π Lecture 29 β Atkinson & Shiffrin Model
π Overview: This lecture explains the Atkinson & Shiffrin Model of memory, focusing on how information transfers from sensory register to short-term memory (STM) and then to long-term memory (LTM). It matters because it details the processes of retrieval, recall versus recognition, and the role of practice and priming in memory activation.
ποΈ Topics Covered
The lecture covers the entire memory process per the Atkinson & Shiffrin Model, including transfer from STM to LTM, retrieval from LTM back into STM, recall versus recognition, applications in studying and testing, the role of rote learning, and eyewitness testimony. It also explores experiments on speed of retrieval with practice, spread of activation in semantic networks, and associative priming effects.
π Lecture Summary
Long Term Memory
Cognitive psychology relies on experiments to generate and refine models. An important experiment by Anderson (1976) illustrated how speed of retrieval varies with practice. In the first phase, subjects were drilled on sentences like "The Sailor is in the park" and "The lawyer is in the church" until they knew them by heart. Subjects were then tested to verify if each sentence was among those studied (e.g., "The sailor is in the park" vs. "The sailor is in the church"). Subjects knew the material well enough to be correct almost all the time; the experiment focused only on the speed of their correct recognition judgments.
Results: The findings showed differences due to delay in presentation:
- Short Delay: Less Study (1.11 sec), More Study (1.10 sec)
- Long Delay: Less Study (1.53 sec), More Study (1.38 sec) The time span for Less Study was 0.42 sec and for More Study was 0.28 sec. In the less study and short delay condition, subjects were fastest; in the more study with short delay condition, the rate was similar. However, in the long delay with less study condition, reaction time increased to 1.53 sec. Delay and more study both made a difference.
π Definition β Spread of Activation: The process by which activating one node in a memory network activates associated nodes, with activation taking time to spread through long-term memory. π‘ Why this matters: The implication is that weaker memories take longer to reactivate. To remember a thing, one must study it more.
Spread of Activation
Perlmutter & Anderson conducted an unpublished experiment cited in cognitive psychology textbooks. Subjects were presented with a sequence of words and asked to generate associates that began with specific letters. There were two conditions:
- Priming condition: Pairs of letters and words are associated (e.g., "Dog C" for Cat; "Bone M" for Meat).
- Control condition: Words and letters are unrelated (e.g., "Gambler C"; "Bone M").
Results: Reaction time for the priming condition was 1.41 sec; for the control condition, it was 1.53 sec. This network model explains the difference: through the word "eat," both "dog" and "bone" are linked, showing relationship between words. The priming condition was recalled more easily. Therefore, activating the network structure to answer the first associate helps activate the structure needed for the second. This experiment showed that activation spreads from active portions to other portions of memory, and this spread takes time.
π Example: In the priming condition, "Dog β C" helps activate "Cat" because "Dog" and "Cat" are associated; in the control condition, "Gambler β C" does not prime any related word for "Bone β M".
Associative Priming
Mayer & Schvaneveldt (1971) performed a classic demonstration of associative priming. Subjects judged whether pairs of items were words.
- Positive Pairs: Unrelated (e.g., Nurse Bread) β 940 ms; Related (e.g., Butter x β 855 ms)
- Negative Pairs: Non-word 1st (e.g., Plame Wine) β 904 ms; Non-word 2nd (e.g., Wine Plame) β 1087 ms; Both non-words (e.g., Plame Reab) β 884 ms
If either item in a pair was a non-word, subjects responded "no." When the top item was a non-word, subjects were faster to reject the pair than when only the second item was a non-word (because they did not have to judge the second item). Subjects were fastest in the both non-words condition, faster in the non-word first condition than in the non-word second condition. In positive pairs, subjects were much faster in related pairs than in unrelated pairs.
π Definition β Associative Priming: The facilitation of processing a stimulus due to its association with a previously presented stimulus. π‘ Why this matters: This result indicates that because subjects judged the first item to be a word, activation spread from that word and primed information about the associatively related, second item. Judgment takes place in working memory; the representation of the word must be active in short-term memory. We can read related material faster than non-related, incoherent material. The implication is that associative spreading of activation through memory can facilitate the rate at which words are read.
More Priming
Ratcliff & McKoon (1981) reported a different priming demonstration of spreading activation. Subjects memorized sentences such as "The doctor hated the book." They then had to decide whether the noun "book" was in the studied sentence. Sometimes a prime word like "doctor" was presented just before "book." Ratcliff & McKoon varied the delay between prime and target from 50 to 300 ms. All intervals were too short for subjects to develop any conscious expectations. The decrease in reaction time reflects the growth in the level of activation.
π Formula: Activation spread time β 200 ms (after which it slows down).
Implications:
- There is a strong effect of priming in Long Term Memory β strength of coding.
- Activation spreads quickly up to 200 ms, after which it slows down.
- For recognition, information must first be activated and then inspected. Activation must spread to info in LTM to be brought into STM, which takes time (200 ms).
Short-term memory has less storage (about 7 items at once) and approximately 20 seconds duration of storage. Long-term memory has unlimited capacity and permanent storage. For information to be used in a task such as recognition judgment, it must first be activated and then inspected. When information is in LTM but not currently in working memory, activation must spread to it (taking time, as seen in the McKoon & Ratcliff study). Once activated, the time to inspect the information depends on its level of activation, as illustrated in the Sternberg experiment.
β Key Takeaways
The most critical points are: (1) Memory retrieval speed varies with practice and delay β weaker memories take longer to reactivate, so studying more improves retention. (2) Activation spreads through semantic networks from one associated concept to another, facilitating recall of related information. (3) Associative priming experiments show that related words are processed faster than unrelated words, proving that activation spreads in memory. (4) STM has limited capacity (7 items) and duration (20 seconds), while LTM has unlimited capacity and permanent storage. (5) For recognition, information must first be activated in LTM and then inspected in STM, with activation taking approximately 200 ms to spread.
π§ Quick Revision Questions
- What were the results of Anderson's 1976 experiment regarding speed of retrieval with less vs. more study and short vs. long delay?
- How does the Perlmutter & Anderson experiment demonstrate spread of activation in long-term memory?
- In the Mayer & Schvaneveldt (1971) associative priming experiment, why were subjects faster for related positive pairs than unrelated positive pairs?
- What did Ratcliff & McKoon (1981) discover about the time course of activation spreading in long-term memory?
- Why does it take longer to recognize information from LTM when there is a delay between study and test?
π Lecture 30 β Interference
π Overview: This lecture examines how activation spreads through memory networks and what factors influence this process. It focuses on the Fan Effect as a key demonstration of interference in memory retrieval, showing how the number of facts associated with a concept slows down recognition time.
ποΈ Topics Covered
The lecture covers encoding as a process of transferring information from short-term to long-term memory, the Fan Effect experiment by Anderson (1974) demonstrating how multiple associations slow recognition, network representations of memory with spreading activation, the critical assumption that activation reaching a proposition is inversely related to the number of links, and the concept of limited-capacity activation where source nodes divide fixed activation among all outgoing paths.
π Lecture Summary
Memory
Various factors can affect the amount of activation that spreads to a knowledge structure. From previous experiments we can infer that strength of encoding has an effect such that more strongly encoded information receives greater activation. Another factor is the number of alternative network paths down which activation can spread. Encoding is a process of transferring information from short-term memory to long-term memory through codes.
The lecture shows activation in a network using the word "penguin" as an example. Penguin has two links, one is "bird" and the other is "swimming." Bird and swimming have their own links as well. If we think about swimming, the activation model will become active and we can recall penguin because of its connection with swimming.
The Fan Effect
Anderson (1974) performed an experiment where subjects had to memorize 26 facts. In these statements, some persons were paired with only one location and some locations with only one person. Other persons were paired with two locations and other locations were paired with two persons. Each statement was followed by two numbers, reflecting the number of facts associated with the subject and the location. For instance, sentence 3 is labeled 2-1 because its subject occurs in two sentences (sentences 3 and 4) and its location occurs in one sentence (sentence 3).
The sentences were:
- The doctor is in the bank. (1-1)
- The fireman is in the park. (1-2)
- The lawyer is in the church. (2-1)
- The lawyer is in the park. (2-2)
Subjects were drilled on each sentence. Before beginning the reaction time phase, subjects were able to recall all the locations associated with a particular type of person (e.g., doctor) and all the people associated with a particular location (e.g., park). Then they began a speeded-recognition phase of the experiment, during which they were presented with sentences and had to judge whether they recognized them from the study set. Foil sentences were created by repairing people and locations from the study set.
Results
The recognition time for sentences as a function of number of facts learned about persons and location:
| Specific Location | 1 sentence | 2 sentences |
|---|---|---|
| 1 sentence | 1.11 | 1.17 |
| 2 sentences | 1.17 | 1.22 |
Results show recognition time increases as a function of both the number of facts studied about the person and the number of facts studied about the location.
The Fan Effect: Network
The network representation for sentences 1 through 4 shows that every node has three parts: relations, subject, and location. By applying the activation concept to this representation, we can account for the increase in reaction time. Subject might recognize a probe such as "a lawyer is in the park." First, presentation of terms "lawyer," "in," and "park" serves to activate their representations in memory. Then activation will spread from these nodes to activate the target proposition and enable it to be recognized. Fireman and lawyer interfere with park. But there is no interference between doctor and lawyer, so there is no fan effect.
The Critical Assumption
The amount of activation reaching the proposition is inversely related to the number of links leading from it.
Subjects should be slower to recognize a fact involving lawyer and park than one connecting doctor and bank because more paths emanate from the first set of concepts. In the lawyer and park case, two paths point from each of the concepts to the two propositions in which each was studied, whereas only one path leads from each of the doctor and bank concepts.
π Definition β Fan Effect: The increase in reaction time related to an increase in the number of facts associated with a concept.
π Principle: Activation reaching proposition = Fixed capacity Γ· Number of paths emanating from the source node
π Example: In Anderson's experiment, recognizing "lawyer is in the park" took 1.22 seconds (both concepts had 2 sentences each), while recognizing "doctor is in the bank" took 1.11 seconds (both concepts had 1 sentence each). The difference of 0.11 seconds occurred because lawyer and park each had two links (higher fan) while doctor and bank each had only one link (lower fan).
π‘ Why this matters: This demonstrates that additional knowledge about a concept can actually slow down retrieval of specific facts β more information is not always better for quick memory access.
Activation: Limited capacity
This experiment points to a limited-capacity feature of the spreading-activation process. The nodes from which the spread of activation starts, such as lawyer and park, can be called source nodes. One node can have thousands of connections. This one node supplies energy to other nodes, but the energy of this node is also transferred to others in a very limited amount. Like in the experiment from the last lecture, the word "dog" activates other kinds of dogs, meats, bones, and other animals.
A source node has a fixed capacity for emitting activation. This capacity is divided among all the paths emanating from that node. The more paths that exist, the less activation will be assigned to any one path and the slower will be the rate of activation.
At one time we can make many nodes with one word. Like the word "fish" has many links. We can remember many other things, like other kinds of fish, water, sea, other sea animals, etc. Another example: "gambler" β we can make many links with this word like cards, the pictures on cards, the figures of cards, etc.
π Definition β Interference: The phenomenon where additional information about a concept disrupts memory for a particular piece of information related to that concept.
Interference
The fan effect is the name given to this increase in reaction time related to an increase in the number of facts associated with a concept. It is so named because the increase in reaction time is related to an increase in the fan of facts emanating from the network representation of the concept. The term conveys that additional information about a concept interferes with memory for a particular piece of information. Interference affects a wider range of measures than just recognition time. The term "fan effect" is reserved for interference effects as measured by reaction time.
β Key Takeaways
The Fan Effect demonstrates that the more facts you know about a concept (higher "fan"), the slower you will be to recognize any single fact involving that concept. This occurs because activation from a source node has limited capacity β it is divided among all outgoing paths, so each path receives less activation when more paths exist. The critical assumption is that activation reaching a target proposition is inversely related to the number of links from the source concepts. Interference from competing associations is the underlying mechanism, and the fan effect specifically measures this interference through reaction time. For exam purposes, remember that Anderson's experiment showed recognition time increasing from 1.11 seconds (1-1 condition) to 1.22 seconds (2-2 condition).
π§ Quick Revision Questions
- What is the fan effect, and how did Anderson (1974) demonstrate it in his experiment?
- Why does recognition time increase when a concept is associated with more facts?
- What is the critical assumption about activation reaching a proposition in a spreading activation network?
- How does the limited-capacity feature of activation explain the fan effect?
- In Anderson's experiment, why would "lawyer is in the park" (2-2) be recognized more slowly than "doctor is in the bank" (1-1)?
π Lecture 31 β Interference
π Overview: This lecture examines how interference affects memory retrieval, focusing on the fan effect where increased facts about a concept slow reaction time. It explores interference in both experimental and real-world memories, explains why forgetting occurs, and contrasts recall versus recognition performance. Understanding these mechanisms is crucial for comprehending how memory networks operate and why recognition is typically superior to recall.
ποΈ Topics Covered
The lecture begins with the fan effect, describing how additional information about a concept interferes with memory for specific facts. It then examines historical memories through Lewis & Anderson's (1976) study on fantasy facts about public figures. Next, interference and retention are discussed using paired-associate learning experiments comparing A-D and C-D group designs. Finally, recall versus recognition differences are explained through network activation models.
π Lecture Summary
Interference
The fan effect is the name given to this increase in reaction time related to an increase in the number of facts associated with a concept. It is so named because the increase in reaction time is related to an increase in the fan of facts emanating from the network representation of the concept. The term conveys the fact that additional information about a concept interferes with memory for a particular piece of information. Interference affects a wider range of measures than just recognition time. Fan effect is reserved for interference effects as measured by reaction time.
Historical Memories
Lewis & Anderson (1976) investigated whether the fan effect could be obtained with material the subject knew before the experiment. They had subjects learn fantasy figures, for example, "Napoleon Bonaparte was from India." Subjects studied 0-4 fantasy facts about each public figure. After learning these facts they proceeded to a recognition-test phase. In this phase they saw three types of sentences: fantasy world statements, true statements, and false statements like:
- Statements they had studied in the experiment
- True facts about the public figures (such as Napoleon Bonaparte was an emperor)
- Statements about the public figure that were false both in the experimental fantasy world and in the real world Subjects had to respond to the first two types of facts as true and to the last type as false.
Results Subjects responded much faster to actual truths than to experimental truths. The advantage of the actual truths can be explained, because these true facts would be much more strongly encoded in memory than the fantasy facts because of greater prior exposure. The more fantasy facts one learned about a person, the longer it took them to recognize something they already knew about the person; Napoleon was an emperor.
π Definition β Fan Effect: The increase in reaction time caused by an increase in the number of facts associated with a concept, where additional information interferes with memory for a particular piece of information. π Example: Learning that "Napoleon Bonaparte was from India" (fantasy fact) slows recognition of the true fact "Napoleon Bonaparte was an emperor," and this slowing increases with more fantasy facts learned.
Interference and Retention
Now we will consider what happens as these interfering effects get more extremeβeither because the to-be-recalled fact is very weak or because the interference is very strong. There is evidence that the subject simply fails to remember the information under both conditions. Results showing such failures of memory have traditionally been obtained with paired-associate material, although similar results have been obtained with other material.
Experiment In a typical interference experiment, two critical groups were defined.
- The A-D experimental group learns two lists of paired associates:
- First list is List A-B: cat-43 and house-61
- Second list is List A-D: cat-82 and house-37
- The C-D control group also first studies the A-B list, but then studies a different second list:
- List C-D is bone-82 and cup-37, which does not contain the same stimuli as the first list
After learning their respective second lists, both groups are tested for their memory of their first list, in both cases the A-B list.
Results In general, the A-D group does not do as well as the C-D group with respect to both rate of learning of the second list and retention of the original A-B list. The results are presented in the following figure.
Implications The implication is that failure to recall is the extreme case of a long retrieval time. Thus, it is not the case that the forgotten information is not in memory, but rather that it is in memory but is too weak to be activated in the face of interference from other associations. Paired associate memory is too weak to recall. Forgetting is not actual loss of information but rather loss of ability to activate that information.
π Definition β Paired-associate learning: A memory task where subjects learn to associate one stimulus (e.g., "cat") with a specific response (e.g., "43"). π Formula: A-D group interference = original A-B list memory weakened by competing A-D second list; C-D control = original A-B list memory preserved because second list uses different stimuli (C-D) π‘ Why this matters: This demonstrates that forgetting results from interference preventing activation, not from information being erased from memory.
Recall versus Recognition
Consistent with the hypothesis that there exists in memory information that we cannot recall is the fact that we can recognize many things we cannot recall. This phenomenon suggests that information can be in memory even though it cannot be activated in the recall test situation. The memory network analysis makes clear the reason that recognition often works even when recall fails. So, recognition is generally better than recall.
For example, if there is a question: "Who invented the lenses we use in spectacles?" Then a huge fan of information networks becomes active. We recall a lot of information that is related to glasses or spectacles. For example, someone mentions Ibn-al Haitham invented the lenses we use in spectacles. If we have listened to this before, then we can recall this answer because of strong enough information. So, joint activation helps the second statement.
There are many other possibilities. If with this question we have some options like: Michael, Ibne Batota, Albaroni and Ibn Al Haitham. Now these options interfere with our information. And we become confused. But because of our links or network we can recall correct information. Like spectacles were invented by Muslim scientist so, Michael could not be answer.
So, recognition is typically better than recall because a recognition test typically provides more sources for activating memory. Recognition test is better than recall test. Tip of the tongue phenomenon happens in recall, not in recognition.
For example, if you see a man you say you have seen him before. So you can recognize him. But you are not recalling his name. In our daily life, in our exams, in any test or in other situation we think recognition is our friend and recall is not much friendly.
Another example is if someone asks you when Baber came in Hindustan and invaded Hindustan. The chances are we could not recall. If someone gives us some options like: 712, 789, 1566 and 1020 with the question. Then it becomes easy to recognize when Baber invaded India.
So the conclusion of all that is recognition is a better and easiest task than recall.
π Definition β Recall: The active retrieval of information from memory without cues (e.g., answering "Who invented lenses for spectacles?" without options) π Definition β Recognition: Identifying previously learned information when presented with it again (e.g., choosing Ibn Al Haitham from a list of names) π Definition β Tip of the tongue phenomenon: A state where one feels certain that information is stored in memory but cannot immediately recall it; occurs in recall but not recognition π‘ Why this matters: Recognition provides more retrieval cues than recall, explaining why we can often recognize things we cannot recallβinformation is present in memory but needs stronger activation to be retrieved.
β Key Takeaways
The fan effect demonstrates that increasing the number of facts associated with a concept slows reaction time due to interference. Lewis & Anderson (1976) showed this effect applies even to pre-existing real-world knowledge, where learning fantasy facts about public figures slows recognition of true facts about them. Paired-associate experiments reveal that forgetting is not information loss but rather an inability to activate weak memory traces due to interference from competing associations. This explains why the A-D group (with competing stimuli) performs worse than the C-D control group. Recognition consistently outperforms recall because it provides more activation sources in memory networks, which is why tip-of-the-tongue states occur only during recall, not recognition.
π§ Quick Revision Questions
- What is the fan effect and how is it measured?
- How did Lewis & Anderson (1976) demonstrate the fan effect using real-world knowledge?
- In the paired-associate experiment, why did the A-D group show poorer retention of the original A-B list compared to the C-D group?
- What is the key implication about forgetting according to the interference framework?
- Why is recognition typically better than recall, and what phenomenon occurs only in recall?
π Lecture 32 β Long Term Memory: Recall versus Recognition
π Overview: This lecture explores the relationship between recall and recognition in long-term memory, emphasizing how contextual clues and network activation aid memory retrieval. It also addresses the controversial question of whether forgotten memories are permanently lost, using Penfieldβs experiments as a case study.
ποΈ Topics Covered
The lecture covers eyewitness testimony and how contextual clues influence recognition, activation in memory networks with examples like penguin and swimming, the importance of creating links and elaborate networks for effective learning and recall (including smell and picture associations), and the Penfield (1959) experiment on whether forgotten memories are lost forever.
π Lecture Summary
Eye witness testimony or expert witness
When we recognize others, we make our own ways based on shared features stored in memory. For example, we recognize Chinese people by their small eyes and foreheads, or perceive Black farm people similarly due to common visible traits. This can lead to false testimonyβfor instance, a white man seeing a Black farm man at night might wrongly identify other Black farm people as the same individual because memory uses contextual clues saved in long-term memory to fill gaps.
Our memory is an inference shaped by these contextual clues. We use stored information to fill gaps in recall, which can lead to errors in recognition, especially in eyewitness situations.
π Definition β False testimony: Incorrect identification or recall of an event or person due to reliance on generalized or inferred contextual clues stored in memory, rather than specific accurate details. π Example: A white man sees a Black farm man at night and later incorrectly identifies other Black farm people as the same man because he uses a shared trait (e.g., being Black and from a farm) as a contextual clue.
Activation in network
The lecture presents a figure showing activation in a network. The word penguin has two links: one to bird and another to swimming. Bird and swimming have their own further links. If we think about swimming, the activation model becomes active, and we can recall penguin because of its connection with swimming. Thus, words are linked with many other wordsβby remembering one word, we can recall others.
In daily life, if we want to learn things and make links among this information, it becomes easier to remember or recall. For example, to learn an essay about parents, students can make links like father, mother, love, affection, help, sacrifice, food, etc. Making links among information helps recall information in exams.
Even for historical memories, making links aids recall. For instance, to recall who discovered Sulfuric Acid, learning that it was made by a Muslim scientist and linking that fact makes recall easy.
When students learn lessons by making a network of all material, they can easily recall it. Creating pictures of material also helps because memory for pictures is better and has deeper quality for recall than memory for words. Memory for smells is the longest lasting of all memories. For example, putting one drop of perfume on the page where you are making a network helps remember the information because the association between interesting and boring things makes remembering fast and easy. This is called elaborate networks.
In all sciences (physics, chemistry, biology, mathematics), information becomes easy to learn through elaborate networks. Books include pictures and figures because they help in remembering. Even in History or Islamiyat, hierarchies and tables are provided. Another example of memory for smells is that females often use diaries that have a scent or put perfume on notebooks.
π Definition β Elaborate networks: A learning technique where information is linked through associations (e.g., smells, pictures, hierarchies) to create deeper, more retrievable memory traces. π Principle: Association between interesting (e.g., smell) and boring (e.g., text) things β easier and faster remembering. π Example: Putting a drop of perfume on a page while studying an essay makes recalling that essay easier because the smell is linked to the memory. π‘ Why this matters: Elaborate networks leverage multiple sensory inputs (sight, smell) to strengthen memory retrieval, a key strategy for effective studying.
Are forgotten memories lost forever?
An important issue is whether forgotten memories are lost forever. An interesting possibility is that we never truly lose memoriesβthey are still present but too weak to be retrieved.
Penfield (1959) conducted a neurosurgical experiment. He electrically stimulated parts of patients' brains while they were conscious (the stimulation was painless). Patients reported what they experienced. This helped Penfield determine the function of various brain regions. Stimulation of the temporal lobes led to reports of memories that patients could not recall normally, such as events from their childhood.
It appeared as if Penfieldβs stimulation activated portions of the memory network that spreading activation could not reach. However, it is difficult to verify whether the reported memories were accurate because checking historical events was nearly impossible. Thus, although suggestive, the Penfield experiments are generally discounted by memory researchers. Even so, the question of forgotten memories remains important.
π Definition β Penfield (1959) experiment: A study where electrical stimulation of the temporal lobes in conscious patients triggered vivid, normally inaccessible memories, suggesting memories might not be permanently lost. π Key finding: Stimulation of temporal lobes β reports of childhood memories not accessible during normal recall. π Example: A patient under Penfieldβs stimulation might suddenly remember a specific toy from age 5 that they could not recall before, but the accuracy of this memory is unverifiable.
β Key Takeaways
Students must remember that memory is inferential, using contextual clues that can lead to false testimony in eyewitness situations. Activation in networks shows that words are linked, so making associations (e.g., between penguin and swimming) enhances recall. Creating elaborate networks using pictures, smells, and hierarchies significantly improves memory for any subject, as sensory associations make retrieval easier. The Penfield experiment suggests that forgotten memories may not be lost but simply too weak for normal retrieval, though this idea is controversial due to verification difficulties. Most critically, linking new information with interesting or sensory cues is a powerful learning strategy for exams.
π§ Quick Revision Questions
- What is "false testimony," and how do contextual clues create it in eyewitness situations?
- How does activation in a network explain how recalling "swimming" can help you recall "penguin"?
- What are "elaborate networks," and give one example of how smells can improve memory.
- What did Penfieldβs 1959 electrical stimulation experiment on the temporal lobes suggest about forgotten memories?
- Why do memory researchers generally discount Penfieldβs findings, despite their suggestiveness?
π Lecture 33 β Long Term Memory: Forgotten Memories Exist
π Overview: This lecture explores the fascinating phenomenon that forgotten memories still exist in long-term memory, even when we cannot consciously recall them. Through experimental evidence from Nelson's research, the lecture demonstrates that memories persist below conscious awareness and examines how knowledge is represented in the mind, including digital versus analog coding, propositional networks, dual code theory, and the complex nature of categories.
ποΈ Topics Covered
The lecture covers Nelson's 1971 and 1978 experiments demonstrating that forgotten memories still exist through savings in relearning, both for recall and recognition tests. It then examines representation of knowledge in long-term memory including digital versus analog formats, propositional networks versus images, dual code theory, and finally explores categories and concepts β including the problems with defining features and necessary and sufficient features across different knowledge domains.
π Lecture Summary
Forgotten Memories Exist β Nelson (1971)
Nelson (1971) conducted an experiment demonstrating that forgotten memories still exist in long-term memory. Subjects learned 20 number-noun paired associates until they reached a criterion of one errorless trial. Two weeks later, subjects returned for a retest and recalled 75% of the items. The critical focus was on the 25% of forgotten items.
For these forgotten items, subjects received new learning trials. The missed paired associates were either kept the same (unchanged) or changed (a new response was paired with the old stimulus). For example, if a subject learned "43βdog" but failed to recall the response to "43," they would now be trained on either "43βdog" (unchanged) or "43βhouse" (changed).
Results: After studying the new list once, subjects correctly recalled 78% of the unchanged items but only 43% of the changed items. If subjects had lost all memory for forgotten pairs, there should have been no difference between changed and unchanged pairs. This large advantage for unchanged items indicates that subjects had retained something about the paired associates even though they had been unable to recall them initially. This retained information was reflected in the savings displayed in relearning.
Recognition of Lost Memories β Nelson (1978)
Nelson (1978) extended this investigation to recognition tests. Four weeks after learning, subjects failed to recognize 31% of paired associates they had learned. As in the previous experiment, subjects relearned the missing items. For half the stimuli, the responses were changed, and for the other half, they were left unchanged.
Results: After one relearning trial, subjects recognized 34% of the unchanged items but only 19% of the changed items. Even when subjects fail this sensitive recognition test, there appears to be evidence that a record of the items is still in memory β evidenced by better relearning for unchanged than changed pairs.
π‘ Why this matters: This is a remarkable demonstration where recognition is worse than recall, yet memory is still present. The savings in relearning reveal that memories persist even when conscious retrieval fails.
Representation of Knowledge
Representation of knowledge is a fundamental issue in long-term memory that deals with what form knowledge is stored. The main issues include:
- Digital versus analog representations
- Propositional networks versus images
- Dual Code Theory
- Categories and concepts
- Defining features
- Necessary and sufficient features
Digital versus Analog
The lecture introduces the distinction between digital and analog representations of knowledge, though this section is briefly mentioned without full elaboration.
Propositional Networks versus Images
In propositional analysis, only the meaning of an event is represented (like the fan effect). Unimportant details β details that humans tend not to remember β are not represented. In this network, information, relations, and arguments are connected to each other, forming a network.
π Definition β Propositional Network: A representation system where only the meaning of information is stored, with connections between concepts, relations, and arguments forming a network structure.
Dual Code Theory
The Dual Code Theory proposes two distinct coding systems:
Propositional Network Code: For example, "my house is in Lahore" β this sentence is stored in memory. If someone asks about the distance between the house and a hotel, we create different sentences, make links between roads and stops, and can then tell the distance. These sentences are stored in memory. Information about abstract concepts like intelligence, love, and honesty is also stored in this propositional form.
Images are stored: For example, the image of the house is stored as a visual representation. Different scenes, presentations, smells, or tastes of food are also stored in memory. We can recall them even when they are not present.
π Definition β Dual Code Theory: The theory that knowledge is represented in two distinct formats β verbal/propositional codes for meaning and abstract concepts, and imaginal codes for visual, sensory information.
Categories
Cognitive psychology is a merger or meeting point of different types of knowledge including philosophy, computer science, artificial intelligence, psychology, and social work.
The category fruits groups many different kinds of objects that have essential features in common while excluding objects belonging to other categories such as vegetables. Traditionally, categories have been defined as having necessary and sufficient features.
π Definition β Category: A group of objects sharing the same essential features. Examples: bird, furniture, fruit, robin.
Dictionary Definition of Fruit: "The edible product of a plant or a tree consisting of a seed and its envelop. The envelop is juicy and pulpy." Examples: apple, orange, plum.
Necessary and sufficient features for fruit: Edible, contains seed, juicy/pulpy envelope.
Dictionary Definition of Vegetable: "Edible plant product eaten raw or cooked." Examples: carrots, spinach, tomatoes.
Biological classification: "Fruit is that part of the plant which develops out of a flower and nurtures seeds."
Necessary and sufficient features for vegetable: Edible, plant products.
Problem with Categories
Categories are not as neat and clean as philosophers would have liked them to be. Several problems arise:
Problem 1: In biology, tomato is a fruit (it develops from a flower and nurtures seeds). However, for a chef, tomato is a vegetable. The category "vegetable" does not even exist in biology.
Problem 2: Is chicken a bird or an animal? In biology, chicken is both an animal AND a bird. However, for a layperson, birds must be able to fly β flying is considered a necessary feature of the category "bird." So people consider chicken an animal because they eat it, but biologists consider chicken both a bird and an animal.
Importance of Categories: Categories are very important. Letters are categories. If we know the categories of things, we can recognize them even when we don't know the exact thing. For example, there are many types of dogs. We may have seen only specific dogs in our area, not a German Shepherd. But when a German Shepherd appears, we can recognize it as a dog because of the category of dog and its essential and sufficient features. Pattern recognition is a part of category recognition.
π Definition β Necessary Features: Features that must be present for something to belong to a category.
π Definition β Sufficient Features: Features that, if present, are enough to determine category membership.
Categories of Language
Categories are critical to our understanding of information processes. They help us know how we think about things around us. Language is also very important in category recognition because words have different meanings in different languages. In one country like Pakistan, many languages are spoken: Sindhi, Balochi, Punjabi, Pashto, among others.
Dialect versus Language: Some languages are considered dialects while others are considered proper languages. For example, some people say Punjabi is a dialect, while others say it is a complete language.
β Key Takeaways
The most critical points from this lecture are: (1) Forgotten memories still exist in long-term memory, demonstrated by the savings effect β when subjects cannot recall or recognize previously learned information, they still show better performance on unchanged versus changed items during relearning. (2) Knowledge representation involves multiple formats including propositional networks (storing meaning and connections) and imaginal codes (storing visual and sensory information), as proposed by Dual Code Theory. (3) Categories are defined by necessary and sufficient features, but these definitions often break down across contexts β as shown by tomato being a fruit in biology but a vegetable in cooking. (4) Categories are essential for pattern recognition and cognitive processing, allowing us to recognize novel instances of familiar categories. (5) The distinction between dialect and language demonstrates that categories are influenced by social, cultural, and contextual factors, not just objective features.
π§ Quick Revision Questions
-
In Nelson's 1971 experiment, why did subjects show better recall for unchanged paired associates (78%) compared to changed ones (43%), even though they had initially failed to recall the forgotten items?
-
What does the savings effect in relearning demonstrate about the nature of forgotten memories in long-term memory?
-
According to Dual Code Theory, what are the two distinct formats in which knowledge can be represented in memory?
-
Why does the biological classification of tomato as a fruit conflict with the culinary classification of tomato as a vegetable, and what does this reveal about the concept of necessary and sufficient features?
-
How does pattern recognition relate to category knowledge, and why can we recognize a German Shepherd as a dog even if we have never seen one before?
π Lecture 34 β Representation of Knowledge: Categories
π Overview: This lecture explores how humans organize knowledge through categorization, explaining the definition, benefits, and levels of categories. It also distinguishes between concepts, exemplars, and prototypes as different ways the mind represents categories. Understanding these structures is essential for grasping how memory and cognition organize the vast information we encounter daily.
ποΈ Topics Covered
The lecture covers the definition and benefits of categorization, including how it reduces complexity and aids learning. It then explains the three levels of categories: superordinate, basic, and subordinate. The discussion moves to concepts as mental representations, necessary and sufficient features, mental experiments, exemplars, and prototypes.