Measurements of real data traffic have been widely recognized as an important aid in performance modeling, design and development of telecommunication networks. Newly emerging services and new applications of data communications demand more exhaustive measurement studies of data traffic to provide foundations for the development of telecommunication networks of the next century. A survey of publications on data traffic measurements before 1988 was presented at the 12th ITC Congress [PAW88]. Most of the reported research activity in this area is user- or system-oriented and emphasises the dependence of data traffic on network architecture and/or communication protocols. Little attention has been paid to studying the influence of various data transformations on the statistical properties of data traffic. In this paper we investigate the changes in length of data packets caused by data compression. A specific feature of this data transformation lies in its inherently random output rate. Even a regular stream of bytes is compressed into a bursty stream of bits. The output rate of a compressor will be lower than the input rate on average, but is unpredictable for each byte. Several input characters may cause only one bit to be added to the output, while the next character could generate several bits. At the packet level the sporadic output from the character level is moderated, but does not dissappear. The resulting burstiness in streams of compressed packets can be observed regardless of whether packets are of equal size before compression (as in the OSI architecture, where compression is applied to packets after they enter a packet switching network, that is, below the transport layer) or of random size (as in many local data communication systems) [BELL89]. In both cases some packets are compressed well, while others do not, and some of them may even be expanded. In this paper we are concerned with the changes caused by compression in streams of packets of random length in local interactive data communication systems.