sure. Anderson et al. illustrated the 3 mechanism of forgetting. Synaptic plasticity also gives a straight forward answer: if one does not use it, it decays, that is also the underlying cause of the tip of the tongue effect. Connection strength is too weak, so it takes some time to retrieve it. When even more time passes, memory can be entitrely forgotten. But what is more important: memory is NOT static, any retrieval changes the memory. It makes it labil, reconsolidation can rewrite and erase a memory. Here the neural mechanisms have been studied, roughly similar to encoding, a few molecules in the later cascade differ though. A nice illustration is the paper by Monsfils et al.

sure. Anderson et al. illustrated the 3 mechanism of forgetting. Synaptic plasticity also gives a straight forward answer: if one does not use it, it decays, that is also the underlying cause of the tip of the tongue effect. Connection strength is too weak, so it takes some time to retrieve it. When even more time passes, memory can be entitrely forgotten. But what is more important: memory is NOT static, any retrieval changes the memory. It makes it labil, reconsolidation can rewrite and erase a memory. Here the neural mechanisms have been studied, roughly similar to encoding, a few molecules in the later cascade differ though. A nice illustration is the paper by Monsfils et al.

We provided a definition of memory based on circuits of cortical columns. In this case, downstream columns are activated via synaptic connections. In relation to "forgetting," there would be a failure of downstream columns to activate as a result of weakened synaptic connections. However, I think the gist of the question can better be addressed by location of processing and memory in the human brain. We suggested that declarative ("conscious") memory refers to the "verbal Interpreter" (located in the left frontal operculum and the source of internal verbal dialogue) being aware of a memory stored elsewhere in the cortex. This region has no direct connections to the right posterior cortical regions where sensory non-verbal emotional memories are hypothetically stored. This means that there can be a "verbal" lack of awareness when those memories are activated. The attached article discusses these aspects and the referenced 2012 paper also available on my contribution page discusses the columnar model in detail.

Article Psychotherapy and the Brain: The Dimensional Systems Model a...

It seems the observations and discussions in literature fit well with the forgetting mechanisms in STM or the working memory. However, they didn’t explain the forgetting mechanisms of LTM if there is.

Actually, the fundamental question on where the LTM is in the brain was not well investigated. Further, according to the knowledge representation theory in the brain in neuroinformatics [Wang, 2009; Wang et al., 2011; Wang & Fariello, 2012; ICIC, http://www.ucalgary.ca/icic/]], any healthy individual cannot afford to forget things once they are permanently (hard coded) in LTM because many other existing and newly acquired items of knowledge are mutually dependent on them in the network. The loss of any node in the concept network of LTM will result in a damage of a significant portion of the entire knowledge network. That is the nature why there is no built-in forgetting mechanism in LTM, although there is one in the STM.

If you don’t think so, try to consider the following practical question: Can one explain how “1+1=2” in a kid’s LTM may be forgotten intentionally?

I think we have to be careful with the term "forgetting a memory" because it is the source of confusion. It is been used recently too frequently without specifying what do people mean by "forgetting" and "memory". I have to say that I personally do not agree with the current trend.

Forgetting:

do people mean the loss of access to a memory? or they mean complete erasure of a memory?

Memory:

it is important to determine which type of memory, STM/Working memory? or an established LTM?

An established LTM can not be erased completely. it is likely that we have lost access to the memory only. Although literature gives some conflicting findings, majority of the key papers done by key behaviorists show that when an established memory is disrupted or "forgotten" it can be reinstated, recovered by time and re-detected by a reminder or saving tests (Ref: Miller and springer, Science 1990; A Russain article published in 1997 and translated in 1999: Radyushkin and Anokhin, Rossiiskii Fiziologicheskii Zhurnal; Monfils et al., Science 2009; and see comment by Nadel and Land: memory trace revisited).

Another evidence for the lack of forgetting (i.e. erasure) of an established memory comes from AD patients. despite the sever neurodegeneration they still exhibit recovery periods, during which they recall certain memories from the past including recent past.

Thus, I agree with Yingxu Wang, unless you surgically remove it, an established memory can not be forgotten (i.e. erased or completely gone).

We might actively update an established memory and make the access to the new update easier (extinction and cognitive therapy studies). But I do not believe that under such scenario we have erased the old memory.....rather we weakened the access.

It is the STMs and/or working memory that are vulnerable to forgetting (complete erasure). In this case forgetting is the failure to consolidate these memories into established LTMs. So the literal meaning of forgetting (completely gone) applies to STMs.

In our daily life we face many situations indicating that we really have forgotten an established memory. I think the major source of confusion comes from our lack of understanding of STMs/WM and their time windows. We believe that they last from seconds to hours. While we defined LTM as recalling 24hrs later. Thus, when we recall certain information within 24-48 hrs later we think that it is already an established LTM. Based on that when we fail to recall the same memory days or weeks later we think that we forgot the memory. In fact, we can not exclude the possibility that the first recall happened when the memory still was within the STM time window and later the memory failed to get consolidated into an established LTM and thus was forgotten completely.

I think forgetting an LTM you mean interfering it. If so, then it is possible.

Most articles I found are focusing in PTSD and some suggested that erasing can happen in the labile state during awake (not SWS sleep) in which retrieval occurs (http://www.nature.com/neuro/journal/v14/n3/full/nn.2744.html).

Another group showed the time frame of 10 minutes after reactivation of memory for erasing it (http://www.sciencemag.org/content/337/6101/1550.long).

Early this year, it was shown that stress, or corticosterone, have effects to the extinction of memory in reconsolidation (http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0059075).

It is possible to deliberately interfere long-term memory, considering their reconsolidation mechanisms. To achieve this purpose, the different types of memories, differing in their content and the amplitude of their distribution throughout the brain, should be considered. During memory retrieval, memory become labile again and at this precise moment it could be modified or deleted. Usually the reconsolidation process takes place and it becomes long-term memory.

There are evidences in animals showing that immediately after long-term memory evocation, the administration of protein synthesis inhibitors or PKZ blockers are capable of deleting a long-term memory. As mentionated North Akara, there are also evidences in humans showing that the traumatic memory could be interfered.

So, it seems that memory can be deliberately interfered and deleted considering their reconsolidation mechanisms.

There has been some computational learning work done specifically on long term learning that tackles the problem of stability-plasticity. For a brief overview of some of this work please see https://www.researchgate.net/publication/242025424_The_Consolidation_of_Task_Knowledge_for_Lifelong_Machine_Learning

Conference Paper The Consolidation of Task Knowledge for Lifelong Machine Learning

Most of the work on Long-Term-Memory vs Short-Term-Memory, to the best of my knowledge, stems from the studies on hypnosis.

Researchers noticed that people under hypnosis did not remember what they were told during hypnosis but if the memory was triggered afterwards (through the use of "unlikely coincidences" and recall techniques), they remembered it either consciously, as a memory created during hypnosis (and became aware of having been hypnotized), or unconsciously, as a memory with suggestive power.

This lead to the short term/long term dichotomy.

Popular culture in North America such as Hollywood movies and conspiracy theories have spoken of "memory creation" or "mind control".

As far as I'm aware, most reconsolidation literature in humans concerns removing the emotional association attached to certain memories, e.g. using propanolol (Brunet et al. 2008). This doesn't 'delete' a memory - just reduces it's emotional impact. However. a recent study from MIT http://www.pnas.org/content/early/2013/05/16/1218472110.abstract has shown that if you retrieve a 'consolidated' memory, then learn new, interfering, information then this affects later recall of that memory.

As for the neurological mechanisms, according to Larry Squire's 'Standard' model of consolidation, the hippocampus and other medial temporal lobe (MTL) structures initially interact with the neocortex to retrieve memories. Eventually, given enough time, the memories can be retrieved from the neocortex without the aid of the MTL. The hippocampus has also been implicated in reconsolidation, although so far I think this is exclusively based on animal work (http://www.annualreviews.org/doi/full/10.1146/annurev.psych.093008.100523).

See mindspaces.org for information on erasure of the negative emotions mediated in the lateral amygdala. As noted above, a memory is labile when in consciousness and can easily be erased. This is true in the amygdala, but not, apparently in the hippocampus, which is a much more advanced brain module