The contextual interference (CI) model hypothesizes random practice (high CI) leads to inferior acquisition, but superior retention and transfer capabilities compared to blocked practice (low CI). These phenomena are well established in laboratory settings. However, the transfer to applied settings, particularly in sports practice is still under discussion. The aim of this systematic review and meta-analysis is to evaluate the possible generalisability of the CI phenomenon in sports-based contexts with regard to performance outcomes. Up to April 16th, 2022, five electronic databases (PubMed, Web of Science, Scopus, Taylor & Francis, and SciELO) were used to search for relevant studies investigating differences between blocked and random schedules at post-acquisition, retention, and/or transfer tests. Using the PICOS criteria, a total of 933 records were screened. Outcomes evaluating the acquisition, retention, and/or transfer performances in sports skills in healthy participants were included. The quality of the selected studies was scored using the PEDro scale. Thirty-seven studies were selected, of which twenty-seven were of good quality and the remaining ten were rated as fair quality. Out of 205 overall pooled outcomes, only 43 performance outcomes (21%) agreed with the CI phenomenon after the acquisition (18 out of 103), retention (19 out of 84), and/or transfer (6 out of 36) phases. No statistically significant overall difference between blocked and random practice was detected at post-acquisition (effect size (ES) = 0.1, p = 0.154), retention (ES = −0.159, p = 0.141) or transfer testing (ES = −0.243, p = 0.071). The subgroup analysis showed inferior acquisition and superior retention following random practice only in individuals aged 20–24 years (ES = 0.282, p = 0.030 during the acquisition, and ES = −0.405, p = 0.011 during retention), with no difference at transfer testing for this specific age group. No similar significant effects have been concurrently identified during both acquisition and retention phases in any of the remaining subgroups (e.g., examined based on experience level, sports, skills, and testing protocol categories). The present results suggest CI effects can only be confirmed under very limited conditions, which seriously challenges the extension of CI effects to the sport context in general. Problematic conclusions for children's learning are discussed, as well as future research strategies to better understand counterintuitive learning approaches.

情境干扰 (CI) 模型假设随机练习（高 CI）导致较差的习得，但与阻塞练习（低 CI）相比，保留和转移能力更强。这些现象在实验室环境中得到了很好的证实。然而，转移到应用设置，特别是在运动实践中仍在讨论中。本系统回顾和荟萃分析的目的是评估 CI 现象在基于运动的背景下对绩效结果的可能普遍性。截至 2022 年 4 月 16 日，五个电子数据库（PubMed、Web of Science、Scopus、Taylor & Francis 和 SciELO）被用于搜索调查在获取后、保留和/或传输测试。使用 PICOS 标准，共筛选出933条记录。包括评估健康参与者运动技能的获得、保留和/或转移表现的结果。使用 PEDro 量表对所选研究的质量进行评分。选择了 37 项研究，其中 27 项质量良好，其余 10 项被评为质量一般。在 205 个总体汇总结果中，只有 43 个绩效结果 (21%) 在收购（103 个中的 18 个）、保留（84 个中的 19 个）和/或转移（36 个中的 6 个）阶段后同意 CI 现象。在获得后（效应大小（使用 PEDro 量表对所选研究的质量进行评分。选择了 37 项研究，其中 27 项质量良好，其余 10 项被评为质量一般。在 205 个总体汇总结果中，只有 43 个绩效结果 (21%) 在收购（103 个中的 18 个）、保留（84 个中的 19 个）和/或转移（36 个中的 6 个）阶段后同意 CI 现象。在获得后（效应大小（使用 PEDro 量表对所选研究的质量进行评分。选择了 37 项研究，其中 27 项质量良好，其余 10 项被评为质量一般。在 205 个总体汇总结果中，只有 43 个绩效结果 (21%) 在收购（103 个中的 18 个）、保留（84 个中的 19 个）和/或转移（36 个中的 6 个）阶段后同意 CI 现象。在获得后（效应大小（在收购（103 个中有 18 个）、保留（84 个中有 19 个）和/或转移（36 个中有 6 个）阶段后，只有 43 个绩效结果 (21%) 同意 CI 现象。在获得后（效应大小（在收购（103 个中有 18 个）、保留（84 个中有 19 个）和/或转移（36 个中有 6 个）阶段后，只有 43 个绩效结果 (21%) 同意 CI 现象。在获得后（效应大小（ES)  = 0.1，p  = 0.154)，保留 ( ES  = −0.159，p  = 0.141) 或转移测试 ( ES  = −0.243，p  =  0.071 )。亚组分析显示仅在 20-24 岁的个体中随机练习后获得较差的获得和较高的保留（ES  = 0.282，p  = 0.030 在获得期间，ES  = -0.405，p = 0.011 在保留期间），在这个特定年龄组的转移测试中没有差异。在任何剩余的子组（例如，根据经验水平、运动、技能和测试协议类别进行检查）中，在获取和保留阶段都没有同时发现类似的显着影响。目前的结果表明 CI 效果只能在非常有限的条件下得到确认，这严重挑战了 CI 效果在一般体育环境中的扩展。讨论了儿童学习的有问题的结论，以及未来的研究策略，以更好地理解反直觉的学习方法。