I've posted on memory a few different times, but thus far I've shied away from going into great molecular detail; in fact, I've pretty much avoided molecular and cellular neuroscience altogether on this "blog." This sidestepping results, to be honest, from laziness; it is easier to make gambling and ventriloquism widely appealing than it is to spice up intracellular mechanisms like gene regulation and protein folding, although I believe the latter two are actually quite intriguing and wholly relevant to understanding the mind.
Glossing over molecular details is actually somewhat at odds with my attitude towards neuroscience, a field which has appealed to me since middle school because it links causal, physical mechanisms with delightfully wondrous things like memorizing pieces of music (the actual moment of inspiration occurred while I was playing the piano). Since then, I have been fascinated by the idea that the biology--proteins, molecules, genes etc--of individual cells is directly related to the complexities of human thought, from kicking a soccer ball and catching a dodgeball to learning a language and dreaming.
In the days since middle school, however, I've come to realize that by attempting to bridge molecules to behavior, neuroscience is both marvelously exciting and incredibly problematic. In between these two levels are, in increasing levels of organization: the cellular, the intercellular (synaptic), the circuit (networks/pathways), the regional (e.g. fMRI studies), and the systems (e.g. motor systems), and a wide range of inter-level hierarchies upon which I won't begin to touch. Because of the enormous distance one must travel from specific molecules to the human mind, many cognitive neuroscientists dismiss "reductionism" as analyzing mechanisms which are too far removed from behavior to be directly relevant; they believe each level must be bridged before making any larger connections.
I agree that the mind cannot be understood by looking solely at the simplest biological components, but I also feel that knowledge of neural networks, etc., is meaningless unless we understand the biological basis. In other words, the cellular approach is necessary, but not sufficient, for understanding the brain. Most cognitive processes, in particular memory formation, have much to gain from molecular and cellular analyses.
Memory formation is endlessly fascinating on all levels. Conceptually, memory is (to quote Eric Kandel), "a form of mental time travel [which] frees us from the constraints of time and space"; mechanistically, it is a result of the brain's ability to embody, retain, and modify information in neural circuits. To further define 'memories' using neuronally (i.e. biologically)-relevant vocabulary, it is helpful to distinguish it from closely-related 'knowledge' and 'learning.'
'Knowledge,' in neuronal terms, is the perceived world converted into a neuronal form; it exists as "internal representations." These representations issue from the activity and connectivity of neurons (forming an assembly of neurons: a 'neural circuit'), and is thus inextricably linked to the biological properties of those neurons, particularly of their functional interconnections (i.e. synapses).
'Learning' is then the experience-dependent creation or modification of these internal representations; i.e. changes in the way the neurons are connected to each other in specific circuits, particularly the strength of their synapses ('synaptic plasticity'). 'Memory' is thus the retention of the aforementioned experience-dependent modifications. The salient idea is that specific biological properties must be altered, (in particular, those of the synapse) in order for memory to be established; moreover, these properties are products of universal cellular and molecular mechanisms that are employed throughout the body and the living world.
(So what are these cellular and molecular mechanisms? For those in need of some scientific background information on neuronal communication, this site is quite clear and comprehensive, or for a briefer version I've given a summary here.)
And now, leaping and bounding back through the conceptual hierarchy of neuroscience, these biological processes are linked to functional changes occurring within neural circuits, which latter ultimately guide behavior. Thus, although reductionist techniques attempt to experimentally link molecules directly to behavior, the overarching theoretical goal involves bridges between and amongst all levels. Modern molecular techniques are all the more powerful when combined with other levels of analyses: after intervening at the cellular or molecular level, well-accepted psychological and behavioral paradigms can be employed to determine whether a given biological process is correlated with (or necessary, or perhaps even sufficient for) the occurrence of a behavioral phenomenon.
One elegant example of the power of reductionism (which motivated this post, in particular the somewhat lyrical wax of an introduction) was just published online in Nature Neuroscience. The study, carried out by a group from UT Southwestern, manipulated a neuronal protein to assess its role in learning and memory, thus attempting to bridge the behavioral and the molecular pathway levels directly. I will go into more detail on the paper, and the neurobiology of memory, in the following post.