An empirical assessment of a linear-stochastic perspective for Canadian macroeconomic time series is presented. The methods used are based on the mathematics of 'chaos'. Present evidence suggests that low-order deterministic chaos does not provide a satisfactory characterization of the data. The absence of significant nonlinear structure for the investment and unemployment series is of particular note in light of past findings with American data. The degree to which the use of a time trend can impose a pseudo-structure on the data is illustrated.