On-farm computing is increasingly becoming an integral part of farm decision systems. To aid the development of appropriate systems and their efficient use it is important to understand the changing farm computing scene. The study reported here contributes to these objectives in the New Zealand case. The situation is likely to be similar to other western countries. The basis of the study was data from a postal survey over summer 1997/98, together with the results from previous similar surveys. The postal survey of 3,021 randomly selected New Zealand primary producers enabled exploring the penetration of on-farm computers and details of their use. The response rate (49.5 %) was exceptional with 1,437 valid replies being received by the mid-April 1998 cut-off date. Computer penetration was 42.72% of the sample compared with 6% in 1986 and 24.40% in 1993. The comguter farms tend to be larger than non-computer farms, the managers tend to have higher levels of formal education, they tend to be younger, and they tend to be involved in more off-farm businesses. From ownership/intended ownership details it appears the uptalce rate is probably at a maximum. Computer use is around 20 hours per month with word processing, financial recording and analysis as well as financial budgeting continuing to be the important uses. The farm manager and his or her spouse are the main business use operators (78.5%). Most users (89%) believe a comguter is an economic investment. Of increasing importance is the use of the Internet with some 3 hours/month spent on Intemet access and communication. Currently 28% of computer users have a connection, but a further 40% indicate they will connect in the next two years. E-mail is the main use of the Internet but entertainment and fun as well as technical information gathering are important uses. Some 47% believe the Internet is valuable or better with 37% still being neutral or undecided. Generally, there are few differences when the data is divided by farm type, suggesting most managers view a computer similarly for all production types. Of major significance is the conclusion that computer owners and non-owners are not inherently different in their objectives. While M e r work on a wider range of variables is necessary, this suggests training programmes and software need not be markedly different for each group.