PSYCH 243- Cognitive Processes (Notes)

Chapter 7: Mental Imagery & Cognitive Maps

Mental Imagery

• Perception uses previous knowledge to attain and interpret the stimuli registered by the senses.
  • Involves both top-down and bottom-up processes.
• Mental Imagery: mental representations of stimuli when they are not physically present.
  • Useful in creative tasks and problem-solving.
  • Involves top-down processing only.
• Imagery extends beyond visual stimuli to include auditory, tactile, olfactory, and gustatory experiences.
  • Lemon experiment: participants vividly imagine slicing a lemon and bringing it close to their mouth. Participants frequently experienced strong multi-sensory reactions.
• The Imagery Debate (Kosslyn et. al., 2006): Are mental images represented and processed in the brain as analog or propositional code?
• Analog code: images resemble perception.
  • Encoded as pictorial representations.
  • Supported by research involving mental rotations and spatial tasks.
• Propositional code: images resemble language.
  • Encoded as an abstract language-like representation. (e.g. “triangle”)
  • Supported by studies with complex or ambiguous figures.
• Neuroimaging research indicates that imagining an object and actually seeing the object both activate the primary visual cortex.
• Research on visual imagery is difficult to conduct because mental images cannot be directly observed and fade quickly.
  • However, if we use analog coding, people should make judgments about mental images of objects similar to how they do about physical objects. This involves four broad characteristics: rotation, distance, size, and shape.
• Mental Rotation Tasks (Shepard & Metzler, 1971)
  • Mental rotation of 3D objects increased with the number of rotations (similar to physical objects), supporting the analog view.
  • Deaf individuals fluent in ASL were more skilled in mental rotation tasks due to experience mentally rotating signs. (Emmorey et al, 1998)
  • Kosslyn et al (2001) demonstrated that mental rotation engages the motor cortex, supporting the analog encoding view.
    • One group physically rotated objects. Another group passively watched objects being rotated. Both later mentally rotated objects, but only the first group displayed activation of the motor cortex.
• Imagery and Distance (Kosslyn et al, 1978)
  • Scanning time on mental maps increased with distance, as if they were looking at an actual map.
• Imagery and Size (Kosslyn, 1975)
  • Mental images of objects differ in perceived detail based on relative size. (e.g., The mental image of a rabbit had more detail when imagined next to a fly versus an elephant.).
• Imagery and Shape (Paivio, 1978)
  • Judgments about angles (clock hands) took longer when angles were similar, consistent across both physical and mental stimuli.
  • Participants were further tested based on their mental imagery skills. High imagery ability participants were able to make quicker angle-comparison judgments.
• To summarize, these experiments demonstrate that mental imagery behaves similarly to real-world perception, suggesting analog coding:
  • When people rotate a mental image, a large rotation takes them longer, just like when they rotate a physical object.
  • People make distance judgments in a similar fashion for mental images and physical stimuli for visual and auditory stimuli.
  • People make decisions about shape in a similar fashion for mental images and physical stimuli (angles formed by hands of a clock).
• Mental imagery and physical images can interfere with one another due to competing overlapping cognitive resources between perception and mental imagery.
  • Segal & Fusella (1970) demonstrated this: visual imagery interferes visual perception, auditory imagery interferes auditory perception.
• Ambiguous figure tests suggest:
  • We use analog coding for simple stimuli.
  • We use propositional coding for complex or ambiguous stimuli.
• Imagery and ambiguous figure
  • Participants shown an ambiguous figure (Star of David) were were unable to identify a simpler sub-shape (a parallelogram) using mental imagery, suggesting it was encoded propositionally. (Reed, 1974)
  • Participants were shown an ambiguous figure (duck-rabbit) and could not later give another interpretation from mental imagery. However, participants could reinterpret the image after drawing it from memory. This suggests propositional encoding for ambiguous or complex figures. (Chambers & Reisberg, 1985)
• Propositional coding helpful when stimuli are complex. For example, verbally describing a jigsaw puzzle piece.
• Participants imagined the letter ‘X’ superimposed on the letter ‘H’ and reported new shapes, like a bowtie or a right triangle, supporting analog coding. (Finke et al, 1989)
• Concrete words are generally easier to imagine than others. For example, “elephant” is encoded both propositionally and analogously, while “honesty” is only propositional.
• Neuroimaging research shows overlapping brain activation (especially the primary visual cortex) during both visual perception and mental imagery, further supporting analog encoding. (Kosslyn et al, 1996)
• People with lesions in the primary visual cortex both fail to register visual images and fail in visual imagery tasks, indicating deficits in visual imagery and perception are linked.
• Summarizing the analog vs propositional debate:
  • The majority of scientists support analog coding. However, Pylyshyn argues for propositional coding.
  • Resolution: Dual Coding Hypothesis. We probably do a mixture of bot — analog coding for most stimuli but propositional for complex or ambiguous stimuli.

Cognitive Maps

• Cognitive Maps: Mental representations of geographic information and our environment.
• Spatial Cognition: Broader interdisciplinary field studying thoughts about spatial relationships, navigation, object tracking, and cognitive maps.
• Strategy for remembering where you parked your car: Face the car as if you’re returning, note landmarks / cues.
• Factors influencing our estimate of distance: number of intervening cities, semantic categories, landmark vs. non-landmark destinations.
• Number of intervening cities distortion: People perceive distances on a map as longer when more intervening cities were present. (Thorndyke, 1981)
• Semantic categories distortion: Buildings that are semantically or functionally related are perceived as physically closer together.
• Landmarks vs. Non-landmarks distortion: We estimate that landmarks (sites known to us) are closer than non-landmarks. (e.g., From NYC, Philadelphia might seem closer than an unknown town in New Jersey.)
• Cognitive maps apply heuristics to shape and position, regularizing small inconsistencies and creating more idealized maps.
• 90-degree heuristic: Angles in cognitive maps are regularize. Street intersections are remembered as closer to 90 degrees.
• Symmetry heuristic: Geographical features and curves are imagined as more symmetric and balanced than their true forms. (e.g. Recalling the curve of a coastline as smoother and more symmetrical.)
• Rotation heuristic: Geographic features (states, rivers, etc.) are mentally rotated to appear more vertical or horizontal.
  • Example: Reno, Nevada, is perceived as east of San Diego, California, though it is actually west.
• Geographic features are mentally aligned, making locations seem more directly aligned than they are.
  • Philadelphia is perceived as further north than Rome. We tend to align the US with Europe more than it actually is.
  • Windsor, Canada is perceived further north than Detroit, Michigan. We tend to think of Canada as north of Michigan.
• Distinguishing between Alignment Heuristic and Rotation Heuristic:
  • Alignment involves adjusting two or more structures relative to each other. (e.g., aligning Europe and the U.S.)
  • Rotation involves rotating one structure to appear more vertical or horizontal. (e.g., rotating Nevada and California mentally)
• People have reasonable self-awareness (metacognition) regarding their spatial abilities, though systematic errors remain in estimating distances and angles.
• Edward Tolman (1948) first proposed cognitive maps based on experiments with rats navigating mazes.
  • Later research extended Tolman, showing humans also use cognitive maps based on one’s position relative to the environment.
• Cognitive maps can also derive from verbal descriptions.
• The Spatial Framework Model (Franklin & Tversky, 1990) explains how we mentally organize space around our bodies, emphasizing certain directions over others.
  • Subjects read narratives describing the location of objects around a either a standing or reclining observer and then asked to identify what was located beyond head or feet, front-back, or right-left.
  • In general, quickest response time for above-below, then front-back, and slowest for right-left.
  • This is because we naturally prioritize vertical space (due to gravity and bodily orientation) and front-back directions (linked to vision and forward movement), whereas left-right distinctions are less instinctive and thus slower to process.

---

April 9, 2025

Segal & Fuseslla (1970) slide will note be on exam if it’s not on the study guide. Professor didn’t have the accompanying slides to explain it—there’s some sound that goes with it.

Imagery and an ambiguous figure (Reed, 1974). Reed tested people’s ability to decide whether a specific visual pattern was part of a design they had seen earlier. For example, the Star of David was shown. They were later asked if the star contained a parallelogram (it does). Most people respond, “No.” The big debate is if they’re coding it as an analog image. If they did, they would be able to re-visualize it and see the parallelogram. Propositional (sentential) or analog (iconic) format.

Ambiguous figure (Chambers & Reisberg, 1985). Fifteen participants were shown the duck-rabbit and were asked to create a clear mental image of it. They were later asked to give another interpretation of the figure. None could do it, suggesting they could not consult a stored mental image. Then, participants drew the figure from memory and all fifteen participants could reinterpret the redrawing as the opposite of their prior interpretation. This suggests we may use propositional codes for ambiguous or more complex figures and use analog codes for simple figures. i.e. We use propositional codes like “duck” or “rabbit” for more complex figures.

Ambiguous figure tests suggest:

• for simple figures we store analog
• For more complex we store propositionally

Finke et al (1989) asked participants to imagine the letter ‘H’ superimposed by the letter ‘X’. Could people see new shapes in this configuration? “Yes” supports analog encoding.

Some words are easier to visually imagine than others. For example, elephants vs. honesty.

The Dual Coding Hypothesis:

• The analog/propositional debate continues, but most believe both encodings are used.
• But the majority of research supports the analog viewpoint (Kosslyn & colleagues).
• Pylyshyn’s propositional viewpoint argues that analog images are not necessary, central component of mental imagery.
• At present, the analog code explains most stimuli and most tasks. However, for some kinds of stimuli, people may use propositional codes (for complex, ambiguous stimuli).

People with prosopagnosia cannot visually imagine faces.

Kosslyn and co-authors (1995) used PET scans and found that imagining letters activated the primary visual cortex— the same areas activated when viewing the letters.

April 23, 2025

Chapter 7, Mental Imagery

• Cognitive Maps: Our metacognition about our spatial ability is generally good, but we also make incorrect judgments about distance and angles.
• Rotation Heuristic: involves rotating one structure or location clockwise or counterclockwise to make it more vertical or horizontal than it really is. San Diego and Reno. (This will be on exam.)
• Alignment Heuristic: mentally manipulating two or more structures or locations relative to each other. We tend to align the US with Europe more than it actually is, assuming Philadelphia is north of Rome. We tend to think of Canada as north of Michigan, therefore assuming Windsor is north of Detroit. (This will be on exam.)
• Edward Tolman (1948) first proposed we create cognitive maps. Proposed that rats make cognitive maps based on features and cues relative to each other in their spatial environment.
• We also use verbal descriptions to make cognitive maps. e.g. giving directions to someone based on street names and landmarks.
• Spatial framework model:

Chapter 8: Cognitive Processes

General Knowledge and Semantic Memory

• General knowledge is fundamental information that informs cognitive processes.
• Semantic memory stores organized knowledge about the world, such as general facts. (Distinct from episodic memory, which is about personal experiences.)
  • Encyclopedic knowledge: general facts (e.g., Albany is the capital of New York).
  • Lexical knowledge: knowledge about words, meanings, and grammar.
    • There are about half a million to a million English words.
  • Conceptual knowledge: categories, concepts, and their interrelations, enabling reasoning and inference. (e.g., knowing a chair is for sitting).

Concepts and Categories

• Categories and concepts are essential components of semantic memory, enabling us to divid the world into categories and make sense of our knowledge.
• Categories group similar objects (e.g., fruits: apples, oranges).
• Concepts are mental representations of categories, enabling us to:
  • Code and store information efficiently by combining a variety of similar objects under one concept, reducing storage space.
  • Make inferences about new examples within known categories (e.g., understanding a new type of fruit is edible, a new type of chair can be sat in).

Models of General Knowledge Storage

• Prototype: comparison to an idealized representative of a concept.
• Exemplar: comparison to multiple concrete examples of a concept.
• Network: interconnected nodes (concepts), facilitating retrieval.

Prototype Approach (Rosch, 1973)

• Items categorized by comparison to a “prototype” or typical example (ideal representative).
• Prototypes vary culturally and temporally (e.g., the prototype of a cell phone has evolved).
• Prototypicality describes how typical an item is within its category. (e.g. robins are more typical than penguins or ostriches for the category of birds.)
• Prototypes share attributes in a family resemblance—category members share overlapping attributes without necessarily sharing a single defining feature.
• Levels of categorization: objects are hierarchically grouped at superordinate-level (most general), to basic-level (moderately specific), to subordinate-level (most specific.)
  • People most commonly identify items at the basic-level.
  • Example: Toy > Doll > Ragdoll
• Neuroscience research (PET scans) suggests different levels of categorization activate different regions of the brain. (Kosslyn et al, 1995)
  • Superordinate terms (e.g., “toy”) activate the prefrontal cortex, associated with language and associative memory.
  • Subordinate terms (e.g., “ragdoll”) activate the parietal region, associated with visual attention.

Exemplar Approach

• Categorization based on comparing new stimuli to multiple specific examples (exemplars) of a concept (chair), rather than a single prototype.
• Exemplars:
  • Are actual instances stored in memory, rather than abstract averages.
  • Better handle atypical examples (penguin as bird).
  • Likely preferred with increased familiarity.

Comparing Prototype and Exemplar Approaches

• Prototypes are simpler but rigid. Exemplars are more flexible and capacious.
• We probably use both exemplars and prototypes for categorization.

Network Models of Semantic Memory

• Focus on network-like interconnectedness among concepts rather than strict categorization
  • e.g. The concept of a chair depends on the concepts to which it is connected: legs, back, seat, etc.
• Three network models: Collins & Loftus (1975), Anderson’s ACT Theory (1983, 2000), Parallel Distributed Processing

Collins & Loftus’ Network Model (1975)

• Knowledge structured as a “cognitive network” of interconnected “nodes” representing concepts.
• Spreading activation: activating one node triggers “linked” nodes, facilitating memory retrieval.
• A spreading activation network can be represented as a network of nodes and edges, with shorter edges indicating ideas are more closely related.
  • The sentence verification task supports this: “A robin is a bird.” is processed more quickly than “A robin is an animal.” because the second sentence requires more travel time through the network.
• The typicality effect: faster reaction time when an item is a typical member of a family
  • explained by link strength: frequent activation strengthens connections.

Parallel Distributed Processing (PDP) Model

• Cognitive processes distributed across parallel neural networks.
• Knowledge stored as patterns of neural activity, not in singular nodes, enabling simultaneous (parallel) processing rather than sequential (serial) processing.

Schemas and Scripts

• Schemas: generalized knowledge about common situations or events.

• Scripts: specific schemas outlining structured sequences of familiar activities (e.g., grocery shopping steps).

• Better recall if schemas/scripts identified upfront, though schemas can lead to memory errors.

Schemas and Memory

• Memory Selection: Schemas guide recall, often enhancing memory for schema-consistent items (expected in context) but also schema-inconsistent items if they’re surprising.

• Boundary Extension: People tend to remember viewing fuller scenes than actually seen, extending boundaries based on their schemas.

• Memory Abstraction: Retaining the general meaning rather than exact details. Two approaches:

  • Constructivist: General meaning prioritized.

  • Pragmatic: Precise wording prioritized when details matter.

• Memory Integration: Final memory formation stage where schemas integrate information, sometimes leading to distorted memories consistent with schemas after a delay.

Summary of Key Points

• Schemas: Generalized knowledge, usually accurate heuristics, central to cognitive processing.

• Scripts: Structured sequences within schemas aiding organized recall.

• Boundary Extension: Memory often expands scene boundaries based on schemas.

• Memory Abstraction and Integration: Schemas influence how we abstract and integrate memories, sometimes causing memory inaccuracies.

Chapter 8, General Knowledge

• Most people will use the basic level before the ordinate or superordinate level.
• We probably use a combination of exemplars and prototypes for categorization.
• Compare PDP, prototypical, exemplar, and network.
• Do people remember information that’s inconsistent with schemas? Yes, you often remember things contrary to expectation. e.g. A dog running through a theatre.
• Schemas are a kind of heuristic, a general rule or strategy that is typically accurate.

April 30, 2025

• Know difference between fluent, non-fluent, and aphasia.