Abstract

Introduction

Which properties and happenings can humans see? It’s fairly uncontroversial that we can see the shapes of things and their movements. It’s also reasonably uncontroversial that we can’t see properties like market value or processes like radioactive decay. People may sometimes talk about seeing these things, but it seems plausible that what they strictly and literally see are only characteristic visual indicators of market value or radioactive decay.

What about simple causal interactions like launchings, burstings, blockings and supportings—can we see these things? Consider two balls colliding. What do we see? Do we see only one motion followed by another or do we see some kind of causal interaction?

David Hume appears to have held that we perceive only one motion followed by another:

“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.”¹

Albert Michotte, on the other hand, held that we can see certain types of causal interaction in the same sense that we can see shape or movement:

“There are some cases … in which //p. 271// 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.”²

Several philosophers and psychologists agree with Michotte that causal interactions are visible in the same sense that shape and motion are. Among the philosophers are David Armstrong³, and Peter Strawson who says:

“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).”⁴

The psychologists Brian Scholl and Patrice Tremoulet agree that we can see causal interactions:

“just as the visual system works to recover the physical structure of the world by inferring properties such as 3-D shape, so too does it work to recover the causal … structure of the world by inferring properties such as causality.”⁵

While agreeing that we can see some causal interactions, Michotte, Strawson and Scholl adopt different approaches to thinking about perception. Scholl is driven by David Marr’s question, “What kind of information is vision really delivering?”⁶ Strawson focuses on descriptions of perceptual content that describe how the world is presented to the perceiver. And Michotte is concerned with perception as a source of causal concepts.

The variety of approaches to perception can easily complicate the task of answering questions about what we can perceive. What could ensure that, in making claims about what can or can’t be seen, we are genuinely agreeing with or contradicting others who make similar sounding claims rather than merely using terms like ‘see’ in ways that fail to connect with their positions? Michotte, Strawson and Scholl all hold that we can see causal interactions in the same sense that we can see shape, whatever sense that is. The appeal to shape perception ensures genuine agreement despite any differences on the nature of perception.

In this paper I argue that we can see some causal interactions in the same sense that we can hear speech, whatever sense that is. I will also tentatively suggest that speech perception is importantly different from shape perception. To see the shapes of things is, in a wide range of naturally occurring conditions if not always, to be aware of those shapes and in a position to know things about them; but hearing speech and seeing causings are not so related to awareness of the things perceived. If so, Michotte, Strawson and Scholl are right that we can see some causal interactions but wrong that we can do so in the same sense that we can see shape.

Note that I will only consider visually presented stimuli. Some researchers hold that we perceive causal interactions through touch or experiences of agency. These views will not be considered here. Following common use, I adopt ‘causal perception’ as a label for whatever perceptual experiences are associated with visual stimuli involving or standing in for causal interactions. Understood in this way, our first question is whether the objects of causal perception are causal interactions or just movements.

I.

Although we have quite a lot of data on causal perception, the interpretation of this data is not straightforward.

Let’s take Albert Michotte’s experiments on launching as our starting point. The launching sequence is illustrated schematically below.⁷ When seeing this sequence, most subjects describe their experiences in causal terms, as the experience of a collision or launching. They also distinguish this experience from the experience of seeing one movement followed by another. (The latter experience is reported when there is a delay between the two movements). Michotte studied launching in great detail and we will discuss his findings below.

	Michotte’s launching display8

Peter White and Elizabeth Milne showed subjects various other kinds of schematic animation (see figure below). They found that subjects reliably reported seeing pulling, disintegration and bursting.⁹

	Pulling, disintegration and bursting stimuli used by White and Milne10

There’s no doubt, then, that people have distinctive experiences for a range of stimuli which mimic different types of causal interaction, and that people regularly describe these experiences in causal terms.

Elizabeth Anscombe,¹¹ Curt Ducasse¹² and Peter Strawson¹³ have all observed that there are also many everyday situations in which its is natural to describe what we see in causal terms. These include seeing a bird bending a branch by alighting on it and a boulder crushing a hut as it rolls down a mountainside. Reading these authors, it may seem tempting to conclude from what people unreflectively say about their experiences that we can perceive causal interactions.¹⁴

I do not think this conclusion follows. Consider what people say they see when shown an animation involving simple geometric shapes of the sort used by Heider and Simmel and illustrated below.¹⁵ Many people report seeing things such as the large triangle intimidating the small one, or the small triangle defending the little circle as the big triangle tries to attack it. This suggests that there may be nothing that people won’t talk about seeing given the right conditions. So if there is a substantial question about whether we can see some types of causal interaction, it can’t be correct to answer this question on the basis of what people say they see.

	A frame from Heider and Simmel’s moving picture-film16

I am not suggesting that people who talk about seeing intimidating or defending are mistaken or confused, only that our commonsense talk about perception serves a range of practical interests which probably do not include, and are in any case not limited to, answering theoretical questions about the nature of vision. We shouldn’t take what people say about their perceptions at face value any more than we should take what they say about the meanings of their utterances at face value.

In cases like intimidating or defending it seems plausible to say that we see at most visible indicators of these things, not the interactions themselves. While there may be special phenomenological effects characteristic of intimidating or defending, it doesn’t follow that we can see these things happen (and of course Heider and Simmel do not claim that we can). This is not to say that future research couldn’t show that subjects really do see intimidating, only that nothing so far considered supports this conclusion. Now those who follow Hume in denying that we can perceive causal interactions can say the same thing about findings like Michotte’s and White and Milne’s. They can admit that various stimuli give rise to characteristic phenomenology which many people unreflectively describe in causal terms but deny that we experience causal interactions.

Some philosophers seem to have missed this point. In Chapter 4 of Intentionality John Searle objects to an analysis of causation in terms of regularities on the grounds that it “flies in the face of our common-sense conviction that we do perceive causal relations all the time.” Searle may be wrong to suppose that there is any such common-sense conviction; for reasons just given he’s certainly wrong to suppose that such a conviction would support a claim about what we perceive. Searle then notes that “[t]he experience of perceiving one event following another is really quite different from the experience of perceiving the second event as caused by the first.”¹⁷ All by itself this observation is unhelpful because something similar is true of experiences of perceiving intimidating or defending. Searle’s final point is that “the researches of Michotte and Piaget would seem to support our common-sense view.”¹⁸ As we’ll see by taking a closer look at Michotte, this is not exactly straightforward. Michotte himself acknowledges this. At one point he imagines Hume learning about his experiments and writes:

“it is probable that his philosophical position would not have been affected in the least.”¹⁹

In short neither Michotte’s nor more recent research (more of which is considered below) straightforwardly shows that we can see causings. In order to defend the claim that we can see causal interactions, we need a way of getting beyond what some people unreflectively say about their experiences.

II.

How can we get past disagreement over merely intuitive descriptions of visual experiences? By finding something that needs explaining.

Susanna Siegel suggests considering carefully matched pairs of experiences. For example, she imagines two ways of experiencing a situation in which a ball lands in a plant pot just before the lights go out.²⁰ It might seem to you compelling to suppose that the ball’s landing causes the lights to go out (even though you would presumably not judge this to be so), or you might experience the two events as unconnected. Siegel claims that these are two distinct experiences that the same stimuli can give rise to. Her question is then what explains the difference between the two experiences. She argues that the difference is best explained by supposing that the two events are experienced as causally connected in one case but not the other.

Siegel’s ‘matched pairs’ approach looks promising because it identifies something that needs explaining, a difference between two experiences. But can we apply Siegel’s approach to Michotte’s launching stimulus? There would have to be a pair of occasions on which a subject saw the same type of launching sequence, experiencing successive movements on one occasion and experiencing launching on the other. Unfortunately, it seems there are no such pairs because, in the case of launching and similar stimuli, the characteristic phenomenology described as causal appears to be mandatory if you’re paying attention. So we cannot apply Siegel’s approach here.

This problem relates to a defect of Siegel’s approach: it does not distinguish perceptual phenomenology from the non-perceptual phenomenology associated with the feeling of being struck by an idea. When a ball drops just as a light goes out, are we struck by the thought that one caused the other or do we somehow perceive this? Siegel’s approach does not distinguish these two possibilities, which undermines her claim to have shown that “causation is represented in visual experience.”²¹ When Michotte discusses the sort of cases Siegel focuses on, he denies that causation is perceived but notes that “a causal interpretation is urgently called for.”²² This seems at least as plausible as Siegel’s claim that we perceive a causal interaction when a ball lands in a plant pot just before the lights go out. Indeed, it is tempting to think that genuinely perceptual phenomenology is always mandatory for subjects attending to the relevant aspects of stimuli. If so, the only phenomenological effects Siegel’s ‘matched pairs’ approach can discern are non-perceptual because the required pairs of experiences do not exist when the phenomenology is mandatory.

Our aim is to support the claim that some causal interactions can be seen by finding something that needs explaining.

Brian Scholl and Ken Nakayama offer a promising candidate that needs explaining: illusory causal crescents. Consider a display just like Michotte’s launching stimulus except that the two objects completely overlap before the second starts to move (see left half of the figure below).²³ In this case almost all subjects report seeing a single object moving in front of or behind a stationary object; some report that the moving object changes colour.²⁴ However, it is possible to enhance this display to induce subjects to report seeing the overlap event as a launching event rather than as the movement of a single object (see right half of the figure below).²⁵ And here’s the fact to be explained: when subjects report seeing the overlap event as launching, they also report seeing an illusory crescent as if the shapes did not completely overlap and they systematically underestimate the degree of overlap between the two objects.²⁶

	Standard overlap event (left) and overlap event with contextual launching event (right).27

Does explaining this fact require appeal to the claim that we can see some types of causal interaction? Let’s allow, as Scholl and Nakayama claim, that subjects are seeing illusory causal crescents and that the experience characteristic of launching affects the perception of shape.²⁸ Given this, Scholl and Nakayama offer ‘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.”²⁹

If this is the best explanation of the illusory causal crescents, then we can indeed see at least one type of causal interaction.

This argument appears to be a promising way of attacking the idea that the experience characteristic of launching involves only some special phenomenological effect indicative of collisions. After all, why would such a phenomenological effect be associated with the illusory perception of crescents? Scholl and Nakayama’s position is well supported insofar as explaining the association between seeing launchings and seeing crescents requires appeal to some state with causal content.

But must that state be perceptual? On the face of it, Scholl and Nakayama’s finding is equally well explained by either of two hypotheses: (a) there is a visual perception of a collision; or (b) there is some non-doxastic but also non-perceptual cognition of a collision. Since there may well be ways to exclude (b), I do not claim that Scholl and Nakayama’s argument is incorrect. But I cannot make use of this argument here without ruling out (b) and that would seem to require understanding how perception is distinct from and relates to other forms of cognition. Also, Scholl and Nakayama’s strategy may not generalise to cases other than the launching stimuli. We would ideally like to be able to say more generally which types of causal interactions can be perceived rather than just that at least one type is perceived (for examples see White and Milne’s stimuli above). For these reasons I want to introduce and then pursue a third and final approach to showing that we can see some types of causal interaction. (The illusory causal crescents finding will be useful later.)

Michotte’s work provides the most promising candidate for a fact in need of explanation. The first phase of Michotte’s work involves discovering the precise conditions under which the launching effect occurs, such as the relative speeds of the two objects, the delay between the first and second objects’ movements, and the trajectories of the two objects. For instance, consider what happens if there is a pause between the first object’s movement and the second object’s movement. How long can the pause be before the launching effect vanishes? Michotte’s findings are presented in this table:

Illustration of Michotte’s study of the conditions under which experiences characteristic of launching occur30

Contemporary philosophers and psychologists sometimes write as if all Michotte did was study, in minute detail, the conditions under which the launching effect occurs. But this is to ignore the crucial second phase of Michotte’s work. He wasn’t directly concerned to know that pauses longer than 70ms dramatically reduce the impression of launching. His question was why the launching impression occurs under precisely these conditions and not others. He writes that after identifying conditions under which the launching effect occurs:

“we were confronted with a second task, that of ‘understanding’ the phenomenon, of … seeking to find out why such and such conditions were necessary for its production, and why it possessed such and such properties.”³¹

Michotte’s idea, then, is that we can get past superficial ideas about the perceptual experiences characteristic of launching by first (a) identifying the precise conditions under which these experiences occur and then (b) explaining why they occur under these conditions. If it turns out that the best explanation at stage (b) requires the claim that we see causal interactions, then we will have an argument for that claim.

The key fact to be explained, then, is that the experience characteristic of launching occurs under certain conditions. Having come this far with Michotte, I now want to deviate from his views. To develop the argument that we can see causal interactions, we need to compare causal with speech perception.

III.

One of the questions set for this volume is: What methods can we use to determine the admissible contents of experience? One way to answer this question is to study cases where researchers have been successful, and speech is one of those cases.

According to Alvin Liberman and Ignatius Mattingly’s “motor theory” of speech perception, when we encounter speech we perceive not sounds but gestures. More exactly, we perceive “the intended³² phonic gestures of the speaker.”

In the case of causation our question is, Do we perceive merely a sequence of movements or do we perceive causal interactions? In the case of speech, a parallel question is, Do we perceive sounds or intended phonic gestures?

Naturally not everyone accepts the motor theory of speech perception.³³ For our purposes, it is most important that Liberman and Mattingly have a valid argument for the conclusion that the objects of speech perception are intended gestures; whether the premises of this argument are true is less important. After all, even if their claims about speech were false, parallel claims might be true of causal perception.

Liberman and Mattingly’s argument starts from the premise that speech perception is categorical. To illustrate, consider the following representation of twelve sounds:

ba-da-ga from http://www.columbia.edu/itc/psychology/rmk/T2/T2.2b.html

Each sound differs from its neighbours by the same amount as any other sound, at least when difference is measured by frequency. Most people would not be able to discriminate two adjacent sounds, except for two special cases (one around -3 to -1 and one around +1 to +3) where the discrimination is easier; here people hear the sound change from da to ga or from ga to ba. This pattern of heightened discrimination is the defining characteristic of categorical perception.³⁴ Small changes to stimuli can make large differences to perception, large changes to stimuli can make small differences to perception, and the stimuli can be ordered and sorted into categories such that discriminating nearby pairs of stimuli on either side of a category boundary is dramatically easier than discriminating pairs from within a category.

The existence of these category boundaries is specific to speech perception as opposed to auditory perception generally. When special tricks are used to make subjects perceive a stimulus first as speech and then as non-speech, the locations of boundaries differ between the two types of perception.³⁵

Speech perception also exhibits constancy (otherwise called invariance): the location of the category boundaries changes depending on contextual factors such as the speaker’s dialect³⁶ or the rate at which the speaker talks³⁷; both factors dramatically affect which sounds are produced. This means that between two different contexts, different stimuli may result in the same perceptions and the same stimulus may result in different perceptions.³⁸

Later I will defend a distinction between categorical and other forms of perception. It’s essential for drawing this distinction that I have characterised categorical perception in terms of abilities to distinguish stimuli and not in terms of how stimuli appear to perceivers. For example, I have not talked about pairs of colours or sounds being perceived as similar. In my view, which I take to be the standard view,³⁹ categorical perception is not a matter of how things appear at all. ⁴⁰

Given that speech perception is categorical and exhibits constancy, we can ask where the category boundaries fall. Which features of the stimuli best predict category membership? Liberman and Mattingly show that, in the case of speech, category boundaries typically correspond to differences between intended phonic gestures. The existence of category boundaries and their correspondence to intended phonic gestures needs explaining.

Following Liberman and Mattingly, we can explain this by postulating a module for speech perception. Anything which is potentially speech (including both auditory and visual stimuli) is passed to the module which attempts to interpret it as speech. It does this by attempting to replicate stimuli by issuing the same gestures that are also used for producing speech (this is the ‘motor’ in ‘motor theory’). Where a replication is possible, the stimuli are perceived as speech, further auditory or visual processing is partially suppressed, and the module identifies the stimuli as composed of the gestures that were used in the successful replication. Accordingly we can say that the stimuli are perceived as a sequence of phonic gestures.

One line of response to this argument involves attempting to show that the category boundaries correspond to some acoustic property of speech at least as well as to intended phonic gestures.⁴¹ If such a correspondence were found, it might be possible to give a explanation better than Liberman and Mattingly’s for the existence of categories corresponding to intended phonic gestures. Reasons for doubting any such explanation exists include the constancy effects already mentioned and also coarticulation, the fact that phonic gestures overlap (this is what makes talking fast).

In outline Liberman and Mattingly’s argument for the claim that the objects of speech perception are intended phonic gestures has this form:

(1) There are category boundaries … .

(2) … which correspond to intended phonic gestures

(3) Facts (1) and (2) stand in need of explanation.

(4) The best explanation of (1) and (2) involves the claim that the objects of speech perception are intended phonic gestures.

This illustrates how we might establish claims about the objects of perception without relying too directly on what people say they perceive and without presupposing very general theoretical claims about what it means for something to be perceived.

I am not suggesting we can take the case of speech perception as a model for understanding perception generally. In my view speech perception, like other forms of categorical perception, is importantly different from ordinary, non-categorical perception. By contrast, some philosophers have developed general accounts of perception which appear to be supported primarily by cases of categorical perception such as speech and colour. For instance, both Fred Dretske and Jerry Fodor have taken speech perception as a paradigm case for very general theories about what it is for something to be an object of perception.⁴² Mohan Matthen has recently gone even further by supposing that all perception is a matter of assigning stimuli to categories.⁴³ For reasons tentatively outlined later, these authors may be mistaken to take categorical forms of perception as paradigm cases of perception generally. The suggestion I develop next is only that the categorical perception of speech is a reasonable model for causal perception of visual stimuli.

IV.

Is speech perception a good model for causal perception? I want to suggest that it is, and that we can use an argument like (1)–(4) above to show that we perceive causal interactions.

The first step is to show that causal perception exhibits the key features of categorical perception. As already mentioned, whether or not the experience characteristic of launching occurs depends on interactions among various factors including the relative speeds of the two objects, the delay between the first and second objects’ movements, and the trajectories of the two objects. Importantly, the difference between an experience characteristic of launching and an experience of two distinct motions often depends on small changes to parameters—changes which are smaller than can normally be discerned in perception. For instance, consider trials in which there is a short pause between the movements of two objects, A and B. When two trials differ only in that the gap is 28ms longer on one of the trials, most observers probably could not tell the trials apart, except in the special case where 28ms makes the difference between the experience characteristic of launching and the experience characteristic of two movements (see figure below). As in the case of speech, then, there are points around which exceptionally fine-grained discriminations are possible. This is the key indicator that causal perception is categorical.

	Effect of the interval between movements on whether subjects report experiencing direct launching or two distinct movements.44

As well as such things as relative trajectories and speeds, the experience of launching also depends on high-level features of stimuli. For example, Michotte found that if the second ball in a launching sequence is initially seen moving backwards and forwards on its own account, then the experience distinctive of launching does not occur.⁴⁵ So the experience characteristic of launching depends not only on simple spatial and temporal parameters but also on more complicated structural features of the stimuli.⁴⁶ Scholl and colleagues have shown, further, that in some situations the experience characteristic of launching depends how events unfold shortly after the moment of collision.⁴⁷ This suggests that the relation between experiences and visual stimuli is complex in the same way as the relation between speech experiences and speech stimuli: in both cases there is no simple mapping from features of the stimuli to features of the experience unless one characterises the stimuli in terms of gestures or causes.

In short, the experience characteristic of launching involves both the heightened discriminability and the sensitivity to complex features of stimuli that are characteristic of categorical perception. Research on other causal stimuli such as pulling and bursting is less well developed but also provides some evidence that causal perception may be a form of categorical perception.⁴⁸

To say that perceptual experiences characteristic of launching and other causal stimuli are categorical is not yet to say where the category boundaries fall. The experimental findings appealed to so far rely on what subjects say about their experiences only as a way of distinguishing whether on two occasions they have the same or different types of experience. At this point, what the experiences are experiences of is an open question.

We can take a first step to answering this question by finding things the categories correspond to—that is, by inquiring which features of stimuli reliably predict whether they will be assigned to a given perceptual category. Of course this is only a first step. Facts about what perceptual categories correspond to do not straightforwardly determine what those categories are categories of.

Michotte argued that the conditions under which the experience characteristic of launching occurs correspond to physical laws governing collisions.⁴⁹ In fact, Michotte thought the correspondence so close that at one point he even suggests it would be possible to learn about the physics of collisions by studying their perception:

“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.”⁵⁰

Opposing Michotte, Peter White and Elizabeth Milne suggest that the conditions under which the perceptions occur do not correspond to conditions required for actual pulling or bursting. Rather, they suggest that they may correspond to laws implicit in our naive conceptions of causal interactions.⁵¹

Evaluating these hypotheses requires painstaking empirical research on category boundaries (just as in the case of speech). To illustrate, consider Michotte’s finding that the experience of launching occurs when there is a pause between two objects’ movements of around 80ms or less. How is this consistent with the laws of mechanics—surely no pause can be tolerated? 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.⁵² 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.

This is the kind of detailed investigation needed to support (or refute) claims that conditions for experiences characteristic of launching correspond to conditions for a certain kind of collision. Such research may eventually enable us to make fine-grained distinctions between competing hypotheses about which causal interactions are perceived and how they are perceived. As things stand it is already reasonable to conclude that there is some correspondence between the conditions under which experiences characteristic of various types of causal interaction occur and conditions under which the causal interactions themselves occur.

This needs qualification because few of the stimuli used in experiments on causal perception involve real causal interactions. Strictly speaking we need to hedge the claim that conditions for experiences characteristic of launching correspond to conditions under which collisions occur by restricting it to the perceiver’s natural environment and excluding factors to which the relevant perceptual systems are insensitive.⁵³ In case this hedge seems like a fudge, note that exactly the same considerations apply in the case of speech. Earlier I mentioned that category boundaries in speech perception correspond to differences in intended phonic gesture. But as speech perception is reliably fooled into treating all kinds of synthetic sounds as speech, this is not strictly true. As in the causal case, it is at best true taking into account the limits of perceptual systems, their natural environments and the types of information they are concerned with processing.

If a simple covariational or causal theory of perceptual content were correct, the need to hedge claims about correspondence would show that we do not perceive causal interactions (and that we do not perceive intended phonic gestures). I am assuming that such theories are false and appealing to correspondence not as a necessary or sufficient condition on perception but only as something to be explained. Given this, hedging would only be objectionable if done to such an extent that the correspondence in question were no longer a fact that needs explaining. After all, even imperfect correspondences need explanation (for example, a correspondence between weather forecasts and actual weather conditions).

So far, then, we have covered the first three steps of an argument parallel to the one about speech:

(1') There are category boundaries …

(2') … which correspond to certain types of causal interaction.

(3') Facts (1') & (2') need explanation.

This brings us to the fourth and final step of the argument:

(4') The best explanation of (1') & (2') involves the claim that the objects of perception are causal interactions.

In defence of (4'), I conjecture that the perceptual experiences characteristic of causal interactions are best explained as by-products of perceptual mechanisms whose function is to track three-dimensional objects and their movements. Causal perception is part of object perception.

At least three pieces of evidence suggest that perception of objects is closely linked to causal perception. First, Michotte argued for such a link on the basis of his finding 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.⁵⁴ It is plausible that other types of causal interaction also involve conflicts between cues to object identity. Second, as we saw above Scholl and Nakayama have shown that when a sequence involving complete overlap between two objects is perceived as a launching event, subjects perceive an illusory crescent as if the overlap were only partial.⁵⁵ They conclude that “the perception of causality can also affect other types of visual processing—in this case the perception of spatial relations among moving objects.”⁵⁶ Third, Krushke and Fragassi have shown that the object-specific preview effect vanishes in launching but not in various spatio-temporally similar sequences.⁵⁷ Since the object-specific preview effect is 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.⁵⁸

These three pieces of evidence indicate that causal perception interacts with object perception. It is not that we first see sequences of events involving objects and then assign them causal significance. Rather, the perception of something as a causal interaction is bound up with what we perceive as objects in the first place.

This suggests a simple explanation for the correspondence between the perceptions characteristic of various kinds of causal interaction and conditions under which those causal interactions occur. The correspondence exists because, given contingent facts about how it actually works, 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 perception and this is why we enjoy categorical perception of those types of causal interaction.

If this explanation is correct, we can conclude that causal interactions are objects of perception in the same sense that intended phonic gestures are. But this explanation is not as satisfying as the one offered in the case of speech—that one explained how intended phonic gestures could be perceived whereas this one only says that causal interactions need to be perceived given how object perception works.

Where can we find a deeper explanation? If, as I have been suggesting, object perception and causal perception are one and the same process, a theory about how object perception works might also explain how causal interactions are perceived. Consider Elizabeth Spelke’s theory: object perception is best understood as an inference where the premises are descriptions of stimuli in terms of surfaces and their layout; conclusions about objects are inferred from these premises by means of a set of principles describing how objects behave. Here, for example, are two of Spelke’s four principles of object perception:

boundedness—“two surface points lie on distinct objects only if no path of connected surface points links them … the boundedness principle implies that … two distinct objects cannot interpenetrate.”

no action at a distance—“separated objects are interpreted as moving independently of one another if such an interpretation exists.”⁵⁹

These principles bear not only on object identity but also on causal interactions between objects. In the case of launching, the no action at a distance principle is clearly relevant; for example, it would explain why there is no launching effect without spatial contact. The boundedness principle is also relevant, as it explains why the experience characteristic of launching is typically absent when there is complete overlap between two objects. In general, Spelke holds that “object perception reflects basic constraints on the motions of physical bodies.”⁶⁰ Developing this idea, Renee Baillargeon, Su-hua Wang and their colleagues have argued that a single set of principles are responsible for object segmentation and for the perceptual identification of certain events as physically possible or impossible.⁶¹ If this is right, the principles which explain object perception also explain how causal perception works.

This completes my argument for the claim that we can perceive certain types of causal interaction. This argument was outlined in (1')–(4') above. Clearly the argument is far from decisive since it depends on a complex mix of theoretical and empirical issues. Even so, the argument demonstrates how to respond to Hume’s claim that we do not perceive causal interactions. Given the evidence we have so far, it is more plausible that we perceive causal interactions than that we do not. Equally importantly, the comparison with speech perception provides us with some insight into what it means to perceive causal interactions: they are perceived categorically and they are objects of perception in the sense that intended phonic gestures are.

Note that strictly speaking this argument establishes only that we perceive causal interactions when presented with visual stimuli, not that we see them. Causal perception may be amodal, as speech perception is sometimes held to be.⁶² Even so, I’ll continue to write about seeing causes and hearing gestures for simplicity. Nothing said here contradicts David Marr’s view that “the quintessential fact of human vision—[is] that it tells about shape and space and spatial arrangement.”⁶³

IV.

So far I have argued that we can see some causal interactions, and that seeing causal interactions is part of seeing objects (seeing objects move is interdependent with, not prior to, seeing causal interactions). The argument followed a well-understood model from the case of speech perception. First we found a feature of causal perception in need of explanation, the existence of categories and their correspondence to situations in which causal interactions occur. Then I argued that the best explanation for this feature involves the hypothesis that we perceive causal interactions: the feature occurs because, given how they actually work, perceptual systems responsible for identifying and tracking objects must also discriminate certain types of causal interaction.

As mentioned at the start, many of those who agree that we can perceive causings also claim that we perceive causings in the same sense that we perceive shape, whatever sense that it. Full evaluation of this further claim would involve considering issues about perception beyond the scope of this paper. On the other hand, some evaluation of this claim is necessary for understanding what it means to perceive causings. In a preliminary and tentatively way, I suggest that the further claim is false and conclude with some considerations distinguishing causal and speech perception from shape perception.

Here are two claims about shape perception. First, in a wide range of naturally occurring circumstances, perceiving an object’s shape is a way of being aware of its shape, that is, of being in a position to think about and act on its shape.⁶⁴ Second, to describe what someone perceives as including a particular shape is to describe a characteristic of their experience which is, often enough, introspectable.⁶⁵ I assume that these claims are true and constitutive of shape perception and argue that parallel claims about causal or speech perception are false. If these claims are indeed constitutive of shape perception but false of causal perception, then we do not perceive causal interactions in the same sense that we perceive shape.

Consider the relation between facts about the objects of speech perception and its introspectable character. Not even the best trained, most conceptually sophisticated experts can use introspection to resolve conflicts about whether the objects of speech perception are sounds or gestures. The issues that determine the conflict concern the nature of the stimuli (what invariants are there?) and processes (does passive speech perception involve the motor cortex?). I suspect introspection tells us nothing about what we perceive when we perceive speech; certainly it has no bearing on conflicts about the objects of speech perception. So speech and shape perception differ with respect to introspection.

The parallel point about causal perception is less clear because some philosophers and psychologists seem to have held that introspection can reveal what we perceive when we perceive causal interactions. While in principle there might also be perceptual experiences facts about the objects of which are knowable through introspection, the type of perceptions Michotte studied and that have been the focus of this paper are not such perceptions. As in the case of speech perception, discerning the objects of causal perception depends on careful investigation of the conditions under which they occur and the reasons why they occur under these conditions. Introspection is inessential.

What about the relation between perception and awareness? In the case of speech researchers distinguish categorical perception of phonemes from ‘phonological awareness’, the ability to intentionally manipulate and reason about phonemes. Phonological awareness is measured by a range of tasks that require sorting words according to their initial phoneme, segmenting or completing words and blending or eliding phonemes.⁶⁶ Although these tasks might seem to test obscure abilities, phonological awareness has enormous practical value in facilitating the acquisition of literacy.

Several factors jointly make it necessary to distinguish perception from awareness of phonemes. Most importantly, infants from four months or earlier enjoy categorical perception of phonemes,⁶⁷ but it takes them until around four years before they can think or reason about phonemes. Infants and younger children who hear ‘sat’ or ‘bat’ can distinguish the words by virtue of perceiving distinct intended phonic gestures but cannot act on or think about this difference.⁶⁸ Perception and awareness also differ with respect to the way they develop and the factors that influence their development. Unlike phoneme perception, children’s acquisition of phonological awareness develops gradually over several years, varies systematically depending on their oral language (e.g. Turkish vs. French), and is facilitated by learning a writing system where some types of writing system help more than others (e.g. syllabaries vs. alphabets). Furthermore, children find certain types of phoneme harder to distinguish in thought than others (e.g. those that differ only with respect to voicing are harder to distinguish than those that differ only with respect to articulation), whereas they have no corresponding difficulties perceiving distinctions between phonemes. In principle there might be several ways of explaining these developmental findings. But there is a natural explanation which is generally taken for granted by researchers in this area: in becoming aware of phonemes we have to rediscover them, lacking access to them as objects of perception; and we become aware of phonemes as things which sometimes differ with respect to their identities and properties from the intended phonic gestures we perceive.

If this is correct, phoneme and shape perception differ in that perceiving phonemes is not a way of being aware of them. What about causal perception? As in the case of speech, there is a long gap between perceiving and the onset of reasoning abilities. Infants enjoy categorical perception of causal interactions from no later than 9-months,⁶⁹ whereas children are two-and-a-half- or three-years-old before they can engage in even the simplest forms of causal reasoning.⁷⁰ Much as in the case of speech, children’s ability to make judgements about visually presented causal interactions appears to improve gradually over several years up to around 5-years-old,⁷¹ and they have difficulty integrating perceptual information with their reasoning.⁷² In the causal case, some investigators regard perception and awareness as closely connected despite the developmental gap, arguing that it can be explained by extraneous factors such as methodological deficiencies⁷³ or task demands⁷⁴. Given the uncertainty in this area, I suggest that the points of comparison between speech and causal perception make plausible as a contender the alternative hypothesis that the developmental gap between perception and awareness is explained by the fact that causal perception is not a way of being aware of causal interactions.

These considerations indicate that causal perception differs from shape perception with respect to introspection and awareness. The role of categorical perception of causings is only to distinguish events, not make manifest to us how the events are distinguished or the natures of the events so distinguished. From the point of view of the perceiver, categorical perception is simply a signal of sameness or difference with distinctive phenomenological effects. In this respect, causal perception is comparable to sensation: perceiving causal interactions is no more or less a way of being aware of them than experiencing a distinctive stinging sensation is a way of being aware of a nettle. If this is right, causal perception may also resemble sensation with respect to epistemology: seeing causings is not a way of gaining knowledge about them and can be a source of knowledge only because we learn to associate characteristic phenomenology with the kind of events that give rise to it. Certainly we should reject the claim that causal perception is a source of causal concepts.⁷⁵

In philosophy of perception there is a divide between those who take the claims about introspection and awareness as defining the phenomenon and those who offer theories that are at odds with, or at least hard to reconcile with, these claims. Fred Dretske’s views illustrate the second school of thought:

“Seeing objects is a way of getting information about them. […] What makes it X (rather than Y) that we see is that the information these internal events carry is information about X (rather than Y).”⁷⁶

On the face of it, an internal event’s carrying information about something does not necessarily involve either our being aware of that thing or the event’s having a particular introspective character. Something like Dretske’s way of thinking about perception might be true to causal and speech perception, and perhaps categorical perception more generally. But since these forms of perception may differ from shape perception with respect to awareness and introspection, it doesn’t follow that the same approach is true to shape perception. Equally, although the first view cannot be true to causal or speech perception because on that view it makes no sense to distinguish sharply perception of an object from awareness of it (on this view, perceiving something sometimes just is being aware of it), it doesn’t follow that the view is false when restricted to non-categorical forms of perception.

In conclusion, we do perceive causal interactions and doing so is part of perceiving objects. In this respect Hume was wrong, but he may have been more right than wrong. For perceiving causal interactions may differ from perceiving shape with respect to how such perceptions relate to awareness and introspection. If these relations are defining characteristics of the kind of perception involved in seeing shapes, Michotte, Strawson, and Scholl and colleagues are wrong to regard causal and shape perception as of a kind. We see causal interactions in the same sense we hear intended phonic gestures.⁷⁷