计分单位对单音节识别测试的影响

The effects of scoring unit on monosyllabic word recognition test

目的比较以整字和音位为单位计分进行单音节字识别测试的结果,探索更加合理的计分方式。方法将自行编制的每张词表包含25个单音节字的汉语单音节短词表共30张制作成测试CD。60名听力正常的受试者参加实验,男女各30名。测试采用拉丁方设计。每张词表在6个强度上进行测试。选择受试者听力较好耳进行测试。反应方式为受试者复述测试词。记录每个测试项的正确和错误模式。分别以整字和音位(声母、韵母、声调)为计分单位记录每张词表的测试项正确数目。使用Matlab对各张词表进行logistic曲线拟合,绘出每张词表在两种计分方式下的PI曲线,计算每张词表的识别阈值和PI曲线的斜率。使用SPSS11.5统计软件对数据进行统计学处理。首先使用配对t检验分别比较6个测试强度上两种计分方式得到的总测试项正确率;然后使用配对t检验比较两种计分方式所得的各词表PI曲线斜率和阈值。结果以音位为计分单位得到的50%识别阈明显低于以整字为单位(P<0.001),20%～80%斜率(P<0.001)和50%斜率(P<0.001)则明显小于以整字为单位得到的结果。按音位计分时在各个强度上的识别正确率均显著高于按整字计分(P<0.001),且得到的结果离散度较大。结论以音位为计分单位,测试信度有一定程度的增加,但同时测试项同质性变差,降低了测试的灵敏度。临床上使用短词表进行单音节识别测试时使用整字作为评分单位较为合理。

Objective To study the application of different scoring methods in monosyllabic word recognition test through comparing the results of two scoring units： word scoring and phoneme scoring. This will help establish a more appropriate methodology of word recognition examination. Methods Thirty monosyllabic word lists of mandarin words were developed, 25 test items in each list. Sixty subjects with normal hearing took part in this test, 30 male and 30 female. Each list was presented at 6 different presentation levels. Signals were delivered to the better ear and the subjects were asked to repeat the words that they thought they had heard and to write down the error patterns. The total number of correct test items was recorded in two ways： correctly repeated words in total and correctly repeated phonemes in total. PI function curves for each list were presented to show the relationship between recognition score and word presentation intensity level, using a logistic regression tool by Matlab. The recognition threshold （dB HL） and PI curve slope （%/dB） were calculated and further studied. Results The 50% recognition thresholds obtained from phoneme scoring units were lower than those of word scoring units （P〈0.001）, and 20%-80% slope （P〈0.001） and 50% slope （P〈0.001） smaller than those of word scoring units. At each presentation level, phoneme units scored more than word units （P〈0.001） and phoneme scores spreaded widely as opposed to word scores. Conclusion The phoneme scoring method can improve the reliability, while the homogeneousness of test items gets worse, thus reducing the sensitivity of the test. Word scoring method is recommended as scoring unit for monosyllabic word test, especially in short test lists.