The Perception Of Word Boundaries

Published Date: 02 Nov 2017

Individual Essay

"The Perception of Word boundaries in a Second Language"

EVELYN ALTENBERG

PART ONE

This essay is a review of Evelyn Altenberg’s "The Perception of Word Boundaries in a Second Language" (2005).

There has been limited exploration of speech segmentation by second language learners. Lamminmaki (1979) looked into the perception of English Juncture by younger Finnish learners of English and discovered the task he had set for them was difficult (However, may be unreliable due to lack of English Natives in the control group and no statistical analysis of data). Other background information into the realm of Word Boundaries would include studies from Yavas (1998) who suggested that with all factors equal, and in terms of perceiving; something unexpected and marked will stand out more than an unmarked-expected segment. Thus, something that is marked will be more perceptionally visible and smooth. Speakers should be more successful at segmentation on a juncture that is more marked. However, perception does not rely on the marked-ness alone. It is unlikely listeners rely on just one cue when segmenting any speech. Speakers use their lexical (both syntactic and morphological) and segment speech after recognizing a familiar word. For example, Altenberg suggests "by segmenting speech after English verbs with –ing endings, speakers use their acoustic-phonetic knowledge, segmenting syllables before an aspirated voiceless stop". It would appear that some of these phonetic cues are more important than others within speech segmentation. Nakatani and Dukes (1977) suggested the "strongest cues for juncture are glottal stops, laryngealization, aspiration on voiceless stops and /l/ /r/ allophones." Obviously not all languages use these same phonetic cues which then becomes problematic for linguists. Cutler (1986) suggests all speakers of all languages have different segmentation strategies "depending on the languages rhythmic characteristics." In Altenbergs study, she decides to compare English Speakers with Spanish speakers; however Machpherson and Goldstein (1975-2001) cement the idea that Spanish doesn’t have aspirated voiceless stops like the English language. Along with not having the same phonetic cues, it’s often reported that Language learners have difficulties with speech segmentation, suggesting the second language is spoken too quickly. McAllister (1996) claims that: "Highly proficient non-native speakers have a globally perceptual foreign accent".

In order to research these suggestions, Evelyn proposed the following statements as a hypothesis. Firstly, the aspiration stimuli will be more difficult for L2 learners to segment than the glottal stop stimuli. Secondly, segmentation will be best for the double cue stimuli. Thirdly, if clusters are more perceptually smooth as they increase in marked ness; segmentation should be better on the second group than on the first, and better on the third group than the second. She hoped to address other questions within her study such as; will the learners be able to acquire second language acoustic phonetic cues? Are some stimuli easier for learners to segment and what factors account for this? How do language learners compare to native speakers in ability to use the acoustic phonetic cues to segment speech into words? Is there evidence of transfer at the allophonic levels and of marked-ness in L2 segmentation?

The following is a summary of the Methods and Participants used within Evelyn’s study. The participants were both Native and Non Native speakers. 19 Native English speakers, whom were mostly female, had previously studied languages within school. They ranged in age from 18-22. Non Native Speakers came from community colleges in the metropolitan New York area; they had been placed into either intermediate or advanced speakers of English based on previous classifications from their teachers. 42 Pairs of phrases were recorded in a quiet room using a professional analogue tape recorder, used by an adult male, university educated native speaker of New York English. There were 3 groups of stimuli recorded (aspiration, glottal stop and double cue), and subjects heard each phrase only once, with 6 seconds between each stimuli. They chose their answers from an answer sheet. Thus, the independent variable in this study is both the stimuli and phrases (What is varied during the experiment) the dependant variable is the participants response to the stimuli (What is measured) and the controlled variables are those held constant, for example: that everyone heard the clips in the same place at the same volume and pace. Evelyn concluded that Spanish listeners perceived the word boundaries less accurately than the English speakers, and their performance on the aspiration pairs were significantly worse than glottal stop pairs. They were also unable to use the allophonic cues available. No differences were seen in the performance of Native English speakers. L1 transfer and marked-ness appears responsible for ease and difficulty for use of L2 word segmentation cues.

Some positives of this study, and it’s contribution to the world of Second Language Acquisition would be that the delivery of the test itself was flawless- testing participants in a quiet room with fixed intervals to respond on answer sheets, with the volume of the tape recorder so all participants could here. The use of natural stimuli allowing for a test that’s closer to real world processing. The natural speech used in the recordings gave natural results also. As for SLA: other researchers are currently using this experimental paradigm to pursue future investigations

Some criticisms made of the study are that clearly cross language research into word segmentation and allophonic cues is very limited, thus there is not a lot of previous information to compare with. There is much more research needed to observe similar difficulties to listeners of other languages and then exploring their factors in processing allophonic cues. Evelyn herself attempted to minimize lexical effects when constructing the stimuli, but it may not be possible to ever eliminate the effects of lexical bias from word segmentation studies. By using natural stimuli rather than synthesized, makes it more difficult to separate factors such as transfer and marked-ness which play a role in participant’s perceptions. A lot of criticism of in this study has been focused around the participants: Native speakers of English were used as a pre-test, selecting the phrases they found most natural which puts the Spanish participants at a disadvantage. They also considered themselves to be Native speakers, which may not be accurate as there was no proof. The Spanish speakers were also found to actually be Dominicans, Mexicans and Panamanians. All participants placements into intermediate and advanced speakers were based on their performance in written English exams, no phonological criteria played a role. Finally, four participants omitted at least one question in the questionnaire they were given, which gives inaccurate results.

PART TWO

INTRODUCTION

Altenberg achieved great results and useful findings within the field of Second Language acquisition. What she achieved is now being replicated across the world with many different languages. However, it was completed in 2005, which is now 8 years outdated. I wish to explore a language more similar to English. I think this will be interesting to compare alongside with Evelyn’s research into Spanish speakers. The language I have chosen is Dutch. English and Dutch are both Germanic languages.

English and Dutch share very similar words for example:

English: Sleep/Ship, eat/that/out, make/book

Dutch: Slaap/schip, eet/date/uit/, maak/boek

[Livius Dutch History Webpage]

With Spanish speakers not being able to produce the allophonic cues properly due to lack of practice within their language, I decided to research what sort of sound clusters Dutch and English share.

"In English, the hard /g/ was replaced by /y/ this never reached Germany. Dutch was influenced, so the Dutch /g/ became a voiced affricate /y/ that is similar to the /ch/ in Swiss German. Today this is spelled /g/, similar to German, but the pronunciation is different. (This is why foreigners have much difficulty with the Dutch gutturals, and often call the language a nearly fatal throat disease." [Livius Dutch History Webpage]

My hypothesis would be that both English and Dutch speakers will perform equally well in the stimuli lexicon decision task, with Dutch speakers also performing better in comparison to Evelyn’s Spanish speakers.

Interestingly, whilst searching for background data I came across a Reverse study: "English and Dutch share very similar object attributes. To determine which attribute selection can be done in a language dependent way additional experiments with less similar languages are necessary." Koolen (2010). This would appear to be looking for a comparison study such as Spanish speakers!

OVERVIEW

The instructions given to both sets of participants must be clear and concise, and in simple terms. There must be no confusion as to whether this is an acoustic phonetic test or a language exam. As this will be quite a long test, there may be a need for filler items or distractors to combat priming, strategic responses and fatigue. I must also take into account uncontrolled variables: such as the time taken for words to be read and understood, Evelyn left 6 seconds between stimuli and participants suggested that worked well.

PARTICIPANTS AND METHODS

There will be 40 participants altogether. There will be 20 Native Speakers of English (Found at a British/International School in the Netherlands) their ages will range from 16-21. Preferably 10 will be female and 10 will be male. 20 participants will be Native Dutch speakers and learners of English, 16-21 years old, half male and half female. The test will take place in a school room or hall with good acoustics so that every participant has a fair chance of hearing the stimuli. There will be just 30 pairs of stimuli (Less than Altenberg) as it was claimed that 40 pairs of stimuli is too much and the participants are subjected to boredom and are less likely to give accurate results. The frequency of lexical decision is important, they must be stimuli that both groups will be familiar with, and must be plausible not abstract.

By selecting participants of School age (with the right consent) I feel it will be easier to decide through proper exam results their exact level of English. It will be optional to take part in the study, and there will be no money reward as in Evelyn’s study, I feel this will distract the students instead of encouraging them to take part. It would be unrealistic to test all members of a population which is why we must try to determine properties of these participants using a sample. However, within a school we must also remember variations that will affect results such as gender, age, education level, fatigue and mood. I plan to get as much data as I can, individuals vary so much, so collecting individual data: testing subjects on standard psychological tests, e.g. verbal fluency tests, memory tests, vocabulary tests, etc.

CONTROLS AND CONCLUSIONS

When Altenberg started her experiment, she could not be sure of what the results would be; this is because linguistic research involves a lot of experimentation. This experiment will be most similar to a natural observation, thus there will be minimal interference with the normal process of events. As it is also introspective, there will also be background pre judgements and my own knowledge of syntax and semantics. Introspective involves working with a lot of singular judgements, eliciting responses from participants and questionnaire work. I will make sure to record everything that happens, so that a statistical conclusion (how many participants scored correctly) can be met. "Linguistic research is like the sciences generally": Chomsky (2011) Thus an experiment design is so very important. By designing, we can try to decrease experiment bias, increase participant variation, balance chance variation and make sure it’s replicable for other linguists "Informal experiments are hard to replicate test because they are not usually recorded and are often heavily susceptible to item variation" (Sprouse & Almeida). One bias I am particularly aware of within this follow up study would be stimulus framing via linguistic authority, so "we’re good friends, I think you’d agree you can’t say this English word" "Regardless of how great the differences are between our groups on whatever dependent measure we are using, it is always possible that the results are due to chance fluctuations." Ray & Rivizza (1985)