Bilger R C, Nuetzel J M, Rabinowitz W M, Rzeczkowski C
J Speech Hear Res. 1984 Mar;27(1):32-48. doi: 10.1044/jshr.2701.32.
All 10 forms of the test of Speech Perception in Noise (SPIN) were presented to 128 listeners who had some degree of sensorineural hearing loss. Presentation of the speech track was at 50 dB above the estimated threshold for the babble track. Signal-to-babble ratio was 8 dB. Half of the subjects listened through headphones and half via loudspeaker. Half were tested in a single session and half in two sessions spaced 2-4 weeks apart. Two markers independently scored every test session. Statistical analyses indicate that transducer, number of visits, and order of test form presentation have little or no effect on test scores, and differences between markers, although significant, are quite small. The subtests consisting of items with strong contextual cues generate an average reliability coefficient of .91, whereas the value for the low-context subtests is .85. The 10 forms do not, however, constitute a set of equivalent forms, and there are large differences in mean performance on the low-context portions.
对128名患有某种程度感音神经性听力损失的听众进行了噪声中言语感知测试（SPIN）的全部10种形式的测试。言语音轨的呈现比估计的嘈杂音轨阈值高50分贝。信号与嘈杂音的比率为8分贝。一半受试者通过耳机聆听，另一半通过扬声器聆听。一半受试者在单次测试中进行测试，另一半在间隔2至4周的两次测试中进行测试。两名评分员对每个测试环节进行独立评分。统计分析表明，换能器、测试次数和测试形式呈现顺序对测试分数影响很小或没有影响，评分员之间的差异虽然显著，但非常小。由具有强烈上下文线索的项目组成的子测试产生的平均信度系数为0.91，而低上下文子测试的值为0.85。然而，这10种形式并不构成一组等效形式，低上下文部分的平均表现存在很大差异。
