The Speech Transmission Index (STI) is routinely applied for predicting the intelligibility of messages (sentences) in noise and reverberation. Despite clear evidence that the STI is capable of doing so accurately, recent results indicate that the STI sometimes underestimates the effect of reverberation on sentence intelligibility. To investigate the influence of talker and speaking style, the Speech Reception Threshold in noise and reverberation was measured for three talkers, differing in clarity of articulation and speaking style. For very clear speech, the standard STI yields accurate results. For more conversational speech by an untrained talker, the effect of reverberation is underestimated. Measurements of the envelope spectrum reveal that conversational speech has relatively stronger contributions by higher (> 12.5 Hz) modulation frequencies. By modifying the STI calculation procedure to include modulations in the range 12.5 - 31.5 Hz, better results are obtained for conversational speech. A speaking-style-dependent choice for the STI modulation frequency range is proposed.