Elizabeth,

I think there might be two distinct questions here. First, what is the minimum # of strides for a subject to achieve steady state walking, and second, from a statistical perspective, how many data are required to accurately quantify a given temporal parameter.

The first question is usually answered using judgment or perhaps by convenience, however one paper has actually looked into this:

http://www.ncbi.nlm.nih.gov/pubmed/20362453

You might need to answer the second question experimentally. 'Steady-state' could mean different things depending on which variables you are studying and how repeatable this variable is from step-to-step. Subjects walking can actually exhibit quite a bit of variability in their gait, particularly if they are asked to do something novel like wearing different shoes, or walking on a treadmill. You could measure your variable of interest over several different time periods and then statistically assess how many steps are required for a given precision.

The answer to when steady state is achieved may depend on the population being studied in addition to the type of footwear.

http://www.ncbi.nlm.nih.gov/pubmed/17383185

This paper also discusses how the number of strides influences variability of gait speed and stride length.

Thank you Sean and Eric for the great responses and links to literature.