Investigating a Dual-Process Account of Learning, Memory and Cognition.

他的项目旨在回答是否有令人信服的证据表明人类存在联想系统的问题，并以此解决更大的理论性争论，即单过程（完全命题，见Mitchell， De Houwer, & Lovibond, 2009）还是学习和记忆的双过程（见McLaren等人，2014）是人类认知的更好模型。这种争论并不局限于学习，因为双过程理论已经被用来解释许多认知功能，如推理和决策（Evans & Stanovich, 2013）。这个问题的答案将产生真正的后果。Lovibond（参见2018年参考文献中的一个例子）认为，临床治疗应该减少对行为部分的强调，转而关注更具认知性的“谈话”部分，这是基于对联想过程对人类精神生活的贡献产生怀疑的证据。鉴于此，该项目不仅有潜力为科学文献做出重大贡献，而且还有助于我们将对学习的理解应用于焦虑和恐惧症等心理健康问题，并为教育实践提供信息。我的方法将是同时使用经验和计算建模的角度来研究人类的峰移等现象。在经验方面，我将采用行为和神经科学方法，如脑电图（EEG）和经颅直流电刺激（tDCS）。脑电图可以用来检查伴随任何影响的神经特征，但本质上是相关的。然而，使用tDCS的神经刺激将允许我测试每个过程（联想和命题）是否可以独立调节，并使我能够推断任何影响的因果机制。这些经验技术将与计算模型相结合，从而分析控制学习的具体过程和方程，以及刺激和对象如何在大脑中呈现。这可以用来建立在现有的联想（例如，McLaren和Mackintosh， 2002）和认知（例如，Lee， Hayes和Lovibond， 2018）模型的基础上。一个有助于区分双过程理论和单过程理论的研究例子是一个测试两个维度的研究。在本研究中，亮度（从暗到亮，用X表示）和颜色（从蓝到绿，用Y表示）将构成两个维度。因此，刺激值表示为（X，Y），范围从深蓝色（1,1）到浅绿色（11,11），中点为中等亮蓝绿色（6,6）。参与者最初将接受两种刺激——3,5和9,7——的训练，这两种刺激分别与左和右两个类别有关。由于这些刺激明显比另一个更亮或更暗，这将导致归纳出控制亮度维度的规则（3和9）。然后，他们将接受另外两种刺激——5,5和7,7——的训练，这两种刺激的颜色与最初的训练刺激相同，但亮度和黑暗程度没有那么极端。这有助于进一步建立亮度维度的规则，也将确保足够的联想强度积累到颜色维度。如果参与者关注亮度维度，因为这在训练中最明显相关，那么当在这个维度（刺激颜色保持不变）上进行测试时，双过程解释将预测基于规则的反应。此外，当参与者在颜色维度上进行测试时，这个解释也可以预测峰值移位。参与者随后将接受采访，以确定他们是否使用了任何规则或注意到刺激和类别之间的任何关系。本实验也将进行计算模拟，并与模型结果进行比较。

his project seeks to answer the question of whether there is compelling evidence for an associative system in people, and in doing so resolve the larger ongoing theoretical debate of whether single process (entirely propositional, see Mitchell, De Houwer, & Lovibond, 2009) or dual-process accounts of learning and memory (see McLaren et al., 2014) are the better model for human cognition. This debate is not limited to learning, as dual-processes accounts have been used to explain many cognitive functions such as reasoning and decision making (Evans & Stanovich, 2013). An answer to this question would have real consequences. Lovibond (see 2018 reference for an example) has argued that clinical therapies should de-emphasise the behavioural component and focus instead on the more cognitive, "talking" part based on evidence casting doubt on the contribution made by associative processes to human mental life. Given this, the project has not only the potential to significantly contribute to the scientific literature, but to also help us apply our understanding of learning to mental health issues such as anxiety and phobias, as well as informing educational practice. My approach will be to use both an empirical and a computational modelling perspective to study phenomena such as peak shift in humans. Empirically, I will employ both behavioural and neuroscientific methods, such as electroencephalography (EEG) and transcranial direct current stimulation (tDCS). EEG can be used to examine the neural signatures that accompany any effect but is essentially correlational in nature. Neurostimulation using tDCS, however, will allow me to test whether each process (associative and propositional) can be independently modulated and enable me to infer the causal mechanisms responsible for any effect. These empirical techniques will be combined with computational modelling that will allow analysis of the specific processes and equations that govern learning, as well as how stimuli and objects are represented in the mind. This can be used to build upon existing associative (e.g., McLaren & Mackintosh, 2002) and cognitive (e.g., Lee, Hayes and Lovibond, 2018) models.An example of a study that would help distinguish between dual and single-process theories is one in which two dimensions are tested. In this study, both brightness (ranging from dark to bright, denoted by X) and colour (ranging from blue to green, denoted by Y) would constitute the two dimensions. Henceforth, the stimulus values will be represented as (X,Y) and range from dark blue (1,1) to light green (11,11), with a midpoint of medium bright blueish-green (6,6). Participants would initially be trained with two stimuli - 3,5 and 9,7 - that are associated with two categories, left and right, respectively. As these stimuli are obviously brighter and darker than one another, this will lead to induction of a rule governing the brightness dimension (3 and 9). They will then be trained with two more stimuli - 5,5 and 7,7 - which are the same colour as the initial training stimuli, but not as extreme in their brightness and darkness. This serves to further establish a rule on the brightness dimension and will also ensure sufficient associative strength has accumulated to the colour dimension. If participants are attending to the brightness dimension, as this was most clearly relevant during training, a dual-process account would predict rule based responding when tested on this dimension (where the stimulus colour remains constant). Furthermore, this account would also predict peak shift when participants are tested on the colour dimension. Participants would be interviewed after to ascertain if they were using any rules or noticed any relationships between the stimuli and categories. This experiment would also be computationally simulated, and the results would be compared with that of the model.