It is shown that k(p+3)/2 + p-2 letter comparisons suffice to determine whether a word is a member of a lexicographically ordered dictionary containing 2^p-1 words of length k. This offers a potential savings (compared to worst case complexity of binary search) that asymptotically approaches 50 percent.