action in step function not changing

hi - i had expected the action parameter in the step function (reinforcement learning) to at least change reasonably frequently for a discrete action set of [-1 0 1] but I found that it does not. The signature of the step function is

function [nextObs,reward,isDone,loggedSignals] = stepFunction(action,loggedSignals)

This behaviour is true for both rlPGAgent and rlACAgent. Is there a flag that I need to set somewhere?