We choose to count just the comparison operation for keys. We can take the instance characteristics to be the size of our data base. Best time for successful search: 1 Best time for unsuccessful search: n Worst time for (un)successful search: n Average successful search, assuming all keys have equal probability to appear ${1\over n}$\sum _{i=1}^n i = (n+1)/2$ Average unsuccessful search: n Hence, best and worth time are n, and the average approaches this value. We can go down to $n/2$ by assuming sorted lists