Information overload usually refers to a situation in which a decision-maker's performance is harmed by too much input information. Experiments have shown that better decisions can occur when selected information is withheld from the decision-maker. This paper suggests that the notion of information overload can apply also to purely mechanical processes. It is presumed that a paired-comparison measurement procedure which attempts to detect as many ordered pairs as possible in a set which is linearly ordered by some unidimensional attribute can identify only part of the underlying linear order L. This part is in the form of a partial order P which is a subset of L. An algorithm can then be applied to P in an attempt to identify ordered pairs in L which are not in P. Simulation experiments have shown that significantly more than 50 percent of the ordered pairs added to P by the construction process of the algorithm agree with their orders in L. However, it has also been discovered that the construction process is expected to yield better results (in a sense specified precisely in the paper) when its input is taken to be a subset of P which is obtained by removing selected ordered pairs from P. Thus, by withholding certain known and correct information from the input to the construction process, a better output is expected.
