memory 2

We all have encounters, where we meet someone, are sure we have met the person before,
but have no clue where or when we met them, or what his/her name is. This is an example of
facial memory and of familiarity. Recall the study where subjects were asked to judge whether
a word represented an animate object, or whether the object was large or small. Upon the
recollection stage they were required to judge whether they had seen the word before
(episodic memory), whether it was printed in green or red (source memory) and how confident
they were about their judgement. Confidence judgements are related to familiarity. Sorting
fMRI data relative to confidence (familiarity) judgments showed that neither the hippocampus
nor the parahippocampus were activated. However, the entorhinal cortex showed significant
activation which covaried with confidence judgements. This study shows a double
dissociation: activation in the hippocampus and parahippocampus that correlates with
retrieval of episodic (source) memory, and activation of the entorhinal cortex correlating with
familiarity judgements.
Is the hippocampus the storage site for long term episodic memory?
While some of the previously described studies could suggest that long term episodic memory
is stored in the hippocampus, the findings from patient H.M. (and from other patients) argue
against this idea, as their amnesia was largely anterograde, less so retrograde. fMRI studies
again provided valuable insight into this question. To answer which areas are active during
memory retrieval subjects were shown a word (e.g. ‘Bell’) and either an associated picture or
a sound which they should remember. 2 days afterwards they were shown various words and
were asked to retrieve the associated picture or sound (whichever was paired with the word).
Retrieval of auditory information activated higher order auditory cortical areas, while retrieval
of visual information activated higher order visual areas. Thus, it appears that items stored in
long term memory are stored (at least partly) in those areas which originally encode them, it
may be similar to reactivation of the original concept (object) representation in higher sensory
areas, which replicates a ‘perception’, i.e. causes a memory.
Encoding and retrieval of memory consistently activates the frontal cortex. It seems that the
left frontal cortex is more involved in semantic retrieval and episodic encoding, while the right
frontal cortex is more active during episodic retrieval, i.e. an example of hemispheric
specialization. However, this distinction is not perfect, i.e. there is some overlap. Imaging
studies have demonstrated that the left frontal cortex is more active during encoding of words,
there is bilateral activation in frontal cortex when nameable objects are encoded, while the
right frontal cortex is more active when faces are encoded into memory. This demonstrates
that the frontal cortex shows material specific activation during encoding. However, there are
many competing theories which aim to explain quite diverse findings, and the issue about
laterality and specificity of the frontal cortex during encoding and retrieval has not been finally
resolved.
The storage of semantic (factual) knowledge
As seen from H.S. long term memory of semantic knowledge was still possible to some extent
after bilateral removal of the hippocampus. Semantic knowledge is built through

2
associations over time. For example our concept of an elephant is built through accumulated
knowledge over time. We know what they look like (we see many different pictures over time),
what sounds they make (we hear different sounds over time), where they live, and we know
what the right word is to identify them (i.e. elephant). Interestingly, the different aspects that
contribute to this knowledge are not stored in one location, but they are stored in a distributed
manner. This knowledge was first derived from lesions in patients. Lesions to the occipital and
especially temporal cortex can result in very specific agnosias. Patients may still be able to
accurately copy a picture of an item that is well known to them, but they are no longer able to
name the item (called: associative agnosia). Conversely apperceptive agnosias leave
patients unable to copy a visual item, which they otherwise can still name. Memory of faces
is stored in different locations to memory of e.g. animals. Animate objects are stored in
locations that differ from locations where in-animate objects are stored.
Fine structure of anatomical organization of the MTL
So far we have learned a fair amount about the coarse organization of the medial temporal
lobe. Looking at the more detailed anatomy, and the input-output structure reveals to some
extent why and how the hippocampus might aid in long term memory storage.
Sensory information flows from early sensory areas (where feature extraction occurs) to
higher order sensory areas (where unimodal objects or categories are represented). It can
then either directly pass the information to the MTL or take a detour through multimodal
association areas (multimodal object representations or categories), and then pass the
information to the medial temporal lobe. The information is either passed to parahippocampal
cortex or to the perirhinal cortex. Both structures are connected to the entorhinal cortex, which
is the main interface between cortical (and subcortical) areas and the hippocampus. The main
output from the entorhinal cortex to the hippocampus is through the perforant pathway to the
dentate gyrus. The mossy fibre output from the dentate gyrus projects to CA1 of the
hippocampus, which in turn by means of Schaffer collaterals connects to area CA1 of the
hippocampus. The main output of the hippocampus originates in CA1 which projects to the
subiculum, which in term is connected to the entorhinal cortex. Reciprocal connections
between entorhinal cortex and parahippocampal and perirhinal cortex ensure that processed
information can be returned to the unimodal and multimodal association areas of the cortex.
Cellular mechanisms of learning in the hippocampus
During learning a variety of changes at the neuronal level take place. These include changes
related to transmitter release, changes related to postsynaptic sensitivity to transmitter
release and morphological changes of the pre- and postsynaptic neurons (increases in the
numbers of synapses between neurons).
Additional (necessary) background information: Before we can look at these mechanisms
it is necessary to provide some additional background information regarding neuronal
communication that has not been covered in PSY1005. You have learned (PSY1005, Lecture
3) that an action potential travels down the axon, when a neuron gets sufficiently excited. At
the synapse the action potential releases the neurotransmitter, which diffuses across the
synaptic cleft, to bind to receptors at the postsynaptic site and there causes depolarization- or
hyperpolarization (dependent on the transmitter released and/or the postsynaptic receptor
involved). In relation to the cellular mechanisms of learning, it is necessary to understand a
little bit more of the mechanisms by which the transmitter is released (synaptic transmission),
it is necessary to understand a little bit more about the different types of post-synaptic
receptors that exist, and a little bit more about what is called second messenger systems
(which act as intracellular messenger molecules). The relevant reading can be found in
chapter 2, pages 47-51, Cognitive Neurosciences, Gazzaniga, Ivry, Mangun, Third edition.
The hyperlinks embedded above give additional information.

3
Synaptic transmission: In brief, when the action potential arrives at the chemical
synapse, the associated depolarization causes voltage gated Ca++ channels to open, and
Ca++ enters the cell. While an increase in positive ions obviously results in further
depolarization, Ca++ acts as an intracellular messenger molecule, which triggers a
biochemical chain reaction. It is this biochemical chain reaction, which allows the intracellular
vesicles (which contain the transmitter) to fuse with the cell membrane, and release the
transmitter into the synaptic cleft. The important message here is that Ca++ induces a
biochemical reaction. Ca++ can also cause biochemical reactions which results in altered
efficacy of receptors and ion channels, it can induce protein synthesis in the cell, and thereby
induce growth of additional presynaptic and/or postsynaptic terminals. Thus Ca++, once
inside the cell, is an important messenger molecule which can change the microanatomy of
neurons and the efficacy of neuronal communication.
Receptor types: In PSY1005 you have learned about voltage gated receptors and
ligand gated receptors in relation to the action potential and synaptic transmission. The Ca++
receptor mentioned in relation to the synaptic transmission (see above) would be an example
of a voltage gated receptor, which is expressed in high density at the presynaptic terminal.
The voltage gated Na+ and K+ receptors that have been mentioned in PSY1005 are involved
in the action potential generation and are examples of different types of voltage gated
receptors. Another receptor type is a ligand gated receptor. These can be separated into two
main groups, directly coupled (ionotropic) or indirectly coupled receptors (metabotropic).
Directly coupled receptors change their conformation when the ligand binds to them and
thereby allow ions to flow through the receptor. Indirectly coupled receptors trigger an
enzymatic cascade (a biochemical chain reaction) which alters properties of specific protein
and thereby alters the cell properties. These altered cell properties are important for learning.
Glutamate as a neurotransmitter: The most prominent neurotransmitter in the brain is
glutamate. It can act on a variety of different pre- and postsynaptic receptors. It plays a key
role in the excitation of neurons at the postsynaptic site through ionotropic glutamate
receptors. Importantly these come in a variety of different forms. Broadly, it is possible to
distinguish NMDA receptors and non-NMDA receptors. These differ in two key aspects. Non-
NMDA receptors are ligand-gated receptors which open whenever glutamate binds to the
receptor. This will then allow Na+ to enter the cell and depolarize it. AMPA receptors are
usually fairly selective for Na+, i.e. they will not permit Ca++ to enter the cell (but there are
exceptions). Conversely, the NMDA receptor is a mix between a ligand gated receptor, and a
voltage gated receptor. If glutamate binds to the receptor the ion channel will only open if the
cell itself is fairly depolarized (>-50mV). Thus, nothing will happen unless other channels (e.g.
AMPA channels) also open. However, if the cell is depolarized and glutamate binds to the
NMDA receptor it will open and it will allow NA+ and Ca++ to enter the cell. Thus, the NMDA
receptor is able to sense that the presynaptic cell was active (glutamate release) and that the
cell it lives in is equally active (depolarization has occurred). Donald Hebb has postulated that
such sensing should be a key mechanisms in learning and neuronal plasticity (Hebbian
learning). As described above, a key step is also the entry of Ca++ into the cell. Ca++, once
inside the cell causes further depolarization, but it also acts as an intracellular messenger,
triggering a cascade of events which can result in increased sensitivity of the cell to incoming
events at future times.
One of the most studied electrophysiological phenomena associated with learning is long
term potentiation. It has been demonstrated to occur at all stages of the processing in the
hippocampus (at the synapses of the perforant pathway, of the mossy pathway, as well as of
the Schaffer collaterals). Long term potentiation reflects the finding that a short burst of
electrical stimulation applied to any of these pathways will change the response of the
recipient cell to transmitter release from the stimulated terminals. Potentiation reflects the fact
that response becomes larger, i.e. the depolarization of the cell (when postsynaptic receptors
bind transmitter) increases. Long term depression reflects the opposite finding, whereby
responses get smaller. Along the different processing stages different types of long term
potentiation can occur, namely associative and non-associative potentiation. Non-associative
long term potentiation describes the finding that there is no need for the pre- and postsynaptic

4
sites to be active simultaneously. The changes that occur are solely due to changes in the
presynaptic terminal. They occur because the tetanic electrical stimulation results in strong
influx of Ca++ ions. Ca++ ions are necessary for transmitter release, but they are also
involved in many second messenger cascades. Broadly speaking second messenger
systems involve cascades of biochemical (enzymatic) reactions which can alter the cell
properties and make it more susceptible for certain incoming information (see above). Large
amounts of Ca++ entering the presynaptic terminal will set in motion these enzymatic
reactions which result in increased transmitter release for extended periods of time, thus
potentiating the response of the post synaptic cell. Non-associative long term potentiation
does not require the involvement of NMDA receptors. Conversely associative long term
potentiation requires the pre and postsynaptic cells to be active simultaneously, and it
requires the involvement of NMDA receptors. As described above, NMDA receptors bind
glutamate, which allows them to open, but Mg++ will block the open pore, unless the cell is
already depolarized (this relates to the voltage gating of NMDA receptors mentioned above) .
It senses that the presynaptic and the postsynaptic sites are active simultaneously. The
opening of the NMDA receptor will further depolarize the cell as NA+ and Ca++ enter, but the
Ca++ will set in motion a biochemical cascade which results in phosphorylation (:=adding a
phoshate group to a protein, which alters is biochemical properties) of non-NMDA receptors
(an abundant class is the so called AMPA receptor). This phosphorylation, which is fairly
enduring, increases the sensitivity of the AMPA receptors to glutamate, ensuring that more
receptors open upon transmitter release. It also sets in motion a retrograde messenger
system, which results in increased transmitter release from the presynaptic site in the future.
LTP can also be divided into two phases. An early phase does not require new protein
sythesis, but involves modification of the existing receptors and transmitter release
machinery. The late phase is more long lasting and involves protein synthesis and structural
changes to the pre- and postsynaptic neurons.. All of these effects ensure that the
postsynaptic neuron is now more sensitive to incoming information from a specific input
(neuron), thus when that input/information arrives, it produces a more vigorous response.
Since it is widely believed, that memory is (at least partly) a function of different synaptic
connectivity strength in a highly connected network, changing the weights of the connections
can alter memories and engrain new memories into the network.
Long term potentiation in slices is a rather artificial means to alter neuronal connectivity
strength. But the involvement of NMDA receptors in learning can also be demonstrated in
vivo. This has been demonstrated in hippocampal place cells. Hippocampal place cells are
cells that fire when a mouse (or rat) enters a certain location within a given environment. The
place preference develops very quickly, when the animal is put into a new environment and is
stable thereafter, which is evidence for quick and lasting learning and memory at the cellular
level. A place cell has different place preferences for different environments, i.e. it can store
information about different environments. Elegant genetic modification of mice, which resulted
in selective loss of NMDA receptors in the CA1 region of the hippocampus, has demonstrated
that place cell preference does not develop to the same extent in mutant mice, and
importantly place field locations change daily in these animals. It can further bee shown that
mutant mice of this type do not show long term potentiation, and also are strongly impaired in
the water maze task (Morris water maze). In this task they have to learn (by try and error) the
location of an invisible platform which they can stand on. If the platform is visible, mutant and
wild type mice swim to the platform at equal rates. Thus NMDA receptors in CA1 of the
hippocampus are not directly involved in the swimming task when overt detection is possible.
If the platform is invisible, then wild type mice learn the location, and on subsequent test will
swim towards it. Mutant mice fail to learn it.