Unmodeled defects must be fortuitously detected by test pattern sets to achieve low defective part levels. However, fortuitous detection of many complex defects becomes less likely if test sets are structurally biased toward particular circuit configurations. In the past, we have demonstrated that test sets that observe sites multiple times with good excitation balance can fortuitously detect more defects than test sets with similar numbers of observations and less balance. Here we extend our analysis of structural bias to the path segments through which circuit sites are observed. We investigate observation diversity for all possible patterns and for particular test sets. Furthermore, we explore how this diversity relates to excitation balance.