In a small-scale New-Keynesian model with a hybrid Phillips curve and IS equation, the paper is concerned with an arbitrary frequency of the agents' synchronized decision making. It investigates the validity of a fundamental methodological precept according to which no substantive prediction or explanation of a well-defined macroeconomic period model should depend on the real time length of the period. While this principle is basically satisfied as the period goes to zero, the impulse-response functions of the high-frequency versions can qualitatively as well as quantitatively be fairly dissimilar from their quarterly counterpart. The result proves to be robust under variations of the degree of price stickiness. The main conclusion is that DSGE modelling may be more sensitive to its choice of the agents’ decision interval.