classical versus instrumental conditioning CS-US versus stimulus-response-reinforcer similarities in induction contingency association prediction difference in role of response classical: preparatory instrumental: determines whether reinforcer will occur common characteristics effects of practice extinction when contingency is eliminated spontaneous recovery dependence on time interval stimulus control blocking configural learning generalization and discrimination associative bias WHAT IS ASSOCIATED? association is stimulus-response-reinforcer, rather than stimulus-response test by devaluating reinforcer (Colwill and Rescorla) push rod to left, food push rod to right, sugar solution taste aversion, pair one reinforcer with lithium chloride decreased rate of response associated with devalued reinforcer associations between responses and neutral stimuli bar press-noise (w/o reinforcement) expt group: noise-food (w/o bar) control group: light-food (w/o bar) lever was reintroduced (but no food was given) expt group pressed bar more often can learn relationship between response and stimulus that follows, even if stimulus is not reinforcing analogy to sensory preconditioning critical for learning complex chains of responses only last of which involves reinforcement example: maze-running secondary reinforcement reverse order: noise-food, then bar press-noise analogy of second-order conditioning noise has become a secondary reinforcer, or conditioned reinforcer many examples in human life money, letter grades WHAT IS THE CONDITIONED STIMULUS? stimulus generalization generalization gradients training: reinforce particular frequency test on range of frequencies, response peaked at training frequency do for several training frequencies negative gradient: don't reinforce particular frequency stimulus discrimination train to peck in response to light, 1000 Hz tone relatively flat generalization gradient (tone doesn't matter) then give light alone without reinforcement sharp gradient to generalization now train to respond to 1000 Hz but not to 950 test on range of frequencies sharpening of generalization gradient shift of maximum to 1050 Spence's theory of discrimination learning additive of positive and negative gradients prediction of peak shift test by train to discriminate between 160 and 256 cm^2 test with 256 and 409, animal chooses 409 because of peak shift this interpretation is opposed by Gestalt explanation: animal learns relation (choose larger stimulus) errorless discrimination learning red-green discrimination (pigeon must peck at key when it is red) difficult to stop pigeon from pecking at green takes many thousands of trials natural tendency of pigeons to peck at lighted objects for food gradual training (Terrace, 1963) red key only alternate with darkened key (which pigeons do not peck) substitute dark key with dim green key increase intensity of green gradually finally pigeon makes proper red-green discrimination learning can occur with little or no error dimensional learning reversal shifts (switch colors) easy because same stimulus dimension difficult because opposite response humans and higher apes find reversal shifts easier young children and nonprimates find nonreversal shifts easier category learning pigeon learning of concept of a tree (Herrnstein et al., 1976) 700 slides of trees and non-trees learned more easily than simple one-dimensional problems able to generalize to new pictures between versus within categories key1: 10 cats key2: 10 cats key3: 10 flowers key4: 10 flowers discriminations between categories are easy discriminations within are hard opposite of a computer WHAT IS THE CONDITIONED RESPONSE? is a specific response being learned? Macfarlane (1930) taught rats to swim through a maze for food able to run through the maze for food Lashley (1924) taught monkeys to solve manipulation problem with one hand could generalize to other hand if first was paralyzed animals are capable of generalizing to other responses Skinner's definition of operant: equivalence class of responses that lead to same reinforcement maze learning radial 8-arm maze place food at end of each arm rats visit about seven out of eight arms in their first eight choices good ability to avoid revisiting arms no specific order of arms rather, good memory of environment instinctive drift train pig to take large wooden coin and place it in a piggy bank learned well given food reinforcement after a few weeks behavior changed would drop the coin root it (dig or turn it up with its snout) toss it up in the air part of natural food-gathering behavior of pigs training frustrated by predispositions autoshaping (training facilitated by predispositions) illuminate key, food pigeons learn to peck, even though there was no contingency reinforce with food or water food: pecked with open beak water: pecked with closed beak interpret as classical conditioning CS (lit key), US (food/water), CR (pecking) ASSOCIATION: CONTIGUITY OR CONTINGENCY? contingency rather than contiguity test by manipulating probability of reinforcement partial reinforcement takes longer to extinguish after training on partial reinforcement resistance to extinction increases as reinforcement rate is lowered because learning conditions are similar to extinction conditions possible features being learned nonreinforced response to eventual reinforcement feeling of frustration to reinforcement familiar in everyday life "superstitious learning" food made available regularly at 15 sec intervals (Skinner) no contingency between behavior and reinforcement pigeons developed highly routinized behaviors counterclockwise turn thrust head into upper corners of its cage interpretation: learned from accidental contiguity speculation: might be cause of superstitious behavior by humans later expts suggested two stages interim behaviors terminal behaviors (always involved some variety of pecking) new interpretation terminal behaviors are autoshaping example: fixed schedules of human eating interim behaviors (watching TV) terminal behavior (going to kitchen and setting table) learned helplessness dogs given painful shocks at random times control: could avoid shocks by pushing panel expt: nothing dogs could do to avoid shock both groups placed in same condition could avoid shock by jumping over barrier after hearing tone control group learned expt group whined and yelped but made no attempt to escape explanation of math phobia associative bias pigeons more readily learn to peck to receive food than to avoid shock readily learn to flap wings to escape shock example: difficult to lean forwards when skiing, due to natural tendency to lean backwards