Chapter 3: Hiding a Cut
3.4 Suppress attention
3.4.1 In the blink of an eye
188.8.131.52 The Cut = A Cinematic Blink
Almost forty years before O’Regan and colleagues were exploiting the potential of blinks for hiding visual disruptions, the film director John Huston was discussing exactly the same technique in relation to film.
“All the things we have laboriously learnt to do with film, were already part of the physiological and psychological experience of man before film was invented…. Let me make an experiment – maybe you will understand better what I mean. Move your eyes, quickly, from an object on one side of the room to an object on the other side. 36 Implicit change detection is usually indicated by an increased frequency of eye movements to the changed object.
84 In a film you would use a cut. Watch! There- you did exactly what I expected: in moving your head from one side of the room to the other, you briefly closed your eyes. Try it again, in the other direction. There! You see, you do it automatically. Once you know the distance between the two objects, you blink instinctively. That’s a cut….. In the same way, almost all devices of film have a physiological counter- part” (John Huston quoted in; Bachmann, 1965)
This quotation takes the idea of film as an analogue for attention (previously seen in section 3.2) and expands on it by adding the observation that our eyes blink during an eye movement. This observation is correct, blinks do frequently accompany eye movements but not always (Fogarty & Stern, 1989). The actual cause of the perceptual separation of a saccadic eye movement into two static views of the visual scene is saccadic suppression. This will be discussed in more detail in the next section (3.4.2).
However, the idea that a blink could be used to hide a cut struck a chord with some editors. The most significant of these was the editor of Apocalypse Now (Francis Ford Coppola, 1979) and The English Patient (Anthony Minghella, 1996), Walter Murch. In his book: “In The Blink of an Eye: A Perspective on Film Editing” (Murch, 2001) Murch outlines his intuitions about how the blink represents the juncture between film and our experience of reality. Murch’s thesis is not outlined in a systematic manner (this is not to be expected from such a discursive book on film editing) instead he recounts his intuitions about blinks and their relationship to film in an anecdotal fashion. However, by examining his discussion his main arguments can be extracted:
1. A blink functions as punctuation to cognitive events, i.e. thoughts (page 61-62). 2. Blink rate increases as cognitive activity increases (page 62).
3. A cut is the cinematic equivalent of a blink (page 62-63).
4. Therefore, the rate of cutting should match the rate of blinking were the viewer to experience the action in real-life (page 62).
85 5. We blink in synchrony with other people when we are sharing their thoughts
6. When all members of the audience are engaged with the film they will blink in unison (page 71).
7. If all of these factors are taken into consideration, blinks can be used to hide cuts (pages 59 and 69).
This is an interpretation of Murch’s ideas and it is not known whether he would agree with this interpretation or believe that these effects would occur across the majority of cuts. However, as a thought experiment it is useful to take these ideas as hypotheses and try to find some supporting evidence.
Hypotheses 1 and 2 suggest that blinks can be used as a measure of cognitive activity. Murch himself referred to the work of “Dr. John Stern of Washington University in St. Louis” to validate this claim and a look at Stern’s publications does provide supporting evidence (Fogarty & Stern, 1989; Fukuda, Stern, Brown, & Russon, 2005; Nesthus & Stern, 2002). Stern and colleagues have presented evidence that blink duration and frequency increase as the complexity of cognitive processing required to perform a task increases (Fogarty & Stern, 1989; Fukuda et al., 2005). This relationship is believed to occur only with tasks that do not involve the acquisition of visual information. When a task requires attention to be concentrated on external visual stimuli, such as reading, blink rate decreases significantly (Bentivoglio et al., 1997; Holland & Tarlow, 1975). By comparison, when engaged in a conversation speakers have a tendency to blink in between phrases and at the end of sentences (Holland & Tarlow, 1975). Whether this is due to the blink acting as punctuation to the speaker’s thought processes or as a form of social attention used to emphasise the speaker’s point (e.g. in combination with a gaze shift towards the listener) is not currently known. However, Stern does believe that changes in blink duration and frequency can be used as a measure of cognitive activity (Fogarty & Stern, 1989; Fukuda et al., 2005).
86 Murch’s third hypothesis equates cuts with the individual thoughts a viewer would have if they were observing the same action in real-life. This relates to the idea of film as an analogue for the viewer’s attention and thought processes. Murch references another quote from John Huston to summarise this idea:
“To me, the perfect film is as though it were unwinding behind your eyes, and your eyes were projecting it themselves, so that you were seeing what you wished to see. Film is like thought, it’s the closest to thought process of any art.” (John Huston quoted in Sweeney, 1973).
Whilst blinks do show a degree of dependency on cognitive activity they are not the most informative ocular activity. Pupil dilation, as previously discussed (see section 3.3.2) can be used as a direct measure of cognitive load (Hess & Polt, 1964) and saccadic eye movements provide a real-time measure of which visual information is being processed and how it is being related to other parts of the visual scene (Ballard, Hayhoe, Pook, & Rao, 1997). In fact, the most reliable association of blinks to cognitive activity is actually seen in their co-occurrence with saccadic eye movements. When viewers make large saccadic eye movements they are usually shifting from processing one part of the visual scene to processing another. The probability that a blink co-occurs with a saccade increases as the length of saccade increases (Fogarty & Stern, 1989). Therefore, Murch’s hypothesis that “A cut is the cinematic equivalent of a blink” (Murch, 2001; page 62-63) might actually express his intuition that a cut should represent a significant change in viewpoint, analogous to a large saccadic eye movement, and will therefore co-occur with a blink. The blink is a by-product of the change in viewpoint which serves as an external manifestation of the internal saccadic suppression. It is this saccadic suppression that is actually responsible for making the viewer blind to the changing scene during the saccade, not suppression due to the blink. If this blink-saccade combination is actually what Murch is referring to then every cut would have to coincide with a saccade if the viewer were to be blind to the cut. Whether this occurs will be discussed in the next section.
87 Murch’s fourth hypothesis builds on this idea that a blink is an index of cognitive activity and suggests that a scene’s cutting rate should reflect the blinking rate a viewer would have if they were to engage in the same activity in reality. This seems like a valid claim if the blinks Murch is referring to are those that occur during large saccades. If a viewer needed to make numerous large saccades whilst observing a scene then, if the filmmaker’ aim was to replicate this experience in the film they would also need to use the same number of shots to capture the scene. Each large saccade will probably be accompanied by a blink making it analogous to the cut between shots. This relationship between blinks and cuts is useful to an editor as it emphasises the viewer’s comprehension of the scene. If a shot is presented for less time than a viewer would choose to focus an analogous part of a real-world scene there is the chance that they will fail to fully comprehend the shot. Of course, a limited or skewed comprehension might be the intention of the editor but to accomplish this the editor must first understand how long the viewer needs to fully comprehend a shot. These time constraints are dependent on the ocular activity required by the shot (e.g. fixation or visual search) and the associated patterns of attention and suppression.
Murch’s fifth and sixth hypotheses claim that we blink in synchrony with another person when we are deeply engaged with them. The survival benefits of such behaviour are easy to imagine. If we are confronted with an aggressor we do not want to provide them with a period of time during which we are unable to respond to an attack even if this is only 150ms. If we timed our blinks to coincide with that of an aggressor we could ensure that our eyes received the cleansing they require whilst not letting down our guard. There may also be a social advantage to mirroring another person’s blinks. Our ability to automatically replicate other people’s facial expressions is well known (e.g. Miller, 2005) and it is believed to be an important tool for expressing intimacy and developing social relationships. Mirroring another person’s blinks might serve a similar purpose. However, a quick survey of the social intelligence and facial expression literature has produced no empirical evidence that such mirroring of blinks occurs. This could be because a targeted study has never
88 been performed meaning that Murch’s insight might actually exceed that of social psychologists. Future studies must address this question.
However, there is evidence that blinks do not coincide with edit points. It has previously been found that the breakpoints between visual events (or just after) are valid edit points and are identified as such by editors (see section 3.3). These breakpoints have been associated with the encoding of the old event in memory (Newtson & Enqguist, 1976) and the acquisition of information about the new event (Smith et al., in press; Whitwell, 2005). These changes in cognitive activity resemble those identified by Murch as being associated with blinks. If blinks do coincide with edit points as hypothesised by Murch there should be a noticeable increase in the frequency of blinks around the time of breakpoints. In an eye tracking study of event segmentation behaviour (carried out in conjunction with the author of this thesis), no such relationship was found (Smith et al., in press; Whitwell, 2005). Breakpoints were accompanied by a contraction of the pupils (indicating a decrease in cognitive load) following by an increase in saccade frequency (indicating visual search) but there was no significant increase in blink frequency or duration. If all viewers blinked at the same time (as proposed in Murch’s sixth hypothesis) this would show up as a clear increase in blink frequency.
However, the absence of blinks does not completely invalidate Murch’s hypotheses as the stimuli used for this study were composed so that the actor’s faces, expressions, and blinks were mostly indiscernible (Long Shots were used). If blink synchrony is a social behaviour then closer shots that place more emphasis on actors’ faces or involve the viewer more (e.g. dialogue sequences rather than action) might lead to an increase in blink synchrony. This possibility must be explored in future studies.
Murch’s final hypothesis ties all other hypotheses together and proposes that blinks can be used to hide cuts. This hypothesis can be assumed to be true based on the findings of Change Blindness. It is not assumed that viewers will fail to detect changes of all size, change blindness experiments usually change an individual
89 object where as a cut changes the entire visual scene, but the evidence does seem to indicate that even if the change is eventually detected it will not capture attention immediately after the blink. Therefore, even given the absence of current evidence that blinks do coincide with cuts, the potential for blinks to hide cuts does exist. If an editor tries to coincide cuts with blinks the chances of achieving an “invisible” cut will increase.