Lecture 6: Clark Hull Flashcards
Clark Hull: Macrotheorist
Hull was the outstanding learning theorist in the US during the 1930’s-40’s
His is an all-inclusive theory that deals with everything
Psych historians contend that only two people have created inclusive theories: Hull and Freud
Hull wrote the theory so well, it was widely exposed to researchers, who then showed it didn’t work, since it couldn’t be manipulated, because it had such specific predictions
At Yale University, he was very influential: Bandura, Dollard, Miller, Mowher, all trained in Hull’s lab
1943 Principles of Behavior: An introduction to Behavior Theory
His goal was to “use an objective systematic manner..to examine the molar principals of behavior.. …differences are due to different conditions under which habits are set up and function”
Written in the form of theorums and postulates, hypotheses
Very formalized, micromanaging—extremely popular with grad students, just go through book and do the experiments as he instructs
Hull’s approach in creating his theory
Since psychology was not yet sufficiently developed to provide laws, Hull felt justified guessing at nature of psychological laws, stating them as true, and then attempting to prove his laws through experiments
Hull felt it was possible to reconcile views of Pavlov and Thorndike:
- Pavlov’s work was only a case of Thorndikian learning
- Pavlov’s US [food] was also acting as a reward
Restatement of the Law of Effect
Whenever a response is closely followed by the diminuition of the drive stimulus [drive reduction], there will be an increment between strength of the bond between response and stimuli present at the time the response is initiated
In other words, strengthening occurs between any stimuli present when a response results in drive reduction
Hulls main emphasis is on primary reinforcers [food, water, sex, relief from pain etc]
Saw himself as expanding upon Thorndike, but as a more scientific objectifier
Hull Postulate 1: Primary Reinforcers
There is no learning unless a drive [physiological need] is reduced
For a drive to be reduced it must be present
Therefore motivation is essential to learning
A drive may not need to be fully eliminated, it may only need to be diminished
You don’t need enough food to be satiated, just enough to reduce the drive
Learning will proceed in increments
Hull Postulate 2: Secondary Reinforcers
Any stimulus that is present when a primary reinforcment is being supplied will take on the characteristics of the primary reinforcer→ secondary reinforcer
*e.g. mother talks to baby while feeding it, the sound of the mothers voice becomes a secondary reinforcer
Secondary reinforcers increase strength to S-R bonds just like primary reinforcers do
*e.g.white rats fed in a white box, later with no food, they chose the white box with no food over a black box with no food
Type and variety of secondary reinforcers is unlimited
Hull Postulate 3: Secondary Drives
Secondary drive: any stimulus that is present when at the time a primary drive is activated and rapidly reduced can take on the properties of the basic drive and serve as a basis for future learning
Thus if a NS preceded a painful stimulus, that NS will come to cause a drive that functions like pain
Thinking in Pavlovian terms [* this helps me conceptualize but I may be way off target]:
Primary drive =~ UCS
Primary Reinforcer = any response which reduces drive
Secondary Drive = ~ CS+, produces similar but not identical drive which functions like UCS.
Hull Postulate 4: Continuous and Cumulative Learning
Learning is continous and cumulative, every reinforcement adds strength to learning
One implication to this is there is a regular, progressive increase to any learning
*Learning Curves
Learning Curves, Spence, Reinforcement
Theoretically, a learning curve shows a regular pattern
However, pattern is actually erratic for any individual
*Hull is a Thorndike theorist creating Thorndike learning curves
Spence found it necessary to divide groups of learners into slow / med / fast groups to get regular learning curves
Called “Hull-Spence” theory
Reinforcement based theory with contiguity added to it
Hull dismissed contiguity, but Spence brought contiguity into the theory as experiments proved it was needed
Variables that affect course of learning [learning curve]
What is being learned
Readiness of learner
Individual differences
Drive states
Incentive values
Features of performance being measured
Probability of Response Evocation formula
sEr = (D * sHr * k * m)*2/3 – (Ir * sIr)
response evocation =
drive x habit strength x incentive x magnitude x 2/3
minus
reactive inhibition x conditioned inhibition
[original formula sEr = (D * sHr * k) – (Ir + sIr)]
RO: Works as a model, but doesn’t really predict
sEr
probability of this response evocation
s, r, D, k, m
s→ stimulus conditions
r→ response
D→ drive
k→ amount of reinforcement [incentive]
m→ magnitude: variable difference in class of reinforcement • e.g. fast food vs nice restaurant
sHr
habit strength
how often response has been reinforced
Ir
Reactive Inhibition
Function of fatigue or boredom [even though you’re still being reinforced]
**Temporary state—dissipates over time
E.g. ditch digging, you get $10 for each swing of the pickaxe. Eventually muscle fatigue even though you can still make money
After some rest, the next day you’ll be out digging the ditch again
sIr
conditioned inhibition→ learned inhibition
boredom or learned fatigue
It feels good when a response is stopped
Thus, the situation becomes a learned stimulus for wanting to leave
*Does not dissipate over time
**Each time reactive inhibition dissipates, a small part transfers to conditioned inhibition
[this is the presumed method of extinction of tics offered by Yates–(I think)]
Threshold
A threshold is established and then test whether sEr number will get a response
The total must be higher than the threshold
The threshold is not constant and it oscillates depending on internal and external factors
Thus, it impossible to accurately predict the response
RO: “copout clause” [like Guthrie’s Dynamic Situations]
Hull’s T-Maze
Rats and T-maze, learning which way to turn based on food
Associations are learned for any number of stimuli paired with food: e.g. symbols, colors etc
Skinner’s criticism of T-Maze apparatus
T-maze is a fixed trial apparatus like Thorndike’s puzzle box
It does not represent natural responding–animals are free operants, they do not get picked up and reset for a new trial
Skinner was only concerned with rate of response—the only thing that living organisms can be measured on
Hull’s Gradient of Reinforcement
In order to explain the initial movement before turning the corner, Hull proposed a gradient of reinforcement
The easiest, and most immediate learned behavior is that which produces the quickest reward
A chain occurs here, with end goal (r1) having influenced prior responses in an effort to reach this goal (r2, r3, ri)
Little r-g = the goal-related behavior
[“Fractional antedating goal response “]
Little s-g = the feeling related to performing that behavior
*becomes a secondary reinforcer
rg and sg
Attempts to explain a goal-related long chain
Also used to explain foresight, planning, long-term behavior
The response that is easiest to learn is the one next to the reinforcer
Secondary reinforcer but only reinforcing in this particular chain
sg does not have to be related to rg, could be related to something in the apparatus
Hull vs. Tolman: “run” the maze
Hull: rat learns to “run” the maze
Tolman taught rats to run the maze then floods the maze, they cannot run the maze, and they swim the maze and still get to the food box
They did not learn to “run” the maze
They learned something: Tolman—a cognitive map of the maze
Hull vs Tolman
Tolman disproved many of Hull’s premises—said he couldn’t explain the groupings of the chain—all these things all lead to the same goal and that’s how they became associated
i.e. when the rat reaches the goal, it reinforces relationship between all of of the behaviors, similar to Guthrie’s idea that all change stimulus complex
• E.g rat doesn’t need to relearn how to open the door
In response, Hull came up with “habit family hierarchy”
Habit Family Hierarchy
Set of alternate habits integrated by a common goal stimulus and arranged in a preferential order, from the most common response to the least common
Latent Learning
Hull: without motivation, animal will not learn the maze
Tolman disproves this by having them simply explore the maze
Then after a number of trials food is put in both goal boxes
After 2 trials, no difference in performance from the rats who trained with food
therefore, the rats learned maze without food
They could not know it well enough to learn in two trials unless they learned something—latent learning
Hull response: there was internal reinforcement, yes, but without incentive/external reinforcement, there would be no change in behavior
