Art imititates life?
March 1, 2010 10:29 PM Subscribe
Bringing New Understanding to the Director’s Cut (NYT) Art imitates life? Neuroscientists studying vision have observed a 1/f distribution in the natural scenes we encounter everyday. A new study shows movies have a similar 1/f distribution of scene pacing as natural scenes we encounter in daily life.
“For two days straight, I went through the movie, ‘Spies Like Us,’ with Dan Aykroyd and Chevy Chase,” said Christine E. Nothelfer, who worked on the project as an undergraduate intern. “I went through it frame by frame, I knew where every single cut was.”
Aren't there ethical guidelines for neuroscience research?
posted by Blazecock Pileon at 11:36 PM on March 1, 2010 [11 favorites]
Aren't there ethical guidelines for neuroscience research?
posted by Blazecock Pileon at 11:36 PM on March 1, 2010 [11 favorites]
I've been looking at the literature on 'processing fluency' in experimental aesthetics- and this just fits right in there.
I think it helps to explain why there are certain films that just felt wrong to me-like Dogma, did anyone else get that?
posted by leibniz at 11:46 PM on March 1, 2010 [1 favorite]
I think it helps to explain why there are certain films that just felt wrong to me-like Dogma, did anyone else get that?
posted by leibniz at 11:46 PM on March 1, 2010 [1 favorite]
Aren't there ethical guidelines for neuroscience research?
Sure, for human subjects. Not for student interns.
Also, pink noise is way more relaxing than white noise
posted by demiurge at 11:56 PM on March 1, 2010
Sure, for human subjects. Not for student interns.
Also, pink noise is way more relaxing than white noise
posted by demiurge at 11:56 PM on March 1, 2010
That's cool. For my own thesis, I'm working on something I call the Caine/Hackmann Theory, which stipulates that at all times, 24 hrs. a day, one can find either a Michael Caine or Gene Hackmann film on television.
Or wait, is that just something I saw in a movie ....
posted by mannequito at 12:10 AM on March 2, 2010 [2 favorites]
Or wait, is that just something I saw in a movie ....
posted by mannequito at 12:10 AM on March 2, 2010 [2 favorites]
Good lord, this paper is insufferably smug.
In no way do we claim that there is any intention on the part of filmmakers to develop a 1/f film style, even if they knew what that might be.
Oh, those filmmakers! If only they knew!
Instead, we claim that, as explorations and crafting of film have proceeded for at least 70 years, film narrative has fallen naturally into 1/f shot structure as the myriad of other considerations in filmmaking have played against each other in shaping film form.
Yes, that's the ticket. But somebody's probably gonna ask what would cause that kind of evolution... Hmm...
Thus, we suggest that the mind can be "lost" (Kael, 1965) most easily in a temporal art form with that structure. That is, setting the actual narrative aside, perhaps being engrossed in a film is, in part, to allow its 1/f temporal structure to drive the mind exogenously.
If 1/f structure is more engrossing, then we'd expect films to evolve towards it. Films do seem to be moving in that general direction. Therefore, 1/f structure is more engrossing. That works, right? Shit, where'd I put that logic textbook?
I mean, I guess it's kind of interesting to look at the statistics for this kind of thing. But what's up with the overclaiming?
posted by equalpants at 12:19 AM on March 2, 2010
In no way do we claim that there is any intention on the part of filmmakers to develop a 1/f film style, even if they knew what that might be.
Oh, those filmmakers! If only they knew!
Instead, we claim that, as explorations and crafting of film have proceeded for at least 70 years, film narrative has fallen naturally into 1/f shot structure as the myriad of other considerations in filmmaking have played against each other in shaping film form.
Yes, that's the ticket. But somebody's probably gonna ask what would cause that kind of evolution... Hmm...
Thus, we suggest that the mind can be "lost" (Kael, 1965) most easily in a temporal art form with that structure. That is, setting the actual narrative aside, perhaps being engrossed in a film is, in part, to allow its 1/f temporal structure to drive the mind exogenously.
If 1/f structure is more engrossing, then we'd expect films to evolve towards it. Films do seem to be moving in that general direction. Therefore, 1/f structure is more engrossing. That works, right? Shit, where'd I put that logic textbook?
I mean, I guess it's kind of interesting to look at the statistics for this kind of thing. But what's up with the overclaiming?
posted by equalpants at 12:19 AM on March 2, 2010
let me see if im reading the NYT link correctly:
in movies there are fewer long camera scenes than there are short continuous shots?
as for the NYT graphic: im see a bunch of dots and then the 1/f line kind of nowhere near the middle 10% confidence band.
i think the real story here is how Dr. Cutting actually gets funding....
posted by dongolier at 12:27 AM on March 2, 2010
in movies there are fewer long camera scenes than there are short continuous shots?
as for the NYT graphic: im see a bunch of dots and then the 1/f line kind of nowhere near the middle 10% confidence band.
i think the real story here is how Dr. Cutting actually gets funding....
posted by dongolier at 12:27 AM on March 2, 2010
If 1/f structure is more engrossing, then we'd expect films to evolve towards it. Films do seem to be moving in that general direction. Therefore, 1/f structure is more engrossing. That works, right? Shit, where'd I put that logic textbook?
Uh, what do you think is wrong with that "logic"? Do you think it's reasonable to say that films are evolving towards a 1/f structure for no reason? It's certainly logically possible but not likely.
You would expect statisticians to analyze events using probability theory, rather then propositional logic.
posted by delmoi at 12:30 AM on March 2, 2010
Uh, what do you think is wrong with that "logic"? Do you think it's reasonable to say that films are evolving towards a 1/f structure for no reason? It's certainly logically possible but not likely.
You would expect statisticians to analyze events using probability theory, rather then propositional logic.
posted by delmoi at 12:30 AM on March 2, 2010
That said, there is a natural tenancy for some things to fall into a Zipf law distribution, of which 1/f is the most common form.
There was a big discovery that word frequencies were governed by a Zipf law curve, and people thought it had to do with some internal laws of the brain or something, but it turns out that totally random text will also have word frequencies that follow zipf's law.
So it's possible that the time frequencies of these cuts will naturally fall into a zipf's law curve without any actual reason.
posted by delmoi at 12:41 AM on March 2, 2010 [3 favorites]
There was a big discovery that word frequencies were governed by a Zipf law curve, and people thought it had to do with some internal laws of the brain or something, but it turns out that totally random text will also have word frequencies that follow zipf's law.
So it's possible that the time frequencies of these cuts will naturally fall into a zipf's law curve without any actual reason.
posted by delmoi at 12:41 AM on March 2, 2010 [3 favorites]
Uh, what do you think is wrong with that "logic"? Do you think it's reasonable to say that films are evolving towards a 1/f structure for no reason? It's certainly logically possible but not likely.
Well, here's what I'd say is missing from their suggestion:
1. consideration of other possible reasons (besides attention) why shot structure might approach 1/f (they say: "When discussing cognitive emissions of a 1/f signal, Gilden (2001) focused on memory and interference. Without denying their importance in this context, we choose to focus on attention.")
2. effects of other properties of films (besides shot structure) on attention
3. some kind of explanation or guess as to how 1/f structure on a feature-length time scale would produce an engrossing effect (they say: "How might 1/f shot patterns entrain attention over periods of 1 to 3 hr? Current theories of attention provide little guidance.")
4. evidence that people find films with closer-to-1/f structure more engrossing
posted by equalpants at 12:57 AM on March 2, 2010
Well, here's what I'd say is missing from their suggestion:
1. consideration of other possible reasons (besides attention) why shot structure might approach 1/f (they say: "When discussing cognitive emissions of a 1/f signal, Gilden (2001) focused on memory and interference. Without denying their importance in this context, we choose to focus on attention.")
2. effects of other properties of films (besides shot structure) on attention
3. some kind of explanation or guess as to how 1/f structure on a feature-length time scale would produce an engrossing effect (they say: "How might 1/f shot patterns entrain attention over periods of 1 to 3 hr? Current theories of attention provide little guidance.")
4. evidence that people find films with closer-to-1/f structure more engrossing
posted by equalpants at 12:57 AM on March 2, 2010
Yeah, I just realized that there could be other reasons for 1/f distributions in movies simply due to random distributions. Perhaps older movies lacked it due to limitations caused by shooting with small film rolls and editing by hand.
posted by delmoi at 1:56 AM on March 2, 2010
posted by delmoi at 1:56 AM on March 2, 2010
That's cool. For my own thesis, I'm working on something I call the Caine/Hackmann Theory, which stipulates that at all times, 24 hrs. a day, one can find either a Michael Caine or Gene Hackmann film on television.
You have misunderstood the data. Michael Caine, Gene Hackman, and Gérard Depardieu are all the same person, from three separate universes. Their films are how the three "vibrational states" communicate with each other. This is why you have never seen them together. The universe would have ended.
posted by GenjiandProust at 3:02 AM on March 2, 2010 [2 favorites]
You have misunderstood the data. Michael Caine, Gene Hackman, and Gérard Depardieu are all the same person, from three separate universes. Their films are how the three "vibrational states" communicate with each other. This is why you have never seen them together. The universe would have ended.
posted by GenjiandProust at 3:02 AM on March 2, 2010 [2 favorites]
Radiolab had an episode that covered this. Unfortunately their search function is even more broken than MeFi's and I can't find it. It was discovered by the editor of the film The Conversation (starring, oddly enough, Gene Hackman), Walter Murch.
posted by DU at 4:46 AM on March 2, 2010
posted by DU at 4:46 AM on March 2, 2010
second delmoi on zipf and power laws. If you're not careful, you can see uninformative power laws everywhere.
posted by honest knave at 4:48 AM on March 2, 2010
posted by honest knave at 4:48 AM on March 2, 2010
Might be interesting to compare two cuts of essentially the same material, one much 'pinker' than the other (is that possible?). At the moment I'm finding it hard to accept the implied conclusion that Rocky IV is way more engrossing than The Seven Year Itch.
I do sort of feel that the use of technical methods to render films almost infinitely engrossing has already been dealt with more than adequately in the seminal work of James O. Incandenza.
posted by Phanx at 5:20 AM on March 2, 2010 [2 favorites]
I do sort of feel that the use of technical methods to render films almost infinitely engrossing has already been dealt with more than adequately in the seminal work of James O. Incandenza.
posted by Phanx at 5:20 AM on March 2, 2010 [2 favorites]
Michael Caine and Gene Hackman: A Bridge Too Far. There was no other overlap between the three, other than the Academy Awards. My thesis calculated the likelihood of being able to watch Scrubs at any given time of day (the number is never lower than 99.994%), including the possibility of concurrently-aired episodes (max is 4, observed in the wild).
posted by AzraelBrown at 5:34 AM on March 2, 2010
posted by AzraelBrown at 5:34 AM on March 2, 2010
Radiolab: Why Do We Blink?
This week, we ask a question that we thought was a no-brainer: why do we blink? Film editor Walter Murch tells us about a strange discovery he made years ago while working on The Conversation – could something as small as a blink actually be the trick of his trade? We also talk to Japanese researchers Tamami Nakano and Shigeru Kitazawa about the experiment they conducted to understand how we see the world, when we choose not to, and why.
posted by (Arsenio) Hall and (Warren) Oates at 5:38 AM on March 2, 2010 [4 favorites]
This week, we ask a question that we thought was a no-brainer: why do we blink? Film editor Walter Murch tells us about a strange discovery he made years ago while working on The Conversation – could something as small as a blink actually be the trick of his trade? We also talk to Japanese researchers Tamami Nakano and Shigeru Kitazawa about the experiment they conducted to understand how we see the world, when we choose not to, and why.
posted by (Arsenio) Hall and (Warren) Oates at 5:38 AM on March 2, 2010 [4 favorites]
as for the NYT graphic: im see a bunch of dots and then the 1/f line kind of nowhere near the middle 10% confidence band.
Looks fine to me. Whether or not it falls in the 10% confidence band is beside the point. The data's noisy and that's fine. The trend definitely seems to be progressing upwards toward the 1/f line which is the point.
posted by tybeet at 5:41 AM on March 2, 2010
Looks fine to me. Whether or not it falls in the 10% confidence band is beside the point. The data's noisy and that's fine. The trend definitely seems to be progressing upwards toward the 1/f line which is the point.
posted by tybeet at 5:41 AM on March 2, 2010
AzraelBrown Michael Caine and Gene Hackman: A Bridge Too Far. There was no other overlap between the three, other than the Academy Awards.
If you study A Bridge Too Far carefully, you will notice that the director carefully places other actors (especially James Caan, Laurence Olivier, and Sean Connery, with occasional reliance on Denholm Elliott and Ryan O'Neal) to create a counter-vibration and maintain stability of the universe. The Academy Awards, of course, take place in a separate reality where the laws of nature (and logic) are different from our own. But I suspect you already knew this.
posted by GenjiandProust at 6:29 AM on March 2, 2010
If you study A Bridge Too Far carefully, you will notice that the director carefully places other actors (especially James Caan, Laurence Olivier, and Sean Connery, with occasional reliance on Denholm Elliott and Ryan O'Neal) to create a counter-vibration and maintain stability of the universe. The Academy Awards, of course, take place in a separate reality where the laws of nature (and logic) are different from our own. But I suspect you already knew this.
posted by GenjiandProust at 6:29 AM on March 2, 2010
For those who can't view it from the second link above, the full article is available online here. (Or as pdf.)
posted by wanderingstan at 6:57 AM on March 2, 2010
posted by wanderingstan at 6:57 AM on March 2, 2010
I suspect the Times chose to run this piece solely so they could use "pulsatile" instead of the equivalent and more recognized "pulsating."
Twice.
posted by Naberius at 7:10 AM on March 2, 2010
Twice.
posted by Naberius at 7:10 AM on March 2, 2010
They looked at only 150 "popular" Hollywood films since 1935? And this is evidence of trends in filmmaking? I'm not a statistician, so maybe so, but I'd be interested in reading a lot more about how a representative sample was chosen, which genres are represented, how closely the selections correspond with popular and/or critical acclaim, etc. At any rate, 150 seems like a pathetically small sample.
I'd think you could come up with some interesting data by looking at editorial decisions in, say, 150 action movies since 1935, or 150 musicals. But this seems way scattershot.
posted by Joey Bagels at 7:18 AM on March 2, 2010
I'd think you could come up with some interesting data by looking at editorial decisions in, say, 150 action movies since 1935, or 150 musicals. But this seems way scattershot.
posted by Joey Bagels at 7:18 AM on March 2, 2010
Radiolab: Why Do We Blink?
I think a better question (and also on the theme of the post) is why Radiolab is edited and mixed so idiosyncratically. I've never heard a public radio program with so much overlapping and jump-cut dialogue--my reaction is similar to was leibniz said earlier about Dogma--it's almost like the style drowns out the substance.
posted by kittyprecious at 7:34 AM on March 2, 2010 [1 favorite]
I think a better question (and also on the theme of the post) is why Radiolab is edited and mixed so idiosyncratically. I've never heard a public radio program with so much overlapping and jump-cut dialogue--my reaction is similar to was leibniz said earlier about Dogma--it's almost like the style drowns out the substance.
posted by kittyprecious at 7:34 AM on March 2, 2010 [1 favorite]
I think a better question (and also on the theme of the post) is why Radiolab is edited and mixed so idiosyncratically.
To make it more interesting to listen to. And to replicate natural conversation and idea-sharing. Obviously opinion differs, but I think Radiolab is the best radio show and/or podcast out there.
posted by (Arsenio) Hall and (Warren) Oates at 7:36 AM on March 2, 2010
To make it more interesting to listen to. And to replicate natural conversation and idea-sharing. Obviously opinion differs, but I think Radiolab is the best radio show and/or podcast out there.
posted by (Arsenio) Hall and (Warren) Oates at 7:36 AM on March 2, 2010
I can see your point, though, kittyprecious. I liken Radiolab to an Altman film, with all of the overlapping dialogue, and that pleases me, I get more out of it and I feel like the producers treat me like someone who can keep up with a more fast-paced show. But on the other hand, I don't like how Errol Morris cuts his films (excepting the first three) for the same reason you mention, and I can see how you'd find it annoying in a radio format.
posted by (Arsenio) Hall and (Warren) Oates at 7:42 AM on March 2, 2010 [1 favorite]
posted by (Arsenio) Hall and (Warren) Oates at 7:42 AM on March 2, 2010 [1 favorite]
GenjiandProust: If you study A Bridge Too Far carefully, you will notice that the director carefully places other actors (especially James Caan, Laurence Olivier, and Sean Connery, with occasional reliance on Denholm Elliott and Ryan O'Neal) to create a counter-vibration and maintain stability of the universe.
Dude, it blows my mind: I had heard that Schrödinger had dabbled in film, but I had no idea his theories had moved beyond the theoretical. Stabilising a single acting force in two states simultaneously was a mere child's dream not too long ago. It must have proven to difficult to include the z-dimensional Depardieu factor. I mean, just look at Depardieu; it's a miracle his waveform remains as stable as it does.
posted by AzraelBrown at 8:01 AM on March 2, 2010 [1 favorite]
Dude, it blows my mind: I had heard that Schrödinger had dabbled in film, but I had no idea his theories had moved beyond the theoretical. Stabilising a single acting force in two states simultaneously was a mere child's dream not too long ago. It must have proven to difficult to include the z-dimensional Depardieu factor. I mean, just look at Depardieu; it's a miracle his waveform remains as stable as it does.
posted by AzraelBrown at 8:01 AM on March 2, 2010 [1 favorite]
I want very much to like this, but even after reading the paper I'm completely confused what the research even shows. Roughly speaking, it sounds like they're doing a discrete Fourier transform of the (normalized) shot length as a function of shot order, and they showed that the distribution of frequencies follows a 1/f law. OK. They also did some analysis that showed a greater degree of serial correlation in shot-length for more recent movies. How are these two concepts related? Anyone?
My thought process reading this paper was like: "Ok, sure, yeah, sure, What the holy hell?!"
posted by albrecht at 9:05 AM on March 2, 2010
My thought process reading this paper was like: "Ok, sure, yeah, sure, What the holy hell?!"
posted by albrecht at 9:05 AM on March 2, 2010
That said, there is a natural tenancy for some things to fall into a Zipf law distribution, of which 1/f is the most common form.
There was a big discovery that word frequencies were governed by a Zipf law curve, and people thought it had to do with some internal laws of the brain or something, but it turns out that totally random text will also have word frequencies that follow zipf's law.
So it's possible that the time frequencies of these cuts will naturally fall into a zipf's law curve without any actual reason.
posted by delmoi at 2:41 AM on March 2
You're just not looking deep enough, maaaan! (hey, pass me that J)
posted by symbioid at 9:21 AM on March 2, 2010
There was a big discovery that word frequencies were governed by a Zipf law curve, and people thought it had to do with some internal laws of the brain or something, but it turns out that totally random text will also have word frequencies that follow zipf's law.
So it's possible that the time frequencies of these cuts will naturally fall into a zipf's law curve without any actual reason.
posted by delmoi at 2:41 AM on March 2
You're just not looking deep enough, maaaan! (hey, pass me that J)
posted by symbioid at 9:21 AM on March 2, 2010
Looks fine to me. Whether or not it falls in the 10% confidence band is beside the point. The data's noisy and that's fine. The trend definitely seems to be progressing upwards toward the 1/f line which is the point.
But the upward trend only starts around 1960, and was decreasing before that. So the data looks pretty close to random to me unless you think that from 1930 to 1960 directors were trying to drive audience attention away from their films.
My guess would be that there is a coincidental correspondence between the conventions of modern action movie shot editing and a 1/f distribution. Though I doubt anyone could prove it one way or the other.
posted by afu at 9:30 AM on March 2, 2010
But the upward trend only starts around 1960, and was decreasing before that. So the data looks pretty close to random to me unless you think that from 1930 to 1960 directors were trying to drive audience attention away from their films.
My guess would be that there is a coincidental correspondence between the conventions of modern action movie shot editing and a 1/f distribution. Though I doubt anyone could prove it one way or the other.
posted by afu at 9:30 AM on March 2, 2010
wait, there's a directors cut of "Spies Like Us"?? where can i find it?
posted by marienbad at 10:07 AM on March 2, 2010 [1 favorite]
posted by marienbad at 10:07 AM on March 2, 2010 [1 favorite]
I want to like and understand what's going on here, but I still don't understand what 1/f means. I'm a filmmaker and an editor, I know all about shot lengths and so on, but even reading the article and the wikipedia page, I don't understand what this 1/f thing is.
What are we talking about? I get that it has something to do with series of shots of similar lengths being cut together within scenes or sequences, but I don't understand what that has to do with this math function. Can someone explain it to me so that it makes sense to the non-mathematician?
I'm always trying to use things that perceptual psychology discovers in my own filmmaking, so if this is something useful, I'd love to utilize it, or at least experiment with it, but I don't understand the math.
posted by MythMaker at 10:18 AM on March 2, 2010
What are we talking about? I get that it has something to do with series of shots of similar lengths being cut together within scenes or sequences, but I don't understand what that has to do with this math function. Can someone explain it to me so that it makes sense to the non-mathematician?
I'm always trying to use things that perceptual psychology discovers in my own filmmaking, so if this is something useful, I'd love to utilize it, or at least experiment with it, but I don't understand the math.
posted by MythMaker at 10:18 AM on March 2, 2010
Have you tried this link yet? I found it to be the clearest explanation of what 1/f is in layman's terms. You'll have to scroll down a bit to see the pink noise/1/f bit.
posted by longdaysjourney at 11:44 AM on March 2, 2010 [1 favorite]
I actually like (some of) the editing that RadioLab does. What I like less is the time they spend on the "human interest" at the cost of the science. There are already plenty (PLENTY) of shows out there about a guy who realizes he should have spent more time with $WHOEVER. Just tell me the theorem he proved.
posted by DU at 11:54 AM on March 2, 2010
posted by DU at 11:54 AM on March 2, 2010
People love tautologies because tautologies are lovable.
posted by srboisvert at 12:50 PM on March 2, 2010 [2 favorites]
posted by srboisvert at 12:50 PM on March 2, 2010 [2 favorites]
people love tautologies or people do not love tautologies
posted by dongolier at 1:44 PM on March 2, 2010
posted by dongolier at 1:44 PM on March 2, 2010
Tautologies are tautologies.
posted by kittyprecious at 2:16 PM on March 2, 2010
posted by kittyprecious at 2:16 PM on March 2, 2010
I put a tautology in your tautology so that you can tautology your tautology.
posted by Blazecock Pileon at 5:10 PM on March 2, 2010
posted by Blazecock Pileon at 5:10 PM on March 2, 2010
Trying to wrap my head around this...
As I understand it now, and please throw a brick through my window if I've got this math confused, but. If I wanted to make a perfectly 1/f film that was 38.5 seconds long, I would make it with shots of these lengths:
1 shot 0.1 seconds long
2 shots 0.2 seconds long
3 shots 0.3 seconds long
...etc...
10 shots 1.0 second long
(the longest shot length is 1.0 second so that becomes the "1" in "1/f" and the "f" is the frequency of cuts, as in a .1 second shot has a frequency of 10 per second. so in this case a 10-per-second shot should occur 1/10th as often as the longest length occurs. ) The sequence also is important, in that this pattern should hold when analyzing a fraction of the film, not just the whole
So if this is correct, then this is what I make of it. These scientists are saying that our brains expect this kind of variety in stimulus. Perhaps this is a survival adaptation, perhaps raids from neighboring tribes come out to a 1/f pattern as well. In any case we look for something else to hold our attention if a film isn't meeting our brain's expectations.
So then going over this chart The 60s see a burst in experimentation as studios search for ways to reach boomers, and we see a low point in relation to the 1/f pattern. And then... Jaws! Hollywood discovers the blockbuster and dedicates itself to chasing them. There is even a kink in that chart exactly the year Jaws came out! A couple years later The Empire Strikes Back nails the 1/f ratio (too bad jaws is not on there). The behemoth crawls 1/f-wards.
May be utter nonsense, but it sure makes a good story in my head.
posted by thetruthisjustalie at 1:15 AM on March 3, 2010
As I understand it now, and please throw a brick through my window if I've got this math confused, but. If I wanted to make a perfectly 1/f film that was 38.5 seconds long, I would make it with shots of these lengths:
1 shot 0.1 seconds long
2 shots 0.2 seconds long
3 shots 0.3 seconds long
...etc...
10 shots 1.0 second long
(the longest shot length is 1.0 second so that becomes the "1" in "1/f" and the "f" is the frequency of cuts, as in a .1 second shot has a frequency of 10 per second. so in this case a 10-per-second shot should occur 1/10th as often as the longest length occurs. ) The sequence also is important, in that this pattern should hold when analyzing a fraction of the film, not just the whole
So if this is correct, then this is what I make of it. These scientists are saying that our brains expect this kind of variety in stimulus. Perhaps this is a survival adaptation, perhaps raids from neighboring tribes come out to a 1/f pattern as well. In any case we look for something else to hold our attention if a film isn't meeting our brain's expectations.
So then going over this chart The 60s see a burst in experimentation as studios search for ways to reach boomers, and we see a low point in relation to the 1/f pattern. And then... Jaws! Hollywood discovers the blockbuster and dedicates itself to chasing them. There is even a kink in that chart exactly the year Jaws came out! A couple years later The Empire Strikes Back nails the 1/f ratio (too bad jaws is not on there). The behemoth crawls 1/f-wards.
May be utter nonsense, but it sure makes a good story in my head.
posted by thetruthisjustalie at 1:15 AM on March 3, 2010
thetruthisjustalie: I think you have it backwards.
You'd have one 1 second shot and 10 0.1 second shots. But if you think about it the 1 one second shot and the 10 0.1 second shots take the same amount of time
Think about it this way. A movie has N frames. So in a 2hr movie you'd have 172800 frames. Now randomly partition those frames into 'shots'
Any randomly generated 1 second shot would have 240 frames. And any randomly generated set of 10 0.1 second shots would also have 240 frames. So 0.1 second shots would be 10 times more common then 1 second shots.
viola, 1/f distribution, caused by nothing at all.
posted by delmoi at 5:01 AM on March 3, 2010
You'd have one 1 second shot and 10 0.1 second shots. But if you think about it the 1 one second shot and the 10 0.1 second shots take the same amount of time
Think about it this way. A movie has N frames. So in a 2hr movie you'd have 172800 frames. Now randomly partition those frames into 'shots'
Any randomly generated 1 second shot would have 240 frames. And any randomly generated set of 10 0.1 second shots would also have 240 frames. So 0.1 second shots would be 10 times more common then 1 second shots.
viola, 1/f distribution, caused by nothing at all.
posted by delmoi at 5:01 AM on March 3, 2010
But the upward trend only starts around 1960, and was decreasing before that. So the data looks pretty close to random to me
Yes, but the consistent increasing trend since then far outweighs the decrease you mention. You can see this more clearly if you draw a line through the data, like so, so it's clearly not random fluctuation.
posted by tybeet at 5:13 AM on March 3, 2010
Yes, but the consistent increasing trend since then far outweighs the decrease you mention. You can see this more clearly if you draw a line through the data, like so, so it's clearly not random fluctuation.
posted by tybeet at 5:13 AM on March 3, 2010
And any randomly generated set of 10 0.1 second shots would also have 240 frames. So 0.1 second shots would be 10 times more common then 1 second shots.
viola, 1/f distribution, caused by nothing at all.
I don't think that's what meant here by "1/f distribution." As I understand it, the analysis was not on the distributions of shot-lengths, but rather on the power distribution in the discrete Fourier transform of the sequence of shot-lengths. That is, they're considering not only the lengths of shots, but how those lengths are organized in a series. If there were no serial correlation between the lengths of shots from one to the next, then the "signal" would look something like white noise, as in the link that longdaysjourney posted. If there were very strong correlation, it would look like "brown noise." Instead, as they argue, it's somewhere in between, and it's moving closer to "pink noise" in recent years. The "1/f distribution" here refers to the fact that pink noise has power at each frequency that's proportional to the reciprocal of the frequency. The wikipedia page on pink noise also has a fairly good explanation.
I guess what's interesting about this, potentially, is that this kind of noise has been observed in other areas of human cognition, so you can make a vague hand-wavy argument about how Hollywood is converging on some kind of optimal editing formula. Or, as my friend put it, "Result? Press coverage, because you have a paper about math, movies, and the brain. Only not really." It seems much more likely to me that any observed variation in movie editing structure over time can be explained by changes in technology, larger budgets, and the growing popularity of action as a genre, none of which were controlled for in the study.
posted by albrecht at 8:26 AM on March 3, 2010
viola, 1/f distribution, caused by nothing at all.
I don't think that's what meant here by "1/f distribution." As I understand it, the analysis was not on the distributions of shot-lengths, but rather on the power distribution in the discrete Fourier transform of the sequence of shot-lengths. That is, they're considering not only the lengths of shots, but how those lengths are organized in a series. If there were no serial correlation between the lengths of shots from one to the next, then the "signal" would look something like white noise, as in the link that longdaysjourney posted. If there were very strong correlation, it would look like "brown noise." Instead, as they argue, it's somewhere in between, and it's moving closer to "pink noise" in recent years. The "1/f distribution" here refers to the fact that pink noise has power at each frequency that's proportional to the reciprocal of the frequency. The wikipedia page on pink noise also has a fairly good explanation.
I guess what's interesting about this, potentially, is that this kind of noise has been observed in other areas of human cognition, so you can make a vague hand-wavy argument about how Hollywood is converging on some kind of optimal editing formula. Or, as my friend put it, "Result? Press coverage, because you have a paper about math, movies, and the brain. Only not really." It seems much more likely to me that any observed variation in movie editing structure over time can be explained by changes in technology, larger budgets, and the growing popularity of action as a genre, none of which were controlled for in the study.
posted by albrecht at 8:26 AM on March 3, 2010
Here's a theory why some films have a higher correlation to pink noise.
You can simulate pink noise by beginning with an initial low frequency waveform of random amplitude and then adding progressively smaller random amplitudes of higher frequencies on top of it. This will give you a downward slope from left to right on a spectral graph. http://en.wikipedia.org/wiki/Colors_of_noise
So... Let's say the editor has edited the first pass of a film. It's a basic edit, not too much action, long scenes to make it dramatic and artsy. Now the director sees it. Wants a little bit more action. The producers see it. A little bit more action, please. A few more tweaks. This all translates into smaller amounts of higher frequency noise just by having multiple people express their opinions on it.
In the last 20 years, this expression of opinion has become easier and cheaper to accomplish with the advent of computer-based editing tools. People take it for granted now.
So I posit: The more editorial passes that are taken on film from different people, the more the film's cuts will approach 1/f noise.
QED.
posted by hanoixan at 9:04 AM on March 3, 2010
You can simulate pink noise by beginning with an initial low frequency waveform of random amplitude and then adding progressively smaller random amplitudes of higher frequencies on top of it. This will give you a downward slope from left to right on a spectral graph. http://en.wikipedia.org/wiki/Colors_of_noise
So... Let's say the editor has edited the first pass of a film. It's a basic edit, not too much action, long scenes to make it dramatic and artsy. Now the director sees it. Wants a little bit more action. The producers see it. A little bit more action, please. A few more tweaks. This all translates into smaller amounts of higher frequency noise just by having multiple people express their opinions on it.
In the last 20 years, this expression of opinion has become easier and cheaper to accomplish with the advent of computer-based editing tools. People take it for granted now.
So I posit: The more editorial passes that are taken on film from different people, the more the film's cuts will approach 1/f noise.
QED.
posted by hanoixan at 9:04 AM on March 3, 2010
I hadn't noticed the NYT link to the actual paper [pdf] that James Cutting published, which confirms that I totally don't get the math. They did analyze from both local and global perspectives though. And a film that is 1/f should hold people's attention as far as the cut pacing goes.
posted by thetruthisjustalie at 11:50 AM on March 3, 2010
posted by thetruthisjustalie at 11:50 AM on March 3, 2010
« Older Woodworks Library | Return of the Fungi Newer »
This thread has been archived and is closed to new comments
Geisler et al. (2001)[15] took this a step further by exploring the inferential power of edge ensembles with respect to predicting the path of contours in natural images. By analyzing the co-occurrence statistics of pairs of edges, not only did Geisler et al. confirm the observations of Sigman et al. about cocircular contours, but they precisely quantified the inferential power of the related across-domain statistics.
Alrighty, then.
posted by empath at 11:31 PM on March 1, 2010 [3 favorites]