PSY260 Lecture 8: Procedural Memory, Skill Learning & Expertise

Squire’s Model of Long-Term Memory

• LTM → Nondeclarative + Declarative memory

• Nondeclarative memory → procedural (motor) memory + implicit memory

• Declarative memory → episodic + semantic memory

Explicit Learning vs. Implicit Learning

• We can learn with/without conscious effort

• Explicit learning: learned with conscious effort

  • Ex. studying for an exam

• Implicit learning: learned without conscious effort

  • Happens passively, with repeated exposure

  • Ex. picking up song lyrics without trying

Implicit Learning Study

• Participants were given a cue then pressed an associated button as fast as possible

  • Cues were usually in random order, but a 10-item sequence sometimes occurred

    • Participants became very fast at the 10-item sequence → could anticipate each subsequent key

  • Participants weren’t aware that they were learning the pattern → implicit learning

Procedural Memory in Amnesia

• Amnesiac patients can acquire new skills

  • HM’s mirror drawing performance improved

• Dissociation between “knowing how“ (procedural) and “knowing that you know how“ (episodic)

  • HM got better and better at the mirror reading task (↑ procedural)

  • BUT didn’t remember doing the task (↓ episodic)

• Amnesiacs acquire this skill at a similar rate to controls → procedural memory stays intact

  • BUT don’t remember doing it → episodic memory is damaged

Skill

• An ability to perform a task that has been honed through learning

  • Reading

  • Kicking a ball

• Skill memory, aka procedural memory

Skill memory vs. Episodic/Semantic memory

• Similarities

  • Improves with practice

  • Can become long-lasting

• Distinctions

  • Often hard to verbalize

  • Can be learned and retrieved without conscious awareness

  • Always requires repetition for goog learning

  • Distinct brain substrates

Kinds of Skill Memory

• Operant responses

  • Perceptual-motor skills

  • Cognitive skills

Perceptual-motor skills

• Learned movement patterns guided by sensory inputs

  • Ex. how to hit a curve ball, how to spike in volleyball

  • Closed: rote sets of movements

    • Same every time

  • Open: require adjustments based on environment

Cognitive skills

• Habits of problem solving

  • Ex. reading, composing a paper

• Not as embodied but get better with practice

Perceptual-motor skills vs. Cognitive skills

• Although these categories can be helpful:

  • Most skills involve both cognitive and perceptual-motor components

    • Ex. writing

  • Learning for both types of skills seems similar

Stages of Skill Learning (Lövden et al.)

1. Expansion/Exploration

   1. Test a broad pool of candidate neural circuits for performing the job

   2. Start wide → multiple pathways

2. Selection

   1. Choose the most promising circuit

   2. Best/most efficient pathway

3. Refinement

   1. Further training of selected circuit

   2. Strengthen that chosen circuit → increase speed, efficiency, automaticity

Paul Fitts: Stages of Skill Learning

1. Cognitive stage

   1. Performance is based on rules that can be verbalized

   2. Thinking quite hard

      1. Ex. using written instructions to set up a tent

2. Associative stage

   1. Actions become stereotyped

   2. Muscle memory with same steps

      1. Ex. setting up tent in a fixed sequence, without instructions

3. Autonomous stage

   1. Movements seem eutomatic

   2. Without thinking/focus

      1. Ex. setting up a tent while having a conversation

Practice: Overview

• Research shows that not all practice is good practice

  • Possible to reinforce wrong circuit

• Practice is usually ineffective without quality feedback

  • Feedback shouldn’t be too frequent

• Spaced out practice > massed practice (cramming)

• Varied practice (mixed skills) > constant practice (single skill)

Practice: Feedback

• Feedback is when you gain knowledge of results during training

  • Indication of how well you are performing the skill

• Feedback is essential for most types of learning

Practice: Feedback (Thorndike)

• Asked participants to draw lines exactly 3 inches long while blindfolded

  • Half the participants were given feedback

  • Other half wasn’t

• Despite equal practice → only feedback group improved

Practice: Feedback (Frequency)

• Frequent feedback is helpful initially, but doesn’t lead to high long-term performance

  • Could become reliant on it

• Infrequent feedback causes a slower start, but usually leads to better long-term performance

  • Not enough guidance to begin with

• Best: a mix

  • More frequent feedback in the beginning

  • Less feedback afterwards

Power Law of Learning

• With effective practice, performance increases

• Remarkably, the pattern of performance gain is similar across tasks and species

• PLL:

  • Gains are rapid at first, but decrease proportionate to what has already been learned

  • Same shape as Ebbinghaus curve

Practice: Spaced vs. Massed

• Spaced practice is much more effective

• Postal workers trained on a sorting machine:

  • Conditions: 1hr/day, 2hr/day, 4hr/day

  • 1hr condition got better in a shorter total # of hours, but more spread out

  • Spacing = less studying needed to get better

Practice: Constant vs. Variable

• Constant practice: focused on a single skill

  • Ex. play the G major scale 20 times

• Varied practice: alternates between a set of skills

  • Ex. play 4 different scales 5 times each

• Variable practice is often (but not always) more effective

  • May very well feel to the participant that they are doing worse, though → focusing on multiple may feel like you’re not progressing

  • However, constant practice is sometimes good or better → unclear how to pick the best practice schedule for a given skill

Habits

• Behaviours that become automatic and routinized

  • Generally develop through operant conditioning processes

• Habits can be:

  • Adaptive/beneficial

  • Neutral

  • Maladaptive/harmful

• Individual differences

  • Different people form habits in different ways and over different spans of time

  • Some habits may require lots of repetition, others can develop quickly

Tolman, Ritchie, Kalish: Mouse Maze

• Two start locations: S1 and S2

• Two goal locations: G1 and G2

  • Food for mice

• Rats trained in one of two ways:

  • Response learning

    • A particular response was always enforced

    • Ex. always turning right, regardless of S1/S2

  • Place learning

    • A particular location was always reinforced

    • Ex. always going to G2 regardless of S1/S2

• Place learning develops much quicker

Dual Solution Task

1. Training

   1. Rats are trained starting the same arm (S1/S2) and always find food at a consistent location (G1/G2)

   2. Could solve task in two ways:

      1. Learn the correct response → always G1

      2. Learn where to go → always turn left

2. Testing

   1. Removed food from maze to see where rats would go

   2. 8 days of training later: place learning dominates (always going to G2)

   3. 16 days of training later: response learning dominates (always turning left)

      1. Indicates habit → automatic

Expert

• Someone who has developed a skill better than most people

Expertise

• Seems to involve perceptual learning that lets the expert see the world differently than a novice

  • Ex. chess masters scan the board differently than novices

Talent

• Ability to master a skill with little effort

  • Ex. some people have a knack for languages

• Those who start off performing a skill well are more likely to have an easier time becoming experts

  • Get better feedback earlier → positive feedback → stick with it

Becoming an Expert

• How does one become an expert?

  • Through extensive practice

  • Through talent

• Twin studies → show that both expertise and talent matter

Chunking theory

• Experts organize information into larger, more meaningful units called chunks

• Allows for more efficient processing and recall of domain-specific information

Template theory

• Extension of chunking theory

• Experts develop complex cognitive structures (templates) that include both fixed and variable information

Deliberate Practice Theory

• Expertise is developed through highly structured, effortful practice

  • Must be very deliberate and intentional

• Focused on improving specific aspects of performance with immediate feedback

Cognitive Load Theory

• Experts have automated many domain-specific skills, reducing cognitive load (automaticity → saves cognitive load)

  • Pattern recognition: can quickly recognize patterns in their domain

  • Specialized production rules: allow for rapid, automatic responses in domain-specific situations

• Allows them to tackle complex problems more efficiently

Basal Ganglia

• Sits at the base of the forebrain

  • Striatum (largest structure) involved in conjunction of movement and reward

  • Regulated velocity, direction, and amplitude of movement → movement characteristics, fine tuning

• Collects input from throughout cortex

• Outputs to:

  • Thalamus → motor cortex

  • Brainstem (+ spinal cord)

    • Motor areas to refine movement

Damage: Basal Ganglia vs. Hippocampus

• Basal ganglia:

  • Selectively impairs skill learning

  • Spares episodic memory

• Hippocampus:

  • Impairs episodic memory

  • Spares skill learning

Basal Ganglia: Roles

• Particularly important in forming new skill memories

• Precise role of the basal ganglia is unclear

  • Basal ganglia is involved in movement → lesion effects on skill learning could be due to movement problems

    • Not specific to learning

  • Unclear if the basal ganglia are involved in consolidation/storage of skill memory, or both

  • Some evidence suggests the ultimate site of storage is the cortex

Dual Solution Task: Basal Ganglia and Hippocampus

• Injected saline (control) or lidocaine (sedative- temporary brain damage) into rat brains at each day to see how response/place learning is affected

• At Day 8:

  • Injecting lidocaine in basal ganglia → no effect

  • Injecting lidocaine in hippocampus → place learning disrupted

• At Day 16:

  • Injecting lidocaine in basal ganglia → response learning disrupted

  • Injecting lidocaine in hippocampus → no effect

• Summary:

  • Place strategy dominates (involves hippocampus) →

  • Response strategy dominates (involves basal ganglia - becomes habit)

Cerebellum and Skill Learning

• Patients with cerebellar damage take twice as long to trace images compared with controls

  • Rate of learning is similar to controls

  • Damaged patients just start at a lower point

Experience-Dependent Plasticity: Functional Changes

• Functional changes (neural efficiency):

  • When a skill is first learned, more areas of the cortex become active in representing the skill

  • Like first part of skill learning: expansion/exploration or cognitive stage

    • Testing many pathways to see what works best

  • With practice, expert brains become more efficient at processing relevant information

    • This is reflected in reduced activation of certain brain areas

    • This indicates specialization of brain pathways (refinement/autonomous stage) so only necessary pathways are needed

Cerebral Cortex and Semantic Memory

• Ex. participants touch each finger to thumb in a fixed sequence as quickly and accurately as possible

• Improvement rapid at first, then slower rate (power law)

  • After training → expanded region of motor cortex active during task

Experience-Dependent Plasticity: Structural Changes

• Long-term engagement of an area can lead to structural changes the areas of the brain related to that skill

• For example…

  • Musicians → ↑ grey matter volume in auditory/motor areas

  • Divers → ↑ cortical thickness in areas involving body awareness and motor control

London Taxi Drivers

• Receive extensive training to memorize the city’s complex street layout

  • Can get anywhere from anywhere, have an entire cognitive map of London in their heads

• Show larger posterior hippocampi

  • Size of posterior hippocampus correlates with amount of time spent as a taxi driver

  • Suggests ongoing plasticity