Replicability of representational similarity and its role in successful memory retrieval

S62 and S63 for a project registered on OSF (remaining data and files are found there) Abstract: Studies of hippocampal pattern similarity during event encoding and its relationship to subsequent memory retrieval have revealed inconsistent results. Our lab recently found evidence that differences in cognitive processing during encoding can modulate the relationship between hippocampal pattern similarity and recognition success (Lim et al., 2023). This finding is consistent with the theoretical proposal that hippocampal representations have a dynamic relationship to memory retrieval in which cognitive goals are influential (Brunec et al., 2020). However, there have been few attempts to replicate representational similarity findings from fMRI and, to our knowledge, no evidence of successful replication in either the hippocampus or subsequent memory studies. In order to draw strong theoretical conclusions from our findings, or others in the literature, it is important to demonstrate that those findings are robust. The current study attempted a direct replication of our prior experiment with the exception of minor modifications in the neuroimaging parameters, which were intended to assess the degree to which representational similarity analyses are influenced by reasonable technical differences in data collection. We did not replicate the finding that cognitive variability interacts with future recognition success in the hippocampus. Overall, we failed to replicate nine out of twelve significant F-test results from the Lim et al. study. The three findings that were replicated can be explained by minor visual differences in stimulus presentation on variable cognitive context trials. In addition, we found three new significant effects in the current study that did not previously appear in our earlier study. Therefore, we conclude that fMRI studies using representational similarity analysis of subsequent memory performance are sensitive to minor methodological variation. Further tests of the replicability of these findings are needed prior to drawing theoretical conclusions from their results. Preregistered Stage 1 protocol: https://osf.io/njzhq (date of in-principle acceptance: 08/01/2025)本项目相关的S62与S63数据集已在开放科学框架（Open Science Framework, OSF）注册，其余数据与文件可于该平台获取。 摘要：针对事件编码阶段海马模式相似性（hippocampal pattern similarity）及其与后续记忆提取的关联的研究，所得结果并不一致。本课题组近期发现，编码阶段的认知加工差异可调节海马模式相似性与再认成绩之间的关联（Lim等，2023）。该发现与一项理论假说相符：海马表征与记忆提取存在动态关联，其中认知目标发挥着重要影响（Brunec等，2020）。然而，目前鲜有研究尝试重复功能磁共振成像（functional magnetic resonance imaging, fMRI）中的表征相似性研究结果，且据我们所知，尚无研究在海马体或后续记忆相关研究中证实此类重复成功的证据。为了从本团队及已有文献的研究结果中得出可靠的理论结论，验证这些结果的稳健性至关重要。本研究尝试对先前的实验进行直接重复，仅对神经影像学参数进行了小幅调整，旨在评估数据采集过程中合理的技术差异对表征相似性分析的影响程度。本研究未能重复“认知变异性与海马体中未来再认成绩存在交互作用”这一发现。整体而言，我们未能重复Lim等（2023）研究中12项显著F检验结果中的9项。得以重复的3项研究结果，可归因于可变认知情境试次中刺激呈现存在小幅视觉差异。此外，本研究还发现了3项此前研究中未出现的新的显著效应。因此，我们得出结论：采用表征相似性分析探究后续记忆表现的功能磁共振成像研究，对方法学上的细微差异十分敏感。在基于此类研究结果得出理论结论之前，仍需对这些发现的可重复性开展进一步验证。已预注册的第一阶段实验方案：https://osf.io/njzhq（原则性接受日期：2025年1月8日）