\def \ititle {04: Can Humans Perceive Causal Interactions?}
\begin{center}
{\Large
\textbf{\ititle}
}
\iemail %
\end{center}
three videos
Brian Scholl
what did you see?
[I'm about to talk about illusory causal crescents. I first show them two videos.
This is a full overlap video.
You can drag the slider to show them that it's full overlap, but first ask them what they see.]
Brian Scholl
what did you see?
[This is a causal capture video with full overlap.
Focus on the top sequence. Tell me what you saw!
You can drag the slider to show them that it's full overlap, but first ask them what they see.]
Brian Scholl
what did you see?
Normally in this case people report the impression that the top sequence collided.
That is, they didn't pass, they collided.
Contrast shape with monetary value.
People sometimes talk about being able to detect monetary value.
Presumably we perceive visual characteristics and infer monetary value.
Consider causal interactions.
Where do they fit in?
Are they perceptually
experienced in whatever sense the shape and motions of things are?
‘When we consider these objects with the utmost attention, we find only
that one body approaches the other; and that the motion of it precedes
that of the other, but without any sensible interval.’
\citep[p.~77]{Hume:1739lj}
Hume 1739-40 [1978]: 77
‘the account flies in the face of our common-sense conviction that we do
perceive causal relations all the time. The experience of perceiving one
event following another is really quite different from the experience of
perceiving the second event as caused by the first’
‘the researches of Michotte and Piaget would seem to support our
common-sense view’
\citep[pp.~114-5]{Searle:1983tx}
Searle, 1983: pp. 114-5
Sometimes ‘a causal impression arises, clear,
genuine, and unmistakable, and the idea of cause can be derived from it
… in just the same way as the idea of shape or movement can be derived
from the perception of shape or movement.’
\citep[p.~270--1]{Michotte:1946nz}
Michotte 1946 [1963]: 270–1
‘the causal perception is the perception of the work of a mechanical force,
just as the impression of the movement of a car is the perception of its
displacement in physical space’
\citep[p.~228]{Michotte:1946nz}
Michotte 1946 [1963], p. 228
This looks like a straightforward contradiction, but we have to be careful ...
‘This causal impression, however, would have been for him [Hume] ...
nothing but an illusion of the senses, as is shown by his
views with regard to the feeling of effort. ... [I]t is probable that his
[Hume’s] philosophical position would not have been affected in the
least.’
\citep[p.~256]{Michotte:1946nz}
Michotte, 1946 p. 256
‘In a great boulder rolling down the mountainside and flattening the wooden hut in its path we see an exemplary instance of force … these mechanical transactions … are directly observable (or experienceable)’
\citep[p.~118]{Strawson:1992yh}
Strawson 1992, p. 118
‘just as the visual system works to recover … physical structure … by inferring properties such as 3-D shape, so too does it work to recover … causal … structure … by inferring properties such as causality’
\citep[p.~299]{Scholl:2000eq}
Scholl & Tremoulet 2000, p. 299
?
‘we seem to be as far as ever from deciding whether the hypothesis is
true: whether we perceive launchings rather than recognizing them by
means of stored patterns in long-term memory.’
\citep[p.~92]{rips:2011_causation}
Rips 2011, p. 92
How to Get Beyond Intuition?
\section{How to Get Beyond Intuition?}
\section{How to Get Beyond Intuition?}
Can humans perceive causal interactions?
Debate based on intuitions appears to be going no where.
Is there a way to answer this question (or a refinement of it)
without relying only on what philosophers say about their experiences?
The question for this section is,
Can humans perceive causal interactions?
Let my try and show you stimuli that were used in an experiment (without yet telling you
anything about the experiment). What do you see?
[If the animation doesn't work, there's a static version on the next slide.]
Thines et al (1991)
OK, so adults: (a) verbal reports. So what?
‘There are some cases … in which a causal impression arises, clear, genuine, and unmistakable,
and the idea of cause can be derived from it by simple abstraction in just the same way as the
idea of shape or movement can be derived from the perception of shape or movement’
\citep[p.\ 270--1]{Michotte:1946nz}
Adults will also report experiencing causal interactions including pullling, ...
Scholl & Tremoulet 2001, figure 2
... disintegration ...
... and bursting.
Heider & Simmel 1946, figure 1
Can humans perceive causal interactions?
So can humans, adult and infant, perceive causal interactions?
So far I don't think we have strong reasons to accept that they do.
In infants we have discrimination and in adults we have verbal reports.
But we shouldn't trust verbal reports.
After all, people will say all kinds of things about their experiences.
This is nicely illustrated by a famous experiment on apparent behaviour by \citet{Heider:1944ts}.
How to get beyond intuition?
Michotte: the experience of launching depends on interactions among various factors including
the relative speeds of the two objects
the delay between the first and second objects’ movements
the spatial gap between the two objects
the trajectories of the two objects.
But how does this help us? Importantly, tiny variations in the parameters will make
big differences in the experiences reported.
Let me illustrate this for the delay between the objects' movements.
adults: (b) they can discriminate between short gaps and long gaps.
That is, the can discriminate gaps of around 50ms.
Michotte 1946 [1963], p. 115 table IX (part)
Maybe this is clearer as a figure.
Michotte 1946 [1963], p. 115 table IX (part)
People can distinguish between stimuli that differ only in that the gap between two movements
is approximately 50ms longer in one than the other.
A 50ms difference makes the difference between reporting launching and reporting two movements.
We need to do more to understand the effect, ...
NB: Rips 2011 notes that missing out the third category of judgement
makes the case for categorical perception seem different than it is.
Consider an encounter with three two-object movements where the delays between movements are 50, 100 and 150ms.
1. The phenomenal difference between the first two encounters is larger than the phenomenal difference between the second two.
2. This difference in differences is a fact in need of explanation.
3. The fact cannot be explained by perceptual experience of objects or their motion.
4. The best explanation for (1) is that we perceptually experience causal interactions.
An alternative argument ...
‘… why it is that in our experiments certain particular conditions were
found necessary in order to give rise to a causal impression. They
correspond to the different characteristics of reproduction. …
anyone not very familiar with the procedure involved in framing the
physical concepts of inertia, energy, conservation of energy, etc., might
think that these concepts are simply derived from the data of immediate
experience’
\citep{Michotte:1946nz}
Michotte, 1946
(1) A distinctive experience occurs under certain conditions.
(2) The best explanation for (1) is that the experience in question is the experience of a collision.
The question was how we can get beyond intuition in understanding the verbal reports.
I don’t think the argument offered about how to explain the conditions under which the
launching effect works is obvious successful.
But I do think it’s right to ignore verbal reports in favour of an attempt to get at the
nature of the processes that explain the launching effect.
How to get beyond intuition?
The launching effect: detecting a 50ms difference in the delay between two movements.
Part of the answer is this. We don't worry about the content of the verbal reports.
We just focus on the fact that their content changes depending on a tiny, 50 millisecond
difference in the delay between two movements.
Call this \emph{launching effect}.
This doesn't tell us what people are detecting.
But it does tell us that the effect is not merely confabulation or making it up.
So we have taken a tiny step beyond intuition. But we also have to answer two questions.
How is launching detected? For example, does it involve perceptual processes?
Why is a delay of up to around 70ms consistent with the launching effect occuring?
The Launching Effect and Perceptual Processes
\section{The Launching Effect and Perceptual Processes}
\section{The Launching Effect and Perceptual Processes}
How to get beyond intuition?
The launching effect: detecting a 50ms difference in the delay between two movements.
How is launching detected? For example, does it involve perceptual processes?
Why is a delay of up to around 70ms consistent with the launching effect occuring?
The question is,
How is launching detected? For example, does it involve perceptual processes?
Three bits of evidence are relevant ...
1. Illusory causal crescents
2. Adaptation
3. Apparent motion (Kim et al)
Kim et al 2013, S6
Kim et al 2013, figure 2
Kim et al 2013, figure 3
The question is,
How is launching detected? For example, does it involve perceptual processes?
Three bits of evidence are relevant ...
1. Illusory causal crescents
2. Adaptation?
3. Apparent motion
[see appendix]
The question is,
How is launching detected? For example, does it involve perceptual processes?
Three bits of evidence are relevant ...
1. Illusory causal crescents
2. Adaptation
3. Apparent motion (Kim et al)
Guess how the launching effect works! A natural thought is this: first you
perceive objects, then you identify causal interactions based on contiguity
etc. This turns out to be completely wrong.
Guess how the launching effect works ...
The impression of launching is judgement-independent.
So it can't be a consequence of thinking about the interaction.
Still, it might be a consequence of perceiving objects in certain relations to each other.
However a key finding shows that this is wrong.
Surprisingly, we don't first perceive objects and then get the launching effect;
rather, the launching effect is tied up with perceptual process of identifying objects' surfaces.
judgement-independent
Thines et al (1991)
[I'm about to talk about illusory causal crescents. I first show them two videos.
This is a full overlap video.
You can drag the slider to show them that it's full overlap, but first ask them what they see.]
[Static images follow in case video doesn't work.]
what did you see?
Normally, if the two balls overlap completely, subjects report seeing a single object
changing colour.
[This is a causal capture video with full overlap.
Focus on the top sequence. Tell me what you saw!
You can drag the slider to show them that it's full overlap, but first ask them what they see.]
[Static images follow in case video doesn't work.]
what did you see?
Normally in this case people report the impression that the top sequence collided.
That is, they didn't pass, they collided.
causal capture
Causal capture is described by \citep{Scholl:2002eb}.
As I said,
normally, if the two balls overlap completely, subjects report seeing a single object
changing colour.
But if we show subjects a sequence like the launching effect but where the first square
overlaps the second's position before it moves. When this event is shown is isolation
almost all subjects see it as a single object changing colour. But when the event is
shown with an
unambiguous launching effect nearby, almost all subjects now see the 'overlap' event as
a launching.
Causal capture means that we can show subjects a sequence with complete overlap and
still have the report a causal effect.
Why do we care about causal capture? Because it gives rise to illusory causal
crescents ...
[Now I'll explain illusory causal cresecents.]
Scholl and Nakayama 2004, figure 2 (part)
Here's a static image representing the sequence you saw first, when there was full overlap.
‘when there is a launching event beneath the overlap (or underlap event) timed such that
the launch occurs at the point of maximum overlap, observers inaccurately report that
the overlap is incomplete, suggesting that they see an illusory crescent.’
\citep[p.\ 461]{Scholl:2004dx}
Why does the illusory causal crescent appear? Scholl and Nakayama suggest a
‘a simple categorical explanation for the Causal Crescents illusion: the visual system,
when led by other means to perceive an event as a causal collision, effectively
‘refuses’ to see the two objects as fully overlapped, because of an internalized
constraint to the effect that such a spatial arrangement is not physically possible.
As a result, a thin crescent of one object remains uncovered by the other one-as
would in fact be the case in a straight-on billiard-ball collision where the motion
occurs at an angle close to the line of sight.’
\citep[p.\ 466]{Scholl:2004dx}
*here or later?
Contrast Spelke’s view.
‘objects are conceived: Humans come to know about an object’s unity, boundaries, and
persistence in ways like those by which we come to know about its material composition
or its market value.’
\citep[p.\ 198]{Spelke:1988xc}.
Scholl and Nakayama 2004, figure 4
‘Figure 4. Observers' accuracy in experiment 1 at judging the amount of intersection in the test event, expressed as the magnitude of the resulting underestimatesöie the actual intersection minus the perceived intersection. These data are categorized by context condition and amount of actual intersection, and plotted as percentages of the fixed width (ie the diameter; 2.87 deg) of the object. The magnitude of the illusory causal displacement is thus the underestimate in the Launch condition minus the underestimate in the No Context condition.’
Scholl and Nakayama 2004, figure 5
(*This figure just shows when the overlap event was perceived as causal.)
The question is,
How is launching detected? For example, does it involve perceptual processes?
Three bits of evidence are relevant ...
1. Illusory causal crescents
2. Adaptation
3. Apparent motion (Kim et al)
Actually illusory causal crescents seem to take us further
Illusory causal crescents fit perfectly with Spelke’s brilliant insight.
‘object perception reflects basic constraints on the motions of physical bodies …’
(Spelke 1990: 51)
\citep[p.\ 51]{Spelke:1990jn}
‘A single system of knowledge … appears to underlie object perception and physical reasoning’
\citep[p.\ 175]{Carey:1994bh}
(Carey and Spelke 1994: 175)
I think there's something here that should be uncontroversial, and
something that's more controversial.
It can’t be knowledge, but let’s not worry about that because there’s a worse objection.
Rips’ objection (2011, p. 92) [in appendix]
‘This possibility, though, is not one that advocates of Michotte’s
hypothesis have taken up. The favored view is one in which separate
modules are responsible for descriptions of objects and of their
mechanical interactions, with central mechanisms then resolving potential
conflicts between them.
For example, Leslie (1988) argues that perceptual
modules do not by themselves settle inconsistencies between the spatial
positions of objects and their causal interactions. In certain
illusions, adults tend to perceive solid objects passing through each other
(Leslie cites the Pulfrich double-pendulum illusion; see Wilson &
Robinson, 1986). Thus, the module for object tracking doesn’t prohibit
interpretations of physical events that are causally impossible. Of
course, people realize that such events cannot really be taking place and
are surprised when they perceive them, but that is because they have
principles stored in long-term memory—for example, the principle that two
solid objects cannot be in the same place at the same time—that classify
these events as illusions or anomalies. Leslie’s argument suggests
that people individuate objects and calculate their causal relations by
means of separate mechanisms; thus, we can’t count on causal constraints
being part of the object-tracking module. If Leslie is right, there is
reason to question Butterfill’s (2009) conjecture that ‘object
perception and causal perception are one and the same process’’ (p.
421).’
Response: it was too simple to say that
‘object perception and causal perception are one and the same process’ or to
talk about ‘object perception and physical reasoning’ because object perception
is a complicated thing with many aspects.
What Carey and Spelke (and following them, Butterfill) should have said is that
there is a system of object indexes which assigns indexes in ways that reflects
physical constraints such as solidity on objects’ motions, and this is what gives
rise to the launching effect.
The question is,
How is launching detected? For example, does it involve perceptual processes?
Three bits of evidence are relevant ...
1. Illusory causal crescents
2. Adaptation
3. Apparent motion (Kim et al)
How to get beyond intuition?
The launching effect: detecting a 50ms difference in the delay between two movements.
How is launching detected? For example, does it involve perceptual processes?
Why is a delay of up to around 70ms consistent with the launching effect occuring?
Answering the second question gives us a hint about this ...
Ingeniously, Michotte compares launching with the movement of a single object. The single
object moves half way across a screen then pauses before continuing to move. Michotte found
that the longest pause between the two movements consistent with subjects experiencing them
as a single movement is around 80ms, exactly the longest pause consistent with experiences
characteristic of launching \citep[pp.\ 91--8, 124]{Michotte:1946nz}. Accordingly, the
experience characteristic of launching appears to require that the two movements be
experienced as uninterrupted—--this is why they can be separated by a pause of up to but
no longer than 80ms.
But what is the significance of the fact that you get a launching effect only when
the movements are perceptually consistent with the movements of a single object?
It‘s that no delay is perceived.
Ok, but what is the nature of these perceptual processes?
Object indexes ...
Object Indexes
\section{Object Indexes}
\section{Object Indexes}
Perception involves a system (at least one) of indexes which attach to objects
These object indexes can be thought of, roughly, as mental analogues of the
pins that an old fashioned logistician sticks into a map in keeping track of
supply trucks.
When things go well, the movements of trucks on the ground are mirrored by
the movements of pins on the map.
The key characteristic of the pins is this: ignoring re-use, if you have
the same pin at two times, then the trucks it points to at those times are
one and the same truck.
Michotte thought of the launching effect as bound up with the perception of
objects and motion.
[This motivates thinking about object indexes ...]
‘the movement performed by object B appears simultaneously under two
different guises: (i) as a movement (belonging to object A), (ii) as a
change in relative position (by object B)’
\citep[p.~136]{Michotte:1946nz}
Michotte 1946 [1963], p. 136
‘the physical movement of the object struck gives rise to a double
representation. This movement appears at one and the same time (a) as a
continuation of the previous movement of the motor object, and (b) as a
change of relative position (a purely spatial withdrawal) of the projectile
in relation to the motor object.’
\citep[p.~140]{Michotte:1946nz}
Michotte 1946 [1963], p. 140
Suppose you are shown a display involving eight stationary circles, like
this one.
Four of these circles flash, indicating that you should track these circles.
All eight circles now begin to move around rapidly, and keep moving unpredictably for some time.
Then they stop and one of the circles flashes.
Your task is to say whether the flashing circle is one you were supposed to track.
Adults are good at this task \citep{pylyshyn:1988_tracking}, indicating that they can use at least four object indexes simultaneously.
(\emph{Aside.} That this experiment provides evidence for the existence of
a system of object indexes has been challenged.
See \citet[p.\ 59]{scholl:2009_what}:
\begin{quote}
`I suggest that what Pylyshyn’s (2004) experiments show is exactly what they intuitively
seem to show: We can keep track of the targets in MOT, but not which one is which.
[...]
all of this seems easily explained [...] by the view
that MOT is simply realized by split object-based attention to the MOT targets as a set.'
\end{quote}
It is surely right that the existence of MOT does not, all by itself,
provide support for the existence of a system of object indexes.
However, contra what Scholl seems to be suggesting here, the MOT paradigm
can be adapated to provide such evidence.
Thus, for instance, \citet{horowitz:2010_direction} show that, in a MOT paradigm, observers
can report the direction of one or two targets without advance knowledge of which
targets' directions they will be asked to report.)
Pylyshyn 2001, figure 6
There is a behavioural marker of object-indexes called the object-specific preview benefit.
Suppose that you are shown an array of two objects, as depicted here.
At the start a letter appears briefly on each object.
(It is not important that letters are used; in theory, any readily
distinguishable features should work.)
The objects now start moving.
At the end of the task, a letter appears on one of the objects.
Your task is to say whether this letter is one of the letters that appeared at the start or whether it is a new letter.
Consider just those cases in which the answer is yes: the letter at the end is one of those which you saw at the start.
Of interest is how long this takes you to respond in two cases: when the letter appears on the same object at the start and end, and, in contrast, when the letter appears on one object at the start and a different object at the end.
It turns out that most people can answer the question more quickly in the first case.
That is, they are faster when a letter appears on the same object twice than when it appears on two different objects
\citep{Kahneman:1992xt}.
This difference in response times is the
% $glossary: object-specific preview benefit
\emph{object-specific preview benefit}.
Its existence shows that, in this task, you are keeping track of which object is which as they move.
This is why the existence of an object-specific preview benefit is taken to be evidence that object indexes exist.
Kahneman et al 1992, figure 3
The \emph{object-specific preview benefit}: ‘observers can identify target
letters that matched the preview letter from the same object faster than
they can identify target letters that matched the preview letter from the
other object’ \citep[p.\ 2]{Krushke:1996ge}.
Kahneman et al 1992, table 4
object index / file
To clarify terminology,
I should say that whereas I’m talking about object indexes,
researchers more typically interpret this research in terms of object
files.
I’m sticking to object indexes rather than object files for
reasons of simplicity and caution.
If you believe in object files then you can interpret what I’m saying
as referring to object files.
And if you have doubts about object files, you might still have reason
to accept that a system of object indexes exists.
Assignments of object indexes can conflict with beliefs and knowledge states.
Scholl 2007, figure 4
In this scenario,
a patterned square disappears behind the barrier; later a plain black ring emerges.
If you consider speed and direction only, these movements are consistent with there being just one object.
But given the distinct shapes and textures of these things, it seems all but certain that there must be two objects.
Yet in many cases these two objects will be assigned the same object index \citep{flombaum:2006_temporal,mitroff:2007_space}.
So one signature limit of systems of object indexes is that information about speed and distance can override information about shape and texture.
This system of object indexes
does not involve belief or knowledge
and may assign indexes to objects in ways that are inconsistent with
a subject’s beliefs about the identities of objects
\citep[e.g.][]{Mitroff:2004pc, mitroff:2007_space}
What principles characterise how object indexes are assigned and maintained?
To see the need for principles,
return to the old-fashioned logistician who is keeping track of supply trucks.
In doing this she has only quite limited information to go on.
She receives sporadic reports that a supply truck has been sighted at one or another location.
But these reports do not specify which supply truck is at that location.
She must therefore work out which pin to move to the newly reported location.
In doing this she might rely on assumptions about the trucks’ movements being constrained to trace continuous paths, and about the direction and speed of the trucks typically remaining constant.
These assumptions allow her to use the sporadic reports that some truck or other is there in forming views about the routes a particular truck has taken.
A system of object indexes faces the same problem when the indexed objects are not continuously perceptible.
What assumptions or principles are used to determine whether this object at time $t_1$ and that object at time $t_2$ have the same object index pinned to them?
[ducks picture]
Is one object index assigned or two?
Assigning object indexes requires segmentation.
Spelke, 1990 figure 2a
cohesion—‘two surface points lie on the same object only if the points are linked by a path of
connected surface points’
rigidity—‘objects are interpreted as moving rigidly if such an interpretation exists’
Consider a stick moving behind a screen, so that the middle part of it is occluded.
Assigning one index even though there is no information about continuity of surfaces
may depend on analysis of motion.
principle of continuity---
an object traces exactly one connected path over space and time
Franconeri et at, 2012 figure 2a (part)
[Here we’re interested in the issue rather than the details: the point is just that
continuity of motion is important for assigning and maintaining object indexes.]
Suppose object indexes are being used in tracking four or more objects simultaneously and one of these objects—call it the \emph{first object}—disappears behind a barrier.
Later two objects appear from behind the barrier, one on the far side of the barrier (call this the \emph{far object}) and one close to the point where the object disappeared (call this the \emph{near object}).
If the system of object indexes relies on assumptions about speed and direction of movement, then the first object and the far object should be assigned the same object index.
But this is not what typically happens.
Instead it is likely that the first object and the near object are assigned the same object index.%
\footnote{
See \citet{franconeri:2012_simple}. Note that this corrects an earlier argument for a contrary view \citep{scholl:1999_tracking}.
}
If this were what always happened, then we could not fully explain how infants represent objects as persisting by appeal to object indexes because, at least in some cases, infants do use assumptions about speed and direction in interpolating the locations of briefly unperceived objects.
There would be a discrepancy between the Principles of Object Perception which characterise how infants represent objects as persisting and the principles that describe how object indexes work.
But this is not the whole story about object indexes.
It turns out that object indexes behave differently when just one object is being tracked and the object-specific preview benefit is used to detect them.
In this case it seems that assumptions about continuity and constancy in speed and direction do play a role in determining whether an object at $t_1$ and an object at $t_2$ are assigned the same object indexes \citep{flombaum:2006_temporal,mitroff:2007_space}.
In the terms introduced in the previous paragraph, in this case where just one object is being tracked, the first object and the far object are assigned the same object index.
This suggests that the principles which govern object indexes may match the principles which characterise how infants represent objects as persisting.
Object Indexes and the Launching Effect
\section{Object Indexes and the Launching Effect}
\section{Object Indexes and the Launching Effect}
A wild conjecture: Causal interactions are detected, or otherwise treated specially, by
perceptual processes
involved in segmenting and tracking objects.
How to get beyond intuition?
The launching effect: detecting a 50ms difference in the delay between two movements.
How is launching detected? For example, does it involve perceptual processes?
Why is a delay of up to around 70ms consistent with the launching effect occuring?
Ok, but what is the nature of these perceptual processes?
Object indexes ...
object indexes require tracking some causal interactions ...
But before we get to that, what is this about causal interactions?
It's reasonably obvious that object indexes require segmentation (otherwise how would they
attach to *objects*), and the evidence that they can survive occlusion is relatively easy to
understand. But the point that having object indexes involves tracking causal interactions
is less straightforward. So let's consider that ...
Object indexes are linked to causation.
In order to track objects, a perceptual system has to be sensitive to be causal interactions
Why is this true?
Because when you have a causal interaction, there's a conflict between principles of object
perception e.g. distinct surfaces=>two objects, vs good continuity of motion=>one object
The perceptual system needs to know when conflicts should be reconciled and when they should
be written off.
We get perceptual effects of causal interactions when there are conflicts among cues of object
identity.
This is a point Michotte made. He found that launching occurs when there is a conflict between
cues to object identity: good continuity of movement suggests a single object whereas the
existence of two distinct surfaces indicates two objects.
It is plausible that other types of causal interaction also involve conflicts between cues
to object identity.
preliminary evidence
Kruschke and Fragassi, 1996 figure 2 (part)
Krushke and Fragassi (1996) have shown that the object-specific preview effect vanishes
in launching but not in various spatio-temporally similar sequences. Since the object-specific
preview effect is regarded as diagnostic of feature binding, this is evidence that in launching
sequences, features of the second object (such as motion) remain bound to the first object for
a short time after the second object starts to move.
Kruschke and Fragassi, 1996 figure 2
Kruschke and Fragassi, 1996 figure 3
‘Unfortunately, this preliminary study has not received any published follow-up work, to our
knowledge, so it is not clear if it would also generalize beyond highly specific contrasts.’
\citep{choi:2006_measuring} on Kruschke and Fragassi:
‘one previous study suggested that causal events gave rise to diVerent
patterns of object-specific priming than non-causal events (Kruschke &
Fragassi, 1996). Unfortunately, this preliminary study has not received any
published follow-up work, to our knowledge, so it is not clear if it would
also generalize beyond highly specific contrasts’ \citep[p.~108]{choi:2006_measuring}.
Choi & Scholl, 2006 p. 108
This finding fits (by design) almost perfectly with Michotte’s phenomenological analysis ...
‘the movement performed by object B appears simultaneously under two
different guises: (i) as a movement (belonging to object A), (ii) as a
change in relative position (by object B)’
\citep[p.~136]{Michotte:1946nz}
Michotte 1946 [1963], p. 136
‘the physical movement of the object struck gives rise to a double
representation. This movement appears at one and the same time (a) as a
continuation of the previous movement of the motor object, and (b) as a
change of relative position (a purely spatial withdrawal) of the projectile
in relation to the motor object.’
\citep[p.~140]{Michotte:1946nz}
Michotte 1946 [1963], p. 140
\emph{Causal Object Index Conjecture}:
Causal Object Index Conjecture:
Effects associated with the ‘perception of causation’
are consequences of errors (or error-like patterns) in the assignments of object indexes and
their phenomenal effects.
Predictions:
i. Where there is perception of causation, there will be errors (or error-like patterns)
in the assignments of object indexes.
ii. Factors that can influence how object indexes are assigned or maintained can influence
perception of causation.
There is already some evidence which hints at this ...
The conditions that affect launching are consistent with it being a consequence
of how object indexes are updated and maintained ...
‘Michotte and his followers worked out many of the factors which mediate
the perception of causality, such as the role of absolute and relative
speeds, spatial and temporal gaps in the objects' trajectories, differences
in the durations and angles of each object's trajectory, etc ...
‘This research has generally shown that many different spatiotemporal parameters are critical for perceiving causality, but that featural parameters (eg colors, shapes, sizes) play little or no role.’
\citep[p.~456]{Scholl:2004dx}
Scholl and Nakayama 2004, p. 456
Assignments of object indexes are not generally sensitive to featural information
(*check: perceiving expressions of emotion paper / dev.mind).
Scholl 2007, figure 4
Causal Object Index Conjecture:
Effects associated with the ‘perception of causation’
are consequences of errors (or error-like patterns) in the assignments of object indexes and
their phenomenal effects.
Predictions:
i. Where there is perception of causation, there will be errors (or error-like patterns)
in the assignments of object indexes.
ii. Factors that can influence how object indexes are assigned or maintained can influence
perception of causation.
There is even more evidence which hints at this ...
If you've read Rips (2011) you’ll know about
dissociations between the launching effect and causal reasoning.
dissociations between the launching effect and causal reasoning (Rips 2011)
Schlottmann and Shanks (1992, Experiment 2) ... The point of this
experiment was to distinguish between cues to causality provided by
perceptual features within a display (launching vs. delay) and by
statistical contingencies (between Object A moving and Object B moving) as
observed across trials. If Michotte’s hypothesis is correct, statistical
contingencies should not affect perceptual judgments of whether a
launching has occurred. Conversely, perceptual features should not affect
judgments of whether A’s movement is statistically necessary for B’s
movement. ... The results of the study were clear cut ...: Launching versus delay
affected ratings of how convincing the collisions looked but not ratings of
necessity. However, the presence of a predictive statistical relation
affected ratings of necessity but not ratings of convincingness. These
results are consistent with Michotte’s hypothesis: Statistical
contingencies over trials do not affect the cause detector on the right of
Figure 3A, which is sensitive only to purely spatio-temporal properties of
the individual displays.
\citep[p.~87]{rips:2011_causation}
Schlottman & Shanks (1992, p. 340):
‘Despite the fact that a colour change of Object B more reliably predicted B’s movement than did the
impact of Object A, where was no impact of the predictive relationship on subjects’ ratings of
perceived causality. In other experiments asking subjects to make causal judgements, however, such
alternative predictive signals are known to reduce subjects’ ratings dramatically (Shanks, 1986).’
‘The clearest dissociation among the neuropsychology studies comes from
Roser, Fugelsang, Dunbar, Corballis, and Gazzaniga (2005), who tested two
split-brain patients. ... The patients were able to discriminate the
launching sequence from the others only if their right hemisphere processed
the event. By contrast, the same patients were able to solve a causal
reasoning problem—figuring out which of two switches controlled a
light—only if the problem appeared to their left hemisphere.’
\citep[p.~88]{rips:2011_causation}
Causal Object Index Conjecture:
Effects associated with the ‘perception of causation’
are consequences of errors (or error-like patterns) in the assignments of object indexes and
their phenomenal effects.
Predictions:
i. Where there is perception of causation, there will be errors (or error-like patterns)
in the assignments of object indexes.
ii. Factors that can influence how object indexes are assigned or maintained can influence
perception of causation.
Causal Object Index Conjecture:
Effects associated with the ‘perception of causation’
are consequences of errors (or error-like patterns) in the assignments of object indexes and
their phenomenal effects.
Potential objections:
i. Adaptation effects [see appendix]
ii. Rips-Leslie argument from the Pulfrich double-pendulum illusion [see appendix]
How to get beyond intuition?
The launching effect: detecting a 50ms difference in the delay between two movements.
How is launching detected? For example, does it involve perceptual processes?
Why is a delay of up to around 70ms consistent with the launching effect occuring?
Very good, but what does this tell us about the question with which I started?
I.e. can humans perceive causal interactions?
The Launching Effect and Metacognition
\section{The Launching Effect and Metacognition}
\section{The Launching Effect and Metacognition}
If the launching effect is a consequence of the operation
of a system of object indexes, why does it have phenomenal consequences?
One possibility is that conflicts in assigning object indexes give rise to metacognitive feelings
of surprise
which subjects have learnt to interpret as impressions of causation.
Back to the question I started with ...
Consider causal interactions.
Where do they fit in?
Are they perceptually
experienced in whatever sense the shape and motions of things are?
perceptual process vs perceptual experience
What if anything do the findings tell us about phenomenology?
Causal Object Index Conjecture:
Effects associated with the ‘perception of causation’are consequences of errors (or error-like patterns) in the assignments of object indexes andtheir phenomenal effects.
What, if any, experiential effects do these errors have?
feeling of surprise
There is a feeling of surprise which has features characteristic metacognitive feelings.
‘the intensity of felt surprise is [...] influenced by [...] the degree of the event’s interference with ongoing mental activity’
Reisenzein et al, 2000 p. 271; cf. Touroutoglou & Efklides, 2010
In particular,
‘the intensity of felt surprise is not only influenced by the unexpectedness of the
surprising event, but also by the degree of the event’s interference with ongoing mental
activity, [...] the effect of unexpectedness on surprise is [...] partly mediated by
mental interference’ \citep[p.~271]{reisenzein2000subjective}
That is, the feeling of surprise is a sensational consequence of mental
interference. (This can be tested by increasing cognitive load: this intensifies feelings
of surprise without, of course, making the events themselves more suprirsing. But see
\citep{reisenzein:2017_cognitiveevolutionary} for an alternative interpretation of such
findings.)
So whereas the feelings of agency and familiarity are both consequences of
unexpected fluency of processing, the feeling of surprise is supposed to be the
opposite: it is a consequence of unexpected interference in processes.
\footnote{%
An alterantive is proposed by \citet[p.~79]{foster:2015_whya}:
‘the MEB theory of surprise posits that: Experienced surprise is a metacognitive assessment
of the cognitive work carried out to explain an outcome. Very surprising events are those
that are difficult to explain, while less surprising events are those which are easier to
explain.’
\citet{foster:2015_whya} is about reactions to reading about something unexpected, whereas
\citet{reisenzein2000subjective} measures how people experience unexpected events (changes to
stimuli while solving a problem). The latter is much closer to what I’m after.
}
object index assignment error
↓
metacognitive feeling of surprise
+
learnt association
↓
judgement of causality
(And disposition to judge causes people to say they see causings.)
Recall this argument ...
Consider an encounter with three two-object movements where the delays between movements are 50, 100 and 150ms.
(1) The phemomenal difference between the first two encounters is larger than the phenomenal difference between the second two.
(2) This difference in differences is a fact in need of explanation.
(3) The fact cannot be explained by perceptual experience of objects or their motion.
(4) The best explanation for (1) is that we perceptually experience causal interactions.
The research on object experiences suggests this is false.
The causes are causal, but the experiences is of objects.
Or rather, of an interruption to the way objects are perceived.
The launching effect is all about how motion is processed perceptually,
so it is natural to suppose that the phenomenology reflects this.
Michotte went to great lengths to argue that the experience of one object
having another’s movement amounted to experiencing causation (‘ampliation of the
movement’), but this appears unjustified.
conclusion
In conclusion, ...
Can humans perceive causal interactions?
Causal Object Index Conjecture:
Effects associated with the ‘perception of causation’
are consequences of errors (or error-like patterns) in the assignments of object indexes and
their phenomenal effects.
object index assignment error
↓
metacognitive feeling of surprise
+
learnt association
↓
judgement of causality
Working hypothesis: Causal interactions are detected, or otherwise treated specially, by
perceptual processes
involved in segmenting and tracking objects.
The perceptual system responsible for identifying objects must also concern itself with certain
kinds of causal interaction in order to reconcile conflicting cues to object identity.
In slightly more detail: one function of our perceptual systems is to identify and track
objects; this is done by means of various cues; sometimes the visual system is faced with
conflicting cues to object identity which need to be resolved in order to arrive at a
satisfactory interpretation; when certain types of causal interaction occur there is a
conflict among cues to object identity; these conflicts must be treated differently from other
conflicts because they do not indicate failures of object identification and so do not
require resolution or further perceptual processing.
So object perception depends on sensitivity to certain types of causal interaction and this is
why the launching effect occurs.
This is unexpected insofar as perception is often supposed to be limited to features of the
world less abstract that causal interactions. Indeed, the notion that perceptual processes
represent three-dimensional objects rather than mere surfaces was at one time controversial.
The research we have reviewed shows that perceptual processes represent not only
three-dimensional but properly physical objects, that is, objects capable of causally
interacting with each other.
Conjecture: These detections trigger metacognitive feelings,
which subjects have learnt to interpret as impressions of causation.
Appendix: Adaptation and Launching
\section{Appendix: Adaptation and Launching}
\section{Appendix: Adaptation and Launching}
A challenge to the conjecture ...
A question was:
Can humans perceive causal interactions?
Now I think we have achieved an answer:
Working hypothesis: Causal interactions are detected, or otherwise treated specially, by perceptual processes involved in segmenting and tracking objects.
Rolfs et al, 2013 figure 1 (part)
Rolfs et al, 2013 figure 2B
Arnold et al 2015, figure 1
‘participants were asked to judge whether simulated bounces of a disc against a static wedge were
‘‘hard’’ (like a pool ball bouncing on concrete) or ‘‘soft’’ (like a squash ball bouncing).’‘ Adaptation suggesting repeated ‘‘rigid’’ collisions resulted in more ‘‘soft’’ bounces
being reported, whereas adaptation suggesting repeated elastic object collisions resulted in more ‘‘hard’’
bounce reports—an object elasticity aftereffect.’
adaptation
This seems hard to reconcile with object indexes.
But (A) note the questions in the commentary; and (B) Hubbard on Kiritani (1999)
for launching when the contact is occluded; and context effects (Scholl on postdiction)
and (D) evidence for effects of
acoustic stimuli on launching, suggesting that the effect may not be modality-specific
(The evidence on this point is controversial. Sekuler and colleagues show that when subjects observe
an ambiguous visual display consistent with either a collision or a passing event, the timing of a
tone can control whether subjects report seeing a collision or pass- ing, and argue that this is a
multisensory phenomenon: R. Sekuler et al., ‘Sound Alters Visual Motion Perception’, Nature,
(), p. . Watanabe and Shimojo extend this finding by showing that not any event (or
non-event) which draws attention at the moment of a collision will disambiguate the display; they
argue that the tone’s effect on the perception of a col- lision is a ‘genuine audiovisual effect,
not an audiovisual effect that results from auditory effects’: K. Watanabe and S. Shimojo, ‘When
Sound Affects Vision: Effects of Auditory Grouping on Visual Motion Perception’, Psychological
Science, (), pp. –. Guski and Troje, on the other hand, show that features which carry
no information about causation, such as a blink, can also influence whether subjects report seeing a
collision or a passing. These authors conclude that auditory influences on the perceptual of
collisions are ‘no true cross-modal phenomenon’: R. Guski and N.F. Troje, ‘Audiovisual Phenomenal
Causality’, Perception and Psychophysics, (), pp. –, at p. .)
Appendix: The Pulfrich Double Pendulum Illusion
\section{Appendix: The Pulfrich Double Pendulum Illusion}
\section{Appendix: The Pulfrich Double Pendulum Illusion}
Are object perception and causal perception are one and the same process?
Not if causal perception arises from the fact that solid objects cannot be
perceived to move through each other and Leslie’s informal report about
Wilson and Robinson’s (1986) version of
the Pulfrich double pendulum illusion is correct.
Wilson & Robinson 1986, figure 1
The Pulfrich double pendulum illusion \citep{wilson:1986_impossibly}.
figure caption: ‘The double Pulfrich pendulum: (a) construction of the
pendulums; (b) what observers do not see. The pendulums are set swinging in
opposite phase and observed with both eyes, one eye covered by a neutral
density filter ( ~ 2 log units). Observers see the pendulum bobs following
each other round in a horizontal ellipse. They do not see the arms twist
round each other as they should if the seen movements of the bobs were
veridical. Observers are disconcerted by this inconsistency.’
LESLIE: ‘What Wilson and Robinson do not describe, however, is what observers
see happening to the rods. They say that observers do not see them twisting
round each other, but they do not say what observers do see.’
‘I have therefore investigated this myself ... Equally striking is the clear
perception of the rigid solid rods passing through each other. Most observers
were able to find an angle of view where even the pendulum bottles appear to
pass through one another despite their large size and marked surface texture.’
\citep[pp.~198--9]{Leslie:1988ct}
Note that there is a dearth of information about whether assigning and
maintaining object indexes really does take into account solidity.
Mitroff, Scholl and Wynn 2005, figure 2
Compare Scholl and Mitroff’s bouncing/streaming study. This study:
(i) suggests solidity may be a constraint, at least sometimes;
(ii) indicates that phenomenology and object indexes may not perfectly align
Mitroff et al p. 74: ‘we induced a strong bias to consciously perceive
streaming, by using smooth, constant, and reasonably fast motion. This
allows us to provide the clearest possible situation in which to evaluate
the relationship between conscious percepts and object files.’
Note that:
The claim that some system of object indexes sometimes gives rise to
phenomenal expectations does not imply that object indexes and phenomenal
expectations are always aligned. Mitroff et al. (2005) construct a situation
involving two objects which simultaneously undergo temporary occlusion. In
this situation, perceivers’ verbal reports imply the objects’ paths crossed
whereas measuring an object-specific preview benefit implies that the objects
bounced off each other. The object indexes underpinning object-specific
preview benefits are unlikely to be informing phenomenal expectations about
objects’ movements in this situation. Object indexes and phenomenal
expectations can come apart in some situations.
Mitroff, Scholl and Wynn 2005, figure 3
Return to the claims and Rips’ objection ...
‘object perception reflects basic constraints on the motions of physical bodies …’
(Spelke 1990: 51)
\citep[p.\ 51]{Spelke:1990jn}
‘A single system of knowledge … appears to underlie object perception and physical reasoning’
\citep[p.\ 175]{Carey:1994bh}
(Carey and Spelke 1994: 175)
I think there's something here that should be uncontroversial, and
something that's more controversial.
Rips’ objection (2011, p. 92)
Leslie’s informal report on the Pulfrich double pendulum illusion (which
appears never to have been published) suffices to show that
it was too simple to say that
‘object perception and causal perception are one and the same process’ or to
talk about ‘object perception and physical reasoning’.
But it remains possible that the launching effects are consequences of the
ways that object indexes are assigned and maintained (although this is far
from the only conclusion compatible with the limited available evidence).
Rips: ‘This possibility, though, is not one that advocates of Michotte’s
hypothesis have taken up. The favored view is one in which separate
modules are responsible for descriptions of objects and of their
mechanical interactions, with central mechanisms then resolving potential
conflicts between them.
For example, Leslie (1988) argues that perceptual
modules do not by themselves settle inconsistencies between the spatial
positions of objects and their causal interactions. In certain
illusions, adults tend to perceive solid objects passing through each other
(Leslie cites the Pulfrich double-pendulum illusion; see Wilson &
Robinson, 1986). Thus, the module for object tracking doesn’t prohibit
interpretations of physical events that are causally impossible. Of
course, people realize that such events cannot really be taking place and
are surprised when they perceive them, but that is because they have
principles stored in long-term memory—for example, the principle that two
solid objects cannot be in the same place at the same time—that classify
these events as illusions or anomalies. Leslie’s argument suggests
that people individuate objects and calculate their causal relations by
means of separate mechanisms; thus, we can’t count on causal constraints
being part of the object-tracking module. If Leslie is right, there is
reason to question Butterfill’s (2009) conjecture that ‘object
perception and causal perception are one and the same process’’ (p.
421).’
