This paper describes a method for predicting speech intelligibility in reverberant sound fields. The RASTI method, which utilizes the MTF (Modulation Transfer Function) as the basic parameter, is commonly used for predicting speech intelligibility in sound fields. This paper proposes a weighted MTF, obtained by weighting the echograms with an exponential time window before calculating the MTF. Intelligibility tests using a variety of computer-generated echograms confirm that the proposed method is better for predicting speech intelligibility in reverberant sound fields than the ordinary MTF (RASTI) method.