Abstract: A time assignment speech interpolation system includes a variable length speech buffer for temporarily storing sampled signals until a transmission facility becomes available. Signals on input speech channels are converted into digital samples. If a transmission facility is available, the digital samples are reconstituted into analog signals which are transmitted over the facility. If no facility is available, digital samples are stored in a digital memory for a variable length of time until a transmission facility becomes available. Digital samples are retrieved from the memory in chronological order of their time of arrival at the memory. The variable length memory accommodates a short term overload of input signals to provide flexible buffering.