In this vein, recent research has often shown that reactions are not always based upon objective or portrayed characteristics of the speaker but rather upon characteristics expected by the listener (Aboud et al., 1974). Williams et al. ( 1971) also provide data supporting this sort of wish fulfillment. In one of their experimental conditions, they asked teachers to evaluate the speech patterns of a biack child whom they saw and heard on a videotape recording. Despite the fact that another (white) child's middle-class speech patterns were superimposed on the tape, the black child was nevertheless perceived as sounding "ethnic-non-standard" and was also rated as low in "confidence- eagerness" (see also below). In other words, teachers had biased their perceptions of the speech (and hence their evaluation of the child) in the direction of their stereotyped vocal expectations. Thus, even when listeners' speech stereotypes are disconfirmed in the form of speakers' actual vocal characteristics, discrepancies need not always arise at the listeners' level of cognitive awareness. Instead, listeners may organize the linguistic input according to their own predetermined and valued cognitive structures (Street and Hopper, 1982). Moreover, additional information about a speaker's class and ethnic background can not only bias one's reconstructions of their past performances but can also influence the perceptions of their subsequent behaviour (Ball et al., 1982; Thakerar and Giles, 1981). Of course, the pragmatic implications of such findings are enormous; even if the speech of non-standard dialect children is "improved", there is no guarantee that it will be recognized as such by teachers and others anyway.