M1
What is Cognition?
"…The term "cognition" refers to all the processes by which the sensory input is transformed, reduced, elaborated, stored, recovered, and used. It is concerned with these processes even when they operate in the absence of relevant stimulation, […]. Such terms as sensation, perception, imagery, retention, recall, problem-solving, and thinking, among many others, refer to hypothetical stages or aspects of cognition (Neisser, 1967 p. 4)".
The Roots of The Study of Mind
Philosophy
Plato (Meno); Aristotle (De anima); Descartes (Mediations Metaphysiques); Locke (An essay concerning human understanding); Hume (An enquiry concerning human understanding); Kant (A critique of pure reason)
Their ideas about the mind were vastly different, but they all agreed that it was important to study
Early Scientific Work
In science, researches use a set of discipline-specific techniques devised to generate empirical data with the goal of testing theories about how the world works
Fechner (1860): What is the relation between physical stimuli and sensation?
Donders (1868): How quickly can people react to a single stimulus? What happens if they must choose?
Ebbinghaus (1885): How do learning and memory vary with time and practice?
Wilhelm Wundt and Introspection
Wundt, the scientific method, and introspectionism
In 1879, Wundt established the first laboratory of psychology
One of the methods developed for studying cognition was introspection (assessing your own feelings)
Introspection
Looking within
Study of the contents of consciousness and of mental processes through inner observation
Titchner was a strong proponent of this method in America
Problem: “You can’t really get people to look within and see whats going on” → Most mental processes are not cognitively penetrable
This major difficult and the Zeitgesit:
The theory of evolution
Positivism
Pavlov’s Conditioned reflexes (classical conditioning)
Led to behaviorism (do not need to study the mind to understand thinking)
Evolution
Charles Darwin
Traits that help an animal survive or reproduce are passed down more often, which means they become more common over time
Natural Selection
Members of species are variable
There is a struggle for existence
Therefore, some variations must confer an advantage for survival
These adapted individuals will tend to survive
These adapted individuals will have offspring with similar characteristics vis inheritance
Hence, there will be a natural selection
Logical Positivsm
Philosophers were fed up of arguing about useless things
All knowledge must be based on empirically verifiable facts
Operationalization
To establish a clear relationship between the theoretical construct and its empirical basis in the operations producing scientific data
Operationalization is simply turning fuzzy, abstract ideas into concrete, measurable numbers or actions
Classical Conditioning
Ivan Pavlov
Russian physiologist who was initial interested in the digestive system
A type of associative learning during which two events (a stimulus and a response) become interconnected such that one may activate the representation of the other
Classical Conditioning
US (Unconditional Stimulus): naturally causes a reaction (Meat)
UR (Unconditional Response): natural reaction (Salivation)
CS (Conditional Stimulus): learned trigger for a reaction (Metronome + Meat → Salivation)
CR (Conditional Response): learned reaction (Metronome → Salivation)
Extinction Learning
Extinction in psychology is a procedure where a learned behavior gradually weakens and disappears because it is no longer reinforced
Behaviorism
A theory in the philosophy of mind which maintains that talk of mental events should be translated into talk about observable behaviour
It makes the term “psychology” problematic
Psychology means study of the mind but behaviourism says do not study the mind
James Watson
His perspective deemed psychology an objective, natural science focused strictly on observable behaviors, environmental stimuli, and conditioning
Behaviorism’s Assumptions about Psychology
“Tabula rasa” → Mind is a blank slate
All learning stems from forming associations (classical or operant)
Theories about the mind are not needed in a complete explanation of psychology
To behaviourists all they need to know is the stimuli and the response
Operant conditioning
Positive Reinforcement: Adding a rewarding stimulus to increase a behavior (e.g., giving a bonus for high performance).
Negative Reinforcement: Removing an unpleasant stimulus to increase a behavior (e.g., turning off an alarm by buckling a seatbelt).
Positive Punishment: Adding an unpleasant stimulus to decrease a behavior (e.g., assigning extra chores for breaking rules).
Negative Punishment: Removing a pleasant stimulus to decrease a behavior (e.g., taking away toys)
Problems with Behaviorism
Tabula rasa (no innate dispositions)
Breland and Breland: The Misbehaviour of Organisms
A raccoon's refusal to deposit coins into a box—instead opting to rub and "wash" them
Animals did not come into this world with a blank state, they have some prior knowledge
Taste Aversion Learning
A conditioned taste aversion learned if ingestions of a novel flavor is followed by an aversive consequence such as indigestion or food poisoning
Your brain links a specific flavor with getting sick, causing you to avoid that food in the future. It is a survival mechanism to prevent you from eating dangerous or poisoned food again.
One trial learning
Aversion learning takes place even when some time elapses between the CS + US pair and the UR
Garcia and Koelling (1966)
The rats did not avoid the water in all conditions. They only learned associations that "made sense" to their survival instincts:
This study found there is a predisposition for forming associations
Taste cues → Food and being Sick
Tactile Cues → Brightened noise and shock
More problems with behaviourisn
All learning stems from forming associations
Theories about the mind are not needed in a complete explanation of psychology
Verbal Behaviour
According to Skinner, operant conditioning is sufficient to explain the development of language
Organisms produce sound. Words are reinforced while non-words are not. These words are associated together to form sentences. Labels are associated with objected to create meaning…
It is not necessary to “look inside the organism” to understand language
Chomsky’s Review
Noam Chomsky: A review of B. F. Skinner’s Verbal Behavior
Behaviorism (and its theoretical constructs) are inadequate to explain language acquisition
One must appeal to intermediate variables, to cognitive constructs
Chomsky’s Review
Stimuli, responses, associations, and reinforcement are well-defined constructs in experiments with Skinner boxes, but is this case for language acquisition?
Quine’s induction problem: How can a problem as complex as language be solved if there are no innate mental constraints?
How can non-reinforced linguistic behaviour occur?
Chomsky’s Conslusion
Langugae requires more than operant conditioning to explain
A complete explanation of language must include innate dispositions based on rulebased learning
Chomsky’s review was instrumental in launching the cognitive revolution
Behaviourism’s Legacy
Behaviorist researches established the methodological rigor that is necessary to the scientific study of psychology
Classical and operant conditioning are two of the most reliable, important and applicable concepts in psychology
Associative learning explains a large portion of all animal behaviour The Cogntive Revolution
Behaviourism cannot fully explain all of psychology
The mind is back, but how do we explain it?
Cognition
Refers to the mental processes involved in acquring knowledge
Involves thinking or conscious experience
What is a Computer?
Device that manipulates information through the use of rules
“Maybe human thinking works a bit like these computers”
Information-processing Idea
Computers are thinking machines, so maybe humans work something like computers
Computers use logical rules → “If-then”
Programmers use a flow chart to try the logically map out what the computer will do, then make it into code
If humans think like computers, then thought must involve input, storage, processing and output
The Mind is a Rule-based Information Processing Program
The mind is a computer program that runs on a physical computer called a brain
If humans think like computers, then thought must involve input, storage, processing and output
The work of the cognitive psychologist is to understand the mind as a program that runs on the brain
The Information Processing Paradigm
Postulates (from Howard, 1983):
(1) Between stimulus (input) and the response (output), there are stages of processing which require time
(2) When inputs are processed, they undergo transformations in form and content
Behaviorists firmly believe if you put the proper reinforcers in the environment, you can shape someone to be whatever
It is possible to take measurement of what is going on in the mind, and that allows to study it scientifically
(3) Some stages of processing have a limited capacity: There is a limit on the amount of information that can be processed simultaneously
(4) Processing is serial: Processing at one stage is dependent on the completion of processing at the previous stages (you can only do one thing at a time with steps)
Parallel processing: Being able to consider and process multiple pieces of information simultaneously
Sternberg Task Demonstration
High-speed scanning in human memory
A classic psychology experiment demonstrating how people retrieve information from short-term memory
Goal of the experiment: How is symbolic information retrieved from recent memory?
Experiment - Sternberg
Participants are asked to view a set of 1-6 digits (i.e., the memory set) for 1.2 seconds
Then, they are shown a probe (i.e., a test digit)
The task is to say as quickly as possible, while minimizing mistakes, whether the test item is in the original set of digits
Method
Eight participants took part in the study
The participants has 24 practise trials and 144 test trials
A Four Stage Process Model for Memory Scanning
“Underlying the paradigm of these experiments is the supposition that if the selection of a response, requires the use of information that is in memory, the [response times] will reveal something about the process by which the information is retrieved
Encode test item → Scan and compare with memory set items → Binary (yes/no) decision → Execute motore response
The only not constant is “scan and compare with memory set items”
Sternberg Hypothesis
Serial self terminate: Comparing all the numbers to the number, if you find the number you record it right away
Serial exhaustive: Look at each item one at a time: Even if you find the number, you go through all the numbers till the end regardless
Most common
Parallel processing: Considering all things simultaneously (the memory set is seen by the minds eye and you can pick out the number)
Results - Sterberg
Response time = 397 ms + 38 per item
The Importance of the Information-Processing Paradigm
Experiments such as Sternerg’s scanning task are conducted only if researchers believe that cognitive processes:
Take time
Transform the input
The foal of the information processing approach is to build models of human cognition
A third Paradigm for Studying Cognition
Behaviourism (1900-1960): Associationsim + no study of mental representations
Information processing (1960-today): Rule based learning + the study of mental representations
Connectionism (1986-today): Associationism + the study of mental representations
Connectionism
Connectionists believe that it is not necessary to postulate a “program” level to fully understand cognition
Postulates that a good model of cognition is a model of the brain
Mind = Brain
What is the Brain
A network of highly interconnected neurons that processes information in parallel and that learns by forming associations
Connectionist Network
A network of highly interconnected abstract units that processes information in parallel and that learns by forming associations
Mathematically described
Computer simulated
3 layers
Input units: Receive information from environment
Hidden units: Systems thinking, what happens in the mind
Output units: Systems response
How a Connectionst Network works
Simulations start with random associations (tabula rasa)
Starts getting input
Ouput nonsense
Each time it provides an answer, the system is given feedback
Using this feedback, it starts adjusting connections among the units (associations) so that it can do better next time
Hebbian learning rule: Neurons that fire together, wire together
With training and feedback-driven learning, the networks can learn object recognition (and the past tense of verbs and categories)
Connectionist Networks are Sub-symbolic
Sub-symobolic: No node or unit in the system has a specific meaning. It is the pattern of activation across the network that has a meaning.
All knowledge consists of association
There are no rules or symbols
Connectionist Networks
Why prefer connectionist models to information processing (symbolic) models?
Biological plausibility: These models are built on how the brain works
Autonomous learning with minimal “innate” knowledge (can learn on their own)
Some kinds of learning are more easily explained by connectionist networks (e.g., perceptual tasks)
Together, information-processing and connectionism may provide a more complete explanation of mental processing
Steven Pinker on Language
Language is a distinct, evolved biological "instinct" or mental module, not a cultural invention
Kids use language rules and generalization from the moment they begin to speak
Chomsky: Children are pre-wired with a universal grammar
One More Approach
Behaviourism (1900-1960): Associations + no study of mental representations
Information processing (1960-today): Rule-based learning + the study of mental representations
Connectionism (1986-today): Association + the study of mental representations
Cognitive neuropsychology (80’s-today): Behavioural and neurological meaures are employed to theorize about the mind
Cognitive Neuroscience
The study of cognition by using behavioural methods and neuroscience techniques
Theories of cognition:
Behavioral measures
Task accuracy
Response time
COnfidence judgement
Recogntion
Verbal protocols
Neurological measures
fMRI and positorn emission sonography
Event-related potentials
Electroencephalography
Transcranial magnetic stimulation
Cognitive Neuroscience
Psychologists seek to understand the mind, and cognitive neuroscientists are concerned with understanding how the mental processes take plain in the brain
Embodied Cognition
Sensory and motor processing are involved in people’s understanding of the world
Hauk, et al → How do humans understand words
Uses brain
Embodied cognition is a plausible hypothesis that word meanings in the brain are strongly associated with cortical areas that are related to action
Passive Reading: Participants read action verbs related to the face ("lick"), arms ("pick"), or legs ("kick")
Motor Localizer: Participants performed actual movements with their tongue, fingers, or feet
Words like "lick" activated areas in the motor strip adjacent to or overlapping with actual tongue movements.Words like "pick" activated areas related to finger/arm movements.Words like "kick" activated areas related to foot/leg movements.
Glenberg and Kaschak
Not using brain
Method: Participants were given a sensibility judgment task. They read sentences to determine if they made sense (e.g., "Close the drawer" or "Open the drawer")
Setup: A special button box was used with three buttons: one close to the body, one in the middle, and one far away
Task: The participant rested their finger on the middle button. To answer "Yes," they had to move their hand to either the near or far button.
Directional Stimuli: Sentences implied action either toward the body ("Open the drawer" / "Liz told you the story") or away from the body ("Close the drawer" / "You told Liz the story")
Facilitation (Faster): When the sentence direction matched the movement direction (e.g., a "towards" sentence combined with a "towards" move), participants were faster.
Interference (Slower): When the sentence direction conflicted with the movement direction (e.g., an "away" sentence, but they had to move toward their body to press "Yes"), reaction times were slower
Conclusion
Action sentence compatibility (ACE): Language understanding is grounded in bodily action
Is the Interpretation of the Data Correct?
Hickok: The myth of the mirror neuron
People can understand actions that they can’t perform (e.g., fly), actions that don’t involve the motor system (e.g., sweat), and abstract concepts unrelated to action (e.g., love)
Patients ith amyotrophic lateral sclerosis (a disease that desroys motor neurons) still understand action words very well
Meta-analyses have found motor cortex activation for non-words
Motor cortex activation related to the processing of action words simply reflect associative learning in the brain
Weisber et al.
Irrelevant neuroscience information changed people’s evaluations of psychologica explanations
Conclusions on Cognitive Neuroscience
Advantages
Provides additional DVs to test psychological theories
Develops our understanding of the brain
Disadvantged
Often expensive
Interdependence between behavioural and neurological measurement
If the goal is to develop psychological theory, cognitive neuroscience is not the ultimate answer
Neurosceintific study of the mind requires vast knowledge and expertise
Cognitive Psychology
Scientific way of trying to study the mind and thinking
Cognitive Science
Multidisciplinary study of the min dBottom-up Processing
First, the system may rely on the physical input
Using things in your environment to build an idea and ultimately recognize it
Bottom-up or datadriven processing:
Processing which is driven by he stimulus pattern, the incoming data
Template matching
The input is compared with memorized exemplars (templates) until a good fit is found
Problems:
Parsimony: How many templated would be necessary? → A lot
Flexibility: can the templated account for the variability?
Pandemonium Architecture
Solves issue of flexibility
Uses bottom-up processing
Pandemonium is a data-driven recognition model based on feature analysis; that is, objects are recognized from an analysis of their components
Pandemonium is composed of four type of recognition units (demons)
(1) Image demon: Transforms information in the environment into something the system can process (transduction. They record the initial image of the external signal.
(2) Feature demons: Look at specific features only. They look for a particular characteristic pattern.
(3) Cognitive demons: Listen to the different feature demons, and seek out one particular pattern.
(4) The decision demon: Listens to the pandemonium created by the demons. Its anser is determined by the demon that is yelling the loudest.
Empirical Support for Pandemonium
Power:
With a finite set of feature detectors, pandemonium can recognize a potential infinite number of objects
It will recognize in spite of changes in size, orientation and other distortions
People’s performance in identification studies:
In letter identification tasks, people will mistakenly call “O” a “C”, “G” or “Qu” but never a “H” or “T”
Neuropsychological studies
Certain brain cells respond only to specific line orientations
Problems with Pandemonium and Top-down Processing
Gestalt Principles of Organization
When we look at things, we tend to see stuff thats not even there
the whole is different than the sum of its parts
Similarity, closure, continuity, proximity
Top down or Conceptually Driven Processing
Using your knowledge of the world to deduce what you’re seeing
Processing which is influenced by the context and higher level knowledge
Empirical Evidnece for top-down Processing
The word superiority effect (reicher, 1969)
People are more accurate in identifying a letter when it is part of a word
Paticipants: Nine extensively trained subjects took part in the forced-choice letter identification task
Material: Three types of stimuli were used (4 letter word, 4 letter non-word, letter)
Procedure: The participants were shown one of the three display types for approx. 35ms, 60ms or 85mc
Then, they were asked to say which of two letters had appeared in the cued location
The Interactive Activation Model
McClelland and Rumelhart: Asked how does the knowledge that we have interact with the input?
The Interactive Activation Model is a pandemonium-like system that includes both bottom-up and top-down processing
The Interactive Activation Model: General Assumptions
Features, letters and words were build from the alphabet presented on the right
The system was equipped with 1179 four letter words
The Interactive Activation Model: Representation Assumptions
Perceptual processing takes place within a system in which there are several levels of processing, each concerned with forming a representation of the input at a different level of abstraction (features, letters and words)
Visual perception involves parallel processing
Visual processing occurs at several levels at the same time
The Interactive Activation Model: Representation Assumptions Cont
There is a node for each word and each leter (in each letter position)
The nodes are organized into levels
The nodes are connected to all other nodes (parallel processing) within levels r between adjacent levels
Connections may be excitatory (nodes yell “me”) or inhibitory (nodes yell “not me”)
The Interactive Activation Model: Operating Assumptions
Upon the presentation of a stimulus, a set of featural input is made available to the system
Initially, all nodes are in a quiscent state
The features activate letter, which is turn activated words, which finally send back activation to the letters
The answer of the system is the letter with the most activation at a given time
Features → Letters → Words → Best match wins
The Interactive Activation Model: Conclusions
With a limites number of clearly defined postulates, the interactive activation model formally provides a sufficient explanation of the word superiority effect
This model is a direct precursor of contemporary connectionist models
James McClelland
Influential cognitive psychologist
He elaborated the interactive activation model and led the parallel distributed processin research group
He was responsible for associationism’s return to the mainstream