The spark-chamber analysis was performed in two parts: firstly an automatic pattern recognition program assigned the most likely gamma-ray events (usually less than 4% of all recorded events), secondly these events were subjected to a visual scan to check for correct track assignment and classification. The editor could make reassignment of the tracks and reject the obvious background triggers mimicking the gamma-ray pattern.
Over 2 million events have been manually edited by about a dozen operators from 3 institutes during 8 years, hence the editing standard is somewhat variable. To measure the size of this variability the anticenter observation, 39, was edited in all three institutes. The differences in source intensity derived for the Crab and Geminga in the three data sets was () although on an event by event basis the data showed technical differences. The major effect of the different editing standard is in the background rejection, some institutes being more discriminating than others. Differences in angular resolution between the different establishments cannot be excluded: although are probably of second order.
With this understanding of what can cause temporary variations in the COS-B performance, it is recommended that a background sample from the same observation period be examined before claims of source variability be made.