A Preliminary ReportWAESOL 2013 October 20, 2012 Nancy Hiser, MA nhiser@EnglishTalkShop.com We need to better understand the issues that underlie non-native speakers of English pronunciation so we may more accurately assess and address them. To determine which issues a non-native speaker needs to address to improve English pronunciation, students are often evaluated. An assessment commonly includes asking students to read aloud. Reading aloud allows control of vocabulary, sentence structure, and speech targets. It can yield a comprehensive view of all of the phonemes (in all positions within words, in all possible phonetic contexts, or in a realistic frequency of occurrence), and a wide range of suprasegmentals (stress, intonation and linking). Another advantage of this method is that it is repeatable, so students can be reassessed at any future time to document progress. However, is reading aloud a valid measure of pronunciation? If native speakers read aloud, they don’t sound like they are speaking, do they? No, volume and inflection tend to equalize—that is, there seems to be an “evening out” of loudness and pitch changes so the result is more monotone. Rate may be slower or faster but the rhythm seems to be more even than in their spontaneous speech. Any differences in native speakers’ pronunciation between reading aloud and speaking are not so obvious simply because they tend not to have mispronunciations but certainly the presence of a regional accent in speaking would be expected in reading aloud. Is non-native speakers’ reading aloud similarly affected? One important difference to consider when analyzing non-native speakers’ reading aloud is, of course, that they may not be familiar with the vocabulary or sentence structure in the written sample. And, what about pronunciation? Do the mispronunciations noted in reading aloud reflect the same mispronunciations that are found in their speaking? Are the percentages of occurrence similar? When reading aloud is used to establish a baseline, targets are often modified by instructors after listening to these students talk. Priorities shift. This suggests reading aloud is not a fully adequate measure on which to base pronunciation training. A better understanding of the differences in reading aloud and spontaneous speech would be valuable. Experts acknowledge that perceiving or recognizing a spoken target (phonemes, pitch, stress, etc.) is often a problem for non-native speakers. “Perception precedes production” is a well-accepted expression although research results are more complicated. (Studies will be cited and summarized in the final paper.) How important is it to address this prior to or in conjunction with production training? Does this listening discrimination skill have predictive power?
PurposeThe purpose of this study was to measure and correlate these features of ESL speakers:
- The presence of plural markers in spontaneous (prompted) speech; the presence of plural markers in reading aloud; the accuracy in discriminating a plural vs. a singular word in a listening task.
- The accuracy of production of the voiced TH sound (“this”) in spontaneous (prompted) speech; the accuracy of production of the voiced TH sound (THH) in reading aloud; the accuracy in discriminating the voiced TH sound (THH) in words from words with the D or Z sounds in a listening task.
- The accuracy of production of the vowel sound EE (“eat”) in spontaneous (prompted) speech; the accuracy of production of the vowel sound EE in reading aloud; the accuracy in discriminating the vowel sound EE in words from words with the vowel sound I in a listening task.
SubjectsESL students from five college/university settings where they were interviewed. They were asked by campus instructors to voluntarily participate in a 20- to 30-minute interview but were not given any information about the purpose of the study. Each student was offered American Speechsounds pronunciation software in appreciation of their participation. For confidentiality, each subject was identified by the letters of the institution they attended along with their initials, such as PCC (for Portland Community College) HB (name initials). Subjects were eliminated from the study for self-reported hearing or speech difficulties. Information collected included: age, gender, native language, other languages studied, age at which the subject began to speak English, how spoken English was learned, how long the subject has lived in an English-speaking environment, and the amount and type of pronunciation training the subject has received. Personal information was kept private and not shared with anyone except the other investigators in this study.
StudyFour parts were completed for each subject:
- A brief interview – Exhibit A
- Recorded conversation – Exhibit B
- Recorded reading aloud – Exhibit C
- Listen and Choose exercises in American Speechsounds Professional Version – Exhibit D
ScoringRecordings were scored blindly. Subjects’ recordings of reading aloud was scored by one examiner; subjects’ spontaneous speaking was scored by two examiners independently. Both examiners are speech pathologists with 20+ years of experience working in accent improvement. In cases of differing results, recordings were reviewed to get consensus.
Results (See Graphs)
|Presence in speaking||46% (29.27)||0-90%|
|Presence in reading||94% (8.83)||70-100%|
|Accuracy in discriminating||88.3% (6.72)||68%-96%|
|Accuracy in speaking||55.6% (31.92)||0-100%|
|Accuracy in reading||63% (33.26)||0-100%|
|Accuracy of discriminating||70.4% (15.73)||52-90%|
|Accuracy in speaking||82% (12.4)||50-100%|
|Accuracy in reading||75% (15.73)||40-100%|
|Accuracy of discriminating||68.75% (13.95)||39-90%|
Reading Aloud Accuracy vs. Speaking Accuracy
- Plurals: All 20 subjects had higher reading aloud accuracy than speaking accuracy. Reading aloud accuracy was significantly better than speaking. (p=0.00; range:100-10%)
- THH: 13 subjects had equal to or higher reading aloud accuracy than speaking accuracy and the mean difference in accuracy was not significant. (p=0.31)
- EE: 9 subjects had equal to or higher reading aloud accuracy than speaking accuracy; however, the mean difference in accuracy was not significant. (p=0.11)
Listening Discrimination vs. Speaking Accuracy
- Plurals: 18 subjects had higher accuracy in discrimination of plurals than in speaking accuracy. Accuracy of discrimination was significantly better than speaking. (p=0.00)
- THH: 12 subjects had higher accuracy in discrimination of THH than speaking accuracy. However, the mean difference in accuracy was not significant. (p=0.07)
- EE: Only 5 subjects had higher accuracy in discrimination of plurals than in speaking accuracy. Accuracy in discrimination was significantly worse than speaking accuracy (p=0.00).
- Plurals: Accuracy in speaking was positively, though not significantly, correlated with accuracy in reading aloud (r=0.279, p=0.233). However, there was no relationship between accuracy in speaking and accuracy in auditory discrimination (r=0.009, p=0.969)
- THH: Accuracy in speaking was positively correlated with accuracy in reading aloud (r= 503; p=0.024). However, there was no relationship between accuracy in speaking and accuracy in auditory discrimination (r=0.006, p=0.782)
- EE: Accuracy in speaking was not related to accuracy in reading aloud or auditory discrimination (ps>0.05)
- Subjects with 100% pronunciation of plurals in speaking: none
- Subjects with 100% accurate pronunciation of THH in speaking: three. Their scores for pronunciation accuracy in reading aloud were 100, 90, 90; scores for listening were 57, 85, 80.
- Subjects with 100% accurate pronunciation of EE in speaking: one. This subject’s score for pronunciation accuracy in reading aloud was 70 and for listening was 72.Subjects with 100% pronunciation accuracy in reading aloud of plurals: twelve. All had pronunciation errors in speaking: scores ranged from 0 to 90% accurate.
- Subjects with 100% pronunciation accuracy in reading aloud of THH: two. One had 10% pronunciation accuracy in speaking; the other had 100% accuracy.
- Subjects with 100% pronunciation accuracy in reading aloud of EE: one. Score for accuracy in speaking was 90%.No subjects had 100% accuracy in any listening task.
CommentsWhile there were several statistically significant positive correlations between reading aloud and speaking measures, a larger sample size is needed to confirm/extend findings. Data from 17 additional subjects whom have been interviewed and recorded will be added to this study and reported at a later date. Additional testing for other speech targets (stress, pitch, etc.) is desirable. Testing of native speakers would yield important data. As any of these variables improve over time, it would be helpful to know what occurs in the other variables: Pronunciation, reading aloud, listening discrimination?
InvestigatorsElinora Jane Cater, M.A. (students from Lake Washington Technical College) Nancy Hiser, M.A. (students from Portland Community College) Anne Magnan-Park, Ph.D (students from Indiana University at South Bend) Catherine Moore, M.A. (students from California State University at Fullerton)
SponsorEnglish Talk Shop whose contribution was in the collaborative design of the study and in providing the American Speechsounds CD’s. No funding was provided.
SUBJECT INTERVIEW INFORMATION
ID:____________________________________ (school initials + subject initials)
Age: <20____ 20-25____ 25-30____ 30-35____ 35-40____ 40-50____50+___
Gender: F____ M ____
Other Languages Learned:_____________
Began Speaking English Age:_____
Learned to Speak English:
from books ____
from TV/Audio media ____
from an instructor:
-native (Amer. Eng.) instructor _____
living in an English speaking environment
other English-speaking country__________
audio DVD’s/CDs ____
Do you have any hearing or speech problems in your native language? ____ (If yes, discontinue protocol)
- Say to the subject: “Tell me some things you find in the campus bookstore.”-Subjects must use 10 regular plurals.-If they don’t, prompt them with questions such as “What else?” or “Tell me more.” Keep them talking until they have used 10 regular plural nouns. (This does not require their pronunciation of the plural ending because that is the information we are collecting.)Example of response: “The bookstore has book and supplies for classes.” (3). Tell me exactly what kind of supplies. “Well, notebooks, pencil, and pen.” (6) Yes, what else? “There is food, too.” What kind of food? “Just snack and sodas.” (8) Who shops in the bookstore? “Mostly students and sometimes teacher.” (10)
- Say to the subject: “Show me how to use your cell phone.”-Subjects must use 10 words with the voiced TH sound.-If they don’t, prompt them with questions until they have used 10 words which have the voiced TH sound.Example of response: “You push this (1) button to turn it on. Then (2) this (3) light comes on. You press these (4) other (5) buttons to call someone. You hold it up to your ear and talk.” How do people call you? “They (6) dial my number and the (7) phone rings”. How do you retrieve your voice mail messages? “I do this (8) and then (9) this.(10)” How do you change your ring tone?” etc.
- Ask the subject to tell you about their countries or what they like/don’t like/are surprised about in this country. Continue asking them questions until they have used 10 words with an EE sound.Example: “In my country (1) people (2) work outside on farms or in offices.” Tell me more. “Children go to school until they are 13 (3) and then they decide if they want to go to college.” Tell me what subjects you study. “ We (4) study (5) math, history (6), geography (7-8), English.” Tell me about how people dress. “We usually (9) wear these (10) clothes”. Be careful not to count words that can be pronounced with the schwa (“the,” “decide,” etc.)
Recorded Reading Aloud
Ask subjects to read these sentences aloud.
Libraries ask students for IDs before letting them check out books or journals. Most materials can be checked out for 2 weeks. The main library is open 7 days a week and the hours are posted on the doors.
Is this one further away than that one? I can’t see either of them very clearly. Move the blue one to the right. Then, move the other one to the left.
Please complete each of these forms and return them to me. We need your student ID on line 3. Write clearly.Score responses
Libraries ask students for IDs before letting them check out books or journals. Most materials can be checked out for 2 weeks. The main library is open 7 days a week and the hours are posted on all the doors. _____/10
Is this one further away than that one? I can’t see either of them very clearly. Move the blue one to the right. Then move the other one to the left. ____/10
Please complete each of these forms and return them to me. We need your student ID on line 3. Write clearly. _____/10
Completed Listen & Choose in American Speechsounds softwareAsk each subject to complete Listen & Choose using headphones, and without using the Replay button or the Listen buttons. They will hear the stimulus only once. Make sure the volume is comfortable for the subject before beginning. The score will appear when the subject completes each exercise. Note which words the subject made an error on.
Endings: Bells and Whistles. % score_________ Error words: (32 randomly presented pairs of words*)
Consonant THH: % score ________ Error words: (21 randomly presented pairs of words)
Vowel EE: % score ________ Error words: (22 randomly presented pairs of words)