You are on page 1of 53

Guidelines for promoting

intelligibility
John M. Levis
Iowa State University
jlevis@iastate.edu

Why I am exploring this topic?
Intelligibility is widely agreed to be the most
important goal for spoken language
development, for both listening and speaking
It is the most important goal for ESL settings
and for non-ESL settings (both where NNSs
will interact with NSs and where they interact
primarily with other NNSs)
Intelligibility is a moving target, depending on
the interlocutors, situation, register, and other
elements of context. Thus, context-sensitive
principles are needed to make decisions.
Overview of talk
What is intelligibility?
Why is it important?
What is thought to promote intelligibility?
The segmentals/suprasegmentals
debate
Guidelines for promoting intelligibility
Nuanced intelligibility
Recommendations

What is intelligibility?
General definition
Intelligibility may be broadly defined as the
extent to which a speakers message is
actually understood by a listener (Munro &
Derwing 1999, p. 289)
This broad definition implies at least two
different types of understanding
Successfully identifying words
Understanding a speakers intended meaning
Successfully identifying words
Intelligibility (technical definition)
The ability of listeners to accurately decode
individual words in the stream of speech
or, The ability of a speaker to say words in such a
way that listeners can decode them
Pronunciation deviations do not necessarily
impair the ability to decode, e.g.
Dialect pronunciations
English as a Lingua Franca (Nonnative) speaker
pronunciations

Understanding intended meanings
Comprehensibility (two definitions)
The accuracy with which a speakers intended
meaning is perceived (this implies a way to
measure comprehension)
Hahn (2004)
The perception of how easy it is to understand a
speaker (this implies a more global view of
comprehension that trusts listeners intuitions)
Derwing and Munro; Munro & Derwing (various
references)
Why is this important?
Teaching for intelligibility/comprehensibility
implies a principle of differential importance
Some pronunciation errors are more likely to affect
understanding than others
Some pronunciation teaching topics should be
emphasized while others should not
Theres a practical reason as well. It is rare to
have courses devoted to pronunciation
instruction. So we need to make changes
quickly and effectively.
Triage (Judy Gilbert)
Native-like accents and intelligibility
Intelligibility assumes that native-like
pronunciation is not an important goal;
Rather, its important to be understandable
even if accented. Why not a native accent?
It doesnt seem to be possible for most learners
Its not necessary (unless youre a spy)
Language proficiency does not depend upon
having a native-like accent
Everyone, even native speakers, has an accent.
Being native-like usually means privileging one
accent above other appropriate accents.
What promotes intelligibility?
Some proposals
A focus on suprasegmentals
a short-term pronunciation course should
focus first and foremost on
suprasegmentals, as they have the greatest
impact on the comprehensibility of learners
English (McNerny & Mendelsohn, 1992,
186)

An emphasis on the big picture
The Zoom Principle
A pronunciation syllabus should begin with
the widest possible focus [i.e., general
speaking habits] and move gradually in on
specific problems (Firth 1992, 173)
Attending to errors that affect NNS
understanding
Most speakers of English in the world are NNSs
who speak English with other NNSs.
Multiple Englishes imply a need for an
internationally understandable norm
if we are to provide appropriate pedagogic
proposals for EIL pronunciation, then these must be
linked directly to relevant descriptions of NNS
speechin terms of what constitutes optimum
productive competence and what learners need to
be able to comprehend (Jenkins, 2002, 84)
Segmentals vs. Suprasegmentals
The traditional debate
Suprasegmentals are more likely to promote
comprehensibility
But, segmentals are obviously important
But, suprasegmentals are more likely to
reveal common problems across a range of
first language backgrounds
But.


Why the debate is not useful
You cant have one without the other
Rhythmic structure (a suprasegmental) and vowel
quality (a segmental) are interdependent, e.g.
rcord vs. recrd
Rhythmic structure and consonant clarity are
closely connected (e.g., aspiration of initial stops
rappl vs. rpid; deletion of /h/ in unstressed
syllables Did he do it? are affected by stress
patterns
Differential importance applies within
categories as well as across categories
Final consonant errors in Vietnamese-
accented English impair listener
understanding more than initial
consonant errors (Zielinski, 2006)
Field (2005) studied listeners ability to
understand 2-syllable word stress errors with
and without changes in vowel quality. He
found that there is a significant decrement in
intelligibility when stress is shifted to an
unstressed syllable without an accompanying
change in vowel quality (p. 414). When the
stress shift was accompanied by a change in
vowel quality (from weak to full) the loss of
intelligibility was considerably less marked
(p. 415)
Seven Guidelines for teaching for
intelligibility
Derwing & Munro (2005) call for
decisions about pronunciation teaching
to be based on research. This is
important, but as they admit, there is not
enough research yet to base all
decisions on it. So,
Some of my guidelines come from research
Some come from practice
These are offered in no particular order
of importance and show some overlap
Seven guidelines
Functional load
Frequency
Potential for penalty
Probability of offense
Lexical importance
Processing constraints
Learnability

Guideline 1: Functional load
[Functional load] is a measure of the
work two phonemes do in keeping
utterances apart (King, 1967, as cited in
Munro & Derwing 2006, 522). Functional
load is measured partly by
# of initial minimal pairs two sounds have
# of final minimal pairs two sounds have
Likelihood that the distinction is enforced in
all varieties of English
Munro & Derwing 2006
Tested NS subjects listening to sentences
with high and low functional load errors
High functional load errors
/l/-/n/ (light-night), /s/-//(sell-shell), /d/-/z/ (ride-
rise)
Low functional load errors
//-/d/ (then-den), //-/f/ (three-free)
Subjects rated accentedness and
comprehensibility of the sentences
Accentedness (on a scale of 1-9)
Perceived comprehensibility (on a scale of 1-9)
Low FL errors and accentedness
the presence of one, two, or three low FL
errors resulted in significantly worse
judgments of accent than the presence of
no errors (527)

High FL errors strongly affect
comprehensibility
high FL errors had a significantly greater
effects on the listeners ratings [for
comprehensibility] than did low FL errors.
Even sentences that contained only one
high FL error were rated significantly worse
for comprehensibility than sentences
containing three low FL errors (527)
Conclusions
Errors in phonemes that carry a high
functional load are more likely to affect
listeners ability to understand than are
errors with sounds that carry a low
functional load
Guideline 2: Frequency
Base Belief: Speech that contains more
phonetic and phonemic errors will be less
understandable than speech that contains
fewer
unintelligibility[is] the cumulative effect of
many little departures from the phonetic norms
of the language. A great many of these may
be phonemic; many others are not. Under
certain circumstances, any abnormality of
speech can contribute to unintelligibility
(Prator & Robinett, 1985, xxii)
Munro & Derwing 2006
There is some evidence for and against this
concept, again from the previous study on
functional load
Low FL errors and frequency
the presence of one, two, or three low FL errors
resulted in significantly worse judgments of accent
than the presence of no errors. However,
sentences with two or three low FL errors were not
rated as more accented than sentences that
contained a single FL error. In other words, there
was no evidence of a cumulative effect of low FL
errors on accentedness (527)
Frequency of high FL errors and accent
Althoughthe presence of one or two high
FL errors led to a significant increase in the
perception of accentedness over the no-
error condition, sentences containing two
high FL errors were rated as significantly
more accented than sentences containing
only one high FL error. In other words, a
cumulative effect of high FL errors was
seen (527)
Frequency of high FL errors and
comprehensibility
Sentences with one and two high FL errors
were equally comprehensible
It may bethat numbers of segmental
errors alone do not account fully for
variability in accentedness or
comprehensibility. Rather, the nature of the
errors may affect their performance (530)
Guideline 3: Potential for Penalty
Certain contexts of use have higher stakes for
the speaker and listener than others. If youre
selling in a shop in an area where ethnic
shops are the norm, your needs for
understandable pronunciation are lower than if
you are a doctor or a nurse. Some high
stakes areas:
Education (International teaching assistants)
Health (Medical personnel)
Translation (Spoken language translators)
While this principle is important, it also
opens the very real possibility of
prejudicial judgments of speech that
have nothing to do with being
understood
Rubin (1992)
Lippi-Green (1997)
Munro (2003)
Guideline 4: Probability of offense
When mispronunciations sound like
taboo words
Beach, sheet, piece (/i/ vs. /I/)
Taboo sound-alikes can come up in very
unexpected places
French class speaking about silverware,
students were being unresponsive, teacher
changed to English A fork [sounded like
fuck]! Havent you ever had a fuck?]
focus
These kinds of mistakes carry the possibility
of extreme distraction or embarrassment
and need to be addressed, either by
instruction or avoidance
Taboo sound-alikes fit with a concept
related to intelligibility/comprehensibility,
irritation. Irritation can occur whenever a
listener finds speech understandable but
unpleasant for some reason, such as type
of accent.
Guideline 5: Lexical importance
Some words carry key content more than
others This is especially important in high
stakes communication contexts.
Ability to guess meaning from context in
reading comprehension is impaired when
fewer than 95% of the words are known.
When fewer than 80% of the words are known,
the ability to understand is very low (Nation
1990)
If these are the figures for reading, where the
permanent nature of the text is a significant
help, what must they be for listening?
Example: International Teaching Assistant
instruction
Most ITA training has an emphasis on pronouncing
key technical vocabulary correctly
There is also often attention to pronouncing key
sub-technical vocabulary (words that cut across
disciplines, such as develop) understandably
When content is unfamiliar, understanding is
impaired both by the subject matter and the
way the content is packaged (the spoken
qualities of the message)
Guideline 6: Processing constraints
Unfamiliar messages will take longer to
process than will familiar messages
Familiarity in content
Familiarity in speech style
Heard coROLLary in a talk by Wilga Rivers when I
expected COroLAry It took me 45 seconds to unpack
the segmentals
NNS listener perceptions (Jenkins 2002) let cars and
clay houses)
NS perceptions (Munro and Derwing)
Expected mistakes vs. unexpected ones
When knowledge of the world (top-down processing)
and the understanding of the speech details (bottom-
up processing) do not match, NS listeners will first try
a top-down interpretation that makes sense. If that
does not work, they will try to process from a bottom-
up perspective. Or they will give up.
If there is insufficient knowledge of the world (top-
down knowledge) then listeners must rely more
heavily on bottom-up processing.
There is evidence that L2 learners rely more heavily
on bottom-up processing than do L1 learners in their
native language
When processing constraints interact
with high-stakes listening where the
potential for penalty is great, the
problems can be enormous
ITAs teaching in any college field
Miranda warnings and word frequency

Reaction time research
The work of Anne Cutler and her colleagues (see
Cutler, Dahan, & van Donselaar 1997 for a review)
consistently shows that unexpected elements in
speech affect listeners ability to process speech
Other research, such as John Fields study,
measure the cases where intelligibility is entirely
lost (2005, p. 415). This kind of research is less
sensitive to processing demands than reaction time
research
Munro & Derwings research shows that
loss of comprehensibility (where the
speakers intended meaning does not seem
clear) is far more common and probably
more serious than loss of intelligibility
(where a word cannot be understood). This
is likely due to processing difficulties.

Guideline 7: Learnability
Some features of pronunciation are more
learnable
Jenkins (2002)
[Besides not being a cause of unintelligibility in
Jenkins data] as many pronunciation teachers
are aware, some of these features seem to be
unteachable. That is, no matter how much
classroom time is spent on them, learners do not
acquire them (97)
Some areas not included in Jenkins Lingua
Franca core (2002)
// (thank), // (then), // (will)
Weak forms, especially the use of // (schwa) in
words like to, and, from. In EIL, the full vowel
sounds tend to help rather than hinder intelligibility
(98)
Final Pitch movement
Levis (1999) - Pitch movement differences on certain
types of grammatical forms (yes-no questions) are not
important to teach

Further evidence that some features
may not be learnable
Pennington and Ellis (2000)
Recognition tasks for several aspects of
intonation/stress for Cantonese speakers
learning English
Contrastive sentence focus (Is HE driving the bus?
Vs. Is he driving the bus?)
Final pitch movement on tags (Hes going, isnt he?
(rising vs. falling)
Phrasing (The fight is over, Fred vs. The fight is over
Fred)
Internal phrase structure (Shes a lighthouse keeper
vs. Shes a light housekeeper)
Recognition tasks in two conditions:
When there was no previous instruction,
and when there was.
Subjects performed well on recognizing the words
and grammar of previously heard sentences when
prosodic form was not tested
Subjects performed poorly on recognizing
previously heard sentences if prosodic form was
included. This was especially so when there was
not previous explicit focus on form.
Training with explicit focus on prosodic form
increased recognition ability only for
contrastive sentence focus. The other aspects
of intonation/stress were not amenable to
instruction
Why did only contrastive sentence focus show
improvement?
certain aspects of prosody --- such as the
relatively universal relationship of enhanced
prosody and marked meaning, as contrasted with
neutral prosody and unmarked meaning --- can be
more readily taught than some other more
language-specific aspects [of prosody] (p. 387)


Recommendations
Take a nuanced view of any target. It is
likely that all phonological categories
include more and less important
features, e.g.,
Consonants (Some targets are important,
some are less so e.g., /l/-/n/ is more
critical than //-/f/ or //-/d/)
Vowels (phonetic length may be more
important than phonemic quality Jenkins
2002)

Intonation (Not all kinds of intonation are
likely to be equally important. Sentence
focus is likely to be important (Hahn 2004;
Pennington and Ellis 2000) while final
intonation, especially on certain
grammatical structures, is much less so
(Levis 1999; Pennington and Ellis 2000)
Word Stress (e.g., Rightward misstressing
affected intelligibility more than leftward
(Field 2005)

Distinguish between listening and speaking
Listening improvement can lead to production
improvement
Learning to hear the /l/-/r/ distinction can lead to
better production for Japanese learners even
without practice (Bradlow, Pisoni, Akahane-
Yamada, and Tokhura1997)
Even when features are not considered learnable
in normal classroom instruction, teaching the
feature for reception may be critical for later
acquisition outside the classroom (Jenkins 2002)
Building a range of tolerance for understanding
(listener training) will develop flexibility


Recognize that not all learners need to
function in the same contexts
Some need to understand and be understood by
NS interlocutors
Some need to understand and be understood by
NNS interlocutors
There is compelling evidence that ELF [i.e.,
NNS] interlocutors engage in communication
strategies and accommodation processes that
may conflict with the ways in which NSs
typically negotiate understanding (Pickering
2006, 227)
Recognize that pronunciation is more
than listening and speaking
Visual support can be critical
Body language (e.g., for negation)
Visual support (e.g., ITAs)
Circumlocution is useful for any speaker
Oral spelling or restatement of numbers can
quickly disambiguate many situations
What might be our priorities?
Functional load and Frequency
Word Stress
Consonants, including high functional load
consonants, aspiration and final consonants with
grammatical meaning
Vowel lengthening and vowel quality
Weak forms and fast speech phenomena (for
listening)
Potential for penalty, Probability of offense,
and Lexical importance
Key vocabulary for speaking needs
Processing constraints
Sentence Focus
Word Stress
Weak forms and fast speech phenomena
(for listening, especially in ESL contexts)
Learnability
Sentence focus
General speaking habits


The PowerPoint slides for this talk will be
available after March 26 at

jlevis.public.iastate.edu/intelligibility.ppt

References
Bradlow, A., Pisoni, D., Akahane-Yamada, R. & Tokhura, Y. (1997).
Training Japanese learners to identify /l/ and /r/: IV: Some effects of
perceptual learning on speech production. Journal of the Acoustical
Society of America,
Brown, A. (1989). Functional load and the teaching of pronunciation.
TESOL Quarterly, 22 (4): 593-606.
Cutler, A., Dahan, D. & van Donselaar, W. (1997). Prosody in the
comprehension of spoken language: A literature review. Language and
Speech, 40 (2):141-201.
Derwing, T. & Munro, M. (2005). Second language accent and
pronunciation teaching: A research-based approach. TESOL Quarterly,
39 (3): 379-397.
Field, J. (2005). Intelligibility and the listener. The role of lexical stress.
TESOL Quarterly, 39 (3): 399-423.
Firth, S. (1992). Pronunciation syllabus design: A question of focus. In
Teaching American English pronunciation, Oxford University Press, pp.
173-183.

Hahn, L. (2004). Primary stress and intelligibility: Research to motivate the teaching
of suprasegmentals. TESOL Quarterly, 38 (2): 201-223.
Jenkins, J. (2002). A sociolinguistically-based, empirically-researched pronunciation
syllabus for teaching English as an international language. Applied Linguistics,
23: 83-103.
Levis, J. (1999). The intonation and meaning of normal yes-no questions. World
Englishes, 18 (3): 373-380.
Lippi-Green, R. (1997). English with an accent. Routledge.
McNerny, M. & Mendelsohn, D. (1992). Suprasegmentals in the pronunciation class:
Setting priorities. In Teaching American English pronunciation, Oxford University
Press, pp. 185-196.
Munro, M. (2003). A primer on accent discrimination in the Canadian context. TESL
Canada Journal, 20 (2): 38-51.
Munro, M. & Derwing, T. 1999. Foreign accent, comprehensibility, and intelligibility in
the speech of second language learners. Language Learning, 49 (supp. 1): 285-
310.
Munro, M. & Derwing, T. (2006). The functional load principle in ESL pronunciation
instruction: An exploratory study. System, 34: 520-531.

Nation, I. S. P. (1990). Teaching and learning vocabulary.
Newbury House.
Pennington, M. & Ellis, N. (2000). Cantonese speakers memory for
English sentences with prosodic cues. Modern Language
Journal, 84 (3): 372-389.
Pickering, L. (2006). Current research on intelligibility of English as
a lingua franca. Annual Review of Applied Linguistics, 26: 219-
233.
Prator, C. & Robinett, B. 1985. Manual of American English
pronunciation, 4
th
ed.. Rinehart Holt Winston.
Rubin, D. (1992). Nonlanguage factors affecting undergraduates
judgments of nonnative English-speaking teaching assistants.
Research in Higher Education, 33: 511-531.
Zielinski, B. (2006). The intelligibility cocktail: An interaction
between listener and speaker ingredients. Prospect: An
Australian Journal of TESOL, 21 (1): 22-45.

You might also like