A portion of an exploratory investigation conducted recently at Lockheed Missiles and Space Company, Sunnyvale, Calif., into the comparative effectiveness of various prediction techniques for data compression is described. The comparisons were made by simulation with an IBM 7094 digital computer with the use of approximately 150,000 samples of actual vehicle telemetry data received during a typical satellite launching, in addition to synthetic telemetry data. The compression techniques discussed employed zero-, first-, and second-order polynomial predictors, or modifications thereof. The results showed that one of the zero-order techniques (one which would be relatively simple to implement) was the most effective in removing redundancy from the data. Although these data were somewhat limited both in scope and quantity, the results tend to support the oft-repeated objection to polynomial predictors that an inordinate amount of emphasis is given to sample points far removed from the point to be predicted, compared to that afforded the most recent samples.