A new procedure for simulating the acoustical response of a receiver to a source in a reverberant room is proposed. Convolving such an impulse response with an input signal, e.g., speech, is useful in applications where an acoustical environment must be highly reproducible, easily controllable, or is unavailable for an actual recording. The proposed procedure is an extension of the image method by Allen and Berkley [J. Acoust. Soc. Am. 65, 943-950 (1979)]. The new features include the possibility to simulate simple source and microphone directivity, as well as head-shadow effects, which can be considerable for subjects with microphones mounted close to the head. A rigid sphere diffracting a plane wave provides a good approximation for the effects of speaker as well as listener head shadow and allows simple yet realistic simulations of many relevant situations. Implementational issues are discussed and the method is verified by comparing simulation results with recordings in actual rooms. It was found that by including source directivity and head shadow, transfer functions become significantly more realistic.