Animal Learning and Cognition - University

Darwin started the modern study of animal behaviour. Provided a mechanism by which evolution could occur and provided evidence of evolutionary continuity between animals and humans

Romanes relied on anecdotes from less reliable resources and concluded that animals were rational,empathetic and reasoning creatures. He saw a simple progression between intellectual capabilities and the animal's supposed position on the evolutionary scale.

Lloyd Morgan opposed Romanes antropomorphism and called for more careful interpretations. Lloyd Morgan's Canon: "In no case may we interpret an action as the outcome of the exercise of a higher faculty, if it can be interpretted as the outcome of one which stands lower in the psychological scale."

Thorndike studied cats' ability to escape puzzle boxes. There was no sudden decrease in escape trials suggesting that cats did not reason or suddently gain insight. Concluded that animals learn through trial and error. Thorndike's Law of Effect states that if a response leads to a satisfying outcome then it will be strengthened

Watson argued against introspectionism as it was subjective. Instead, he studied the science of behviourism, the study of observable phenomena.

Skinner developed a technique called response shaping, to train animals to perform a behaviour to gain a reward, he called this operant or instrumental learning. Also developed schedules of reinforcement to keep animals performing a response without always recieving a reward. Also developed the conditioning chamber.

Pavlov demonstrated pavlovian conditioning. Conditioned stimulus and unconditioned stimulus pairings leads to the CS alone being enough to produce a conditioned response. During acquisition CR increases to asymptote, when CS is presented alone CR drops off, this is extinction, after a 24-hour rest spontaneous recovery occurs as CS alone produces CR showing relationship is not forgotten in extinction. Also showed that CR was sensitive to properties of the CS, response was lower the more the tone deviated from the original CS. This is a result of stimulus generalisation, when response is lower due to different sstimulus it is known as generalisation decrement.

Banks and Flora ranked animals in a linear progression of intelligence from fish, chickens, sheep, cows, horses, cats, dogs, apes and humans. Justified this progression due to appearance, evolution, aristotle's great chain of being and brain size

However, much of an animals brain is used for motor control so brain size cannot justify more/less intelligence. Jerison controlled for body size and generalised the cephalisation index. Humans and dolphins have a high CI, chimps and elephants slightly lower, dogs and cats then sheep, cows and rats at the bottom.

Anaxorgoras reasoned that all animals are equally intelligent but some are just better able to express it.

McPhail mirrors this. Concluded that results could be explained by differnce in factors like motivation or perception or motor skills rather than just intelligence

Problem: Unexpected between-species differences in speed of learning

• Skard found that rats were as quick as humans to learn how to make an error free run through a maze
• Angermeiser found that in one test fish were quickest but in another they were slowest. Concluded it it difficult to equate perceptual and motivational demands of the test for all species.

Problem: Differences within species as a result of biological revelance of learning relationships between stimuli

• Garcia & Koelling showed that some cues were more biologically relevant than others. Rats easily predict association with saline and illness but not footshock and vice versa with light and shocks

Problem: May distract from other important intellectual capabilities

• Now consider intelligence a combination of memory, reasoning, navigation & communication

Problem: Unexpected between-species differences in speed of learning

• Skard found that rats were as quick as humans to learn how to make an error free run through a maze
• Angermeiser found that in one test fish were quickest but in another they were slowest. Concluded it it difficult to equate perceptual and motivational demands of the test for all species.

Problem: Differences within species as a result of biological revelance of learning relationships between stimuli

• Garcia & Koelling showed that some cues were more biologically relevant than others. Rats easily predict association with saline and illness but not footshock and vice versa with light and shocks

Problem: May distract from other important intellectual capabilities

• Now consider intelligence a combination of memory, reasoning, navigation & communication

Vaughan and Greene showed pigeons pairs of photos; they got a reward for pecking on one but not the other, positions were changed so they couldn't learn position patterns. After training birds could discriminate 320 photos, evidence of a high memory capacity

Animals are able to remember information about time. Periodic timing is the ability of animals to respond to a specific timing. Interval timing is the ability of animals to respond on the basis of specific durations.

Brannon and Terrace showed that monkeys could remember numerosity represented by particular stimuli. Monkeys trained to touch squares with different numbers of dots in numerical order. Used test trials to rule out any other confounding variables (used novel shapes and sized stimuli)

Hernstein, Loveland and Cable showed pigeons photos; half with trees half without. Pigeons eventually able to accuratley discriminate between the 2 categories. Led to the proposal that pigeons could acquire concepts and categorise information.

Fodor proposed innate categories; appropriate for face recognition in humans but not plausible for discimination in pigeons.

Exemplar Learning suggests that animals remember every instance or exemplar of each category. Again, this is not a very plausible explanation.

Feature Learning suggests that if a pigeon learns features commin in all rewarding pictures it will form a positive association with similarities and a negative one with unique features. Explains  why they are good at novel pictures but would expect them to be just as good in training which they were not.

Exemplar Learning with Stimulus Generalisation explains that an animal shows a strong CR to items similar to a trained stimulus and responding decreases as simplarity does.

Recognition test used to test whether an animal has retained information about a previous stimulus

Whitlow presented a loud tone to rabbits and measured change in blood flow to their ears and found that response was stronger to S1 than to S2 presented 60seconds later

Wagner found that if the interval changed from 60-150secs then response to S2 was just as high as to S1. Said that the representation can be found in 3 states; A1 where it is at the centre of an animal's attention; A2 where it is in the periphery of an animal's attention and Inactive where memory is not able to influence behaviour.  A2 state has to decay to inactive before returning to A1 and so if S2 is presented before delay is complete (150s) and matches S1 then habituation is observed

If physiological system is fatigued (Effector Fatigue) then it will not be able to respond to any stimulus. Whitlow explained that when properties of stimulus was different there was no habituation. Since behviour wasn't impaired, effector fatigue was ruled out

Whitlow introduced a distractor as interference and showed that S2 was just as well recieved ruling out receptor fatigue.Showing that habituation involves memory and exists in many species.

Animal visits 4/8 arms of maze and ther is food

All 8 are then opened  and animals have to avoid previously visited ones to find food in remaining arms.

Beatty & Shavalia trained rats to complete normal arm maze with a delay after the 4th arm was entered. During interval, rats completed a second radial arm maze in a different room. This seemed to have no effect on memory for first task

Roberts was only able to produce retroactive interference by making rats complete 3 distractor arm mazes

Memory is relatively short lived at 4 hours.

Miller & Berk trained tadpoles to move between compartments to avoid an electric shock. Animals remembered to avoid the black compartment even after their metamorphis to frogs.

Has 3 phases; consolidation, retention and retrival in which the animal forms a memory trace, stores information and acts on memory when required to.

Consolidation theories state that forgetting is due to inadequate trace laid down or due to the passing of time. Hebb said that memory depends on the virtually permanent formation of circuits of interconnected neurons (consolidation and rehearsal). Duncan supports this, trained rats to move to other side of box after light to avoid shock. ECS immediatley after trial prevented consolidation and impaired rats' ability to remember significance of the light.

Retirval theories state that the formation of a memory is instantaneous and once formed stays intact and that forgetting is due to animal's inability to retrieve information. Dewer, Sara and Hars trained rats to run through a maze with 6 choice points. At beginning it took rats 300s, by the end only took 30s. 1 group given 25-day delay and it took them 150s to complete. 2nd group had 25-day delay but placed in cage next to apparatus for 90s beforehand and it only took them 40s to complete. Suggests reactivation treatment facilitated the retrival process that consolidation theorists thought unimportant

Before Pavlovian conditioning the presentation of a neutral stimulus produced no learnt effect

During conditioning neutral stimulus is paired with and unconditioned stimulus Presentarion of US causes animal to perform a UR After a number of CS-US pairings animal produces CR whenever CS is presented before presentation or even in the absence of the US

Can say that CS actiivates memory or representation of UsCan't assume animal has mental experience of remembering US because that is unmeasurable

Peter Holland and collegues believe CS is able to activate some of the perceptual mechanisms normally activated by the US.

Konorski believed USs have 2 distinct characteristics; Specific which make the US unique; Affective, characteristics which USs have in common, these can be aversive or appetitive. When CS activates representation of specific qualities we can expect CR to mimic UR, known as stimulus substitution. When CS activates representation of affective properties we can expect a prepatory response. If affective property is aversive we may expect a withdrawal of responses kknown as conditional suppression. Criticised for being too simplistic

Siegel demonstrated analysis of drug tolerance by injecting 2 groups of rats with morphine. All rats showed tolerance to drug after a few injections. Group MPM then recieved 12 injections of saline where as group MRESTM recieved no injections. Both groups then injected with morphine again in stage 3. Drug had greater effect on MPM rats as their tolerance had diminished. CS-US pairings of injection drug is broken for MPM rats and so compensatory CR goes into extinction.

Drug tolerance in humans can then be broken down. CS is the context/location that drug is taken. CR compensates for the drug effect when body feels stimulation from injection in context (CS). Addict has to take more of drug for it to have an effect. If drug is taken in a new context means there is a lower compensatory CR because CS is different. So drug effect is greater in a new context.

Contiguity is the idea that events that occur close together will become readily associated.

Rescorla used truly random control and showed that unless their was a clear predicitive relationship then no learning will occur

Garcia and Koelling found that rats learn that a light and clicker signal footshock but not the onset of illness. Also, that flavour of saline can predict illness but not foootshock. Findings were despite that CS-US pairings were contiguous.

Kamin showed that for successful conditioning, US had to be suprising. Group suprise had a more intense footshock in stage 2 so new learning about significance of light occured. Controls had no prior learning so light and noise were both learnt about. Another group had prior training so no learning occured.

Latent inhibition is known as the importance of selective attention in info processing in humans. Less attended to cues means slower learning. If light signalled nothing in first condition then rats were slower to learn it signalled a shock in the second condition than controls who started in C2

Wagner - Attention will be high to novel stimuli

Mackintosh - Attention will be high to novel stimuli

Pearce-Hall - Attention will be high whilst learning is ongoing and low when learning is complete due to distinction between controlled and automatic processing

Kaye and Pearce conducted an experiment with 3 groups; group none where light signalled no outcome; group continuous where light always signalled delivery of food and group partial where light signalled food for only half the trials. Measured orientating response, how often rats looked at light to provide index of how much attention is being paid to the light. OR was initially high for all groups, dropped for groups none and continuous but remained high for group partial.

Pearce-Hall is only theory theory consistent with the results of this experiment. Once learning is complete for none and continuous, attention is low. Animals in partial can't learn what the light signals so their attention remains high.

Reasoning is the ability to combine 2 or more pieces of information to draw a novel concusion. Reasoning can be split i two ways;; deductive is said to occur when the conclusion is neccessitated by the premises and Inductive which is when the conclusion is likely to form the premises - a conclusion drawn from facts.

Dead Reckoning is naviagating by taking account of own body movements. Wehner and Srinivasan showed that ants use dead reckoning. After being moved, ants navigated as if they had not been displaced.

Animals also navigate using landmarks. Collet and collegues trained gerbils to find food a fixed distance and direction from a landmark. When the food was removed for test trial, gerbils searched in incorrect place in respect to the landmark.

Ability to plot course to hidden goal using landmarks is known as piloting and can take place with many landmarks

A cognitive map is something on which an animal identifies their own position and uses it to plan route to goal. Would mean animal can take novel routes, shortcuts and detour around obstacles

Morris trained rats to locate hidden in a pool of water. Experimental group released from new place and appeared to find platform quicker than controls who had no prior experience. However, nearly all rats began in the wrong direction. Possible that they only altered direction once they had explored pool and come across path they had previously used in training

Tolman and collegues conducted many experiments to find evidence of cognitive maps but all seemed to have a flaw

There does not seem to be enough evidence for cognitive maps

Gillian and Premack trained chimp to use language by pointing to symbols on a board. Learnt symbol for "same as" and given a symbolic reasoning task. Sarah understood by analogy the correct answers. Its possible experimenter inadvertedly cued Sarah. Reasoning can only be proven if results were to be replicated.

Caledonian crow given a wire to use to retrieve bucket containing food. After some experience crow was able to retrieve bucket by creating a hook. However, previous training or experience may be important to consider in the production of novel behaviours

Epstein, Kirshuit, Lanza and Rubin found that birds with prior experience were quicker to move box under banana to stand on, to reach for banana then those with no experience

Kohler showed that chimps were able to use poles and boxes to get to food which was out of reach. Said it was evidence of insight, that behaviour came after chimps had internally reasoned. But the chimps already had experience. A later experiment showed that without prior experience they could not reporduce this behaviour

Von Frisch showed that bees transfer information about the location of food using dances. The round dance is used when food is less than 100m away and encourages random searches. The Waggle Dance is used when food is more than 100m away and tells both the direction and distance. Angle between verticle and direction of straight run is the same as the angle bwteen position and food. Length of straight run and proportional number of waggles and how long it takes to complete figure of 8 indicates distance.

Seyfarth and Cheney described alarm calls in vervet monkeys which signal different predators. These are not innate but learnt in childhood.

Chomsky said that because all language conforms to rules of grammar humans have a language acquisition device which wouldn't be found in other animals

McPhail argued that there is no intelligence difference of humans and other species other than human's ability to use language

Hockett argues that, to be considered a language, communication must meet this criteria; Discrete arbituary units, semanticity, displacement and productivity through syntax

Furness, Hayes & Hayes had limited success training orangutans and chimps to speak, partly due to physical limitations of apes' ability to use speech

Gardener & Gardner attempted to train Washoe to use ASL. After 4 years could produce 132 words and string simple sentences together. He saw a swan (which he hadn't seen before) and signed "water bird". Also adopted Loulis who learnt 22 words due to interactions with Washoe  and there is some evidence that Washoe actively trained Loulis.

Premack trained chimp to use symbols or cards which meant different words. Learnt 130 words and had to place them in order to be rewarded (evidence of syntax). Could answer questions correctly. Demonstrates use of discrete and arbituary units and semanticity and displacement.

Terrace trained Nim to use ASL, recorded over 19000 multiword utterances of 5235 types over 18-months. Said there was no grammatical structure. Copied sentences and altered 1/2 words so it had correct syntax but Nim did not elaborate sentences, mean length did not increase

Savage-Rumbaugh trained a chimp and bonobo to use lexigram to create sentences 

Some researchers use comprehension tasks to explain language shown to understand grammar rules