Understanding how people remember and why they forget is a perennial research topic in psychology. A classic finding is that recalling information from memory is known to increase the probability of being able to retrieve that information again later, and strengthens long-term memory. This benefit has been shown to persist for months after the initial recall attempt. This phenomena is known as the testing effect. More recently, there have been attempts to apply this research finding in educational contexts.
Although decades of research may intuitively suggest that there is enough information to apply this testing effect in practical contexts, it turns out that the structure of the materials that are tested may have an affect on the retrieval process.
In most classic studies of the effect, participants learn lists of words. After an initial reading of the list, one group will then be asked to recall the words that they can remember. Another group of participants will be allowed to reread the list. Generally, the group that is asked to recall the words from memory remembers more words later than the group that reread the word list.
One issue with this procedure that may limit classroom application is that, unlike more educational material, the words on these lists have typically been unrelated and didn’t allow any higher-level reasoning. Lists of unrelated words contrast sharply with complex prose passages, wherein each sentence and idea relates to the others in an organized manner.
While there have been some recent findings that the testing effect works on complex materials, such as passages of text, the results have been mixed. This has highlighted some important issues. For instance, how does testing reinforce the relationships between the items to be learned? Do items that are related need to be tested together in order to improve memory for that relation?
My research aims to determine how the complexity of information interacts with the testing effect, and how the particular sequencing of units of information (such as ideas in a passage of text) may increase (or decrease) memory for their relations. Additionally, since effective education is more about helping students to draw inferences than memorize facts, my dependent measures will require participants to draw inferences based on the information they learned. For example, if a participant learned “Students can commute by walking, biking, or driving” and “Freshmen are not allowed to drive.” certain conclusions result from relating these sentences. A test question such as “How can freshman commute to school?” would require relating the first two sentences in order to be answered correctly.
Generous funding provided by the Dartmouth Graduate Student Graduate Alumni Research Award allowed me to collect preliminary data to investigate whether and how the testing effect influenced inferential ability. Prior research has shown that while testing does indeed improve memory of such sentences, it does not necessarily improve inferential ability related to the information in those sentences. I first replicated this original finding, and then altered the implementation of the testing to attempt to determine the limiting factor. My subsequent experiments have found that testing can indeed improve inferential ability as well as memory, but that the testing procedure must allow relational processing of sentences that may need to be used together on later inferences.
One limitation of these findings is that it is difficult to quantify the relations between such sentences, which precludes calculation of exactly how difficult one inference is relative to another. In future work I plan to switch from using sentences to using spatial relations (e.g. House is west of Supermarket; Supermarket is north of Post office). This will allow me to precisely quantify the relative difficulty of an inference, because the difficulty of the inference “Where is the house relative to the post office?” can measured by counting the number of locations between them. Additionally, I will be able to manipulate the order in which these spatial relations are presented, which may shed light on how sequencing and testing are related. Thanks to this research award I aim to determine the optimal conditions under which the testing effect can be used for learning educationally relevant materials.
