Investigating the Relationship Between Statistical Learning and Rule-Switching Flexibility: An Individual Differences Approach with Behavioral and fMRI Evidence

NCU Institutional Repository > 生醫理工學院 > 認知與神經科學研究所 > 博碩士論文 > Item 987654321/98471

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: https://ir.lib.ncu.edu.tw/handle/987654321/98471

题名:	Investigating the Relationship Between Statistical Learning and Rule-Switching Flexibility: An Individual Differences Approach with Behavioral and fMRI Evidence
作者:	詹巧恩;Chan, Chiao-En
贡献者:	認知與神經科學研究所
关键词:	統計學習;規則轉換;功能性磁振造影;Statistical learning;Rule-switching flexibility;fMRI
日期:	2025-08-25
上传时间:	2025-10-17 12:49:03 (UTC+8)
出版者:	國立中央大學
摘要:	統計學習與規則轉換彈性是人類適應環境的兩項核心認知能力。前者使我們能從經驗中歸納出環境的規律，進而穩定地應對外在世界；後者則賦予我們在環境變化時調整策略的靈活性。理想上，一位高效的學習者應同時具備優異的規則提取能力與靈活的策略轉換能力。然而，若從認知資源有限的觀點出發，當資源集中投入於某一歷程時，可能會相對削弱對另一歷程的支持。因此，統計學習與規則轉換兩種能力之間是否存在交互作用，以及其神經基礎為何，仍是有待釐清的重要問題。本論文透過一系列行為實驗（實驗一與實驗二）以及一項功能性磁振造影研究（實驗三），系統性地探討此議題。實驗一（含1a與1b）整合三項經典統計學習作業以全面測量受試者的統計學習能力，包括視覺統計學習作業（VSL）、天氣預測作業（WP）以及連續性雙結構序列反應時間任務（SDS-SRTT），並使用體系轉移偵測作業（RSDT）作為規則轉換敏感度的指標。在實驗二中，我們將 SDS-SRTT 改為「交錯式雙結構序列反應時間典範」（IDS-SRTT），使舊規則的消失與新規則的出現變成一個連續漸進的過程，從而更真實地模擬動態環境下的規則改變。行為結果一致顯示，在動作序列學習的框架中，統計學習能力與規則轉換彈性呈現出明顯的權衡關係（trade-off）：個體在統計學習上表現越佳，反而在面對規則轉換時表現出較低的彈性，暗示兩者可能競爭相同的認知資源。為進一步探索其神經基礎，實驗三透過功能性磁振造影觀察受試者在 IDS-SRTT 作業中的腦部活動。我們在學習單一規則階段比較結構序列與隨機序列的大腦反應，發現在面對結構序列時，皮質運動區與小腦的活化下降，反映了學習序列後神經連結的效率提升；同時，額頂葉網絡（FPN）與警覺網絡（SN）在學習期間也展現相同的活化反應。相對地，在進入新規則的轉換階段，FPN、SN的活化顯著上升，反映其在偵測並適應新規則時扮演積極角色。其中，警覺網絡在轉換初期高度參與以偵測新訊號，而額頂葉網絡則提供持續支持以建構與穩定新目標，顯現出清晰的動態分工。整體而言，本論文不僅在行為層面揭示統計學習與規則轉換彈性之間的潛在權衡關係，更在神經層次上提出整合性觀點：一套共通的認知控制系統，能夠透過調整其功能狀態，在「穩定學習」與「彈性適應」兩種需求之間取得平衡。這套以額頂葉與警覺網絡為核心的動態適應機制，為理解如「資訊覓食」（Information Foraging）等更複雜的適應性行為，提供了重要的神經科學基礎。 ;Statistical learning (SL) and rule-switching flexibility are two core cognitive abilities essential for adapting to the environment. While SL allows for the efficient extraction of regularities, flexibility enables individuals to adapt when those regularities change. Ideally, an effective learner should not only acquire new regularities efficiently but also maintain a high degree of flexibility to switch when the situation changes. However, when cognitive resources are predominantly allocated to one ability, this may come at the expense of another. Despite their theoretical importance, the relationship between SL and rule-switching flexibility has rarely been systematically investigated. To address this gap, a series of experiments employing the serial reaction time task (SRTT) in combination with the functional magnetic resonance imaging (fMRI) technique in Experiment 3 are conducted. Specifically, Experiments 1a and 1b employed a comprehensive battery of conventional SL paradigms, including a triplet segmentation task of geometric shapes in the visual modality (VSL), a weather prediction task (WP), and a sequential-dual-structure serial reaction time task (SDS-SRTT), alongside a regime-shift detection task (RSDT) to assess sensitivity to environmental changes. Experiment 2 introduced a novel interleaved-dual-structure serial reaction time task (IDS-SRTT), which better simulated the gradual emergence and replacement of regularities. Across eight sessions, participants were exposed to a shift from an initial structure (Structure 1) to a novel one (Structure 2). Behavioral results across Experiment 1 and 2 consistently revealed a negative correlation between SL performance and rule-switching flexibility in the motor domain, suggesting a potential trade-off between learning stability and adaptive flexibility. In other words, heightened sensitivity to repeated patterns may, in fact, reduce the flexibility necessary for adapting to new regularities. In Experiment 3, we employed fMRI to identify the neural substrates associated with SL and rule-switching flexibility in the IDS-SRTT paradigm. During SL, reduced activation in the cortical motor areas and cerebellum was observed when comparing structure to random trials, consistent with the notion of neural efficiency following successful learning. Additionally, the frontoparietal network (FPN) and the salience network (SN) also showed reduced activation during learning of structure. In contrast, during the transition to a new structure, both FPN and SN exhibited increased activation, reflecting their involvement in adapting to rule changes. Notably, SN engagement diminished across the switching phase, whereas FPN activity remained consistently elevated, suggesting a shift from conflict detection to sustained goal-directed control. In sum, this thesis elucidates the behavioral trade-off between SL and rule-switching flexibility and, critically, presents an integrated neural account of how the brain navigates these competing demands. Our findings indicate that a shared cognitive system, centered on the FPN and SN, dynamically reconfigures its functional role. This neural flexibility allows the system to shift between modes of ′exploitation′, efficiently executing learned patterns, and ′exploration′, searching and adapting to new information, thereby managing the inherent tension between stability and change. This model of a dynamic, common control system provides a novel and foundational framework for investigating complex adaptive behaviors, from skill acquisition to information foraging in uncertain worlds.
显示于类别:	[認知與神經科學研究所 ] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	3	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....