2007thepsychologyoffilmperceivingbeyondthecut.pdf

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/7549692

The psychology of film: Perceiving beyond the cut Article  in  Psychological Research · August 2007 DOI: 10.1007/s00426-005-0025-3 · Source: PubMed

CITATIONS

READS

36

865

2 authors, including: Filip Germeys KU Leuven 36 PUBLICATIONS   318 CITATIONS    SEE PROFILE

All content following this page was uploaded by Filip Germeys on 16 May 2014. The user has requested enhancement of the downloaded file.

Psychological Research (2007) 71: 458–466 DOI 10.1007/s00426-005-0025-3

O R I GI N A L A R T IC L E

Filip Germeys Æ Ge´ry d’Ydewalle

The psychology of film: perceiving beyond the cut

Received: 19 April 2005 / Accepted: 4 September 2005 / Published online: 8 October 2005
Springer-Verlag 2005

Abstract First-order editing violations in film refer either to small displacements of the camera position or to small changes of the image size. Second-order editing violations follow from a reversal of the camera position (reversed-angle shot), leading to a change of the left–right position of the main actors (or objects) and a complete change of the background. With third-order editing violations, the linear sequence of actions in the narrative story is not obeyed. The present experiment focuses on the eye movements following a new shot with or without a reversed-angle camera position. The findings minimize the importance of editing rules which require perceptually smooth transitions between shots; there is also no evidence that changes in the left–right orientation of objects in the scene disturb the visual processing of successive shots. The observed eye movements are due either to the redirecting of attention to the most informative part on the scene or to attention shifts by motion transients in the shot. There is almost no evidence for confusion and/or for activities to restore the spatial arrangement following the reversal of the left–right positions.

The psychology of film: perceiving beyond the cut Most current perception research involves the presentation of rather simple and typically static visual events while our daily interaction with the environment entails the visual perception of complex moving stimuli. The present study investigates dynamic displays involving movement of objects in the same scene or in the presentation of successive scenes. Such issues are central to the processing of a film. A movie typically involves acF. Germeys Æ G. d’Ydewalle (&) Department of Psychology, University of Leuven, 3000 Leuven, Belgium E-mail: [email protected]

tions in a scene and a succession of scenes from one part of the story to the next. Surprisingly, not much research has been carried out on the perceptual processing of film segments (for an overview, see Hochberg, 1986; Hochberg & Brooks, 1996). Kraft (1986) investigated the effects of cutting on viewers’ evaluation and retention of film and whether the cuts could be recalled but no attention is given to the direct perceptual effects of the cuts. In Cowen (1988), recall of action and reconstruction of the linear order of events were associated with the degree of montage linearity. The present study focuses on the direct perceptual effects of successive shots. We are particularly interested in the transition between shots from the same scene, when the position of the camera is changed. Films consist of a series of shots edited together to make a coherent visual story. A shot is a single run of the camera, while a cut is the transition between the end of one shot and the beginning of the next. In the story of the movie, the chronologically linear sequence of events is often systematically changed, as for example when flashbacks are occasionally inserted. Moreover, a film is constructed from a large number of shots, each shot typically lasting only 10 s. For example, d’Ydewalle, Desmet, and Van Rensbergen (1998) analysed a 10-min segment of the middle part of a recent movie, The Remains of the Day: there were 71 shots, giving an average duration of 8.45 s per shot. Films are on the average one hour and half long; this means that they usually consist of more than 500 cuts. Despite the appearance of several cuts each minute, the perception of a movie is typically experienced as continuous. In order to create movies, in which the sense of continuity is preserved, film makers follow editing rules (Bordwell & Thompson, 1979; Giannetti, 1982; Richards, 1992; Salt, 1992). However, for several decades, there has been a debate among film makers about the principles of film editing. The Formal Editing Principle or the so-called Hollywood concept of editing (see Bretz, 1962; Mascelli, 1965; Reisz & Millar, 1968; Vorkapich, 1972; Wurtzel, 1983) recommends that successive shots

459

should minimize perceptually disruptive transitions. The rules, which are derived from the Principle, can be standardized independently of the content of the movie. The modern viewpoint (see Godard, 1966; Wurtzel, 1983), on the one hand, stresses the consistency of the narrative structure (the Narrative Editing Principle); the consistency of the story line will overrule the perceptually disturbing effects in the transition between successive shots as attention will primarily be directed to grasping the succession of significant events in the story. On the other hand, the modern viewpoint also stresses the importance of deliberately inserting editing violations, in order to highlight occasionally critical parts of the story. d’Ydewalle and Vanderbeeken (1990) presented a classification of editing rules and accompanying editing violations. Editing violations of the first order are referred to as either small displacements of the camera position or small changes of image size, disturbing the perception of apparent movement and leading to the impression of jumping. Editing rules of second order involve a spatial-cognitive schema of the displayed scene. An editing violation in this case (sometimes also called 180-rule violations) means a switch in the left– right location of the objects in the scene. Two types of second-order editing violations are to be distinguished: the static and dynamic reversed-angle shots. A static reversed-angle shot involves the transition between shots taken on two sides of the gaze axis. For example, when a displayed actor looks to the right in the first shot, but due to a change of camera position, looks to the left in the following shot. When more than one object is in view, this results in a change in relative position of the displayed objects. Left and right positions are being switched in this case and the background is also completely different. In order to restore the perceptual continuity, the observer may have to rotate the scene. In the case of a dynamic reversed-angle shot, the object or actor leaves the screen in a first shot, moving for example from left to right on the screen. Direction continuity on the screen requires the same direction in the next shot, again from left to right. With the dynamic reversed-angle shot the constancy is violated; the actor reappears on the screen, this time moving from right to left. Third-order editing rules are meant to preserve and reinforce the narrative continuity of a story. The several parts of the story need to refer sufficiently to each other, allowing the observer to integrate them into a single chronological sequence of events. Such an integrative activity is obviously important when flashbacks (a thirdorder editing violation) occur. d’Ydewalle and Vanderbeeken (1990) and d’Ydewalle et al. (1998) manipulated the transition between successive shots in a movie, in order to unveil the processes involved for obtaining a smooth perception of the movie. In the experimental conditions, the three levels of violations were inserted at various positions of the movie; such editing violations did not occur in the control conditions. In d’Ydewalle and Vanderbeeken

(1990), the eye movements of the observers were recorded during two time intervals: zero to 200 ms, and 200 to 400 ms after a critical cut. d’Ydewalle et al. (1998) recorded the eye movements in five time blocks: 200 to 0 ms before the cut, or 0–200 ms, 200–400 ms, 400–600 ms, and 600–800 ms after the cut. In both studies, the amount of eye movements was measured by analysing the standard deviations of the x- and y-values of the observer’s point of regard in the visual field. In d’Ydewalle and Vanderbeeken (1990), the editing violations decreased the extent of eye movements but there were more eye movements following second-order than first-order editing violations during the 200–400 ms interval following the cut. The findings were interpreted as evidence for a sequence of two steps in the processing of the editing violation. There is first an immediate (i.e., during the 0–200 ms time interval) effect of the editing violations, narrowing (or focussing) attention to the changed parts of the screen. Thereafter, second-order editing violations increase the eye movements, revealing some confusion of the observers by the reversed-angle shots and/or their attempts to rotate the picture into the right axis. d’Ydewalle et al. (1998) confirmed only the second stage. Second-order editing violations increased the eye movements at the 200–400 ms time interval while there was no evidence for the first focussing stage: At the 0– 200 ms interval following a cut, eye movements were not more restricted following an editing violation than following a correctly edited shot, nor were they more restricted compared to the time interval preceding the cut. Narrative discontinuity also increased the eye movements at the 200–400 ms interval following a cut, and this occurred independently from the effects of the second-order editing violations as no significant interactions emerged between narrative structure (continuous vs. discontinuous) and the second-order editing manipulations. In the two studies, there were more eye movements 200 to 400 ms following a cut which involved a secondorder editing violation. Tentatively, the increase was explained as expressing either confusion due to the rather important change in the picture and/or cognitive activities to rotate the new visual input into the axis of the preceding shot. Scrutinizing and attempting to explain the increased eye variability at the 200–400 ms interval following a cut with a reversed-angle shot (as well established in d’Ydewalle et al. 1990, 1998) are the focus of the present study. Is the increase due to more noise in the eye data, reflecting disturbances and/or confusions following a reversed-angle shot? Or is the increased amount of eye movements related to attempts of the observers to redirect attention following an unexpected rotation of the whole scene? Despite the fact that attention and eye movements are not necessarily perfectly correlated (Klein, 1980; Klein & Pontefract, 1994), it is safe to assume that in conditions of movie watching larger eye movements do mainly reflect attention shifts. Therefore, a new movie was constructed

460

involving only two people in a discussion at the two ends of a table. On the screen, attentional areas could easily be defined: one for each speaker, and the third one outside the region of the two speakers (the background); the background was kept deliberately homogeneous in order to keep the perceptual changes limited to the two speakers after reversing the position of the camera. Only static reversed-angle shots were used as editing violations.

Method Participants Twenty-five students from various departments at the University of Leuven (Belgium) volunteered to participate in the experiment. All participants ranged between the ages of 18 to 24 years and had normal vision. Five participants with whom calibration difficulties occurred were excluded. The remaining 20 participants were randomly divided into two groups of 10 people (each time, five male and five female). The first group watched a movie with static reversed-angle shots. The second group watched the same movie without editing violations. All participants were unaware of, and uninformed about the specific purpose of the experiment. Materials and equipment The movie involved two seated people (one female and one male) in conversation facing each other. The issue being discussed concerned the making of a movie introducing the university to incoming first-year students. The movie of the experiment consisted of 22 shots and took approximately 5 min. It was filmed in an artificially created room with green panels surrounding the actors. This was done in order to keep the background constant while filming from either side of the action axis (defined by a virtual line between the two speakers). The conversation shifted continuously between the two speakers. Each shift in speaker was always accompanied by a change in camera position. So, in each shot only one actor was talking, though both actors were always in view in the target shots. Correct- and static-reversed-angle versions of the movie were constructed. Both versions of the movie were identical with the exception of five target shots (at Position 4, 9, 14, 18, and 22 of the 22 successive shots). In order to keep the presented information and layout of the reversed-angle target shots as similar as the ones in the correct-angle target shots, the five reversed-angle target shots were filmed to be near-perfect mirror images of the correct-angle versions. All correct- and reversedangle target shots, except correct-angle Shot 4 and its reversed-angle version, were filmed simultaneously by the use of two cameras. For correct-angle Shot 4 and its

reversed-angle version this was not the case, because one of both cameras was always in view of the other in this particular shot. To circumvent the problem the correctand reversed-angle version were filmed separately (the actors performing this shot twice). The camera position in the shot preceding a correctangle target shot was always at the same side of the virtual line between the two actors; this was not the case when the following target shot was a reversed-angle one (see Fig. 1). In Target Shots 1, 2, and 3, the two cameras (shooting for the correct- and reversed-angle images, respectively) were at about 45 from the two sides of the virtual line; the visual angle to the virtual line in Target Shot 5 was much smaller leading to over shoulder pictures. In Target Shot 4, the position of the two cameras were at each side of, and orthogonal to the virtual line. The movie was shown on a BARCO monitor with a screen of 40 · 50 cm2, standing at about 150 cm from the ground. Participants were seated at a distance of about 150 cm from the screen, subtending a visual angle of approximately 15 · 19 degrees. An eye-movement registration system (DEBIC 90) was used, which is based on the cornea-reflection pupil-centre method. The system has a sampling rate of 50 Hz; accordingly, every 20 ms the x- and y-value (expressed in 256 by 256 TV lines) of the participant’s point of regard in the visual field as well as the pupil diameter and a time code are stored on a computer. The system does not include any head-gear and does not impose other constraints on the head movements of the viewers. The equipment needs to be calibrated for each participant, and this takes approximately 2 min. The results of the eye-movement recording can be seen on a separate screen; a cross hair moves constantly over the scene of the film and indicates participant’s point of regard in the presented visual field for every 20 ms. The moving cross hairs and the visual field were taped. Procedure The experiment started with the positions of the participants’ right eye in the visual field being calibrated. The instructions requested the participants to watch a movie of about 5 min. The content of the movie was briefly outlined. After the movie, the participants were questioned to be sure they were unaware of the purpose of the experiment. The session was closed by explaining the true nature of the study.

Results In order to unravel the shifts of attention, the scene was divided into three areas: the speaker area (the actor talking in the target shot), the area of the non-speaking actor, and the remaining parts of the scene. A frame by frame analysis of the videotape with the cross hair was made to

461 Fig. 1 Male and female speakers around the table, as a function of the virtual line (dashed line), PS preceding shot; CA and RA correct- and reversed-angle view; OS over shoulder view

462

detect attention shifts (as expressed in the eye movements) of the observers. As we wanted to relate the findings of this analysis with the ones in d’Ydewalle et al. (1998) on the variability of the eye movements, the frame by frame scanning (with each frame lasting 20 ms) was restricted to the first 800 ms following a cut. Five time blocks were defined for each shot: 200 to 0 ms before the starting of the shot, or 0–200 ms, 200–400 ms, 400–600 ms, and 600– 800 ms after the starting of the shot. Figure 2 describes all the findings. We counted the number of observers who, after the starting position, made an eye movement to the speaker area (black bars in Fig. 2), to the non-speaker area (grey bars) or to the speaker area after moving first to the non-speaker area (white bars). This was done separately for the five time blocks. Whenever a new shot is presented, observers are still looking at a position on the screen corresponding to the information available in the previous shot. We will call this the starting position. It is important to observe that all but one (this one being in reversed-angle Shot 2) of the participants’ eyes were always focussed on the area where the speaker of the preceding shot was located; that is, at the end of the preceding shot, almost all observers were indeed focussing on the speaker of that shot. Due to the changed position of the camera, the starting position at the cut could be located at three different positions of the target shot: either in the area of the new speaker, the non-speaker (i.e., the speaker of the preceding shot), or in-between the two actors. In correctangle Shots 1, 2, 3 and 4 and reversed-angle Shots 3, 4 and 5, the starting point was situated in-between the two actors; in correct-angle Shot 5 and reversed-angle Shots 1 and 2, the starting point was situated in the speaker area. When the starting position was in-between the two actors (correct-angle Shots 1, 2, 3 and 4 and reversedangle Shots 3, 4 and 5), the observers needed to redirect attention to the new speaker in order to follow the displayed action. This mainly happened at the 200– 400 ms interval. Only a small number of shifts to the non-speaker area were detected, except in reversed-angle Shot 4. Here all observers (two observers at the 0– 200 ms interval; eight at the 200–400 ms interval) shifted attention to the non-speaker area. So, in reversed-angle Shot 4 all observers initially made an eye movement to the wrong area. Soon after the shift to the non-speaker area, all 10 observers made an additional shift to the speaker area in the 400–600 ms interval. After close examination of reversed-angle Shot 4, however, we noticed that the non-speaker made a rather quick movement of the head (to remove some hair out of her face) directly after the cut. This movement must have attracted observers’ attention causing an eye movement to the non-speaker. It should be reminded that the correctand reversed-angle versions of Shot 4 were the only two shots filmed separately. The actors performed the shot twice with a different camera position. In the correctangle version the sudden head movement was not made by the non-speaker.

When the starting position was situated in the speaker area (correct-angle Shot 5 and reversed-angle Shots 1 and 2), there was no need for the observers to move their eyes in order to follow the action on the screen. In correct-angle Shot 5, observers indeed remained fixated on the speaker area during the entire period. In reversed-angle Shot 1, most of the observers kept their eyes in the speaker area during the first 600 ms following the cut. As time went on, especially in the 600– 800 ms interval, a few observers made an eye-movement shift to the non-speaker area. In reversed-angle Shot 2, almost the same pattern was observed, although a few observers made this kind of eye movement to the nonspeaker much earlier (during 0–200 ms interval). In order to link the present pattern of findings to the findings in d’Ydewalle et al. (1998), the amount of eye movements was measured by analysing the raw data of the x- and y-values of the observer’s point of regard in the visual field. For each time block (200 to 0 ms before the starting of the shot, or 0–200 ms, 200–400 ms, 400– 600 ms, and 600–800 ms after the starting of the shot), we took as a measure of the variance of eye movements the standard deviations of the x- and y-values (the horizontal and vertical movements), separately for each participant. Analyses of variance were carried out separately for the standard deviations of the x- and y-values. Each time, the independent variables were the two groups of participants (with correct-angle vs. reversedangle shots), the target shots, the starting position (inbetween the two actors vs. the speaker of the new shot), and the five time blocks. In the analysis, the data from the reversed-angle Shot 4 were excluded, as the quick movement of the non-speaker likely attracted attention, which affected the eye movements of the observers. The analysis of variance on the vertical movements (y-value) did not produce significant main and interaction effects. The absence of reliable effects needs not to be further discussed. The critical manipulation of the experiment (correct-angle vs. reversed-angle shots) indeed produced horizontal shifts of the position of the speakers on the screen with almost no vertical shifts; therefore it was not surprising to see that the vertical eye movements were not affected by the change of camera position. In the analysis of the horizontal eye movements, it is important to mention that the nature (correct-angle vs. reversed-angle) of the target shot was never involved in significant main or interaction effects. However, critical was the starting position which interacted significantly with all other variables (except target shots), separately and together. The highest order significant interaction involved the starting position, the two groups of participants, and the five time blocks, F(4, 72) = 3.94, MSE = 16.277, p <0.006. Figure 3 describes the averages involved in the interaction. In shots with the starting position in-between the two actors, there was a peak at the 200–400 ms interval following the start of the shot; this peak was significantly different from the averages at the other time intervals,

463

Fig. 2 Number of observers moving to the speaker (black bars), to the non-speaker (grey bars), or back to the speaker (white bars), as a function of the five shots and time blocks, with correct- and reversed-angle shots (sp starting position in speaker vs. in-between two actors)

464

Fig. 3 Mean standard deviation of the X-coordinates of the eye movements as a function of the starting position and the five time blocks, with correct- and reversed-angle shots

either with correct-angle shots, F(1, 18) = 25.76, MSE = 10.095, p < 0.0001, or with reversed-angle shots, F(1, 18) = 10.62, MSE = 10.095, p <0.005. When the starting position was in the speaker, a peak at the 200–400 ms was not observed. With correct-angle shots, the averages as a function of the time blocks did not differ significantly from each other; with reversed-angle shots, there was a peak at the last time block (600–800 ms) which differed significantly from the four other means, F(1, 18) = 9.08, MSE = 35.303, p <0.008. Generally speaking, the significant findings on the horizontal eye movement variability, as described in Fig. 3, are well explained by the number of observers moving to the speaker area as a function the starting position. In all shots with a starting position in-between the two actors (see Figure 2: correct-angle Shots 1, 2, 3, and 4, and reversed-angle Shots 3 and 5), most observers moved their eyes to the speaker area at the 200–400 ms interval, causing the significant peak variability at that interval. No such peak was observed at the correct-angle Shot 5; as already noted, the starting position in the correct-angle Shot 5 was already in the speaker area. The same is true for the reversed-angle Shots 1 and 2: at the 200–400 ms interval there was no peak eye-movement variability, but also no observers moving to the

speaker for the simple reason that the starting position was already in the speaker area. With the starting position in the speaker, observers moved attention to the non-speaker area at later time intervals. This happened in the reversed-angle Shots 1 and 2. We cannot explain why this did not occur in correct-angle Shot 5: observers indeed remained fixated on the speaker area during the entire period.

Discussion The experiment showed that the eye movements were determined almost entirely by the most informative parts (the speakers) before and after the cut, whether they were filmed from the same side of the axis between the two speakers or crossing it. There was no evidence of confusion from axis crossing. Following a cut, d’Ydewalle and Vanderbeeken (1990) noticed an initial decrease of eye movements which was then followed by an increase 200–400 ms following the cut, and this was particularly evident when there was a complete switch of the left-right position of the objects (or people) in the scene (reversed-angle shots). They proposed a two-stage model: first, observers focus attention to the new relevant part of the scene;

465

thereafter follow confusion and/or an attempt to reconstruct and to connect the current spatial arrangement with the preceding one (possibly involving a mental rotation). The increased eye variability at the 200–400 ms, but not the initial decrease, was confirmed in d’Ydewalle et al. (1998). The present experiment was designed in order to unveil what is going on during the 200–400 ms interval following the cut. The findings in the present study clarify when such a peak in eye variability at the 200–400 ms is to occur, and why such a peak emerges. The peak seems entirely to be a function of the distance between the position of the most informative part from the preceding shot in the current shot (defining the starting position) and the new most informative part in the current shot. If there is no distance, no eye movements are apparent, and accordingly there is no peak at the 200–400 ms interval (see correct-angle Shot 5 in Fig. 2, and reversed-angle Shots 1 and 2 in Fig. 2). When there is a distance, the increase at the 200–400 ms interval is always observed (see correct-angle Shots 1, 2, 3, 4 and reversed-angle Shots 3 and 5). The present data do not provide evidence for confusion and/or mental rotation following a changed camera position, except perhaps in the reversed-angle Shot 2: there we find an increase of eye variability immediately after the cut (0–200 ms interval), with four observers moving attention away from the speaker. In the reversed-angle Shot 4, all observers moved the eyes from the speaker area to the non-speaker area at either the 0–200 ms interval (two observers) or 200– 400 ms interval (eight observers), leading to a first increase of eye movements. At the next interval (400– 600 ms interval), they then all moved back the eyes to the speaker. After close examination of reversed-angle Shot 4, however, we noticed that the non-speaker made a rather quick movement of the head (to remove some hair out of her face) directly after the cut. The movement must have attracted observers’ attention to the nonspeaker. In the correct-angle version the incidental head movement was not made by the non-speaker. The preceding paragraphs explain the obtained peak, or its absence, at the 200–400 ms interval following a cut. The increased eye variability is entirely due to shifting the eyes to the current speaker, as far as one is not yet fixating the speaker already; the different pattern of results in the reversed-angle Shot 4 is due to an unexpected movement in the picture. Clearly, our former studies (d’Ydewalle et al., 1998) were wrong to assume that the increased eye variability at the 200–400 ms interval following a cut with a reversed-angle shot is due to disturbances and/or confusions following a reversedangle shot. In d’Ydewalle and Vanderbeeken (1990), the main actor remained at about the same position on the screen after a correct-angle camera change; however, this was not necessarily true in d’Ydewalle et al. (1998). Accordingly, d’Ydewalle and Vanderbeeken did not obtain a peak variability after a new correct-angle shot

while a small but reliable increase in variability was observed in the same condition in d’Ydewalle et al. (1998). In the reversed-angle shots of both studies, the peak variability at the 200–400 ms interval was observed. Moreover, the peak in the reversed-angle shot was considerably larger than in the correct-angle shot in d’Ydewalle et al. (1998). In both studies, the position of the main actor was indeed vastly different following a reversed-angle shot. In d’Ydewalle and Vanderbeeken (1990), the videos were recorded in a single, regular living room; in d’Ydewalle et al. (1998), the movie displayed various actions during a walk. Accordingly, the reversed-angle shots from the two preceding studies involved a complete change in background. The movie in the present study, however, was recorded in an artificial laboratory room, with homogenous green panels at all sides; therefore, the background did not differ after a reversedangle shot. It is not clear whether this difference between the present and former studies is critically important. On the one hand, two of our pilot studies (Plees, 1998; Stellamans, 1997), also with correct- and reversed-angle shots, suggest that the object changes in the background are largely left unnoticed by the viewers. Using short videos, Levin and Simons (1997) showed that even the person might be changed across a cut, without being noticed. On the other hand, an unexpected movement in the periphery attracted attention of the viewers in the present study, suggesting that the background is being processed up to a certain extent. Moreover, experiments on the perception of objects in static scenes clearly reveal the importance of peripheral information from a scene in identifying the objects in central vision (see, e.g., De Graef, 1998; De Graef et al., 1990, 1992). Basically, the experiment shows that movie viewers move the eyes to the current speaker in a depicted conversation. If the cut causes the speaker to move to a different area of the screen, the eyes move to get there; the only exception in the experiment occurs when there is an unexpected movement in the periphery of the action. It suggests that important changes in the visual scene, due to the changed camera position, do not really disturb the viewer. The experiment provides evidence against the hypothesis that violating editing rules causes large degrees of confusion or ruins the representation of a scene. In terms of visual perception, the Formal Editing Principle (Hollywood editing rules), which emphasizes smooth transitions between shots, does not need to be followed strictly; the findings are thus more in agreement with the modern viewpoint which states that perceptual inconsistencies between shots are easily overcome by the narrative structure of the movie. Despite the fact that we clearly failed to obtain evidence for perceptual confusion when the camera crosses the action axis between two shots, we need to end with a few cautionary notes. First, the present study looked only at the successive regions (speakers, non-speakers, and background) where the eyes were directed. More finegrained analyses could discover violation effects of editing

466

rules. It is indeed common to observe increased fixation durations to a stimulus that violates expectations. However, detailed analyses on individual fixations could not be performed here as the eye-movement equipment had a sampling rate of 50 Hz which also included missing data. Despite the restricted range of actions in the movie (a discussion around the table in an artificial room), there were still many transient motions also. Second, we defined the speaker as the informative region because the discussion at the table is then under the speaker’s control. However, there is an alternative explanation. Given that a speaker is likely to produce more motion transients than the non-speaker because the mouth moves and there are more gestures, observers’ attention would likely be drawn to the speaker location by the transients. The conclusion is consistent with the analysis of reversed-angle Shot 4 in which a motion transient created by the non-speaker seems to attract the eyes. If observers wanted to direct attention to the speaker, all they would need to do is wait for the motion transient after each cut. If so, we should expect no difference in the locations of the eyes for violations and normal cuts. Given that the background of the scene was constant regardless of whether or not there was a violation, observers would have to rely on a combination of such transients and gaze direction to determine the optimal location of the eyes. In this case, the film contained a number of violation cuts, perhaps leading to an expectation that gaze angle was unpredictive of speaker location. Consequently, observers just waited for the transients to draw their attention rather than trying to use gaze direction as they normally would. We conclude that the consistency of the film narrative structure overrules the perceptually disturbing effects of cuts in the transition between successive shots. This is consistent with the literature on film segmentation and event perception. For example, Schwan, Garsoffky, and Hesse (2000); see also Zacks, Tversky, & Iyer, (2001) showed that segmentation behaviour depends primarily on the occurrence of content-defined breakpoints and is largely unaffected by the occurrence of cuts. Acknowledgements This research is supported by a G.O.A.-grant from the Flemish Government, Convention N 98/01, and by a IAP P4/19 grant from the Federal Government of Belgium. We thank Johan Van Rensbergen and Willem Stellamans in helping to set up the experiment. We are also grateful to Daniel J. Simons and Keith Rayner for helpful comments on earlier versions of this article.

References Bordwell, D., & Thompson, K. (1979). Film art: an introduction. Reading, MA: Addison-Wesley. Bretz, R. (1962). Techniques of television production. New York: McGraw-Hill. Cowen, P.S. (1988). Manipulating montage: Effects of film comprehension, recall, person perception, and aesthetic responses. Empirical Studies of the Arts, 6, 97–115.

De Graef, P. (1998). Prefixational object perception in scenes: Objects popping out of schemas. In G. Underwood (Ed.), Eye guidance in reading and scene perception (pp. 313–336). Oxford: Elsevier. De Graef, P., Christiaens, D., & d’Ydewalle, G. (1990). Perceptual effects of scene context on object identification. Psychological Research, 52, 317–329. De Graef, P., De Troy, A., & d’Ydewalle, G. (1992). Local and global contextual constraints on the identification of objects in scenes. Canadian Journal of Psychology, 46, 490–509. d’Ydewalle, G., & Vanderbeeken, M. (1990). Perceptual and cognitive processing of editing rules in film. In R. Groner, G. d’Ydewalle & R. Parham (Eds.), From eye to mind: Information acquisition in perception, search, and reading (pp. 129–139). Amsterdam: Elsevier Science Publishers B.V. (North-Holland). d’Ydewalle, G., Desmet, G., & Van Rensbergen, J. (1998). Film perception: The processing of film cuts. In G. Underwood (Ed.), Eye guidance in reading and scene perception (pp. 357–367). Oxford: Elsevier. Giannetti, L. (1982). Understanding movies. Englewood Cliffs, NJ: Prentice-Hall. Godard, J.-L. (1966). Montage, mon beau souci [Editing, my great concern]. Cahiers du Cine´ma in English, 3, 45–46. Hochberg, J. (1986). Representation of motion and space in video and cinematic displays. In K.R. Boff, R. Kaufman & J.P. Thomas (Eds.), Handbook of perception and human performance: Vol. 1. Sensory processing and perception (pp. 22–1 to 22–64). New York: Wiley. Hochberg, J., & Brooks, V. (1996). The perception of motion pictures. In M.P. Friedman & E.C. Carterette (Eds.), Cognitive ecology (pp. 205–292). San Diego, CA: Academic Press. Klein, R. (1980). Does oculomotor readiness mediate cognitive control of visual attention? In R.S. Nickerson (Ed.), Attention and performance: Vol. 8. (pp. 259–276). Hillsdale, NJ: Erlbaum. Klein, R.M., & Pontefract, A. (1994). Does oculomotor readiness mediate cognitive control of visual attention: Revisited! In C. Umilta` & M. Moscovitch (Eds.), Attention and performance: Vol. 15. (pp. 333–350). Cambridge, MA: MIT Press. Kraft, R. N. (1986). The role of cutting in the evaluation and retention of film. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 155–163. Levin, D.T., & Simons, D.J. (1997). Failure to detect changes to attended objects in motion pictures. Psychonomic Bulletin & Review, 4, 501–506. Mascelli, J.V. (1965). The five Cs of cinematography: Motion picture techniques simplified. Hollywood, CA: Cine/Graphic Publications. Plees, K. (1998). Inferentie van bewegingsrichting in filmmontages [Inference of movement direction in film montage]. Leuven, Belgium: Laboratory of Experimental Psychology. Reisz, K., & Millar, G. (1968). The technique of film editing. New York: Hastings House. Richards, R. (1992). A director’s method for film and television. Boston, MA: Focal Press. Salt, B. (1992). Film style and technology: History and analyses. London: Starword. Schwan, S., Garsoffky, B., & Hesse, F.W. (2000). Do film cuts facilitate the perceptual and cognitive organization of activity sequences? Memory & Cognition, 28, 214–223. Stellamans, W. (1997) Het effect van montagefouten op het perceptueel verwerken van film en video. [The effect of editing errors on the perceptual processing of film and video]. Leuven, Belgium: Laboratory of Experimental Psychology. Vorkapich, S. A. (1972). A fresh look at the dynamics of filmmaking. American Cinematographer, 53, 182–195. Wurtzel, A. (1983). Television production. New York: McGrawHill. Zacks, J.M., Tversky, B., & Iyer, G. (2001). Perceiving, remembering, and communicating structure in events. Journal of Experimental Psychology: General, 130, 29–58.

View publication stats