Skill Acquisition 1

Question 1

expert vs novice

Answer 1

better/faster component of skills

different representation

Question 2

procedural vs declarative knowledge

Answer 2

knowing how vs knowing that

skills, rules vs facts, concepts

hard to explain, implicit vs easily made explicit

fast, automatic vs slow

can be learnt by amnesics vs easily forgotten by amnesics

Question 3

Rabinowitz & Goldberg (1995) alphabet arithmetic task

Answer 3

addition, subtraction, match

declarative: 12 equations 36 times
procedural: 72 equations 6 times

declarative: become faster more rapidly, unaffected by size of digit, better transferred to subtraction

procedural: better at transferring to new addition equations, worse at subtraction
- > learnt procedures specific to direction, domain specificty

Question 4

effects of amnesia on learning

Answer 4

HM: no hippocampus, amygdala or rhindal cortex

severe global anterograde amnesia

could still improve on mirror drawing task even if he doesn’t recognise it, procedural knowledge preserved

Question 5

attention and skill

Answer 5

well-learnt skills become automatic:
-more proceduralised
-Require less attention or cognitive resources
-Less interference from secondary task
-Less affected by stress (unless you pay
attention to procedures, Beilock & Carr, 2001)
-Hard to modify

eg. experienced drivers take less time to
shift gears, and are less affected by
secondary task

Question 6

Shinar, et al (1998) study of experienced vs inexperienced drivers

Answer 6

Put experienced and inexperienced drivers
in automatic or manual cars in Tel Aviv
–Measured success at detecting signs (eg, “slow down”).
– For experienced drivers type of car had no effect
– For inexperienced drivers 13% fewer signs detected when in manual car.

Extra attention required for gear changes
decreased that available for other tasks.

Question 7

proceduralisation

Answer 7

as we learn, declarative knowledge is made into production rules.

These are specific to that skill.
IF (this is situation), THEN (do that).

Only execute rule when conditions are met.

Declarative and procedural knowledge may coexist side by side.
– e.g., speak foreign language fluently but also remember grammar rules.

Question 8

three stages of acquiring a skill

Answer 8

Skill requires shift from declarative to
procedural knowledge

often this also means shift from explicit to
implicit, but not necessarily.

1. Cognitive stage
2. Associate stage
3. Automatic stage.

Question 9

cognitive stage

Answer 9

Develop declarative encoding of skill (from
self or coach).
– Try to understand
– Go through steps (e.g., pilots, juggling)

Rehearse as skill is performed.
– e.g., memorizing position of gears (“up, left”) and the sequence for engaging the clutch and moving stick shift.

Inadequate for highly skilled performance.

Question 10

associate stage (knowledge compilation)

Answer 10

Two things occur:
1) Errors in initial understanding are
gradually detected and eliminated.
– e.g., learn to coordinate clutch and accelerator.

2) Connections between elements are
strengthened.
– e.g., no longer have to sit and think about how to get into second.

Result is a successful procedure for performing the skill.

Question 11

autonomous stage (procedural)

Answer 11

Procedures become more automatic.

Execution of skill is better,
– e.g., faster, more accurate, like a reflex
– Changes in patterns of brain activation, e.g., for
motor skills less cerebellum and more temporal
lobe activation (Lindquist & Guadagnoli, 2008).

Declarative knowledge of skill may be lost.

Can still be “bugs” in productions.

No sharp distinction between associative and autonomous stages: may be a continuum.

Question 12

Bryan & Harter (1899) Morse code learning study

Answer 12

learning progresses from recognizing letters to recognizing words.

So different stages of proceduralization.

Improvement was continuous, but became
slower

Question 13

power law of practice

Answer 13

T = A + FP-b (P raised to a power) 
or log(T) = log(A) - b log(FP)

time decreases fastest with increasing number of trials at the start (learning rate highest at the start)

continues to improve at slower rate

Question 14

evidence for power law of practice

Answer 14

improvement can keep occurring
forever, though at a diminishing rate.

Blackburn (1936): continuing
improvement by two subjects performing
10,000 addition problems.

Power-law applies to many skills, even
complex skills such as justifications for
geometry-like proofs (Neves & Anderson,
1981).

Question 15

physical limitation

Answer 15

Physical constraints will place limits on
improvement.

Crossman (1959) studied Cuban cigar roller.
– Speed followed power-law for ten years.
– Then reached physical limits of her machinery

Question 16

criticisms of power law of practice

Answer 16

We don’t know whether improvement can or cannot go on forever.
– Hard to definitively test.

Heathcote et al (2000) argue that learning is an exponential function.

Aggregate data may follow law but not always individuals (Haider & Frensch, 2002).
– May be due to strategy changes.

Logan (1988) proposed that it is not due to
progressively better procedures, but due to
memory-based processes.

Question 17

Ericsson, Chase & Faloon (1980) study of random digit sequence

Answer 17

sequences of random digits to read then
recall.
– If sequence was correct, then increase by 1, otherwise decrease by 1.
– After randomly selected trials he reported his thoughts
– Spent one hour a day

Test at start showed he could recall 7 digits

Around the fourth session Steve had
reached 9 digits, but felt he could go no
further.
– Science said he was right.

But then he realized he could group digits
into meaningful chunks of 3 or 4
– He was a runner, so used running times.
– Eg, 907 represented a good two mile time
– In this way he could use his working memory as an index to his long-term memory

Spent one hour a day, 3-5 days a week for 20 months (approx. 230 hours).

Increased digit span from 7 to 79 digits

At end of session could recall 80% of all the digits from the hour session.

Better than any memory expert at time. Up to 82 digits

Performance dropped after 2 months when he was deliberately presented with sequences that did not fit to running times.
– But then supplemented with ages, eg, 893 “89 point 3, very old man” and dates, eg, 1944 “near end of WW2

Question 18

going further on digit recall test

Answer 18

Mnemonic associations could at best get him to 28 digits, so needed retrieval structures
– Started by created two running times then rehersing them and the rest of the digits.
– Then developed retrieval structures in which 4 digit chunks became subgroups within supergroups, then another level.

Did his basic memory capacity increase?
– Groups never more than 4 digits, never rehearsed more than 6, level never had more than 3 groups.
– When switched to letters, dropped back to 6 consonants

Question 19

implications of digit recall study

Answer 19

Practice did not change basic memory capacities, but improved used of meaningful chunks and structure, which seem to be basis of our memories.
– Probably true of all memory feats, eg, World Record for Pi is now 70,030 digits, 3115 for random digits

Develop better mental representation for task.

Joshua Foer, journalist who became memory champion (“Moonwalking with Einstein”)

Critical aspect of expertise could be gaining the right representation

Question 20

qualitative differences between expert and novice

Answer 20

Experts are obviously better than novices
and have practiced much more.

May differ qualitatively, not just quantitively

De Groot (1965) studied chess grandmasters

Question 21

De Groot (1965) study of selecting chess move

Answer 21

De Groot (1965) performed protocol
analysis of chess players selecting a move.
– Masters did not search deeper
– Did not examine more moves
– But they looked at the right moves from the beginning

better intuition

Also had better memory for chess positions

Question 22

Chase & Simon (1973) study of chess pieces

Answer 22

Compared chess master and novice recall
for normal and random placement of chess
pieces

masters remember normal chess boards better, but not better at scrambled chess boards

after each glance masters recreated a
meaningful chunk of the position.

So better mental representation: Simon & Gilmartin (1973) suggest master acquire a
“vocabulary” of 50,000 chunks.
- Takes at least 10 years.

Question 23

benefits of better mental representation

Answer 23

1. improved speed
2. links to knowledge
3. better anticipation
4. better planning
5. focus on the right elements
6. better learning

Question 24

1. improved speed (chess)

Answer 24

If perceptual difference between experts and novices then speed may make a difference.

Chess is played both slow (40 moves in 2.5 hours) and fast (all moves in 5 minutes).

Burns (2004): performance in fast play
correlated between .78 and .90 with predicted performance based on slow play.

stronger relationship between fast and slow play as expertise level increases.

What distinguishes players at the highest
level of skill seems to be what they do in
first few seconds.
– Does not mean that the rest of the time is
wasted
– Crowley & Byrne (2004): more errors in fast chess play
– Less analysis by both players, but does not change their relative performance

Finding the right moves at the start seems
result of having the right representations.

Question 25

1. improved speed (sports)

Answer 25

In sports time is clearly critical.

Savelsbergh et al (2002) compared 7 semi-pro to 7 recreational soccer goalkeepers.

Shown film of lead up to penalty kick and start of flight, and moved joystick to anticipated location.

experts more likely to move to the right place (more penalties stopped, higher proportion of correct height and side, lower proportion of corrections), faster to initiate joystick movement, but reaction time not faster

difference: novice focus on kicking foot, experts focus on non-kicking foot

Question 26

2. links to existing knowledge (physics)

Answer 26

Chi, Feltvich & Glaser (1981) found that
physics experts classify problem differently.
– Novices (physics undergraduates) use surface features, eg. Incline plane, pullies, friction.
– Experts (postgrads) used deep feature, eg, conservation of energy, Newton’s 2nd law.

Question 27

2. links to existing knowledge (medical)

Answer 27

Lesgold, et al (1988) presented medical students and radiology experts with the following x-ray.

Sail-like area in the right lung is a shadow.
– Students interpreted this as a tumor as tumors are the most common cause of shadows.
– Experts better at correctly interpreting it as a collapsed lobe of the lung: Focused on richer features, such as lobe’s boundaries.

Staff radiologists used more reasoning, and
developed coherent model of patient.
– Like other experts, build a mental representation of situation

Skill development not monotonic as 1st/2nd year resident more often correct then 3rd/4th.
– Seem to start out learning to identify particular features, but initially may make more errors as start to use deeper recognition-triggered reasoning.

Patel & Groen (1991) found experienced doctors more accurate but had poorer recall for what info was used in diagnosis (because they use more holistic approach, less on details)

Experts think about their domain differently.

Question 28

3. better anticipation

Answer 28

Ward, Ericsson & Williams (2013) showed video of a soccer game to players then stopped it and asked them what would happen next.
– Dribble? Shoot? Pass?
– More accomplished a players was the more often they predicted successfully.
– Also had better memory for where players were.

Better prediction helps you respond better and faster.
– Maybe one reason why game seems to “slow down” for best players.

Question 29

4. better planning

Answer 29

Bläsing, Güldenpening, Koester, and Schack
(2014) showed that expert rock climbers could automatically identify what grips were needed at each point on the wall

This allowed them to plan out a climb in their heads before they set out.

Evidence that their brain was already signalling the right grip to the hands.

Question 30

5. focus on the right elements

Answer 30

Schools train chicken sexers who after 2-
6 years of practice can classify 1000 day
old chicks per hour with 98% accuracy

Biederman & Shiffrar (1987) had expert identify critical areas in pictures, then
developed a set of basic features.

Gave sheet to naïve subjects
– 84% accurate, basically as good as professionals for same pictures.

Knowing what elements to focus on could substitute for months/years of practice

Question 31

6. better learning

Answer 31

Sikes (2013) showed that what differentiates best musicians from lesser is better mental
representation.
– These provides a guide to practice and ultimately performance.
– Lesser players rely on feedback from teachers

McPherson & Renwick (2001) found that better musicians made fewer errors but corrected more of them.
– Poorer had 11/min errors on average first time through a piece, then 70% of the same errors second time
– Best had 1.4/min errors, but corrected 80% second time through.