认知与发展心理学研究室知识库

人们对环境中的序列信息进行编码，或者从正在进行的动作中学习正确的序列，是适应环境的一种非常重要的能力。研究者发现，序列学习也可以是一种无意识的、偶然的、程序性的学习过程。在内隐序列学习的研究中，序列反应时范式是使用最为广泛的研究范式。多年来，研究者们采用口头报告的主观标准或者再认测验等客观标准，分离人们在内隐学习中获得的意识和无意识知识。然而，这些方法面临敏感一一污染的问题，如主观测验可能会因低估人们获得的无意识知识，而对意识知识不敏感，而客观测验则会因为无意识知识对成绩的贡献，而高估意识知识。有研究者提出了“意识知识可控，无意识知识不可控”这一逻辑方法，由此发展出了过程分离程序和赌博任务。然而，研究者们采用过程分离程序得到的实验结果并不一致，有关序列学习与意识的关系仍不清楚。此外，有关人们在序列学习中获得知识的类型，如具体知识(如组块)还是抽象知识(如类别)或者运动知识还是知觉知识，还存在争议。

以上有关序列学习的研究，刺激多是在阈上呈现的，人们对刺激的知觉是有意识的，只是人们对刺激间关系的知识是无意识的。为了探讨人们是否可以对阈下刺激进行序列学习，本研究将双眼竞争的实验范式引入序列学习任务中，深入探讨阈下序列学习的可能性、知识类型和神经机制。研究一考察阈下知觉序列学习的无意识性。实验1采用阈下序列学习范式(CFS范式)和阈上序列学习范式(Non-CFS范式)考察知觉序列知识是否能够无意识的获得，结果发现，CFS组与非CFS组被试都能够获得三联体结构的无意识的序列知识。实验2采用短时阂下序列呈现的方法考察知觉序列学习中获得的知识类型(具体、抽象)和特点(一阶、二阶)，结果发现，阈下序列学习组被试只能学习到具体知识的一阶序列，而阈上序列学习组被试能够学习到具体序列的一阶和二阶序列，但是两组被试都没有获得抽象序列知识。实验3采用长时阈下序列呈现的方法考察长时间的阈下与阈上序列学习中习得知识的类型。结果发现，延长阈下序列学习的时间，被试仍只能获得获得一阶和二阶的具体序列知识，并不能获得抽象序列知识。

研究二考察运动序列和知觉序列在阂下序列学习中的作用。本研究将突破连续闪烁抑制范式(b-CFS)与经典的序列反应时任务(SRT)结合，将突破抑制反应时作为因变量分别考察阈下运动序列学习和阈下知觉序列学习的无意识知识的动态变化。在阈下运动序列学习中，要求被试对所看到的具体刺激按相应键进行反应，运动序列中包含了运动和知觉的信息输入，二者都可促进整体运动序列的学习;在阂下知觉序列学习中，要求被试对所有刺激都按空格键，人们只能进行知觉序列的学习。结果发现，随着学习的进行，两组被试的反应时均显著下降，并且当刺激遵循的二阶序列变为不同的二阶序列时其反应时显著增加，而当学习的二阶序列再度呈现时被试的反应时又显著下降，说明被试在两种条件下都学到了二阶序列知识。并且，两种条件下的序列学习成绩之间没有显著差异，说明纯粹的知觉学习在阈下序列学习中起主要作用。

研究三采用脑电实验考察阈下知觉序列学习的时间进程。实验6考察了学习序列改变引起的b-CFS组和非b-CFS组脑电成分的变化。行为结果发现，只有b-CFS组获得了知觉序列知识。ERP结果发现，学习序列的改变引起b-CFS组前中部P3波幅的增加，但没有引起非b-CFS组波形的显著变化;与此相一致，时频分析结果发现，学习序列的改变只在b-CFS组引起了p和Y能量的显著增加。这些结果表明，只有阈下条件下进行的序列学习才能够引起P3波幅与p和α能量的变化。

本研究系统考察了阈下序列学习的可能性、知识类型及神经机制，发现人们在阈下序列学习中只能习得具体的一阶和二阶知识，知觉序列在阈下序列学习中起主要作用，并且阈下与阈上序列学习具有不同的神经机制。这些研究发现有助于加深对阈下序列学习的联接强度及知识类型的理解，并为揭示序列学习中无意识知识产生的神经机制提供了启示。

It is a very important ability to adapt to the environment by coding sequence information or learning the correct sequence from ongoing actions. Researchers found that sequence learning can also be an unconscious, incidental and procedural learning process. In the study of implicit sequence learning, the Sequence Reaction Time paradigm is the most widely used. Over the years, researchers have used objective criteria such as subjective criteria for oral reporting or recognizing tests to isolate implicit and explicit knowledge gained in implicit learning. However, these methods might be polluted, such as the main observations may underestimate the unconscious knowledge that people obtain, while the objective observations may overestimate the conscious knowledge. Some researchers have proposed a logical method of "conscious knowledge is controllable and unconscious knowledge is uncontrollable". Process Dissociation Procedure and gambling tasks have been developed. However, the experimental results obtained by the researchers using the process dissociation procedure are not consistent, and the relationship between sequence learning and consciousness is still unclear. In addition, there is controversy about the type of knowledge people have in sequence learning, such as specific knowledge (such as chunks) or abstract knowledge (such as categories) or motor knowledge or perception knowledge.

In the above studies on sequence learning, the stimuli are mostly presented on the conscious level. People aware of perception of stimuli, but their knowledge of the relationship between stimuli is unconscious. In order to explore the possibility of subliminal sequence learning, this study combines the experimental paradigm of sequence learning and binocular rivalry, and deeply explores the types of knowledge obtained in subliminal sequence learning. Study 1 examined the unconsciousness of subliminal perceptual sequence learning. Experiment 1 used the subliminal sequence learning paradigm (CFS paradigm) and the non-CFS paradigm to investigate whether the perceptual sequence knowledge can be obtained unconsciously. It was found that both the CFS group and the non-CFS group can obtain the triplet structures unconsciously. Experiment 2 used the method of short-time subliminal sequence presentation to investigate the types of knowledge (specific, abstract) and characteristics (first-order, second-order) obtained in the learning of perceptual sequence. It was found that the subjects can acquire the specific sequence instead of abstract sequence. There was no significant difference in the response of the rule sequence to the irregular sequence. Experiment 3 used long-term subliminal sequence representation to investigate whether longer CFS learning and non-CFS learning can obtain abstract unconscious knowledge. It was found that after a long-term subliminal sequence learning, the subjects obtained the first-order and second-order specific sequence knowledge, but did not obtain abstract sequence knowledge.

The first study contains 3 experiments. We investigated unconscious perceptual sequence learning by using continuous flash suppression (CFS) paradigm. The results of Experiment 1 showed that visual sequence learning can be unconsciousness in comparison to non-CFS condition. The results of Experiment 2 showed that the detection latencies were shorter when the targets followed FOC and SOC structures than not in the visible condition, while the detection latencies were shorter only when the targets followed the FOC structure than not in the invisible condition. The results of Experiment 3 showed that after lengthened the familiarization phase, the detection latencies were shorter for targets following FOC and SOC structures than not even in the invisible condition. However, there were no any significant learning effects for reversal structures in both conditions. The results indicated that the chunk-based rather than abstract-based structures could be acquired in subliminal visual perceptual sequence learning.

Study 2 examined the role of motion sequences and perceptual sequences in subliminal sequence learning. This study combined breaking continuous flash suppression paradigm (b-CFS) with the classical sequence reaction task (SRT), and investigated the dynamic changes of unconscious knowledge in subliminal motor sequence learning and perceptual sequence learning using breaking time as the dependent variable. Experiment 4 required the subjects to respond to the corresponding key according to the specific stimulus. Both the motor sequence and the perceptual sequence can facilitate sequence learning. As a result, it was found that the breaking time decreased significantly as the learning progressed, and increased significantly when the second-order training sequence became another second-order sequence. The breaking time decreased significantly when the training sequence was presented again indicating that subjects learned the knowledge of the second-order training sequence. Experiment 5 required participants to press the same button for all stimuli that people can only learn perceptual sequence. As a result, the same reaction time change pattern as in Experiment 4 was found, indicating that subjects learned a pure second-order perceptual sequence. Moreover, there was no significant difference between motor sequence learning and perceptual sequence learning in Experiment 4 and Experiment 5, indicating that pure perceptual learning plays a major role in subliminal sequence learning.

Study 3 used EEG experiments to investigate the temporal progression of subliminal perceptual sequence learning. Experiment 6 examined the changes in brain electrical components of the b-CFS group and the non-b-CFS group caused by changes in learning sequences. The behavioral results showed that only b-CFS group learned unconscious sequence Knowledge. The ERP results showed that subliminal sequence learning caused increase of the P3 amplitude in the frontal central area in b-CFS group, but did not cause significant changes in the waveform of the non-b-CFS group. In consistent with this, the time-frequency analysis found that the learning sequence caused a significant increase in beta and gamma energy only in the b-CFS group. These results indicated that subliminal sequence learning does have a different neural mechanism than non-subliminal sequence learning.

This study systematically investigated the possibility of subliminal sequence learning and revealed the type of knowledge type and neural mechanism. It was found that people can only acquire specific first-order and second-order knowledge in subliminal sequence learning. Perceptual sequences play a major role in subliminal sequence learning. Subliminal sequence learning has different neural mechanisms with unconscious sequence learning. These findings help to deepen the understanding of the strength and type of knowledge of subliminal sequence learning, and provide insights into the neural mechanisms involved in sequence learning and unconscious knowledge.