First, and I apologize for harping on this issue so much, if we concentrate on supplying comprehensible input where the focus is on the message and not on the form, this will in of itself contribute to a low filter. If the topic being discussed is at all interesting, and if it is comprehensible, much of the "pressure" normally associated with a language class will be "off", anxiety will be lowered, and acquisition will result. As mentioned earlier, I think a desirable goal is that the student "forget", in a sense, that the message is actually encoded in another language.

Second, we will be able to keep the filter low by not insisting on too-early production, before the student is "ready". Language teachers (and students) associate progress in second language acquisition with speaking fluency ("Do you speak French?"), and the logical consequence of this is that we want our students to talk from the beginning. My personal view is that forcing early production, before the student has built up enough competence through comprehensible input, is perhaps the single most anxiety-provoking thing about language classes! While some students may want to talk as soon as possible, others may feel less secure until they have built up more competence. In other words, the length of the silent period is variable (see Note 12); Asher's ten-hour estimate may be "average", but it is not carved in stone. A safe procedure is simply not to force production and let the student decide when to start talking.

Closely related to speaking readiness for production is the question of error and error correction. Second language acquisition research tells us clearly that errors are inevitable, and that they will be plentiful in early stages. To give the reader an estimate, in an experiment we conducted at Queens College, we found approximately one error for every five words in compositions written by ESL students in a placement examination for our extension course (Krashen et al., 1978). The better students averaged about one error for every ten words, and the least proficient about one error for every two words! A sure method of raising the filter is attempting to correct errors, especially in beginning stages and especially in spoken language! Error correction is, unfortunately, the profession's typical reaction to error, and in my view it has been a serious mistake. There are several reasons why it is a mistake. We focus here on what is probably the most serious flaw in error correction, its effect of the affective filter.

74