A transient memory lapse in humans 1–3 h after training

Learning retention, the ability to recall previously learned information, exhibits various patterns in humans: constant performance, decline after a delay, or initial improvement followed by decline or constancy. However, a U-shaped pattern—a transient memory lapse—characterized by a temporary decline in retention followed by improvement, has been observed in numerous non-human species. This phenomenon, often called the "Kamin effect," has been documented across various species (rats, mice, apes, birds, fish, invertebrates) and learning types (avoidance learning, classical conditioning, operant conditioning, extinction learning, discrimination learning, non-associative learning, episodic-like learning, social learning). Proposed explanations for these lapses include transitions between memory phases (short-term, intermediate-term, long-term memory) and temporary retrieval failures due to fluctuating internal cues. This study aimed to investigate whether such transient memory lapses also occur in humans, focusing on perceptual learning, a type of skill learning that differs from previously studied learning types in its extensive training requirements and the nature of the acquired memory (a skill rather than a memory of an experience). The study hypothesized that if transient memory lapses reflect a fundamental aspect of learning, they should be prevalent in humans and occur across diverse learning types.

The Kamin effect, first observed in rats during avoidance learning (Kamin, 1957), demonstrates a U-shaped retention curve where performance is initially high, drops significantly at intermediate times (0.5–6 hours), and then recovers. Subsequent research extensively documented transient memory lapses in various non-human species and across numerous learning paradigms. These lapses often present as slower relearning or a return to pre-training performance levels. The timing of the lapses varies, but typically occurs within tens of minutes to several hours after training. Two primary explanations exist for these lapses: They may reflect transitions between different memory phases (short-term, intermediate-term, and long-term memory), or they may be caused by temporary retrieval failures due to fluctuations in internal cues present during training. Despite strong evidence in non-human animals, research on transient memory lapses in humans has been limited to a few studies and largely ignored by mainstream human learning literature.

This study employed two perceptual discrimination tasks: interaural-level-difference (ILD) and interaural-time-difference (ITD) discrimination. 192 young adults with normal hearing were divided into seven groups. Six groups received 300 trials of training on ILD discrimination, followed by testing on either ILD (to assess learning) or ITD (to assess generalization) at three time points: 30 minutes, 1–3 hours, and 6–10 hours post-training. A seventh control group received no ILD training and was tested on ITD discrimination. Discrimination thresholds were estimated using a three-down-one-up adaptive procedure. Data analysis involved linear mixed-effects regression (LMER) to assess mean improvement, offline learning, offline-learning maintenance, and within-session learning for both tasks. The analyses considered group (time point) and subject ID as fixed and random factors, respectively, with additional within-subject factors (bout or block) depending on the specific analysis.

The results showed a non-monotonic pattern of performance on both tasks. For the trained ILD task, performance was superior to pre-training levels at 30 minutes and 6–10 hours, but significantly lower at 1–3 hours post-training. The magnitude of improvement was greater at the earlier and later time points compared to the 1–3 hour period, confirming a transient memory lapse. For the untrained ITD task, performance at 30 minutes and 6–10 hours post-training was better than the naïve control group, indicating generalization. However, performance at 1–3 hours post-training did not differ from the naïve group, mirroring the lapse observed on the trained task. Analyses of block-by-block performance revealed that the lapse on the trained ILD task manifested as a worsening of performance during the testing session itself, while the lapse on the untrained ITD task resulted in performance consistent with naïve levels. Individual-level analyses revealed that the transient decrease in learning magnitude 1–3 hours post-training was present across a range of initial thresholds for both tasks. Notably, the 1–3-hour time range showed no transition to steadily poor performance, instead representing a transient lapse period itself.

This study provides two examples of transient memory lapses in human perceptual learning, a form of skill learning. The lapses observed at 1–3 hours post-training cannot be attributed to adaptation, habituation, or perceptual deterioration. The different patterns of lapses for trained and untrained tasks suggest that the specific manifestation of the lapse is determined during the testing period rather than during training. The two proposed explanations for transient memory lapses—transition between memory phases and retrieval failure due to mismatched cues—are discussed in the context of these findings. The results suggest a possible transition between different forms of perceptual memory 1–3 hours after training, aligning with the short-term, intermediate-term, and long-term memory phases identified in non-human species. Alternatively, the lapse could indicate a mismatch between internal cues during training and testing. The study’s results highlight the importance of considering transient memory lapses in human learning and suggests the possible involvement of similar memory phases and molecular mechanisms across diverse learning types.

This study provides strong evidence for the existence of transient memory lapses in human perceptual learning, extending the observation of this phenomenon to a new learning type and memory system. The findings support the hypothesis that these lapses are a fundamental aspect of memory formation. Future research should explore the neural mechanisms underlying these lapses and investigate their implications for understanding memory disorders and optimizing learning strategies. Further investigation is needed to determine the prevalence of these lapses across different types of learning tasks and individual differences.

The study focused on a specific perceptual learning task and a limited age range. The generalizability of the findings to other learning types and populations requires further investigation. The relatively small sample sizes at certain time points could limit the statistical power of some analyses. Also, future research should explore whether the lapse is related to incomplete learning of the task.