Amee P. Shah - firstname.lastname@example.org
Research Laboratory in Speech Acoustics and Perception
Department of Speech and Hearing Cleveland State University
2121 Euclid Avenue, MC 431-B
Cleveland, OH 44115 (216) 687-6988
Popular version of Paper 4aSC1
Presented Thursday morning, November 18, 2004
148th ASA Meeting, San Diego, CA
Please listen to this speech sample:
Does this sound like it is spoken by a nonnative speaker of English, a "foreign-accented"
speaker? My guess is, you, the listener, are not only able to correctly decide
that it is indeed a sample of foreign-accented speech--but many of you may even
be able to try and identify the native language of the speaker, i.e.,
Spanish in this case. It appears (as also confirmed by numerous research studies)
that native speakers of a language are not only able to reliably detect, but
also differentiate between different accents. The interesting question is: How
are we able to (so reliably) tell when speech is "accented," or different
from the native standard form? Are accents in the "tongue of the speaker"
or the "ear of the listener", or some combination thereof?
An examination of studies to date, in conjunction with findings from some of my own research, suggests that there may be three sets of factors, i.e., a three-dimensional interaction, that apparently influence the production and perception of foreign-accented speech. These three dimensions are akin to the communication components seen in the "Speech Chain Model," i.e., variables related to the speaker, the speech per se, and the listener, all of which form the chain from production to perception of speech, and as will be shown in this paper, interact in the perception of accentedness.
First, extensive research findings substantiate the influence of speaker-related variables in the presence of a foreign accent. Among these, differences in the ages at which the nonnative speakers learn the second language are found to predict whether the speaker will show the presence of a foreign accent in their speech. The younger one learns a second language, the more native-like one sounds in that language. Additionally, beyond a certain age range (the "critical" or the "sensitive" period), usually by adolescence, learning a second language will invariably leave the speaker with a trace of "foreign-ness" in their speech, though they may have managed to master other aspects of language, such as vocabulary and grammar, with native-like proficiency.
Another factor, while not as definitive as age of acquisition but also quite significant in predicting the presence and the degree of foreign-accentedness, is the length of period the nonnative speaker has spent in the region/country of the second language. For example, the longer a Russian-accented speaker spends in the U.S., the more native-like he/she will sound in English. However, it has been shown that a stay beyond a 5-7 years' period of residence does not necessarily predict further progress towards native-like patterns of speech since learning may have "stabilized" at that point.
Of course, the length of residence in the nonnative country is often intertwined with other psycho-social variables that determine whether the nonnative speakers will gain ultimate native-like pronunciation. These are, to name a few, motivation-level of the second-language learners; relatively greater amount of interaction with native speakers of the language and correspondingly, relatively lesser amount of interaction with other nonnative/ foreign-accented speakers; the gender of the speakers; and other special abilities to perceive and produce the nonnative sounds, including imitating and mimicking sounds, discriminating pitches and so on. Thus, we can see that accented speech is not an exclusive listener-dependent phenomenon, rather it is certainly also influenced by the differences in the "tongues of the speakers."
Turning to the second dimension, the speech-related variables, predicting the presence of "accentedness" has been studied rather extensively in the field. Findings show that the interlanguage differences in the phonetic patterns of the speech of the nonnative speakers influence listeners' perception of accentedness of nonnative speech.
Findings show that the speech of the nonnative speakers learning the second language has systematic "interlanguage" features. That is, certain sound properties borrowed from one's own language, and certain others that are different from one's own language and more and more like those of the native speakers of the second language. Interestingly, nonnative learners tend to be consistently similar within the same language group, and markedly distinct from those of a different language group. For example, Spanish-accented speakers have consistently similar phonetic patterns as a group, and those are distinctly different from German-accented speakers. It is, presumably, this systematic difference that helps listeners characterize the differences across different accents and influence their perception of accentedness.
While we now know that accentedness has a physical basis, and is not simply a perceptual illusion or a psychological construct shaped "in the ear (or mind) of the listener," differences across listeners do tend to influence the perception of accentedness, which is the third dimension that completes the three-dimensional triad of perceived accentedness. Research indicates that differences across listeners, in that whether they are native or nonnative speakers themselves of the language in question and whether they share the language group with the speakers they are listening to, will greatly influence their sensitivity in detecting and differentiating a foreign accent. Relatedly, if the listeners are native listeners of the language in question, the fact that they may have received prior listening experience with foreign-accented speakers, in general, and the specific accent in particular, will likely influence their perception of accentedness. Further, the amount of exposure to a specific foreign accent will also likely reduce the listeners' perception of the degree of the accentedness for that speaker. Over time, listeners tend to become better familiarized with the idiosyncrasies of that speakers' accented speech, and may not only be able to understand the speaker better, but also may think that the speakers' speech has become "less accented"! Yet another variable that determines listeners' acuity in perceiving accented speech is the listening condition in which they hear the accented speech; the noisier the conditions, the greater the difficulty listening to accented speech, and the stronger will the accent be perceived to be. Finally, listeners' preferences and biases for certain language and/or ethnic groups will tend to affect how they perceive the accented speech that is spoken by those nonnative speaker groups. On the other hand, special listening skills and phonetic training, if any, increase the sensitivity with which listeners can discern the variations in accented speech.
This collective evidence from research in each of the three sets of variables, speaker-related, speech-related and listener-related, has raised many questions for the theoretical understanding formed within the scientific community. Controversies have ranged over areas such as the importance of critical period in predicting the ultimate attainment of second-language proficiency, the linguistic environment that shapes the perception for a second language, children's perceptual shaping in the second language over the developmental period, to name a few. Similarly, extensive research in each of these areas has yielded findings that have served as the bases to plan training approaches in ESL classrooms, or accent-modification programs. That is, questions such as whether to target the speech sounds per se or the prosody (rhythm) of speech in the training programs in order to make the foreign-accented speaker perceived to be more intelligible to listeners. Finally, the theoretical constructs and practical implications have served in speech technological advances as well. Automatic speech recognition systems, speech processing models, voice-activated devices incorporating speaker differences due to foreign accents have become areas of active research and development.