This study aimed to derive functions to predict the intelligibility of passages from the compact disc version of the Connected Speech Test (CST) when they are heard through different filters and at different signal-to-noise ratios. A frequency-importance function (FIF), a performance-intensity function (PIF), and an audibility-index (AI) transfer function (TF) are reported. They were derived from the scores of 60 normalhearing adults who had listened to filtered CST passages masked with talker-spectrum-matched (TSM) noise. The FIF, in 1/3-octave bands, is a bimodal curve. The PIF in TSM noise and the TF are both asymmetric S-shaped curves. Comparisons between the results of this study and others reconfirm that different speech materials have different AI functions. The FIF for the CST overlaps l/3-octave band functions for continuous discourse and average speech but does not have the same shape as those functions. The TF indicates that CST passages are generally more intelligible than isolated monosyllabic words (NU6 lists) and somewhat less intelligible than continuous discourse. The former result is probably due at least partly to the effects of context, whereas the latter result may be due primarily to how clearly the talkers pronounced the speech materials.