The hippocampus via the trisynaptic circuit serves to act as a comparator of old and new data.  Old information travels from the entorhinal cortex (at ECII) via the perforant path to the dentate on to CA3 and final is passed on unchanged to CA1 from which it is sent via the subiculum back to EC.  But, if there is new information then CA3 is inhibited while new information from primary sensory cortices is sent via association cortices to ECIII which projects to CA1 where this new information drives the firing of CA1 pyramidal cells which in turn send this revised input signal back to EC via the subiculum.  The next time the information enters CA3 from EC it will have been modified to reflect what was recently seen or heard.

If the activity of this circuit is dysregulated as occurs with age-related amnestic mild cognitive impairment wherein there is loss of inhibitory control over CA3 subregion then learning and memory function is impaired.  

Check out our paper:

http://www.ncbi.nlm.nih.gov/pubmed/25941121

From my point of view, there is no doubt that the hippocampus store information on a long term basis, at least several years in human, as demonstrated by retrograde amnesia of such duration (2 to 10 years) in patient suffering from hippocampal lesion or dysfunction.The fact that hippocampal LTP could last at least one year in the rat (that is more or less half of the animal life span), comes in support of the view that hippocampus has such abilities (Abraham et al, 2002, J Neurosci).

Its pivotal role as a promoter of information storage in the neocortex is obviously well known, as it has been the classical view of the hippocampus role since Scoville and Milner description of patient Henry Molaison anterograde amnesia (see the fantastic book "Permanent Present Tense", from Suzanne Corkin, for the whole story of HM). In accordance with a crucial involvement of the hippocampus in early learning phase,  our team showed a spectacular increase of activity in the dentate gyrus (probably as the result of a processing through a hippocampal-entorhinal loop) when rats learn an associative olfactory learning (Truchet et al, 2002, Hippocampus, http://www.ncbi.nlm.nih.gov/pubmed/12440576).

Fortunately, more and more colleagues now consider the involvement of the hippocampus in retrieval as well (see for instance Suzuki et al, 2011, Hippocampus, or Rugg and Vilberg, 2013, Curr Opi Neurobio for a review), and I have some difficulties to imagine a retrieval process that will purely and simply bypass the hippocampus complex.

Hippocampal LTP merely increases the probability that CA1 fires upon stimulation by CA3 thereby overriding the influence of novel extraneous stimuli.  But I do not believe this is actually where the data we think of as a memory is is stored per se.  Higher order cognitive domains must be reactivated for us to remember what something sounds of looks like.  The observation that LTP occurs in other brain regions supports this thesis.

For example see:

http://www.ncbi.nlm.nih.gov/pubmed/21123576

Thanks for your comment, Marcia,

Indeed, LTP can be induced in many brain areas, including neocortices, but the fact that it could last at least one year in the DG is a strong indication that hippocampus is equipped to store information for longer periods of time than needed just for allowing storage at a neocortical level. Don't get me wrong, though : it is very unlikely that the long term storage of a complex information, say an episodic event, rely only on hippocampal formation, but surely it is involved, even for retrieval.

The observation that blocking glutamate mediated LTD in the CA1 subregion interferes with consolidation of memories supports the importance of facilitating information flow through this pathway.   I just do not want anyone to think that they could loose their language function for example by destroying the hippocampus.  The fact that HM could still speak and engage in a conversation is just one example of acquired information being "stored and retrieved" from brain regions outside of the hippocampus. But on the contrary this previously learned information could not be reconsolidated without the hippocampus  which is why he kept forgetting his doctors names.  Like I said, the hippocampus is important for encoding, consolidating and reconsolidating of memories but it is not where the information is stored.  

Well, when it comes to HM, things are not so simple as the classical view "no hippocampus = anterograde amnesia".

First, from what I remember from Corkin's book, HM could remember with all details only two episodes of its life before surgery. One of them was a plane flight, a highly emotional event, which is not surprising.Second, HM autopsy showed that the hippocampus removal was less extensive than expected, and a large part of its memory deficit could maybe be attributed to lesion of the entorhinal cortex :

http://www.nature.com/ncomms/2014/140128/ncomms4122/full/ncomms4122.html

About langage, of course you are right :), but lesions to hippocampal formation or, worse, medial temporal lobe, could affect semantic memory and therefore, communications skills, don't you think ?

Semantics requires "comparing" one word to another and choosing which word to use thus, it can be a dynamic process.  Not the same as encoding/consolidating a name per se for the first time.  Anomia which is an early sign of AD might lead one to think that names are stored in the hippocampus when in reality there is a failure of the comparator to assign the correct name to the item being seen.  It is not that the memory of what a harmonica for example looks like has been lost per se but rather that the stored memory for the word cannot be related to the image currently being seen.  Thus, when the comparator fails the dynamic aspects of learning and memory function (e.g., encoding/consolidation/reconsolidation) are disrupted.

Part of the problems with current knowledge of the hippocampus is that our models of how it works with the brain are too simplistic. We can literally see that the terms of storage in the hippocampus are less than the terms of amnesia especially retrograde amnesia reported for H.M. This suggests that there is something going on beside storage in the hippocampus, but our models just suggest that the hippocampus is somehow rewriting the cortex memory, not how exactly this works. I have seen evidence that there is a 2 year development pattern in children's learning that might correspond to the 2 year retrograde amnesia seen in H.M.  if DG LTP is only 1 year, where is this second year coming from?

careful : one year LTp is in the rat, we have no idea, for obvious technical/ethical reasons, how it could last in human.

About your comment, Marcia : of course, I'm not saying that semantic information is stored only, or at all,  in the hippocampus. I am just convinced that you need the hippocampus for optimal retrieval for most memories. It doesn't mean that hippocampus is necessary to retrieve all kind of complex information, but we need it to perform at our best.

Well Bruno, you might be right, 1 year is only 1/2 of what I am proposing, but if Karmelloff - Smith's 2 year latency in language learning in childhood is correct, it might indicate that human LTP runs in a 2 year pattern. H.M. would seem to corroborate this. This might suggest that in fact there is a latency in learning while memory is kept in the hippocampus. The hippocampus may act as a substrate for reprogramming the cortex memory after the 2 year latency.

Sorry that's Karmiloff-Smith not Karmelloff-Smith.

I have attached a new brief article explaining how our cortical column theory (Dimensional Systems Model) explains hippocampal involvement in memory that was described in the Moss et al. (2012) and Moss & Moss (2014a &b) journal articles.

Article The hippocampus and memory: The binding of parallel cortical circuits.

AIMS Neuroscience just published an article on the hippocampus and creativity

http://www.aimspress.com/article/10.3934/Neuroscience.2016.1.105

Dear Robert,

As to discuss the role of the hippocampus, I should rather prefer to avoid the ontological perspective suggesting that storage and association might be dissociated. They are intricate as the two faces of a single coin. Hence I should prefer a review paper  linking these conjoined twins!

yours sincerely,

Françoise

This was just published at AIMS Neuroscience in response to the question.

Article A Theory on the Singular Function of the Hippocampus: Facili...