The benefit of testing on the retention of verbal materials has been studied quite extensively, however very little attention has been put on establishing when the best time to actually apply a test is. The present study investigated the effect of testing (immediate and delayed) on the learning of novel words in English as a foreign language (EFL). The participants were 20 students of EFL enrolled on a 5-year teaching programme. They learned the meaning of 20 matched novel words presented with images, sentences, and exercises during learning. The experiment took place over a week. On day 1 participants learned a list of 10 words, and a day later (day 2), they learned another set of 10 words and were then immediately tested on all 20 words. On day 8, participants were tested again on all the words they had learned. A semantic categorisation task was used for the purpose, consisting of classifying newly learned words into living or nonliving things. The results showed that participants classified more accurately and responded faster to newly learned words tested immediately after training than words learned a day earlier, and these effects were stable over time. These results can be explained by interference theories or by the alternative retrieval route theory because when testing is applied immediately after learning, it acts as an instant shield that protects newly learned words from interference, or strengthens their retrieval routes.