A Conjecture on Accent (1997)

A Conjecture on Accent in a Second Language 
Stephen Krashen
In: Z. Lengyel, J. Navracsics, and O. Simon (Eds.) 1997. Applied Linguistic Studies in Central Europe, vol 1. Department of Applied Linguistics, University of Veszprem, Hungary.

Scientists use the term "conjecture" when their generalization is based on such flimsy evidence that it does not deserve the label "hypothesis." This is such a case. My conjecture is that accurate pronunciation in a second language, even in adults, is acquired rapidly and very well. We simply do not use our best accents because we feel silly.

Restated in more respectable terms, we have an "output filter," a block that keeps us from doing our best, from "performing our competence." This block is powerful and it is difficulty, maybe impossible, to lower or weaken it with conscious effort. (The output filter differs from the affective, or input filter. The affective filter prevents input from reaching the language acquisition device. The output filter prevents us from using what we have acquired.)

Here is the flimsy evidence. Much of it is based on my own experience, but I suspect, after presenting these ideas to a number of audiences and getting reactions, that others have had similar experiences.

1. Variability: Our accents in second languages vary, depending on how we feel. We are influenced by the situation, especially whether we feel we are being evaluated. When I speak French to someone who doesn’t speak English (or at least not very well), where there is no audience, and I am comfortable with that person, I must say that my accent is not bad. On other occasions I have been told that I speak French without a traced of a French accent.

Here is an example of the latter, an experience I hope some readers can identify with. I was visiting Ottawa in the early 1980’s, meeting with former colleagues, discussing, in French, our work on sheltered subject matter teaching which had begun when I worked there a few years before. I was very comfortable with the group I was talking with; they included a close friend and my former French teacher. I was doing very well. While I was at the chalkboard, making a point, a stranger entered the room. My mind raced: This man is probably a native speaker of French, or at least much better than I am, and he probably thinks my French is terrible. My accent and fluency deteriorated immediately and involuntarily. In other words, my output filter went up.

One of the most accomplished polyglots in the world, Dr. Kato Lomb of Hungary, reports that she has had similar experiences. Now 88, Dr. Lomb has acquired 17 languages and is now working on Hebrew. I visited Dr. Lomb several times recently, and we spoke English (her English is excellent). On one visit, my wife and daughter came with me.  Dr. Lomb remarked to me that she felt her accent in English had been better when we were alone. She explained that she felt quite comfortable with my wife and daughter, but the fact that she did not know them as well as she knew me caused a small amount of self- consciousness and hurt her performance. Dr. Lomb is an enormously successful language acquirer and an experienced interpreter; if she feels the effects of the output filter, we can be sure others do.

2. Our ability to imitate other dialects of our first language, as well as foreign accents. Given sufficient input, we can all do these things to at least some extent. The point is that we do not, because we would feel uncomfortable doing so. The output filter holds us back.

I can imitate, to some extent, a British accent. I have acquired the rules for doing so subconsciously, and have no idea what kind of articulatory adjustments I am making when I do it. I do not, however, use a British accent when speaking to someone from London. My perception is that it would be rude, and even ridiculing, as if I were making fun of his speech, or as if I were representing myself as someone I am not.

Similarly, we can imitate foreign accents in our first language. Obviously, we do not do this in ordinary conversation. It would, we feel, be perceived as rude.
There are domains in which the use of these accents is permitted, in plays and jokes, for example. Even in these situations, however, their use is sensitive. In plays, dialects must be rendered very accurately, and in jokes their use can be demeaning.
Our ability, yet reluctance to use accents and dialects again shows that we do not perform our competence fully and that there are powerful affective forces holding us back.

3. The alcohol study. Guiora, Beit-Hallahmi, Brannon, Dull, and Scovel (1972) asked subjects to drink different amounts of alcohol after eating a candy bar. Not unexpectedly, they reported that subjects’ short-term memory decreased with greater consumption. Accent in a foreign language, however, was best after subjects drank 1.5 ounces of bourbon. It was less accurate with both less and more than this amount of alcohol. There was, in other words, an optimal point of inebriation. As most of us know, alcohol has the effect of lowering inhibitions. My interpretation of these results is that alcohol lowers the output filter, at least temporarily. Too much alcohol, however, disturbs control of the speech apparatus.

4. Stevick’s example. Stevick (1980) describes a Swahili class he taught at the Foreign Service Institute that had three students in it. One was at a significantly higher level than the others. When the top student had to drop the class, the number two student suddenly showed a dramatic improvement. My conjecture is that his output filter lowered, freed from the inhibiting influence of the better student.

Discussion

To understand what factors are at work here, we need to consider what language is for. Sociolinguists tell us that language has two functions: To communicate and to mark the speaker as a member of a social group. A part of language that plays a major role in marking us as members of a social group is accent. Accent has little to do with communication; we can communicate quite well in another language having acquired only some of the sound system. Accent tells the hearer who you are, where you are from, in some cases your social class, and in other cases your values. When we identify with the members of a group, we talk the way they do.

Beebe’s review (Beebe, 1985) confirms this. We do not always imitate the speech we hear the most. Children usually talk the way their peers talk, not the way their parents or teachers talk. (In some cases, children do talk like their parents; these children identify with adult values, rather than those of other children, confirming that it is group membership that counts.)

My conjecture is that accent is acquired rapidly but is not performed, because we do not feel like members of the group that uses it; we are not members of the club (Smith, 1988). Either we do not wish to be members or have not been invited to be members. And even after we feel we are at least partly in the group, we can feel suddenly excluded, resulting in a stronger output filter.

If this conjecture is correct, it has interesting implications for pedagogy. Despite the numerous "accent improvement" courses available, there is no evidence that second language accent can be permanently improved by direct instruction. Even if we could improve accent through instruction, however, the effect might be harmful. Getting people to talk like members of groups they do not belong to may be similar to convincing someone to wear inappropriate clothing - a tuxedo at an informal lunch or a jogging suit at a formal dinner.

This conjecture does not suggest that all those with accents in their second language who live in the country where the language is spoken have failed to become members of society. In fact, it suggests the contrary. Most second language acquirers have good accents. Listen to them carefully. They are rarely perfect if they began the second language as adults, but they typically acquire an impressive amount of the sound system. They certainly do not speak the second language using only the sound system of their first language. The problem is that we usually make "all or nothing" judgments with respect to accent. Either it is native-like or "accented." In reality, many second language acquirers acquire substantial amounts of the second language accent. In addition, it is likely that we hear them under less than optimal affective conditions: with lower output filters, they may sound even better.

If this conjecture is correct, another conclusion we can draw is that only our "best" accents, produced under optimal conditions, should be considered when judging accent quality or when discussing the limits of adult acquisition of pronunciation.

References

Beebe, L. 1985. Input: Choosing the right stuff. In Gass, S. and Madden. C. (Eds.) Input in Second Language Acquisition. New York: Newbury House. pp. 404-414.
Guiora, A., Beit-Hallahmi, B., Brannon, R., Dull, C. and Scovel, T. 1972. The effects of experimentally induced changes in ego status on pronunciation ability in a second language. Comprehensive Psychiatry 13: 421-428.
Smith, F. 1988. Joining the Literacy Club. Portsmouth, NH: Heinemann.
Stevick, E. 1980. Teaching Languages: A Way and Ways. New York: Newbury House.