The spread of Classical Swine Fever (CSF) virus (strain Lorraine), originally isolated in the first CSF infected herd of the 1993-1994 Belgian epizootic, was examined in an isolation unit with three adjacent pens and 15 weaner pigs per pen. Virus was introduced through experimental inoculation of one weaner pig in the middle pen (pen 2). The experimentally inoculated pig became viraemic 4 days post-inoculation (dpi) and the pen mates at 12 (n = 9) and 14 dpi (n = 5). The first viraemia in pens 1 and 3 was observed 18 dpi. Pigs were found to be seropositive in pens 1, 2, and 3 from 24, 20, and 22 dpi onwards, respectively. The reproduction ratio (R0) for the pigs in pen 2, estimated according to the martingale method, was 81.3 (s.e. = 109.54). The rate ratio (Cox proportional hazard) of the first pigs to become viraemic in pen 3 (airborne contact plus contact via contaminated clothing and footwear with pen 2) versus pen 1 (airborne contact with pen 2 only) was 1.60 (P = 0.3342). Thus, the additional contact of contaminated clothing did not affect transmission of the CSF virus. The survivor function (Kaplan-Meier survival analysis) did not significantly differ per pen. The time from first detection of virus in plasma to death was not significantly different between pens. The mean rectal temperature of pigs in a pen increased 3 to 4 days prior to detection of virus. The proportion of seropositive pigs per pen (p) from the day the first weaner pig in a pen became viraemic (dpf) was examined as a function of time with a logistic regression model. The model parameter estimates did not differ between pens. Hence, the data from the three pens were pooled. The regression equation of the seroprevalence over time for the pooled data was p = 1/[1+e(4.65-0.39 *dpf)].