Transcript Chap8bAlt

PSY 402
Theories of Learning
Chapter 8 – Stimulus Control
Space and Time in Humans and Animals
Visual Perception in Pigeons


Pigeons recognize objects in the same way as
humans do.
Biederman’s geons are recognized by both
humans and pigeons.


Geon (geometric ion) – a primitive component of
an object.
When geons were preserved, recognition was
better than when they were disrupted.
Conjunction of Features


Both pigeons and humans experience “pop
out” of single features, but not for the
conjunction of features (combined features).
Treisman’s studies were replicated:


Conjoined features require additional attention
and more processing for pigeons & humans.
Peck location was determined using a
touchscreen.
Same-Different Detection


Pigeons can make decisions based on
variability of a pattern.
Pigeons were trained to make same-different
discriminations:




Rows need not be perfectly aligned (straight).
The number of items in the array doesn’t matter.
Entropy (number of different items) is important
– 8 items are needed for accuracy.
Is detection based on texture or evaluation?
Attention Processes



Differential reinforcement of different aspects
of a stimulus causes pigeons to pay more or
less attention to them (dot color vs tone).
Attention to the features of cryptic prey
increases with reinforcement and decreases
with split attention.
Priming improves attention to important
features in runs of trials, but disappears as the
task becomes automatic.
Pigeon Working Memory

Pigeon working memory is very short.




Delayed matching to sample task (DMTS) –
performance is at chance with a 6 sec delay.
Increased exposure to the sample improves
performance.
Practice improves performance (60 sec retention)
Interference hurts performance – changing
anything before or after the task interferes.
Rat Working Memory

An elevated radial maze was used to test rat
memory for the location of food.




Rats are very accurate at remembering which
arms they have already visited, and not repeating.
Extra-maze (outside) cues are used to remember.
When cues were attached to a curtain and then
rotated, the rats had to rotate their visits.
Rats learned to avoid the arms that never had
food, using long term memory.
Two Explanations


Retrospective code – the rat might remember
where it has been before.
Prospective code – the rat might remember
places it has not yet visited (looking forward).


Memory load is highest at the task’s beginning.
Both rats and pigeons use both kinds of codes.

Delays inserted into the task were most disruptive
in the middle, not at beginning or end.
Reference (Long Term) Memory

It is difficult to determine which kind of long
term memory is used in conditioning because
animals cannot speak.


Procedural, declarative, semantic, episodic.
Episodic memory was demonstrated in jays
using mealworms & peanuts.

They learned that over time, the mealworms
might “spoil” and so searched for peanuts instead
Behavior is Sensitive to Time

Animals have an internal sense of time.




When activities occur at certain times, animals
can use time of day cues as CSs or occasion
setters to predict or locate food.
Circadian rhythms govern activity.
Zeitgeibers link rhythms to the environment.
Animals can also detect time intervals:


Inhibition of delay in classical conditioning.
Scallops in fixed interval (FI) operant schedules.
Interval Timing Experiments


Temporal generalization gradients can be
produced by reinforcing animals for
responding to a signal of a specific duration.
Peak procedure – a stimulus is presented, then
the first response after some time interval is
reinforced (the animal must wait until then).

Shorter intervals were judged more accurately.
Scalar Property

The ability of animals to judge longer and
shorter intervals equally well is an example of
Weber’s law.



Perceived differences are a constant proportion of
the value being judged.
The amount of error in longer intervals is
proportionate to the length of the interval.
The animal may be comparing the lapsed time
to the previously reinforced interval.
Temporal Bisection

Skinner gave rats a task where they had to
press the right lever after a short interval and
the left lever after a long interval.



Rats pressed the right and left levers equally often
when intervals in the middle were presented.
This result was the same when 2/4 sec and 4/8 sec
intervals were used.
Regardless of intervals, the rats always
behaved as if they were comparing ratios.
How do they do it?

Internal clock model – clock, memory and a
comparator.



Information processing approach.
Multiple oscillator model – oscillators start
with the task and are monitored in working
memory.
Multiple-time-scale model – memory strength
is associated with the reinforcer.
Cues Guiding Spatial Behavior



Dead reckoning – an internal sense of
direction guides behavior.
Beacons – cues near the goal that guide
behavior.
Landmarks – cues not near the goal but with a
fixed relationship to it.

Humans use landmarks differently than pigeons.
Geometric Models

Chang found that rats use geometric
information in addition to cues to find food.



Preference for the opposite corner (a mistake)
suggested that cues alone were not being used.
When cues were removed, preference for the
opposite corner (a mistake) increased.
One short wall and one long one exists for both
the correct choice and the wrong one.
Evidence for Mental Maps

Radial maze – rats orient toward arms then
make microchoices about which to enter.


Memory guides “blind” choices obscured by
doors.
Water maze – rats must swim to find a
submerged platform.

Rats found the platform even with the start place
changed – landmarks from the room were used.