Entry

The effect of Entry on attention across matched-exit/entrance cuts was investigated under the pursuit initiation hypothesis. Based on existing evidence that an object needs to reappear from behind an occluder for it to be perceived as the same object (Gibson et al., 1969; Michotte, 1991; Spelke et al., 1995; Xu & Carey, 1996). The

171 hypothesis stated that subjects would expect the object to re-enter the screen from a position of full occlusion after the cut. If the object appeared partially or fully on the screen after a cut it was predicted that viewers would have difficulty allocating attention to the object (reflected in low correct response rates immediately after the cut) and perceive the motion of the object as more erratic (rated less “smooth” on the questionnaire).

In general, the results of this study indicated that the opposite was true: locating the object fully on the screen immediately after the cut (100% Entry) led to highest correct response rates and highest “smoothness” ratings. The results indicate that 100% Entry produces the shortest period of attention withdrawal under 0%, 50%, and Random Exit conditions. The condition predicted by the pursuit initiation hypothesis to produce the best performance, 0% Entry, actually leads to the longest periods of attention withdrawal. In the 100% Exit group there does not appear to be a benefit of any Entry condition as all produce a very short period of attention withdrawal. In fact, the only decrease in correct response rates for the 100% Exit group occurs in the frame immediately after the cut (cue position 1). This indicates that when subjects are repeatedly presented cuts with 100% Exit they perform saccades across the cut with an average duration less than 83ms. Saccades with such short durations must be reflexive (Sheliga et al., 1995; Walker et al., 2000).

Previously, the benefit of 100% Exit was associated with the voluntary withdrawal of attention prior to the cut (see 4.4.1) and 50% Exit was seen as the main benefactor of reflexive saccades. This evidence indicates that 100% Exit also seems to benefit from reflexive saccades. A voluntary saccade performed in anticipation of a cut would be expected to show a decrease in performance lasting 150-200ms (Rose et al., 2002). The 100% Exit subject group show a decrease lasting an average of 83ms. This occurs even if the focal-object is fully occluded after the cut (0% Entry) and if there are signs of saccade preparation prior to the cut (seen in the 50% Entry condition). Both these conditions indicate that the reflexive saccade begins before the focal- object has relocated across the screen. If the reflexive saccade was only seen in response to a cut with 100% Entry the speed of the saccade could be attributed to

172 attention capture by the suddenly appearing object (see 2.3.2). The focal-object would be functioning as an attentional pull cue. However, when the focal object relocates to a position under the opposite screen edge (0% Entry) it cannot attract attention. This suggests that it is the focal-object’s occlusion by the screen edge that is redirecting attention. Based on this evidence the screen edge can be viewed as an attentional push cue similar to a gaze shift.

What cannot be known, given the design of this study, is where the screen edge redirects attention to. The convention of relocating the focal-object to the opposite screen edge after a matched-exit/entrance cut seems to suggest that attention is pushed directly across the screen. This could be seen as the opposite to Hirsch’s spatiotemporal continuity during occlusion. Spatiotemporal continuity during normal occlusion is believed to function by viewers projecting the focal-object’s path prior to occlusion forwards across the space of the occluder to predict where the object will re-appear (Hirsch, 1982). If the same projection occurs in reverse when a focal- object is occluded by the screen edge it would predict that the object would reappear at the point at which this projection intersected with the opposite screen edge. Unfortunately, in this study the focal-object always appeared at the same position along the screen edge so no effect of location (other than Entry) can be seen. To examine if viewers had a preference for the focal-object’s re-entry position, attention would need to be probed at various positions around the screen edge immediately after a matched-exit/entrance cut. Alternatively, eye tracking could be combined with an unexpected Time Gap to detect if viewers shifted there gaze to a specific screen location without the focal-object being present.

An alternate explanation of the rapid saccades observed after 50% and 100% Exit could be that a matched-exit/entrance cut functions both as an attentional push and pull cue. Occlusion by the screen edge may initially disengage attention from the focal-object, pushing attention back on to the screen in a distributed form. When either the focal-object or a reaction time cue suddenly appears on the screen the viewer’s attention is then captured (“pulled”) by the new object. Such a combination of pushing and pulling of attention would enable the editor to redirect the viewer’s

173 attention to any part of the visual scene, not just the screen edge directly opposite the occluding edge. An empirical study such as that described above would be required to investigate this explanation.

The big question raised by this evidence of reflexive saccades after 100% Exit is: Do these reflexive saccades maintain the perception of existence constancy created by occlusion? The sudden, unnatural relocation of an object through space should violate the expectation of spatiotemporal continuity created by occlusion (Hirsch, 1982; Michotte, 1991; Spelke et al., 1995; Xu & Carey, 1996). However, the primitive status of matched-exit/entrance cuts within the continuity editing styles seems to suggest that they do result in the perception of “continuity”. Whether the editor’s concept of “continuity” can be interpreted as the same concept referred to by perceptual and developmental psychologists will be discussed in Chapter 5:.

4.4.2.1

Flash-lag, Fröhlich, and Edge Effect

One final result emerging from the Entry manipulations should be discussed before moving on: the “edge-effect”. Across a lot of experimental conditions, when the reaction time cues were presented 125ms after the cut (cue position 4) a decrease in correct response rates was observed which deviated from the smooth increase in correct response rates either side i.e. cue positions 3 and 5 (see section 4.3.4). This difficulty in reacting to the cues presented so soon after a cut can be attributed to an interaction between two established visual phenomenons: the flash-lag and fröhlich effect.

The Flash-lag effect refers to the illusion that when a flash and a moving object are presented at the same location the flash is perceived as located behind the moving object (Eagleman & Sejnowski, 2000). In this study the effect occurred whenever a reaction time cue was flashed on top of the focal-object. The impression was of the cue lagging behind the focal-object83. Given that the velocity of the focal-object was 83 _{This illusion was the source of many hours of head scratching for the experiment’s designer who} checked and rechecked the positioning of the cues before being informed of the flash-lag effect.

174 consistent throughout its screen time, any negative effects the flash-lag effect might have on performance should have been consistent across all cue positions.

However, the flash-lag’s effect on performance may have increased immediately after saccades due to its interaction with another visual illusion: the fröhlich effect (Fröhlich, 1923). When subjects are asked to determine the position at which a fast moving stimulus first enters a screen, they will typically mislocate the entry position further along the stimulus’ path (Fröhlich, 1923). This phenomenon occurs when the moving target is viewed peripherally (Müsseler & Aschersleben, 1998), pursued foveally after an unexpected onset, or saccaded to after the onset (Yarrow et al., 2005). In this experiment, the Fröhlich Effect would have occurred whenever the focal-object entered the screen edge. The impression would have been of the focal- object accelerating on to the screen. This distortion of the focal-object’s location combined with the perceived displacement of cue and focal-object position caused by the flash-lag effect may have combined to make response to the cue very difficult. In addition to this, cue position 4 for 0% Entry also lay directly on top of the screen edge. The disadvantage this appears to have caused was an unfortunate side-effect of the methodology used in this study. However, the ability to measure attention at so many specific screen locations offsets this disadvantage.