Lecture 3: Combining Neuropsychology and Neuroimaging to Understand Learning and Memory in Development
Declarative Memory
Declarative memory refers to memories that can be consciously described or verbally expressed. Endel Tulving proposed that declarative memory contains two distinct forms of conscious experience:
episodic memory
semantic memory.
This distinction is important because the two systems develop differently, rely on somewhat different brain regions, and are affected differently by brain injury and ageing.
Episodic Memory
Episodic memory involves remembering specific personal experiences and mentally reliving past events.
Key features include:
remembering what happened
remembering where and when it happened
subjective recollection
“mental time travel”
highly personal experiences.
Example:
“I went to the park at lunch time and saw a really cute dog.”
Episodic memory depends heavily on the hippocampus because the hippocampus binds together all aspects of an experience into a single memory trace.
This type of memory develops slowly across childhood. Young children often cannot remember detailed early-life experiences even when those experiences were emotionally important. Early episodic memory appears immature and underdeveloped because hippocampal systems continue developing postnatally.
In older adulthood, episodic memory difficulties become more pronounced. This is associated with age-related hippocampal changes and hippocampal atrophy.
Laboratory measures of episodic memory include:
recall tasks
associative memory tasks
“remember” judgments.
Semantic Memory
Semantic memory refers to general knowledge about:
language
vocabulary
concepts
facts.
Example:
“There are often dogs in the park.”
Semantic memory is not linked to a specific event or personal experience.
Young children acquire semantic memory very rapidly:
vocabulary expands quickly
language develops efficiently
concepts are learned with ease.
Semantic memory is relatively preserved in older age compared with episodic memory. Older adults may forget recent events but still retain:
vocabulary
factual knowledge
conceptual understanding.
This preservation is thought to depend on distributed cortical systems rather than the hippocampus alone.
Laboratory measures include:
vocabulary tests
information/general knowledge tests.
Developmental Differences Between Episodic and Semantic Memory
The developmental trajectories of episodic and semantic memory differ substantially.
Semantic memory:
develops early
is highly efficient in young children
remains relatively stable into older age
depends strongly on cortical systems.
Episodic memory:
develops slowly
depends heavily on hippocampal maturation
declines more noticeably in older adulthood
is vulnerable to hippocampal injury.
This developmental dissociation strongly supports the idea that episodic and semantic memory are partially separate systems.
Developmental Amnesia (DA)
Developmental amnesia occurs following early hypoxic-ischaemic injury, often during traumatic birth events where oxygen supply to the brain is reduced. Developmental Amnesia (DA) is a rare memory disorder that emerges in early childhood.
The injury particularly affects the hippocampus.
Children with developmental amnesia typically show:
severe episodic memory impairment
preserved semantic memory
preserved language development
relatively normal IQ
apparently normal early development.
The memory problem is often not noticed immediately because infants do not yet rely heavily on episodic memory. Families usually begin noticing difficulties around age 4–5 when children start failing to remember recent experiences or important events.
Common signs include:
repeatedly asking the same questions
forgetting recent conversations
inability to remember birthday parties or daily events.
Despite these episodic impairments, children with DA often acquire extensive semantic knowledge. Some children develop rich knowledge about topics such as Pokémon, facts, language, or hobbies while being unable to remember what happened yesterday.
This creates a striking dissociation between:
preserved semantic memory
impaired episodic memory.
Evidence Supporting Separate Memory Systems
Neuropsychological evidence strongly supports Tulving’s distinction between episodic and semantic memory.
Children with developmental amnesia often show:
normal or above-average vocabulary scores
preserved conceptual understanding
catastrophic episodic memory impairment.
This is theoretically important because it suggests:
semantic learning can occur without intact episodic memory
episodic memory is not strictly necessary for acquiring concepts and language.
The findings challenge intuitive assumptions that people must remember experiences in order to learn from them.
People do not need to rembember the experience to learn facts from it.
Neuroimaging and the Memory Networks
Functional neuroimaging studies complicated the simple separation between episodic and semantic memory because the two systems show substantial overlap in brain activation patterns.
Semantic Memory Network
The semantic network includes:
inferior frontal gyrus
temporal cortex
posterior cingulate cortex
dorsomedial prefrontal cortex
ventromedial prefrontal cortex.
These regions support:
language
conceptual processing
meaning
factual knowledge.
Episodic Memory Network
The episodic network includes:
hippocampus
parahippocampal cortex
retrosplenial cortex
angular gyrus
medial prefrontal cortex
middle temporal gyrus.
Many cortical regions overlap between episodic and semantic memory networks.
This overlap created a major theoretical issue:
if the systems overlap neurally, why do patients with hippocampal injury show such selective episodic deficits?
A possible explanation is that:
both systems recruit shared cortical processes during normal development
the hippocampus acts as a particularly important node for episodic recollection
rare developmental cases reveal the dissociation more clearly.
Declarative Memory is Multifaceted
Episodic memory is not a single process. It contains multiple component processes that may develop at different rates.
Different memory tasks may therefore measure different aspects of episodic memory rather than a single unified system.
Encoding Processes
1. Perceptual binding
Links different aspects of an experience together into one memory:
people
places
objects
emotions
context.
The hippocampus is important for binding these details into a coherent episodic memory.
2. Pattern Separation
Allows similar experiences to be stored as separate memories.
Prevents interference and confusion between overlapping events.
Example:
remembering today’s lunch separately from yesterday’s lunch.
LONG TERM STORAGE
Encoded memories are stabilised and consolidated over time, especially during sleep.
This is important in developmental amnesia because recognition-based learning effects often emerge after delays, suggesting sleep consolidation strengthens weak memory traces.
Retrieval Processes
1. Pattern Completion
A partial cue triggers retrieval of the full memory.
Example:
hearing “birthday” may reactivate the cake, decorations, people, and emotions associated with the event.
fMRI studies showed this can still occur neurally in developmental amnesia despite impaired conscious recollection.
2. Subjective Recollection
The conscious feeling of mentally reliving an experience.
This is often severely impaired in developmental amnesia.
3. Recall
The behavioural expression of remembering, such as verbally describing an event.
Recall depends not only on memory but also:
language
attention
executive function.
Memory Tests Are Not Process-Pure
Memory tasks always involve multiple cognitive systems, including:
attention
language
executive functioning.
This explains why patients may perform differently across memory tasks and why preserved recognition can occur alongside impaired episodic recall.
Associative Memory And Recognition Development Study
Researchers studied children with congenital heart disease because reduced oxygen supply early in life may affect hippocampal development and episodic memory.
Children were assessed longitudinally at:
Time 1 (T1): mean age 3.3 years
Time 2 (T2): mean age 5.1 years.
Researchers measured:
cognitive development
language abilities
motor abilities
associative memory.
Associative memory was tested using the “Birthday Party Story,” a child-friendly episodic memory task designed for children aged 3–6.
Children:
encoded information by helping characters prepare for a birthday party
experienced a short delay while hearing about the party
later remembered which items belonged to each character.
The task measured associative binding because children had to link together:
characters
objects
contextual information.
Associative binding is an important component of episodic memory and is strongly associated with hippocampal function.
The prediction was that if the hippocampus develops gradually after birth, associative memory performance should improve with age.
Children performed above chance even at age 3, showing that some associative memory abilities are already operating in early childhood.
Importantly, age predicted associative memory performance even at Time 1 after controlling for general cognitive development, language, and motor abilities.
This means:
older children within the T1 group performed better than younger children
associative memory development could not simply be explained by general intelligence or language development.
Performance also improved further by Time 2 around age 5.
Together, these findings suggest that associative binding processes continue developing across childhood alongside hippocampal maturation.
During the pandemic, the Birthday Party Story was adapted into an online tablet/smartphone game.
The online version added:
a recognition memory control task
parent ratings of attention, engagement, and listening.
Recognition memory involved simple familiarity judgments such as:
“Which shoes have you seen before?”
Associative memory still required contextual binding:
“Which snack belongs to Snoop?”
Researchers predicted that associative memory would improve with age, whereas recognition memory would not because familiarity processes mature earlier and are less dependent on hippocampal development.
Regression analysis showed that age still predicted associative memory even after controlling for:
recognition performance
parent ratings
attention and engagement.
Recognition memory showed much less developmental change than associative memory.
This strengthened the conclusion that hippocampal associative processes have a prolonged developmental trajectory and continue maturing across childhood.
Key Takeaway
Associative memory develops gradually across childhood because hippocampal systems mature slowly after birth. Children as young as 3 already show some associative memory ability, but performance improves with age as hippocampal-dependent binding processes become more developed.
The studies showed that this improvement is specific to associative memory rather than general cognitive development, attention, engagement, or simple familiarity-based recognition. Recognition memory remained relatively stable, while associative memory continued improving across childhood.
This supports the idea that:
familiarity develops earlier
hippocampal associative binding develops more slowly
episodic memory maturation has a prolonged developmental trajectory.
Six-Year-Old With Developmental Amnesia
A six-year-old child with major hippocampal reduction demonstrated:
preserved language
preserved IQ
impaired associative memory
preserved recognition memory.
This produced a very strong dissociation between:
associative episodic retrieval
simple familiarity-based recognition.
The child also demonstrated preserved semantic understanding of birthdays and could discuss them in detail.
This reinforces the idea that semantic knowledge can develop despite profound episodic impairment.
Recognition Memory and Familiarity
Recognition memory refers to the feeling that something has been encountered before without recollecting contextual details.
Example:
recognising a familiar face but not remembering where you know the person from.
Recognition is thought to rely more on cortical familiarity systems and less on the hippocampus.
familiarity/recognition → more cortical systems
episodic recollection → more hippocampal systems.
Patients with hippocampal injury can therefore:
recognise information
experience familiarity
fail to retrieve associated episodic context.
Pattern Completion Deficit in Developmental Amnesia
The major deficit in DA appears to involve pattern completion.
Children with DA can:
encode information
experience familiarity
store partial traces.
However, they struggle to:
retrieve contextual associations from partial cues
reconstruct coherent episodic memories.
This prevents:
full episodic recollection
subjective remembering.
Hippocampal Function in DA
The hippocampus in DA is usually reduced rather than completely absent.
Residual hippocampal activity sometimes remains, and patients may occasionally produce isolated episodic memories.
This suggests:
hippocampal functioning is degraded rather than eliminated
episodic retrieval is unreliable rather than impossible.
Interestingly, sometimes partial hippocampal damage may disrupt memory more than severe damage because the remaining hippocampal tissue can produce weak or inefficient memory signals that interfere with compensatory cortical systems. In severe damage, the brain may rely more consistently on alternative cortical pathways for familiarity and semantic learning, allowing compensation to occur more effectively.
Important Developmental Implication
An important theoretical idea emerging from these findings is that semantic frameworks may develop before mature episodic memory.
Young children may first:
develop conceptual structures
organise knowledge semantically
before developing the ability to form detailed autobiographical episodic memories.
This may explain:
infant learning
preserved semantic development in DA
why early childhood episodic memories are sparse despite extensive learning.
Unconscious Associative Memory in Developmental Amnesia (fMRI)
Individuals with developmental amnesia (DA) often show a striking dissociation:
newly learned information can sometimes be recognised
but it cannot be consciously recalled.
This raises several important possibilities:
the memory trace may be too weak
associative binding may be impaired
consolidation may be inefficient.
A major limitation of behavioural testing alone is that it only reveals what a person can consciously report. Neuroimaging allows researchers to investigate whether memory-related processing is still occurring in the brain even when the individual cannot consciously access the memory.
This is important because some memory processes in DA may resemble normal learning and memory processes that are also used in normal early childhood before mature episodic recollection fully develops. If relying on behavioural measures.
fMRI Scene Reinstatement Paradigm
Researchers designed an fMRI paradigm to investigate whether associative memory representations could still be reactivated in DA despite impaired recall.
Encoding Phase
Participants studied:
180 word-picture pairs
each word paired with either:
a meaningful scene
a scrambled scene.
Scene images included:
urban scenes
rural scenes.
Scrambled scenes acted as a control condition because they lacked meaningful spatial information.
The scenes were specifically chosen because they reliably activate scene-sensitive cortical regions during perception and memory retrieval.
Participants made semantic judgments about whether the object fit within the scene. This encouraged associative binding between:
the word
the object
the contextual scene.
Only one presentation of each word-scene pair occurred, making the task highly dependent on rapid associative encoding.
Retrieval Phase
The retrieval phase occurred inside the MRI scanner.
Participants first answered:
“Did you see this word before?”
This mainly tested familiarity and recognition.
Then participants answered:
“Was it paired with an urban scene?”
This required:
episodic retrieval
associative binding
contextual recollection.
Example:
“LEMON” had actually been paired with a rural scene
“CHAIR” had been paired with a scrambled scene.
Patients with DA were expected to perform relatively well on simple recognition judgments because familiarity processes are often preserved. However, they were expected to struggle with recalling the associated scene context.
Scene Reinstatement
The critical question was not simply whether patients answered correctly, but whether the brain showed evidence of reactivating scene-related information during retrieval.
This process is called scene reinstatement.
Scene reinstatement occurs when retrieving a memory reactivates cortical regions originally involved in perceiving that experience.
If seeing the word “LEMON” later reactivates scene-processing regions, this suggests that the associated scene memory is still represented neurally, even if the participant cannot consciously report it.
Brain Regions Involved
Researchers focused on scene-sensitive cortical regions including:
parahippocampal cortex
retrosplenial cortex.
These regions are strongly associated with:
scene perception
spatial processing
contextual memory.
Two forms of localisation were used:
anatomical masks
functional localisers.
The functional localiser compared:
viewing real scenes
viewing scrambled scenes.
This identified regions selectively activated by scene processing.
Control Participants
Healthy controls showed clear scene reinstatement.
When words previously associated with scenes were retrieved:
scene-processing regions became active
scrambled-scene trials did not show the same activation.
This demonstrated that the brain was reactivating contextual scene information even though no scene was physically present during retrieval.
The cortical activity therefore reflected memory retrieval rather than visual perception.
Unexpected Findings in Developmental Amnesia
Patients with DA showed surprisingly strong scene reinstatement despite severe behavioural impairments.
Behaviourally:
they often failed to correctly report the scene context
they appeared unable to consciously recollect the associations.
Neurally:
scene-related cortical regions were strongly reactivated
associative memory processing appeared relatively normal.
This was highly important because it suggested:
memory traces still existed
associative processing still occurred
the major problem may involve conscious access to memory rather than complete memory loss.
This led to the phrase:
“when the brain, but not the person, remembers.”
Retrieval Goal Manipulation
Researchers introduced a second task to test whether scene reinstatement occurred automatically or depended on retrieval goals.
Instead of asking about scene context, participants answered:
“Was the word presented on the left or right side?”
In this task:
scene information was irrelevant
participants only needed spatial location information.
Importantly:
scene reinstatement disappeared in both controls and DA patients
scene-processing regions were no longer strongly activated.
This demonstrated that:
reinstatement was not automatic
it depended on the retrieval demands of the task
DA patients could still strategically reactivate contextual information when prompted.
It means the brain only reactivated the scene information when the task required remembering the scene.
So the word itself (“LEMON”) did not automatically make the brain replay the scene every time.
Instead:
if the participant was trying to remember the scene → scene brain regions became active
if the participant was only trying to remember left/right position → scene brain regions stayed inactive.
This shows memory retrieval is goal-directed. The brain reactivates the information that is relevant to the current question or task.
The key idea is:
Pattern completion = the brain can reconstruct associated information from a partial cue.
But whether that reconstruction fully happens depends on the retrieval goal and attention.
So when participants were asked about the scene:
the brain used the word cue (“LEMON”) to reactivate the associated scene
this is evidence of pattern completion.
But when the task only asked about left/right position:
the brain did not need the scene information
so it did not strongly reactivate the scene representation.
This means pattern completion is partly goal-directed, not constantly triggered at full strength every time a cue appears.
Interpretation of the Findings
The findings suggest that many associative memory processes remain surprisingly intact in DA.
Processes that appear relatively preserved include:
perceptual binding
pattern completion
cortical reinstatement.
However, subjective episodic recollection appears disrupted.
Patients could not consciously reconstruct rich autobiographical experiences even though the underlying neural processing appeared relatively normal.
This suggests DA may function partly as a disconnection syndrome:
memory processing occurs
cortical reactivation occurs
but conscious awareness and narrative recollection fail.
Pattern Completion
Pattern completion refers to retrieving an entire memory representation from a partial cue.
Example:
The word “LEMON” triggered reactivation of the associated rural scene.
This demonstrates that pattern completion processes may still occur in DA.
However:
subjective recollection did not emerge
participants could not consciously describe or relive the scene.
This dissociation is central to understanding hippocampal memory function.
Precision and Binding Model
The findings support the Precision and Binding model.
According to this model, the hippocampus contributes to:
precise memory representations
high-resolution associative binding
detailed recollection.
In DA:
memory representations may still exist
but they may be imprecise or weak
conscious retrieval may therefore fail.
This explains how:
familiarity can remain intact
some cortical reinstatement can occur
episodic recollection is still severely impaired.
Memory Processes Across Encoding and Retrieval
The model separates memory into multiple stages:
Encoding
perceptual binding
pattern separation
long-term storage.
Retrieval
pattern completion
subjective recollection
recall.
The evidence suggests:
earlier cortical processing stages may remain functional in DA
later conscious recollection stages are disrupted.
Key Takeaway
The study showed that many associative memory processes in developmental amnesia remain relatively intact despite severe episodic memory impairment. Patients still showed:
perceptual binding
pattern completion
cortical reinstatement
even when they could not consciously recollect the memory.
This suggests the main problem in developmental amnesia may not be complete loss of memory traces, but impaired conscious episodic recollection and reduced memory precision. Semantic learning may still be supported through implicit memory signals and cortical reactivation.
Example:
A child with DA may not remember learning a fact about Vikings in class, but repeated brain reactivation and familiarity may eventually allow them to “know” the fact anyway.
Semantic Learning in Developmental Amnesia
Semantic memory is often a relative strength in DA.
Children with DA often achieve:
normal vocabulary scores
strong factual knowledge
good academic skills in some domains.
However, they still struggle in education because learning new information is difficult without episodic memory support.
Researchers proposed that:
recognition
familiarity-based learning
may support semantic learning in DA.
Case H
Case H was an 8-year-old child with developmental amnesia caused by meconium aspiration at birth.
Features included:
typical infant development
memory problems emerging during preschool years
repeated questioning
inability to remember birthdays.
MRI showed:
approximately 51–52% reduction in hippocampal volume bilaterally.
Despite this:
IQ was high
vocabulary was strong
academic attainment was relatively preserved.
However, severe impairments were observed in:
verbal recall
episodic learning
delayed recall.
Recognition memory remained much stronger than recall.
Recognition-Based Semantic Learning
Researchers tested whether recognition learning could support semantic learning in DA.
Case H watched educational videos about topics such as:
Vikings
Egyptian nomads.
Two learning conditions were compared:
Recall-Based Learning
After watching the video:
the child answered free recall questions.
Performance remained very poor even after repeated learning cycles.
Recognition-Based Learning
After watching the video:
the child completed multiple-choice recognition tests.
Performance improved substantially because recognition relied on familiarity rather than free episodic recall.
Delayed Learning Effects
The major improvement emerged after a one-week delay.
Recognition-based learning produced:
better later recall
reduced impairment
more stable semantic knowledge.
This suggests:
sleep-dependent consolidation may strengthen weak memory traces
cortical familiarity systems may gradually support semantic learning over time.
Interesting Error Patterns
Case H often produced unusual responses during recall:
information appeared fragmented
contextual details were inconsistent
isolated facts were sometimes preserved.
Example:
The child repeatedly remembered:
“brought the cattle inside to protect them from the cold”
even when failing other questions about Viking homes.
This suggests:
some semantic fragments consolidate
retrieval remains noisy and unstable
memory signals fluctuate in accessibility.
Educational Implications
Multiple-choice testing may function as a practical classroom intervention for children with DA.
Advantages:
easy to implement
compatible with online learning systems
relies on preserved familiarity processes
reduces demands on episodic recall.
Repeated recognition opportunities may gradually strengthen semantic representations through cortical consolidation.
This may explain how:
infants learn before mature episodic memory develops
semantic knowledge can emerge despite hippocampal impairment.
Semantic Precision
The Precision and Binding model proposes that the hippocampus supports precise, high-resolution memory representations in both:
episodic memory
semantic memory.
Example:
“dog” = broad semantic category
“Rhodesian Ridgeback” = highly precise semantic representation.
Researchers propose that people with DA may rely on imprecise semantic representations that are still sufficient for learning.
Semantic Precision and Aging
Adults with acquired amnesia also show imprecise semantic memory.
Older adults commonly demonstrate:
reduced episodic precision
less detailed semantic responses
age-related hippocampal atrophy.
Recognition learning appears to reduce these age-related memory deficits as well.
This suggests shared mechanisms between:
developmental amnesia
normal cognitive ageing.
Overlapping Regions in semantic and episodic neural network
Medial Prefrontal Cortex (mPFC)
This appears in both networks.
Semantic role:
supports conceptual knowledge
integrates meaning
links information with existing knowledge.
Episodic role:
supports autobiographical retrieval
self-referential remembering
organising memories into coherent narratives.
So the mPFC may help integrate information into broader knowledge structures in both systems.
Posterior Cingulate / Retrosplenial Region
These regions are closely linked and often overlap functionally.
Semantic role:
supports retrieval of meaningful conceptual information.
Episodic role:
strongly involved in contextual recollection
scene reconstruction
autobiographical memory retrieval.
This region is important for internally directed cognition and mental reconstruction in general.
Temporal Cortex / Middle Temporal Gyrus
The semantic network includes temporal cortex broadly, while the episodic network specifically includes the middle temporal gyrus.
These areas support:
conceptual representations
stored knowledge
integration of information across experiences.
In episodic memory, they may help reactivate stored details associated with past experiences.
Regions More Specific to Episodic Memory
These are much more strongly tied to episodic recollection and hippocampal processing:
hippocampus
parahippocampal cortex
angular gyrus.
These regions are especially important for:
associative binding
contextual detail
pattern completion
subjective recollection.
This explains why hippocampal damage selectively disrupts episodic memory more severely.
What overlap perhaps suggest
The overlap suggests that episodic and semantic memory are not completely separate systems.
Instead:
they share cortical processing networks
semantic and episodic memory interact continuously
semantic knowledge may partly develop from repeated episodic experiences.
However, the hippocampus appears uniquely important for:
vivid recollection
detailed contextual binding
conscious episodic remembering.
So when the hippocampus is damaged:
shared cortical systems can still support semantic learning
but rich episodic recollection becomes severely impaired.
That is why patients with developmental amnesia can:
acquire vocabulary
learn facts
build conceptual knowledge
while still being unable to consciously relive personal experiences.