The results reviewed here indicate that the neoHebbian framework has explanatory power in the understanding of memory. A key insight to emerge from physiological experiments is that LTP at short times is relatively insensitive to dopamine, whereas LTP with longer delays is sensitive. Moreover, other properties of LTP (weak/strong; penumbra) appear also to have predictive power in explaining properties of memory.
The idea that long-term memory depends on more than the Hebbian condition makes sense, given the likelihood that too much information written into memory might lead to overwriting (and thus degradation) of pre-existing memories [77
]. According to the neoHebbian model, the default condition is that information is stored automatically in the hippocampus by early LTP based on a local Hebbian process. Generally, this newly stored information will fade away as early LTP declines. Only if a systems-wide computation determines that there is a high level of novelty or motivational salience will dopamine be released, allowing the biochemical processes of late LTP to incorporate the information into long-term memory.
The conditions that lead to activation of the human dopamine system, as judged by fMRI signals, are similar to the conditions leading to activation in non-primate mammals. It therefore seems likely that these principles may apply across all mammals. Nevertheless, experiments that further test the neoHebbian ideas in humans are warranted. An important prediction is that long-term memory would be affected positively (but possibly with an inverted U-shape pattern) by drugs that increase dopamine availability, such as levodopa (L-DOPA), which might enhance the burst-evoked release of dopamine. There has been remarkably little work on the effect of dopaminergic manipulations on human long-term memory. The long-term safety of available dopaminergic drugs is an important problem that needs to be overcome. Alternatively, progress may be made by taking advantage of genetic diversity in the human population with respect to genes that strongly affect dopaminergic function, which has already proven valuable for studying the human reward system (eg. [78
A better understanding of the human dopaminergic system is relevant to several human neurological or psychiatric disorders and to aging. Parkinson’s disease (PD) usually begins with loss of nigrostriatal dopamine cells in the ventral tier SNc, leaving the dorsal tier projection neurons to the hippocampus, ventral striatum, and cortex intact [79
]. It is thus understandable that severe episodic memory problems are not a symptom of early PD [80
]. However, at late stages, when all types of dopaminergic projection neurons are affected, memory problems do become severe. In schizophrenia, the enhanced activation of the hippocampal system may increase dopamine release and skew behavioral choice towards repetitive behavior (see Box 3
). In aging, there is now converging evidence from human imaging studies for a structural degeneration of the VTA/SN (for reviews, see [54
]). Consistent findings have also been recently reported in aged rats [82
]. The possible negative effect of such a decline on the ability of information to enter into long-term memory has not yet been adequately considered.
Understanding the role of dopamine in memory may lead to methods that can be used to improve learning and teaching methods. First, based on the penumbra hypothesis, exposure to novel or salient information could be used to improve the long-term persistence of other information given in temporal proximity. Second, strategies based on the ability of external reward or reward anticipation to release dopamine might prove to be useful in increasing the retention of learned material. Finally, just as with external performance positive feedback, memory persistence might benefit from internally generated reward signals. For instance, the ability to recollect newly acquired information may be intrinsically rewarding. In fact, the study of human learning has revealed an interesting puzzle; long-term retention is not helped by simple re-exposure to recently learned material but is greatly helped by retesting even when subjects already know the answer [83
]. One interesting possibility is that retesting provides an opportunity to generate intrinsic reward signals, thereby enhancing long-term persistence of newly learned material.