The intelligibility of speech as a function of the context of the test materials.

Abstract

Words and sentences were read over a noisy communication system and recorded by listeners. The percentage of words recorded correctly is the articulation score. Articulation scores depend upon the context in which the test items occur. When the context permits the test item to be any one of a wide range of alternative items, the articulation score is much lower than when the context restricts the item to one of a small number of alternatives. Contextual constraints upon the number of alternatives were introduced by (1) using different materials, (2) varying the size of the test vocabulary from 2 to 1000 different monosyllables, (3) comparing words in sentences with the same words in isolation, and (4) repeating every item 3 times. (1) The ten digits can be correctly recorded at a signal-to-noise ratio 17 db worse than is required for nonsense syllables. (2) Two-word test vocabularies required 18 db less intensity than 1000-word vocabularies. (3) Sentence context improved intelligibility by 6 db. (4) Immediate repetition of the test item produced improvements of only 1 or 2 db. Thus the articulation score depends upon what the test stimulus might have been as well as upon what it actually was. A spoken word can be perceived correctly in one context, and yet the identical acoustic stimulation may not be discriminable in another context. As the number of alternative test items increases, the amount of information necessary to discriminate an item correctly also increases; therefore, the noise level must be decreased to permit a more accurate discrimination among the larger number of alternatives.