modifying stockwell & bowen s hierarchy of difficulty ......1. introduction the hierarchy of...

31
Modifying Stockwell & Bowens Hierarchy of Difficulty within the Framework of Contrastive Analysis William Keegan Storrs Dr. Gillian Lord, Advisor Department of Linguisitcs May 1, 2014

Upload: others

Post on 23-Sep-2020

7 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Modifying Stockwell & Bowen’s Hierarchy of Difficulty

within the Framework of Contrastive Analysis

William Keegan Storrs

Dr. Gillian Lord, Advisor

Department of Linguisitcs

May 1, 2014

Page 2: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Abstract

The present study seeks to further investigate one of the landmark theories in the field of

linguistics and Second Language Acquisition: the Contrastive Analysis Hypothesis (CAH, Lado,

1957). Specifically, it will examine the Hierarchy of Difficulty derived from the CAH, as

proposed by Stockwell and Bowen (1965). This hierarchy seeks to make predictions, based on

the first language (L₁), about how ‘easy’ or ‘difficult’ the process of acquisition will be for any

sound in the L₂. Using data collected from college-age learners of Spanish, we provide evidence

that the hierarchy of difficulty is not, in and of itself, sufficient to predict a particular phone’s

level of difficulty with consistency, but that the tenets of the CAH can be used to explain learner

errors retrospectively. Finally, a new kind of hierarchy will be proposed, which allows sounds to

be classified, perhaps more accurately, on a scale of difficulty based on the means by which

learners go about restructuring their L₂ phonemic inventories.

Page 3: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

1. Introduction

The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based

on the theory of Contrastive Analysis popularized by Lado (1957) asserts that the ‘easiness’ or

‘hardness’ -- that is, the level of difficulty -- of the sounds in the phonetic inventory of a second

language (L₂) can be predicted based on a system of comparing the categorization of the sounds

in the first language (L₁). The sounds of the two systems are to be classified into one of three

categories in both the L₁ and the L₂: ‘obligatory,’ ‘optional,’ or ‘null.’ According to the L1 and

L2 categorizations, the sounds can thus be assigned to different difficulty levels in the overall

hierarchy of sounds. Table 1 below shows Stockwell and Bowen’s original hierarchy, as

proposed in their seminal work in the middle of the last century.

L₁ L₂ 0 Obligatory

0 Optional

Optional Obligatory

Obligatory Optional

Obligatory 0

Optional 0

Optional Optional

Obligatory Obligatory

Table 1: Stockwell and Bowen’s (1965) Hierarchy of Difficulty

Note that Stockwell and Bowen use the term ‘categories,’ but in modern terms this

concept would most probably be labeled as “phonemic,” “allophonic” or “non-existent.” For the

Harder

Easier

Page 4: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

purposes of this paper, we can update this chart, giving us Table 2, which shows the

corresponding modern-day terminology, as well as the L₁ and L₂ examples that will be

investigated in the present study.

English Spanish Example

Non-existent Phonemic /r/

Non-existent Allophonic [β]

Allophonic Phonemic /ɾ/

Phonemic Allophonic [ð]

Phonemic Non-existent N/A

Allophonic Non-existent N/A

Allophonic Allophonic None exist

Phonemic Phonemic /f/ /l/ /e/

Table 2: The Hierarchy of Difficulty, using modern-day terminology and showing the phones to

be tested by the present study.

In the most difficult category, for example, we see that the Spanish trill /r/ (as in the word

perro /pero/) is an example of a sound that is phonemic, or obligatory, in the L₂ but is non-

existent in the L₁. Only slightly easier is the voiced bilabial approximant [β] (as in the word

labio [laβi̯o]), a sound that is allophonic, or optional, in the L₂ (as an allophone of the phoneme

/b/) but, like /r/ is non-existent in the L₁. Predicted to be somewhat easier is the alveolar tap /ɾ/

(as in Spanish para ti /pa.ɾa.ti/ and English pot of tea [pʰɑ.ɾə.tʰi]), an example of a sound that is

phonemic (obligatory) in the L₂ but allophonic (optional) in the L₁, occurring as an allophone of

/t/ and /d/. Moving towards sounds that are predicted to be easier still, we find the voiced dental

approximant [ð] (as in Spanish lado [la.ðo] and English lather /læ.ðɚ/), a sound that is

Harder

Easier

Page 5: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

allophonic (optional) in the L₂ as an allophone of /d/, but phonemic (obligatory) in the L₁. We

have marked the next two categories as N/A, because these levels represent sounds that are not

included in the L₂ sound inventory and, therefore, we cannot account for them in an experiment

that tests for L₂ production. For the allophonic/allophonic category (optional/optional), there are

no sounds that fall under the category of allophone in both English and Spanish inventories.

Finally, we arrive at /f/, /l/, and /e/ (Spanish faz /faθ/, animal /a.ni.mal/, and peso /pe.so/ and

English face /fejs/, animal /æ.nɪ.mlˌ/, and peso /pej.sow/) which all fall under the easiest level of

the hierarchy, as they are all phonemic (obligatory) in both the L₁ and the L₂.

Although many linguists are well acquainted with the theoretical assertions of Lado and

of Stockwell and Bowen, surprisingly little empirical work has been done to investigate the

validity of this hierarchy. Therefore, this paper will focus on expanding the body of work that

puts this theory to the test. To do so, the elicited speech of 24 university-age, English-speaking

learners of Spanish is systematically examined for accuracy in the production of the seven target

phones described above, which represent the various applicable levels of the Hierarchy of

Difficulty: /r/, [β], /ɾ/, [ð], /f/, /l/, and /e/. By calculating the extent to which speakers with

different levels of proficiency produce these phones in a native- (henceforth, “accurate”) or non-

native-like (henceforth, “inaccurate”) manner, we make the assumption that those sounds which

are most often produced accurately are the easiest to acquire, while those that are more often

produced inaccurately are more difficult. We recognize of course that rate of production does not

automatically indicate relative ease of acquisition. However, using these cross-sectional study

data we believe that we can approximate the general route of acquisition of the sounds in

question which, in turn, allows us to hypothesize which sounds require longer to be acquired and

as such can be assumed to be more problematic for L2 learners.

Page 6: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

The following sections of this paper explore what previous work there is on the subject,

and will then detail the methodology of the study. Results are then presented for each of the

sounds in question, as well as in terms of overall accuracy. Generally speaking, we will show

that the Hierarchy of Difficulty was able to predict apparent patterns of difficulty for only four of

the seven sounds, while the patterns found for three sounds were not predicted by the hierarchy.

Therefore, the results section explores both why these predictions were not possible using solely

the Hierarchy of Difficulty and how the hierarchy might be modified to more accurately model

the observed data of the participants. Namely, we will propose that Stockwell and Bowen’s

hierarchy is insufficient because it accounts only for the phonemic categorization of phones, and

that we must not only consider this categorization, but also the processes by which the various

phones are restructured and incorporated into the inventory of the L₂.

2. Previous Work

For several decades, the role of transfer has been heavily scrutinized as a major source of

interference in the field of second language acquisition (SLA) and learning. Indeed, Major

(2008) points out that early research on language transfer can even be traced back to the Müller’s

(1861) work on the role of transfer in pidgins and creoles. However, it was not until the middle

of the last century that the role of language transfer truly found its footing in the domain of SLA

theory and research. Weinrich (1953) laid the foundation for our modern understanding of the

various forms transfer can take, naming seven different types of sound transfer: (1) sound

substitution, (2) phonological processes, (3) underdifferentiation, (4) overdifferentiation, (5)

reinterpretation of distinctions, (6) phonotactic interference, and (7) prosodic interference. Table

3 provides definitions of and examples for these types of transfer-based interference.

Page 7: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

While Weinreich identified different kinds of interference in SLA, it was not until a few

years later that Robert Lado (1957) proposed his Contrastive Analysis Hypothesis (CAH), which

attempted to explain how these different kinds of interference played a role in SLA.

Process Definition Example

Sound substitution

Substitution of an L₂ sound

for one that is similar in the

L₁.

Use of English alveolar [d]

for Spanish dental [d̪]

Phonological process Interference caused by

differences between L₁ and

L₂ allophonic structure.

Use of English allophone [ɫ]

in syllable-final positions in

Spanish.

Underdifferentiation L₂ includes distinctions not

present in L₁.

English allophones [t] and [ɾ]

are separate phonemes in

Spanish.

Overdifferentiation L₁ includes distinctions not

present in L₂

English phonemes /d/ and /ð/

are allophones of the same

phoneme in Spanish.

Reinterpretation of

distinctions

“Reinterpret[ation] of

secondary or concomitant

features as primary or

distinctive features” (Major,

2008, pg. 67)

Interpretation by L₁ Spanish

speaker of English aspiration

([p] vs. [pʰ]) as a voicing

distinction (/p/ vs. /b/)

Page 8: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Phonotactic interference Use of L₁ syllable structure

norms in the L₂

L₁ Spanish production of

English ‘stop’ as /es.tap/

Prosodic interference Use of L₁ prosodic norms in

the L₂ (e.g., intonation, stress,

etc.)

L₁ English production of

Spanish /a.ni.ꞌmal/ as

/ꞌa.ni.mal/

Table 3: A Summary of Weinreich’s (1953) Types of Sound Transfer

The assertions Lado made paved the way for decades of research using the CAH.

Specifically, he claimed that transfer is the major problem that needs to be accounted for in SLA,

and that it can both predict and explain all errors committed by language learners. It should be

mentioned, too, that while Lado’s publication is considered the seminal work of the CAH, his

views were significantly influenced by the writing of Fries (1945), who had written a book for

language teachers presenting claims about how the first language can have influences on the

second language is acquired and how to modify curricula to account for this occurrence.

While Lado may be considered the father of the Contrastive Analysis Hypothesis,

Stockwell and Bowen (1965) are the linguists who fleshed it out and popularized it as a learning

theory (see also Stockwell, Bowen, & Martin, 1965). Their book, The Sounds of English and

Spanish, set forth hierarchies of difficulty that attempted to predict which sounds would be easy

to acquire and which sounds would be more difficult (for further explanation, see section 1:

Introduction)

Just five years after Stockwell and Bowen’s hierarchy of difficulty had been published,

Wardhaugh (1970) responded to the supposed predictive power of the CAH, claiming that it

“raises many difficulties in practice, so many in fact that one may be tempted to ask whether it is

Page 9: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

really possible to make contrastive analyses” at all (p. 2). Wardhaugh went on to address

Stockwell and Bowen’s hierarchy of difficulty, stating that it is “based more on their experience

and intuition than on an explicit theory for predicting difficulties” (p. 8). However, Wardhaugh

did not completely discredit the CAH, but rather proposed the creation of a strong version and a

weak version, with the former asserting that transfer allows for the prediction of errors and the

latter merely admitting that transfer facilitates explanation of L₂ production errors. While he

made it no secret that he lent little to no credence to the strong version, Wardhaugh affirms that

“in its weak version [… the CAH] has proved to be helpful and undoubtedly will continue to be

so as linguistic theory develops” (p. 12). Indeed, as we will see in the present study, CAH

cannot fully predict the difficulty of acquisition for all sounds (see especially /l/ and /e/ in Figure

3), but it makes it easier to explain difficulty in production from a post hoc standpoint. Major

(2008) agrees, noting that “if prediction is defined as an absolute occurrence or nonoccurrence of

phenomena for all individuals, then [Contrastive Analysis (CA) …] can be easily falsified.

However, if prediction is defined in the probabilistic sense, then CA is alive and well today” (p.

65).

If, then, language transfer and, in turn, the CAH, are successful in explaining errors, but

not in predicting errors that will occur, even over-predicting errors that do not occur, then what is

the source of these discrepancies and unexpected errors? Selinker (1972) made the novel

assertion that perhaps not all errors are due to transfer and interference from the L₁. His work on

Interlanguage systems and fossilization led him to propose that universals and other

“developmental” factors similar to those that cause errors in first language learning in infancy

can also play significant roles in the acquisition of a second language. For example, free

variation and systematic variation show incomplete acquisition of rules, sociolinguistic factors

Page 10: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

typically stemming from a struggle to form identity in the L₂, psychological factors like attention

and memory, and affective factors like stress and anxiety.

Largely due to Selinker’s findings, the linguistic research community veered away from

transfer-based theories and shied away from making claims that interference played a strong role

in L₂ learning. During this period, it was unpopular for linguists to claim that transfer played a

significant role in the SLA and Second Language Learning (SLL) processes. However, a decade

later, Andersen (1983) dared to return transfer to the limelight with his “Transfer to Somewhere”

Hypothesis. This hypothesis “specified [the] conditions under which transfer can and cannot

operate” (Major, 2008, p. 66): namely, that the input from the L₂ that the learner receives must

have a “corresponding structure” that is (mis-)generalizable to patterns of the L₁. For example,

in the present study we see that the English pattern of /l/ [ɫ] in syllable-final contexts is mis-

generalizable for L₂ Spanish learners, who might produce [tal] as [taɫ] as a result of this mis-

generalization. Kellerman (1995) responded later with his “Transfer to Nowhere” Hypothesis,

asserting that SLA transfer is not dependent on similarity to occur, but rather that it is

unpredictable and can happen between any set of structures between two languages.

Around the same time, linguists working from a generativist Universal Grammar

approach were attempting to account for the role of transfer by proposing a full gamut of

theories, including the Fundamental Difference Hypothesis (Bley-Vroman, 1989, 1990), the Full

Transfer Full Access Hypothesis (Schwartz & Sprouse 1994, 1996 and Sprouse & Schwartz

1998), the Valueless Features Hypothesis (Eubank 1993/1994, 1994, 1996), and the Minimal

Trees Hypothesis (Vainikka & Young-Scholten 1994, 1996a, 1996b). These theories attempt to

explain not only what degree and type of access second language learners have to Universal

Grammar, but also the extent to which transfer affects SLA. These theories are typically viewed

Page 11: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

by modern-day linguists as more credible and more legitimate than the CAH, due to the fact that

they are based on a psychological theory of learning and can account for more factors than

simply just transfer, such as memory, L₁ acquisition, and the differentiation of linguistic

competence from linguistic performance.

Although the CAH may no longer be in its heyday (as it was during the times of Anan,

1981; Palik, 1977; Redard, 1973; Soudak, 1977; and Tomaszczyk, 1980; as listed by Major,

2008, p. 68), it “continues to be invoked in much second language speech acquisition research

today” (Zampini, 2008, p. 219). Major’s (p. 68) central premise that many L₂ errors result from

breakdowns in transfer from the L₁ is still heavily influencing research in linguistics today in

almost every aspect of phonology, from individual phones to phonotactics, to prosodic features,

and even broader concepts, such as dialectology and connectionist theories. Thus, while the

world of SLA may have graduated from the belief that transfer is the only factor that influences

acquisition to one that acknowledges that language learning is a complex task that involves

variables like age of acquisition, markedness, sociolinguistic factors, and orthography, we must

still admit that transfer undeniably plays a role. We need not assume, though, that that role is as

straightforward or predictable as we once believed it to be.

While Stockwell and Bowen’s hierarchy difficulty is familiar to most linguists who have

studied the Contrastive Analysis Hypothesis, there has been little to no research that empirically

investigates the validity of its claims. The present study seeks to add to the body of research

about the role of transfer, interference, and linguistic theory, specific in the phonology domain,

by methodically examining each level of the hierarchy and by evaluating both its successes and

its shortcomings.

Page 12: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

3.Methodology

3.1 Research Questions

As explained above, the present study aims to explicitly test the claims of the Contrastive

Analysis Hypothesis through Stockwell and Bowen’s Hierarchy of Difficulty by examining the

degree to which they can accurately predict ease or difficulty of specific phones in L₂

phonological acquisition. To that end, the specific questions that guided this investigation are as

follows:

RQ1: How accurately does Stockwell & Bowen’s Heirarchy of Difficulty predict

patterns of acquisition of Spanish L2 sounds by L1 English speakers?

RQ2: How accurately does Stockwell & Bowen’s Heirarchy of Difficulty explain

patterns of acquisition of Spanish L2 sounds by L1 English speakers?

3.2 Participants

Data were originally collected from 119 participants at a large university in the

southeastern United States, all of which were enrolled in one of the Spanish courses selected for

study by the researcher: Beginning Spanish 1 (SPN1130, n = 17); a 5th semester bridge-the-gap

conversational course (SPN 2240, n = 36); a 7th semester composition and grammar course

(SPN3301, n = 38), and an advanced stylistics and writing course (SPN 4420, n = 28). At the

university where the study was carried out, the first-semester course is generally populated with

students who need to fill a language requirement, while the other courses are electives for

majors, minors and other students interested in continued language study.

Page 13: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

3.3 Tasks

All participants completed a written language background questionnaire (reproduced in

Appendix A) that elicited information regarding languages spoken, ethnic identity, degree of

exposure to Spanish (via courses, media, friends, etc.) and linguistic background. They then

moved on to the data collection portion of the experiment, in which they used a headset

microphone to record themselves reading a text aloud, two times through (see Appendix B). For

data analysis purposes (see Section 3.5 below), only the second recording of each speaker was

considered. By considering the first reading a practice reading, participants were able to

familiarize themselves with the text and thus minimize reading mistakes in the second. We thus

assume that the second reading is a more accurate demonstration their linguistic competence

with respect to the L2 sound system.

The reading text was created by the researcher and was designed to sound like a natural

reading passage and include multiple tokens of each of the seven phones targeted for study in

this experiment: /r/, [β], /ɾ/, [ð], /f/, /l/, and /e/. While the text included dozens of tokens for

many of the phones, ten examples of each phone were selected within the recorded reading text

(as indicated by brackets in Appendix B) and appeared in a variety of contexts, including onset,

coda, and clusters of phones (whenever possible). In this way, we avoided confounding

variables related to specific phonological context and instead are able to get a glimpse of the

overall accuracy of each sound.

3.4 Data Selection

Once the corpus of data had been gathered, the researcher used the language background

questionnaires to select a sub-group of participants who fit the profile for participation in the

Page 14: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

experiment, namely, non-heritage second language learners without experience in other

languages. Therefore, for our purposes we removed speakers who reported use Spanish at home,

who spoke a language besides English and Spanish (either as an L₁ or an L₂), who had ever

received speech therapy, or who had studied or lived abroad for an extended period of time.

After the group of data had been refined to the smaller group of eligible speakers, which

represented only about half of the original number of recordings, the researcher randomly chose

six recordings from each level of experience: three from female speakers and three from male

speakers, for a total of twenty-four recordings. These selected recordings were then assigned

code numbers by a third party to disguise the level of the participants to the judges.

3.5 Data Analysis

The researcher and another judge evaluated the accuracy of production of the phones,

judging each as either categorically “native” or “non-native.” Both judges had previously

completed a Spanish phonetics course with high grades and have received training in the field of

Spanish linguistics and phonology. Each judge listened to all 1,680 tokens, using a mixture of

Praat analysis and aural perception and made comments on the accuracy of production of each

token before meeting to discuss the results. When the judges agreed on the accuracy of the

production of a phone, no further investigation was deemed necessary. However, where the

judges disagreed, the lead researcher used Praat acoustic analysis software to make a final

judgment on the disputed token. The participants were, therefore, awarded an accuracy score for

each phone under investigation. One point was awarded for each correctly produced token in the

passage, resulting in scores ranging from 0 (no correct utterances) to 10 (all target tokens

produced correctly) for each phone. The results of the analysis are discussed below.

Page 15: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Participant /r/ [β] /ɾ/ [ð] /f/ /l/ /e/ total

gender

group

avera

ge

group

average S.D.

1131 F1 0 2 5 0 10 6 4 27

34.66

30.5 10.48

1131 F2 0 1 7 1 10 7 1 27

1131 F3 7 6 7 7 10 8 5 50

1131 M1 0 1 2 3 10 7 2 25

26.33 1131 M2 8 4 2 7 10 5 1 37

1131 M3 0 0 0 0 10 6 1 17

Total 1131 15 14 23 18 60 39 14 183

2240 F1 0 3 9 7 10 10 3 42

42.66

42.5 5.59

2240 F2 2 4 6 3 10 7 2 34

2240 F3 5 7 8 8 10 10 4 52

2240 M1 0 4 9 5 10 9 2 39

42.33 2240 M2 9 4 8 1 10 10 4 46

2240 M3 4 2 10 3 10 8 5 42

Total 2240 20 24 50 27 60 54 20 255

3301 F1 0 7 1 1 10 7 3 29

35.33

43.83 12.09

3301 F2 0 9 9 8 10 10 6 52

3301 F3 0 0 2 3 10 8 2 25

3301 M1 0 7 7 9 10 9 7 49

52.33 3301 M2 5 7 10 8 10 10 3 53

3301 M3 5 6 8 10 9 8 7 55

Total 3301 10 36 37 39 59 52 28 263

4420 F1 7 5 5 9 10 9 6 51

40.66

42.5 8.42

4420 F2 2 2 10 3 10 6 3 36

4420 F3 6 0 5 1 10 10 3 35

4420 M1 0 3 6 3 10 9 3 34 44.33

4420 M2 7 2 6 4 10 9 5 43

Page 16: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Table 4: Accuracy Results for Each Participant

4. Results and Discussion

Table 4 below shows the results for each participant considered in data analysis. It shows

total number of correct tokens produced (out of ten) under each phone, the total number of

correct phones produced per speaker, the average of each experience level grouping, the average

of each gender within those level groups, and the standard deviation for each group.

First, overall accuracy was calculated by averaging the total number of correct tokens

produced by the members of each group, resulting in an average for each group. The total

possible for this average is 70 (10 tokens x 7 phones). A series of two-tailed t-tests was then run

to determine statistically significant differences between experience levels for the overall

accuracy of each group.

Results indicate a statistically significant difference between the group average of SPN

1131 and SPN 2240 (p = .033). The differences between the groups SPN 1131 and SPN 3301 (p

= .069) and between SPN 1131 and 4420 (p = .054) are approaching statistical significance. No

other group differences proved to be statistically significant (SPN 2240/SPN 3301 p = .812; SPN

2240/SPN 4420 p = 1.000; SPN 3301/SPN 4420 p = .829). We must then conclude, based on the

experiment data, that second-semester learners performed significantly less accurately than their

more advanced counterparts, but we cannot at this time make definitive claims about differences

between these more experienced groups.

4420 M3 5 8 10 9 10 9 5 56

Total 4420 27 20 42 29 60 52 25 255

Page 17: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Similar two-tailed t-tests were run for the differences between each group for each of the

seven phones individually. Figure 1 shows in graphical form the differences among experience

groups for each phone.

Figure 1: Observed Accuracy of Each Group for Each Sound

The only statistically significant differences between groups are between SPN 1131 and

SPN 2240 for the phone /ɾ/ (p = .007) and between SPN 1131 and all other groups for the phone

/l/ (p = .004, .008, and .012, respectively). There are differences approaching statistical

difference between SPN 1131 and SPN 3301 for the phone [β] (p = .057), between SPN 1131

and SPN 4420 for the phone /ɾ/ (p = .066), between SPN 1131 and SPN 3301 for the phone /e/ (p

= .073), and between SPN 1131 and SPN 4420 for the phone /e/ (p = .068). These statistical

results suggest that there is some degree of learning occurring for the phones [β] /ɾ/, /l/, and /e/,

but that, overall, there are no remarkable patterns of growth over time among the groups of

0

1

2

3

4

5

6

7

8

9

10

/r/ /β/ /ɾ/ /ð/ /f/ /l/ /e/

Acc

ura

cy S

core

Sound Segment

Actual Data for Accuracy by Group

SPN 1131 SPN 2240 SPN 3301 SPN 4420

Page 18: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

students. We will, therefore, now turn to discussion of the data as a whole, without respect to

differences between groups.

Figure 2 shows accuracy predictions for each phone, as derived from Stockwell and

Bowen’s Hierarchy of Difficulty. The x-axis gives the seven observed phones, while the y-axis

gives the “total tokens correct” score, calculated by adding the scores of each participant together

for the phone listed on the x-axis. Figure 3 depicts the observed data from the present study,

with important differences from the predictions made in Figure 2 highlighted in yellow.

Figure 2: Predicted Accuracy for Each Sound (based on Stockwell & Bowen’s Hierarchy of

Difficulty)

0

48

96

144

192

240

/r/ /β/ /ɾ/ /ð/ /f/ /l/ /e/

Tota

l To

ken

s C

orr

ect

Sound Segments

C.A. Model for Whole-Group Accuracy

Page 19: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Figure 3: Observed Accuracy for Each Sound (based on collected data)

The most significant problems with the predictions made by the hierarchy are centered on

the phonemes /l/ and /e/. Where the CA hierarchy of difficulty predicts that these sounds (along

with /f/) should be the easiest to acquire for L₁ English/L₂ Spanish learners and should all be

produced with a very high degree of accuracy due to their statuses as obligatory in both L₁ and

L₂. However, what we observe is that all three are produced at radically different rates of

accuracy, with /f/ reaching ceiling (100%), /l/ hovering around 82%, and /e/ just breaking 36%.

To return to RQ1, this question cannot be supported by the present findings, with the

Hierarchy of Difficulty achieving a predictive accuracy of only 60%. However, the results for

RQ2 are more interesting. While it probably should not be claimed that the Hierarchy of

Difficulty can per se explain the data observed in this project, we find that the underlying

principle of the CAH (i.e.: that difficulties arise where the first and second languages differ) still

has explanatory power.

While Stockwell and Bowen’s Hierarchy of Difficulty and the CAH had no way of

predicting the discrepancies evidenced in the data, it is easy enough make a contrastive analysis

0

48

96

144

192

240

/r/ /β/ /ɾ/ /ð/ /f/ /l/ /e/

Tota

l To

ken

s C

orr

ect

Sound Segments

Actual Data for Whole-Group Accuracy

Page 20: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

in retrospect and hypothesize that these differences come from underlying complexities in

allophonic structure of the L₁ (English). The phoneme /f/ in English has just one allophone, [f],

while both /l/ and /e/ have two non-co-articulatory allophones ([l]/[ɫ] and [ej]/[ə]). Inclusion of

the allophones of /l/ that are governed by patterns of co-articulation (such as [ l̟ ] and [ l̪ ]) was

deemed irrelevant by the researcher because the phonotactic rule of co-articulation is common to

both languages and is therefore only a complicating factor to performing contrastive analyses.

We should note, too, that in Standard American English [e] never occurs by itself, but rather co-

occurs with the off-glide [j] to form a diphthong.

But what, then, are we to say about [ð], where no explanation, even in retrospect, is

offered by the Contrastive Analysis Hypothesis? This interference could come from any number

of likely sources, among them complications of orthography, frequency, and the “noticing” of

key differences. The confluence of the English phonemes /d/ and /ð/ into one singular Spanish

phoneme /d/ is made even more difficult for learners by the fact that the English graphemes ‘d’

and ‘th’ must also be collapsed into the grapheme ‘d.’ With respect to frequency, it seems

logical that L₁ English/L₂ Spanish learners would have more problems with /d/ than with /ɾ/,

given that /ɾ/ will virtually always be realized as [ɾ] (and therefore almost never as English [ɹ]),

but /d/ shows variation, at times appearing as [d] and at times as [ð]. Finally, “noticing” (e.g.

Schmidt, 1990), a vital component of Flege’s (1992, 1995) Speech Learning Model, implies that

a failure to notice key differences can result in an “equivalence classification” and could prove to

entail strong negative effects on learners’ control of a phoneme like /d/. That is, “equivalence

classification” would claim that if a learner fails to notice the subtle difference between [d̪] and

[ð], they will consequently not be able to produce an accurate distinction between the two phones

in their own L₂ speech.

Page 21: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

These results, I believe, could better be explained by a different kind of hierarchy. One

that takes into account not only a sounds status as “optional,” “obligatory,” or “null” in the L₁

and L₂, but also the complexities of the deeper phonemic and allophonic structures of the target

sounds. Table 5 shows, describes, and exemplifies these five proposed types of change. The key

advantage to a hierarchy like the one proposed here is that it allows for consideration of the

processes utilized by learners in restructuring their sound inventories and the phonemic

structures therein. This key addition to the theory is most easily observed in the re-ordering of

the levels of difficulty for the sounds /l/, /ɾ/, and [ð]. Rather than simply considering the fact that

/l/ and /f/ do not change categories (phoneme phoneme), and that /ɾ/ and [ð] do change

categories (respectively: allophone phoneme; phoneme allophone), this new hierarchy

accounts for the fact that the entire phonemic structure of /l/ must be altered (i.e.: the allophone

[ɫ] must be eliminated from the possible variants) while the structure of the /f/ phoneme is

preserved. Moreover, it includes tiers that consider the fact that some sounds must be separated

from their L₁ phonemic structures (like /ɾ/ from English /t/ and /d/) and sounds that are separate

in the L₁, but must be “melded” or fused together into a single L₂ phoneme. This process of

phonemic restructuring, we believe, is one that carries with it great importance and should be

thoroughly accounted for in theories that attempt to predict difficulty of individual structures in

the field of SLA.

Page 22: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Type of Change

(from easiest to hardest)

Description Example(s)

No change Phonemic and allophonic

structures are the same in L₁ and L₂.

/f/

Collapsing of differences Allophones are removed from

the L₁ phonemic structure.

/l/

(collapsing of [l] and [ɫ])

Change of category An allophone in the L₁ becomes a separate phoneme

in the L₂

/ɾ/

(separation from phonemes /t/

and /d/)

Melding of categories Two L₁ phonemes are

combined to form one

singular L₂ phoneme.

[ð]

(melding of /d/ and /ð/)

Creation of category A category for a new sound is

created in the L₂ [β], /r/, /e/

(/e/ without the mandatory

off-glide)

Table 5: Proposed Revisions to the “Hierarchy of Difficulty”

5. Conclusions

We have seen, therefore, that the data from this experiment seem to support the idea that

the Contrastive Analysis Hypothesis and the Hierarchy of Difficulty cannot accurately predict

which sounds will be easy and which sounds will be hard for L₂ learners of a given language.

Even though it did serve as an accurate predictor for four of the seven examined phones, we have

seen that there are significant holes in the theory, as far as the extent to which it considers (or,

perhaps more accurately does not consider) all of the relevant factors in the acquisition of an L₂

phonetic inventory.

In response, we have proposed a different hierarchy, one that is informed by the various

types of restructuring of phonemic categories that can take place during the acquisition of a

second language. In the future, more work on the topic will be necessary to make more

Page 23: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

definitive claims about the precise effects of transfer and phonemic inventory restructuring on

difficulty of acquisition in a second language. First, more phones should be examined under

each level of the hierarchies, both old and new, to ensure the validity of the claims made above.

Secondly, it would be wise to examine data across a variety of registers and styles of speech and

to compare these data with the data of native speakers. In fact, we hope in the future to continue

this project by examining not only the patterns of native speaker speech on the reading task, but

also the patterns of students who participated in a Spanish phonology course at the same

institution, allowing for the analysis of explicit instruction on difficulty in L₂ phonology.

Finally, in order to make these findings even more generalizable, it would be prudent to test the

claims of the hierarchies and the roles of interference across a variety of first and second

languages.

It is hoped that the new hierarchy proposed here can prove to be more effective in its

ability not only to explain errors in L₂ phonology that stem from interference from the L₁, but

also to predict phonic learning difficulty for a range of first and second languages.

Page 24: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Works Cited

Anan, F. 1981. A study of intonation reference (translation from Japanese). The Bulletin, The

Phonetic Society of Japan 167: 9-13

Anderson, R. 1983. Transfer to somewhere. In S. M. Gass & L. Selinker (eds.), Language

transfer in language learning, Rowely, MA: Newbury House, 177-201.

Bley-Vroman, R. 1989. What is the logical problem of foreign language learning? In S. Gass &

J. Schachter (eds.), Linguistic perspectives on second language acquisition, Cambridge:

Cambridge University Press, 4-68.

Bley-Vroman, R. 1990. The logical problem of foreign language learning. Linguistic Analysis

20: 3-49

Eubank, L. 1993/1994. On the transfer of parametric values in L2 development. Language

Acquisition 3: 183-208.

Eubank, L. 1994. Optionality and the initial state in L2 development. In T. Hoekstra & B. D.

Schwartz (eds.), Language acquisition studies in generative grammar: Papers in honor

of Kenneth Wexler from the 1991 GLOW Workshops, Amsterdam: John Benjamins, 369-

388.

Eubank, L. 1996. Negation in early German-English interlanguage: More valueless features in

the L2 initial state. Second Language Research 12: 73-106.

Flege, J. E. 1992. The intelligibility of English vowels spoken by British and Dutch

talkers. In R. D. Kent (ed.), Intelligibility in speech disorders: Theory, measurement, and

management (Vol. 1), Philadelphia: John Benjamins, 157-232.

Flege, J. E. 1995. Second language speech learning: Theory, findings, and problems. In W.

Page 25: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Strange (ed.), Speech perception and linguistic experience: Issues in cross-linguistic

research, Timonium, MD: York Press 233-277.

Fries, C. C. 1945. Teaching and learning English as a foreign language. Ann Arbor: University

of Michigan Press.

Kellerman, E. 1995. Crosslinguistic influence: Transfer to nowhere? Annual Review of Applied

Linguistics 15: 125-150.

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

Major, R. C. 2008. Transfer in Second Language Phonology. In J. G. H. Edwards and M. L.

Zampini (eds), Phonology and Second Language Acquisition, Philadelphia: John

Benjamins Publishing Company 63-94.

Müller, M. 1861/1965. Lectures on the science of language. New Delhi: Munshi Ram Manohar

Lal.

Palik, Y.-J. 1977. Pronunciation difficulties of Korean students learning English as a second

language. Dhak Yonku 13: 177-189.

Redard, F. 1973. Contrastive phonemics: Notes and exercises. Neuchâtel, Switzerland: Centre de

Linguistique, Universite de Neuchâtel.

Selinker, L. 1972. Interlanguage. International Review of Applied Linguistics 10: 209-231.

Schmidt, R. 1990. The role of consciousness in second language learning. Applied Linguistics,

11, 129-158.

Schwartz B. D. & Sprouse, R. A. 1994. Word order and nominative case in nonnative language

acquisition: A longitudinal study of (L1 Turkish) German interlanguage. In T. Hoekstra

& B. D. Schwartz (eds.), Language acquisition studies in generative grammar: Papers in

Page 26: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

honor of Kenneth Wexler from the 1991 GLOW Workshops, Amsterdam: John

Benjamins, 317-368.

Schwartz, B. D. & Sprouse, R. A. 1996. L2 cognitive states and the full transfer/full access

model. Second Language Research 12: 40-72.

Soudak, L. I. 1977.Error analysis and the system of English consonants. ELT Journal 31: 125-

130.

Sprouse, R. A. & Schwartz, Schwartz, B. D. 1998. In defense of full transfer in German-English

and French-English interlanguage: Comparative L2 acquisition research. Proceedings of

the Annual Boston University Conference on Language Development 22: 726-736.

Stockwell, R. & Bowin, J. 1965. The Sounds of English and Spanish. Chicago: University of

Chicago Press.

Stockwell, R., Bowin, J., & Martin J. W. 1965. The grammatical structures of English and

Spanish. Chicago: University of Chicago Press.

Tomaszczyk, J. 1980. On accented speech: The Polish of Polish Americans. Studia Aglica

Posnaniensia, 12: 121-137.

Vainikka, A. & Young-Scholten, M. 1994. Direct Access to X’-theory: Evidence from Korean

and Turkish adults learning German. In T. Hoekstra & B. D. Schwartz (eds.), Language

acquisition studies in generative grammar: Papers in honor of Kenneth Wexler from the

1991 GLOW Workshops, Amsterdam: John Benjamins, 265-316.

Vainikka, A. & Young-Scholten, M. 1996a. The early stages of adult L2 syntax: Additional

evidence from Romance speakers. Second Language Research 12: 140-176.

Vainikka, A. & Young-Scholten, M. 1996b. Gradual development of L2 phrase structure. Second

Language Research 12: 7-39

Page 27: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Wardhaugh, R. 1970. The contrastive analysis hypothesis. TESOL Quarterly 4: 124-130.

Weinrich, U. 1953. Languages in contact. New York: Linguistic Circle of New York.

Zampini, M. L. 2008. L2 speech production research. In J. G. H. Edwards and M. L.

Zampini (eds), Phonology and Second Language Acquisition, Philadelphia: John

Benjamins Publishing Company 219-249.

Page 28: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Appendix A: Background Questionnare

Last name: ____________________ First Name: ____________________ Middle Initial: ____

Gender: M F Age: ________ Place of Birth: _____________________________

For each of the following, please indicate the languages you speak, the age you began

learning it, what percent of your total language use (spoken and in print) is conducted in

that language, and whether it is spoken in your home.

Dominant Language: _______________ Age: _____ Usage: ______% Home: Y / N

Second Language: _______________ Age: _____ Usage: ______% Home: Y / N

Third Language: _______________ Age: _____ Usage: ______% Home: Y / N

Other languages (please indicate degree of exposure): __________________________________

______________________________________________________________________________

Please indicate any cultures or ethnicities with which you identify, ranking the strength of

that identification on a scale from 1 to 10 (one being low and ten being high) [eg: Hispanic

(10), Caucasian (6), Indigenous American (3)]

______________________________________________________________________________

______________________________________________________________________________

Have you ever lived in an environment where Spanish is the majority language (including

study abroad)? If yes, please provide details of location(s), duration of stay, language of

school(s) attended, language of host family, etc.

______________________________________________________________________________

______________________________________________________________________________

Page 29: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

In which of the following were you exposed to the Spanish langauge: (check all that apply)

Elementary school (ages 5-10)

Middle school (ages 11-14)

High school (ages 14-18)

College (ages 18-22)

Please list all college-level Spanish classes previously completed (at UF or other institution

of higher education):

______________________________________________________________________________

______________________________________________________________________________

On a scale from 1 to 10, please indicate your self-perceived level of proficiency in the

Spanish language in the following areas:

Reading: _____

Writing: _____

Speaking: _____

Comprehension of spoken language: _____

On a scale from 1 to 10, please indicate how much the following factors have contributed to

you learning Spanish:

Interacting with friends: _____

Interacting with family: _____

Reading: _____

Language tapes/self-instruction: _____

Watching TV: _____

Listening to the radio: _____

Language labs: _____

Language classrooms: _____

On a scale from 1 to 10, please indicate how much you are currently exposed to Spanish in

the following contexts:

Interacting with friends: _____

Interacting with family: _____

Reading: _____

Language tapes/self-instruction: _____

Watching TV: _____

Listening to the radio: _____

Language labs: _____

Language classrooms: ____

Page 30: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

On a scale from 1 to 10, how would you evaluate your pronunciation of the Spanish

language? (Please provide a brief explanation)

______________________________________________________________________________

______________________________________________________________________________

Are you currently or have you previously been enrolled in any linguistics classes at UF? (If

yes, please list them)

______________________________________________________________________________

______________________________________________________________________________

Have you ever received direct training in Spanish phonology, phonetics, or pronunciation?

(If yes, please explain)

______________________________________________________________________________

______________________________________________________________________________

Have you ever undergone speech therapy or any pronunciation adjustment? (If yes, please

explain)

______________________________________________________________________________

______________________________________________________________________________

Have you ever had a vision problem, hearing impairment, language disability, or learning

disability? If yes, please explain (including any corrections):

______________________________________________________________________________

______________________________________________________________________________

Are there any other factors that could contribute to your use or command of Spanish?

______________________________________________________________________________

Page 31: Modifying Stockwell & Bowen s Hierarchy of Difficulty ......1. Introduction The Hierarchy of Difficulty proposed by Stockwell and Bowen (1965) and based on the theory of Contrastive

Appendix B: Text Used to Collect Oral Data

El sol brillaba [f]uertemente, con una [l]uz blanco que me pica[b]a los ojos. Me levanté

muy d[e]spacio, notando la escena ente[ɾ]a. En la tie[r]a (tierra) dura, ha[b]ía poca hie[ɾ]ba y

na[ð]a de [f]lor[e]s, sino mi[l]es de [r]aíces que [f]orma[b]an una r[e]d con[f]usa, como una

telaraña. Más a[r]i[b]a (arriba), vi arbustos ve[ɾ]des y los troncos ma[r]ones (marones) de

árboles [e]normes. Miré al cielo y me [ð]i cuenta de que llovía. Ent[ɾ]e las nub[e]s oscuras,

destelló un [r]e[l]ámpago y escuché el rui[ð]o del trueno. A mi la[ð]o derecho, había un sendero

estrecho que desapareció en el bosque. Después de unos mom[e]ntos, decidí segui[ɾ][l]o.

Cuando yo había camina[ð]o po[ɾ] un rato, yo vi un [r]ío y [b]ajé [f]ác[i]lmente del

sendero para llegar a la orilla. De la orilla, pu[ð][e] ver todo lo que escondía los altos árboles del

bosque. C[e]rca del horizonte había un volcán, una to[r]e (torre) sa[l]vaj[e] de pie[ð][ɾ]a y

[f]uego. Entré en el agua para [b]añarme un poco, y me sorprendí que no esta[b]a [f]río. De

repente, escuché un susurro en los arbustos y los pája[ɾ]os se [f]ueron vo[l]ando de las [r]amas

allá encima. Yo permanecí inmóvil. El animal emergió lentamente de la vegetación. Cuando

po[ð]ía ver la mitad del anima[l], me di cuenta de que tenía la [f]orma de muje[ɾ]. Era

[f][l]aquísima y me parecía completamente a[b]u[r]ida (aburrida). Me sentía muy [ð]é[b]il

mientras ella me aproximaba. El nervioso sentido de catástrofe me llenó. La ninfa abrió [l]a

[b]oca y habló con su voz dulce: «No p[e]rmito que na[ð]ie se bañe en mi río.» Empezó a [r]eír

y yo g[ɾ]itaba con horror. Bruscamente, me desperté seguro en mi cama.