Difficulty: Introductory to Intermediate | Prerequisites: Working Memory and Short-Term Memory study notes (Part 1).
These notes cover the second half of the working memory lecture: how memories are formed, stored, and retrieved. This is where the course moves from "what can STM hold?" to "how does information get into long-term memory, and how do we get it back out?" The material introduces the levels-of-processing framework, the serial position curve, encoding specificity, state-dependent recall, and the reconstructive nature of retrieval. These concepts are heavily tested and form the foundation for everything the course covers about long-term memory, false memories, and eyewitness testimony. You should already be comfortable with the STM and working memory material from Part 1.
Memory involves three processes: encoding, storage, and retrieval. Deep (meaning-based) processing at encoding produces stronger memories than shallow (surface-level) processing, and this matters more than whether you intend to learn the material. At retrieval, memory works best when the context matches how the information was originally encoded (encoding specificity), and retrieval itself is reconstructive, meaning you piece together an answer from stored traces and inferences rather than replaying a recording.
Encoding
The process of converting information into a form that can be stored in memory. Think of it as the "writing" stage: how well you encode something determines how well you can find it later.
Storage
The process of maintaining encoded information over time. In simple terms, this is where the memory sits between when you learn it and when you need it.
Retrieval
The process of accessing stored information when you need it. Think of it as searching through a filing system: the quality of your search cue determines whether you find what you are looking for.
Serial position curve
The U-shaped pattern in free recall: items at the beginning of a list (primacy effect) and the end of a list (recency effect) are remembered better than items in the middle.
Primacy effect
Better recall for items presented early in a list, likely because they received more rehearsal and were encoded into long-term memory.
Recency effect
Better recall for items presented at the end of a list, likely because they are still active in short-term memory at the time of recall. This effect fades with a delay.
Levels of processing (Craik and Lockhart, 1972)
The framework proposing that deeper, more meaningful processing at encoding leads to better memory than shallow, surface-level processing. In simple terms, thinking about what something means helps you remember it far better than thinking about what it looks like.
Intentional memory
Memory formed when you are deliberately trying to learn the material.
Incidental memory
Memory formed without any intention to learn, as a byproduct of processing the material for another purpose.
Elaboration
The process of adding meaning, associations, and connections to material during encoding. In simple terms, the more you connect new information to things you already know, the easier it is to retrieve.
Encoding specificity
The principle that memory is best when the context at retrieval matches the context at encoding. Think of it as a lock-and-key relationship between how you learned something and how you try to remember it.
State-dependent recall
The finding that memory retrieval is better when your physiological or mental state at test matches your state at encoding. This refers to context outside the material itself (mood, substance state, physical environment).
Recall
A memory test that asks you to produce the learned items with minimal cues (e.g., "Tell me all the items you saw"). The cue is weak: the learning context itself.
Recognition
A memory test that presents items and asks whether you saw them before (e.g., "Did you see this item?"). The cue is strong: the item itself.
Confabulation
Filling in memory gaps with information that feels correct but may be fabricated, without awareness that the memory is false. Seen in Korsakoff's syndrome and other conditions affecting retrieval and executive control.
Every memory system, whether human, animal, or machine, must handle three processes:
Encoding: Getting information into the system
Storage: Maintaining information over time
Retrieval: Getting information back out when needed
Animal models are frequently used in memory research because these three processes are universal across species. When you study a memory phenomenon, a useful first question is: is this about encoding, storage, or retrieval?
STM and LTM differ on three key dimensions:
Representation: STM uses acoustic/linguistic coding; LTM uses meaning-based (semantic) coding.
Duration: STM lasts less than 10 seconds without rehearsal; LTM can last days, months, or years.
Retrieval: STM retrieval is relatively effortless (the items are "right there"); LTM retrieval requires a search process guided by cues.
When people are given a list of items and asked to recall them in any order (free recall), a characteristic U-shaped curve emerges:
Primacy effect: Items at the beginning of the list are recalled well. This is consistent even with a delay between study and test. The likely explanation is that early items receive more rehearsal and are transferred to LTM.
Recency effect: Items at the end of the list are recalled well. This effect fades with a delay, suggesting these items are still sitting in STM at the time of recall.
Items in the middle of the list are recalled worst.
The Glanzer and Cunitz study demonstrated this pattern across 0-second, 10-second, and 30-second delay conditions. With delay, the recency effect disappeared but the primacy effect held steady.
Some researchers interpret the serial position curve as evidence for a slot theory of working memory: items occupy discrete slots, and items at the beginning and end of a list have privileged access to those slots.
The central claim: the deeper and more meaningfully you process information at encoding, the better you remember it. "Deep" processing means engaging with the meaning of the material. "Shallow" processing means attending to surface features (e.g., what it looks like, how many letters it has).
Intentional vs. incidental memory:
Intentional memory: Deliberately trying to learn items.
Incidental memory: Remembering items you were not trying to learn, as a byproduct of processing them for another task.
Hyde and Jenkins (1969) study:
Two groups studied a list of 24 words, each with a different orienting task:
Shallow group: Count the number of e's in each word.
Deep group: Judge how pleasant each word is (a semantic, meaning-based task).
Results (words recalled out of 24):
Incidental learning, pleasantness task: 16.3
Incidental learning, counting e's task: 9.9
Intentional learning (told to memorise): 16.1
Incidental pleasantness with intentional instruction: 16.6
Incidental counting e's with intentional instruction: 10.4
Key implications:
Deep processing (pleasantness judgement) produced recall nearly as good as intentional memorisation. Intention to learn added almost nothing on top of deep processing.
Shallow processing (counting e's) produced poor recall even when participants were told to memorise. Trying hard is not enough; what matters is the type of processing.
Rehearsing or rereading material without engaging with its meaning leads to weak memories.
Elaboration of material is very helpful: connecting it to what you already know, generating examples, thinking about why it matters.
Retrieval is a search. Long-term memory contains a vast amount of stored information, and finding a specific item depends on the quality of the retrieval cue.
Recall vs. recognition:
Recall ("Tell me all the items you saw"): The cue is weak. The learning context is essentially the only cue.
Recognition ("Did you see this item?"): The cue is strong because the item itself is presented.
Recognition is generally easier than recall, but it can be made difficult by using semantically related foils. If you studied "cupcake, balloon, candle, buttercream" and are asked whether "birthday" was on the list, you are likely to say yes, because your memory grouped the items by meaning. This error reveals that LTM stores information semantically.
Levels of processing and retrieval cues:
Consider the test word "elephant." Shallow encoding ("it has two e's") produces a cue that does not distinguish elephant from many other words. Semantic encoding ("elephants are terrifying, they have trunks and crush people") produces a rich, distinctive memory trace that is far easier to find at retrieval.
Encoding specificity (Klein and Murphy, 2000):
Participants studied word phrases (e.g., "wrapping PAPER," "tennis SHIRT," "tasty CHICKEN") and had to remember the capitalised words. At test, they decided whether capitalised words had appeared before, in three conditions:
Same context (tennis SHIRT at study and test): 79% accuracy
Same word meaning, different context (tasty CHICKEN at study, broiled CHICKEN at test): 64% accuracy
Different word meaning (wrapping PAPER at study, daily PAPER at test): 56% accuracy
The closer the test context is to the study context, the better memory is. The way you initially encode a word has a large effect on how you remember it. If the context at retrieval matches the context at encoding, memory will be better. This is encoding specificity.
State-dependent recall extends the encoding-specificity principle to context outside the material itself. If you learn material in an unusual physiological or mental state, you may retrieve it better when you are in the same state again.
Baddeley's diving study: Participants who learned material underwater recalled it better when tested underwater than on land, and vice versa.
Eich's marijuana study: Participants learned a word list either under the influence of marijuana (M) or in a control state (C), then were tested four hours later in either state.
Those who studied under M and were tested under M recalled more than those who studied under M but were tested under C.
Those who studied under C and were tested under C recalled the most overall.
The key finding is the interaction: recall is best when the study state and test state match, regardless of which state that is.
State-dependent recall refers to context outside the material (mood, substance state, physical setting), whereas encoding specificity refers to the context of the material itself (the meaning or associations activated during encoding).
Memory retrieval is not like replaying a video. You reconstruct a memory by assembling the best answer you can from whatever traces and cues are available.
Recall = stored information + inferences. You fill in gaps based on what makes sense.
Confabulation occurs in some forms of amnesia (e.g., Korsakoff's syndrome): the person fills in memory blanks with plausible but incorrect information, and crucially, they are unaware that the memory is false. This involves a breakdown in both retrieval and executive control.
The general principle applies to healthy memory too. When you retrieve information, you are building it back up from pieces, not accessing a perfect copy. This makes memory vulnerable to distortion, suggestion, and interference.
Students often think that trying harder to memorise something guarantees better recall. The Hyde and Jenkins study shows that the type of processing matters more than the intention to learn. Shallow processing with full intention to memorise still produces poor recall.
Students often confuse encoding specificity with state-dependent recall. Encoding specificity is about the match between the material's encoding context and the retrieval cue. State-dependent recall is about the match between your physiological or environmental state at encoding and retrieval.
Students often assume recognition is always easy. Recognition can be quite difficult when the foils are semantically related to the studied items (e.g., "birthday" as a foil after studying party-related words).
Students often think of memory as playback, like rewinding a recording. Memory is reconstructive: you piece together an answer from stored traces and inference, which is why memory errors, distortions, and confabulation are possible.
⚠️ The levels-of-processing framework and the Hyde and Jenkins study are core exam material. Be able to explain why deep processing beats shallow processing, and why intention to learn adds little beyond deep processing.
⚠️ Know the three components of the serial position curve (primacy, recency, middle dip) and how delay affects each.
⚠️ Be prepared to explain encoding specificity with the Klein and Murphy (2000) study: the three conditions, the results, and what they demonstrate.
⚠️ Distinguish between encoding specificity and state-dependent recall. Exam questions often present a scenario and ask which principle applies.
⚠️ Understand the difference between recall and recognition in terms of cue quality. Be ready to explain why recognition can sometimes be difficult (semantically related foils).
⚠️ Confabulation and the reconstructive nature of memory are frequently tested. Know what confabulation is, which condition it is associated with (Korsakoff's syndrome), and what it reveals about retrieval processes.
True or False: Intentional memorisation always produces better recall than incidental learning. (False: deep incidental processing can match or beat shallow intentional learning.)
Fill in the blank: The ______ effect describes better recall for items at the beginning of a list. (Primacy.)
True or False: The recency effect persists even after a 30-second delay filled with a distractor task. (False: the recency effect fades with delay.)
Fill in the blank: Encoding specificity predicts that memory is best when the context at ______ matches the context at ______. (Retrieval, encoding.)
True or False: Confabulation involves knowingly making up false memories. (False: the person is unaware the memory is fabricated.)
Q: Explain the levels-of-processing framework. What does "deep" processing mean, and why does it lead to better memory?
A: The levels-of-processing framework (Craik and Lockhart, 1972) states that the depth of processing at encoding determines memory strength. Deep processing involves engaging with meaning (semantic processing), which creates a richer, more distinctive memory trace. Shallow processing deals with surface features (e.g., visual appearance, letter counts). Deep processing leads to better memory because the resulting trace has more retrieval routes and is more distinguishable from other memories.
Q: Describe the Hyde and Jenkins (1969) study. What were the key findings, and what do they imply about the role of intention in learning?
A: Two groups processed a word list with different orienting tasks: one counted e's (shallow) and one judged pleasantness (deep). The deep-processing group recalled about 16 of 24 words, nearly matching the intentional-learning control group (16.1). The shallow group recalled only about 10 words even with intentional instructions. This implies that the type of processing, not the intention to learn, is the primary driver of memory encoding.
Q: What is encoding specificity? Describe the Klein and Murphy (2000) study and its results.
A: Encoding specificity is the principle that memory is best when the retrieval context matches the encoding context. Klein and Murphy had participants study words in phrases (e.g., wrapping PAPER) and then tested recognition in three conditions: same context (79%), same meaning but different context (64%), and different meaning (56%). This demonstrates that even for the same word, changing the encoding context reduces recognition accuracy.
Q: How does state-dependent recall differ from encoding specificity?
A: State-dependent recall concerns the match between external/physiological context (mood, substance state, physical environment) at encoding and retrieval. Encoding specificity concerns the match between the internal context of the material (the meaning and associations activated). Both predict that matching conditions help memory, but they operate on different aspects of context.
Q: Why is memory described as "reconstructive"? What does confabulation illustrate about this?
A: Memory retrieval involves piecing together an answer from stored traces and inferences, not replaying a recording. Confabulation (as seen in Korsakoff's syndrome) illustrates this clearly: when retrieval fails, the person fills in gaps with plausible but false information, without awareness that the memory is fabricated. This shows that retrieval is an active construction process, and when executive control breaks down, the construction goes unchecked.
Levels of processing connects directly to study strategies: the research suggests that active engagement with meaning (self-testing, generating examples, explaining concepts to someone else) produces far stronger memories than passive rereading or highlighting.
Encoding specificity and state-dependent recall connect to eyewitness testimony research later in the course. The conditions under which a witness encodes an event (stress, focus of attention, time of day) constrain what they can retrieve, and mismatches between encoding and retrieval contexts help explain why eyewitness accounts are often unreliable.
The reconstructive nature of memory is the foundation for the false-memory literature, including the DRM paradigm, misinformation effects, and the work of Elizabeth Loftus on leading questions and memory distortion.
Memory processes, encoding, storage, retrieval, serial position curve, serial position effect, primacy effect, recency effect, Glanzer and Cunitz, levels of processing, depth of processing, Craik and Lockhart 1972, shallow processing, deep processing, semantic processing, Hyde and Jenkins 1969, intentional memory, incidental memory, incidental learning, elaboration, elaborative encoding, encoding specificity, Klein and Murphy 2000, state-dependent recall, state-dependent memory, Baddeley diving study, Eich marijuana study, context-dependent memory, recall vs recognition, retrieval cue, cue quality, foils, false recognition, confabulation, Korsakoff syndrome, reconstructive memory, cognitive psychology, long-term memory, LTM