generalization and discrimination

generalization

the transfer of past learning to novel events and problems

when similar stimuli predict similar outcomes

discrimination learning

the process by which animals or people learn to respond differently to different stimuli

concept formation

the process by which we learn about new categories of entities in the world, usually based on common features

generalization gradient

a curve showing how changes in the physical properties of stimuli correspond to changes in responding

a generalization gradient shows that an animal’s response changes in a graded fashion

depending on the degree of similarity between a test stimulus and the original training stimulus

after training in which a single stimulus has been reinforced repeatedly, generalization gradients

show a peak (point of maximal responding) that corresponds to the original stimulus on which the animal was trained

an upside down U generalization gradient

suggests that animals expect the chance that two stimuli will have the same consequence drops off sharply as the stimuli becomes more distinct

generalization gradients can predict

the likelihood that the consequences of one stimulus will be the same as that of other similar stimuli

consequential region

a set of stimuli in the world that share the same consequence as a stimulus whose consequence is already known

consequential region

a set of stimuli in the world that share the same consequence as a stimulus whose consequence is already known

simple network model

has a single input node for each of five possible colors of light

stimulus representation

the form in which information about stimuli is encoded within a model or brain

discrete-component representation

each individual stimulus (or stimulus feature) corresponds to one element (node) in the model

discrete-component representations fail

in cases where stimuli have a high degree of physical similarity since the models then produce unrealistic generalization gradients

distributed representation

information is coded as a pattern of activation distributed across many different nodes

the stimulus “yellow” is represented by the combined activity of three nodes

it’s representation is distributed across multiple nodes

the are fixed weight between input nodes

and internal representation nodes

there are associative weights (modified with learning)

between internal nodes and output node

stimulus- generalization gradient of the trained distributed-representation network model generates

peak responding to the trained stimulus (yellow)

decreased responding for stimuli that are increasingly different from the trained stimulus

stimulus control

the influence of cues in the world on an organism’s behavior

first group of pigeons (discrimination learning experiment)

trained to respond to a 1000 Hz tone with a key peck for food

second group of pigeons (discrimination learning experiment)

trained to respond to 1000 Hz tone with a key peck for food

but response to 950 Hz tone would result in no food

interdimensional discrimination

the two stimuli in the experiment differ within a single dimension (tone)

extradimensional discrimination

when animals learn to discriminate between stimuli that differ across multiple dimensions (like tones and light)

sensory preconditioning

training in which presentation of two stimuli together as a compound results in a later tendency to generalize what is know about one of these stimuli to the other

sensory preconditioning is usually tested in three phases

assocition created between tone and light during phase 1. in phase 3 the animal generalizes from the light to the tone (OUTCOME BASED GENERALIZATION)

acquired equivalence

a learning and generalization paradigm in which prior training in stimulus equivalence increases the amount of generalization between two stimuli, even if those stimuli are superficially dissimilar

negative patterning

a behavioral paradigm in which the appropriate response to individual cues is positive, whereas the appropriate response to their combination (pattern) is negative (no response)

negative patterning and eyeblink conditioning

each condition is trained separately (tone-airpuff, light-airpuff, tone+light- no airpuff)

with training, the negative-patterning task

can be mastered by many animals, including humans

in the primary auditory cortex (A1)

neurons respond to auditory stimuli of different frequencies

tonotopic representation

spatial arrangement of where sounds of different frequency are processed

tones close to each other in terms of frequency

are represented in neighboring regions of the brain

low to high frequency gradient from one end

of the primary auditory cortex to the other

richard thompson established a direct relationshop between behavioral properties of auditory generalization

and certain anatomical and physical properties of the auditory cortex

animals with lesions to primary auditory cortex

can still respond appropriately to a conditioned auditory stimulus

suggesting that primary auditory cortex is not needed

lesions to primary auditory cortex results in animals responding to all frequencies

overgeneralization- flat generalization gradient

pontine nuclei receives input from a variety of cortical areas

including auditory cortex

plasticity of cortical representations experiment

pair 2,500 Hz tone with shock

record single neurons in primary auditory cortex before and after training

after training, the response of an auditory cortex neuron changed from being most responsive to a 1,000 - Hz tone

to being most responsive to tones (stimulus must be meaningful) nearer to the training frequency (cortical remapping)

presenting tone and shock separately

results in no change

tone alone

results in decrease in coding of that frequency (habituation)

several brain regions determine

whether a stimulus merits cortical remapping

nucleus basalis

a small group of neurons located in the basal forebrain; these neurons deliver acetylcholine to the cortex

acetylcholine is important for

cortical plasticity

tone paired with nucleus basalis electrical stimulation (rather than shock)

results in cortical remapping

damage to the basal forebrain causes

anterograde amnesia

sensory preconditioning is blocked in animals

with lesions to the hippocampus

neophobia

fear of new things

thigmotaxis

wall hugging

problems with hinman’s experiment

rats tended to sit at one reward port and try to get reward after every tone (attention and temporal structure)

how hinman gets the rats to pay attention (turn into a discrete-trial based paradigm)

have the rat initiate the trial by touching a sensor

shaping- nose poke triggers tone + food

touch sensor plays a tone (S) → food available at correct reward port