Reliability of the Speech Sounds Perception Test

Reliability and Item Analysis of the Seashore Rhythm Test

The internal reliability of the Seashore Rhythm Test was examined for a sample of 376 patients. Two different combinations of items yielded reliability coefficients of approximately .75. Repeated-measures analysis of variance indicated significant differences in the mean numbers of errors occurring on the three subtests. Item analysis showed considerable variability in the rate at which the items were passed. Items with low pass rates were identified among all three subtests which suggests that factors other than length of the patterns to be discriminated contribute to difficulty of items.

Detecting Insufficient Effort Using the Seashore Rhythm and Speech-Sounds Perception Tests in Head Injury

This study examined the capacity of the Seashore Rhythm Test (SRT) and the Speech-Sounds Perception Test (SSPT) to detect insufficient effort in a clinical sample. Forty-six participants with financially compensable mild head injury who obtained scores indicative of insufficient effort on multiple measures were compared to 49 participants with brain injury who were not involved in litigation. Receiver operating characteristic (ROC) curve analysis indicated that both the SRT (AUC = .84) and SSPT (AUC = .80) were significant (p < .001) predictors of insufficient effort. Maximizing sensitivity and specificity, the optimal cutoff scores were 8 errors on the SRT and 10 errors on the SSPT. Combining both variables into a logistic regression function increased the diagnostic efficiency.

Speech-sounds perception test: long- and short-forms reliability adjusted for item difficulty

The internal consistency reliability of the Speech-Sounds Perception Test was computed for the long form (.9159), short form (.8410), and the six series (.6109 to .7310). The total sample (N= 632) consisted of normal, diffusely brain-damaged, and undiagnosed patients referred for assessment.

Item analysis of Halstead's Speech-Sounds Perception Test: quantitative and qualitative analysis of errors

The pattern and type of errors on the Speech-Sounds Perception Test was examined in a sample of 513 patients referred for neuropsychological evaluation. There was considerable variability in the rate of correct response. Groups of patients with independently documented lesions were compared in regard to number and type of errors. It was found that patients with left-hemisphere lesions made a higher percentage of ending errors and a lower percentage of beginning errors than patients with right-hemisphere or bilateral lesions. These differences were maintained when the total error score was included as a covariate. It was also found that a cutoff score based on the percentage of ending errors correctly classified the unilateral-lesion patients at a statistically significant level. The potential clinical utility of these results were discussed.

Error pattern and item order on Halstead's Speech Sounds Perception Test

Examined the possibility that the previously reported pattern of errors on the Speech Sounds Perception Test could be explained on the basis of learning (N = 30). This possibility was examined by altering the order of presentation of subtests so that the series with the highest mean error scores in the previous study were moved to the positions in which fewest errors had been observed previously. It was found that in spite of the alteration of order of presentation, the subtests that had the greatest number of errors in the previous study again had the largest mean errors. Conversely, the subtests with the lowest mean errors previously again had the lowest mean errors. It was concluded that order of administration did not influence the number of errors on each subtest and that therefore the observed pattern could not be explained on the basis of learning. Further data were presented that suggested a relationship between the mean number of errors on each subtest and the number of items in each subtest with correct choices that were phonetically similar to real words.