The instruction frequencies for the 4 data sets are plotted in Figure 8.
To assess the relationship between instruction frequency and selective pressure, we compare the instruction frequencies of the terminal taxa for the same set of 4 selective pressures.
However, the instruction frequencies for the three Avida data sets that experience natural selection are clearly not uniform.
When comparing the instruction frequencies, we see very little difference between the two from the overlapping and non-overlapping reward structures despite the significant difference in MP's reconstruction accuracy.
The instruction frequencies for nand and IO, both of which are critical for performing Boolean logic tasks, are also significantly above 3.86% in the overlapping and non-overlapping reward structure data sets where there is a selective pressure to perform Boolean logic tasks.
Given that the instruction set size is 26, the average instruction frequency should be 3.86%.
We then illustrate how selective pressures cause changes in both ASRV and instruction frequency (which is related to the character substitution model).
Using socio-historical constructivist theory to guide instruction interventions, instruction frequency and questioning were intentionally manipulated during one-half of each course.
We use instruction frequency as a proxy for accepted substitution percentage because instruction frequencies are easier to interpret given the 26 instructions in the Avida instruction set.
To help determine which factor, ASRV or non-uniform accepted substitution probabilities, is the more influential byproduct of natural selection on the performance of maximum parsimony, we plot the ASRV and instruction frequencies that result from a variety of selective pressures.
