In automatic milking systems (AMS), there is variability in milking intervals (MI) within and between cows. Extended MI (particularly greater than 16 h in pasture-based systems) have a negative effect on milk yield (MY). Having cows that tolerate extended MI without negative effects on MY could improve overall system performance. The aim of this study was to describe MI of cows with contrasting production milked in pasture-based AMS. A database containing records of milking events for 917 multiparous cows for one year (July 2015 – June 2016) from 2 AMS farms in Australia was used. Each record contained farm, cow, lactation, days in milk, MI and MY. Daily yields were then calculated as the sum of milking events. Lactation curves were adjusted with an incomplete gamma function (Wood, 1967) with a random intercept. Daily yields were expected to be auto-correlated. The adjustments were made with PROC NLMIXED from SAS. Predicted curves of average daily production according to lactation (2 vs 3 or more) and calving season (warm vs cool) were obtained. The best linear unbiased prediction (BLUP) allowed categorization of cows and lactations as having either high or low milk production (positive and negative BLUP, respectively). Then, each MI were categorized as belonging to short (<16 h) or long (≥16 h) MI. Differences between categories were compared using the binomial test of proportions. There were significant differences in the distribution of MI within each production level (P < 0.0001). High production cows had 34% of milking events with intervals greater than 16 h and were 4.13 times more likely than low production cows to have low MI (<16 h). Preliminary results also indicate that high production cows do not have such a negative effect of longer MI on MY as the low production cows do. This study indicates that there are cows that have long MI and still maintain high levels of production. Identifying and selecting for these cows should enable improve robot performance.