In this report, we illustrate the considerable impact of researcher degrees of freedom with respect to exclusion of participants in paradigms with a learning element. We illustrate this empirically through case examples from human fear conditioning research, in which the exclusion of ‘non-learners’ and ‘non-responders’ is common – despite a lack of consensus on how to define these groups. We illustrate the substantial heterogeneity in exclusion criteria identified in a systematic literature search and highlight the potential problems and pitfalls of different definitions through case examples based on re-analyses of existing data sets. On the basis of these studies, we propose a consensus on evidence-based rather than idiosyncratic criteria, including clear guidelines on reporting details. Taken together, we illustrate how flexibility in data collection and analysis can be avoided, which will benefit the robustness and replicability of research findings and can be expected to be applicable to other fields of research that involve a learning element.

It has been proposed that memory retrieval can destabilize consolidated memories, after which they need to be reconsolidated in order to be retained. The presentation of relevant information during memory reconsolidation could then result in the modification of a destabilized memory trace, by allowing the memory trace to be updated before being reconsolidated. In line with this idea, Schiller et al. (2010) have demonstrated that memory retrieval shortly before extinction training can prevent the later recovery of conditioned fear responding that is observed after regular extinction training. Those findings have been the subject of considerable controversy, due in part to theoretical reasons but also due to a number of failures to obtain similar results in conceptual replication attempts. Here, we report the results of a highly powered, direct, independent replication of the critical conditions of Schiller et al. (2010, Experiment 1). Due to misrepresentation of the exclusion criteria in the original Schiller et al. (2010) report, data collection was considerably delayed. When we eventually managed to attain our pre-registered sample size, we found that we could not observe any benefit of reactivation-extinction over regular extinction training in preventing recovery of conditioned fear. The results of the present study, along with the mixed findings in the literature and the misreporting in Schiller et al. (2010), give cause to question whether there is robust evidence that reactivation-extinction prevents the return of fear in humans.

Experimental evidence in humans and non-human animals suggests that the administration of propranolol shortly after the retrieval of an emotional memory can lead to an attenuation of its later expression, a phenomenon known as post-reactivation amnesia. Using more potent amnestic drugs, post-reactivation amnesia has been shown in animals to be reversible by re-administration of the drug prior to memory retention testing. The latter finding suggests that, at least under some circumstances, post-reactivation amnesia may not reflect a disruption of reconsolidation (i.e., a memory storage deficit) but an acquired state-dependency of memory expression (i.e., a memory retrieval deficit that is relieved when the drug state is recreated during testing). We conducted a double-blind, placebo-controlled study to investigate whether the previously established amnestic effects of post-reactivation propranolol administration on memory retention in humans may similarly reflect a retrieval deficit. In four groups of participants, fear memories were first established through differential fear conditioning. One day later, a single presentation of the CS+ without shock was used to reactivate the memory in three of the four groups, followed by the administration of 40 mg Propranolol HCl (Groups PrPl and PrPr) or placebo (Group PlPl). Memory was not reactivated in the fourth group (Group NR). Another 24 h later, Propranolol HCl (Group PrPr) or placebo (Groups PrPl, PlPl, and NR) was again administered, followed by a test of memory retention (extinction testing) and recovery (reinstatement testing). We did not observe any effects of post-reactivation propranolol on memory retention; conditioned responding was similar for all groups at the start of retention testing and similarly sensitive to recovery through reinstatement. We did observe an acute effect of propranolol administration on fear-potentiated startle responding during retention testing in Group PrPr, where participants exhibited attenuated startle responses during extinction testing but similar sensitivity to reinstatement as participants in the other groups. While our findings fail to corroborate previous reports of propranolol-induced post-reactivation amnesia in humans, they do point to acute effects of propranolol administration on extinction performance.

The generalization of learned behavior has been extensively investigated, but accounting for variance in generalized responding remains a challenge. Based on recent advances, we demonstrate that the inclusion of perceptual measures in generalization research may lead to a better understanding of both intra-and interindividual differences in generalization. We explore various ways through which perceptual variability can influence generalized responding. We investigate its impact on the ability to discriminate between stimuli and how similarity between stimuli may be variable, rather than fixed, because of it. Subsequently, we argue that perceptual variations can yield different learning experiences and that interindividual differences in generalized responding may be understood from this perspective. Finally, we point to the role of memory and decision-making within this context. Throughout this paper, we argue that accounting for perception in current generalization protocols will improve the precision of obtained generalization gradients and the ability to infer latent mechanisms. This can inspire future attempts to use generalization gradients as a (clinical) predictor or to relate them to individual traits and neural correlates and, ultimately, may lead to new theoretical and clinical insights.

In a highly influential report, Schiller and colleagues (2010) demonstrated long-lasting fear reduction in humans when conducting extinction training shortly following fear memory reactivation. While trying to experimentally replicate the critical conditions of Schiller et al. (2010, Experiment 1), we discovered several irregularities in their paper. Criteria for participant exclusion and the number of excluded participants were misreported; qualitative experimenter decisions actually determined their participant inclusions. Moreover, their statistical analyses were internally inconsistent. After corresponding with the original authors, we received their original data files, allowing us to replicate the reported analyses to verify their results. Here, we report the results of seven separate sets of analyses, three replicating the analyses reported by Schiller et al. (2010) and four applying more principled approaches to participant exclusion, thus including different subsets of the total datasets available, to deduce the influence of specific exclusions and experimenter decisions on the results. For Experiment 1, we were mostly able to replicate the analyses contained in the original report when applying the same qualitative exclusions. However, we found that all of the differences in fear recovery between reactivation-extinction and regular extinction reported by Schiller et al. (2010) were dependent on the qualitative exclusions that they made. With any of the principled approaches to participant exclusion, the degree of fear recovery was highly similar between groups. For Experiment 2, a similar analysis was not possible due to a lack of available data for the excluded participants. Hence, we conducted a verification analysis on the original sample only, which failed to confirm the differences in fear recovery reported by Schiller et al. (2010). Together with the re-analyses, we report a number of additional issues with the way Schiller et al. (2010) processed, analyzed, and reported their data that indicate that their results are unreliable and flawed.

In this report, we use fear conditioning research as an example to illustrate the considerable impact of researcher degrees of freedom with respect to exclusion of participants. In human fear conditioning research, the exclusion of substantial numbers of participants as ‘non-learners’ and ‘non-responders’ is common - despite a lack of consensus on how to define these groups.We illustrate the substantial heterogeneity in exclusion criteria based on a systematic literature search and highlight potential problems and pitfalls of different definitions through case examples based on re-analyses of existing datasets. Based on this, we propose a consensus on evidence-based rather than idiosyncratic criteria for the definition and treatment of ‘non-learners’ and ‘non-responders’ including clear guidelines on reporting details. Taken together, we illustrate how flexibility in data collection and analysis can be avoided in the field of fear conditioning, which will benefit the robustness and replicability of research findings.

Ample evidence suggests that memories enter a labile state upon retrieval, requiring reconsolidation processes in order to be retained. During this period of instability, various interventions can be applied to modify problematic memories. A novel behavioral intervention was designed, aimed at disrupting amygdala-based cognitive processing following the retrieval of a conditioned threat memory, in order to prevent its reconsolidation. We fear-conditioned participants on day 1, and reactivated their memory on day 2. Following reactivation, the reactivation plus emotional working memory task (R + EWMT) group completed an EWMT, while the reactivation only (RO) group served as a no-task control. On day 3, all participants were tested for memory retention, followed by a test for sensitivity to reinstatement. We observed successful acquisition and reactivation in fear-potentiated startle responding, skin conductance responding and US expectancies in both groups. Differential fear responding was fully preserved in the R + EWMT group relative to the RO group at the beginning of retention testing, and both groups were comparably sensitive to reinstatement. Thus, we failed to obtain any evidence that the execution of an EWMT after threat memory reactivation impairs reconsolidation. Further research is indicated to clarify whether threat memory reconsolidation can be disrupted by taxing relevant WM resources.

No abstract