The "Cardassian Expansion Scenario" was recently proposed by Freese & Lewis as an alternative to a cosmological constant in explaining the current accelerating universe. In this paper we investigate observational constraints on this scenario from recent measurements of the angular size of high-z compact radio sources compiled by Gurvits and coworkers. We show that the allowed intervals for n and zeq, the two parameters of the Cardassian model, are heavily dependent on the value of the mean projected linear size l. However, the best fit to the current angular size data prefers the conventional flat cold dark matter model to this Cardassian expansion proposal, though the latter is cosmologically credible and compatible with the-z diagramfor some values of l.