The study of the perception of speech is aided by instruments capable of representing physical patterns which are complex in frequency, time, and intensity. The sound spectrograph developed by Potter and co‐workers provides visual patterns which are highly suggestive for the isolation of the distinctive aspects of auditory patterns. However, comparison and judgment by ear are desirable, and this requires additional instrumentation to play back modified spectrograms, or completely synthetic spectrographic patterns which differ only in the particular characteristic under examination. The spectrograph developed for these studies records on film, portraying a dynamic range ofca.40 db by a linear density variation of 2.0. The use of a Photoformer to control the recording light permits compensation for non‐linearities of other components, or the intentional introduction of high contrast (compression), black‐to‐white reversal, or intensity‐contour characteristics into the spectrograms. The pattern playback uses these spectrograms directly, or after retouching or printing, to control the modulation at syllabic rates of an optical scanning beam which is then converted into sound. Synthetic spectrograms, hand‐drawn on a transparent medium, are used in a similar manner. Spectrograph and playback were designed for specific application to speech‐like sounds and to studies of the perception of such sounds. Design considerations and performance data are discussed.