician was able
to set the recorder to the optimum peak levels to insure a
near-perfect sound recording. After the instructions, I
asked if there were any questions and we proceeded.

5.2 Tasks

For the first task, each student was presented with 14
pictures of simple vocabulary items. As in the pilot study,
this task was given to elicit responses withjout having to
introduce orthography. This allows for a closer study of

the acquired final devoicing rule without much interference.

34

Each subject was also given ample time to focus on form.

The names of six pictures employed the final devoicing rule,

two for each devoiced stop.

grapheme <b>, <d> or <g>.

These words end with the

The other eight words were used

to disguise the experiment's purpose from the subjects being

tested.

Some changes have been made in the words/pictures

that were chosen in the pilot study to avoid any possible

English cognates.

I avoided words which had obvious English

cognates with final voiced obstruents, e.g. flung, English

hand-

In the picture recognition task, approximatly four

seconds were alloted for each picture and response. The

following is a list of the pictures that were presented, in

the order in which they were presented to all subjects.

rot

gelb (devoiced)
Mann

StraBe

Geld (devoiced)
Stirn

Berg (devoiced)
Buch

Ball

Rad (devoiced)
Frau

Zug (devoiced)
Katze

Staub (devoiced)

red
yellow
man
street
money
forehead
mountain
book
ball
tire
woman
train
cat

dust

The second task introduced orthographical interference but

continued to allow time for the student to focus on form and

recall previous word final devoicing instructions.

task the students read words out loud from cards.

For this

The words

were written on cards to avoid the reductions and

assimilations that might take place if the words were read

35

too quickly from a list or a text as in the pilot study.
All of the words which were presented in the picture
recognition task and required final devoicing were also
presented in the one—word reading task. The distractor
words were different to avoid subjects' detection of the
experiment's agenda. Two of the new distractor words also
required final devoicing. I did not consider these words in
the data analysis because of their absence in the picture
recognition task. The following is a list of the words in
the order that they were presented in the word-reading task.

blau

Stuhl

Staub (devoiced)

Baum

Geld (devoiced)

Zeit

Wand

Hund

Berg (devoiced)

Vogel

gelb (devoiced)

Tisch

Post

Zug (devoiced)

Stern

Schrei

Rad (devoiced)

After the word-reading task, each student was given a
short text to read aloud. I predict orthographical
interference in this task, without the time to focus on
form. The text had 6 words that employed the final
devoicing rule. The following is a copy of the text and its

translation. The words with final devoicing are Jérg, Tag,

halb. Emma. seams: and list:-

36

Mein Name ist J6rg. Ich spiele gern FuBball. Es

macht mir SpaB. Ich stehe jeden Tag um halb sechs

auf, um FuBball zu spielen. Mein Freund und ich
spielen oft zusammen, um gesund zu bleiben. Ich

habe auch eine Freundin, die Sabine heiBt. Sie

ist sehr lieb.

My name is George. I like to play soccer. It's

fun. I get up every day at five thirty to play

soccer. My friend and I often play together in

order to stay healthy. I also have a girlfriend

named Sabine. She is very kind.

Only words with the voice/voiceless distinction of the
stops /b/-/p/, /d/-/t/ and /g/-/k/ were tested. As in the
pilot study, I did not believe that the two groups of
students had vocabularies which warranted the study of the
/v/-/f/ and /z/-/s/ distinction.

5.3 Place of Articulation

The place of articulation is the third factor that I
considered when testing the data. I also wanted to see if
this played a role as to whether the final devoicing rule
was utilized. A difference in the amount of devoiced velar
stops (more devoicing) as compared to the bilabial stops
(less devoicing) would be quite revealing. A significant
difference in this area would support the presence of the
articulatory factor presumed to be reflected in Ruhlen's
"voice warp". A larger percentage of devoiced velar stops
would suggest that the size of the cavity into which the
prereleased, postglottal, pressurized air is released, plays

a role in determining whether certain stops will be easier

to devoice.

Chapter 6

Results

6.0 Introduction

The results are mainly reported in Tables 6-1 and 6-2.

The

The 'x' represents the voicing of the final phoneme.

The '*' represents either

'0' represents proper devoicing.

Table 6-1

a failure to respond or an incorrect response.

presents the data from the beginning-level students and

te-level students

1a

termed

in

Table 6-2 those from

6.1 Data

Table 6-1 Beginning Level Results

Text
B B D D G G

Picture Word
B B D D G G

B B D D G G

Student

XXXXXX
OXXXXX
XXXXXX
OOXXXX
XXXXXX
OOXOOX
XXXXXX
XXXXXX
XXXXXX
XXXXXX
XXXXXX
XXXXXX

XXXXXX

**XXXX

XXXXXX
XXXXXX
XXOXXX

vat
vxs
ova
* *
ova

XXXXXX
OOOXOO
XXXOXX

nus
vxs
OVA
0...
01.

o * o * o o
x * x x x x
x x x x x *

XXXXXX
XXXXXX
XXXXOX

X*XXOX

XXXXXX

ova
vxs
xv“
* *
v.*

10
11

**X*O* XXXXXO

12

XXXXXX

XXXXXX
XXXXXX
XXXXXX
XXXXXX

* * x * x x
* * x * x x
s e x e x s

13

XXXXXX

14
15
16

XXXXXX
XXXXXX

X * X X * *

XXXXXX
XXXXXX
XXXXXX
XXXXXX

XXXXXX
XXXXXX

X*XXX*
X*XXX*

17
18
19
20
21
22
23

XOXXXX
XXXXXX

O*XX*X

**xxx*
X*XXXX
O*OXX*

XXXXXX
XXXXXX
XXXXXX

XXXXXX

OXOOOX

xxxxxx

0 * x x x x
* x x x x x
x * x * o x

XXXXXX

XOOXXX
XXXXXX

24
25

XXXXXX

37

38

Table 6-2 Intermediate Level Results

Picture Word Text
Students B B D D G G B B D D G G B B D G G
1 x x x x x x x x x x x x x x x x x x
2 x * x x o x x x x x o x x x x x o x
3 o * o x o o o o o x o o o o o o o o
4 o x x o x x x x o o x x x o x x x x
5 x o o o o x x x o o o x o x o x x x
6 o o o x o x x x x x x x o x x x x x
7 x * x x x x o o x x o x x o x x x x
8 o * o o o o o x o o o o o o o x x o
9 x * x x x x x x x x o x x x x x x x
10 * * o o o o o o o o o o o o o o o o
11 * * o o o o o x o o o x o o x x o x
12 * * x x o x x x x x x x x x x x o x
13 x x o x o * o x o o o o x x x x o x
14 o * x x x x x x x x x x x x x x x x
15 x * x x x * x x x x x x x x x x x x
16 * * o x x x o x x x o x o o x x x x
17 o * o o o o x o o x x o o o x x x o

A quick glance at the data shows that, even for the
intermediate level students, the acquisition of the final
devoicing rule was incomplete. Two intermediate level
students, numbers one and fifteen, showed no evidence of any
knowledge of the final devoicing rule. Thus, over 88% of
the students showed some evidence of final devoicing. The
results from the beginning level students showed only 12
students out of 25 giving evidence of final devoicing, which
is just below 50%. These percentages can be misleading
because in both cases there were some students that devoiced
properly in only one or two instances. If we were to
ascribe stages to the acquisition of the final devoicing
rule, we would find that these students with the low rate of

success in devoicing have only begun to acquire the rule.

39
They would be in the beginning stages of acquisition. The
following tables show the correlation between the number of
correct responses and the number of students that achieved
those responses. The numbers on the x axis of both tables
do not represent the total number of students in that level,
but rather the number of subjects that responded correctly f

the indicated number of times. i

_-_

Table 6-3 Beginning Level Frequency r_
13 X
12
11

Number 10

of 9

Students 8
7
6
5 X
4 X
3
2
1 X X X

 

 

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Number of Devoiced Responses

Table number 6-3 shows that the students in the beginning level
German class generally have not acquired the final devoicing
rule. Table 6-4 shows the intermediate level students spread
very evenly over the scale, showing that most have at least begun

to acquire the final devoicing rule.

40
Table 6-4 Intermediate Level Devoicing Frequency

13
12

Number
of
Students

X X X
X X X X X X X

 

 

O 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Number of Devoiced Responses

6.2 statistical Interpretation
6.1.0 Introduction

An initial glance at the data suggests the presence of
significant differences or interaction between the factors
task, students' year of study, and place of articulation. A
multivariate analysis (including chi-square and probability
tests) was done on all the above data. Notice that factor
group #1 (Task) refers to the manner of presentation.
Factor group #2 (Level) is the students' year of study.

Factor group #3 (Item) refers to the place of articulation.

41

6.1.1 STEP'ISE VARIABLE RULE ANALYSIS: Three Factor Groups

Table 6-6 First VARBRULE Data

Factor group i 1 Task
P = picture cue
W a word
T = text

Factor group # 2 Level

x = German 103
Y = German 203

Factor group # 3 Place
B a bilabial
D alveolar

G = velar

Number of cells: 18 Total no. of factors: 8

Group # l p: 0.615 w: 0.503 t: 0.382
Group # 2 x: 0.300 y: 0.700

Group f 3 b: 0.566 d: 0.436 g: 0.498
Log likelihood = -314.791

Cell Total Apps Expected Error
pxb 20 7 3.893 3.079
pxd 41 4 5.151 0.294
pxg 37 7 5.752 0.321
wxb 50 5 6.648 0.471
wxd 50 6 4.177 0.868
wxg 50 5 5.228 0.011
txb 50 5 4.272 0.136
txd 50 1 2.631 1.067
txg 50 1 3.321 1.738
pyb l8 8 10.224 1.120
pyd 34 15 14.917 0.001
pyg 32 15 16.011 0.128
wyb 34 11 15.463 2.363
wyd 34 14 11.271 0.988
wyg 34 15 13.208 0.397
tyb 34 16 11.457 2.717
tyd 34 6 7.889 0.589
tyg 34 10 9.487 0.038

Total Chi-square = 16.326 Chi-square/cell = 0.907

42

According to VARBRULE, factor group 3, place of
articulation, has no effect in raising or lowering the
probability of proper final devoicing. Notice that the
probability weights conglomerate around the 0.5 score. The
furthest deviation was 0.066. It was, therefore, deleted.
While this does not support the hypothesis that the larger
area between the place of articulation and the vocal cords
might be instrumental as to which stops are easier to
devoice, it does not show negative evidence in contradiction
of the " voice warp", which is simply a description of a
fact about many of the languages of the world.

After the 3rd factor group was deleted, a second
VARBRULE analysis was performed on the data. The factor
groups for this analysis (mode and level) are identical to
the factor groups 1 and 2 of the first run.

6.1.2 STEPWISE VARIABLE RULE ANALYSIS: TWO Factor Groups

Table 6-7 Second VARBRULE Data
Group # 1 p: 0.609 w: 0.511 t: 0.381
Group # 2 x: 0.298 y: 0.702

Log likelihood = -314.367

Cell Total Apps Expected Error

px 98 18 14.590 0.936
wx 150 16 15.749 0.004
tx 150 6 9.721 1.523
py 84 38 41.380 0.544
Wy 102 40 40.224 0.002
ty 102 32 28.335 0.656
Total Chi-square = 3.666

Chi-square/cell = 0.611

43

The results on probability were very similar to the
first run but error scores were lower (.907 vs .611), which
removed any suspicion of interaction between the variables.
The different probability weights in group one continue to
show that a student, regardless of the level, who receives
stimulus in the picture mode, is more likely to devoice
properly (p:.609). The same item presented in the word mode
has about equal chances of being devoiced properly. In
other words, the student is as likely to devoice as not to
devoice (w:.511). In the text stage the student has the
greatest probability to not devoice properly (t:.381). A
probability weight of around 0.5 would show that the mode of
presentation has no effect on the phenomenon being tested.
Any score higher than 0.5 would show that the mode of
presentation plays a part in eliciting the correct response.
Conversely, any score lower than 0.5 would show that the
mode of presentation plays a part in eliciting the incorrect
response. Students from the first year of German study will
more than likely not devoice properly (x:.298). Students
from the second year of German study will more than likely
devoice properly (y:.702). Factor group #2 is "heavier"
than #1. This means that there is a greater distance
between the factor weights. The factor difference from
beginning level (.298) to intermediate level (.702) is 0.4,
whereas the difference of the extreme factor weights in

group 1 is only .228, from .609 in the picture recognition

44
mode to .381 in the text reading mode.

The results give evidence for the relative difficulty
in acquiring the final devoicing rule. Only about 12% of
the students from the intermediate level German class failed
to show any evidence of final devoicing in their responses,
whereas over 50% of the students in the beginning level
failed to devoice in all three tasks. These last two
percentages show that what is statistically predicted is
also the case individually.

Some pedagogical advice that might stem from these
results is that, in order to acquire the final devoicing
rule, it is quite important to study German for at least a
year. One question is whether the explicit teaching of the
final devoicing rule is helpful in the acquisition of German
as a second language? It might be helpful to give the
students some visual representations along with the formal
rule. For example: When introducing the word lag, the
phonemic transcription could be presented simultaneously,
i.e. Tag /tak/. The final grapheme, <g> could then be
explained as one of those letters that loses its voice in
the word-final position. A follow-up study could then be
done to determine the benifits of such instruction.

The results show that there is no significant
interaction between the variables, i.e. the chi-square score
was insignificant. The intermediate students make mistakes

in the same pattern as the beginning students. They just

45
don't make them as often.

The difference in probability weights by task might
suggest that the amount of time a student has to think about
an item is a possible variable that needs to be considered.
During the word-recognition phase each student has more time
to think about the proper pronunciation than during the
text-reading phase. Although not universally accepted,
Krashen's (1983) Monitor Hypothesis might be applicable
here. Only one factor in the Monitor hypothesis is known
for certain to be involved, i.e. the amount of time each
subject has to respond. Whether each individual subject was
focusing on form and whether he/she knew the rule of final

devoicing were not part of the evaluation.

Chapter 7
Conclusions

7.0 Introduction
After the pilot study, I changed certain things in the test
format to make up for certain factors that could influence
the outcome of the study. For example, I discarded the
words that were both finally devoiced and had English
cognates. I also introduced a new phase into the study,
word reading, to isolate the variables being tested. After
the second experiment, more shortcomings and different
phenomena, as well as possible research areas in the future
were discovered.
7.1 Suggestions for Future Research

First, a variable accounted for in the word-recognition
phase was ignored in the text-reading phase. Each word
presented initially in the picture-recognition phase was
also introduced in the word-recognition phase. However, for
the text-reading phase different words with underlying final
voiced phonemes were used to elicit the voiced/devoiced
response. I propose two different modifications for future
studies in this area. First, different words employing the
final devoicing rule could be selected and used randomly
throughout the experiment. The other possibility would be
to use the same words in each phase to insure exact
comparison. The danger with the second modification is that
it might be difficult to measure how extensive the final

46

47
devoicing rule has been learned.

Second, there was no control group in this experiment.
I assumed that none was needed and tested only American
students of German. However, for reasons of scientific
integrity and result generalizability I would reccommend a
control group for any future study. I would recommend two
control groups: one group of native English speakers and the
other group of native German speakers. The English speaking
group would be to make sure the population studied was not
beginning to devoice English. Thomas Veatch (1989:AB) has
evidence that the final devoicing of fricatives is a "pan-
English" phenomenon. The German group would be used to
study the extent of final devoicing of the words in the
experiments.

Third, all words must be closely examined to make sure
that they employ the final devoicing rule. In retrospect, I
see no reason except for spelling to say that the word Jfizg
ends in an underlying voiced stop. There are no alternates
of this word, e.g. with vocalic suffixes in which a voiced
stop appears. No inflections can be attached to the name to
bring about a voiced velar stop.

While I still believe that the text-reading task is
valuable, certain aspects should be changed. All words
requiring final devoicing need to be strategically placed in
order to expect even a native German speaker to devoice

them, i.e. they should be followed by a pause or a word that

48
begins with a voiceless consonant. After the text is
prepared, trial runs should be made with native German
speakers to make sure the predicted obstruents appear.

Part of the proper acquisition of the final devoicing
rule is the ability to retain voicing of the obstruents of
the devoiced words when certain inflections are added. If a
speaker were to devoice certain obstruents regardless of the
presence of an inflection, we could not say that he/she has
acquired final devoicing but rather considered the
underlying form of certain lexical obstruents to be
voiceless.

7.2 Additional observations

As I was compiling the data, I made a point to observe
other phonological phenomena that are unique to German. The
observations I made can be split into two categories: those
that don't involve orthographical interference and those
that do. Only 2 students made an attempt at the velar /r/
of ‘rot'. No students had yet acquired the light /1/ of
'Ball' and ‘Gelb'. For these two examples there is no
orthographical interference. The <r> is still an /r/ and
the <l> is still an [1]. The only difference from the
English version is in the place of articulation. The most
interesting observations are found in the examples where
orthographical interference is involved. All students who
did not produce the initial /ts/ of 23g also did not devoice
final stops. The initial [8/ of ‘§_§;n' and ‘ELIQQQ' were

49

also quite elusive. Of these two examples the first one
proves to be most intriguing. The <z> of German is
pronounced without voice while the <z> of English is
pronounced with voice. There are other differences, but the
difference of voice is of special interest to the author.
By observing which German phonological rules have been
acquired, we might be able to postulate some order of
acquisition. In order to make some attempt at postulating a
rate or an order of acquisition, the most proper form of
research would be a longitudinal study.

Altenburg & Vago (1987:154) gave another reason for
pronunciation errors: "Spelling Pronunciation errors".

Since Hungarian spelling is phonetic in character,

it is particularly natural that some Hungarian

speakers should look to English spelling to aid

them in pronunciation.
Although English letters have complex phonetic
correspondents (Lado 1957:94), it is reasonable to assume
that beginning students would pronounce their German
according to English spelling rules. There were five
students who pronounced the picture of a cat as /kats/
instead of [katsdl Katze. It could possibly be explained,
post hoc, that they were pronouncing this word as they had
seen it written, even though this was in the picture
recognition stage. These students failed to devoice in
every situation.

Another variable that might be considered is whether

particular subjects are "letter readers" (phonological) or

50
"word readers” (global) (Goswami and Bryant:1990). The
letter reader might be more prone to reading each letter,
therefore failing to devoice or continuing to voice as the
grapheme would dictate. The word reader might be more
likely to devoice correctly. I did not consider this to be
a factor in my research. I assumed that since all students
were studying at the university, there were none that were
still using the phonological strategy. The phonological
strategy is theoretically one that children might adopt only
when learning to read.
7.3 Conclusion

I proposed in this thesis to achieve some insight into
a unique area of linguistics, one that has caused
frustration for many students of German as a foreign
language. For me personally, the writing of this thesis has
two results.

First, as a former instructor of German as a foreign
language, the findings with respect to final devoicing and
orthographical interference have provided me with a better
understanding of what to predict in the classroom. In
addition, some thoughts about how to overcome the obstacle
of final devoicing have emerged and can be tested in future
German classes for which I may be responsible.

The experiment, while perhaps still inconclusive, did
clearly show an effect of orthography on the acquisition of

final devoicing by American students. Thus, the second

51
result is an impetus for me to continue research in this
area, expanding my focus to include a comparison of native
English speakers learning German and native German speakers
learning English. I predict that such study will provide
even more valuable insight into final devoicing and the area

of orthographical interference.

LI ST OF REFERENCES

.. Tmfi-Z‘EV‘W If." I

LIST OF REFERENCES

Altenburg, Evelyn and R. Vago. 1987. Theoretical
implications of an error analysis of second
language phonology production. In G. Ioup et al.
eds., Interlanguage Phonology. 148-164. Cambridge:
Newbury House Publishers.

Anderson, Stephen R. 1985. Phonology in the TWentieth
century. Chicago: University of Chicago Press.

Benware, W. A. 1986. Phonetics and Phonology of Modern
German. Washington D.C. Georgetown University
Press.

Brakel, A. 1983. Phonological Markedness and Distinctive
Features. Bloomington, Indiana: Indiana University
Press.

Brown, T. and M. Haynes. 1985. Literacy background and
reading development in a second language. In T.
H. Carr. ed., The development of Reading Skills.
(New Directions for Child Development, no. 27.)
19-34. San Francisco: Jossey-Bass

Dinnsen, D. A. 1985. A re-examination of phonological
neutralization. Linguistics 21:265-279.

Dinnsen, D. A. and Jan Charles-Luce. 1984. Phonological
neutralization, phonetic implementation and
individual differences. JOurnal of Phonetics
12:49-60

Dinnsen, D. A. and F. R. Eckman. 1975. A functional
explanation of some phonological typologies. In
R. Grossman et al. eds., FUnctionalism, 126-134.
Chicago: Chicago Linguistic Society.

Eckman, Fred. 1977. Markedness and the contrastive analysis
hypothesis. Language Learning 27:315-330.

Eckman, Fred. 1981. On the naturalness of interlanguage
phonological rules. Language Learning 31:195-216.

52

53

Eckman, Fred. 1984. Universals, typologies and
interlanguage. In W. Rutherford (ed.), Language
Universals and Second Language Acquisition,
79-104. Philadelphia: John Benjamins Publishing
Company.

Flege, James. 1982. English speakers learn to suppress stop
devoicing. In K. Tuite, R. Schneider and R.
Chametzky (eds.), Papers from the Eighteenth
Regional Meeting Chicago Linguistic Society (18),
111-122. Chicago: Chicago Linguistic Society.

- ‘ 'z’kewrfi'

Gamkrelidze, T. 1975. On the correlation of stops and
fricatives in a phonological system. Lingua
35:231-262

Goswami, Usha and Peter Bryant. 1990. Phonological Skills
and Learning to Read. East Sussex, United Kingdom:
Lawrence Erlbaum Associates Ltd., Publishers.

Greenberg, Joseph H. (ed.) 1963. Memorandum concerning
Language Universals, universals of Language
xv-xxvii. Massachusetts: The M.I.T.
Press.

Hulstijn, J.H. and W. Hulstijn. 1984. Grammatical errors as
a function of processing constraints and explicit
knowledge. Language Learning 34:23-43

Ingram, David. 1986. Phonological development: production.
In Paul Fletcher and Michael Garman (eds.),
Language Acquisition, 223-229. Cambridge:
Cambridge University Press.

Jakobson, Roman. 1962. On ancient Greek prosody. In Selected
writings, 262-271. Netherlands: Mouton & Co.

Katamba, Francis. 1989. An Introduction to Phonology. New
York: Longman Inc.

Krashen, Stephen D. and Tracy Terrell. 1983. The Natural
Approach. Hayward, California: Alemany

Lado, R. 1957. Linguistics Across cultures. Ann Arbor,
Michigan: The University of Michigan Press.

Moulton, W. 1962. The Sounds of English and German.
Chicago: The University of Chicago Press.

O'Dell, M and R. Port. 1983. Discrimination of word-final
voicing in German. Paper presented at the
Acoustical Society of America, Cincinnati.

54

Port, R., F. Mitleb & M. O'Dell. 1981. Neutralization of
obstruent voicing in German is incomplete. Paper
presented at the Acoustical Society of America,
Miami.

Ruhlen, M. 1986. The voice warp: A phonological universial.
In Joshua A. Fishman et al. The Fergusonian
Impact, 17-19. New York: Mouton de Gruyter.

Selinker, L. 1969. Language transfer. General Linguistics
9:67-92

Stockwell, Robert and D. Bowen. 1965. The Sounds of English
and Spanish. Chicago: University of Chicago Press.

Tarone, E. 1988. variation in Interlanguage. London: Edward
Arnold.

Trubetzkoy, N. 1958. Grundzfige der Phonologie. Gettingen:
Vandenhoeck & Ruprecht. [Translation]: C. Baltaxe,
1969. Principles of Phonology. Berkeley, Univ. of
California Press.

Veatch, T. 1989. Final devoicing of fricatives of English.
Paper presented at the Annual Meeting of the
Linguistic Society of America. Washington, DC.

Vennemann, T. 1972. On the theory of syllabic phonology.
Linguistische Berichte 18:1-18

iiiiiiiiiiiii