人工语法学习中意识加工的渐进发展

The Development of Graded Consciousness in Artificial Grammar Learning

内隐学习研究致力于探讨学习活动的意识加工程度,当前相关的研究逻辑从传统的意识二分观转向渐进意识假设。本研究通过对人工语法范式进行改造,并借鉴加工分离程序的有关思想,在学习阶段引入双重测量任务,并根据双重任务成绩计算分离出相应学习时段的意识与无意识成分的贡献分数,从而考察在内隐学习过程中两种加工成分的动态变化特征。研究结果支持渐进意识假设,在学习进程中,无意识成分和意识成分都呈现出渐进发展的趋势;而随学习的深入,两种成分呈现出不同的变化模式,中后期意识加工快速增长,无意识加工则保持平缓发展,渐进意识系统整体呈现向外显学习推进的发展态势。

Consciousness has always been much concerned in cognition science and, with the discovery of implicit learning in artificial grammar learning（AGL） researches, influences of unconscious processes on human cognition have been unprecedentedly highlighted. Dozens of empirical investigations have distinguished two different types of learning, i.e. explicit learning and implicit learning which involve conscious and unconscious processing respectively. Hence their critical attributes and interactive patterns have undergone tentative explorations and several theoretical frameworks have been proposed to demonstrate the underlying mental mechanisms, most of which take the side of dualistic logic. Empirical data tend to indicate that, rather than stand in dichotomy, so often they co-exist only with a quantitative difference. The academic dilemma is now broken through by the graded consciousness hypothesis, as a result of which new perspective arises from the graded consciousness dimension to deeply investigate implicit learning. In the present research the dynamic mental evolution pattern has been explored in artificial grammar learning, concerning both consciousness and knowledge representation. The current research adopts hypothesis of distributive representation and representation rehearsal which indicates that the ever-optimizing distributive representation dominates the learning process. Thus graded consciousness comes into form due to the increasing contribution of conscious processing. A dual-task design was introduced in current study with reference to the PDP paradigm formerly adopted in implicit memory researches. The innovative paradigm adopted a ‘slow＇ learning task and a ‘quick＇ task and the contribution patterns of consciousness and unconsciousness are different in the two tasks, as a result of which the contributions of conscious and unconscious processes could be extracted dynamically along the learning course. This innovative paradigm also makes possible a direct investigation of graded consciousness during the learning phase. The results of current study show that, in the implicit phase of artificial grammar learning, the contribution of conscious processing exhibited a slow-first-fast-later growing pattern while a slow-first-stable-later pattern was found with unconscious processing. In the beginning unconscious contribution prevailed but was eventually exceeded by conscious process in the subsequent blocks. However, in the first half of learning phase, consciousness contribution suffered an undulating performance while unconsciousness more stable. While entering the second half of learning phase, unconsciousness contribution exhibited a stagnating pattern and the uneven mode faded with consciousness process. These findings, compatible with those empirical researches assuming dichotomy logic, not only define distinguishable features for implicit learning and explicit learning, but also demonstrate the graded consciousness in artificial grammar learning and make possible a learning continuum progressing throughout the learning phase of artificial grammar. In the current study which assumed graded consciousness perspective the structures of knowledge representations have also been discussed indirectly. Conclusions are that the developing of unconscious representations precedes that of conscious representations, considering the undulation in the first half phase for consciousness and the stagnation in the second half phase for unconsciousness, which indicates a possible synergic effect between conscious and unconscious processing.