Clive Shepherd writes, "I'm going to take some convincing that Kirkpatrick isn't a reasonable starting point for assessing effectiveness, both qualitative and quantitative." And yet, he notes, "in practice, we don't do that much of it - according to an ASTD survey, 91% of organisations evaluate at level 1 (reactions), 54% at level 2 (learning), 23% at level 3 (behaviour change) and 8% at level 4 (impact on results)" (Level 5, ROI, as he correctly notes, was added later). The problem with Kirkpatrick's measure - or any similar measure, for that matter - is that it is trying to measure a complex phenomenon as a linear result. A change in behaviour associated with learning, for example, may be the result of a perceived change in threat level from management, and have nothing to do with the learning. You don't get at such things with simple measures. Related: James Kariuki, ROI of e-learning.