Introduction to Cognitive Science — Ch1: Foundations and Models of Mind

What Cognitive Science Is

Cognitive Science:
→ The interdisciplinary study of how mind and intelligence work
→ The 1956 "cognitive revolution": behaviourism → cognitivism
  The MIT symposium: papers by Chomsky, Miller, Newell and Simon
→ Core question: how is knowledge represented, processed, and used?

Contributing disciplines:
→ Cognitive psychology: perception, attention, memory, language, thought, problem solving
→ Artificial intelligence (AI): computational implementations of intelligent behaviour
→ Linguistics: linguistic knowledge and language processing
→ Philosophy: the nature of mind, representation, consciousness, free will
→ Neuroscience: the brain and the neural basis of cognition
→ Anthropology: the relationship between culture and cognition

The cognitive science hexagon:
→ Five disciplines linked to each other with differing strengths
→ AI at the centre: the tool that implements and tests cognitive theory

The information-processing paradigm:
→ Mind = an information-processing system (the computer metaphor)
→ Input → encoding → representation → processing → output
→ Serial processing vs parallel processing
→ Flow diagrams: making the processing stages visible

Representation:
→ The assumption that internal mental representations stand for the outside world
→ Propositional representation: linguistic and abstract (if-then rules)
→ Imagistic representation: perception-like, analogue
→ Distributed representation: connectionist, patterns of activation
→ Embodied cognition: the body takes part in representation

Perception and Attention

Perception:
→ The process that turns sensory input into meaningful experience
→ Bottom-up (data-driven) processing: starts from the properties of the stimulus
→ Top-down (conceptually driven) processing: filled in by expectation and knowledge

Gestalt principles:
→ Proximity: things close together are grouped
→ Similarity: things that resemble each other are grouped
→ Continuity: smoothly continuing lines are preferred
→ Closure: incomplete figures are perceived as complete
→ Figure-ground: separating an object from its background

Perceptual constancy:
→ Perception stays stable even as the stimulus changes
→ Size constancy: an object looks the same size at different distances
→ Colour constancy: colour looks the same under different lighting
→ Shape constancy: the same shape is seen from different angles

Attention:
→ Selectively allocating a limited pool of processing resources
→ Functions: selection, focus, vigilance, executive control

Early selection theory (Broadbent):
→ A bottleneck: sensory store → filter → only one channel passes
→ Selection happens first on physical properties (frequency, location)

Late selection theory (Deutsch and Deutsch):
→ Every stimulus is processed for meaning before selection occurs
→ The cocktail party effect: your own name is heard even on an ignored channel

Attenuation theory (Treisman):
→ Unselected channels are weakened rather than fully blocked
→ Familiar or meaningful stimuli can still get through

The spotlight model of attention:
→ Spatial attention illuminates a location the way a spotlight does
→ Exogenous shifts: automatic, driven by the stimulus
→ Endogenous shifts: deliberate, driven by goals

Inattentional blindness:
→ Without attention, even an obvious stimulus goes unseen
→ The gorilla experiment (Simons and Chabris): counting passes, missing the gorilla
→ Change blindness: failing to notice that a scene has changed

Feature integration theory (Treisman):
→ Preattentive processing: basic features (colour, shape) handled fast and automatically
→ Attentive processing: binding features together requires focused attention
→ Predicts the difference between pop-out and conjunction search

Memory Architecture and Processing

The multi-store model (Atkinson and Shiffrin, 1968):
→ A serial model: sensory memory → short-term memory → long-term memory

Sensory memory:
→ A very brief buffer for sensory input
→ Vision: iconic memory, about 0.5 seconds, large capacity
→ Hearing: echoic memory, 2–4 seconds
→ Sperling: whole report vs partial report experiments established its existence

Short-term memory:
→ Limited capacity (7±2, Miller), duration 15–30 seconds
→ Lost without rehearsal; forgotten quickly under interference
→ Prefers an acoustic code (Conrad)

Working memory (Baddeley and Hitch):
→ Recasts short-term memory as an active processing system
→ Components:
  Phonological loop: verbal information, subvocal rehearsal
  Visuospatial sketchpad: visual and spatial material
  Central executive: control, coordination, strategy
  Episodic buffer (2000): integrated storage

Long-term memory:
→ Unlimited capacity, potentially lifelong duration
→ Explicit (declarative) memory:
  Semantic: facts, concepts, knowledge about the world
  Episodic: personal experience, tied to time and place
→ Implicit (nondeclarative) memory:
  Procedural: motor skills and habits
  Priming: unconscious influence of prior experience
  Conditioning: Pavlovian and fear conditioning

Encoding, storage, retrieval:
→ Levels of processing (Craik and Lockhart):
  Deeper (semantic) processing → better memory
→ Elaboration: connecting to what you already know
→ Context-dependent memory: retrieval is better in the original context
→ Encoding specificity: the context of learning should match the context of recall

Forgetting:
→ Decay theory: unused traces weaken
→ Interference theory:
  Proactive interference: earlier learning disrupts new learning
  Retroactive interference: new learning disrupts what came before
→ Retrieval failure: the information is there but the cue is missing (tip-of-the-tongue)
→ Repression (Freud): threatening memories held unconsciously (heavily disputed)

Where Cognitive Science Is Heading

Connectionism:
→ Cognition as patterns of activation across a distributed network
→ PDP (parallel distributed processing, Rumelhart and McClelland)
→ Learned weights instead of explicit rules
→ The theoretical ancestor of deep learning

Embodied cognition:
→ Mind arises from brain, body and environment interacting, not the brain alone
→ Mirror neurons: watching another's action activates your own motor cortex
→ Embodied metaphor: up = good, down = bad (Casasanto and Lakoff)

Situated cognition:
→ Cognition is only meaningful inside a real context
→ Artefacts and the environment are part of the cognitive process
→ Learning must be situated as well, or it will not transfer

The extended mind (Clark and Chalmers):
→ A phone or laptop is part of the cognitive system
→ External storage as the cognitive equivalent of internal memory

Predictive coding:
→ The brain is constantly predicting its next sensory input
→ Only prediction error is passed upward
→ Perception as a process of inference
→ Applied to understanding hallucination, illusion and schizophrenia

Consciousness:
→ The hard problem (Chalmers): why is there subjective experience at all?
→ Global workspace theory (Baars): consciousness as broadcast
→ Integrated information theory (Tononi): consciousness as integrated information (phi)
→ Searching for the neural correlates of consciousness with fMRI and EEG

4E cognition:
→ Embodied: the body
→ Embedded: the environment
→ Enacted: action
→ Extended: tools

Frequently Asked Questions

Q. Is multitasking possible, and how does cognitive science explain it? A. In cognitive terms, genuine multitasking is impossible in most cases. What feels like multitasking is really task switching — moving rapidly between activities. Because there is only one pool of central executive resources, handling two cognitively demanding tasks at once produces a switch cost. Preparation time and error rates both rise at each switch. The exception is when one task has become automatic — walking while holding a conversation, for example — because it no longer draws on conscious resources. Using a phone while driving is dangerous precisely because driving also demands cognitive resources; going hands-free does not help much, since it is the conversation itself, not the handset, that divides attention.

Q. Why does memory get distorted, and can it be trusted? A. Memory does not store the past accurately the way a camcorder does. It is reconstructed according to present knowledge, expectations and emotion. Loftus’s research showed that eyewitness memory is easily distorted by information supplied afterwards through leading questions: in the car-collision study, participants who heard the word “smashed” estimated a higher speed than those who heard “contacted”. Distortion has several sources. Schemas fill in or delete details to fit expectations. Source monitoring errors leave people confused about where they heard something. Strong emotion sharpens the central event while making peripheral details less reliable, not more. None of this means memory cannot be trusted at all. Memory is an evolutionarily “accurate enough” system; what matters is recognising these limits and compensating for them wherever the stakes are high, such as legal testimony.