We deal with frequency computations in polynomial time, or more general with resource bounded
frequency computations. We investigate the first non-trivial case of the Hinrichs-Wechsung conjecture,
which states that as soon as we have at least *2 ^{d+d}* inputs to be queried, it does not
become harder to get an answer with at most

*d*errors, if we increase the number of inputs to be queried. This conjecture can easily be seen to hold for cases

*d*less than

*3*, and it seems very hard to prove in general. We solve the problem affirmatively in the case

*d=3*.