133
531

 

111111111111111111111

LIBRARY 3 1293 01L86

Michigan State
University

 

 

 

 

 

 

 

 

This is to certify that the

thesis entitled

MARKEDNESS AND THE PERCEPTION OF
SPANISH OBSTRUENTS

presented by

Michael David Pasquale

has been accepted towards fulfillment
of the requirements for

M.A. degree in Linguistics

 

 

haw

Major professor

 

Date gfld /S§(

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

 

 

PLACE IN REIURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

V

at“: 1111

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1“ WM“

MARKEDNESS AND THE PERCEPTION OF SPANISH OB STRUENTS
By

Michael David Pasquale

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF ARTS
Department of Linguistics

1998

ABSTRACT
MARKEDNESS AND THE PERCEPTION or SPANISH OBSTRUENTS
By

Michael David Pasquale

This study attempts to explain processing errors that occur in a second language
learning context on the basis of markedness theory and equivalence classification (F lege
1987). Specifically, the context is where native Spanish speakers (N S S) misperceive the
speech of native English speakers (NBS) producing Spanish words.

A theory of markedness is developed which combines universal and language particular
markedness notions. Equivalence classification (EC) deals with the production and
perception of ‘new’, ‘similar’ and ‘identical’ sounds in a second language learning context
and is used to explain errors by NSS.

It was predicted on the basis of markedness and EC that if English speakers
pronounced ESE). ‘all’ as [toDo] then the NSS would perceive this as m ‘bull’ rather
than t_og_<_) because English [D] and Spanish [r] are similar sounds on the basis of Shared
distinctive features and that Spanish [(1] is marked intervocalically.

The results of the test showed that there were misperceptions by NSS. For example.

they heard s_e_r§_ [sera] ‘will be’ for w [se8a] ‘silk’ and t9r_o [toro] ‘bull’ for todo [tofio]

‘all’.

ACKNOWLEDGMENTS

It has been a privilege to work with the chair of my committee Dr. Grover Hudson.
This work has benefited from his helpful comments and suggestions. I would also
like to thank the other members of my committee, Dr. Alan Beretta and Dr. Yen-Hwei
Lin, for their comments. It is also necessary for me to thank Laurie Church for all her
help along the way. This process would have undoubtedly taken much longer without
her help.

This work would not have been possible without the help and encouragement
of many people including colleagues, friends, and family. Iwould like to thank those who
I have considered my mentors: Dr. Henry Osborn, Dr. Floyd Elmore, and Dr. Gary
Hauck. I appreciate the encouragement they have given me to follow this path into
linguistics.

I also would like to thank Gabriela Alfaraz and Monica Bolet for their patience with
me as I tapped into their knowledge of Spanish and for their help with my data collection.

I would like to thank my parents and family for their wonderful support in my
academic endeavors these many years. Last but certainly not least, I would like to thank
my wife Monica for giving me the encouragement to finish this project and for always

believing in me.

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................... v

LIST OF FIGURES .................................................................................................... vi

CHAPTER 1

INTRODUCTION ...................................................................................................... 1

CHAPTER 2

THEORETICAL BACKGROUND ............................................................................ 4
2.1 Markedness Theory ......................................................................................... 4
2.2 Autosegmcntal and Lexical Phonology ........................................................... 11
2.3 Underspecification Theory ............................................................................ 14

2.4 Methods of Analyzing Processing Errors in Second Language Acquisition. 17

CHAPTER 3

SPANISH SPIRANTIZATION ................................................................................ 21
3.1 Introduction to the Relevant F acts ................................................................. 21
3.2 Traditional Analyses ...................................................................................... 22
3.3 Underspecification Analyses ......................................................................... 24
3.4 New Analysis for Spanish Spirantization ...................................................... 27
3.5 The Equivalence of English [D] and Spanish [r] ............................................. 36

CHAPTER 4

TEST INVOLVING PROCESSING ERRORS IN SECOND LANGUAGE

ACQUISITION ......................................................................................................... 3 9
4.] Preliminaries ................................................................................................... 39
4.2 Testing Method ............................................................................................. 39

CHAPTER 5

RESULTS AND ANALYSIS OF TEST
5.1 Predictions on the Basis of Equivalence Classification ................................. 43
5.2 Test Results ................................................................................................... 45
5.3 Analysis ........................................................................................................ 47
5.4 Conclusion ..................................................................................................... 50

REFERENCES ........................................................................................................... 52

iv

LIST OF TABLES

Table 5.1 - Results from Native Spanish Speakers ................................................... 46

Table 5.2 - Results from Native English Speakers ................................................... 47

LIST OF FIGURES

Figure 2.1 - Frequency of English obstruents out of 10,000 spoken words ................ 9
Figure 2.2 - Three frameworks of underspecification ................................................ 15
Figure 3.1 - Spirantization rule application ............................................................... 23
Figure 3.2 - Stop-Formation rule application ............................................................ 24
Figure 3.3 - Continuancy spreading ........................................................................... 25
Figure 3.4 - Continuancy spreading rule application ................................................. 26
Figure 3.5 - Frequency of sounds in 317 languages ................................................... 28

Figure 3.6 - Frequency calculations of certain Spanish phones out of 1,000 spoken

words ..................................................................................................... 28
Figure 3 .7 - English phoneme frequencies out of 10,000 spoken words ................... 29
Figure 3.8 - Markedness rule application .................................................................. 33
Figure 3.9 - Spanish phonemic and phonetic representation comparisons ................ 34
Figure 3.10 - English phonemic and phonetic representation compar'isons ............... 36
Figure 4.1 - Information on Native English Speakers involved in test ....................... 40
Figure 4.2 - Word List used in study ......................................................................... 41
Figure 4.3 - [8~r] minimal pairs in Spanish ................................................................ 42
Figure 5.1 - Derivations of Spanish words ................................................................. 48
Figure 5.2 - Derivations of English words .................................................................. 49
Figure 5.3 - Misperception of NES productions by NSS .......................................... 50

Chapter 1

INTRODUCTION

During the 19503 and 19603, languages were compared to each other using a technique
called Contrastive Analysis (CA). The ease or difficulty in acquiring another language
was thought to be directly related to how similar the two languages were to each other.
The goal of CA was to predict what difficulties the learner would face with a new
language. An example of a CA analysis of English and Spanish is in Stockwell and Bowen
(1965)

This study will look at the production and perception of Spanish obstruents [b, d, g]
in a second language learning context Specifically the context is where native Spanish
Speakers misperceive the speech of native English speakers producing Spanish words.
Previous studies concerning processing errors of Spanish voiced obstruents were done
under CA methods. Generalizations about Spanish-English production errors were given
in Stockwell and Bowen (1965), Politzer and Staubach (1965), and Cardenas (1961). In
these studies, errors were given but it was not explained why they occur.

The goal of this paper is to Show that there may be difficulties in perception by native
Spanish speakers concerning the feature [continuant] in voiced obstruents when non-

native speakers of Spanish use the marked value. It will be shown that [+continuant] can

be the unmarked value in voiced obstruents in the environment following a [+vocoid]
segment.

The perceptions of native Spanish speakers (N SS) were analyzed in this study. The
N88 heard productions of Spanish words by native English speakers (NES). These

Spanish words were minimal pairs such as todo ‘all’ and m ‘bull’. Minimal pairs with a

[8] ~ [I] alternation were chosen in order to test the perception of a NSS.

In Spanish, [8] is pronounced intervocalically while [(1] is used utterance initially and

following a nasal consonant. In English, [d] can be pronounced intervocalically. A

flapping rule applies in English in the following environment: /d/ -> [D] / V”_ V
(Kenstowicz 1994: 68). For example, llaedar/ -> [laeDar] ‘ladder’ and Iraydar/ —)

[rayDer] ‘rider’. If an English speaker pronounces todo ‘all’ as [toDo], then a NSS may

misperceive the word as 3259 ‘bull’. The results of the test Showed that there were
misperceptions by NSS. For example, they heard Q ‘will be’ for gglg ‘silk’ and m
‘bull’ for BBQ ‘all’.

The phonological background will be Autosegmcntal and Lexical Phonology. These are
grounded in the Generative Phonology framework in which phonological structure are
analyzed as segments and distinctive features. A key element that will be developed from
this background is Underspecification Theory. This is the idea that a phone may not be
lexically fully specified for all its features, which are filled in by rules.

The view of Radical Underspecification is that only marked features may be specified
in the lexicon. This study will appeal to markedness and will develop a theory of

markedness.

Second language acquisition (SLA) theory as it relates to second language processing
will also be developed in this study. The production of a second language sound system
and the perception by native speakers of non-native speaker productions are the main
focus of this study. Studies by Flege (1987, 1991) deal with the production and
perception of what he classifies as ‘new’, ‘Similar’ and ‘identical’ sounds. This
classification of second language sound production along with markedness theory will be

used to explain the perception difficulties of Spanish speakers in this study.

Chapter 2

THEORETICAL BACKGROUND

2.1 Markedness Theory
2. 1.1 Historical Perspective

The study of markedness has taken two distinct directions. One focuses on
markedness universally across all languages, and the other direction focuses on
markedness within a particular language. At times these studies have been viewed as in.
opposition to one another.

The theory of markedness itself was founded on work by Roman J akobson and Nikolai
Trubetzkoy. J akobson looked at markedness in a general manner and attempted to deal
with markedness as a universal aspect of language (Andrews 1990:13). Trubetzkoy
limited his focus to phonology. He first used the terms ‘marked’ and ‘unmarked’ in his
article “Die phonologishen Systeme” (1931). The major ideas of the theory of
markedness have never been agreed upon. Trubetzkoy and Jakobson both developed
markedness around the same time but they defined key elements such as ‘marked’,
‘unmarked’ and ‘opposition’ differently (Andrews 1990:13).

Trubetzkoy focused on markedness in application to phonology. He did not favor the
binary system of opposition that Jakobson favored. Trubetzkoy proposed three types

of contrasts of phonemes: (i) privative: two phonemes are identical except that one

contains a “mark” the other one lacks (eg. voicing in /b/ vs. /p/); (ii) gradual: phonemes
have different degrees of a gradient property (cg. height in fi/ ~ /e/ ~ LEI); (iii) equipollent:
each member has a mark that the other lacks (eg. place of articulation in /p/ ~ /t/ ~ lk/)
(Sampson 1980: 108; Andrews 1990:14).

Jakobson extended markedness oppositions from phonology to morphology and
semantics. In phonology, the term ‘distinctive feature’ was used to mean a binary
markedness opposition that Jakobson proposed (Andrews 1990:13). A segment could
have one of only two feature values which were indicated by a + or -.

The distinction between a ‘universalist’ and a ‘particularist’ markedness view began in
the 19605 and still continues today. Joseph Greenberg has been a leading proponent of
universal markedness relations in phonology as well as syntax and semantics. Postal
(1968) and Chomsky and Halle (1968) also developed theories of markedness which
appealed to universal ideas.

The language-particular camp was really started by Jakobson. Early in his studies on
markedness he looked at markedness universally but later he focused on markedness in
particular languages. Battistella (1990) describes the difl’erences between looking at
markedness universally and in a particular language. These differences will be explained in

more detail below.

2.1.2 Markedness Value Determination
2.1.2.1 Universal Markedness Criteria

The criteria for determining markedness values differs when analyzing markedness
from a universal or particular language point of view. Universal markedness studies such
as Greenberg (1966) places emphasis on the high frequency of the unmarked sounds and
the relative complexity of the marked sounds in phonology. Postal (1968) and
Chomsky and Halle (1968) also make decisions based on data from dialect variation and
sound change and physiological / perceptual findings.

First of all, relative frequency plays an important role in determining the markedness
of a feature. Cross-linguistic frequency is given high priority when determining universal
markedness. The unmarked case will have more frequency across the world's languages
than the marked value which may only occur in a few languages. An example is the value
[+Spread glottis] which determines the aspirated forms of consonants. In a cross-
linguistic inventory of voiceless stops in Maddieson (1984227), 291 languages have plain
voiceless stops (91.8%) as compared with 91 languages which have aspirated voiceless
stops (28.7%). The unmarked value [-spread glottis] is more frequent than the marked
value by far.

Greenberg (1966: 17) states that the unmarked values have a higher frequency than
marked values. He gives as an example the feature [nasal]. In Bengali and French,
nonnasalized vowels outnumbered nasalized vowels. In Bengali, the ratio was 50:1 and in
French the percentage was 82.5% to 17.5% in favor of oral vowels. He concludes that

due to relative frequency, [-nasal] is the unmarked feature.

A point of clarification needs to be stated relating to frequency and markedness. A
feature value is not unmarked because it is frequent, it is frequent because the feature is
unmarked. Guitart (1979:54) explains: “Unmarked does not mean ‘more frequent’, or
‘expected in positions of neutralization’, or ‘acquired first’, etc. Unmarked means ‘less
complex’ and Marked, ‘more complex’, and the complexity is either of a physiological or
a perceptual nature.”

The second criterion used to determine markedness is the relative complexity of the
feature value. This is also related to the idea that the unmarked value is more ‘natural’
than the marked value. The notion of ‘complexity’ or ‘simplicity’ was described by
Trubetzkoy as being a markedness distinction: “‘a natural absence of marking’ is
attributed to that opposition member whose production requires the least deviation from
normal breathing” (T rubctzkoy 19692146). Articulatory and perceptual complexity can
both be taken into account for markedness determination. Battistella (1990249) states:

the relevant notion of simplicity in phonology will be one that takes articulator
and acoustic naturalness into account where articulatorily simple features will be
those whose articulation requires the minimal deformation of the vocal tract from a
neutral rest position and acoustically simple ones will be those that are most
easily perceived.
For example, it was shown that the feature [+Spread glottis] was the marked value as
evidenced by frequency but aspiration is also presumably more ‘complex’ than plain
articulation of stops. The feature [+constricted glottis] which is the feature that
determines glotallalization is also more complex. How marked a feature is depends on

how far the feature strays from the ‘ideal’ position (cf. Cairns 1969:878; Jakobson

1962:266).

2.1.2.2 Language Particular Markedness Criteria

The criteria for determining markedness within a particular language are different from
those used for universal markedness considerations. Those who look at markedness from
a language particular point of view will appeal to the distribution of sounds in a particular
language. The frequency of the unmarked sound is given high priority in determining
markedness values. ‘Neutralization’ describes the distribution patterns of the marked
and unmarked in a language (Battistella 1990251).

The frequency of a sound within a language is the focus of language-particular
markedness. Frequency is looked at in different ways depending on a universal or
particular language view. As an illustration of this difi’erence the terms ‘type’ and
‘token’ can be used. Type frequency is used in a universal analysis. This refers to
comparing many languages to see if a certain phone is a part of their phonetic inventories.
The occurrence of a particular sound in a language inventory is counted. For a language
particular analysis, the frequency of tokens is analyzed. The number of times a particular
sound is used in a language is counted, usually as a percentage of usage compared to other
phones in the language. The distribution of phones in a language is important in
determining markedness values language specifically.

To illustrate the difference between type frequency and token frequency, the following
are examples of how markedness is determined in both cases. An example of type

frequency was given above with the feature [+ spread glottis] (Maddieson 1984:27).

 

Out of 3 1 7 languages surveyed, 291 had plain voiceless stops while 91 had aspirated
voiceless stops. The unmarked value of [-spread glottis] is due to the greater frequency
of the value across many languages.

An example of token frequency can be illustrated in English. The relative frequencies
of the features [coronal] to [labial] and [velar] are analyzed. In a survey of 10,000 words

of adult conversation, the following numbers and percentages of use were calculated

 

(figure 2.1).

[+coronal] /t/ 2248 4.615% /d/ 1827 3.767%
[+velar] /k/ 1414 2.903% /g/ 598 1.228%
[+labial] /p/ 694 1.425% /b/ 877 1.800%

 

 

Figure 2.1: Frequency of English obstruents out of 10,000 spoken words. (Carterette
and Jones 19742445)

In Carterette and Jones (1974), the higher frequency of [+coronal] is shown by
percentages of 4.61 5% and 3.767% for /t/ and /d/ respectively. The velar and labial

ob struents have a lower percentage of occurrence in English. This is evidence that
[+coronal] is the unmarked place of articulation in English.

The second criterion for particular language markedness is neutralization. When two
sounds can contrast in a language but fail to do so in an environment, the unmarked value
appears and there is a neutralization of the contrast between the marked and unmarked
values. For example, in English the stop consonants /p, t, k/ have corresponding voiced

stop consonants lb, d, g/. These are all phonemes in English and may occur in the same

position to make minimal pairs such as in initial position: m b_it, .ti_n, din, Q, etc. In the
context following an /s/, the contrast is neutralized. There are s_pi_n, gtgp, and s_lg_i but
there are no s_bir_1, sci—op, or §E- The voiceless consonants remain after neutralization so
[-voiced] would be the unmarked value in the environment # s_ (Battistella uses
[tense] and [lax] in his example, so /p, t, k / (tense) are unmarked while /b, d, g/ (lax) are

marked) (Battistella 1990247).

2.1.2.3 The Relation Between Universal and Language Particular Markedness

This study will follow from the premise that universal markedness values reflect in
general what features are marked or unmarked in languages. Universal markedness values
are expressed in statements as to what is usually found across the world’s languages.
Universal markedness distinctions have been defined by the relative frequency and
complexity of feature values across the world’s languages. These are taken to be the
regular tendencies of language or as general ways of categorizing language as a system.

In this study, universal markedness values will be appealed to and developed through
the criteria of relative frequency and complexity. However, the definition of markedness
of a feature value, for example, will then be applied to a particular language and to specific
environments. The markedness value in these environments may not be the “common” or
“general” markedness value.

Battistella (1990254) addresses the concept of ‘context-sensitive markedness’:

Two considerations are important with respect to phonological markedness
relations: context sensitivity and universality. With respect to the first, it is
important to recognize the markedness relations between phonological features
are contextual in nature; the markedness values of phonological features are not

immutable within a phonological system, but instead are determined according to
the co-occurring features of a phoneme (its simultaneous context) and according to

10

the surrounding phonetic environment (its sequential or syllable-structure context).
An example of a markedness value that is changed due to its sequential context is
shown when /t/ , which is relatively unmarked, would be marked when preceding a /d/ or
an IV. In those contexts, the features of /t/ that were unmarked when analyzed separately

may be marked due to context (Battistella 1990:55).

2.2 Autosegmcntal and Lexical Phonology

Generative Phonology as set forth in Chomsky and Halle (1968) is the foundation for
both Autosegmcntal Phonology and Lexical Phonology. The basis of a Generative
Phonology (GP) framework is a lexical or underlying representation modified by rules.
The basic units in GP are distinctive binary features which classify each phone in a
language. Autosegmcntal Phonology (AP) and Lexical Phonology (LP) are both grounded

in GP ideas but modify certain aspects for better explanation of the facts.

2.2.1 Autosegmcntal Phonology

Autosegmcntal Phonology differs from GP in that phonological representation does
not consist of a Single row of segments, but consists of different levels or tiers. AP was
developed in Goldsmith (1 97 6) as a means of explaining tonal languages. The argument
put forth is that tones are on a separate tier and there is not a lexical one-to-one
relationship between a tone and a segment.

AP has been used to analyze other features of phonology than tones. A characteristic
of AP is a non-linear character of phonological representation. In a non-linear model,

there is not a one-to-one relation between feature specifications and segments of the

11

string. One feature value [F] can be distributed over more than one position just as one
tone can.

Another characteristic of AP that is important to this study is the notion of
assimilation as a spreading process. Spreading in AP was at first studied in regard to
tone. Later, however, it has been proposed that assimilation processes such as place of

articulation assimilation or palatalization should be viewed as the spreading of a feature.

2.2.2 Lexical Phonology

Lexical Phonology is also a further development of GP and has both a phonological
and morphological aspect. The work of Paul Kiparsky (1982) has guided much of the
study of LP. Important aspects of LP include the Strict Cycle Condition, the Elsewhere
Condition, Structure Preservation, and Underspecification Theory.

This study will appeal most directly to Underspecification Theory, which also
developed from the interaction of AP with LP. Reference to the Elsewhere Condition will
also be included in the study and will be extended to markedness theory.

A basic aspect of LPis the distinction of Lexical rules and Post-Lexical rules.

Lexical rules: (i) have access to the internal structure of words, (ii) have exceptions, (iii)
are cyclic, (iv) cannot apply to non-derived forms, (v) are subject to the SCC, and (vi) are
structure preserving.

Post-Lexical rules: (i) have no access to the internal structure of words, (ii) have no
exceptions, (iii) are non cyclic, (iv) apply across the board, (v) may apply to non-derived

forms, (vi) are not subject to the SCC, and (vii) are not structure preserving.

12

Post-Lexical rules always apply and are like allophonic rules. Lexical rules are the
default rules and will be explained more in the section relevant to Underspecification
Theory.

The Elsewhere Condition (EC), explained in Kiparsky (1973), states that when two
rules can apply with different results, the more particular rule overrules the more general
rule. The EC has an important role in markedness theory in that a more particular
language-Specific markedness rule will overrule a more general universal rule if they are in
the EC relationship.

In general, the idea that phonological features can be left unspecified in underlying
representations is Underspecification Theory. There are differing viewpoints as to what
exactly is specified in an underlying or lexical representation and what is not. There is a
general agreement that non-contrastive features are not specified underiyingly and are
specified by rule. The theories differ over the necessity of specifying all contrastive
features. Redundant features are specified by feature filling rules. These redundancy rules
are lexical rules (also called “lexical redundancy rules”). For example, in English the
assignment of [+voice] to those phones specified as [+sonorant] is a lexical redundancy
rule.

Post-Lexical rules: (i) operate crucially across morpheme boundaries or making critical
use of phrasal or syntactic structure; and (ii) fill in, specify, or refer to non-contrastive

features (Kenstowicz 19942209). Spirantization in Spanish is a result of a post-lexical

rule. An obstruent may spirantize word internally (i.e. 1ng [diya]) or across morpheme

boundaries (cf. la gata [layata] and un gato [ungato]).

13

2.3 Underspecification Theory

In underspecification theory, there has been much debate as to what exactly is allowed
to be specified underlyingly. Two main camps have emerged: Restricted and Radical
Underspecification.

A restricted underspecification account (Steriade 1995) will specify underlyingly both
the '+' and '-' values of a contrastive feature. The feature values that are not contrastive
can be unspecified in the underlying representation.

Archangeli (1984, 1988) and Archangeli and Pulleyblank (1989) support a Radical
Underspecification framework. In radical underspecification, generally only the marked
value of a contrastive feature is specified underlyingly in all contexts. The unmarked or
default value of a contrastive feature is unspecified (along with all noncontrastive
features) and is filled in by a context-free default rule. As mentioned above, a default rule
is a lexical rule. In this account, the unmarked value will be assigned.

Kiparsky (1982) supports a Radical Underspecification account. In Kiparsky (1993),
he relies on radical underspecification but he also suggests the idea of context sensitivity
in determining markedness and underspecification.

There have been some reviewers of Kiparsky (1993) that have gone as far as saying
that radical underspecification according to Archangeli et al is “context-free” and
Kiparsky (1993) is changing this to a “context-sensitive” account (cf. Iverson 1993,
Dinnsen & Chin 1995, Dinnsen 1996).

Dinnsen (1996) illustrates the differences between Restricted Underspecification with

C ontext-Free and Context-Sensitive Underspecification. In context-free radical

14

 

underspecification, the marked value of a contrastive feature is specified underlyingly in
all contexts (1996259). The unmarked value is unspecified and is filled in by a context-
free rule (1996259). In context-sensitive radical underspecification, context determines the
underspecification of features. Dinnsen states, “This acknowledges that a given feature
may be the default value in one context and that the opposite value may be the default in
another context.” (1996259). He further states that, “ [t]he consequence is that both
values of a contrastive feature can be specified underlyingly, but not in the same contexts,

preserving strict binarity” (1996:59).

 

 

 

 

Context
Framework Sounds X Y

Restricted underspecification A [+F] [+F]

B l-F] l-F]
Context-free radical underspecification A [+F] [+F]

B O 0
Context-sensitive radical underspecification A [+F] D

B 0 [-F]

 

Sounds A/B are contrastive and differ by [F] only.
‘0’ = underspecified

 

 

Figure 2.2: Three frameworks of underspecification (based on Dinnsen 1996260)

Figure 2.2 illustrates the three frameworks of underspecification. The [-F] feature in

the context-free radical underspecification account is the default value and is filled in by a

context-free default rule. In the context-sensitive account, the [-F] in context X and the

15

[+F] in context Y are the default values and are filled in by a context-sensitive rule
(Dinnsen 1996260).

An example of context-sensitive underspecification has been illustrated in English
quantity alternations in vowels (Kiparsky 1984, Iverson 19942266). In English, vowel
length may be long (paint) or short (tent) before consonant clusters in the same
morpheme. A vowel is only short before Level 1 suffixes which create consonant clusters
such as in the word meant. The vowel is predictably short before the irregular suffix 4
but it is long before other suffixes (meaning, means, etc.) (Iverson 19942266).

The feature [+long] would be given by default rule and vowels which do not fit in the
environment of the precluster shortening rule are underspecified for length. The vowel in
pen would be specified as [-long] underlyingly but the vowel in mean is unspecified for
length. In the word paint, the vowel is Specified as [+long] to avoid the vowel shortening
rule application in precluster position since it is a nonderived context (Iverson 19942266).
In the word tent, a short vowel appears before a consonant cluster. The value for length
may be unspecified and filled in by rule to give [-long]. The resultis that some vowels
are specified underlyingly and some are not. The value [+long] is specified in paint,
[-long] in pen, and unspecified in words like tent and mean. Iverson (1994:266) states:
“this mode of underspecification is still radical since the contexts where plus values occur
(before clusters) never overlap with those where minus values occur (not before clusters);
that is, only one feature value is specified in any given context found among underlying
representations”.

In this study we will refer to “Radical Underspecification” as it is defined by

Archangeli while accepting the idea that underspecification (and markedness) is sensitive

16

to the phonological environment as Kiparsky (1993) suggests. I would suggest that
Kiparsky is not overhauling the theory of radical underspecification but only clarifying it -

based on data which shows markedness is sensitive to context.

2.4 Methods of Analyzing Processing Errors in Second Language Acquisition

Communication in language results from a process of making sounds and listening to
sounds in speech. For a speaker in a second language environment, communication is
more complex due to difficulties in pronouncing sounds like a native speaker.
Communication between a native speaker and a non-native speaker may be made difficult
if both have problems understanding what sounds the other is intending to make. Non-
native speakers may find it difficult to distinguish sounds which are allophones in their
language but are phonemes in the second language. Native speakers may find it hard to
understand a non-native speaker if that person substitutes sounds from their first
language.

In this study, the term ‘processing’ refers to the process of producing sounds and
perceiving sounds in speech. The terms ‘production’ and ‘perception’ refer to the acts
of articulating speech and understanding speech respectively. Processing errors are those

dealing with problems in articulating and understanding speech.

2.4.1 Production Errors in Second Language Acquisition
A foreign accent is inevitable to almost everyone who learns a second language as an
adult. The studies on production errors in second language acquisition (SLA) have

traditionally acknowledged that when second language learners speak a foreign language

17

they will identify phones from the second language (L2) in terms of their native language
(L1) categories. Weinreich (1953) sought to describe what he called ‘phonic transfer’
within a theoretical framework. Weinreich originally attributed the problem of phonic
interference to incorrectly identifying a new phone in terms of one’s first language system
(1953: 14). Weinreich (1957) revises his proposal concerning phonic transfer and the
identification of an L2 sound. He redefined phonic transfer as “the manner in which a
speaker perceives and reproduces the sounds of one language in terms of another”
(19572126). In Weinreich’s analysis, distinctive features are the basis for identification
instead of phonemes (1957227).

Ritchie (1968) and Michaels (1974) focus on using distinctive features to explain
identification problems for an L2 speaker of a language. Michaels (1974) is based on the
markedness conventions from Chomsky and Halle (1968). He states that features in a
second language which are marked in a speaker’s L1 are sometimes replaced by the
unmarked feature in the L1 when speaking the L2. James (1988221) states that in this
case, markedness of a particular feature is limited to a specific language and that an L1
system is imposed on the L2.

The Markedness Differential Hypothesis (MDH) was proposed in Eckman (1977) in
order to appeal to markedness and phonic transfer between two languages. Eckman
defines markedness as “a phenomenon A in some language is more marked than B if the
presence of A in a language implies the presence of B; but the presence of B does not
imply the presence of A” (19772320). The definition of the MDH is:

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, such that,
(a) Those areas of the target language which differ from the native language and are

18

more marked than the native language will be difficult,

(b) The relative degree of difficulty of the areas of the target language which are
more marked than the native language will correspond to the relative degree of
markedness,

(c) Those areas of the target language which are different from the native language,
but are not more marked than the native language will not be difficult (19771321).

The MDH relates to differences in the structure of L1 and L2 but is set forth within a
universal rather than a language specific framework as in Michaels (1974).

Flege (1987, 1988, 1991) joins phonetic research to phonological theory in addressing
the problem of phonic transfer. His studies refer to ‘equivalence classification’ in which
L2 perception and production of ‘identical’, ‘similar’, and ‘new’ sounds are analyzed in a
second language environment.

Flege (1987, 1988) makes a distinction between "new" phones in the L2, and "similar"
sounds between the L1 and L2. He states that sounds that are Similar in the two
languages will be identified and produced in terms of a similar phone of the L1 even if
these sounds are different. Sounds that are new in a second language would be easier to
learn because of the lack of interference from the L1 categories. He states “new sounds
may ultimately evade equivalence classification, thereby permitting the learner to avoid
the limiting effect of previous phonetic experience” (19882369).

Accurate perception and production of ‘similar’ and ‘identical’ phones in the L2 may

be more difficult since they share acoustic similarity and phonologically may be

interpreted as allophones of L1 phonemes (1988:368).

19

2.4.2 Perception Errors in Second Language Acquisition

The equivalence classification proposal can also be applied to perception errors in
SLA. Bohn and Flege (1990) apply equivalence classification to the perception of L2
learners of the productions of L1 speakers. They propose that L2 learners will create
new categories for sounds that are different from those in their L1 system (1990:303).
Equivalence classification does not block the learning of a new sound category since it is a
“new” rather than a “similar” sound. An L2 speaker will classify a sound in the L2
which is Similar to a sound in their L1 as belonging to the same category.

Bohn and Flege (1990) use the notion of equivalence classification in the perceptions
of the L2 learner of L1 productions. This study will apply this use of equivalence
classification in another way -- to the perception of L1 speakers to the productions of L2
language learners. For example, the production of Spanish by native English speakers will
be heard by native Spanish speakers. The perception errors of these native Spanish
speakers will be studied and it will be proposed that they perceive English speakers’

productions of Spanish within a Spanish phonemic matrix.

20

Chapter 3

SPANISH SPIRANTIZATION

3.1 Introduction to the Relevant F acts

This study will concentrate on the voiced obstruents [b, d, g] in Spanish. These
obstruents alternate with the voiced fricatives [[3, 8, 7] which differ with respect to the
feature [+l- continuant]. The stops and fricatives are in complementary distribution and
are allophones. The environments where the voiced stops [b, d, g] occur are (i) at the
beginning of an utterance, (ii) following nasals, and (iii) following /1/ for [(1] only (Harris
1969238). The voiced fricatives [[5, 5, 7] appear, (i) following a segment that is [+vocoid]
(i.e. vowels and glides), and (ii) following a liquid (note exception for [(1] above).

The first context is usually realized intervocalically. The generalization of the
environment in which the fricatives occur is following a segment specified as

[+continuant].

In English, lb, (1, g/ are phonemes and unlike Spanish they may occur intervocalically.

English does not have corresponding voiced fricatives [6, 5, y] as allophones of lb, d, gl.

The phones [[3] and [y] do not appear in the phonetic inventory of English and /5l is a

phoneme which contrasts with ldl.

21

3.2 Traditional Analyses

Traditional grammars of Spanish have posited a Spirantization rule which changes the
voiced stops to voiced fricatives [[3, 8, y] in a certain phonetic environment. In the
traditional analysis, ordinarily the voiced stop obstruents are assumed to be underlying

and are specified for the feature [-continuant], but undergo a change by way of a

phonological rule. Harris (1969) represents Spirantization with the following rule:

{ [-son] )
['SOH, +vcd] —-) [+cont]l { [+cont] } f _ '|
l<[0tcor]>l L<[-0tcor]>_l
(Harris 1969:40)
This rule, however, includes [-sonorant] as an environment which promotes

Spirantization. This presumably is to account for a word such as [aSBerso] ‘adverse’ or

[afifiomen] ‘abdomen’. This part of the rule would not be needed since both obstruents

are preceded by a [+continuant] segment. A more concise Spirantization rule would be:

 

r [+cont] )
[-son, +vcd] -> [+cont]l { < [-cont, -0tcor]> } <[0tcor],>
l J

In this rule, Spirantization occurs in voiced obstruents in either of two environments:
when preceded by a [+continuant] segment or when preceded by a [-continuant] segment

that does not share [coronal] place of articulation. An example of the first context is a

preceding vowel or [r] as in [a8a] hada ‘fairy’ and [perfion] perdon ‘pardon’. An example
of the second context is a preceding [1] such as in [kalBo] calbo ‘bald’. In the case of

shared [coronal] place of articulation, Spirantization would not occur as in [kaldo] caldo

22

 

‘broth’. Harris makes an assumption that lll is {-continuant]. Harris (1969239) states
“...nasals and l comprise just the set of non continuant sonorants in the dialect of Spanish -

[Mexican] under study.” For Harris, the features for lll are [+son, -cont, +cor].

 

 

ldigol ‘I say’ lkalbol ‘bald’ lkaldo/ ‘broth’
Sp. rule: 7 B ----
[diYD] [M1130] [kaldo] 5
i

 

Figure 3.1: Spirantization rule application

 

An alternative analysis to having a Spirantization rule is to posit a stop formation rule
such as the one proposed in Goldsmith (1981). In this type of analysis, the voiced
obstruents are Specified as [+continuant] underlyingly and a rule changes stops to

fricatives in a certain environment. A stop-formation or fortition rule is:

[-son,+vcd] -) {-cont] / || _; {-cont, acor]
[acor] (Goldsmith 1981210)

The nile states that underlying fricatives become stops utterance initially and
following a non-continuant which shares the feature [+l-coronal]. In this analysis also, /1/

in Spanish must be {-continuant] (Goldsmith 1981:11).

23

 

 

l5i‘yol ‘I say’ lkalBo/ ‘bald’ lkalfio/ ‘broth’
S-F rule: (1 ---- d

[diyo] [kalBo] [kaldo]

 

 

Figure 3.2: Stop-Formation rule application

Other stop-formation analyses (such as Nemer 19842248) include homorganicity as
part of the rule (represented as [aplace of articulation]). Goldsmith argues against the
necessity of including homorganicity as a part of the rule since nasal assimilation will

occur before Spirantization. Goldsmith includes the [acoronal] in order to account for [d]

 

following [1].

3 .3 Underspecification Analyses

In her 1979 dissertation, M.C. Lozano analyzes both Spirantization and stop
formation rules in Spanish. She supported a stop formation rule based on empirical
factors but the important conclusion She found was that the Spanish obstruents [b, d, g]
could be underspecified for the feature [continuant]. She proposed that segments such as
[b, d, g] do not derive from underlying fricatives, rather that a stop formation rule, such as
the one proposed by Goldsmith, applies to segments underlyingly unspecified for the
feature [continuant].

Mascaro (1984) agrees with Lozano’s proposal that the underlying obstruents [b, d, g]
are not specified for [continuant]. He proposes, however, that the surface value for
[continuant] is assimilated from the preceding segment. The feature value for [continuant]

is specified by way of continuancy spreading. Following an Autosegmcntal Phonology

24

 

analysis, the assimilation of the feature [or continuant] to the following segment is a

spreading process. This is a single feature assimilation as opposed to complete
assimilation (Kenstowicz 19942150). If the preceding segment is a vowel, glide or a
liquid, the following segment is spirantized because the [+continuant] value is spread. If,
however, the proceeding segment is a nasal consonant the [-continuant] value is spread.
Mascaro (19842291) bases a continuancy spreading rule on a general Spirantization

rule. A continuancy spreading rule:

[-son, +vcd] -> [acont]l [acont]

 

todo ‘all’ tengo ‘I have’
_ -CT +CT +CT -CT +CT -CT +CT
| | l | | I |
t o d o t e n g 0
-CT +CT +CT -CT +CT -CT +CT
| | \ | | l I \ l
t o d o t e n g o

 

 

 

Figure 3.3: Continuancy spreading (cf. Mascaro 19842292)

This continuancy spreading rule is a post-lexical rule which means that the rule can
apply across word boundaries and fill in, specify, or refer to non-distinctive features

(Kenstowicz 19942209).

25

 

 

lkalbol calbo ‘bald’ ldigol digg ‘I say’ ltengo/ tengo ‘I have’
(a) l3 d Y s

[16111130] [diYOI [teng0]

(a) = application of continuancy spreading rule

 

 

Figure 3.4: Continuancy spreading rule application

Figure 3.4 illustrates the application of the continuancy spreading rule and the output

 

of the rule in three words. For the word calbo, the rule applies to [b]. In a continuancy
spreading analysis, /I/ is [+continuant] (see section 3.4.3.2). The [+continuant] value of lll

spreads to [b]. For the word dig, the rule applies to both [(1] and [g]. The [+continuant]
value of the vowel [i] spreads to [g] which results in the output [7]. For the word _te_ngg_,
the rule applies to [g]. The {-continuant] value of lnl spreads to [g].

The cases of [(1] following an l1/ and in utterance initial position afe not covered by the
continuancy spreading rule. The first case would be handled by a properly including rule
such as (a):

(a) [-son, +vcd, +cor] —-> {-cont] / [l] _
The utterance initial case would receive {-continuant] by default since there is no

preceding segment for spreading.

26

3.4 New Analysis for Spanish Spirantization
3.4.1 The Markedness Value for [continuant]

The phonological feature [+l- continuant] is the focus of this study. In section 2.1,
criteria for determining markedness values were presented from both a universal and a
language particular viewpoint. It was proposed that a general ‘universal’ markedness
value could be reversed in a particular environment in a language. This proposal is
supported by the so-called “markedness-reversal” examples given in the past (cf.
Battistella 1990) and also by the Elsewhere Condition (Kiparsky 1973 ). The Elsewhere
Condition states that the particular overrules the more general application of the rule. In
this case the more specific environmental condition may overrule the more general
‘universal’ markedness value.

The frequency of the unmarked value is important in determining both a universal and
a language particular markedness value. As mentioned in section 2.1.2.2, there are
different kinds of frequency that are used. Type frequency counts the number of
language inventories that a particular phone occurs. Token frequency calculates the
percentage of usage of a particular phone in a language.

The value {-continuant] is the unmarked value in general. This value can be

supported by the criteria of relative frequency (i.e. type frequency) and complexity.

In Maddieson (1984:45, 206), the frequencies [b, d, g, B, 6, 'y] are listed on the basis of

surveying 317 languages. The stops [b, d, g] have a higher frequency of occurrence across

the worlds languages than the corresponding fricatives.

27

 

 

[b] 198/317 62.4% [0132/317 10.0%
[d] 195/317 61.5% [5]21/317 6.6%
[g] 175/317 55.2% [7] 40/317 12.6%

 

Figure 3.5: Frequency of sounds in 317 languages (Maddieson 1984245, 206)

On the basis of frequency data and the sense that [+continuant] is more complex since
it ‘adds’ the mark of continuancy to the articulation, the value of {-continuant] is
unmarked and [+continuant] is marked.

It was shown above that stops occurred in greater frequency than fiicatives in a survey
of 3 17 languages. However, the distribution of occurrence in a particular language such as

Spanish gives a different result. In Spanish, the frequency of [b, d, g] and [[3, 5, '7] were

calculated by Navarro-Tomas (1946: 15-30) for Castillian Spanish. His findings are listed

in figure 3.6.

 

 

lb/ 0.29% [B] 2.25%
ldl 1.22% [5] 4.02%
/g 0.08% [7] 0.96%

 

 

 

Figure 3.6: Frequency calculations of certain Spanish phones out of 1,000 words (from
Navarro-Tomas 1946)

The percentages in figure 7 represent token frequencies of Spanish phones out of a

text of 1,000 words. In Spanish, the fricatives occur with higher frequency and

28

[+continuant] can be the unmarked value for Spanish voiced obstruents in the positions in
which they occur.
In the comparison of the token frequencies of stops and fricatives in English, fricatives

do not greatly outnumber stops. In most cases the frequency of stops is slightly higher

or about the same as fiicatives. English does not have [7] or [B] and l5/ is a phoneme, not

 

an allophone of ldl.
/p/ 694 1.425% /U 2248 4.615% /k/ 1414 2.903%
/f/ 692 1.421% lel 338 0.796% lh/ 793 1.628%
lb/ 8771.800% ldl 1827 3.751% lg/ 598 1.228%

/v/ 738 1.515% l5/ 1354 2.780%

 

 

Figure 3.7: English phoneme frequencies out of 10,000 words (Carterette and Jones 1974 :
445)

In general, stops outnumbered corresponding fricatives. The value {-continuant]

is unmarked in English for obstruents which conforms to the universal markedness value

for obstruents.

In Spanish also,then, {-continuant] is unmarked, but [+continuant] is unmarked in
contexts where the voiced fricatives [[3, 8, 7] occur (i.e. obstruents following segments
specified as [+cont]). In a language such as English, there is no allophonic distinction

between ldl and /5/, and [[3, y] do not exist in English. Therefore since a more particular

markedness rule does not apply, the universal {-continuant] unmarked value applies to all

English obstruents.

29

 

3.4.1.2 Classification of Feature [continuant] for Spanish Liquids
3.4. 1.2.1 Classification of Spanish tap lr/

Spanish distinguishes between a voiced alveolar trill lrrl and a voiced alveolar tap lrl.
The tap and trill both occur intervocalically and therefore contrast in that position. An
example of a minimal pair is [perro] perro ‘dog’ and [pero] pero ‘but’. The tap and trill
are both classified as [+continuant]. In Mascaro’s continuancy spreading analysis, the
tap [r] spreads a [+continuant] value (Mascaro 19841290).

In this study, I make the assumption that the English flap [D] is not identical to the
Spanish tap [r] in the way it is produced. Chomsky and Halle (1968) support the claim
that the tap [r] and flap [D] may be produced in different ways. They write:

It may be noted...that the tap [r] may be produced by a different mechanism than
the so-called “tongue flap” [D] which greatly resembles the tap [r]. Whereas the
latter is the result of [subglottal pressure], it is quite possible that the tongue flap
[D] is produced by essentially the same muscular activity that is found in the
dental stop articulation, except that in the case of the tongue flap the movement is
executed with great rapidity and without tension. (1968:318)

Spanish tap [r] will then be assumed in this study to be [+continuant] and produced
differently than the English flap [D]. Section 3.4.2.2 discusses these issues in more depth

in relation to markedness and the misperception of native Spanish speakers to native

English speaker productions.

3.4.1.2.2 The Classification of Spanish lll

Harris (1969) and Goldsmith (1981) assume lll to be {-continuant]. It is preferable,

however, to classify /1/ as [+continuant]. If lll is [+continuant], then [113] and [17] would

be unmarked while leaving only [1d] to be explained as the marked sequence.

30

Padgett (1991) and Kenstowicz (19942488) connect [continuant] with place of
articulation in order to explain why only [d] follows [I] in Spanish. Kenstowicz
(19942488) proposes that homorganicity blocks the spread of continuancy due to an
inalterability effect. Inalterability is from autosegmental phonology and states that when
two segments are linked together by sharing features, that one part cannot be changed
apart from the other.

Kenstowicz and Padgett assume lll to be [+continuant]. The analysis that they
propose is that lll first delinks the feature [+continuant] and then assimilates the coronal
features of the following [d]. Harris (1985) supports this assimilation by stating that IV is
normally alveolar in Spanish but takes on the dental articulation of [d] in the word [kaldo]
‘broth’.

In the cases involving [b] and [g], ll/ does not assimilate to either place of articulation

so there is no inalterability involved. The [+continuant] of lll spreads to [b] yielding

[kalBo]. Kenstowicz (19942488) suggests that lll delinks [+continuant] and assimilates

the {-continuant] of [d]. However, the dental assimilation of IV must precede the

continuancy spreading rule because if spreading occurred first, the output would be

*[kalfio]. Kenstowicz and Padgett posit an underlying {-continuant] for [d] which is

spread to [1] when linked together. If [(1] is underspecified for [continuant], then an ad

hoc rule would have to apply in order for the marked for [kaldo] to be represented.

31

3.4.2 The Application of Markedness Rules
3.4.2.1 Spanish

Chomsky and Ha11e(1968) presented a list of markedness conventions that functioned
like phonological rules to fill in markedness values (19682403). Markedness rules in this
study will be used to illustrate the differences in markedness values for different contexts
in Spanish. Separate markedness rules will be given for each value of [+l- continuant]. In
the following rules, ‘ [u cont]’ means “the unmarked value for [continuant]”.

(a) [u cont] —> [+cont] / [+cont] [-son, _]
(b) [u cont] —> [~cont] / [-son, _]

In both cases, the target of the rule is an obstruent which is underspecified for the
value [continuant] such as [b, d, g] in Spanish. For rule (a), [+ continuant] is the
unmarked value if an obstruent underspecified for [continuant] is preceded by another
segment with a [+continuant] value. Rule (b) functions as an ‘elsewhere’ rule in which
[- continuant] is the unmarked value for those obstruents not following [+continuant], or
in initial position. Markedness rule (b) is a universal rule and would apply in all
languages.

Figure 3.8 illustrates markedness rule application. In the case of [kasa] ‘house’, the
[s] is lexically specified as [-sonorant, +continuant]. Markedness rule (a) applies to the
position following the vowel. The [s] in cg is unmarked for [continuant] since it

follows a [+continuant] segment.

32

 

ltengo/ tengo ‘I have’ lada/ hada ‘fairy’ lkasa/ 9% ‘house’

lnl lg/ lal ldl lal lsl
[-son] [-son] [+son] [-son] [+son] [-son]
{-cont] [Ocont] [+cont][0cont] [+cont][+cont]
/...n_.../ l...a_...l l...as.../
apply rule (b) apply rule (a) apply rule (a)
{-cont] is unmarked [+cont] is unmarked [+cont] is unmarked
output: [tengo] [a5a] [kasa]

 

 

 

Figure 3.82 Markedness rule application
The next environments to analyze are those which have a liquid followed by an

obstruent. If the M preceding an obstruent is [+continuant], then markedness rule (a) will

correctly predict the unmarked forms [B, 5, y].

Environments which cause some difficulties in determining markedness are those
involving lll. In Spanish, [(1] may follow /1/ but not [b] or [g]. Only. the fricatives [B] and
[7] may follow lll. Harris (1969239) and Goldsmith (1981211) assume that lll is

{-continuant]. In this study, /1/ will be [+continuant] and the analysis was given in section
3.4.1.2.2 above.
Assuming that lll is [+continuant], the unmarked value for obstruents following /1/

would be [+continuant]. This would predict that [(1] following /I/ would be marked.

33

 

This study focuses on Spanish ldl and lr/ and misperceptions based on these sounds.

Figure 3.9 compares the phonemic and phonetic feature representations for Spanish ldl

 

and lrl.
Spanish ldl Spanish ldl Spanish lr/
[~sonorant] [-sonorant] [+sonorant]
[+voiced] [+voiced] [+voiced]
[+coronal] [+coronal] [+coronal]
[0 continuant] [0 continuant] [Ocontinuant]

Context: [+cont] _

markedness rule (a)

[5]

[-Sonorant]
[+voiced]
[+coronal]
[+continuant]

Context: Elsewhere

markedness rule (b)

[d]

[~sonorant]
[+voiced]
[+coronal]
{-continuant]

redundancy rule:
[+son] —-> [+cont]

[r]

[+sonorant]
[+voiced]
[+coronal]
[+continuant]

 

Figure 3.9: Spanish phonemic and phonetic representation comparisons

The Spanish ldl is [~son, +vcd, +cor, 0cont]. The markedness rules

(a) [ucont]—> [+cont] / [+cont] [-son, _]

(b) [ucont] —) {-cont] / [-son, __]

fill in the values for continuancy. In the context [+cont] _ , the first rule would apply

resulting in the voiced fricative [5]. The features of [6] are [-Son, +vcd, +cor, +cont].

Elsewhere the second rule would apply resulting in a [d]. The features of the voiced stop

[d] are [-son, +vcd, +cor, ~cont]. These features are the same as English [(1] except that

English [(1] is produced at the alveolar ridge while Spanish [d] is produced dentally.

34

 

The Spanish tap /r/ is [+son, +vcd, +cor, 0cont]. The redundancy rule

[+son] -> [+cont] also applies to the tap lrl. Phonetically, the features of [r] are [~son,

+vcd, +cor, +cont].

3.4.2.2 English
When markedness nrles are applied to English the results differ from Spanish. The
following rules were shown to apply to Spanish obstruents in order to determine the
markedness value for [continuant].
(a) [u cont] —-> [+cont] l [+cont] [-son, _]
(b) [u cont] —-> {-cont] l [-son, _ ].
English violates rule (a) because [-son, +cont] can be specified. English has words
with post-vocalic stops (cf. [dogi] Egg and [kaebi] calyx) which violate rule (a).
English does not violate rule (b). Rule (b) is the more general rule because it refers to a

superset of the cases referred to by rule (a).

The English phones relevant to this study are ldl and lfil. Figure‘3. 10 compares the

phonemic and phonetic representations for English ldl and lfil.

35

 

English ldl

[-sonorant]
[+voiced]
[+coronal]

[0 continuant]

Context: V_ V
flapping rule

ID]

[-sonorant]
[+voiced]
[+coronal]
[+continuant]

 

English ldl

[-sonorant]
[+voiced]
[+coronal]

[O continuant]

Context: Elsewhere

markedness rule (b)

[d]

[-sonorant]
[+voiced]
[+coronal]
{-continuant]

English /5/

[-sonorant]
[+voiced]
[+coronal]
[+continuant]

[5]

[-sonorant]
[+voiced]
[+coronal]
[+continuant]

 

Figure 3.10: English phonemic and phonetic representation comparisons

The English ldl is [-son, +vcd, +cor]. Following radical underspecification,

[continuant] is unspecified. For the phoneme /S/ in English, the feature [+continuant] is

Specified underlyingly. In English a flapping rule ld/—> [D] / V_V applies

intervocalically. The phonetic form for [D] is [-son, +vcd, +cor, +cont].

3 .5 The Equivalence of English [D] and Spanish [r]

The English flap [D] and Spanish tap [r] share the features [+vcd, +cor, +cont] but

they differ with respect to the feature [sonorant]. English [D] is [-Sonorant] while

Spanish [r] is [+sonorant]. English [D] and Spanish [r] are also produced slightly

36

 

differently. According to Chomsky and Halle (1968), the Spanish tap [r] is produced by
subglottal pressure, while the English flap is a rapid articulation of a stop.

Equivalence Classification (F lege 1987, 1988) classifies sounds between L1 and L2 as
‘new’, ‘similar’, and ‘identical’. A sound in an L2 would be classified as ‘similar’ to an
L1 sound due to the sharing of distinctive features. English [D] and Spanish [r] are similar
on the basis of equivalence classification. The sounds share the features [+vcd, +cor,
+cont]. English [D] and Spanish [(1] are also similar sounds Since they share the features
[-son, +vcd, +cor]. English [D] and Spanish [r] differ with respect to the feature
[+l-sonorant]. English [D] and Spanish [d] differ with respect to the feature [+l-
continuant].

In section 2.1, a theory of markedness was shown to be context sensitive. The more
general, universal marked value could be unmarked in a certain environment. In section
3 .4, it was argued that in Spanish [+continuant] is the unmarked value in the context
following a [+continuant] segment. Equivalence classification and markedness are
separate notions. The similarity between two sounds is not affected by markedness per
se. The preference for one of two possible similar sounds, however, can be affected by
markedness. English [D] is similar to both Spanish [(1] and [r]. English [D] differs by
one feature with Spanish [d] ([+l-continuant]) and Spanish [r] ([+l-sonorant]).

Native Spanish speakers (N SS) may prefer the perception of [r] over [d] when a native
English speaker produces [D] because of markedness. The [D] and [r] share the
[+continuant] feature value which is unmarked in Spanish following a [+continuant]
segment. Spanish [(1] is {-continuant] and does not appear in the context following a

[+continuant] segment in Spanish. Therefore, NSS would prefer the unmarked

37

[+continuant] over the marked {-continuant] so [r] would be perceived over [(1] due to

markedness.

38

Chapter 4
TEST INVOLVING PROCESSING ERRORS 1N SECOND LANGUAGE

ACQUISITION

4.1 Preliminaries
The test was designed to collect data in order to duplicate the generalizations of
processing errors as shown in Stockwell and Bowen (1965). It was proposed by

Stockwell and Bowen (1965:44-47), Politzer & Staubach (1965271), and Nunez-Cedeno
(1987:363-3 64) that English L1 speakers may mispronounce Spanish [5] as English [(1].

They also proposed that a native Spanish speaker would hear this as Spanish tap [r] and

misperceive a word if there was a minimal pair (Nunez-Cedeno 1987:364).

4.2 Testing Method
4.2.1 Subjects

The subjects of the test are 5 native English speakers who were learning Spanish at the
college level, and there were 5 native Spanish speakers who acted as a control group for
the study. There were also 3 native Spanish speaker judges who listened to recordings
made by the 10 subjects. The eight native Spanish speakers in this test were either from
Cuba or Mexico and now live in the United States (It was my intention to have all my

subjects speak one dialect but I had to use subjects from two dialects. Though this was

39

not an ideal situation, I did not encounter any unusual problems with my results. There
is no difference in Spirantization between Cuban and Mexican dialects (cf. Harris 1969,
Guitart 1976)). Of the control group, 3 spoke a Cuban dialect and 2 spoke a dialect of
Mexico. Of the judges, 2 judges were from Cuba and one spoke a Mexican dialect. Of
this group there were none who were fluent bilingual speakers of English. Each of the
subjects and judges have Spanish as their first language and Spoke a Cuban or a Mexican
dialect.

The 5 native English speakers ranged in age from 19 to 32 years. All learned Spanish

in the United States and had from 2 to 6 years of instruction.

 

 

 

 

Native English Speakers
Ag; Years of Spanish Study
(1) 32 6 years
(2) 19 4 years
(3) 21 3 years
(4) 25 4 years
(5) 28 2 years

 

Figure 4.1: Information on Native English Speakers involved in test

4.2.2 Testing procedure
The test in this study was designed to elicit processing errors by Spanish and English
speakers. The 10 subjects heard a list of 10 Spanish words read by a native speaker of

Spanish in a Cuban dialect (see figure 4.2). In the list, 5 of the 10 words included the

voiced obstruent [5] between vowels. (Originally the test included 5 additional words

40

 

 

with intervocalic [B]. The English speakers generally heard and pronounced the voiced

bilabial fricative [B] as the English voiced labio-dental fricative [v]. This substitution

labeled the speaker as ‘non-Spanish’ but there was no misperception by the NSS. This

study will focus on data concerning Spanish ldl~[8].)

After each word was read, the subjects wrote down the word they believed they heard.
After they finished listening to the tape and wrote down the 10 words, they turned their
sheet over and listened to the tape again. They were recorded producing each word after
they listened to the word Spoken on the tape. There were 5 words included as
distracters. The recordings were later played for the 3 native Spanish speaker judges.
The native speaker judges listened to all of the recorded words and wrote down the word
they believed they had heard. The list included words which are members of ldl ~ lrl

minimal pairs (see figure 4.3). For example, it was expected that the native English

speakers may incorrectly pronounce Spanish words with l8/ with an English ldl and the

Spanish judges would hear those words as having a Spanish lrl. For example, cada would

be pronounced with English ldl =[D], and this would be heard as Lara.

 

l. mudo 6. hada
2. mujer 7. bueno
3. seda 8. todo
4. gato 9. banco
5. cada 10. perro

 

 

Figure 4.2 Word list used in study (in random order)

41

 

 

(1a)
C23)
C36)
043)
(5a)

cada [kafia] ‘each’
hada [a5a] ‘fairy’
mudo [mufio] ‘mute’
seda [se5a] ‘silk’
todo [tofio] ‘all’

_fiyr minimal pairs

(1b)
(2b)
(3b)
(4b)
(5b)

cara [kara] ‘face’
hara [ara] ‘will do’
muro [muro] ‘wall’
sera [sera] ‘will be’
tom [tom] ‘bull’

 

Figure 4.3 [5~r] rrrinimal pairs in Spanish

42

 

Chapter 5

RESULTS AND ANALYSIS OF TEST

5.1 Predictions on the Basis of Equivalence Classification and Markedness

The background for analyzing processing errors in SLA is ‘equivalence classification’
(EC) which was introduced in section 2.4. Flege (1988) and (1991) uses EC in analyzing
the production errors and the perception errors of the second language learner. This
Study applies EC to the perception of the native speaker to a second language leamer’s
production of the second language.

James (1986) analyzes processing errors that occur between native and non-native
speakers and he states that phonological context can contribute to phonic transfer
(1986: 140). James (1988224) gives an example of how phonological context relates to

phonic transfer. When a native Dutch speaker speaks English as an'L2, the pronunciation

of English /5/ may be affected by context. The specific context involved is the

‘suprasegmental context’ which is the location of the phoneme such as syllable initial or

syllable final within a word. Dutch speakers tend to produce English l5l as a stop [d] or

tap [D] word-initially and as a fricative [z] or [s] in other positions (James 1988224).
This study proposes that when a native Speaker perceives the production of a non-

native speaker, the native speaker will identify the sound in terms of an L1 category.

43

Misunderstanding will result if non-native speakers produce a sound in terms of their L1
and this is perceived as a Similar sound of the L2.

It is predicted by EC that a phone identical in the L1 and L2 will be perceived and
produced the same in the L1 and L2 by non-native speakers. A new sound of the L2 will
be given a new category by non-native Speakers and will be processed as such. A sound
in the L2 which is similar to one of the L1 will be classified as that of the L1 by non-
native speakers. (cf. Flege 1988:367-369).

In the case involving English ldl and Spanish lrl, the following is predicted by
equivalence classification. First, English [D] and Spanish [r] are similar sounds based on
the sharing of the features [+vcd, +cor, +cont]. In an intervocalic environment, English
[D] has a high similarity to Spanish [r] based on the actual equivalence of distinctive
features. As discussed in section 3.5, Spanish [r] is preferred by NSS over Spanish [(1]
due to [r] being unmarked following a [+continuant] segment and [d] being marked in that
position.

Second, because of this high degree of similarity between [D] and [r], a new category
will not be created by a NSS. The similarities between English [D] and Spanish [r] do not
result in the creation or recognition of a new category by a NSS, but rather a presumed
equivalence based on the actual equivalence of distinctive features.

In section 3 .4.1 it was proposed that [+continuant] is unmarked in the contexts
following a segment specified as [+continuant] in Spanish. A stop obstruent such as [b]

or [(1] would be marked in an intervocalic position in Spanish.

44

Therefore due to EC and Markedness Theory, this study hypothesizes the following

results:

(1) if an NES produces the Spanish word gegg he will replace Spanish ldl with
English ldl. This will undergo the English flapping rule so an NES will

produce [seDa].

(2) if an NES produces §e_da_ as [seDa] then a NSS will hear [sera] due to (a) the
perceived equivalence of [D] to [r] and (b) that Spanish [(1] is marked in an
intervocalic position. The feature [+continuant] is unmarked in Spanish
intervocalically so the result is the perception of [r] (which is unmarked

due to being [+cont]) rather than [(1] (which is marked due to being [-cont]).

For example, todo, seda, and mudo would be perceived as toro, sera, and muro

respectively by a NSS.

5.2 Test Results
5.2.1 Results of Native Spanish Speakers

The test included five Native Spanish Speakers (NS S) as a control group. Their results
are listed in Table 5.1. In each table there are 15 cases of each word that were heard by
the NSS judges. This is because each word was spoken by 5 NES’S and each word was
heard by the 3 judges (5 x 3=15). Overall, the judges heard each word 30 times because

there were 15 productions from NES and 15 from the NSS control group.

45

 

 

Spanish word spoken by
L1 Spanish speaker

Word written by native Spanish speaker Judge

 

(1) mudo
(2) $9.2

(3)t_o_d2

(4) £42

(aw

mudo 15/15 100%

seda 12l15 80%

todo 14/15 93.3%

1 unsure

cada 15/15 100%

hada 15/15 100%
se.

muro
cera

tOI'O

hara

0/15

3/15

0/15

0/15

0/15

0%

20%

0%

0%

0%

 

 

Table 5.1: Results from Native Spanish Speakers

When judges listened to the native Spanish speaker control group, in all cases but one,

there were no instances of mishearing [8] as [r]. In the case of gd_a, all three judges heard

the same subject produce this word as gar; instead of seda. This uniformity of responses

would lead us to believe that the subject did in fact pronounce the word with an [r]-like

quality.

5.2.2 Results of Native English Speakers

The five native English speakers (NES) were tested for their production of Spanish

[5]. Table 5.2 gives the results for the five words with a possible misperception of ldl

and lrl for native English speakers.

46

 

 

Spanish word spoken by Word written by native Spanish speaker judge

Ll English speaker

 

 

(1)mudo mudo 14/15 93.3%

(2)_s_e_dg seda; 6/15 40%

(3) todo todo 12l15 80%

(4)cada gag 13l15 86.7%

(5) hada hada 13/15 86.7%
fl

muro

861' a

1/15 6.7%
9/15 60%
3/15 20%
1/15 6.7%
1 unsure
0/15 0%
2 unsure

 

Table 5.2: Results from Native English Speakers

The results of Table 5.2 show that misperception does not occur in every case but did

in some. Even when the correct word with [8],such as cada was chosen by the judge,

there was a difficulty in choosing cada or gag. In two cases, judgesrefused to write an

answer when they were unsure which one was produced (these are indicated with

‘unsure’ in the table above).

5.3 Analysis

5.3.1 Native Spanish Speakers (NSS)

As noted in table 5.1, there was no pattern of misperception between native Spanish

speakers in this study. Figure 5.1, below, illustrates the derivations of the Spanish words

donde, sedg, and m.

47

 

 

 

3 ’

SJranish ‘donde Spanish ‘seda Spanish ‘toro’

ldonde/ lseda/ ltorol
markedness rule (a) ----- 5 -...-
markedness rule (b) d d -..... -..--
[donde] [se5a] [toro]

 

Figure 5.1 Derivations of Spanish words

In the word £19313, markedness rule (b) applies to both instances of ldl to give the
phonetic representation [donde]. Markedness rule (a) applies to £ng to give [sefia].
Neither markedness rule applies in the case of tofl because [r] is a sonorant. These
derivations illustrate the phonology of NSS. No misperceptions resulted from the NSS
because the markedness rules were applied with the NSS control groups productions.
The only case of hearing [r] by the NSS judges was when all three judges wrote that one

subject said sera instead of seda.

5.3.2 Native English Speakers (NES)

The test Showed that NSS did misperceive some NES productions as [r] instead of [5],
however, the results do not conform to the expected results hypothesized in section 5.1.
According to the hypothesis, all cases should be misperceived. Table 5.2 showed that [r]
was misheard for [5] for mu_rq, E, £933, and fl. However, only _cgrg was misheard a
majority of instances (for 9 of 15 words).

There are a few reasons why the results were not as hypthosized. First, the data was

of subjects listening to a native speaker of Spanish and then reciting a word list and not

of conversational speech. Second, the subjects wrote down what word they heard so

48

 

 

 

orthography could have had an effect on their speech. Results would likely be different if
conversational speech was analyzed.

The results did not follow exactly as predicted but that does not mean that nothing
was found. The comparison of results from native Spanish speakers (N SS) and native
English speakers (NES) reveals that markedness may be involved in the production and
perception of Spanish obstruents, which supports the premise behind this study. The
results of NSS perception were different between NSS and NES subjects.

While there were instances of misperception involving a NSS subject, there were also a
few misperceptions involving NES subjects as mentioned above. This suggests that the
L1 of the NES subjects and their phonological systems (comprised of underlying systems
and rules) affect their speaking of an L2 (in this case Spanish).

Figure 5 .2 illustrates the rules that occur in English and gives derivations for rider,

 

 

 

leather, and candy.
Engish ‘rider’ @glish ‘leather’ - English ‘candy’
lrayder/ /l£ 5tr/ lkandi/
flapping rule D —-..- --..-..
markedness rule (b) ----- --.... d
[rayDer] [l a 5 E r] [kandi]

 

Figure 5.2 Derivations of English words

In these words, the flapping rule ldl —-) [D] l V_ V only applies in rider.

Markedness rule (b) applies to the ldl in candy as the universal default rule but does not

apply in rider since the flapping rule precedes and bleeds the markedness rule. In radical

49

 

 

underspecification, l5/ in English [+continuant] would be specified underlyingly since it is

marked in English while {-continuant] is assigned to ldl in 9% by markedness rule (b).
In this study, NES produced Spanish words and those words were heard by NSS
judges. In the cases involving misperception it is proposed that NES applied English
phonological rules when producing Spanish and these were misperceived by NSS due to
markedness and equivalence classification. Figure 5.3 illustrates NSS nrisperception of a

NES production.

 

Spanish todo ‘all’

ltodol
(a) D English flapping rule ldl ->[D]l V__V applied by NES

(b) [toDo] this production with [D] heard by NSS as [+son, +cont, +cor, +vcd]

(c) [toro] markedness theory blocks intervocalic [d] in Spanish, and
[r] is perceived by NSS on the basis of equivalence classification

 

Figure 5.3: Misperception of NES production by NSS

5.4 Conclusion
This study has focused on the misperceptions of native Spanish speakers (N SS).

It has been previously documented that misproductions and misperceptions occur
between English and Spanish speakers concerning [5] and [r] (Stockwell and Bowen
(1965:44-47), Politzer and Staubach (1965171), Nunez-Cedeno (1987:363-364),

Zampini 1994: 471)). These processing errors have been explained in this study by

appealing to markedness theory and to equivalence classification.

50

 

This study has only scratched the surface in applying markedness theory to L1-L2
processing situations. The application of markedness theory and equivalence

classification to L1 perception of L2 productions is an issue for future research.

51

REFERENCES

Alarcos Llorach, E. (1965). Fonologia espanola Madrid: Editorial Gredos, 4a. edicion.
Andrews, E. (1990). Markedness Theory. Durham: Duke University Press.

Archangeli, D. (1984) Underspecification in Yawelmani Phonology and Morphology.
Ph.D. dissertation. MIT, Cambridge, MA.

Archangeli, D. (1988). Aspects of Underspecification Theory. Phonology. 5: 183-208.

‘1

Archangeli, D. and D. Pulleyblank. (1989). Yoruba Vowel Harmony. Linguistic Inquiry.
20:173-217.

Batti stella, E. (1990). Markedness: The Evaluative Superstructure of Language. Albany: I
SUNY Press. ”

 

Bohn, O-S. and J. Flege (1990). Interlingual indentification and the role of foreign
language experience in L2 vowel perception. Applied Psycholinguistics. 112 303-
328.

Cairns, C. (1969). “Markedness, Neutralization, and Universal Redundancy Rules”.
Language. 45:863-85.

Cardenas, J. (1961). Applied Linguisitics: Spanish. Boston: DC Heath and Co.

Carterette, E. & M. Jones. (1974). Informal Speech. Berkeley: University of California
Press.

Chomsky, N. and M. Halle (1968). The Sound Pattern of English. New York: Harper
and Row.

Cressey, W. (197 8). Spanish Phonologi and Morphology: A Generative View.
Washington, DC: Georgetown University Press.

Dinnsen, D. (1996). Context-Sensitive Underspecification and the Acquisition of
Phonemic Contrasts. Journal of Child Language. 23: 57-79.

Dinnsen, D. & S. Chin. (1995). On the Natural Domain of Phonological Disorders.

In Phonological Acquisition and Phonological Theory. J. Archibald (ed). Hillsdale,
NJ: Lawrence Erlbaum Associates, Publishers.

52

 

Eckman, F. (1977). Markedness and the Contrastive Analysis Hypothesis.
Language Learning. 27: 315-330.

Flege, J. (1987). The production of ‘new’ and ‘similar’ phones in a foreign language:
evidence for the effect of equivalence classification. Journal of Phonetics. 15:47-65.

Flege, J. (1988). The production and perception of foreign language speech sounds.
In Human Communication and Its Disorders, Vol. II. H. Winitz (ed.). 224-3 89.
Norwood, NJ :Ablex Publishing.

Flege, J. (1991). Perception and production: the relevance of phonetic input to L2
phonological learning. In Crosscurrents in Second Language Acquisition and
Linguistic Theories. Huebner and Ferguson (eds). 249-289. Philadelphia: John
Benjamins.

Flege, J. & J. Hillenbrand (1984). Limits on pronunication accuracy in adult foreign
language speech production. Journal of the Acoustical Society of America. 76:708-
721.

Goldsmith, J. (1981). Subsegrnentals in Spanish Phonology: An Autosegmcntal
Approach. In Linguistic Symposium on Romance Linguistics: 9. W. Cressey and
D. Napoli (eds) 1-16. Washington DC: Georgetown Univ. Press.

Goldsmith, J. (1990). Autosegmcntal and Lexical Phonology. Cambridge: Blackwell.
Greenberg, J. (1966). Language Universals. The Hague: Mouton.

Guitart, J. (1976). Markedness card a Cuban Dialect of Spanish. Washington, DC:
Georgetown University Press.

Harris, J. (1968). Spanish Phonology. Cambridge: MIT Press.

Harris, J. (1985). Autosegmcntal Phonology and Liquid Assimilation In Havana Spanish.
In Papers from the 13th Linguistic Symposium on Romance Linguistics. L. King and
C. Maley (eds). 127-148. Amsterdam: Benjamins.

Iverson, G. (1993). (Post) Lexical Rule Application. In Studies in Lexical
Phonology. S. Hargus and E. Kaisse (eds) San Diego: Academic Press.

Jakobson, R. (1962). Selected Writings. Vol. I, Phonological Studies. The Hague:
Mouton.

James, A. (1986). Phonic transfer and phonological explanation: some theoretical and

methodological issues. In C ross Linguistic Influence in Second Language Acquisition.
Kellerrnan and Smith (eds). 134-149. Washington DC: Georgetown Univ. Press.

53

James, A. (1988). The Acquisition of a Second language Phonology. Tubingen,
Germany: Narr.

Kenstowicz, M. (1994). Phonology in Generative Grammar. Cambridge: Blackwell.

Kiparsky, P. (1973). “Elsewhere” in phonology. A F estschrifi for Morris Halle.
S. Anderson & P. Kiparsky (eds). 93-106. New York: Holt, Rinehart, & Winston.

Kiparsky, P. (1982). Lexical phonology and morhology. Linguistics in the Morning
Calm. I.S. Yang (ed). 131-175. Seoul: Hanshin.

Kiparsky, P. (1984). On the lexical phonology of Icelandic. Nordic Prosody III.
C.-C. Elert (ed.). 135-164. Stockholm: Almqvist and Wiksell.

Kiparsky, P. (1993). Blocking in non-derived environments. In Studies in Lexical
Phonology. S. Hargus and E. Kaisse (eds) San Diego: Academic Press.

Ladefoged, P. (1993). A Course in Phonetics. New York: Harcourt Brace Publishers.

Lipski, J. (1990). Spanish Taps and Trills: Phonological Structure of an Isolated
Opposition. Folia Linguistica. 24: 153-174.

Lipski, J. (1994). Spanish Stops, Spirants, and Glides: From consonantal to [vocalic].
In Linguistic Symposium on Romance Languages: 23. M. Mazzola (ed.).
67-86. Washington, DC: Georgetown University Press.

Maddieson, I. (1984). Patterns of Sounds. Cambridge: Cambridge Press.

Martinez-Gil, F. (1992). The insert/ delete parameter, redundancy rules, and
neutralization processes in Spanish. In Current Studies in Spanish Lingusitics.
Campos and Martinez-Gil (eds). 495-556. Washington DC: Georgetown Univ.

Press.

Mascaro, J. (1984). Continuant Spreading in Basque, Catalan, and Spanish. In Aronoff
and Oehrle (eds) Language Sound Structure. Cambridge: MIT Press.

Michaels, D. (1974). Sound replacements and phonological systems. Linguistics. 1762
69-81.

Mohanan, KP. (1986). Lexical Phonology. Reidel: Dordrecht.

Navarro-Tomas, T. (1946). Estudios de Fonologia Emafiola. Syracuse: Syracuse Univ.
Press.

Navarro-Tomas, T. (1974). Manuel de pronunciacién espafiola. Madrid: Publicaciones
de la Revista de Filogia Espafiola. Num. [[1

54

Nemer, J. (1984). Stop Formation as a Process. Anthropological Linguistics. 26, 3:
245-269.

Nhfiez-Cedefio, R. (1987). Intervocalic ldl Rhotacism in Dominican Spanish: A Non
Linear Analysis. Hispania. 70:363-369.

Padgett, J. (1991). Stricture in Feature Geometry. Ph.D. dissertation. University of
Massachusetts, Amherst.

Politzer, M. and C. Staubach (1965). Teaching Spanish: A Linguistic Orientation. 2nd
Ed. New York: Blaisdell Publishing Co.

Postal, P. (1968). Aspects of Phonological Theory. New York: Harper and Row.

Ritchie, W. (1968). On the explanation of phonic interference. Language Learning.
18: 183-197.

Sampson, G. (1980). Schools of Linguistics. Stanford: Stanford University Press.

Steriade, D. (1995). Underspecification and Markedness. In Handbook of Phonological
Theory. J. Goldsmith (ed.). 114-174. Cambridge: Blackwell.

Stockwell, R and J.Bowen (1965). The Sounds of English and Spanish. Chicago:
University of Chicago Press.

Tnlbetzkoy, N. (1931). Die Phonologischen Systems. T rauvaux du Cercle Linguistique
de Prague, No. 4. 96-116. Prague: J ednota Ceskolovenskych Matimatikua Fysiku.

Trubetzkoy, N. (1969)[1939]. Principles of Phonology. Translated by Christiane A.M.
Baltaxe. Los Angeles: University of California Press.

Waugh, L. (1979). Markedness and Phonological Systems. In The Fifth LAC US
Forum. W. Wolck and P. Garvin (eds). 155-165. Columbia, SC: Hombeam Press.

Weinreich, U. (1953). Languages in Contact. The Hague: Mouton.
Weinreich, U. (1957). On the description of phonic interference. Word 13: 1-11.

Zampini, M. (1994). The Role of Native Language Transfer and Task Formality in the
Acquisition of Spanish Spirantization. Hispania. 77:470-481.

55