Conrad M
J Neurosci Res. 1976;2(3):233-54. doi: 10.1002/jnr.490020306.
This paper presents a theory of memory and memory mediated learning based on the manipulation of macromolecular conformations. The main features of the theory are: 1) the brain contains primary and reference neurons; 2) inputs from the external environment produce particular patterns of primary firing; 3) the firing of a primary neuron sensitizes certain of its dendrites; 4) the sensitized primaries are loaded by the reference neuron active at the time and in such a way that they fire when called by this reference neuron, thus reconstructing the original pattern of primary activity. The reference neurons may also be loaded by primaries, thus making it possible for the reconstruction process to be initiated by some feature of the initial input. Each reference neuron loads and calls at most one primary pattern of activity, thereby preventing superposition of memories. If the primaries are loadable by sequences of impulses, this makes it possible to increase the connectivity among the various types of neurons by using party-line organization. The loading and calling processes themselves are mediated by call molecules. These are allosteric enzymes, located in the dendrites of primary and reference neurons, whose states are set either by an impulse or sequence of impulses and which catalyze events leading to impulse formation whenever this input recurs. The call molecules are capable of duplicating their setting (or conformation) using either intra- or interneuronal potentials, thereby ensuring stability of the memory trace. The theory allows for general powers of memory manipulation (by rememorization), for the construction of time ordered, content ordered, and associative data structures, and for computation with global representations of the environment. It makes a large number of testable predictions, provides a natural interpretation for the structure of the cerebral cortex, and accounts for: resistance to cooling, differential effects of chemical agents on short and long term memory, distributed character of memory, accessibility of memories by stimulation of specific brain loci, and also the details of classical conditioning and instrumental learning.
本文提出了一种基于大分子构象操纵的记忆及记忆介导学习理论。该理论的主要特点如下：1）大脑包含初级神经元和参考神经元；2）来自外部环境的输入产生特定的初级放电模式；3）初级神经元的放电会使其某些树突敏感化；4）被激活的参考神经元会对敏感化的初级神经元进行加载，使得当被该参考神经元调用时它们能够放电，从而重建初级活动的原始模式。参考神经元也可以由初级神经元加载，这样就使得重建过程能够由初始输入的某些特征启动。每个参考神经元最多加载并调用一种初级活动模式，从而防止记忆叠加。如果初级神经元可由脉冲序列加载，那么通过使用共用线路组织就有可能增加各类神经元之间的连接性。加载和调用过程本身由调用分子介导。这些是变构酶，位于初级神经元和参考神经元的树突中，其状态由一个脉冲或脉冲序列设定，并且每当该输入再次出现时，它们会催化导致脉冲形成的事件。调用分子能够利用神经元内或神经元间的电位复制其设定（或构象），从而确保记忆痕迹的稳定性。该理论允许进行一般的记忆操纵能力（通过重新记忆），构建按时间顺序、内容顺序和关联的数据结构，以及使用环境的全局表示进行计算。它做出了大量可测试的预测，为大脑皮层的结构提供了自然的解释，并解释了：对冷却的抗性、化学试剂对短期和长期记忆的不同影响、记忆的分布式特征、通过刺激特定脑区来获取记忆，以及经典条件反射和工具性学习的细节。
