priming of visual attention by continuity editing in...

of 1 /1
Research suggests that attention in dynamic scenes is biased towards content that differs from what was observed before [1]. While this may be prevalent in many situations, we propose that a preference for repeated information is as well observable. Professional footage contains cinematic cuts, with either continuity editing or discontinuous editing. Continuity cuts repeat major elements from a pre-cut scene in a post-cut scene. Although such cuts constitute abrupt changes, they are often not realized by viewers. In contrast, viewers are more frequently aware of discontinuous cuts [2, 3]. We hypothesize that after cuts attention is captured by visual features repeated from the pre-cut scene. Methods Background Results Continuity had an effect on manual reaction times (RTs) and manual response accuracy. When the movies continued at the same posi- tion, discontinuous cuts were reliably detected. With continuity, the fraction of missed cuts nearly doubled. When movies continued at the different position, performance did not differ between the two types of cuts. However, saccadic reaction time (i.e., relocating gaze to the different position of the task- relevant movie after the cut) was reduced by 26 ms with continuity cuts as compared to discontinuous cuts. Our results demonstrate that the classical technique of continuity editing achieves the desired result of perceptual continuity and a surprisingly large number of cuts is missed. This effect is however not additive with a change of spatial location. We found differences in saccadic latency between continuity cuts and discontinuous cuts, which indicate that information from the periphery contributes to the decision process whether the movies switched positions or not. We propose that re-orientation of attention and gaze after abrupt scene changes, like with cinematic cuts, is primed by repeating visual content from the pre-cut scene. Conclusion Acknowledgements We thank Melanie Szoladatics for her assistance with movie editing and data collection. This work was supported by project number CS11-009 of the WWTF (Wiener Wissenschafts- und Technologiefonds; www.wwtf.at) to Ulrich Ansorge, Otmar Scherzer, and Shelley Buchinger. Manual response accuracy same position different position Rate of correct responses (%) 0 10 75 80 85 90 95 100 Manual RTs same position different position Mean reaction time (ms) 0 900 950 1000 1050 1100 continuity cuts discontinuous cuts Saccades to different position continuity cuts discontinuous cuts Mean saccadic reaction time (ms) 0 550 600 650 700 750 20 viewers participated in an eye tracking study. (EyeLink 1000 Desktop Mount, monocular recording at 1000 Hz) Materials We edited sports movies using either continuity cuts, or discontinuous cuts (324 cuts in total). Example cuts are depicted in Figure 1. Procedure and task In each of 20 blocks, two different sports movies were shown silently side by side on a computer screen (see ‘example block’ in Figure 2). Viewers were instructed to watch one of the movies while ignoring the other and keep their eyes on the movie they were watching. Crucially, the movies switched positions at some of the cuts, so that viewers had to saccade to the new position. Viewers reported by button presses whether the task-relevant movie continued at the same or at the different position. playback cut / different position task-relevant movie starts right cut / same position Example block * ns * ns * * p < .05 References [1] Itti, L., & Baldi, P. (2009). Bayesian surprise attracts human attention. Vision Research, 49, 1295-1306. [2] Smith, T.J. and Henderson, J.M. (2008). Edit Blindness: The relationship between attention and global change blindness in dynamic scenes. Journal of Eye Movement Research, 2(2):6, 1-17. [3] Smith, T. J., Levin, D. T. & Cutting, J. (2012) A Window on Reality: Perceiving Edited Moving Images. Current Directions in Psychological Science, 21, 101-106. Correspondence [email protected] http://attention.univie.ac.at Priming of visual attention by continuity editing in dynamic visual scenes Ulrich Ansorge 1,2 , Christian Valuch 1 & Peter König 2 1 Faculty of Psychology, University of Vienna 2 Department of Cognitive Science, University of Osnabrück viewer watches task-relevant movie task-relevant movie continues at same or different position Figure 1. Example continuity cuts and discontinuous cuts. Manipulation In the task-relevant movie each cut either followed continuity editing rules or not, and occurred at the same time as a cut in the irrelevant movie. Figure 2. Example block (each block lasted until the end of the movies). Figure 3. Results.

Author: others

Post on 15-Jul-2020

6 views

Category:

Documents


0 download

Embed Size (px)

TRANSCRIPT

  • Research suggests that attention in dynamic

    scenes is biased towards content that di"ers from

    what was observed before [1].

    While this may be prevalent in many

    situations, we propose that a preference for

    repeated information is as well observable.

    Professional footage contains cinematic cuts, with

    either continuity editing or discontinuous editing.

    Continuity cuts repeat major elements from a

    pre-cut scene in a post-cut scene. Although such

    cuts constitute abrupt changes, they are often not

    realized by viewers. In contrast, viewers are more

    frequently aware of discontinuous cuts [2, 3].

    We hypothesize that after cuts attention is

    captured by visual features repeated from the

    pre-cut scene.

    MethodsBackground

    Results

    Continuity had an e"ect on manual reaction

    times (RTs) and manual response accuracy.

    When the movies continued at the same posi-

    tion, discontinuous cuts were reliably detected.

    With continuity, the fraction of missed cuts

    nearly doubled. When movies continued at the

    di"erent position, performance did not di"er

    between the two types of cuts.

    However, saccadic reaction time (i.e., relocating

    gaze to the di"erent position of the task-

    relevant movie after the cut) was reduced by

    26 ms with continuity cuts as compared to

    discontinuous cuts.

    Our results demonstrate that the classical

    technique of continuity editing achieves the

    desired result of perceptual continuity and a

    surprisingly large number of cuts is missed. This

    e"ect is however not additive with a change of

    spatial location.

    We found di"erences in saccadic latency between

    continuity cuts and discontinuous cuts, which

    indicate that information from the periphery

    contributes to the decision process whether the

    movies switched positions or not.

    We propose that re-orientation of attention and

    gaze after abrupt scene changes, like with

    cinematic cuts, is primed by repeating visual

    content from the pre-cut scene.

    Conclusion

    Acknowledgements

    We thank Melanie Szoladatics for her assistance with movie editing and data collection.

    This work was supported by project number CS11-009 of the WWTF (Wiener Wissenschafts- und

    Technologiefonds; www.wwtf.at) to Ulrich Ansorge, Otmar Scherzer, and Shelley Buchinger.

    Manual response accuracy

    same position di"erent position

    Ra

    te o

    f co

    rre

    ct r

    esp

    on

    ses

    (%)

    0

    10

    75

    80

    85

    90

    95

    100

    Manual RTs

    same position di"erent position

    Me

    an

    re

    act

    ion

    tim

    e (

    ms)

    0

    900

    950

    1000

    1050

    1100

    continuity cuts

    discontinuous cuts

    Saccades to di"erent position

    continuity cuts discontinuous cuts

    Me

    an

    sa

    cca

    dic

    re

    act

    ion

    tim

    e (

    ms)

    0

    550

    600

    650

    700

    750

    20 viewers participated in an eye

    tracking study.

    (EyeLink 1000 Desktop Mount,

    monocular recording at 1000 Hz)

    Materials

    We edited sports movies using

    either continuity cuts,

    or discontinuous cuts

    (324 cuts in total). Example cuts

    are depicted in Figure 1.

    Procedure and task

    In each of 20 blocks, two

    di"erent sports movies were

    shown silently side by side on a

    computer screen (see ‘example

    block’ in Figure 2).

    Viewers were instructed to watch

    one of the movies while

    ignoring the other and keep their

    eyes on the movie they were

    watching.

    Crucially, the movies switched

    positions at some of the cuts, so

    that viewers had to saccade to

    the new position.

    Viewers reported by button

    presses whether the task-relevant

    movie continued at the same or

    at the di"erent position.

    playback

    cut / di"erent position

    task-relevant movie starts right

    cut / same position

    Example block

    * ns * ns

    *

    * p < .05

    References

    [1] Itti, L., & Baldi, P. (2009). Bayesian surprise attracts human attention. Vision Research, 49, 1295-1306.

    [2] Smith, T.J. and Henderson, J.M. (2008). Edit Blindness: The relationship between attention and global

    change blindness in dynamic scenes. Journal of Eye Movement Research, 2(2):6, 1-17.

    [3] Smith, T. J., Levin, D. T. & Cutting, J. (2012) A Window on Reality: Perceiving Edited Moving Images. Current

    Directions in Psychological Science, 21, 101-106.

    Correspondence

    [email protected]

    http://attention.univie.ac.at

    Priming of visual attention by continuity editing in dynamic visual scenesUlrich Ansorge1,2, Christian Valuch1 & Peter König2

    1 Faculty of Psychology, University of Vienna2 Department of Cognitive Science, University of Osnabrück

    viewer watches

    task-relevant movie

    task-relevant movie continues

    at same or di"erent position

    Figure 1. Example continuity cuts and discontinuous cuts.

    Manipulation

    In the task-relevant movie each cut either followed continuity

    editing rules or not, and occurred at the same time as a cut in

    the irrelevant movie.

    Figure 2. Example block (each block lasted until the end of the movies).

    Figure 3. Results.