Semantic MemoryKnowledge memory

Main questions

How do we gain knowledge?

How is our knowledge represented and organised in the mind-brain?

What happens when we access information?

(Note 2nd and 3rd questions are strongly related.)

Semantic MemoryKnowledge memory

Main questions

How do we gain knowledge?

How is our knowledge represented and organised in the mind-brain?

What happens when we access information?

Test predictions of semantic network model

Semantic Memory

Important tasklexical decision task

make a word-nonword judgement fora letter string

murget

higgle

beer

stout

Semantic Memory

Important tasklexical decision task

make a word-nonword judgement fora letter string

Priming paradigmexamine the influence of one stimulus on the

production of (e.g., naming of) or lexical decision for another stimulus

AnimalBreathes

Skin

Fish

Gills

Tail fin

Trout

Swims

Eel

Perch

Gills

Gills

Gills

Spread of activation means that some nodes willbe more highly activated following the activation of a related concept

Decisions for or naming of some concept (node) should be facilitated by the activation of a related concept

Example: Activation of “fish” should facilitate (speed up)the lexical decision RT for “salmon” compared to anunrelated concept (e.g., chair – salmon)

Test some predictions of the semantic network

Use of word naming as a task(name/read a word as quickly possible)

Fruit

P

Loftus & Loftus (1974)presented fruit – P (subject gives a name of fruit starting with P)investigated idea that priming dissipates with time immediate vs. delayed repetition of category

Reasoning: 1 trial could prime performance on a later trial(priming across or between trials)

If activation decreases with time, then there should be more priming on an immediate trial compared to a delayed trial

Loftus & Loftus (1974)

Prime TargetTrial 1 fruit A (initial trial)Trial 2 fruit P (immediate, lag = 0)

Prime TargetTrial 1 fruit A (initial trial)Trial 2 vehicle TTrial 3 fish STrial 4 fruit P (delayed, lag = 2)

Prediction: Naming RT should be lower (faster) for the immediate trial compared to the delayed trial.

Naming RT as a Function of Condition

Condition

1400

RT (ms)

initial trial lag = 0 lag = 2 Category Repeat

2300

Naming RT as a Function of Condition

Condition

1400

RT (ms)

initial trial lag = 0 lag = 2 Category Repeat

2300

1st trialprimedlater trials

Naming RT as a Function of Condition

Condition

1400

RT (ms)

initial trial lag = 0 lag = 2 Category Repeat

2300

moreprimingfor lag=0than lag=2

Loftus & Loftus (1974)

Prime TargetTrial 1 fruit A (initial trial)Trial 2 fruit P (immediate, lag = 0)

Prime TargetTrial 1 fruit A (initial trial)Trial 2 vehicle TTrial 3 fish STrial 4 fruit P (delayed, lag = 2)

Prediction: Naming RT should be lower (faster) for the immediate trial compared to the delayed trial.Prediction upheld. Conclusion: Activation decreases with time.

Other examples

nurse

doctor

nurse (prime)

doctor (target, right after prime)

(make a lexical decision for “doctor” or pronounce “doctor”)

Does “nurse” prime “doctor”?

Compare RT for

nurse – doctor (related condition)

shoe – doctor (unrelated condition)

nurse – doctor RT < shoe – doctor RT xxx – doctor RT (neutral) blank – doctor RT (neutral)

Difference is a priming score.

Basic semantic priming effect:

Presentation of nurse speeds up (facilitates) lexical decision or proununciation time of doctor compared to a control condition

Basic semantic priming effect:

Is spread of activation automatic?

Evidencemediated priming

effects at short SOAs

effects at short SOAs despite exectancy

Basic semantic priming effect:

Is spread of activation automatic?Evidence Mediated priming (Lorch & Balota, 1986)

nurse related to doctor, doctor related to lawyernurse not related to lawyer

nurse – primelawyer – target

nurse facilitates pronunciation time for lawyer!

Priming is mediated by doctor!i.e., activation spreads from nurse to doctor to lawyer

Basic semantic priming effect:

Is spread of activation automatic?

Choosing the right control condition

Evidenceeffects at short SOAs

effects at short SOAs despite exectancy

Control condition: shoe (prime) - doctor (target)

What if shoe affects decision RT for doctor?!

That is, maybe shoe – doctor is not a good control condition.

Need a better control condition: xxx (not a word)

Does nurse affect decision RT for doctor?Does shoe affect decision RT for doctor?

What if shoe affects decision RT for doctor?

Need a control condition: xxx

Does nurse affect decision RT for doctor?Does shoe affect decision RT for doctor?

Compare RT for nurse – doctor (related condition)

xxx – doctor (control or neutral condition) shoe – doctor (unrelated condition)

Compare RT for nurse – doctor (related condition)

xxx – doctor (control or neutral condition) shoe – doctor (unrelated condition)

nurse – doctor RT < xxx – doctor RT

Facilitation

Compare RT for nurse – doctor (related condition)

xxx – doctor (control or neutral condition) shoe – doctor (unrelated condition)

shoe – doctor RT > xxx – doctor RT

Inhibition

Neely (1976)related, unrelated, and neutral primes

also manipulated prime-target SOA basic idea:

shorter SOA—decision based on automatic processes

longer SOA—decision based on automatic + controlled processes

see OverheadFigure 6.10 from Ashcraft (1998)

RT (in ms) as a function of prime-target condition and SOA

Prime-TargetCondition SOA RT

Neutral 360 592Related 360 575 (17 ms facilitation)Unlrelated 360 611 (19 ms inhibition)

Neutral 600 588Related 600 543 (45 ms facilitation)Unlrelated 600 598 (10 ms inhibition)

Neutral 2000 606Related 2000 554 (52 ms facilitation)Unlrelated 2000 625 (19 ms inhibition)

RT (in ms) as a function of prime-target condition and SOA

Prime-TargetCondition SOA RT

Neutral 360 592Related 360 575 (17 ms facilitation)Unlrelated 360 611 (19 ms inhibition)

Neutral 600 588Related 600 543 (45 ms facilitation)Unlrelated 600 598 (10 ms inhibition)

Neutral 2000 606Related 2000 554 (52 ms facilitation)Unlrelated 2000 625 (19 ms inhibition)

Neely (1976)Results show facilitation and inhibitionResults show an increase in facilitation from SOA of about 350 to an SOA of 600

Why should facilitation increase?

Two types of processes at work?

Maybe initial facilitation is automatic.Maybe extra facilitation is due to controlled

processes.

Neely (1977)related, unrelated, and neutral primesalso manipulated SOA

AND expectancy (thought to reflect controlled processing,which should take some time to be able to be used)

Neely (1977)

Expectancy (thought to reflect controlled processing,which should take some time to be used effectively)

see bird prime, expect a type of bird as a target

bird

robin

bird

dove

bird

arm

Neely (1977)

Expectancy (thought to reflect controlled processing,which should take some time)

see bird prime, expect a type of bird as a target

sometimes get a type of bird as a target

on occasion, get a word from a different category as a target

see Overhead Figure 6.11 (left panel) from Ashcraft (1998)

Only facilitation at first (short SOA), no inhibitionfacilitation effect based on automatic processing(the expectancy hasn’t had time to kick in)

As SOA increases, faciliation and inhibition increaseincreases based on controlled processing(the expectancy has now kicked in and is being used)

Neely (1977)Expectancy (see Table 6.2 from Ashcraft, 1998)(thought to reflect controlled processing,which should take some time)

see body prime, expect part of a building as a target(i.e., expect a category shift)

body

window

body

roof

body

door

body

heart

Neely (1977)Expectancy (see Table 6.2 from Ashcraft, 1998)(thought to reflect controlled processing,which should take some time)

see body prime, expect part of a building as a target(i.e., expect a category shift)

sometimes get part of a building as a target(expected, but not actually related to the prime)

on occasion, get a body part as a target (actually related to the prime, but not expected)

see Overhead Figure 6.11 (right panel) in Ashcraft (1998)

Expected Shift, but Unrelated prime-target (e.g. body – door)

No facilitation at first (short SOA),But facilitation at longer SOAs

No effect based on automatic processingLater facilitation based on controlled processing

(the expectancy)

Related prime-target, but not expected(e.g., body – heart)

Facilitation for short SOA (even though not expected)Inhibition for later trials (even though related)

Initial facilitation effect based on automatic processing(*occurs despite expecting an unrelated target)

Later inhibition effect on controlled processing(the expectancy)

Facilitation in 250-ms SOA condition for related, but unexpected target provides strong evidence for the idea priming is automatic—consistent with the idea of spread of activation.

Evidence from longer SOAsControlled processes can contribute quite a bit to inhibition and facilitation effects.

Semantic Memory

Main questionsHow do we gain knowledge?How is our knowledge represented and organised in the mind-brain?What happens when we access information?

1) Good evidence for automatic activation of related

information (spread of activation). 2) Activation drops fairly quickly. 3) Controlled processes affect performance and need

to be dealt with in terms of theory.