Experimental time-temperature distributions From two different industrial scale retort systems were statistically analysed. The retort temperature was modelled as the sum of a trend value and a residual, with the trend temperatures being simple functions of time. The residuals were modelled using time-series. The resulting impact on the lethality distribution was assessed by calculating the F-value distribution in the centre of cans simulated via a conduction-heating finite element model for 180 simulated temperature histories. Comparing the distributions obtained with those calculated using the actual experimental temperature histories validated the applicability of this approach. The results indicated that the experimental and the modelled average lethalities were statistically similar at 95% confidence. The standard deviation was also similar for the F-value up to the end of holding but larger for the modelled distribution when considering the whole cycle, which was attributed to a correlation between the heating and cooling parameters that was not considered in the model. (C) 2000