The selection rules have been defined so that they implement a reasonable mapping from the sets of descriptor values into the different kinds of relationship strength that the output membership functions represent. What is reasonable in this respect depends on how the occurrence of events in the various time windows correlates with strength of following: it is relatively easy to understand that a large relative number of events in an “Immediate follower” time window means strong relationship, while a more even distribution of events in the various time windows means somewhat weaker relationship and the accumulation of events towards the “passive follower” time window (or the mere occurrence of only relatively few events in any of the time windows) speaks for only a weak relationship. Examples of selection rules are for example:

RULE 1: If “Immediate follower” has SOME and “Passive follower” has FEW, output is GOOD.

RULE 2: If “Active follower” has SOME and “Passive follower” has SOME, output is GOOD.

RULE 3: If “Immediate follower” has SOME and “Active follower” has SOME, output is STRONG.

RULE 4: If “Active follower” has FEW and “Passive follower” has SOME, output is ADEQUATE.