Does Testing Protect Against Retroactive Interference?
Ellen Liu
Washington University in St. Louis
Abstract
Previous studies have found that testing enhances learning. And when it comes to interference, testing has also been shown to protect against proactive interference. However, there has been less consistence in determining the effect of testing on retroactive interference. Therefore, we investigated whether testing could protect against retroactive interference when learning words in foreign languages. We asked participants to learn 20 Lithuanian-English word pairs and manipulated the practice type (retrieval practice vs. restudy) and the post-practice activity (interference vs. no-interference). All participants first studied 20 initial word pairs. They were then assigned to either the interference condition of post-practice activity, where they were asked to learn 20 new word pairs, or the no-interference condition of the post-practice activity, where they were asked to complete an unrelated task. All participants then took a final cued recall test. Our results indicate that there was no significant interaction between practice type and post-practice activity, showing that testing did not, in fact, protect against retroactive interference. These findings can be applied in the context of education, helping educators and students understand that while there are many benefits of frequent testing, testing does not help in retaining previously learned information after learning new information.
Keywords: testing effect, retrieval practice, retroactive interference, interference, learning
Does Testing Protect Against Retroactive Interference?
Oftentimes, learning new information can restrict the capacity to remember older information. For example, students typically have multiple exams over the course of a class. After studying for the second exam, a student is less likely to remember information from the first exam. This difficulty in recalling old information after learning something new is referred to as retroactive interference. In education, one main objective is to decrease interferences in learning in order to expand a student's ability to retain learned information in the long-term. However, this objective of retaining information learned can be applied for everyone throughout daily life. One method that has shown increased memory retention in learning is free-recall testing, known as the testing effect (Roediger & Karpicke, 2006). From research done on the testing effect and research on interferences, the aim in our experiment is to examine whether testing while studying can protect against retroactive interference.
First, the testing effect, when applied to education, suggest that taking a test can be used as a tool to enhance learning and retention (Roediger & Karpicke, 2006). In Roediger and Karpicke's experiments, the study method (either recall test or restudy) and the retention interval (5 minutes, 2 days, or 1 week) were manipulated to test memory and retention. The data showed that for shorter retention intervals, participants who had restudied the material recalled more than those who had taken a test, but participants who had taken the test had recalled more for longer retention intervals. This suggests that the testing study method can be beneficial for long-term retention. They also examined the effects of repeated studying or testing on long-term retention. The data presented similar results as Exp. 1, showing greater forgetting after one week in study methods that had less testing periods. The results of these experiments indicate that testing is a better study method for long-term performance in retaining information, possibly because it acts as practice for thinking in the way one would on a final test.
When looking at learning, there is no doubt that interference plays a role in forgetting information. There are two types of interference: proactive interference and retroactive interference. Proactive interference occurs when previously learned information disrupts the recall of the most recent information learned (Underwood, 1957). Underwood (1957) Conversely, retroactive interference occurs when newly learned information disrupts the recall of previously learned information (Dewar, 2007).
There have been a number of studies conducted in the past few decades on ways to potentially reduce proactive interference. Findings from an experiment conducted by Szpunar, McDermott, and Roediger (2008) suggest that using tests while studying insulates against proactive interference. In their experiment, participants were given vocabulary lists to study, and testing after each list was manipulated (test vs. no test) to test recall. Their overall findings showed that testing previously learned information immediately after learning it can lead to better recall of newer information, protecting against proactive interference. Furthermore, one of their experiments indicated that the benefits of testing against proactive interference can be attributed to retrieval practice, rather than restudying, furthering the Roediger and Karpicke (2006) finding that testing enhances learning.
Studies on the effects of testing on retroactive interference, on the other hand, have not had as consistent of results and it is unclear if testing does guard against this type of interference. In a study conducted by Potts and Shanks (2012), testing was shown to benefit the recall of previously learned information when new information was learned. The experiments in this study tested whether testing previously learned information (i.e. English-Swahili word pairs) could lead to better recall when interfering information (i.e. English-Finnish) was presented immediately after a test was given. The results of the study suggested that participants who were given a reminder test for the English-Swahili word pairs before learning the English-Finnish word pairs had higher recall than participants who were not tested before learning new information.
However, other studies have argued that retrieval practice does not, in fact, protect from retroactive interference (Hupbach, 2014). Results of a study conducted by Hupbach (2014) showed that testing did increase recall of the initial information learned, but when new information interfered with previously learned information, there was a decline in memory performance in all cases (even after retrieval practice). These findings suggest that the effects of retroactive interference outweigh the effect of retrieval practice on memory. Hupbach (2014) proposes that these conflicting results may be due to the strength of the learned information before testing, as the initial information learned was practiced many times before being tested on in the Potts and Shanks (2012) study.
Because of the uncertainties of the effect of testing on retroactive interference, our experiment re-examines whether testing and retrieval practice guard against retroactive interference. We hypothesize that testing does have a diminishing effect on retrieval interference based off of a known testing effect, previously suggested effects of testing on proactive interference, and current literature on this topic. In our experiment, we asked participants to learn Lithuanian-English word pairs obtained from Nelson et. al (2016) and manipulated the practice type (retrieval practice vs. restudy) and a post-practice activity (interference vs. no-interference). All participants first learned Lithuanian-English word pairs using an equal amount both practice types. Afterward, participants were either assigned to the interference condition or the no-interference condition. Those assigned to the interference condition were asked to learn 20 new, interfering Lithuanian-English word pairs, while those assigned to the no-interference condition were asked to complete an unrelated task. Finally, all participants completed a final cued recall test.
Thus, this 2×2 mixed factorial design ultimately had four conditions: retrieval/no-interference (i.e. recall of initial retrieval word pairs when no interference was introduced), retrieval/interference (i.e. recall of initial retrieval word pairs when interference was introduced), restudy/no-interference (i.e. recall of initial restudied word pairs when no interference was introduced), and restudy/interference (i.e. recall of initial restudied word pairs when interference was introduced). We predict that there will be an interaction between retrieval practice and post-practice activity. If testing does in fact protect against retroactive interference as we hypothesize, then we should observe two patterns of results congruent with our prediction. There should be a lower recall in the restudy/interference condition when compared to the restudy/no-interference condition, consistent with previous retroactive interference findings. However, the level of recall should indicate little to no change between the retrieval/no-interference and retrieval/interference conditions, for this would indicate that the effect of retrieval practice outweighs that of interference.
Method
Participants
Thirty-six participants aged 23-63 (Mage = 35.6 and SDage = 9.9 years) were recruited from Amazon's Mechanical Turk, receiving a financial compensation of $4.25. Participants reported having either no familiarity (N=34) or slight familiarity (N=2) with Lithuanian language and were randomly assigned to one of four versions of the experiment, each version comprised of a different combination of conditions and counterbalancing. Random assignment was achieved by having the participants randomly sign up for one of the versions of the experiment, leaving it up to chance which version they signed up for.
Design
The experiment had two independent variables: practice type (IV 1) and post-practice activity (IV 2) manipulated in Phase 1 and Phase 2, respectively. Practice type was measured using two conditions: retrieval practice and restudy practice. Phase 1 had two trials, and all participants underwent both conditions of retrieval practice and restudy practice in each trial. Post-practice activity was also measured using two conditions: interference and no interference. In the interference condition, participants learned new L-E pairs. In the no interference condition, participants answered easy trivia questions.
The experiment had a 2×2 mixed-factorial design. In the first phase, practice type was manipulated within subjects and, in the second phase, post-practice activity was manipulated between subjects. Therefore, the experiment had a total of four conditions: retrieval/no-interference, retrieval/interference, restudy/no-interference, and restudy/interference. The dependent variable was the final cued recall for Phase 1 materials, measured using a final recall test.
Materials
In the experiment, 44 word pairs obtained from Nelson et al. (2016) were used. Two lists of 20 Lithuanian-English (L-E) word pairs were constructed, and four pairs were used as example items during instruction. The first list of 20 word pairs was used during Phase 1 to test practice type, or the method in which a participant studied the vocabulary pairs, in the initial study and practice cycle. To test the two conditions of practice type (retrieval vs. restudy), List 1 was split into two lists of 10 word pairs, List 1a and List 1b, and the pairs in each list served as either retrieval items or restudy items. Practice type was counterbalanced by having List 1a serve as retrieval items and List 1b serve as restudy items for half of the participants, and vice versa for the other half. For retrieval items, participants were asked to type the English word from memory after an initial study. For restudy items, participants were asked to type the English word that is shown.
Phase 2 tested retroactive interference with a post-practice activity of either introducing the participant to 20 new L-E pairs (List 2) in the interference condition or having the participant answer 20 simple trivia questions in the no interference condition. We used simple trivia questions, an unrelated activity, to control for a participant's level of attention, making sure that, while no interfering information was being presented in the no interference condition, each participant in the experiment was still performing a task during phase 2. The trivia questions were made up of the 20 easiest questions out of 100 questions obtained from Weinstein and Roediger (2010) to prevent difficulty of questions from becoming a confounding variable. Therefore, each question asked about a general-knowledge fact, and was answerable with a single word (e.g., a question could read "What is the name of the horse-like animal with black and white stripes?" to which the answer would be "Zebra").
In the brief distractor period between Phase 1 and Phase 2, participants completed a short Big Five Inventory (BFI-S), a 15-item version of the BFI obtained from Lang, John, Lüdtke, Schupp, & Wagner (2011).
The dependent variable, final cued recall for Phase 1 materials, was measured using a final recall test. Scores were calculated by taking the proportion of the times participants answered correctly on the final cued recall test.
Procedure
In Phase 1 of the experiment, participants underwent an initial study and two practice cycles. In the initial study, participants were told that they would study L-E vocab pairs and need to type the English words. After going through four example items, participants studied List 1 (20 L-E word pairs) one at a time and typed the English word, given 5 seconds for each pair.
In Practice Cycle 1, participants were told to either expect restudy trials or retrieval trials. They were given the same list of L-E pairs (i.e. List 1) from the initial study session presented in a random order, one at a time, for 5 seconds each. Half of the pairs were restudy practice trials, where the participant would be given both the Lithuanian and the English translation (e.g., burna – mouth) and was told to type the English word shown. The other half of the pairs were retrieval practice trials, where the participant would only be given the Lithuanian word (e.g. daina – ??) and was told to recall the English word from memory. In both types of trials, the participants were given feedback with the intact pair presented for 3 seconds after each pair. Practice Cycle 2 was identical to Practice Cycle 1, but the pairs were presented in a new random order.
Before the next phase, a short Big Five Inventory (BFI-S) was given to all participants as a brief distractor task for 2 minutes. This task may also act as a way to mask demand characteristics by introducing a task not relevant to the experiment. In Phase 2, post-practice activity was manipulated between subjects. Half of the participants were assigned to the interference condition, where they were told to study 20 new L-E pairs, one at a time, for 5 seconds each, without feedback. They were given both the Lithuanian and the English translation (e.g., burna – mouth) and were told to type the English word shown. Participants went through the list three times, with a new random order each time, for a total of 5 minutes. Presenting new, but similar L-E word pairs as those in List 1 to more accurately test for retroactive interference acted as a manipulation check for IV 2 (post-practice activity). The other half of participants were assigned to the no interference condition, where they were told to answer 20 general knowledge trivia questions. The list of trivia questions had a random order, and the participants were not provided feedback for their answers. They had 15 seconds for each question, and went through the list once, for a total of 5 minutes.
Finally, all participants were given another 3-minute distractor task composed of 15 basic math problems. And in Phase 3, a final cued recall test was given over the 20 L-E word pairs from List 1 (learned in Phase 1). These were all retrieval test trials where the participant would only be given the Lithuanian word (e.g. daina – ??) and told to recall the English word from memory. The final test was conducted one trial at a time for 7 seconds per trial.
In all phases, the experiment controlled for the number of pairs and the length of time each participant had for each pair in each condition. Overall, the experiment took around 25 minutes, after which all participants were provided a short debriefing about the experiment and thanked for participating.
Results
All tests reported were assessed with α = .10 to determine statistical significance.
Figure 1 depicts final test performance for the retrieval/no-interference (M = .817, SD = .134), restudy/no-interference (M = .60, SD = .245), retrieval/interference (M = .65, SD = .275), and restudy/interference (M = .456, SD = .250) conditions. To evaluate these data, a 2×2 mixed factorial ANOVA was conducted, with practice type (retrieval vs. restudy) as a repeated-measures factor and post-practice activity (interference vs. no interference) as a between-subjects factor. The ANOVA revealed no significant interaction and two significant main effects. First, there was no significant interaction between practice type and post-practice activity [F(1, 34) = 0.12, p = 0.731, ⍵2 = 0.0]. Second, participants recalled more words that were practiced via retrieval than words practiced via restudy [i.e., a significant main effect of practice type; F(1, 34) = 41.228, p < .001, ⍵2 = 0.53]. Third, participants recalled fewer words following interference than when there was no interference, [i.e., a significant main effect of post-practice activity; F(1, 34) = 4.86, p =.034 , ⍵2 = .10].
Discussion
In this experiment, we hypothesized that testing has a diminishing effect on retrieval interference based off of a known testing effect, previously suggested effects of testing on proactive interference, and previous literature on this topic. Simple main effects were consistent with previous literature as predicted, revealing that retrieval practice led to higher recall than did restudying and that the presence of interference led to lower recall than having no interference. However, our results indicate that there was no significant interaction between practice type and post-practice activity, showing that testing did not protect against retroactive interference. In fact, there was an effect size of zero, signifying that the difference in recall between retrieval practice and restudying did not change depending on whether there was interference or not.
Limiting Conditions
It is important to note that there was only one experiment conducted in this study, and manipulating some of the control variables (e.g. duration of time between each phase, level of similarity between the initial and interference information, and amount of time spent learning each list) throughout multiple experiments could increase the generalizability of results. Conducting more than one experiment with more variables in this study can create a bigger picture and a more elaborate interpretation of testing's effect on retroactive interference. For instance, manipulating the duration of time between each phase could tell us more about how retention intervals can change how practice type effects post-practice activity. Moreover, the amount of time or amount of trials given in each phase or study period could increase the strength of memory of the information. Like the Roediger and Karpicke (2006b) experiment, the effects of repeated testing could further explain the lack of interaction between practice type and post-practice activity.
Existing Literature
The results from this experiment indicating testing has no effect on retroactive interference are consistent to those of previous studies (Hupbach, 2014), but directly contradict results of Potts and Shanks (2012). In comparison to Hupbach's (2014) study, our study further advances our understanding of testing and interference in the context of word associations and learning in the context of linguistics. The inconsistencies of our results compared to the findings of Potts and Shanks (2012) may be due to the fact that the word pair list we used for our interference (List 2) was in the same language as the initial word pair list (List 1). This can be attributed to previous findings that more similarities between the two lists can increase retroactive inhibition (McGeoch & McDonald, 1931). In Potts and Shanks' study, English-Finnish word pairs were used in the interference condition while English-Swahili word pairs were used in the initial list learned. In this experiment, the similarity in word pairs between the in practice and post-practice phase may have further supported the notion that meaningful relation between information increases retroactive inhibition.
Previous studies have shown that testing and retrieval practice protects against the buildup of proactive interference (Szpunar, McDermott, & Roediger, 2008). However, the results from this study suggest that this may not be the case when it comes to retroactive interference. One possibility for these results may be due to the fact that in proactive interference, competition of information is what prevents recall of memory while retroactive interference is a result of not only competition, but also of "unlearning" (Melton & Irwin, 1940). Therefore, the lack of effect of testing on retroactive interference may be a result of the "unlearning" of information associated with retroactive interference.
Future Directions
An interesting experiment that could potentially shed more light on testing and retroactive interference would be to conduct a similar experiment as this one, but increase the number of conditions in the post-practice activity. Manipulating the similarity between information in the initial study and information in the interference study. For example, instead of only having no-interference and interference conditions, creating conditions on a scale based on degree of interference and meaningful relation between information learned at the different points of the experiment. The finding that meaningful relation increases retroactive inhibition (McGeoch & McDonald, 1931), along with the results indicating that testing protects against retroactive interference in Potts and Shanks' study, where the information learned initially and as interference were dissimilar, may point to an interesting interaction.
Conclusion
Our findings have important implications for future experimentation in the topics of testing and interference, but also for the process of learning. Our results supported the testing effect and its ability to strengthen learning and recalling information. And though our results showed that testing does not protect against retroactive interference, it has served as a helpful baseline for understanding that retroactive interference effects learning and "unlearning" of information. These results can also be applied in the context of education, helping educators and students understand that, while there are many benefits of frequent testing, testing may not help in retaining previously learned information when learning new information. As discussed in the introduction, students take many exams over the course of a class. In many classes, midterm exams are not cumulative, but final exams are. From the findings in this study, students may benefit from adjusting their studying methods to better accommodate for retroactive interference, utilizing a combination of study methods rather than relying merely on testing.
References
Dewar, M. T., Cowan, N., & Sala, S. D. (2007). Forgetting Due to Retroactive Interference: A Fusion of Müller and Pilzeckers (1900) Early Insights into Everyday Forgetting and Recent Research on Anterograde Amnesia. Cortex,43(5), 616-634. doi:10.1016/s0010-9452(08)70492-1
Hupbach, A. (2015). Retrieval practice does not safeguard memories from interference-based forgetting. Learning and Motivation,49, 23-30. doi:10.1016/j.lmot.2015.01.004
Lang, F. R., John, D., Lüdtke, O., Schupp, J., & Wagner, G. G. (2011). Short assessment of the Big Five: robust across survey methods except telephone interviewing. Behavior Research Methods,43(2), 548-567. doi:10.3758/s13428-011-0066-z
Mcgeoch, J. A., & Mcdonald, W. T. (1931). Meaningful relation and retroactive inhibition. The American Journal of Psychology,43(4), 579. doi:10.2307/1415159
Melton, A. W., & Irwin, J. M. (1940). The influence of degree of interpolated learning on retroactive inhibition and the overt transfer of specific responses. S.l.: S.n.
Nelson, S. M., Savalia, N. K., Fishell, A. K., Gilmore, A. W., Zou, F., Balota, D. A., & Mcdermott, K. B. (2016). Default mode network activity predicts early memory decline in healthy young adults aged 18-31. Cerebral Cortex,26(8), 3379-3389. doi:10.1093/cercor/bhv165
Potts, R., & Shanks, D. R. (2012). Can testing immunize memories against interference? Journal of Experimental Psychology: Learning, Memory, and Cognition,38(6), 1780-1785. doi:10.1037/a0028218
Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention . Psychological Science,17(3), 249-255. doi:10.1111/j.1467-9280.2006.01693.x
Szpunar, K. K., Mcdermott, K. B., & Roediger, H. L. (2008). Testing during study insulates against the buildup of proactive interference. Journal of Experimental Psychology: Learning, Memory, and Cognition,34(6), 1392-1399. doi:10.1037/a0013082
Underwood, B. J. (1957). Interference and forgetting. Psychological Review,64(1), 49-60. doi:10.1037/h0044616