Jurchiș Răzvan, Dienes Zoltan
Cognitive Psychology Laboratory, Department of Psychology, Babeș-Bolyai University, Cluj-Napoca, Romania.
School of Psychology and Sackler Centre for Consciousness Science, University of Sussex, Brighton, UK.
Psychon Bull Rev. 2023 Feb;30(1):269-279. doi: 10.3758/s13423-022-02175-0. Epub 2022 Sep 9.
The existence of implicit (unconscious) learning has been demonstrated in several laboratory paradigms. Researchers have also suggested that it plays a role in complex real-life human activities. For instance, in social situations, we may follow unconscious behaviour scripts or intuitively anticipate the reaction of familiar persons based on nonconscious cues. Still, it is difficult to make inferences about the involvement of implicit learning in realistic contexts, given that this phenomenon has been demonstrated, almost exclusively, using simple artificial stimuli (e.g., learning structured patterns of letters). In addition, recent analyses show that the amount of unconscious knowledge learned in these tasks has been overestimated by random measurement error. To overcome these limitations, we adapted the artificial grammar learning (AGL) task, and exposed participants (N = 93), in virtual reality, to a realistic agent that executed combinations of boxing punches. Unknown to participants, the combinations were structured by a complex artificial grammar. In a subsequent test phase, participants accurately discriminated novel grammatical from nongrammatical combinations, showing they had acquired the grammar. For measuring awareness, we used trial-by-trial subjective scales, and an analytical method that accounts for the possible overestimation of unconscious knowledge due to regression to the mean. These methods conjointly showed strong evidence for implicit and for explicit learning. The present study is the first to show that humans can implicitly learn, in VR, knowledge regarding realistic body movements, and, further, that implicit knowledge extracted in AGL is robust when accounting for its possible inflation by random measurement error.
内隐（无意识）学习的存在已在多个实验室范式中得到证实。研究人员还提出，它在复杂的现实人类活动中发挥作用。例如，在社交场合中，我们可能会遵循无意识的行为脚本，或者根据无意识的线索直观地预测熟悉的人的反应。然而，鉴于这一现象几乎完全是通过简单的人工刺激（例如，学习字母的结构化模式）得到证实的，所以很难推断内隐学习在现实情境中的参与情况。此外，最近的分析表明，在这些任务中通过随机测量误差高估了学到的无意识知识的数量。为了克服这些局限性，我们改编了人工语法学习（AGL）任务，并在虚拟现实中让参与者（N = 93）接触一个执行拳击组合动作的逼真智能体。参与者不知道的是，这些组合是由一种复杂的人工语法构建的。在随后的测试阶段，参与者能够准确地区分新的符合语法和不符合语法的组合，表明他们已经掌握了该语法。为了测量意识，我们使用了逐次试验的主观量表，以及一种考虑到由于均值回归可能高估无意识知识的分析方法。这些方法共同有力地证明了内隐学习和外显学习的存在。本研究首次表明，人类能够在虚拟现实中内隐地学习关于现实身体动作的知识，而且，当考虑到随机测量误差可能导致的膨胀时，AGL中提取的内隐知识是可靠的。
