The testing effect occurs with both coherent and incoherent text material

Background: For the last decade it has been discussed whether the complexity of learning materials (e.g., of expository texts) posits a boundary condition for the testing effect. One of few studies in favor of this assumption was published by de Jonge and colleagues (2015). They showed that the long-term benefit of testing compared to restudying occurred only for a low-complexity (incoherent, sentences in a scrambled order) but not for a high-complexity (coherent, sentences in an inferential order) version of an expository text. A fill-in-the-blank test served as practice test and dependent variable. Aims: For a number of reasons, we assumed that the testing effect can be found with text of both high and low in complexity. Therefore, we replicated the critical conditions of their Experiments with more power (Experiment 1). In addition, we ran a conceptual replication with short-answer questions as testing procedure (Experiment 2). Sample: In Experiment 1 participated 111 subjects, in Experiment 2 117 subjects. Most of the participants were university students. Method: While de Jonge et al. (2015) manipulated text complexity across two different experiments, each containing a variation in learning condition and final test delay, we manipulated the complexity within one experiment focusing solely on a one-week delay (and also varying the learning condition). Results: As expected, we found a testing effect which was independent from text complexity (and independently from the test type used). Conclusions: Based on these results, we conclude that the testing effect is not restricted to texts low in complexity.