Midterm 2 (NEW PROGRESS) Flashcards
What are examples of typicality effects?
- We name typical category members before atypical ones
- We are faster to put typical members into categories than atypical ones
- Typical exemplars show larger priming effects than atypical ones
- Infants learn typical category members first
- When producing sentences, we list typical category members before atypical ones
Describe the exemplar theory of categorization
- Rather than storing an abstract prototype in memory to which items can be compared, exemplar theory proposes that we store actual examples of items we have encountered in the past
- Ex: your knowledge of the bird category contains a set of birds that you have seen before
- Categorization occurs by comparing new items to the ones you have in memory and looking for similarity between their features
- If a new item has many similar features with the category members you have in memory and not a lot of features in common with members of other categories, it is placed into the first category
How does the exemplar theory of categorization explain the typicality effects?
- This theory can explain typicality effects because typical items are similar to many other category members so it will be easy (and fast) to retrieve those members from memory
- Atypical members will be harder to retrieve because they are less common
- Ex: on the one hand, an apple is similar to other fruits but dissimilar to vegetables, so it is considered a typical fruit. This means it will be processed quickly because we can think of many other similar fruits. On the other hand, a squash is similar to some fruits, but also similar to some vegetables. This makes it an atypical member of the fruit category and more difficult to identify as a fruit because there are fewer similar examples
How does exemplar theory explain context effects?
- Because it assumes that categorization depends on personal experience
- Ex: a robin is a typical bird in North America because it is similar to many birds that one would encounter there. Similarly, a rainbow lorikeet is a typical bird in Australia because is it similar to many birds that are seen there
Describe Dopkins and Gleason’s (1997) study on the exemplar theory before they introduced the ambiguous exemplars
- They had participants learn to categorize rectangles on a computer screen into 2 different categories
- Participants were not given any rules to help them learn how to sort the rectangles, they were simply told whether they had classified each rectangle correctly
- What participants didn’t know was that the rectangles could be categorized by considering both their length and the position they appeared on the screen -> one type of rectangle was usually wide and near the top of the screen (category 1); the other type of rectangle was narrow and near the bottom of the screen (category 2)
- After many trials, participants were able to accurately categorize the rectangles
- This finding was expected because it has been known for a long time that we are very good at categorization even when we aren’t told how to do it
Describe Dopkins and Gleason’s (1997) study on the exemplar theory when they introduced the ambiguous exemplars
- Participants then had to categorize new, ambiguous exemplars which could theoretically belong to either category (they were medium length in the middle of the screen) and were designed in such a way that prototype theory and exemplar theory made different predictions about how they would be categorized
- The new rectangles were similar to the average, or prototypical, rectangle from category 2, but they were more similar to some of the previously seen individual exemplars from the category 1, meaning that the prototype approach would predict that participants would categorize the new rectangles into category 2, whereas, the exemplar theory would predict that participants would categorize the new rectangles into category 1
- The results showed that participants tended to categorize the new rectangles into the wide and high group (category 1) most often
- They based their categorization on similarity to previously seen exemplars rather than similarity to a prototype
What do the prototype theory and exemplar theory fail to account for?
- Both theories claim that the fact that people provide consistent typicality ratings to items indicates that category boundaries are fuzzy and not rule-based. However, Armstrong, Gleitman, and Gleitman (1983) showed that participants could give consistent typicality ratings to well-defined categories, indicating that perhaps the typicality ratings of other categories, such as birds and fruit, were merely an artifact of the experimental method used and not indicative of fuzzy category borders
- According to these views, we categorize items by comparing the similarity between a new item and either a prototype or many exemplars in memory. However, they don’t specify how we decide which features to compare. This concept of implicit ideas about category membership is not explained by either prototype or exemplar theory, but is addressed by knowledge-based views of categorization
Describe Armstrong, Gleitman, and Gleitman (1983) study on typicality ratings
- They had participants give typicality ratings to naturally occurring categories, such as birds and fruit, but also to categories that have well-defined category membership rules (ex: odd numbers)
- They found that their participants agreed with the idea that numbers are either odd or even and that there really isn’t such a thing as a “more odd” number, (i.e., they conform to rules). However, they still gave odd numbers typicality ratings when asked to do so and participants’ typicality ratings were consistent with each other
Ex: participants agreed that 3 is a more typical odd number than is 447 - Conclusion: If participants could give consistent typicality ratings to well-defined categories, perhaps the typicality ratings of other categories, such as birds and fruit, were merely an artifact of the experimental method used and not indicative of fuzzy category borders
What did Murphy and Medin (1985) point out about categorization?
- That there’s potentially an infinite number of ways that any 2 items could be similar
- Ex: A banana and a television can both be purchased in a store, both are enjoyed by people, both are smaller than a truck, neither of them is purple
- We appear to have an understanding of which features are important for category membership and which are not
- This concept of implicit ideas about category membership is not explained by either prototype or exemplar theory, but is addressed by knowledge-based views of categorization
Describe knowledge-based views of categorization
- This approach assumes we use more than feature similarity to categorize
- Instead, they propose that we rely on our broad knowledge base to explain the reasons for category membership
- Often our ideas about category membership are implicit
- Ex: we know a poodle is a dog because there’s just something “doggy” about it. Even a poodle with three legs that can’t bark is still a dog because it hasn’t lost its “dogginess”
-This type of theory avoids the problem faced by prototype and exemplar theories of explaining which features are important for determining category membership and which are not - According to theory-based approaches, membership isn’t based on features; instead, we use broad theories about essentialism
What did Medin (1989) refer to as psychological essentialism?
- He described this idea that we have implicit theories about the requirements for category membership
- The idea is that all category members possess a fundamental essence that is unique to that category and determines membership
- Ex: Dogs are “doggy,” birds are “birdy,” and fruits are “fruity”
What’s psychological essentialism?
The proposal that categories have a natural underlying true nature that can’t be stated explicitly
What did Murphy and Allopena discover with regards to descriptions of place whose features do and don’t go together?
- They found that participants can learn about the building whose features go together more easily than the second type of building whose features don’t seem to go together
- These results indicate that when we learn about categories, we try to make meaningful connections from our past knowledge to explain the particular combination of features
- According to this view, we rely on categories to teach us about the world, and we use our knowledge about the world to help explain category membership
What’s a consequence of categorization based on psychological essentialism?
- Risk of applying “essential” qualities to social categories in the same way that we do to biological categories
- While it may be reasonable to think all dogs have an essential “doggy” quality, it doesn’t make sense to make the same claims about older adults
- While it’s natural to expect that we put people into categories in the same way that we categorize objects since our brains are designed to categorize the world to help us make sense of it, it’s important to keep in mind that categorizing people that way leads to stereotyping
- There’s a relationship between essentialism and stereotypes
Describe Bastian and Haslam (2006) study on participants’ essentialist beliefs
- They asked participants to complete questionnaires that measured their essentialist beliefs
- The questionnaires measured how much participants believed that people belong to discrete groups (“Everyone is either a certain type of person or they are not”), whether or not qualities are changeable (“Everyone, no matter who they are, can significantly change their basic characteristics”) and whether group membership has a biological cause (“The kind of person someone is can be largely attributed to their genetic inheritance”)
- The results revealed that participants with higher essentialist beliefs were more likely to endorse a variety of stereotypes about different groups of people
Why is it important to not apply essentialism views on people?
- This leads to stereotyping
- While we all categorize people based on our past experiences and knowledge, it’s important to remember that the categories we form about people are shortcuts meant to reduce cognitive load, and not based on essential characteristics of group members
Research about what has been particularly influential in informing psychologists’ understanding of knowledge organization?
Research about human memory and artificial intelligence
What’s a core idea about categories that was first introduced by Rosch et al (1976)?
- He noticed that individual items can belong to multiple levels or hierarchies of categories
- Rosch further suggested that basic level categories are the most cognitively efficient
What are the 3 levels of categories that Rosch named for his idea of the hierarchy of categories?
- Basic level category
- Subordinate category
- Superordinate category
Describe the basic level category
- The level of categorization that people find most natural
- It’s the most cognitively efficient level that is both informative and distinctive
- Level in the hierarchy that seems to most of us to be “just right” -> they provide just the right amount of information about the category to provide useful information (informative), and can be used to distinguish members from members of other categories (distinctive)
- This is the level we typically use to name the category of an item, and it’s also the one that children learn first as they learn to name objects around them
Describe the subordinate category level
- The category level that is below the basic level
- This level is more informative than the basic level but less distinctive
- Informative because they provide a lot of information about the item; however, they are not distinctive because they share many features in common
Describe the superordinate category level
- The category level that is above the basic level
- This level is less informative than the basic level but distinctive
- Ex: animals are quite different from fruits, but knowing something is an animal provides relatively little information compared to knowing that it’s a dog
Rank the different category levels from most broad to most specific
- Superordinate level
- Basic level
- Subordinate level
How can the human mind can be likened to a computer?
Humans are information processors that receive sensory input, use rule-based strategies to manipulate information, and produce a behavioral output, much in the same way as a computer
Where must knowledge be represented and stored?
Knowledge must be represented and stored within an information processing system (either a human or computer)
Why did both early cognitive psychologists and computer scientists develop theories of knowledge organization?
- Because knowledge representation was thought to be essential for information processing
- Many of the early theories of knowledge organization in humans were developed to explain how a computer might be programmed to store information
What became one of the most influential in shaping the way cognitive psychologists think about semantic knowledge?
A model for how to store semantic information in computer memory and train a computer to understand printed text
Describe Collins and Quillian (1969) hierarchical semantic network model
- They suggested that knowledge is stored as concepts within individual units called “nodes”
- These nodes point to properties of those concepts as well as to other nodes so that knowledge is represented in a network of interconnected nodes
- Ex: the concept “CANARY” is stored in a node that has pointers to the properties “CAN SING” and “IS YELLOW” “CANARY” is also connected to the “BIRD” node because a canary is a type of bird
- They believed semantic networks are organized hierarchically with superordinate categories occupying the uppermost level of the network and subordinate categories occupying lower levels
- Because computers (and presumably humans) have limited memory storage, this model demonstrates cognitive economy by only storing a property once at the highest level in the hierarchy
- Semantic networks demonstrate property inheritance as subordinate categories inherit the properties of the superordinate categories they are connected to
- Activation spreads between concepts by pathways connecting them
What’s cognitive economy?
The tendency to conserve cognitive effort and resources
What’s property inheritance?
A characteristic of semantic network models in which nodes inherit the properties of the nodes higher in the hierarchy to which they are connected
How does activation of the hierarchical semantic network model happen?
- A node becomes active as a result of input from the environment
- Ex: from reading the word canary or seeing a picture of a canary in a book
- This activation then spreads from the “CANARY” node to the other nodes that it’s connected to
- An implication of this spreading activation within such a hierarchical model is that nodes that are farther away from each other in the hierarchy will take longer to activate
How did Collins and Quillian test their hierarchical semantic network model in humans?
- By using a sentence verification task
- In this task, participants are presented with sentences, such as “A canary is a bird” and asked to press one button if it is a true statement and another button if it is a false statement
- The results of their experiment confirmed their model
- Participants responded fastest to sentences that could be answered by searching between the fewest levels in the hierarchy (e.g., “A canary is a canary”) and slowest to sentences that required searching through multiple levels in the hierarchy (e.g., “A canary is an animal”)
What’s a sentence verification task?
An experimental task in which participants have to judge whether a sentence is true or false as fast as possible
What problems did other researchers who began testing theories about the organization of semantic memory find with the hierarchical model of semantic knowledge?
- A hierarchical model of semantic knowledge failed to account for typicality effects
- Using a sentence verification task, Rips, Shoben and Smith (1973) demonstrated that participants were faster to respond to “a dog is an animal” than “a dog is a mammal”
- A hierarchical model predicts the opposite results because a search has to travel through more levels in the hierarchy between “DOG” and “ANIMAL” than between “DOG” and “MAMMAL.”
- They also found that participants were faster to respond to “a robin is a bird” than “a chicken is a bird.”
Describe Collins and Loftus (1975) spreading activation model
- A semantic network model in which concepts are organized based on their semantic similarity to each other
- Accounts for typicality effects
- They used the idea of semantic priming to propose a model in which nodes are connected to each other via semantic relatedness instead of using a hierarchical structure
- There’s no hierarchy in the spreading activation model. Instead, concepts are organized by their semantic similarity. The more similar the concepts, the more connections between them and the shorter the distance between them
- Ex: “FIRE ENGINE” is closely related to “TRUCK” and “BUS” but not semantically similar to “FLOWERS”
- This model proposes that when one node becomes active as a result of a stimulus in the environment, activation spreads to all connected nodes -> the farther this activation has to travel, the longer it takes and the weaker it becomes
- Ex: activation spreads quickly and strongly from “FIRE ENGINE” to “TRUCK” but is unlikely to reach “ROSES” because activation will have faded by the time it travels that far in the network
- This idea of spreading activation has been very influential in psychologists’ understanding of semantic knowledge structure
What were Meyer and Schvaneveldt (1971) the first to do?
- They were the first to use a lexical decision task to demonstrate semantic priming between two related words
- That is, they found that participants were faster to respond to the word butter if it was presented with the word bread compared to if it was presented with the word nurse
How does Collins and Loftus (1975) spreading activation model explain typicality effects?
- This model can explain typicality effects because typical exemplars are semantically similar to each other and, therefore, activation will spread quickly between them
- Atypical exemplars are farther apart from other category members because they are less semantically similar to other category members, so activation will take longer to travel between those concepts
What’s a schema?
- A cognitive structure that’s an organized knowledge base about a particular topic
- It includes everything we know about a particular thing, event, person, or situation
- The concept of a schema is broad and may seem vague, but that is because the knowledge within a schema is broad and not necessarily well defined
- Ex: Schema about university includes everything you know about what a university looks like, what you do there, and what people you would find there
What idea did Sir Francis C. Bartlett introduce?
- The idea that knowledge is organized in schemata (schemas)
- He was primarily interested in applied psychology (he was one of the first and most influential psychologists to bring experimental psychology out of the laboratory)
- He wanted to know about the things that affected people’s abilities to remember meaningful, everyday information
- Rather than having participants remember lists of unrelated words or strings of letters, Bartlett wanted to know how people could remember stories and pictures
- He proposed that what we remember is influenced by our past experiences and knowledge
In order to investigate schemata and the role of knowledge on memory, Bartlett devised what method?
The method of repeated reproduction
Describe Bartlett’s method of repeated reproduction
- An experimental technique in which participants must reproduce an item repeatedly from memory
- In this method, participants are shown a stimulus, then asked to reproduce it from memory
- After a delay, participants are asked to reproduce it again from memory, and again, and again, etc.
- Participants are only shown the original stimulus once
- Each reproduction has to be based on memory
- The reproductions become less similar to the original stimulus with each attempt and begin to look more and more like a familiar object
- Over time, details are lost from memory but we can use the information from our schemata to help guide memory retrieval
- Bartlett’s participants did not have past knowledge related to the original stimulus, but they did have schematic knowledge about a face so they used that knowledge to create the best reproduction they could
What did Bartlett’s method of repeated reproduction help him conclude about memory
- Bartlett demonstrated that memory is reconstructive
- Instead of retrieving an exact copy of an event from memory, we rely on our past knowledge and experience to help us reconstruct memory the best we can
- If we don’t have a memory for a specific item, we can use our schematic knowledge to fill-in-the-blanks
What’s implied by the schema theory of knowledge organization and people and their schemas?
- That people with similar experiences will have similar schemata
- In this way, we can talk about shared cultural knowledge and use it to make inferences and predict behavior
Describe Bernsten and Rubin (2004) study on undergraduate students’ schemas about a baby’s imagined life or life scripts
- They asked undergraduate students to imagine a newborn baby and make predictions about what would happen in that baby’s life as they got older
- There was a large amount of overlap across the life events listed by participants, including “start school,” “go to college,” “fall in love,” “get married,” and “have children”
- This showed that part of our knowledge includes expectations about what is expected in a typical life-course within a culture
What do all of the theories based on a “classic” view of cognitive psychology and the computer metaphor of the mind assume about knowledge?
That knowledge must be represented somehow in the mind as symbols that can be manipulated
What’s a problem faced by any information processor that uses symbols to represent knowledge?
The symbol grounding problem
What’s the symbol grounding problem?
- Problem first described by Stevan Harnad
- Relates to how symbols get their meaning in the real world
- The problem is that any symbol system can only replace one symbol with another one, and this process could continue infinitely
- Symbols need to be grounded -> there needs to be some way to connect this symbolic representation to the real world
- Humans aren’t troubled by this problem because we can interact with the environment to ground our symbols
- Ex: we know what an apple is because we can use our bodies to see it, touch it, and taste it
What’s one way that artificial intelligence has addressed the symbol-grounding problem?
- With robotics
- A computer that can receive sensory input from the environment and manipulate objects can address the symbol grounding problem by getting direct access to the environment
Ex: The robot iCub
What has allowed important advances in machine learning and artificial intelligence?
Artificial neural networks (ANNs; also called connectionist models)
Who aided the creation of artificial neural networks and what is the goal of ANNs?
- The creation of ANNs was made possible by computer scientists working to understand how knowledge might be represented in a symbol system, such as a computer or human
- Although, at first glance, connectionist models may seem like a type of semantic network model, their fundamental assumptions are very different from each other
What’s the difference between cognitive psychology and connectionism?
Cognitive psychology was based on a computer metaphor of the mind, but connectionism is based on how neurons in the brain are connected
Describe the structure and function of ANNs
- ANNs are composed of nodes in input, output, and hidden layers (much like the human cortex is composed of layers) that are connected to each other by weighted connections
- Unlike semantic network models, ANNs don’t store knowledge explicitly in nodes. Instead, knowledge is contained in the distribution of weights between the connected nodes. These weights determine which pattern from the inputs (such as an image) will ultimately produce the specific outputs (such as a verbal label)
- Once the network has been successfully trained (by tuning the weights), the “knowledge” is embedded in the behavior of the network, specifically, its ability to respond with the correct output to a given input
- Knowledge is stored as a pattern of activation of the nodes across the entire system
- Different information is stored as different patterns of activation
What happens when a computer is trying to learn whether something belongs to a certain category or not?
- According to the processing of many features of an image, it’ll determine whether something does or doesn’t belong to a category
- Each time the computer guesses correctly, its connections become stronger
- Each time the computer guesses wrong, its connections become weaker
- Through this process, the computer will teach itself to learn what elements of an image correspond to members of a certain category and become quicker and more efficient at identifying them as such
What are some of the benefits of neural network models for human cognition and artificial intelligence?
- For psychologists, because the networks are modeled after the brain, they can explain some phenomena that more traditional cognitive models can’t do as easily
- Because knowledge is stored as a pattern of activity across a large number of units, connectionist networks can withstand some loss of units with limited negative effects -> graceful degradation
What’s graceful degradation?
- A property of networks in which damage to part of the network results in relatively few deficits because information is distributed across the network and no single node contains information
- A network (or brain) doesn’t lose all function as a result of restricted damage, although some limited deficits do occur
- Ex: patients with brain damage as a result of herpes simplex virus encephalitis often display category-specific deficits of semantic knowledge
- Commonly observed with damage to the human brain
What are category-specific deficits of semantic knowledge?
- A specific loss of semantic knowledge from one category but not another as a result of brain damage
- One of the most common category-specific deficits is one in which a patient loses knowledge of living things but retains knowledge of non-living things
- This pattern of deficits can be explained if we think of knowledge as a distributed pattern of activity across cells -> similar items likely result in similar patterns of activation in the brain. As some cells are damaged, the pattern of activation changes in the network, but not all knowledge is lost because there are still many functioning cells
- Because one living thing is more similar to another living thing than it is to a non-living thing, it’s likely that the damaged cells affect items within a category but spare items that belong to a different category
Describe the research by Lambon Ralph, Lowe, and Rogers (2007) on category-specific deficits of semantic knowledge
- The research used an ANN to simulate the pattern of deficits of a group of brain-damaged patients with a living-thing category-specific deficit
- They did this by altering the connection weights to mimic brain damage in an ANN that had previously been trained to recognize pictures of living and non-living things
- This ANN that previously could distinguish living from non-living things, developed a living-things deficit after the artificial “damage.”
- That is, a neural network model that previously was able to name a picture of a fox was no longer able to do so, but could still name a picture of a car
- This supports the idea that human knowledge may be stored as a distributed pattern of activity across neurons, and that connectionist models are a good way to approach knowledge representation in computers
Describe the so-called “black box” problem of neural network models
- Neural networks are notoriously difficult to explain or interpret once they have been trained
- While we can observe the responses of a neural network to a specific input, it’s very difficult to determine why it made the response that it did because the information is represented in the values of distributed weights, not meaningful semantic units
- You might think of this as a modern equivalent to the criticisms of behaviorism: input and output can be observed, but there is no explanation about what is happening in between
- This means that neural networks can end up reproducing the same theoretical challenges as the brains they were meant to replicate
Describe embodied cognition theories (aka grounded cognition)
- They try to capture this grounding of symbols by considering the interaction between the brain, body, and environment in shaping thought
- ## In its weakest form, researchers suggest the body influences cognition
Describe Katinka Dijkstra and her colleagues (2007) study on autobiographical memory retrieval and body position
- They demonstrated that autobiographical memory retrieval can be improved by matching body position at retrieval to body position at encoding (an extension of the encoding specificity principle)
- They asked their participants to describe memories of events that are associated with predictable body positions, such as going to the dentist or waving goodbye
- Participants’ recall accuracy was higher when their body positions during recall matched the typical body position of the event
- Ex: participants more accurately recalled their last visit to the dentist while reclining in a chair compared to when they were standing up
Describe the findings of Williams & Bargh (2008) study on the interaction between the body and thought
They found that participants are more likely to rate descriptions of people as “warm” after holding a hot cup of coffee compared to holding a cup of iced coffee
Describe Barsalou’s (2008) idea of how the body has a causal role in thought
- He proposed an embodied theory in which cognition is grounded in sensory experiences, and that these sensorimotor experiences are used to understand abstract cognitive processes
- According to this view, knowledge uses similar sensorimotor neurons as perception and action
- Ex: our knowledge of an apple relies on a distributed network of the modality-specific neurons that are used for visual, tactile and taste perception of the apple
- In this way, embodied cognition is similar to connectionism
- Knowledge isn’t stored as a single representation of a concept, instead, it is held as a distributed pattern of activity across many neurons
Describe Schnall, Zadra, and Proffitt (2010) findings for their study on the connection between action and thought
In one experiment they demonstrated that participants judged the incline of a hill to be steeper if they had low energy (reduced blood glucose) compared to participants who did not have low energy
Describe the influential experiment by Hauk, Johnsrude and Pulvermüller (2004) that provided support for the idea of distributed, modality-specific knowledge representation
- Using fMRI, they first observed brain activity in a group of participants while they were moving their tongues, their fingers and their feet
- Due to the topographical structure of the motor cortex, motor actions elicited a predictable pattern of brain activity in motor areas of the brain
- Later, they asked participants to silently read words that are associated with actions by the tongue (lick), finger (pick), and foot (kick)
- They found that doing actions and silently reading action words activated similar motor areas of the brain - Participants weren’t performing any actions while silently reading, yet motor areas were still active when reading the action words
- This finding lends support to the claim that knowledge is stored as modality-specific neural activity
How is the motor cortex organized in the brain?
The motor cortex is topographically organized such that adjacent body parts are controlled by adjacent areas of the brain
What’s a major difference between embodied theories of knowledge and classic symbolic cognitive theories?
- That according to embodied theories, knowledge is goal-driven, flexible, and context-dependent
- Classic theories, such as the prototype theory and semantic network models consider knowledge to be abstract and independent of context
- According to such symbolic cognitive theories, all our knowledge about “BIRDS,” for example, is stored together and accessed when we think about birds
- An embodied view suggests that the information we retrieve from semantic memory depends on the context we are in at the moment, such that we more easily access knowledge that is relevant to the current context
- It suggests that the knowledge you access about birds will be quite different when sitting on the beach compared to sitting down for Thanksgiving dinner
Describe Zwaan, Stanfield, and Yaxley (2002) study that provided evidence that knowledge access depends on context
- They had participants read simple sentences about objects (e.g., “The chef saw the egg in the refrigerator”) and then showed them a picture of an object
- Participants’ task was to press one button if the picture had been mentioned in the sentence, and press another button if the picture had not been mentioned in the sentence
- The important manipulation was that 1/2 of the trials, the picture matched the shape of the object implied by the sentence (e.g., a whole egg), and in the other half of the trials, the picture didn’t match the shape implied by the sentence (e.g., a broken egg)
- Participants were faster to respond “yes” if the picture matched the shape that was implied by the sentence
- Ex: they were faster to respond if shown a whole egg rather than a broken egg because an egg is usually found whole in the fridge, not broken
- These results contradict a traditional symbolic cognitive explanation of semantic knowledge because if all knowledge of an item is stored together in a node or category, it shouldn’t matter what shape the object is
- An egg is still an egg if it’s broken, so participants should be able to recognize an egg just as easily regardless of the context
- Instead, it seems as though the sentence acted to make participants think about an egg in the way they would see it in that context -> participants’ knowledge about eggs was context-dependent
Describe Elizabeth Warrington’s (1975) findings on patients who displayed a loss of semantic memory
- These patients were unable to name objects in pictures or describe the characteristics of those objects, but they had normal intelligence, perception, and language abilities
- Today, these patients would be diagnosed with semantic dementia
What’s semantic dementia?
- A type of dementia characterized by a progressive loss of semantic memory leading to deficits in naming ability, comprehension in language, and object use
- Semantic dementia is a progressive neurodegenerative disease characterized by an inability to name objects
-Patients are unable to name objects presented visually, verbally, or by touch because they have a deficit with the knowledge itself, not with processing input from one of the senses - These patients progressively lose all semantic knowledge
- Their speech is fluent but devoid of content because they have lost semantic representations of the world
- Patterson et al. (2007) described a patient with semantic dementia who was able to remember the route to a friend’s house that he hadn’t visited in many years, but asked his wife, “What are those things?” as he passed sheep beside the road on the way there
What part of the brain is associated with semantic dementia?
- Semantic dementia is associated with the degeneration of neurons in the anterior temporal lobe
- Because of the co-occurrence of ATL atrophy and symptoms of semantic dementia, researchers have proposed that semantic knowledge is localized to ATL
Why is the localization of semantic memory in the brain not as simple as saying it’s located in the ATL?
- First, damage to ATL doesn’t always result in the same pattern of semantic knowledge loss
- Ex: patients with neural degeneration resulting from the herpes simplex virus often display category-specific semantic memory deficits. These patients also display ATL atrophy, yet their conceptual knowledge loss is much less severe than that observed in patients with semantic dementia
- If semantic memory is localized to ATL, you would expect all patients with similar damage to display similar behavioral impairments
- Furthermore, neuroimaging research on healthy brains rarely demonstrates ATL activation during tasks involving semantic memory
- Instead, neurologically intact brains display widely distributed activity across a variety of semantic processing tasks
Describe the hub-and-spoke model proposed by Karalyn Patterson, Mathew Lambon Ralph and their colleagues
- A model of knowledge representation
- According to this model, generalized and abstract semantic knowledge is stored in a semantic memory hub in the ATL
- Ex: this is where your general knowledge of “apple” would be stored, including all the places you could find an apple and all its different uses
- In addition to this abstract knowledge hub, context-dependent, and modality-specific detail about items is stored in “spokes” that are distributed across the cortex
- Ex: what different apples look like would be stored in visual processing brain areas, what an apple tastes like would be stored in taste perception cortical areas, and how to hold an apple would be stored in motor cortical areas
- This model can help to explain the seemingly contradictory results from brain imaging studies and patients with semantic dementia, as well as bring together old and new ideas from cognitive psychology about how semantic knowledge is processed
Describe how transcranial magnetic stimulation (TMS) has provided evidence that supports a hub-and-spoke organization of semantic knowledge in the brain
- TMS is a noninvasive technique that uses a magnetic field to stimulate cortical neurons in a localized part of the brain
- When neurons are stimulated using TMS, their normal activity is disturbed, creating what is sometimes called a “virtual lesion.”
- When the magnetic field is removed, neural function returns to normal
- Probic, Jefferies, and Ralph (2010) applied TMS to the ATL and inferior parietal lobule (IPL) of healthy participants while participants named pictures of both living and non-living things
- When the ATL was stimulated, the time it took participants to name all pictures of objects increased
- This supports the role of the ATL as a general semantic hub because it was difficult for participants to retrieve all types of semantic information when this area was not functioning properly
- When the IPL was stimulated, the time it took participants to name non-living things increased, but there was no effect on the naming speed of living things
- On further analysis, it turned out that naming speed was only slowed for non-living things that could be manipulated with the hands (such as a pen) but not for non-living things that aren’t typically manipulated with the hands (such as a sofa)
- These observations support the role of the IPL as a modality-specific spoke for items that we interact with by grasping
- Together, the results provide some compelling evidence that semantic knowledge is stored in both a localized and distributed way in our brains
What is the function of the inferior parietal lobule (IPL)?
IPL is a cortical region that is known to be involved in visually guided hand movements and corresponds to a spoke in the hub-and-spoke model
What was Aristotle’s take on imagery?
Aristotle believed imagery was central to thought and, in fact, wrote that “It is impossible to think without an image”
What did early philosophers question about imagery?
They questioned whether images were mental copies of the world or whether they were something else entirely
What was John B. Watson’s perspective on imagery?
He called imagery “sheer bunk” and instead suggested that what we experience as imagery can be better described as over-practiced language
What’s one difficulty for the scientific investigation of imagery?
- That it’s inherently introspective and cannot be verified by others
- For this reason, the study of imagery was avoided by behaviorists and wasn’t seriously studied again until cognitive psychology was established
What’s mental imagery?
- The experience of mentally creating a perceptual experience in the absence of a physical stimulus
- Imagery is possible for all our sensory modalities
- You can create mental images of stimuli that you have never experienced
What kind of mental imagery are you experiencing when you get a song stuck in your head?
Auditory imagery
What kind of mental imagery are you experiencing when you imagine the soft touch of a feather lightly brushing against your arm?
Tactile imagery
What kind of mental imagery are you experiencing when you imagine the smell of freshly baked bread?
Olfactory imagery
What kind of mental imagery has received the most attention in cognitive psychology, and the associated research has produced prolific results?
Visual imagery
Why do we say questions about the nature of imagery are questions about the nature of knowledge?
- Part of knowing what an apple is, is knowing what it looks like, tastes like, and smells like
- When you think of an apple, you may access abstract knowledge of apple-ness, but you may also create a mental image of an apple
The idea that thought could be represented by words and mental pictures was popularized by what theory from Allan Paivio?
Dual coding theory
Describe Allan Paivio’s dual coding theory
- A theory about knowledge representation that proposes knowledge can be stored as an abstract verbal code or an analog imagery-based code
- Paivio’s theory proposed that human knowledge was represented by 2 separate systems: a verbal system and a nonverbal, imagery system
- One key feature of a verbal system is that it is a type of abstract code
- The dual coding theory was influential in shaping the way cognitive psychologists thought about imagery, however, it wasn’t a theory about imagery, per se
- For Paivio, imagery and language are 2 systems that we used to represent the content of thought, but it didn’t deal with the question of the nature of imagery itself
What’s an abstract code?
- An arbitrary symbol system in which the symbols don’t resemble their real-world referent
- Ex: the word dog in English doesn’t in any way resemble a real-life dog. The letters D-O-G don’t look like a dog, they don’t sound like a dog, they don’t feel like a dog, and so on. They are simply a string of sounds that English speakers have agreed by convention will stand for the four-legged, furry, domesticated animal that is a dog
What are onomatopeia?
- A word that resembles the sound of the item it is referring to
- Ex: “quack” or “boom”
Why do we say onomatopoeia is more of an exception than a rule to the abstract code?
- It can be thought of as an attempt to mimic a sound rather than a word that denotes a particular meaning
- In any case, it is also true that onomatopoeia for the same sound can be quite different in different languages
- Ex: in English a turkey makes a gobble-gobble sound, but in Finnish a turkey says klu klu klu
Describe Paivio’s nonverbal imagery system in the dual coding theory
- It’s based on sensorimotor information and is modality-specific
- The information contained in an image is linked to specific sensory input and motor output in our bodies
- The consequence of this is that images resemble what they stand for in the world -> images are analog codes
What’s an analog code?
- A way to store information that resembles the physical stimulus being represented
- Analog codes retain the perceptual features of the physical stimulus they represent
- Ex: if you create a visual image of a dog, it will “look like” the dogs you have previously seen
What’s the imagery debate?
- A theoretical debate among cognitive psychologists about whether images are stored as pictures in our minds or as propositions
- It has largely been driven by 2 researchers: Stephen Kosslyn and Zenon Pylyshyn
- The imagery debate is not concerned with whether mental images exist or whether we can use imagery for other cognitive tasks
- The major debate concerns the format, or code, that imagery takes in our mind -> whether imagery uses a picture-like code or a language-like code
- Fundamental question: “What comes first: the image or the idea?”
- Both sides agree that propositions are used to represent knowledge, however, Kosslyn argues that imagery is just as fundamental to cognition as are propositions
Describe Stephen Kosslyn’s perspective on the imagery debate
- For Kosslyn, images are depictive representations
- Ex: if you imagine your kitchen, you know where everything is because it is stored in your semantic memory
What are depictive representations?
A type of analog code that maintains the perceptual and spatial characteristics of physical objects
What are descriptive representations?
A symbolic code used to represent knowledge that is abstract and does not resemble a stimulus in the real world
What’s the major difference between depictive and descriptive representations?
- Descriptive representations don’t preserve the perceptual and spatial information of your real kitchen; they contain only the conceptual information
- That is, while you know the toaster is to the right of the stove, only depictive representations maintain the relative distance between objects
Describe Zenon Pylyshyn’s perspective on the imagery debate
- For Pylyshyn our experience of mental imagery isn’t enough to tell us the true format that we use to store knowledge
- He claims that images are epiphenomena, or by-products of more fundamental cognitive processing
- Pylyshyn argues that cognitive processing fundamentally relies on manipulating cognitive symbols called propositions
- Propositional theory that proposes all knowledge is represented as a propositional code
- Propositions are able to describe the relationship between items
- The argument goes that a propositional code is the only code that’s needed for all thought with no need to complicate things with imagery or anything else, for that matter
- A proposition isn’t the same thing as language
- Propositions contain abstract conceptual knowledge which can be conveyed to someone else using language or images, both of which are argued to be secondary to propositions
What’s epiphenomena?
- A by-product that arises from a process but does not have a causal effect on that process
- Ex: imagine you’re typing an essay on your laptop while sitting on the bed. As you work, your laptop generates heat. The longer you work, the more heat is generated. By the end, you have typed a lot of words and generated a lot of heat, however, the heat didn’t cause the essay to get written. It was merely a by-product (or epiphenomenon) of computer processing
What’s a proposition?
An idea unit that can be verified as true or false
What’s one way that researchers have tried to resolve the imagery debate?
- By conducting experiments to observe whether people process images in the same way that they process real stimuli
- The idea is if images are depictive and maintain the perceptual and spatial characteristics of the real world, then people should process images and physical stimuli similarly
- If, on the other hand, images are epiphenomena of abstract propositions, then mental processing would depend on the number of propositions instead of perceptual and spatial characteristics of stimuli
One of Kosslyn’s earliest experiments used what kind of technique to investigate whether images did indeed maintain the spatial characteristics of physical stimuli?
A mental scanning technique
What’s mental scanning?
An experimental technique in which participants are asked to scan their mental images while response time is measured
Describe Kosslyn’s early experiment that investigated whether images maintain the spatial characteristics of physical stimuli
- He used a mental scanning technique
- He reasoned that if visual images are analog codes of physical stimuli, it should take longer to process larger mental distances than shorter distances, just as it would take more time to travel longer physical distances than shorter ones
- First, he presented participants with line drawings to memorize
- All of the objects in the drawings are elongated such that they have an obvious top and bottom or left and right
- Once the pictures had been memorized, participants were told to create a mental image of one of them and focus their attention on a part of the object at either the extreme top or bottom, left or right (specified by the experimenter)
- Next, participants were told to “look on” their mental image for another part of the object and press a button if they “saw” it in their image
Describe the results of Kosslyn’s early experiment that investigated whether images maintain the spatial characteristics of physical stimuli
- Participants’ reaction times to press the button perfectly matched Kosslyn’s prediction: The farther away participants had to shift their attention to find the new part of the object, the longer the search time
- This finding supports the idea that images are depictive representations that maintain the spatial arrangements of physical objects
- There is another possible explanation for the results: perhaps participants were storing information from the line drawings as a list of features and were searching through the memorized list rather than “looking” at a mental image
- The more items on the list to search through, the longer the reaction time, just like Kosslyn found
- The results of Kosslyn’s first experiment could, therefore, be explained equally well as depictive representations or as propositions
How can both a depictive representation theory and propositional theory make the same prediction about reaction time in Kosslyn’s early experiment on the imagery debate?
Because it should take participants longer to identify whether the flower has petals than whether it has leaves, either because there’s a longer distance to travel (depictive representation) or because there are more attributes to search through (propositional representation)
Describe the second experiment Kosslyn conducted on the imagery debated
- Participants memorized a drawing of a map with 7 different landmarks
- Participants were told to visualize one of the landmarks, then to scan their mental image until they “arrived” at another landmark
- In this experiment, the distances between landmarks differed but there were never any additional landmarks or properties in between them
- The results showed that, similarly to Kosslyn’s original experiment, reaction time to mentally travel between landmarks increased as the physical distance between them increased
- Because the distance between landmarks varied but the number of properties remained constant between all landmarks, Kosslyn could conclude that visual images maintained the relative distance of the real picture and were not influenced by the number of landmarks present
Describe Shepard and Metzler’s (1971) mental rotation experiment
- Designed to investigate the time it took to mentally rotate mental images of abstract figures
- They reasoned that if the mind is performing a process that is fundamentally similar to the rotation of real objects then we can make the following prediction: the bigger the angular distance between the two objects that need to be compared, the longer it should take to compare them
- In the real world, the more you have to rotate an object, the more time it takes to do so
- Therefore, if we were doing something like rotating the object in our mind’s eye it should take us longer to respond when the angular distance is greater
- In their experiment, participants saw 2 drawings of 3D objects and were asked to identify whether they were the same or different
- In some of the trials, 2 identical shapes were presented but one of the objects in the pair had been rotated around a vertical axis (“same” objects)
- In the other half of the trials the two objects were different from each other
Describe the findings of Shepard and Metzler’s (1971) mental rotation experiment
- Shepard and Metzler reported a remarkably linear relationship between the amount of angular rotation and the time it took participants to determine that the shapes were the same
- They were able to work out that participants could mentally rotate the objects at a rate of about 60° per second
Why are the findings from mental rotation experiments important for the imagery debate?
Because what participants were doing mentally is similar to what would happen if they had been physically rotating real objects: the greater the distance, the more time it takes to rotate the object
Describe Kosslyn’s (1975, 1978) study on mental scaling
- He asked participants to imagine various animals standing next to either an elephant or a fly -> this was done to establish the scale of the animals
- Kosslyn then asked participants about the properties of the animals
- Ex: “Does the mouse have claws?”
- Participants were faster to answer the questions when the animal was being imagined next to a fly (and was relatively big) compared to when it was imagined next to an elephant (and was relatively small)
- Kosslyn reasoned that participants needed to mentally “zoom in” to “see” the details of the relatively small mouse standing next to an elephant
- This mental zooming takes time, just like walking toward a real mouse to see details does
- If the mouse is next to a fly, the details can easily be seen (both on the real object and image) because it is relatively big, so reaction time is faster
What did Kosslyn do in his mental scaling study to establish that the increased reaction time was caused by the relative size of the image and not something else (like the number of facts known about the animals)?
- He had participants imagine a mouse standing next to an elephant-sized fly and a fly-sized elephant
- This time, the reaction times were completely reversed: participants responded faster to the question about the mouse when it was standing next to the elephant compared to when it was imagined next to the fly
- Reaction time to answer questions about the visual details of animals depended on the relative size of the image
- Participants’ responses were faster for the larger images
- This pattern of responses supports the idea that images are processed similarly to real objects
One of the earliest demonstrations supporting the claim that imagery and perception share mechanisms comes from what?
- A classic experiment by Perky (1910)
- She asked participants to create visual images of everyday items, such as a book or a lemon, while she projected very dim pictures of those same items on a screen in front of participants
- The remarkable finding was that participants described their images to match the pictures being projected (for example imaging a blue book), but they had no idea they were actually seeing the objects
- Participants mistook their perception for imagery
- This isn’t uncommon if the real stimulus is weak enough
Describe the experiment by Segal and Fusella (1970) on the interference of perception by imagery
- They had participants perform a perceptual detection task while simultaneously performing an imagery task
- Participants were told they would either see a picture of a blue arrow (the visual stimulus), hear a note played by a harmonica (the auditory stimulus), or there would be nothing at all
- It was the participants’ task to indicate which stimulus was present, if at all
- Both the visual and auditory stimuli were presented at very low intensities making the detection task quite difficult
- While completing the detection task, participants were instructed to either visualize a tree or the sound of a telephone ringing (back when all telephones made the same ringing sound!)
- Results: detection rates of visual stimuli were lower when participants were imagining a tree compared to when they were imagining the sound of a telephone, and detection rates of auditory stimuli were lower when participants were imagining the sound of a telephone compared to visualizing a tree
- Visual imagery interfered with visual perception and auditory imagery interfered with auditory perception
What do the results of the experiment by Segal and Fusella (1970) on the interference of perception by imagery show?
- These results support the claim that imagery and perception share the same mechanisms
- The idea is that if imagery uses the same mechanisms as perception, then imagining a visual stimulus, for example, would “use up” some of the processing mechanisms available for vision
Describe Farah’s study (1985) on how imagery can facilitate perception
- She showed participants very faint pictures of either the capital letters T or H or participants saw blank slides
- Just like Segal and Fusella (1970), participants performed a detection task while simultaneously performing an imagery task, except this time only visual information was used
- Participants were instructed to create a visual image of either the capital letter T or H while performing the detection task (detect an actual letter T or H)
- Farah found that in this case, imagery facilitated perception
- Participants were more accurate at detecting the same letter that they were imagining compared to when they were imagining the other letter
- This experiment showed that when only visual stimuli were used, imagery could help perception by giving it a boost
What are motion aftereffects?
- A perceptual experience that occurs after exposure to motion in one direction in which a static scene appears to move in the opposite direction of the previously viewed motion
- Visual aftereffects are a type of visual illusion that occur after prolonged viewing of a visual stimulus and are known to result from the activity of cells in the visual system
Describe Winawer and colleagues (2010) study on motion aftereffects
- They had participants imagine motion in one direction for 60 seconds
- They found that imagining motion was enough to later bias participants’ perception of motion and create a motion aftereffect
- Since a motion aftereffect occurs because of activity of cells in the visual system, it can be inferred that imagery makes use of those same cells
How does introspection tell us that we don’t experience imagery and perception in the same way?
- With the exception of a few unusual (or experimentally contrived) situations, we can tell the difference between when we’re imagining something and when we’re actually perceiving it
- Imagery tends to be weaker and more fleeting than perception
What does science rely on?
Falsification
What’s falsification?
- A key principle in science in which theories are tested in order to prove they are false, instead of searching for evidence to confirm a hypothesis
- If a researcher finds evidence that confirms their theory, they can’t be sure if it is always true or just true in that particular experiment
- If, on the other hand, a researcher finds evidence against their theory, they can conclude that the theory isn’t supported
Some of the early evidence against a depictive explanation of imagery comes from research that investigates what?
Research that investigates imagery of ambiguous figures
Describe Reed’s (1974) study on the imagery of ambiguous figures
- He asked participants to memorize pictures of ambiguous figures
- Participants then had to indicate whether new figures were part of the original picture by relying only on their memory
- If participants’ mental images were depictive representations, then they should easily be able to indicate whether the new shapes were part of the original from memory, just as easily as they could if the picture was drawn on a paper in front of them
- Participants were able to accurately indicate whether the new shapes were part of the original in some cases, but accuracy was quite low in other cases
- Reed argued that the results could be explained if participants were giving the picture verbal labels instead of storing their spatial characteristics
- If participants labeled an ambiguous shape as “two overlapping triangles” or “four triangles and a center diamond,” then they can easily identify whether certain shapes (triangles and diamonds) were part of the original
- Participants would have difficulty identifying a parallelogram as part of the original because there’s nothing in the verbal label about a parallelogram
- If participants were storing the image as a depictive representation with the same spatial characteristics as the original picture, then all the new shapes would be equally easy to identify as part of the original
- Because not all of the new shapes were as easily identified as part of the original, this suggested that images may be stored using meaningful verbal labels rather than depictive representations
What do other arguments against depictive representations claim that the findings that support depictive representations are a result of?
- Experimenter expectancy
- Demand characteristics
What’s the experimenter expectancy effect?
An effect in which an experimenter may unconsciously communicate to participants their expectations about what they expect the results to be, and in turn, causing the participant to unconsciously behave according to the experimenter’s expectations
What are demand characteristics?
Subtle cues in experimental tasks or instructions that may bias participants’ behavior
How did Pylyshyn counter-argue Kosslyn’s map mental scanning experiment?
- When Pylyshyn asked participants to imagine that you are standing at the tree with the light on. Now imagine that the light at the tree turns off at the exact same time as the light at the lake turns on, he found there was no relationship between the distance between landmarks and the time it took them to “see” the light turning on at various locations
- He argued that the reason Kosslyn et al. (1978) found a relationship between distance and time in their map mental scanning experiment was because participants assumed they were supposed to act “as if” they were travelling around a map and behaved accordingly
- Participants knew that it should take longer to travel farther distances, so they played along, so to speak
- Pylyshyn argues that because participants’ performance varied depending on the task details, it demonstrates that Kosslyn’s results only supported depictive representations because that’s what participants thought they were supposed to do
-Had the instructions been different, perhaps the results would have been different too
Describe Intons-Peterson (1983) study on mental rotation and image scanning
- She trained 4 undergraduate research assistants to test participants using tasks such as mental rotation and image scanning
- However, she told 2 of the research assistants that she expected reaction times would be slower overall for imagery tasks compared to tasks where participants had the physical stimulus to manipulate
- She told the other 2 research assistants the opposite: that reaction time would likely be faster for imagery tasks compared to when the physical stimulus was present
- Even though the research assistants read all participants the same instructions from a printed script, the results ended up matching the research assistants’ expectations
- Intons-Peterson, therefore, demonstrated clearly that the results of imagery experiments could be affected by experimenter expectancies
When did a major shift in the imagery debate occur?
- When cognitive neuroscience developed ways to observe participants’ brains while engaged in processing tasks
- Now researchers had a way to observe whether imagery and perception actually used the same brain mechanisms
Until the development of modern neuroimaging techniques like positron emission tomography (PET) and functional magnetic resonance imaging (fMRI), the only way to learn about brain function was by doing what?
By studying patients with brain damage
Describe Policardi and colleagues (1996) analysis of patient TC
- Patient TC suffered cardiac arrest after a car accident that left him with damage to his occipital and temporal lobes, 2 areas that are known to be important for visual perception
- As a result of this damage, TC suffered from cortical blindness
- TC was completely unable to distinguish light from dark
- When an object was moving quickly towards his eye, he failed to move his head or even blink
- This loss of conscious vision was accompanied by a loss in visual imagery
- He was unable to provide any visual descriptions of the place he lived, he performed poorly on tasks that required him to answer questions about visual details of objects and animals (ex: “Is the tail of a mouse long relative to the size of its body?”), and was unable to name the color of common objects from memory
- Here we have a case where damage to the brain results in similar deficits in both perception and imagery abilities
Describe Zago and colleagues (2010) analysis of patient PB
- Patient PB had damage to his occipital cortex as a result of a stroke and suffered from cortical blindness similar to patient TC
- However, unlike patient TC, PB performed normally on the same imagery tasks that TC wasn’t able to do
Describe Bartolomeo and colleagues (1998) analysis of Madame D
- Madame D suffered damage to a portion of the brain bordering the occipital and temporal lobes due to multiple strokes
- She was able to see and could copy drawings, however, she could not read or visually recognize objects or faces
- She also complained of a lack of color vision
- Despite these visual impairments, her visual imagery remained remarkably unimpaired
- Although she couldn’t recognize objects by sight, she was able to draw them from memory
-She performed quite poorly on laboratory tests of visual ability but received perfect scores on the analogous imagery tasks - She reported that if she wasn’t able to recognize an object at her home, she would visualize the objects that were typically found in that location to help her recognize the objects actually present -> she used her intact imagery to facilitate her impaired perception
Describe the group of researchers’ analysis of the two patients who lost their ability to use mental imagery after suffering closed head injuries
- Despite deficits in imagery, both patients scored normally on all tests of visual perception, memory and language
- To illustrate the dissociation between perception and imagery in these patients, neither of them were able to draw animals or objects from memory (suggesting difficulties with imagery), however, both were able to accurately copy line drawings of the same objects
What’s one difficulty with drawing conclusions from patients with brain damage?
- The damage is rarely localized to a particular brain region and it’s extremely unusual to find multiple patients with the same pattern of damage
- That makes it difficult to draw conclusions about which brain regions may be required for perception and which also support imagery
