Perceived social and physical causality 2

PERCEIVED PHYSICAL AND SOCIAL CAUSALITY IN ANIMATED ABSTRACT

MOTIONS: SPONTANEOUS REPORTS AND RATINGS Michotte argued that we perceive cause-and-effect, without contributions from

reasoning or learning, even in displays of two-dimensional moving shapes. Two

studies extend this line of work from perception of mechanical to social causality. We

compared verbal reports with structured ratings of causality to gain a better

Anne Schlottmann, Elizabeth Ray, Nathalie Demetriou and Anne Mitchell understanding of the extent to which perceptual causality occurs spontaneously or

depends on instruction or context. A total of 120 adult observers (72 in the main

experiment, 48 in an initial experiment) saw 12 (or 8) different computer animations

of shape A moving up to B, which in turn moved away. Animations factorially varied

Preprint of Schlottmann, A., Ray, E., Mitchell, A. & Demetriou, N. (2006). the temporal and spatial relations of the shapes, and whether they moved rigidly or in Perceived physical and social causality in animated motions: Spontaneous reports and ratings. Invited paper for Acta Psychologica: Special Issue on the a non-rigid, animal-like manner. Impressions of social as well as physical causality Heritage of Michotte in Perception and Cognition Research, 123(1/2), 112-143. doi:10.1016/j.actpsy.2006.05.006 appeared in both free reports and ratings. Perception of physical causality was

stronger than perception of social causality, particularly in free reports. No

differences of this nature appear in infants and children, so the asymmetry may

Department of Psychology, University College London reflect learnt knowledge. Physical causality was relatively unspecific initially, but

Gower Street, London, WC1E 6BT, UK discrimination of causal and delayed control events improved with exposure to

multiple events. Experience seems to affect the causal illusion even over a short

timeframe; the idea of ‘one-trial causality’ may be somewhat misleading. Regardless Correspondence: [email protected] of such effects on the absolute level of responses, the different measures showed

similar patterns of variation with the spatio-temporal configuration and type of

motion. The good fit of ratings and reports validates much recent work in this area. Acknowledgement:

This work was supported by ESRC project grants R000237058 and R000223481.

Thanks to Jane Cownie and Henrik Westin for help with coding. Perceived social and physical causality 3 Perceived social and physical causality 4

Keywords: perceptual causality, mechanical causality, social causality, animacy, 1. INTRODUCTION

launch event, reaction event We organize the events in this world in terms of cause and effect. Sometimes

PsychINFO classification: Visual perception 2323, cognitive processes 2340 this requires effort and problem solving, however, often we simply appear to perceive that A caused B. Such causal perception may be facilitated by knowledge and

experience, but in some motion events it may be a function of simple event features.

Michotte’s (1946/1963) classic work on this issue largely concerns the perception of

physical causality. The present paper, following Kanizsa and Vicario (1968), also

examines whether observers perceive social causality. Our study considers how we

differentiate physical from social causality in simple schematic motion sequences.

We also investigate how spontaneously such impressions arise.

1.1. MICHOTTE’S LAUNCH AND ENTRAINING EVENTS.

The best-known example of perceptual causality is the launch event (Fig.1a).

Most people see the first shape (A) launching the second (B), thus causing its

movement. The entraining event is similar, but instead of stopping upon contact, A

continues forward together with B. People see the first shape pushing the second,

again reporting physical causality. Michotte reported 40-odd experiments on the

conditions under which such perceptual causality obtains. A brief pause at the point of

contact, for instance, destroys it, but other factors, such as speed and path of motion,

are also important. Michotte saw this perceptual causality as a Gestalt property of

particular motion configurations and argued that it provided an intuitive and innate

foundation for the notion of cause, independent of learning and experience.

-- insert Fig.1 about here --

Michotte’s work has been as controversial as it has been influential. Launch and

entraining events appear as simplified collisions and Michotte’s claims are all based Perceived social and physical causality 5 Perceived social and physical causality 6

on data from adults who have much experience with collision events. It is not clear Work using structured response methods, ratings or forced choices, produced why such experience should not affect observers’ causal perception. Indeed, many percentages in better agreement with Michotte (Gordon, Day & Stecher, 1990; have argued over the years that perceptual causality is no different from other, Natsoulas, 1961). Yet such studies involved an important change: Michotte’s topic experience and knowledge-based forms of causal inference (Anderson, 1990; Bruce was immediate perception, which he was at pains to separate from interpretation as

& Green, 1990; Tenenbaum & Griffiths, 2003; Weir, 1978). guided by knowledge. His data, accordingly, were spontaneous perceptual reports. It

There has also been strong criticism of Michotte’s methods (Beasley, 1968; is easier, by far, to analyse structured than unstructured reports, yet the instructions

Joynson, 1970, 1971). He did not report details of procedure and instruction, or even involved could induce particular interpretations and distort the perception. Michotte’s the data, except for some psychophysical studies involving himself and his associates critics thus saw the need for instruction to achieve near universal causal impressions as observers. Other than this, he merely states that he found perceptual causality in as evidence against the claim that causality was perceived.

“all but one or two” of “a large number of subjects (several hundreds)” tested “during Spontaneous report is less likely to be systematically biased, but it is not immune lectures … or in practical demonstration classes.” (Michotte, 1946/1963, p.20, p.40). to higher-level effects. This is why Michotte and others at the time preferred trained

A more quantitatively oriented monograph from his laboratory has not been widely and fully informed observers to naïve observers, believing that experts can report accessible (Crabbé, 1967). Michotte’s approach makes sense from within the tradition pure perceptual experience that intentionally excludes all interpretation, an of trained introspection, but for modern experimental sensibilities his procedures and assumption perhaps difficult to defend. The confounding of perceptual and higher- accounting were far too casual. level processes can also be used to explain the individual differences in naïve

Later researchers adopted more formal experimental methods. Some, like observers’ spontaneous reports: This might merely reflect difficulties with capturing

Michotte, considered spontaneous verbal reports (Beasley, 1968; Boyle, 1960; their causal percepts in words, with instruction facilitating expression. In any event,

Gemelli & Cappellini, 1958; Jansson, 1964). These studies, however, found far fewer modern researchers are well aware of the difficulty of untangling perception, causal responses: Beasley (1968) who had by far the largest sample in the published interpretation and language in structured or unstructured report. literature (181 army recruits) found only 65% causal responses to launch and 45% to Regardless of this issue, experimental studies have clearly confirmed Michotte’s entraining events, when a non-causal event also elicited 45% causal responses. Boyle claim that perceptual causality is a function of the stimulus parameters. Most of the

(1960) screened out half of his original sample because these observers failed to finer-grained analyses focus on temporal contiguity (Gruber, Fink & Damm, 1957; report causality on the first presentation. From such data, there seems little ground for Guski & Troje, 2003; Millar, 1977; Morris & Peng, 1994; Oakes & Kannass, 1999, recent claims that perceptual causality is near irresistible (Scholl & Tremoulet, 2000). Exp 3; Powesland, 1959; Schlottmann & Anderson, 1993; Schlottmann & Shanks,

1992; Shallice, 1964). All this work essentially agrees that standard causal events Perceived social and physical causality 7 Perceived social and physical causality 8

elicit causal responses and that delays of occasionally as little as 60 ms can reduce the techniques that 6-months-olds, who lack much relevant experience, are already causal impression. It is also clear that effective parameters depend on the overall sensitive to the specifically causal, not just spatio-temporal structure of launch events configuration and on practice. (Leslie & Keeble, 1987; Oakes, 1994). From at least 3 years, moreover, children link

What the most effective configurations are is less clear. Relevant other factors, launching to mechanical causality (Schlottmann, Allen, Linderoth, & Hesketh, 2002). e.g., spatial contiguity (Oakes & Kannass, 1999; Schlottmann & Anderson, 1993; Chimpanzees, but perhaps not spider monkeys, may also be sensitive to perceptual

Yela, 1952), the shapes’ velocities (Natsoulas, 1961; Schlottmann & Anderson, causality (O’Connell & Dunbar, 2005). These data support the idea that perceptual

1993), direction (Beasley, 1968), the radius of the action (Boyle, 1961) etc., have causality may be functional very early, with ongoing debate about its basis (Cohen, been studied less and comparison of results is difficult, because stimuli differ Amsel, Redford & Casasola, 1998; Saxe & Carey, 2006; White, 2006a). between studies in more than the focal aspects. Most of Michotte’s demonstrations Data from pre-verbal and non-verbal organisms also suggest that perhaps the have not been reconsidered independently. This includes work on paradoxical cases, difficulty with perceptual report can be overcome with new process measures of so called because they would not obtain in the real world and therefore important for perceptual causality. Recent reaction time (Kruschke & Fragassi, 1986), fMRI claims that perceptual causality is not the result of experience. Many believe that (Blakemore, Boyer, Pachot-Clouard, Meltzoff, Segebarth, & Decety, 2003;

Michotte and subsequent workers clearly demonstrated one-trial, immediate Blakemore, Fonlupt, Pachot-Clouard, Darmon, Boyer, Meltzoff, Segebarth, & Decety, perceptual causality and that the conditions under which it occurs are well delineated, 2001; Fonlupt, 2003; Fugelsang, Roser, Corballis, Gazzaniga, & Dunbar, 2005) and but the data are not as strong as desirable. split-brain studies (Roser, Fugelsang, Dunbar, Corballis, Gazzaniga, 2005) can all be

These issues may seem of historical, rather than current relevance, but perceptual interpreted in line with the idea that perceptual causality is separable from related causality has been of much recent interest in developmental, comparative and neuro- perceptions and other forms of causal thinking. However, such studies have not so far psychology, beginning with Leslie’s seminal work (Leslie, 1982, 1984, Leslie & produced a measure to replace perceptual report (also see Choi & Scholl, 2006).

Keeble, 1987). Researchers trying to understand infants’ earliest ontological While these data and arguments update Michotte’s view that perceptual causality distinctions argue that perceptual causality could help infants learn about the causal provides a unique foundation for the notion of cause, they leave open whether it is structure of the world without requiring causal knowledge (e.g., Baron-Cohen, 1994; later influenced by experience and knowledge (Schlottmann, 2000). Michotte’s own

Leslie, 1988; Mandler, 1992; Premack, 1990). This would help explain how children writings in places suggest that perceptual causality changes with experience, e.g., “ it acquire knowledge about the physical world so fast (Bullock, 1985; Corrigan & sometimes happens that the causal impression does not appear on the first presentation

Denton, 1996), which is difficult to understand on a traditional Piagetian (1971/1974) of the experiment, especially when it is tried on “new” subjects” (p.20), or “practice framework. In line with this are experimental demonstrations using habituation makes experienced observers more critical and the category of launching is very much Perceived social and physical causality 9 Perceived social and physical causality 10

more restricted in their case” (p.111). However, short of sensory adaptation (Gruber et 1.2. KANIZSA AND VICARIO’S REACTION EVENT al., 1957; Powesland, 1959), there have been no successful demonstrations of learning Kanizsa and Vicario (1968) proposed that there was not only perception of effects on perceptual causality (Schlottmann & Shanks, 1992), and the evidence for physical, but also of social causality. They studied an event very similar to launching the dependence of the percept on spatio-temporal parameters, as reviewed above, is (Fig. 1b). Instead of B moving upon contact, it began to move prior to A reaching it, solid as well. Thus Scholl and his colleagues (Scholl & Tremoulet, 2000; Choi & with both shapes moving simultaneously at a distance for a brief period before A Scholl 2004; 2006) argue for an encapsulated, modular process. stopped. Observers now reported that B was reacting to A’s motion, that A was The substantial individual differences in causal responses, on the other hand, and chasing it and B was trying to get away. In this reaction event, as in launching, B’s various context effects on causal impressions (Beasley, 1968; Choi & Scholl, 2004; motion is seen as caused by A’s, but the interaction is social, not physical in nature. Schlottmann et al., 2002; Schlottmann, Ray, Cole, & Hesketh, 2003; Scholl & Kanizsa and Vicario therefore argued for a perception of intentionality or social / Nakayama, 2001) suggest some involvement of knowledge and experience. This is psychological causality, in parallel to Michotte’s perception of physical causality. also implied by work on adults’ memory for launch and non-causal control events

(Choi & Scholl, 2006; Hubbard, Blessum, & Ruppel. 2001; Kerzel, Bekkering, Perception of social causality is less established than perception of physical

Wohlschläger, & Prinz, 2000). The effects can be explained as pre- or post-perceptual causality. For one thing, it might seem that social causality should involve far more

(Choi & Scholl, 2004; 2006), but without a measure of the pure percept separated from complex stimulus conditions (perhaps along the lines of Heider and Simmel’s (1944) such extraneous influences, it remains at least as plausible that perceptual causality famous stimuli) than simple mechanical causality. For another thing, Michotte denied itself is affected by knowledge and experience (Schlottmann, 2000). perception of social causality. While he mentioned finding similar reports to Kanizsa All in all, we have the curious case of a highly productive paradigm that and Vicario, he argued that these reflected interpretation, not perception, often continues to inspire empirical research and theoretical debate, while the basic data are induced by suggestion (1954/1991a, p 41, 1950/1991b, p105). However, again he not clearly documented. One aim of the present study therefore is to provide some gives no data, nor does he say exactly under which conditions he found such reports detailed data. A second aim is to triangulate naïve observers’ spontaneous perceptual (see below for an exception). Kanizsa and Vicario report an initial study with 22 naïve reports with more constrained measures from the same observers, to gain a better observers (3 more were eliminated for never reporting causality), plus understanding of the extent to which instruction and set affect the causal impression. psychophysical studies involving themselves and members of their laboratory.

It is also not clear exactly how the perception of social and physical causality

relate. Michotte (1946/1963) and Yela (1952) reported the perception of mechanical Perceived social and physical causality 11 Perceived social and physical causality 12

causation when the shapes never made contact; social reports are not mentioned. could be done, for instance, through non-rigid rather than rigid motion, as in his

Kanizsa and Vicario (1968) do not comment on this work, but they do consider caterpillar stimulus (Fig.1c). The caterpillar involves translation by rhythmic reports of both physical and social causality, emphasizing the importance of relative expansion-contraction in a way that appears self-generated and animal-like -- but speed for a distinction between the two. However, Schlottmann & Anderson (1993) again there are few independent data (Schlottmann & Ray, 2004). Our study also found substantial individual differences in how observers’ treat velocity differences, considers how causal impressions are affected when reaction and launch events so the present study focuses on spatio-temporal factors instead. involve such non-rigid, animal-like motion.

The case for reaction causality has been substantially strengthened by recent We conducted two experiments, differing only in that one had four additional developmental work: Infants from 6 months are sensitive not only to the causal stimuli and more subjects, and with minor differences in stimulus construction and structure of launch, but also reaction events (Schlottmann, Ray & Surian, 2002; instructions. Both found very similar results, so for brevity we focus on the second,

Schlottmann & Surian, 1999; Schlottmann, Surian & Ray, submitted). It is unclear more extensive experiment, with data from the initial study presented in the appendix. whether pre-verbal infants differentiate two types of causality in these events, but In the main experiment observers saw launch, reaction and delayed control from 3 years, once they talk, children distinguish social causality in reaction events events, all with rigid and non-rigid caterpillar motion. The contrast of launch and from physical causality in launch events, with no difference in the strength of the reaction events allows us to test for a distinction between perceived physical and effects (Schlottmann et al., 2002). These data are consistent with Kanizsa and social causality, but is insufficient to determine the informational basis of the domain

Vicario’s (1968) views and point towards the possibility of perception providing distinction: Launch events involve temporally contiguous contact motion, while tools to help children learn about causal interactions in more than one domain. reaction events involve simultaneous motion-at-a-distance. We therefore varied the

spatial and temporal configuration independently. This added an “entraining” event 1.3. THIS STUDY. (Michotte’s Exp. 60), with simultaneous contact motion, and a “gap” event (Exp. 31), The present work studies launch and reaction events together, considering the with contiguous motion-at-a-distance, again with rigid and non-rigid motion. This perception of social and physical causality in naïve adult observers’ spontaneous allowed us to determine whether the distinction between perceived physical and reports and directed causal ratings. In addition to considering experimental suggestion social causality depends on spatial information, on temporal information or both. via instructions, we also consider experimental suggestion through stimulus design. Developmental theory argues that spatial factors are of paramount importance Michotte reported (1950/1991b, p.105) that social impressions were more frequent if for a distinction between the social and physical world, and that motion-at -a-distance observers were prompted to take the shapes as animate agents and suggested this Perceived social and physical causality 13 Perceived social and physical causality 14

or self-initiated motion is seen as belonging to the social domain (Baron-Cohen, 1994, Materials. Twelve different movies were shown on a 12 inch colour monitor

Leslie, 1988, Mandler, 1992; Premack, 1990). This disagrees, however, with using a Macintosh Powerbook computer. Movies were made with Macromedia experimental practice in the infant literature treating motion-at-a-distance as non- Director using a HyperCard interface. Each lasted about 5 seconds (240 frames). During a trial it repeated continuously, with about 625 ms black (30 frames) causal (Cohen et al., 1998; Oakes, 1994; Oakes & Cohen, 1990), as well as with separating cycles. Michotte’s (1946/1963) and Yela’s (1952) claim that it can appear physically causal Each movie involved two squares, about 2.5 x 2.5 cm (50 x 50 pixels), Red initially to adults. Our study aims to shed some light on this, at least concerning adults. on the left, Green in the middle (initial location varied slightly depending on the spatio- Our study considered three measures of perception. First, we assessed verbal temporal configuration; the stimuli were constructed so that the shapes moved through reports of the very first event encountered, prior to any mention of causality. This the center of the 640 pixel screen at the midpoint of the cycle). Red always moved about approach lacks power, but gives a relatively pure measure of spontaneous perception, 12 cm (240 pixels) towards Green, and Green moved the same distance to the right. In unencumbered by instruction or experience with related stimuli. To avoid suggestion, contact events, contact occurred at the center of the screen. In events without contact we did not ask observers to elaborate if initial responses were insufficient, as common the shapes were about 3 cm (60 pixels) apart at the point of closest approach, in the literature since Michotte (1946/1963, p.306), and unlike Kanizsa and Vicario symmetrical about the center location. In temporally contiguous events, Red stopped (1968), we did not suggest comparisons with objects, animals or people. upon contact or when the point of closest approach was reached and Green moved after Second, we assessed observers’ verbal reports of all stimuli, still without 1 frame. In temporally delayed events, Red stopped upon contact or when the point of instruction, to assess how experience with other events might constrain closest approach was reached and Green moved after about 1.25 s delay (61 frames). In interpretation. Virtually all prior studies used such within-subjects designs. Third, we simultaneous motion events, Red and Green moved together upon contact for about 625 assessed directed rating measures for all events, considering a rating of abstract ms (30 frames) before Red stopped.1 causality, more specific ratings of both physical and social causality, and finally a Contact and no-contact events existed in two versions. In the "square" version rating of whether the shapes’ movements appeared animate. Comparison of these Red and Green moved rigidly at a constant speed of about 9.5cm/s (4 pixels/frame) to measures gives some indication of how spontaneously perceptual causality occurs, cover the 12cm in 1.25s each (240 pixels in 60 frames). In the “caterpillar” version, and of the extent to which such perceptions are affected by expectation. Red and Green moved non-rigidly, with the same average translation speed. The non-

2. METHOD. 1 These values are for the main experiment. In the preliminary study, the cycle was 241 rather than 240 frames long; there were 29 rather than 30 frames simultaneous motion, and shapes in no-contact events were separated by 104 rather than 60 pixels. Perceived social and physical causality 15 Perceived social and physical causality 16

rigid shape first expanded for about 200 ms (10 frames) at a rate of about 19cm/s (8 go back to look at your previous descriptions.” Then the events were shown, for 2 pixels/frame), with the left edge stationary, then it contracted with the right edge minutes each or until all had completed their descriptions. stationary until the original square shape was recovered. These steps were repeated Subsequently the rating task was introduced: “You will briefly see each event twice more, with 2 stationary frames separating steps. To make up for the differences again. This time we are interested specifically in your intuitive impression of why between sequences with/without delay or involving rigid/non-rigid movement, the Green (on the right) moves. Of course, these are just computer animations. stationary periods at the beginning and end of each cycle were adjusted. Objectively, Red and Green move because they are programmed to do so. However,

Design. The overall design was a Spatial (Contact/No Contact) x Temporal we are interested in what the events look like subjectively. For each event you will be

Configuration (Simultaneous/Contiguous/Delayed Motion) x Type of Motion asked four questions. Please answer by marking the [10 cm unmarked] line provided.”

(Rigid/Nonrigid) within-subjects factorial, with stimulus order counter-balanced by This was illustrated for the first rating “Do you have the impression that Red

Latin square. Observers saw the set of events three times: First, they briefly somehow made green move?”3 The instructions explained how to mark this scale: “If described each event in writing. On subsequent runs, they gave four ratings for each you feel strongly that Red made Green move, mark the left end of the scale. If you feel event. 2 that Red made Green move, but this impression is not very strong, mark the scale

Procedure. Subjects were tested in groups of 1 to 4 each, in a 20 to 30 minute towards the left, but not all the way. If you feel strongly that Red did not make Green session, writing their answers in booklets. Subjects sat 1 to 1.5 m away from the move, mark the right end of the scale, etc. If you don’t know or can’t decide, mark the monitor, which initially showed both squares stationary while subjects were talked middle. Use all of the scale to mark the strength of your impression.” through instructions also presented in their booklets: “This study is about how The remaining pages of the booklet contained four scales each, beginning with the people perceive events involving moving objects…. Your task is simply to describe abstract causality rating just described. Second was a social causality rating in what is happening in these events and what these objects are doing. We are interested response to the statement: “Red wants to catch Green. Green moves because it does in your subjective, intuitive impressions. Use everyday language, like you would to not want to get caught.” Third was a physical causality rating: “ Red hits Green. describe the event to a friend…. Use a new page [to describe] each event and do not Green moves because the hit physically set it in motion.” The scale ends were

labelled “yes” or “no”. The fourth rating was whether Red and Green moved like

2 The preliminary study included only 8 events, launch, reaction and delayed control 3 In the first experiment this question was “Do you have the impression that Green’s events, with rigid and non-rigid motion. Also, observers rated one rather than two movement was somehow caused by/a reaction to Red’s movement?” This was the replications of the design. only difference in instructions between the two experiments. Perceived social and physical causality 17 Perceived social and physical causality 18

animals, with scale ends labelled “animals” or “inanimate objects”. Subjects looked (e.g., worm versus ball; he versus it) or type of motion (e.g., slugs versus car drives over these scales and queries were answered, then the first animation was shown. away), with inter-observer agreement of 99% (91% in the initial study).

Description Coding. All written descriptions were categorized by two coders Subjects. The main experiment had 72 observers, 48 female and 24 male, with a blind to the event to which each referred, with 93% agreement between coders (88% mean age of 24 years, ranging from 18 to 52 years. About half were Psychology in the preliminary study). The first coder’s categorization was taken in case of undergraduates, a quarter students with other majors, and the remainder were disagreement. Statements were coded as describing physical causality, social causality professionals educated to at least A-level standard. All observers had normal or or independent, non-causal motion. A few descriptions left the domain of causality corrected to normal vision, and all were fluent English speakers, although 11 had a open or mentioned aspects of both physical and social causality. Some descriptions different first language. None had prior knowledge of perceptual causality.4 lay between social causality and non-causality; these were treated as noncausal for 3. RESULTS AND DISCUSSION. purposes of analysis. (See Appendix A for examples of each category.) 3.1. First Impressions. We coded for causality based on how the interaction between red and green was Table 1 shows the proportion of physically causal, noncausal and socially causal described. A statement was taken as causal if it used a verb with a causal component descriptions that observers gave to the very first animation seen. Appendix A lists the (e.g., hit, chase -- except in 2 cases in which a collision was explicitly separated from descriptions observers gave to the events. B’s movement), or if a causal connective was used (e.g., because), or if the sentence -- insert Table 1 about here -- structure implied a causal link between the motions (e.g., the red moved the green). Observers clearly differentiated between events: Standard launch and entraining The distinction between physical and social causality was also based on the verb events always appeared physically causal, whereas half of the observers saw reaction phrases. Thus “hit” “nudge” and “repel like magnets” imply that B moved due to a events as socially causal. Gap events, as well as delayed control events appeared non- physical force, while “chase”, “run away from” and “instructs to move” refer to goal- causal, except, strikingly, in the case of delayed contact events, which half of the directed links of a more social nature. We did not take into account whether phrases observers still saw as physical. Non-rigid movement decreased physical and boosted identified animate versus inanimate agents. Thus, “red slugs to green, and green social responses; this affected contact events more than no-contact events. Appendix immediately slugs to side of screen” was noncausal, because animate movement alone B shows that the same pattern appeared in the preliminary experiment. does not imply a social interaction of the two shapes. We coded separately, however, whether animate or inanimate agents were described (marked by nouns or pronouns 4 The initial study had 48 observers, 32 female, 16 male, also with a mean age of 24 (range 17 to 48), and of similar composition to the main sample. Perceived social and physical causality 19 Perceived social and physical causality 20

Causal scores, listed on the right in Table 1, were derived for the statistical The lack of differentiation between launch and delayed events may seem counter analysis. Physical reports were assigned a score of 1, noncausal reports a score of 0 to established results that delays of this magnitude destroy the causal impression. and social reports a score of –1. Highly positive or negative scores thus indicate However, Michotte already stated that naïve observers are less critical than experts impressions of physical or social causality, while scores close to 0 indicate and in virtually all published data on delay effects, observers had practice or exposure impressions that are either non-causal or non-specific in the type of causality (but it to multiple events. The present findings thus tally with previous reports, but is clear from Table 1 that response distributions were not bimodal). highlight the need to study causal events together with control events.

The domain distinction led to positive scores for contact events and negative 3.2. Descriptions. scores for no-contact events, with a main effect of the Spatial Configuration, F(1,60) Table 2 shows the proportion of causal and noncausal descriptions for all 12 = 15.427, p < .001, in the Temporal x Spatial Configuration x Type of Motion animations shown to each observer, so this table includes 864 rather than 72 ANOVA. Non-rigid motion led to less positive scores for contact events and to more observations. With exposure to multiple events, observers became less inclusive in negative scores for events without contact, with a main effect for Type of Motion, their causal reports. There were 46% less physical reports to delayed contact events

F(1,60) = 12.350, p = .001. No other effect reached significance, F (2,60) < 2.095, p = with rigid motion, F(1,72) = 15.013, p < .001, which could reflect increased

.132. Even if the descriptions from the initial experiment are included to increase sensitivity to the delay. However, there was also on average a 19 % reduction in power, no effect of the temporal factor reaches significance, F(2,107) < 1.961, p = causal responses to events without delay, F(1,47) = 7.619, p = .008, which suggests

.146. However, in the combined analysis the finding that non-rigid motion affects that a tighter decision criterion was involved as well. contact more than no-contact motion is reflected in a Spatial Configuration x Type of -- insert Table 2 about here --

Motion interaction, F(1,107) = 41.149, p < .001. The response pattern in Table 2 confirms the first impression data from Table 1.

Overall, the descriptions were not uniform, nor did observers perfectly More physical reports appeared for contact events; more social reports appeared for distinguish between events, but the trends agree with Michotte (1946/1963) and events without contact, reflected in a main effect for the Spatial Configuration,

Kanizsa and Vicario (1968): Even with minimal exposure and without instruction F(1,71) = 93.765 p < .001 in the 2 Spatial x 3 Temporal Configuration x 2 Type of observers link contact events and rigid motion to physical causality, but no-contact Motion mixed model ANOVA. events and non-rigid agent motion to social causality. That delayed control events appeared less causal than events without delay is

reflected in the Temporal x Spatial Configuration interaction, F(1,132) = 43.314, p = Perceived social and physical causality 21 Perceived social and physical causality 22

.001. (Greenhouse-Geisser corrections for violation of sphericity were applied in all appeared physical, especially with rigid motion, while reactions and to some extent analyses involving the temporal factor.) The main effect for the temporal gap events appeared social, especially with non-rigid motion. However, social configuration, in contrast, would be non-significant if the temporal factor affected causality appeared less often than physical causality. Thus, perception of social causality is weaker than that of physical causality, or observers are more reluctant or contact and no-contact events equally; scores of 1 and –1 would cancel each other. find it more difficult to express it. Appendix B shows that descriptions in the initial That this effect is nevertheless significant, F(2,128) = 23.723, p < .001, reflects an experiment were very similar. asymmetry in the reports: As is evident in Table 2, there were more physical reports for contact events than social reports for no-contact events. 3.3. Qualitative Data.

Type of motion also affected the descriptions. Although it had little effect on The descriptions provided detailed qualitative information beyond the three delayed events, non-rigid motion made descriptions of the other events less physical categories used for the quantitative analysis; we report these data for the main and and more social. This effect was stronger for contact than no-contact events: Thus the preliminary experiments combined (1240 descriptions total) to increase sample size. causal score for launch and entraining events was far less positive with non-rigid than One issue was the range of situations mentioned. Unsurprisingly, the vast majority of rigid motion, while the score for reaction and gap events was somewhat more physically causal descriptions, 238 of 257, referred to collisions; 8 of these – all for negative with non-rigid motion. This led to a main effect for Type of Motion, F(1, 71) events without contact -- mentioned an invisible barrier between the shapes. The = 93.82, p < .001, a Temporal Configuration x Type of Motion interaction, F(2,135) other 19 physical descriptions referred to magnets; 16 of these were for events = 28.221, p < .001, a Spatial Configuration x Type of Motion interaction, F(1,71 = without contact. 10.648, p = .002 and the 3-way interaction, F(2,137) = 5.326, p =.007. Physical accounts focused on how the first shape (Red) caused the second Despite these differences between event configurations, McNemar tests (Siegel

& Castellan, 1988) confirmed that delayed events attracted less causal responses in shape’s (Green) motion, without concern for what set off Red. Observers did not all cases, with, p < .001 for reaction events with non-rigid and rigid motion, p = .035 mention external forces acting on Red, nor Red’s intentions. Noncausal reports also and p <.001 for gap events with non-rigid and rigid motion, p = .008 and p<.001 for did not mention external causes or either shape’s intention. This is noteworthy, entraining events with non-rigid and rigid motion, and p = .027 and p<.001 for launch because many argue that self-initiated motion is basic for attributions of intentionality events with non-rigid and rigid motion, respectively. Thus perceptual causality, albeit (e.g., Baron-Cohen, 1994; Leslie, 1988; Mandler, 1992; Premack, 1990). weaker in some events than others, appeared throughout. Forty-nine causal statements could not be categorized as physical or social, Overall, events with simultaneous or contiguous motion appeared more causal because they were unspecific about the domain of causation (e.g., “Green is made to than delayed control events, but the events differed in that launching and entraining Perceived social and physical causality 23 Perceived social and physical causality 24

move away”), or described collisions in which it was not clear whether Green was chases Green’, all ambiguous as to whether Red was chasing after Green or chasing it propelled physically or moved to avoid further hits. There were also 2 instances of away. In the other 109 cases, two verbs appeared, one for Red’s and one for Green’s collisions initiated intentionally (“e.g., Red nudged Green and Green complied”.) motion. Twenty-two of these described a goal-directed approach for Red plus a

Of 190 social causality reports, 148 referred to approach-avoidance / chase- simple movement for Green (e.g., Red chases Green and then Green slides off), 37 escape scenarios. Another 17 described a meeting/exchange of information and 25 cases described both chase and escape, i.e., two goal-directed motions, and 50 cases were more idiosyncratic (trying to eat, kiss, hug etc.). described a simple motion for Red plus a goal-directed avoidance for Green. Social accounts, like physical ones, emphasised how Green’s motion was caused, All in all, the majority of observers mentioned Red first and described Green as i.e., that it reacted to Red, with less concern for what set off Red, but this asymmetry reacting to Red. In this respect, reports of perceived social and physical causality are was less pronounced than in physical reports. It is unclear whether this difference is indeed analogous. Social causality might be less frequent than physical causality linguistic or perceptual: In physical causality, inevitably the first motion causes the reports, in part, because it is harder to find concise, yet precise expressions for this. second, which maps directly to phrases such as “Red pushes Green”. Thus temporal, Finally, we considered whether observers described goal-directed interactions or causal and linguistic structures coincide, with many transitive verbs ready-made to whether they elaborated on the agents’ underlying mental states. In 190 social reports express such relations. In social causality, however, this is not the case. Observers could focus on Red’s chasing action, or Green’s running away. Moreover, statements we found 62 references to intentions or other mental states (with 5 references in the such as ‘Red chases Green’ are ambiguous as to the direction of causality: Red might other 1050 reports). Observers might, of course, have thought of how the agents’ chase after Green, with Red reacting to Green, or Red might chase Green away, with mental states mediate the interactions without mentioning them, but in any event,

Green reacting to Red. In the latter case, temporal, causal and linguistic structure perceived social causality does not typically involve making these mental states coincide as for typical collision reports; however, in such accounts both motions are explicit. This is another parallel with perceived physical causality, where observers goal-directed, while in typical collision accounts Red’s motion is left unspecified. also describe mechanical interactions without making the underlying mechanism The descriptions we found reflected the temporal order of the event: Only 16 of explicit. Reports in both domains are consistent with Michotte’s point that observers 148 approach/avoidance descriptions focused exclusively on Green’s motion or see causal links between motions rather than think about the underlying connection. mentioned this first. But although most observers mentioned Red first, there were In sum, the qualitative data highlight parallels and differences between perceived only 3 cases of ‘Red chases Green away’, parallel to a transitive physical physical and social causality. Reports for both showed a causal asymmetry (White, description. There were also 5 cases of ‘Red chases after Green’, and 16 of ‘Red 2006b), with focus on what the first did to the second shape, but most importantly, Perceived social and physical causality 25 Perceived social and physical causality 26

the data confirmed that one appears as collision, the other as approach/avoidance or non-rigid or simultaneous motion alone, without at least one of the other features causality. The suggestions in the rating instructions thus agree with spontaneous produces distinctly lower ratings. perception and are not likely to distort reports. The spatial configuration is more important than the manner of agent motion for

the domain distinction; this in turn is more important than the temporal pattern. This 3.4. Ratings of Causality. cue hierarchy is clearest in the right panel of Fig. 2, showing the difference between Physical and Social Causality. Causal ratings are in Fig. 2 (top left and centre; physical and social ratings. This gives a measure of the distinctiveness of each event, ignore other panels for now). Contact events (solid lines) are rated high in physical, with positive scores indicating physical causality, negative scores indicating social low in social causality. Events without contact (dashed lines) generally elicit slightly causality and scores around 0 indicating non-distinct and/or non-causal events lower ratings than contact events, with reverse pattern. Delayed events received low (analogous to the causal score used for the descriptions). Delayed events, regardless ratings of physical and social causality throughout. Launch and entraining events with of other event features, all score around 0, indicating that this is crucial for non- rigid agent motion elicited the strongest and most distinctive impression of physical causality. Diverging lines for events without delay indicate that other features become causality. Reaction events with non-rigid motion elicited the strongest and most important for the meaning of potentially causal events. distinctive impression of social causality, but reaction events with rigid motion and Greater separation of curves for contact and no-contact events than for rigid and gap events with non-rigid motion were not far behind. The corresponding rating data non-rigid motion indicates that the spatial factor is more important than the manner of from the initial experiment (Appendix C) show a very similar pattern. motion for the domain distinction, but the effects are configural: First, type of motion -- insert Fig. 2 about here -- has more effect in contact than no-contact events, with more separation in the top The data show that all factors considered here contribute to the distinction of than bottom pair of curves. Second, when spatial pattern and type of motion point in non-causality, physical and social causality. Temporal factors are crucial for the the same direction, the temporal factor has little effect, with the top and bottom causal versus non-causal distinction and a delay seems largely sufficient to rule out curves for rigid contact motion and non-rigid no-contact relatively flat. The middle causality. The absence of a delay by itself, however, produces only moderate ratings - curves have more slope: If the spatial pattern conflicts with the agent cue, - a strong impression of causality depends on additional factors. simultaneous motion amplifies the effect of the spatial configuration, perhaps because Contact and rigid motion are both needed for a good physical impression; contact or its absence are visible longer in simultaneous than contiguous motion. whether the motions are contiguous or simultaneous matters less. Each factor alone -- Nevertheless, the curves are not symmetrical about the midline, so non-rigid contact contact without rigid motion or rigid motion without contact -- elicits lower ratings. motion remains non-distinct, while rigid motion without contact can appear social. For a good social impression, in contrast, a combination of no contact with non-rigid or simultaneous motion or both is needed, i.e., this category seems wider. No contact Perceived social and physical causality 27 Perceived social and physical causality 28

An ANOVA on the difference score found a significant Temporal x Spatial comparable. As a result, contact motion is slightly more distinctive, with more

Configuration x Type of Motion interaction, F(2,142) = 39.167, p = 000, with all positive scores, in descriptions than ratings. In contrast, no-contact motion is more lower order effects, F(2,122) > 6.445, p = .004. The 3-way interaction reflects distinctive, with more negative scores, in ratings than descriptions. primarily that delayed events were least distinctive; it was not significant anymore Such differences between ratings and descriptions might reflect the disparity of when delayed events were excluded, F(1,71) = 1.219, p =.273, although all other the measures -- there is no a priori reason why description frequency should be effects recurred, F(1,71) > 6.920, p = .010. Effects of the spatial factor, of motion proportional to impression strength – but we found similar differences between social type and the interaction also appeared in all 3 temporal configurations analysed descriptions and ratings in the main and preliminary experiments and in another separately, F(1,71) > 6.099, p = .016, confirming that physical and social impressions unpublished part-replication, so we are inclined to treat them as real. If so, the rating depend on specific stimulus configurations. instructions seem to boost the level of social responses to all events; they also slightly

When individual events were considered, of 8 events without delay, 7 were more amplify differences in social causality between events. distinctive than corresponding delay events, F(1,71) > 10.475, p = .002; the exception Individual Subjects. Consideration of individual data confirmed the group results: was contiguous, non-rigid contact motion, F<1. Among the highly social events, non- Table 3 shows the number of observers giving higher ratings of one or the other type rigid reaction events were more distinctive than rigid reaction events, F(1,71) = of causality to different events. For events in which the cues agree, i.e., a potentially 5.189, p = .026, and than non-rigid gap events, F(1,71) = 5.660, p = .020; the latter causal temporal pattern and either rigid contact or non-rigid no-contact motion, more two did not differ, F<1. The highly physical events, rigid launching and entraining, than 85% gave domain specific ratings. For conflict events agreement was lower. Rigid did not differ, F(1,71) = 1.833, p = .180. While perceptual causality is not confined to no-contact motion nevertheless appeared social to the majority, but non-rigid contact typical launch and reaction events, it is also clear that observers make subtle motion attracted fairly mixed ratings. This difference might reflect different cue distinctions between the events. strengths: Rigid motion in conflict with no-contact was generally discounted, so it Comparison of Ratings and Descriptions. The bottom panels of Fig. 2 repeat the might be the weaker cue. Non-rigid and contact motion may be closer in strength, in description data from Table 2 in the same format as the ratings. This makes the qualitative similarity in response pattern obvious; events rated higher or lower in contrast, leading to a more ambiguous percept. Finally, for delayed events, ratings causality of one type also elicit more or less descriptions of this type. were mostly unspecific, but a minority continued to give domain-specific ratings.

Fig. 2 also shows that ratings and descriptions differ in the absolute level of -- insert Table 3 about here -- responses: Most saliently, social causality ratings (centre panel) are higher and more Abstract Causality. When asked whether Red somehow makes Green move, spread out than descriptions. For physical causality (left), the spread is more delayed events were rated lower than events without delay (Fig. 3). Simultaneous Perceived social and physical causality 29 Perceived social and physical causality 30

motion was rated slightly higher than contiguous motion (except for rigid contact = 1.380, p = .254. The Temporal Configuration effect appeared for both non-rigid motion) and differences between temporal configurations were more pronounced for motion and rigid motion F(2,139) > 5.812, p = .007, and F(2,139) = 5.813, p = .007, rigid than non-rigid motion motions (squares versus circles). Contact events (solid but the Spatial Configuration effect was only significant for rigid motion, F(1,71) = 15.298, p < .001 and F(1,71) = 2.314, p = .133, as was the interaction, F(2,113) = lines) were rated higher than events without contact (dashed), leading to a Temporal x 3.923, p=.031 and F < 1. Spatial Configuration x Type of Motion interaction, F(2,135) = 5.251, p = .007, with -- insert Fig. 4 about here -- all lower order effects, F(1,71) > 5.798, p =.019. Follow-up analysis found main References to animate agents in the descriptions were also affected by motion effects of the Spatial Configuration and of Type of Motion for delayed, contiguous rigidity, but not by the event context. The 864 descriptions contained 274 references and simultaneous motion events, F(1,71) > 11.817, but the Spatial x Type of Motion to animate agents (including 17 for rigid motion), 177 to inanimate agents (including interaction was significant only for contiguous events, F(1,71) = 13.804, with F<1 for 17 for non-rigid motion.), 27 had both animate and inanimate features (all but 1 for both delayed and simultaneous motion events. non-rigid events.), and 78 reports described non-rigidity without reference to

-- insert Fig. 3 about here -- animacy; 2 of these appeared with rigid motion. All types were equally distributed

Overall, when observers assessed the causality of schematic motion events at an across the four relevant events. All in all, the link between nonrigid motion and abstract level, temporal factors were crucial, but observers also thought that perceived animacy was stronger than that between rigid motion and inanimacy

causality in no-contact events and non-rigid events was somewhat weaker. This 4. GENERAL DISCUSSION corresponds to the previous finding that social causality tends to be somewhat Our results can be summarised simply: Launch and entraining events were seen weaker than physical causality. as physically causal, reaction events as socially causal, but physical was stronger

than social causality. Observers provided causal reports from first exposure and the 3.5. Animacy. data pattern was stable across measures, but experience with the events and type of Non-rigid motion received far higher animacy ratings than rigid motion, but the measure affected the absolute level of responses. Our data confirm a number of event configuration also affected the ratings (Fig. 4). Motion without contact was Michotte’s and Kanizsa and Vicario’s observations, including some that sit uneasily rated as slightly more animal-like than contact motion, but this appeared mainly for with the claim that perceptual causality is independent of learning and experience. non-delayed, rigid motion. The ANOVA showed an Type of Motion x Spatial x 4.1. Spontaneous Causal Perception Temporal Configuration interaction, F(2,135) = 6.280, p = .003, with all lower order This study provided clear data from 120 observers that impressions of physical effects, F(1,70) < 5.141, p =.026, except the Spatial x Temporal interaction, F(2,110) and social causality appear spontaneously, without specific instruction on first Perceived social and physical causality 31 Perceived social and physical causality 32

viewing of schematic events, with close to 100% physical causality responses to type of system involved (Buehner & McGregor, in press; Huber, Schlottmann, & paradigmatic launch and entraining events, and close to 60% social causality Daum, 2004). Perceptual causality provides causal hypotheses for later evaluation by responses to reaction events. This confirms Michotte’s (1946/1963) and Kanizsa and more considered processes, and one might speculate that these hypotheses may

Vicario’s (1968) claims. The high rate of initial causal responses agrees with recent initially be broad, at least with respect to temporal parameters, in anticipation of work with adults (Gordon et al., 1990) and children (Schlottmann et al., 2002), but is calibrating input from experience. Further work is necessary to explore these issues. higher than in older studies (Beasley, 1968; Boyle, 1960), perhaps because observers When observers saw multiple events, they reported causality less often. Physical have become used to “virtual” events. They may also have benefited from clear reports for delayed launching decreased by more than 45%. From this it seems that expectations of what they would see: Observers were shown the stationary shapes and experience with related events increased discrimination, thus affecting perceptual told they would see them move. Lack of such prior understanding may contribute to sensitivity, in line with previous studies of adaptation involving longer stimulus runs initially “chaotic and unorganised” (Michotte, 1946/1963, p.20) impressions. (Gruber et al., 1957; Powesland, 1959). But causal reports for causal launch and

Responses to causal events must be seen, however, against a baseline of responses reaction events also decreased by around 20%. Thus, in addition experience may also to non-causal controls: One-trial perceptual causality is only meaningful if there is have produced a shift in bias, i.e., observers’ general tendency to give causal also one-trial non-causality. But in both studies here, spontaneous discrimination was responses to causal and noncausal stimuli alike decreased. Bias shifts also appeared far from perfect: While delayed no-contact events attracted less than 20% causal in work with children (Schlottmann et al, 2002), suggesting an influence of cognitive responses, strikingly, over 60% were still obtained for delayed contact events. This decision processes on reports of perceptual causality. was not due to a short delay: 1.25 s far exceeds the threshold for non-causality of 60 Michotte (1946/1963, p.111) already stated in passing that more experienced to 200 ms achieved in psychophysical studies; even infants are sensitive to such observers are often more critical than naïve observers. One might thus extrapolate delays. Reduced discrimination was not limited to the first few seconds of viewing: that his reports of close to universal causal perception in hundreds of observers

Events repeated for up to 2 minutes, until observers finished their descriptions. during demonstration classes were based on relatively isolated presentations (again,

Reduced temporal discrimination also appeared in work with children (Schlottmann he does not report details). His psychophysical studies, in contrast, largely involved et al., 2002). It would thus not seem an entirely fleeting phenomenon. experienced observers. In combination, this leads to a slightly misleading picture of

The tendency to see delayed contact events as causal could be due to the media immediate causality, as it is often portrayed in the literature. To the extent that the used here -- people are used to “sticky” computers. Alternatively, it might signal a impression is spontaneous and close to universal, it may not be as specific as more general initial tolerance of delays that rapidly adapts to the range of events. Michotte argued. It does not require much experience for the impression to become as

There is indeed some evidence that tolerance of delays varies dramatically with the specific, but this seems to involve not only sensory adaptation, but also cognitive Perceived social and physical causality 33 Perceived social and physical causality 34

decision elements. Observers’ more critical attitude to non-causal and causal events in no-contact events, improving the social impression in the other conflict case, i.e. for could in part stem from thought about what the events might mean. rigid no-contact motion, and when spatial and motion cues agreed, with stronger social

None of this argues against perceptual causality, but highlights once more that in impressions in non-rigid reaction than gap events. Michotte saw the presence of similar adults this integrates both perceptual and experiential determinants. The strongest motion states (i.e., common fate) as an integrative factor supporting causal perception. evidence for the involvement of a specific perceptual structure is that perceptual This may come into play mainly when the causal impression is not at ceiling. causality appears so early in infancy, prior to much relevant experience (e.g., Leslie Taken together, the temporal configuration seems to modulate strength of the

& Keeble, 1987). But although we know that infants’ percepts are sensitive to the causal impression, while its distinctiveness depends on the spatial configuration and causal structure of events as shown here, their contents remain a matter of conjecture: the type of object motion. Of course, other factors not manipulated here might also

Infants cannot tell us about them, and talking-age children still lack ability to freely play a role. Relative speed, for instance could make the events more distinctive, with express finely nuanced distinctions. Verbal report data from adults may not be faster motion for shape A favouring physical causality (Michotte, 1946/1963), faster conclusive on process, but are essential to explicate the nature of the percepts. motion for B favouring social causality (Kanizsa & Vicario, 1968). Absolute speed of

the objects, in contrast, might contribute to impression strength, with higher speeds 4.2. PERCEPTUAL CONFIGURATIONS FOR PHYSICAL AND SOCIAL CAUSALITY being more integrative (Michotte, 1946/1963). A key result was the good agreement in the pattern of responses between verbal Our data speak to claims that contact versus self-initiated motion without contact reports and ratings. The robustness of the data pattern allows us to characterise the are basic to a perceptual distinction between the physical and social world (Baron- perceptual configuration for physical and social causality with greater confidence: Cohen, 1994, Leslie, 1988, Mandler, 1992; Premack, 1990). For our adults, contact or Delayed motion appeared non-causal; potentially causal events without delay were its absence was clearly not the only cue. This could be a developmental effect, based made distinctive by other event features. Contact and rigid motion pointed towards on learning about other distinctive features of the social and physical world. Recent physical causality, lack of contact and non-rigid motion towards social causality. The studies suggest, however, that self-initiated motion may not be necessary or sufficient spatial factor was more important than object motion. When the cues conflicted with for perception of goal-directed action or social causality in infancy either (Csibra, respect to the domain, the impression was less clear. Gergely, Biro, Koos, & Brockbank, 1999; Schlottmann et al., under review). The role of motion contiguity versus simultaneity was more complex. For contact Our data disagree with Michotte (1946/1963) who argued that contiguous events motion it made little difference: Physical causality was equally strong in standard without contact can still elicit impressions of physical causality -- launching-at-a- launching and entraining. A minor effect appeared only in case of conflict between distance: In a sample of 50 naïve observers, Yela (1952) found 54% causal responses contact and non-rigidity; simultaneous motion appeared slightly more physical than when we found around 20% with a 3 cm gap. The different results may reflect contiguous motion, except on first exposure. Simultaneous motion played a larger role Perceived social and physical causality 35 Perceived social and physical causality 36

differences in instructions or other event features. Yela’s speeds were higher, for On first glance, it might seem that our parameters were sub-optimal. Stronger instance, and with the classic disc projection method the rim of the display slit social impressions might appear if the shapes moved simultaneously for longer, connects the shapes. Both factors might help pull the motions together. perhaps with contingent changes in direction. With Heider and Simmel’s (1944)

More extensive and powerful psychophysical designs with multiple stimulus famous animation, only 1 in 34 naïve observers failed to provide a social description. levels and multiple replications will be necessary for a finer-grained analysis of the Under this approach, however, physical and social events are not directly comparable. effects outlined here. Important at present, however, is the essential similarity in the Social events are more complex, with a more realistic movement sequence. This in patterns for descriptions and ratings. This similarity argues against two typical itself might prompt a social description, rather than the stimulus configuration per se. criticisms found in this area of research: That ratings distort spontaneous perception In the present paradigm, in contrast, launch and reaction events both involve minimal because the instructions introduce biases, while spontaneous reports without specific information for high-level concepts, information of equivalent complexity. direction are insensitive because observers may fail to express important detail. The Another argument points to possibly sub-optimal speeds. Our shapes had equal good agreement of our measures validates much recent work in this area. velocities, but for Kanizsa and Vicario (1968) reaction events elicit better social

The remainder of the discussion will focus on two issues that may further impressions if A is slower than B. However, this does not explain the specific illuminate the role of learning and experience in the causal illusion. The first issue is weakness of social causality, because according to Michotte (1946/1963), launch how to interpret the absolute levels of social and physical responses: Although the events elicit better physical impressions if A is faster. With equal speeds, therefore, pattern of variations across events was similar for reports and ratings, social causality social and physical causality should both suffer. appeared less often than physical causality in free reports; the two were much closer We used equal speeds, firstly, because prior work on effects of relative speed in strength in the ratings. The second issue is that although all three event parameters found extreme individual differences, plausibly related to differences in knowledge manipulated here affected the causal impression, this does not mean that they and experience, in 60 naïve observers (Schlottmann & Anderson, 1993). Michotte, in necessarily operate at the same process level. These points will be discussed in turn. contrast, presented data from 3, Kanizsa and Vicario from 4 experienced observers.

Secondly, demonstrations of perceptual causality in infants and children were 4.3 IS THE SOCIAL ILLUSION WEAKER? successful with equal speeds (Cohen & Amsel, 1998; Leslie, 1982, 1984; Leslie & Less frequent social descriptions could mean that the illusion of social causality Keeble, 1987; Oakes, 1994; Oakes & Cohen, 1990;); our stimuli are very similar to is weaker than that of physical causality or that the stimulus parameters we chose those in Schlottmann and Surian (1999) and Schlottmann et al. (2002). Overall, were more suited to impressions of physical than social causality. A third possibility variation of this factor seemed likely to complicate this study unnecessarily. is that observers are simply less inclined to report social than physical causality. Perceived social and physical causality 37 Perceived social and physical causality 38

If parameter values do not specifically disadvantage social causality, then is the causality perception. That social causality is also more easily modified by rating social illusion weaker? No sign of this appears in development: Three- to 9-year olds instruction underscores this point, as do findings that domain-general, abstract easily identify instances of social and physical causality (Schlottmann et al., 2002). causality ratings are lower for social causality cues (no-contact, non-rigid motion)

With infants as well, perceptual causality for launch and reaction events emerges in than for physical causality cues. In this view, both forms of perceptual causality may parallel from 6 months (Cohen & Amsel, 1998; Leslie & Keeble, 1987; Oakes, 1994; have similar strength early in life, as would appear from the developmental data, but

Schlottmann, et al., 2002). The relative weakness of perceived social causality is late this may change later: The social world involves more complex agents and less emerging, presumably reflecting learning about the nature of the social world. regular actions than the mechanical world, so schematic social and physical events of

Adult observers may report social causality less often, because the animations similar complexity will appear to have quite different degrees of artificiality. Thus seem further removed from real social than real physical situations: People may be experience might weaken the illusion of social causality more. reluctant to anthropomorphize the actions of animated squares, while “physicalizing” Both views agree that perceptual impressions of causality (or at least reports of “blocks” may go undetected. Also, as discussed earlier, more variability in how to thereof) reflect both structural and cognitive, experience-dependent factors, but they describe even simple social events would make it harder for observers to find precise differ in their definitions of perception. On the first view, perception is defined at the expressions, thus further reducing the likelihood of social reports. This leaves open, level of the underlying process, on the second, at the level of the phenomenal of course, whether the perception is weaker or whether observers suppress it more. experience. Both seem valid, but if in addition we want to dissociate process and

Two views are possible in the end: In one, the asymmetry of social and physical experience with the argument that the (modular) process is pure, while reported reports reflects cognitive factors, while the actual perception of social causality is no experience can be cognitively contaminated, then a measure of this process weaker than that of physical causality. The rating task signals that social accounts are independent of the phenomenology is needed (as offered by Kawachi & Gyoba permissible and provides a tool for articulating them. This redresses a knowledge- (2006) for the tunnel effect). In the absence of such a measure for perceptual based imbalance in how easily and willingly social versus physical impressions of causality (Choi & Scholl, 2006; Schlottmann, 2000, it would seem more schematic events are expressed verbally. If perception of high-level concepts such as parsimonious to assume that the perceptual process in adults, like their experience, causality and animacy is tinged by what we know about these concepts, then directed integrates structural and learnt factors. instruction may on occasion be an appropriate antidote to such effects, helping to 4.4. SUGGESTION AND THE EFFECTS OF NON-RIGID MOTION strip the perception back to a less adulterated state. The present studies manipulated two event parameters, the spatial and temporal In the other view, spontaneous report, though imperfect, is still the best measure configuration, as well as one parameter of the objects, namely their manner of of perceptual experience available. Fewer social reports thus indicate weaker social Perceived social and physical causality 39 Perceived social and physical causality 40

motion. All had strong effects on perceptual causality, but as Michotte suggested, from initial encounter with the stimuli. Cognitive effects, in contrast, might become they may not all operate on the same level. stronger upon reflection. Both types of effect might be involved here.

For Michotte, only the event configuration affected perceived physical causality. A perceptual effect may be at work in contact events, i.e., launching and

The nature of the objects did not matter, even when it led to “logically absurd” entraining: In the contiguous event, shape B begins to expand as soon as shape A situations (1946/1963, p.82). Social causality, in contrast, depended on suggestions finishes its last contraction, but it takes about 400 msec until B finally breaks contact

“that human actions were represented” (1954/1991a, p.42), or, for instance, use of his when it in turn begins to contract. Phenomenally, this appears neither as causal, nor caterpillar stimuli (1951/1991b, p.105). For Michotte, the first reflects perception, as independent motion; rather the contiguous motions at the left and right edge of the while the second involves interpretation or “extrinsic meanings” (1954/1991a, p.42). joint shapes distract from their point of contact in the middle; the two objects seem to

In making this distinction he did not, however, seem to apply equal standards to merge temporarily rather than there being a transfer of motion at the point of contact. social and physical causality. When use of caterpillar motion destroyed the launch The shapes also seem to merge in simultaneous contact events, in which A reaches B effect (1946/1963, Exp.71, p.210), he argued that the the nonrigid motion interfered at the end of an expansion; the two then move together with B expanding as A with perception of the interaction. Social reports are not mentioned in his book on contracts, etc. In this view, the perception of launching and entraining is disrupted, causality, but in other writings he characterizes the boost that caterpillar motion may because the edge of motion is not the edge of contact between the shapes. give to social reports as interpretation (1951/1991b, p.105). No rationale is given for In no-contact events, in contrast, this is not an issue, with the objects clearly this contrast between a perceptual account of object effects in perception of physical separated throughout. Indeed, reaction events appear unaffected on initial exposure; causality and a cognitive account in social causality, nor are the data reported. non-rigid motion improves the causal impression only in later descriptions/ratings.

In the present study, non-rigid caterpillar motion indeed led to weaker This may be better explained as a cognitive effect: The non-rigid motion appears impressions of physical causality for launch and entraining events, and to stronger animate, which in turn increases the plausibility of a social interaction. Similarly, impressions of social causality for gap and reaction events. The data thus agree with animate motion in contact events may be interpreted as evidence against a purely

Michotte. Interestingly, the effect on launching and entraining was apparent right physical link. This interpretation may operate together with the purely visual from the initial impressions, in both experiments. The effect on reaction and gap distraction, and would predict even less frequent physical reports after some events, in contrast, was not apparent initially. experience than on first encounter. A further 30% decrease indeed appears for launch,

This difference in temporal development, if reliable, may allow a tentative but not entraining events. Interestingly, children and infants seem less influenced by distinction between a perceptual and a more cognitive account of object effects. If the the non-rigid motion manipulation than the adults (Schlottmann et al., 2002; non-rigid motion interferes with the perception of the interaction, this should appear Perceived social and physical causality 41 Perceived social and physical causality 42

Schlottmann et al., under review). This would seem to fit better with the cognitive Perceptual causality has long been controversial, in part due to the problems with view, but it is not clear why children should escape perceptual distraction. Michotte’s data, in part, because the phenomenon seems redundant with and difficult

This tentative account agrees with Michotte’s claims that non-rigid motion can to separate from other experience-dependent forms of causal thinking. The present have perceptual and cognitive effects, but further work is necessary. The present study confirms and extends work on perceptual causality, but also suggests that interpretation firms up Michotte’s view that perceptual interference might be limited experience contributes to the perception. It may be difficult, if not impossible, to find to contact stimuli. We see little reason, however, to limit cognitive effects to no- conclusive evidence of a specific perceptual structure for causality with adults, contact stimuli, or to consider the perception of social causality more cognitively because its true function may be developmental: Perceptual causality is a tool to aid mediated than that of physical causality. The data disagreed with Michotte, but children’s learning about the causal texture of the world. That the perception gets agreed with Kanizsa and Vicario (1968), in showing that social causality was not overlaid with the knowledge it helps us acquire suggests that it has fulfilled its role. dependent on caterpillar motion or other forms of suggestion. Instead, like physical causality it arises spontaneously, dependent only on specific event configurations.

4.5. CONCLUSION

We perceive both physical and social causality in minimal motion events. This appeared both in the frequency of appropriate causal descriptions in free reports, as well as in structured ratings of the extent of the causal relationships. The distinction between the two forms of causality depended on the spatio-temporal configuration, as well as on the manner of agent motion. The illusion of physical causality was stronger than that of social causality, especially in observers’ descriptions. On the whole, if social causality was suggested through instructions or agent motion, this increased social responses. Suggestions of a physical interpretation had less effect, but experience also affected perception of physical causality, in particular, its specificity improved with more exposure to the events. The general pattern of responses, however, remained the same on initial exposure, after some experience, and regardless of instruction. This affirms Michotte’s point that variations in the perceptual impression depend closely on variations in stimulus conditions. Perceived social and physical causality 43 Perceived social and physical causality 44

Appendix A: The descriptions given by 72 observers in the main experiment to the first (26) The red square stretches across the screen until it reaches the green square then they both stretch for a couple of moves – the red square stops in the centre and the green carries on to the event encountered and the classification for each description. P = physical causality, S = right. (N) social causality A/N = ambiguous social, treated as non-causal for the purpose of analysis, (27) Red worms goes across hits green, green wiggles off. (P/S) N = non-causal. (55) Wriggling like a worm. (N) (56) Red square on left, the green square on right. Red streaks towards green & stops, green streaks away to right and stops Motions are a bit like worms moving. (N) Rigid agent motion (57) The red square is running after the green one, catches it first but then the green square manages Simultaneous contact event to run away again. (S) (10) Red hits green then red and green travel together then separate (P) Contiguous contact event (11) Red pushing Green over short distance. Green offering no resistance. (P/S) (7) Red block moves towards green they meet green moves away (A/N) (12) Red block slides in and bumps into the green block, pushing it along (P) (8) The red square moves along screen until it comes in contact with green square -- pushing the (40) Red is barging the green away. (P) green square across the screen, whilst the red square has stopped (P) (41) Red hitting green (P) (9) Red block moves towards green block When they meet the green block moves away. (A/N) (42) Red bumped into green + pushed it away. (P) (37) Red is running towards and pushing green who tries to run away. (P/S) Contiguous contact event (38) Red one bumps into the green (in an expanding way) one that then starts moving in the same (64) Red square slides in and meets the green square. The push makes the green square slide off. way as the red. (P) (P/S) (39) The red one pushes the green one. they both move like worms. (P) (65) Red block slides into green block pushing it away. (P) Delayed contact event (66) Red block pushes green block. (P) (61) Red caterpillar crawls in, meets the green square, which then caterpillars away (A/N) (70) Red block slides into Green block + pushes it away. (P) (62) Red worm creeps in + stands next to green one for a while, Green worm then creeps to right. (A/N) (71) Red square pushes green square away. (P) (63) The red snake moves to the green snake, they meet and then the green snake moves away. (A/N) (72) Red square pushes green square away. (P) (67) The red worm crawl in + stops adjacent to the green worm. They talk for a while + then the Delayed contact event green worm crawls away. (S) (16) A Red object hitting a stationary green object. Just like total conservation of momentum. Red (68) The red shape moves next to the green one. The green one then moves away. (A/N) stops and green carries on (P) (69) The red worm-like shape moves in from the left and goes next to the green one. After a few (17) Red hits green, green moves in opposite direction at the same speed red was going at. (P) seconds, the green one moves away. (N). (18) Bumping into each other and transferring kinetic energy. (P) Simultaneous no-contact event (46) Red came up to green and then green moved away (A/N) (22) The red Square, is changing in length and pushing along the Green one, which is also gaining (47) The red square moves to the right, and the green square moves after it. (A/N) in length so that the distance between the two is kept equal. (P) (48) Red going to meet green, stay together, then green disappears off (A/N) (23) Red moves first, green moves on then red follows, red stops first, both travel same distance. Simultaneous no-contact event green is smaller than red. (N) (1) Red is chasing the green and green is running away from it. (S) (24) Red block contracts and expands in snake-like manner as if chasing the green also moving in (2) Red block pushes green block away, like magnets (- to - pole) (P) same way. (S) (3) Red block repells green block (like magnets) (P) (52) Red and green squares elongating and shrinking while moving. Looks like red chasing green (31) Red moves to centre + as approaching the green block moves away. Like something has square (S) triggered Green to move once red is 1/4 of the way. (P/S) (53) Red worm chasing green worm (S) (32) Red block chasing green block away. Green starts moves away when red gets near. (S) (54) Red square inches towards green square. they both seem to elongate as they move (reminiscent (33) Red chases green, green runs away before red catches it. Red stops halfway across the screen. (S) of caterpillars!). Green square inches away, when red square gets on square’s width in front of Contiguous no-contact event it and red square stops. (A/N) (19) Red moves to right, before it reaches green, green moves to right at same speed (N) Contiguous no-contact event (20) The red square moves right, towards the green square, then the green square moves right, away (13) Red slides towards green, Green slides away from red. (A/N) from the red square. (N) (14) Red square snaking across to green square (N) (21) Red moves to the middle then green moves to edge from the middle (N) (15) Red blob (caterpillar) moving toward Green, which then runs away. (S) (49) The red block chases towards the green, which then runs away. (S) (43) The elongating red block gets closer to the green one, causing the latter to move away. (P/S) (50) Red object is bumping pushing green. (P) (44) Red one slinks up to the green one, which then slinks off. (N/A) (51) Red block moves towards green which is stationary then red block stops + green block moves (45) Red crawls along – up to green green moves before red can get him (S) to the right (N) Delayed no-contact event Delayed no-contact event (4) Red shuffles towards green. Green shuffles off. (A/N) (28) Red dashes in Green leaves. (A/N) (5) Red creeps up on Green Green moves away (A/N) (29) Red approaches Green Green moves away (A/N) (6) Red moves – stops. then green moves. (N) (30) Red block stops within a short distance of green block, green block moves away. (A/N) (34) Red is crawling towards the green, which causes a chain reaction for the green to crawl. (P/S) (58) Red block moving from left to centre Green from centre to right. (A/N) (35) Red is chasing Green (S) (59) Red and green squares -- like magnetic field – red approaching green – when the red comes in (36) Red moves from left to centre, elongating then contracting like a caterpillar. Green does the a certain range, the green goes away. (P) same, from centre to right (N) (60) Red block slides in; stops, green block slides out after a short pause (N)

Non-rigid agent motion Simultaneous contact event (25) Red box moving, green box moving to the right, each elongating, then becoming equal size boxes, across the screen. (N)

Perceived social and physical causality 45

Appendix C:

Initial Impressions All Descriptions % % Mean P N P N S Causal S Rigid agent motion Score Rigid agent motion Launch event 79 17 4 .745 Delayed contact event 20 66 14 .064 Launch event (contact) 100 0 0 Reaction event 17 49 34 -.170 Delayed contact event 75 17 Delayed no-contact event 4 77 19 -.149 8 Non-rigid agent motion Reaction event (no-contact) 8 33 58 Launch event 32 53 15 .170 Delayed no-contact event 0 67 Delayed contact event 12 75 14 -.021 33 Reaction event 3 51 46 -.426 Non-rigid agent motion Delayed no-contact event 4 75 21 -.170

Launch event 50 17 Overall (N=376) 21 58 21 .005 33 Delayed contact event 20 80 0

Reaction event 0 50 50 Delayed no-contact event 0 83 17

Overall (N = 47) 32 43 26

-- insert Fig. 5 about here --

Perceived social and physical causality 47 Perceived social and physical causality 48

REFERENCES: Anderson, J. R. (1990). The adaptive character of thought. Hillsdale, NJ: Erlbaum. Baron-Cohen, S. (1994). How to build a baby that can read minds: Cognitive mechanisms in mind-reading. Cahiers de Psychologie Cognitive, 13(5), 513-552. Beasley, N. A. (1968). The extent of individual differences in the perception of causality. Canadian Journal of Psychology, 22, 399-407. Blakemore, S. J., Fonlupt, P., Pachot-Clouard, M., Darmon, C., Boyer, P., Meltzoff, A. N., Segebarth, C., & Decety, J. (2001). How the brain perceives causality: An event-related fMRI study. Neuroreport, 12, 3741-3746. Blakemore, S. J., Boyer, P., Pachot-Clouard, M., Meltzoff, A., Segebarth, C., & Decety, J. (2003). The detection of contingency and animacy from simple animations in the human brain. Cerebral Cortex, 13, 837–844. Boyle, D. G. (1960). A contribution to the study of phenomenal causation. Quarterly Journal of Experimental Psychology, 12, 171-179. Boyle, D. G. (1961). The role of the ‘radius of the action’ in the causal impression. British Journal of Psychology, 52, 219-226. Bruce, V., & Green, P. R. (1990). Visual Perception. (2nd Ed.) Hillsdale, NJ: Erlbaum. Buehner, M., & McGregor, S. (in press). Temporal delays can facilitate causal attribution. Towards a timeframe bias in causal induction. Thinking and Reasoning. Bullock, M. (1985a). Causal reasoning and developmental change over the preschool years. Human Development, 28, 169-191. Choi, H., & Scholl, B. J. (2004). Effects of grouping and attention on the perception of causality. Perception and Psychophysics, 66(6), 926-942. Choi, H., & Scholl, B. J. (2006). Can the perception of causality be measured with representational momentum? Acta Psychologica, 123(1), in press. Cohen, L. B. & Amsel, G. (1998). Precursors to infants’perception of the causality of a simple event. Infant Behavior and Development 21 (4), 713-732. Cohen, L. B., Amsel, G., Redford, M. A., & Casasola, M. (1998). The development of infant causal perception. In A. Slater (Ed.) Perceptual development: Visual, Auditory and speech perception in infancy, (pp. 167-209). Hove: Psychology Press. Corrigan, R., & Denton, P. (1996). Causal understanding as a developmental primitive. Developmental Review, 16, 162-202. Crabbé, M. (1967). Les conditions d’une perception de la causalité. Paris: Center Nationale de la Recherche Scientifique. Csibra, G., Gergely, G., Biro, S., Koos, O., Brockbank, M. (1999). Goal attribution without agency cues: The perception of "pure reason" in infancy. Cognition, 72(3), 237-267. Fugelsang, J. A., Roser, M. E., Corballis, P. M., Gazzaniga, M. S., & Dunbar, K. N. (2005). Brain mechanisms underlying perceptual causality. Cognitive Brain Research, 24(1), 41-47. Fonlupt, P. (2003). Perception and judgement of physical causality involve different brain structures. Cognitive Brain Research, 17 (2), 248-254. Gemelli, A., & Cappellini, A. (1958). The influence of the subject's attitude in perception. Acta Psychologica, 14, 12-23. Perceived social and physical causality 49 Perceived social and physical causality 50

Gordon, I. E., Day, R. H., & Stecher, E. J. (1990). Perceived Causality occurs Michotte, A. (1991a). Autobiography. Reprinted in G. Thinès, A. Costall, & G. with stroboscopic movement of one or both stimulus elements. Perception, 19, 17-20. Butterworth (Eds.), Michotte's phenomenology of perception, (pp. 24-49). Hillsdale, Gruber, H. E., Fink, C. D., & Damm, V. (1957). Effects of experience on the NJ: Erlbaum. (Original published 1954). perception of causality. Journal of Experimental Psychology, 53, 89-93. Michotte, A. (1991b). The emotions regarded as functional connections. Guski, R., & Troje, N. F. (2003). Audiovisual phenomenal causality. Perception Reprinted in G. Thinès, A. Costall, & G. Butterworth (Eds.), Michotte's & Psychophysics, 65(5), 789-800. phenomenology of perception, (pp. 66-87). Hillsdale, NJ: Erlbaum. (Original published Heider, F., & Simmel, M. (1944). An experimental study of apparent 1950). behavior. American Journal of Psychology, 57, 243-259. Michotte, A. E. (1963). The perception of causality. (Tr. by T. R. Miles & E. Hubbard, T. L., Blessum, J. A., & Ruppel, S. E. (2001). Representational Miles.) London: Methuen. (Original published 1946). momentum and Michotte’s (1946/1963) “Launching effect” paradigm. Journal of Morris, M. W. & Peng, K. (1994). Culture and cause: American and Chinese Experimental Psychology: Learning, Memory, & Cognition, 27(1), 294-301. Attributions for social and physical events. Journal of Personality and Social Huber, S., Schlottmann, A., & Daum, M. (2004). Judged and perceived Psychology, 67(6), 949-971. causality in visual and action-outcome sequences. Perception, 33, ECVP 2004 Natsoulas, T. (1961). Principles of momentum and kinetic energy in the Supplement, 326. perception of causality. American Journal of Psychology, 74, 394-402. Jansson, G. (1964). Measurement of eye movements during a Michotte Oakes, L. M. (1994). Development of infants’ use of continuity cues in their launching event. Scandinavian Journal of Psychology, 5, 153-160 perception of causality. Developmental Psychology, 30(6), 869-870. Joynson, R. B. (1970). The breakdown of modern psychology. Bulletin of the Oakes, L. M. & Cohen, L. B. (1990). Infant perception of a causal event. Cognitive British Psychological Society, 23, 261-269. Development, 5, 193-207. Joynson, R. B. (1971). Michotte’s experimental methods. British Journal of Oakes, L. M., & Kannass, K. N. (1999). That’s the way the ball bounces: infants’ Psychology, 62, 293-302. and adults’ perception of spatial and temporal contiguity in collisions involving bouncing Kanizsa, G., & Vicario, G. (1968). The perception of intentional reaction. In G. balls. Developmental Science, 2(1), 86-101. Kanizsa & G. Vicario (Eds.) Experimental research on perception. (pp. 71-126) O’Connell, S., & Dunbar, R. M. (2005). The perception of causality in Trieste: University of Trieste. (In Italian). chimpanzees. Animal Cognition, 8, 60-66. Kawachi, Y. & Gyoba, J. (2006). A new response-time measure of persistence Piaget, J. with the collaboration of R. Garcia (1974). Understanding causality. in the tunnel effect. Acta Psychologica, 123(1), in press. (Tr. by D. Miles & M. Miles.) New York: Norton. (Original published 1971). Kerzel, D., Bekkering, H., Wohlschläger, A., & Prinz, W. (2000). Launching the Powesland, P. F. (1959). The effect of practice upon the perception of causality. effect: Representations of causal movements are influenced by what they lead to. Canadian Journal of Psychology, 13, 155-168. Quarterly Journal of Experimental Psychology, 53A(4), 1163-1185. Premack, D. (1990). The infant's theory of self-propelled objects. Cognition, 36, 1- Kruschke, J. K., & Fragassi, M. M. (1996). The perception of causality: 16. Feature binding in interacting objects. In Proceedings of the 18th Conference of the Roser, M. E., Fugelsang, J. A., Dunbar, K. N., Corballis, P. M., Gazzaniga, M. S., Cognitive Science Society, (pp, 44-446) Mahwaj, NJ: Erlbaum. (2005). Dissociating processes supporting causal perception and causal inference in the Leslie, A. M. (1982). The perception of causality in infants. Perception, 11, brain. Neuropsychology, 19(5), 591-602. 173-186. Saxe, R. & Carey, S. (2006). The origin of the idea of cause: Critical reflections on Leslie, A. M. (1984). Spatiotemporal continuity and the perception of Michotte's theory with evidence from infancy. Acta Psychologica, 123(1), in press. causality in infants. Perception, 13, 287-305. Schlottmann, A. (2000). Is perception of causality modular? Trends in Cognitive Leslie, A. M., & Keeble, S. (1987). Do six-month-old infants perceive Sciences, 4(12), 441-442. causality? Cognition, 25, 265-288. Schlottmann, A., Allen, D., & Linderoth, C., & Hesketh, S. (2002). Children’s Leslie, A. M. (1988). The necessity of illusion: Perception and thought in intuitions of perceptual causality. Child Development, 73(6), 1656-1677. infancy. In L. Weiskrantz (Ed.), Thought without language, (pp.185-210). Oxford: Schlottmann, A., & Anderson, N. H. (1993). An information integration approach Oxford University Press. to phenomenal causality. Memory and Cognition, 21(6), 785-801. Mandler, J. M. (1992). How to build a baby II: Conceptual Primitives. Schlottmann, A., & Ray, E. (2004). Perceptual animacy in schematic motion events. Psychological Review, 99, 587-604. Perception, 33, ECVP 2004 Supplement, 308. Millar, D. B. (1977). Perceptual cohesion in simple motion configurations. Schlottmann, A., Ray, E., Cole, K., & Hesketh, S. (2003). Context affects Unpublished doctoral dissertation. University of California, San Diego, La Jolla. perceptual causality in young children. Biennial Meetings of the Society for Research in Child Development. Tampa, FL, USA Perceived social and physical causality 51 Perceived social and physical causality 52

Schlottmann, A., Ray, E. & Surian, L. (April 2002). 6-months-olds’ perception of causation-at-a-distance. International Conference on Infant Studies. Toronto, Canada. Schlottmann, A., & Shanks, D. R. (1992). Evidence for a distinction between judged and Table 1: Observers’ descriptions of the first event encountered. The table shows the perceived causality. Quarterly Journal of Experimental Psychology, 44A(2), 321-342. proportion of physically causal, non-causal and socially causal descriptions (modal Schlottmann, A., & Surian, L. (1999). Do 9-month-olds perceive causation-at-a- response in bold) and mean causal scores for each event. The 7 mixed physical/social or distance? Perception, 28, 1105-1113. unspecific causal descriptions were counted as half answers in both categories. Schlottmann, A., Surian, L., & Ray, E. (2005) Eight- and 10-months-old infants’ perceive action and reaction. Under review. % Mean Scholl, B. J., & Nakayama, K. (2002). Causal capture: contextual effects on the Physical Non Social Causal perception of collision events. Psychological Science, 13(6), 493-498. Rigid agent motion Causal Score Scholl, B. J., & Tremoulet, P. (2000). Perceptual causality and animacy. Trends in Cognitive Sciences, 4(8), 299-309. Simultaneous contact event “Entraining” 92 - 8 .833 Shallice, T. (1964). The detection of change and the perceptual moment hypothesis. Contiguous contact event “Launch” 92 - 8 .833 British Journal of Statistical Psychology, 17, 113-135. Delayed contact event 50 50 - .5 Siegel, S., & Castellan, N. J. (1988). Nonparametric Statistics. Singapore: McGraw- Hill. Simultaneous no-contact event “Reaction” 42 - 58 -.167 Tenenbaum, J. B., & Griffiths, T. L. (2003). Theory-based causal inference. In Contiguous no-contact event “Gap” 17 67 17 0 S. Becker, S. Thrun, & K. Obermayer (Eds.), Advances in Neural Information Delayed no-contact event 17 83 - .167 Processing Systems, 15 (pp. 35–42). Cambridge, MA: MIT Press. Weir, S. (1978). The perception of motion: Michotte revisited. Perception, 7, 247-260. Non-rigid agent motion White, P. A. (1988). Causal processing: Origins and development. Psychological Bulletin,104, 36-52. Simultaneous contact event “Entraining” 8 67 25 -.167 White, P.A. (2006a). Illusions of causality. Acta Psychologica, 123(1), in press. Contiguous contact event “Launch” 58 33 8 .5 White, P.A. (2006b). The causal asymmetry. Psychological Review, 113(1), 132- Delayed contact event - 83 17 -.167 147. Yela, M. (1952). Phenomenal causation at a distance. Quarterly Journal of Simultaneous no-contact event “Reaction” 17 33 50 -.333 Experimental Psychology, 4(4), 139-154. Contiguous no-contact event “Gap” 8 50 42 -.333 Delayed no-contact event 8 67 25 -.167

Overall (N = 72) 34 44 22 .101

Perceived social and physical causality 53 Perceived social and physical causality 54

Table 2: Proportion of physically causal, non-causal or socially causal descriptions, as well as causal scores for each condition, with the modal response in bold. The 38 mixed Table 3: Number of subjects rated each of 12 events as higher in physical causality, physical/social or unspecific causal descriptions were counted as half answers in both as similar in both (ratings differed by no more than 10 points) or as higher in social categories. causality, with the modal category in bold; the median difference between physical and social ratings is also given.

Phys Both Social Median diff % Mean Higher Similar Higher Phys - Soc Physical Non- Social Causal Rigid shape Causal Score Simultaneous contact Entraining 63 7 2 56 Rigid agent motion Contiguous contact Launch 62 6 4 65

Delayed contact 21 39 12 1 Simultaneous contact event “Entraining” 76 18 6 .708 Simultaneous no-contact Reaction 4 7 61 -66 Contiguous contact event “Launch” 83 13 4 .792 Contiguous no-contact Gap 7 16 49 -33 Delayed contact event 8 81 12 -.042 Delayed no-contact 4 39 29 -5

Simultaneous no-contact event “Reaction” 18 57 25 -.069 Non-rigid shape Contiguous no-contact event “Gap” 22 64 14 .083 Simultaneous contact Entraining 36 17 19 9 Delayed no-contact event 5 89 6 -.014 Contiguous contact Launch 15 18 39 -12

Delayed contact 24 33 15 1 Non-rigid agent motion Simultaneous no-contact Reaction 0 4 68 -71

Contiguous no-contact Gap 0 6 66 -62 Simultaneous contact event “Entraining” 31 58 11 .194 Delayed no-contact 2 24 46 -22 Contiguous contact event “Launch” 19 65 16 .028

Delayed contact event 4 81 16 -.125

Simultaneous no-contact event “Reaction” 6 53 42 -.361 Contiguous no-contact event “Gap” 4 71 25 -.208 Delayed no-contact event 4 83 13 -.083

Overall (N = 864) 23 61 16 .075 Perceived social and physical causality 55 Perceived social and physical causality 56

FIGURE CAPTIONS:

A B Fig. 1. (a) Schematic launch event. (Shape A moves towards B which is stationary.

After contact, A is stationary and B moves away.) (b) Schematic reaction event. (B (a) begins to move before contact, slightly before A has reached its final position.) (c) Non-rigid "Caterpillar" motion. (A square shape moves across the screen by first (b) elongating from the right edge, then contracting from the left edge.)

Fig. 2. Domain-specific causality ratings. (The top panels show observers’ ratings of physical (c)

(left) and social (centre) causality for contact (solid symbols) and no-contact events (open symbols) with rigid (solid lines) and non-rigid agent motion (dashed lines). The right panel shows how distinctive the events are, i.e., the difference between the physical and social ratings. The bottom panels repeat the description data from Table 4 in the same format, showing the proportion of uninstructed physical and social reports, as well the difference between these proportions, which is equivalent to the causal score. ) Fig. 1.

Fig. 3: Abstract causality ratings. (Ratings were given for 12 different events in response to the question “Did Red somehow make Green move?")

Fig. 4: Animacy ratings. (Ratings were given for 12 different events in response to the question “Do Red and Green move like animals?”)

Figure 5: Ratings of physical causality (top left), social causality (top right), abstract causality (bottom left) and animacy (bottom right) for the preliminary experiment. The data pattern is very similar to the pattern for the main experiment for the domain-specific ratings, but small differences appear for the abstract rating (attributable to the change in instructions) and for the animacy ratings (unclear origin).

Perceived social and physical causality 57 Perceived social and physical causality 58

physical physical causality social causality distinctiveness (= physical - social) causality “Launch” “Entraining” contact, rigid contact, rigid no contact, non-rigid no contact, rigid contact, rigid

contact, non-rigid contact, non-rigid contact, non-rigid non-distinct/ non-causal contact, non-rigid no contact, rigid no contact, non-rigid contact, rigid no contact, rigid “Gap” “Reaction” no contact, rigid no contact, non-rigid no contact, non-rigid social causality physical causality

contact, rigid

contact, rigid

contact, non-rigid

non-distinct/ contiguous simultaneous non-causal no contact, non-rigid no contact, rigid contact, non-rigid no contact, rigid no contact, non-rigid no contact, rigid Fig. 3 contact, non-rigid no contact, non-rigid contact, rigid social causality

Fig. 2 Perceived social and physical causality 59 Perceived social and physical causality 60

contact, rigid no contact, non-rigid

no contact, rigid

no contact, non-rigid contact, non-rigid contact, non-rigid contact, non-rigid

contact, rigid

no contact, rigid no contact, non-rigid no contact, rigid contact, rigid no contact, non-rigid contact, rigid contact, non-rigid no contact, non-rigid no contact, rigid contact, non-rigid

contiguous simultaneous

contact, rigid no contact, rigid Fig. 4

Fig. 5