Module 2: Flashcards
Perception (vison) is sometimes thought of as being relatively “dumb” (Fodor, 1990).
How “cognitively smart” is perception?
How would we test this?
(Fodor, 1990).
- He argued it is designed to process
what is in the here and now as opposed
to the future or past; meaning that
infants perception is relatively
unsophisticated.
- For example, our perceptual system is
prone to fall for visual illusions. In the
“mullers illusion” even when we know
the lines are the same length our eyes
still perceive them to be different
lengths.
- Is the line between perception and
cognition blurred for infants (like for
adults)? To answer this, we look at the
early link between perception and
cognition in very young infants.
Is there research showing that perception and cognition link from very early on?
- Yes, there is. The best is from infant
imitation tasks.
Meltzoff & Moore (1977)
Test whether babies (1-hour to 3-day-olds) can imitate the facial gestures of their mothers.
Steps:
1. Babies (1-hour to 3-day-olds) seated in
darkened room (with pacifier in mouth).
2. Light illuminate’s adult’s face making
gesture (e.g., tongue protrusion, mouth
opening, lip pursing = modeling
behaviour).
3. Light extinguished (pacifier removed;
will the infant imitate that behaviour?).
Results:
- Babies (as young a 1 hour old)
demonstrated same facial gesture that
was modelled to them = imitation.
Implications:
- This indicates that babies are drawn to
be like the other person (higher-order
level). This social orientation to be like
other people leads them to imitate their
behaviour.
Why?
Imitation has value:
• Imitation supports cognitive
development by helping children learn
about objects in their environment.
• It is valuable to social and cultural
functioning, and cognitive development
(in module 1 where they learned about
objects in their physical world and how
they behave).
• Imitation provides a social-emotional
connection between infants and others,
between people; imitation is about
social connections with other people
(building relationships).
Controversy about Meltzoff & Moore’s (1977) Imitation findings:
§ Interpretation 1:
- Infants’ imitation is just “dumb”
automatic responses, and due to
general arousal and can be triggered
with a pen!
- Original findings difficult to reproduce
(e.g., Oostenbroek et al., 2017). Lack of
replication suggests the effect, if
present, is very weak.
§ Interpretation 2:
- Imitation is social, and flexible! Studies
have confirmed Meltzoff & Moore’s
findings when testing babies from a
variety of countries, and even new-born
monkeys show the effect!
- Not just automatic imitation machines!
18-month-olds can even modulate what
to imitate by what the emotional
repercussions might be (i.e., they do not
just imitate any behaviour blindly, they
modulate or selectively decide what
behaviours to imitate based on the
emotion they believe it will elicit in
others; annoyed = no imitation but
happy = imitation) by Repacholi,
Meltzoff & Olsen (2008).
- Therefore, imitation is an intelligent
process which infants use to socially
connect with others.
Much of the controversy is driven by Meltzoff & Moore’s (1983) conclusion that successful imitation showcases infants’ representational capacity.
§ What do Meltzoff & Moore mean by
representational capacity?
- For successful imitation, infants must be
able to REPRESENT, RETAIN,
REPRODUCE the event.
- Represent: they must be able to hold a
mental representation or concept about
the event they’re witnessing (i.e.,
modelled behaviour).
- Retain: they must be able to remember
or store this mental representation in
their working memory (whilst in the
dark; after a delay).
- Reproduce: how infants work out how
to imitate or reproduce the actions they
see using their own facial muscles.
- Neonates have a representational
system that allowed them to match their
own body transformations to those of
others.
§ Their conclusion refers to infants’
representational capacity at a
behavioural level, but other researchers
have shown it applies to a neural level
as well.
§ At a neural level, the mirror neuron
system (MNS) is a group of neurons (in
ventral premotor and parietal cortex)
that discharge both when people are
executing an action (imitating) and
when they are observing the same
action performed by other individuals
(modelling) = kinaesthetic-visual
matching.
§ Such “mirroring” function enables
people to better understand and
anticipate the goal of others’ actions
and to imitate them.
*neural evidence to support that babies have the ability to intelligently imitate others.
It’s also plausible that the environment can play a role in infants’ ability to imitate behaviour- Motionese
definition
8 characteristics
why people do it
how could it be applied to AI
§ For example, adults tend to modify their
bodily actions (called Motionese) when
interacting with infants (unconsciously;
exaggerated bodily actions; designed
to facilitate infant learning).
§ MOTIONESE (or infant-directed action)
is characterised by:
(1) greater enthusiasm in adult,
(2) closer proximity to infant
(3) larger range of motion
(4) greater repetitiveness
(5) higher interactiveness,
(6) longer attention to face (to see if they’re
paying attention)
(7) greater simplification of actions
(8) more turn-taking
§ Why, do we do it?
- Infants prefer to watch infant-directed
actions than adult-directed actions
(Brand & Shallcross, 2008);
- It helps them make sense of the action
stream, helps understand causal
structure underlying actions (the
meaning of the action),
- the goals of action/purpose;
- infants grasp that ‘this action is relevant
to you’ and can signal to infants what to
imitate.
§ This is beneficial for Artificial
Intelligence (AI) industry, which shows
that robots learn better from motionese
action segments. They can capitalise on
developmental research to more
effectively train their robots (slow and
exaggerated movements to clearly
show behaviour sequence and goal of
action).
remember spatial, occlusion, and contact ways of perceiving causality in our physical world ( = not in view).
what if they occurred in full view. why is this important to consider?
what are three common techniques to use?
But this raises the question:
- Can infants make sense of causality
phenomena that happen in full view? If
they can do it out of view, there should
be converging evidence on infants
learning causal relationships in and out
of view.
- Collision events are useful to test this.’
What are examples of causal phenomena that occur in full view?
1. Pulling:
- Where one object moves and appears
to pull the other objects to follow
behind it.
2. Smashing:
- where one object hits another and
breaks it.
3. Launching:
- The most basic example.
- When one object makes contact with
another, and causes it to move
So, do infants perceive causality in basic launching events?
Leslie & Keeble, 1987
- Infants by 6-mths do perceive launching
event as having causal impression –
infants “see” an object act on another
object.
Steps:
1. Habituation: - Two groups of 6-month-old infants were
habituated to one of two events; 1) the
direct launching event with immediate
reaction, the red ball moves in a straight
line, hits the green ball and immediately
causes the green ball to move (A causes
B to move); 2) the direct launching event
with a delayed reaction the red ball hits
the green ball and after a delay (0.5sec)
the green ball moves (“A moves and
then B moves”) = control group.
2. Test/Dishabituation: - The clip of the direct/delayed launching
event is played in reverse (they watch
the same one they were habituated to in
reverse).
Results: - In immediate Reaction: Will infants regain
interest and pay attention at Test? YES!
(In terms of causality, roles switched: “B
causes A to move). A pushes B, B is
pushed thus when that causal event is
reversed and the roles of the cause-
effect of the objects are reversed it is a
new contingency relationship they can
learn = dishabituation. - In Delayed Reaction: Will infants regain
interest and pay attention at Test? NO.
(Event is absent of causality: “B moves
and then A moves”). There is no cause-
effect contingency learnt (no association
between red/green moving) therefore,
when it is reversed, they do not care =
no dishabituation.
Implication:
§ Infants go beyond tracking basic spatial
movements, they are attuned to
experiencing the world in terms of
causal agents to identify who is doing
what to whom.
§ They were sensitive to the change is
actor/recipient (cause-effect) role
reversal!
§ Supports Agency.
Why is this important? Infants innate interest of identifying causal contingencies in their physical world.
§ Infants’ awareness of causality can play
a role in understanding the workings of
language (e.g., ability to find meaning
from word order). For example, they
need to be able to identify who is the
subject (actor with an active role that
performs an action that has an effect on
the recipient) and who is the recipient
(of the action and is receiving the
action). The difference between the
cow kicked the man vs the man kicked
the cow.
§ Infants that show a sensitivity to
causality in their physical will later on
will be sensitive to subject verb ordering
in language.
Language then elaborates upon these causal contingencies (refines them) and help them map it onto their physical world.
§ Example:
§ Infants are presented with films without sound side by side. A narrative is played over the speaker that matches the action sequence being modeled in one of the videos. The researchers are interested in whether infants would prefer to look at congruent or incongruent visual images. 16-month-old infants with only two words in their vocabulary are able to successfully complete this task and understand that the person completing the action goes in the first position in the sentence (information is ordered in our sentences; word order is used in every language to map the order of events in our physical world; grammar =syntax understanding in 16-month-old infants).
Recap:
§ Infants can accord the status of causal
agents to objects (i.e., in a dynamic
display between objects, infants are
able to attribute agency; giver-receiver
to the objects).
If this is true,
this is called ___ theory or ___
If this is true,
§ then infants should also perceive that a
gents/objects can act to achieve goals (
we [as adults] act to achieve goals.
§ Can infants perceive that agents/objects
perform actions to achieve some
outcome? (i.e., do they understand that
the actions performed by agents are
goal oriented).
§ Researchers theorise that young infants
expect agents to perform actions in a
rational manner given the circumstances – meaning infants expect that:
- agents act to bring about some goal
(outcome towards which an action is
directed).
- agents act by the most efficient actions
available to realise goals.
§ This theory is referred to as the
teleological model of action
interpretation (i.e., the purpose-based
model of action interpretation).
§ What is this model? For example, gif of
people running after a train. Why are
they doing this action? The goal is to get
onto the train. The most efficient way to
do this is to drop their suitcases and run
after it.
§ In infants, when asked why are rocks
pointy they prefer the teleological
explanations “Rocks are pointy so that
they won’t get smashed by animals.”
§ Adults under pressure will place more
weight into teleological theories even
when they have been trained to go
against (i.e., all ages).
§ Other teleological statements are:
§ “Lightning occurs to release electricity.”
§ “Ferns grow in forests because they
provide ground shade.”
§ Q: Do infants track the goal of the
actioned behaviour from the agent to
the recipient?
Important Distinction:
Teleological model of action interpretation does NOT claim
Important Distinction:
§ Teleological model of action
interpretation does NOT claim infants
have expectations that others have
some mental state representation
(intention or belief or desire) towards
the actions that they are carrying out.
§ Goal ≠ a mental state (in the way that
intention, belief or desire are mental
states)
§ Teleological model of action
interpretation only claims that infants
have expectations that:
- agents act to bring about some goal
(outcome towards which a current
stretch of activity is directed).
- agents act by the most efficient actions
available to realise goals.
Do infants track goals in a teleological manner?
Their study shows that infants expect that agents act to bring about some goal.
(Csibra, Biro, Koos & Gergely, 2003)
Steps:
- Habituation
- Infants (18-mths) habituated to watching
a blue circle follow a red circle around
the screen and then disappear off
screen. The blue ball’s goal seems to
be to get/make contact with the red
circle (goal implied as the contact is off
screen). Researchers are interested in
whether infants sensitive to picking up
the goal of blue’s action.
- Test Phase
- Infants were then randomly assigned to
an incongruent goal or a congruent
goal test phase. In the incongruent goal
test phase, the blue moves past the red
ball. In the congruent goal phase, the
blue ball makes contact with the red
ball.
- Researchers are interested in which
phase they look more at the
incongruent and congruent goal phase.
Results:
- Results suggest that infants interpret
others’ actions as goal-directed. This is
because they look longer at the
incongruent goal video than the
congruent phase which suggests they
have interpreted that the blue ball’s
goal was to make contact with the red
ball (i.e., VOE).
Here’s a study of infants expecting agents to act by the most efficient actions available to realise goals
(Csibra, Biro, Koos & Gergely, 2003)
Steps:
- Infants were randomly assigned to
experimental or control group.
(A) Experimental group
1. Habituation:
- Infants (12-months) watch a path with a
barrier in the middle and a red circle and
a yellow circle either side of it. The red
circle makes expansion/contraction
motions and then the small circle runs to
the barrier retreats and then jumps over
the barrier to make contact with the red
circle.
2. Test Phase:
- Infants in the experimental group were
shown either an old action or a new
action. In the new action sequence, there
is no wall between the two circles and
the small circle moves in a straight line
towards the red circle.
- In the old action video, without the
barrier, the small yellow circle still does
the jumping motion to reach the red ball
and make contact.
- Which one will infants look longer at?
3. Results:
- Infants look longer at Old Action than
compared to New Action; why? from a
theological perspective: infants expect
that agents have goals and that they
achieve this in the most efficient manner
(i.e., rational). The old action of jumping
(when no wall) is inefficient while New
Action of straight movement (when no
wall) is efficient.
- Looked longer rational to behave in the
same way when the barrier is wrong. To
infants the action doesn’t make sense
from a teleological manner if the goal is
for the blue ball to make contact with the
red ball.
(B) Control group
1. Habituation:
- In the control group, the video to which
infants were habituated to had the wall
placed to the side of the screen not
obstructing the yellow balls path to the
red ball. Yet the yellow ball still jumps to
reach the red ball.
2. Test:
- Where infants were either shown the old
action (jumps) or the new action (moves
in straight line) to make contact with the
red ball.
3. Results:
- Infants look equally at Old Action and
New Action; from a teleological
perspective if the object makes an
inefficient jumping motion when there is
no need to make said motion. Infants
realise that the object will not act in a
rational manner and can therefore can
move in any motion. The jumping motion
when the wall is not in the way is an
example of this and therefore since they
expect it to behave irrationally they show
no preference for which one they look at.
- This implies that infants from 12-months
are able to identify the goal of agent’s
actions and develop expectation that the
agent will achieve it in a rational and
efficient manner (i.e., use a teleological
framework to interpreting action).
A different question (other than teleology; and inanimate objects)
Inanimate objects behave differently than animate objects. For example, Inanimate objects move only if there is forceful contact, but humans can move without contact (i.e., self-propel). Do infants make this distinction between animate and inanimate objects? Can they suspend their law of contact for cause-effect contingency learning with inanimate objects when interpreting the actions of animate objects?
Spelke, Phillips & Woodward (1995)
Spelke, Phillips & Woodward (1995)
- Provide evidence that 7-month old
infants reason differently about animate
(people) and inanimate objects
(mechanical car) as reflected in infants
ability to view people as being agents
capable of self-propelled motion.
Steps:
- Infants were allocated into an inanimate
or person condition.
(A) Inanimate object condition
1. Habituation:
- Infants were shown a mechanism cart
which moves to the right side of the
screen, going behind a block, and then
after a delay different object (cylinder)
moved out from behind the block
(looking time decrease = habituation).
- The events were then played in reverse.
2. Test:
- The screen was removed, and infants
were either tested with a contact or no
contact test event. In the contact event,
the cart moves and makes contact with
the cylinder and causes it to move. In
the no contact test event, the
mechanical cart goes along the screen,
stops before it makes contact with the
cylinder, and then the cylinder starts to
move.
- They were then played in reverse.
- Which one will infants look longer at?
3. Results:
- Infants are surprised by the no contact
test event which suggests they are
applying the rule of contact in order to
interpret the sequence of events
(contact to propel the inanimate object).
(B) People condition
1. Habituation:
- Same process as before but the objects
are replaced with people.
- Infants were shown a person which
moves to the right side of the screen,
going behind a block, and then after a
delay different person moved out from
behind the block (looking time decrease
= habituation).
2. Test Phase:
- Contact or no-contact events with
people.
- Where the person makes contact with
the second before they start to move
(cause-effect) or no contact is made and
the second person still comes out from
behind the block.
- Which one do they look at?
3. Results:
- No preference in looking time between
the contact vs no contact conditions.
This implies that infants do not apply
the same rule of contact on inanimate
objects onto humans (animate objects).
Therefore, implying that they have an
understanding that humans have the
ability to self-propel (move without
contact) and were not surprised to see
this happen (i.e., VOE). ‘
Meltzoff (1995)
Provides more evidence that 18 month old infants do reason differently about the motions of inanimate object (mechanical) motion and animate (person) motion.
Steps:
1. Infants were assigned to one of two
conditions: human or machine
demonstrator.
2. Habituation:
- In the human demonstration condition,
infants were shown adults holding a toy
dumbbell as they try to pull it a part, but
they drop it.
- In the machine demonstration
conditions, the machine tried to pull the
toy dumbbell, but they dropped it.
3. Test Phase
- Infants were then given the toy
dumbbell to see if they would copy the
behaviour and try to pull it a part as well
(i.e., fulfill the goal of the agent,
machine/person, is their behaviour the
same or different across conditions?).
4. Results:
- Infants are more likely to pull the toy
dumbbell apart when a human
experimenter models their goal directed
behaviour relative to machines. This
suggest as early 18 months of age that
infants are prone to tracking the goal of
human agents relative to machine
agents.
Summary:
§ Previously, we learned that
§ Here we also learned that infants have
other ways of interpreting actions;
§ Previously, we learned that
- Infants can store a remarkable amount
of information that is not purely visual.
- Infants can imitate others’ actions, and
seems to be suggestive of
representational capacity (higher order;
though debated).
- Infants can encode launching events in
terms of causal agency (agent-recipient).
§ Here we also learned that infants have
other ways of interpreting actions;
- Can interpret actions in terms of a
teleological framework, that actions
produced by agents are rational in being
goal-directed and efficient.
- Infants distinguish between the world of
objects and the world of people
(animate/inanimate objects).
- All suggests that infants are engaging in
a knowledge based, conceptual analysis
of the world.
why is numerosity important?
People are surrounded by objects of all
different shapes and sizes in our
physical world. So how do we make
sense of this? How do we group and
count things?
The use of “numbers” to make sense of
objects in our physical world. Are
infants able to compute numerical
computations?
We adults can conceptualise what we see in terms of numerosity (relationships about greater than, less then etc). Do infants compute numerosity?
the most famous evidence of infants’ apparent computation of numerosity.
Wynn (1992)
Steps:
1. Familiarization:
Five-month-old infants we familiarised
with an opening scene of a single
object being placed on the stage. The
screen moves upwards occluding
(hiding) the object. A hand places a
second object behind the screen and
the hands leave empty.
2. Test phase:
Infants saw one of two events; the first
event included seeing the screen lift to
see two objects (possible) or reveals a
single object (impossible). Which one
would infants look longer at? What
would they be surprised by?
3. Results:
Infants looked longer at the impossible
event (1) relative to the possible event
(2). This implies infants are computing
numerosity when observing the events.
How do we know it is due to numerosity or if they just prefer to look at 2 objects instead of one (visual preference). How do we rule this alternative explanation out?
They included a baseline condition
where infants did not show a
preference for looking at one or
multiple objects over the other.
They may have an abstract
representation of numbers (i.e.,
principles of addition, one object plus
one object = 2 objects; they should
know it shouldn’t be three and look
longer at the impossible event; this not
purely perceptual there is an abstract
understanding of numbers).
The debate of Wynn’s (1992) findings:
(A) Interpretation 1:
Maybe infants merely wonder how can
an object that did exist before suddenly
disappeared (physical impossibility)
rather than understand anything about
the impossibility of the arithmetic (1+1 =
1)!
(B) Interpretation 2:
There is a study showing that infants
pay more attention to the impossibility
of the arithmetic and less so to the
impossibility of the physical situation!
Infants do compute numerical
information!
What study shows this? Simon et al.
(1995)
Simon, Hespos & Rochat (1995)
Tested to what extent physical impossibility (identity of objects) is driving Wynn’s classic findings:
Steps:
1. Familiarization:
- Five-month-old infants see a single
Bernie doll on the stage, the screen
goes up occluding the object and the
hand brings out a second doll placing it
behind the screen.
2. Test Phase:
- The screen is lifted, and infants see one
of four outcomes.
(1) a possible test trial (two dolls)
(2) an arithmetically impossible test trial
(1 doll)
(3) an identity impossible test trial
(physically impossible, Bernie and Elmo
doll arithmetic is correct)
(4) identity and arithmetically impossible
test trial (Elmo doll; physical traits and
number of dolls is wrong).
3. Results:
- Looking time is longer in impossible
arithmetic condition which supports
Wynn’s condition (relative to possible
arithmetic condition).
- Infants looking time for the two
impossible conditions are similar
indicating that infants are not
responding to the physicality of the
outcome.
- If infants were responding to the
perceptual physical characteristics of
the event, we would have expected the
identity & arithmetic impossible event to
be higher than arithmetic impossible
event.
4. Implications:
- Infants don’t dwell on physical
impossibility of outcomes; infants
appear to genuinely compute the
numerical results of arithmetical
operations.
Adults can numerically ___ in terms of kind (property) information!
Adults numerically individuate in terms of kind (property) information!
Adults count objects in their physical world based on kind, they are grouped into categories and counted in groups.
Xu and Carey (1996)
Conducted a study to test whether 10-mth-old infants are able to numerically individuate items in terms of their kind or physical characteristics (i.e., property information).
Steps:
Spatiotemporal Condition
1. Infants were assigned to a property
kind or spatiotemporal condition. In the
spatiotemporal (two distinct objects
being tracked in space and time)
condition infants:
- were shown a screen.
- they were shown two objects (cat &
truck) either side of the screen.
- the image fades to black and the
screen comes back on its own and the
toy truck comes out from behind the
screen.
- as the toy truck goes back behind the
screen and the toy cat comes out form
the other side.
- The cat goes back behind the screen
and the steps are repeated.
2. Infants are able to see an expected or
an unexpected outcome. In the
expected outcome the screen drops to
reveal the cat and the truck. In the
unexpected condition the screen falls
to reveal only the truck.
3. 10-month-old infants look longer at the
unexpected outcome relative to the
expected outcome.
In the Property-Kind condition:
Infants aren’t given any cues that there are two distinct objects (harder)
1. Infants are introduced to the screen.
2. The truck comes out from one side.
3. The truck goes behind the screen and
as it does a cat comes out the other
side.
4. This sequence is repeated.
5. In the test phase, the screen is lowered
to reveal a cat and a truck (expected) or
a truck (unexpected).
Results:
- Infants fail at this task and look longer
at the expected condition where there
are two objects.
- Why is this the case? In order for them
to track these two objects well they
have to be mentally equipped to track
the kind of objects (cat vs truck;
assigning objects to kind categories
make it easier for us to track them
across space and time).
- 10-month-old infants have developed
sufficient language skills to do this
without being shown at the start that
there are two distinct objects to track.
- Infants do not use kind membership
(objects do not change kind
membership) until much later to
individuate events – dependent on
learning language (look a truck! look an
animal! Category labels).
But, Absence of evidence is not evidence of absence.
- Some studies show that is very young
infants (4-mth-olds) are shown the
different functions of different objects
(like tools), these young infants are able
to individuate objects by kind
information (the task may be too difficult
to capture the process if present).
