Hello Eoin,

if I understand right, you measured performance on the same objects/persons of that 2 groups for 5 days. If so, then you have to use repeated measures ANOVA (1-way, only for treatment), because the 5-day measurements are correlated if measured on the same objects. Did you?

(2-way ANOVA stays for independent data samples)

Best regards,

Learning is a non-linear process and ANOVA is the least appropriate model.

An appropriate model would be a non-linear multi-level model, using the exponential law of training [1]. This model estimates three parameters per participant, that reflect different aspects of individual learning:

1. previous experience

2. rate of learning

3. maximum reachable performance in the long run

From that on you can sharpen your research question (what aspect is improved? do people learn faster? better in the long run?) and add your treatment factor to the respective parameter.

I'm working a lot with these models in the field of surgery training. If you are interested in collaboration, drop me a note.

--Martin

[1] Heathcote, a, Brown, S., & Mewhort, D. J. (2000). The power law repealed: the case for an exponential law of practice. Psychonomic Bulletin & Review, 7(2), 185–207.

A mixed effects ANOVA model implies some factors are fixed effects and some are random effects. Which factor is fixed effects and which random effects?

A significant interaction effect will obscure main effects if not removed. I therefore support the testing for the interaction between general learning and drug effect. This will make the testing of the main general learning effect and the main drug effect more precise.

When asking a difficult question as this one, it may be advisable to give some data to let people interested in check some features before giving you an answer (or trying to ...).