Roma (2018) Review
9/10
Water across a tiled floor. A reflection of the sky in it. A woman washing a courtyard in a Mexico City house in 1970, who will do this again tomorrow, and whose employers are about to stop being a family.
The 9 is for a film built almost entirely out of camera movement and duration, in which the most significant events happen at the edges of frames, and which is about somebody the form normally puts in the background.
What keeps it off a 10 is a slight imbalance in the last act, where two large set pieces arrive close together and the film’s method briefly becomes spectacle.
The Camera Does Not Take Sides
Almost every shot is either a slow lateral pan or a static wide, and the camera holds at a distance that includes the whole room.
That decision is the film’s ethics. A close-up selects and directs, and this film mostly declines to, so a viewer watching Cleo work is also watching the family, the dog, the street outside and whatever is happening in the next room.
The panning is specific. The camera moves at a constant speed regardless of what it passes, which means a conversation and a wall receive identical attention, and the audience has to decide what matters.
Cuaron shot it himself after Emmanuel Lubezki was unavailable, in 65mm digital, and the resolution is part of the method. A frame this detailed rewards looking at the background, which is where the film keeps putting things.
The 65mm digital capture also means the film holds up at a size nobody watched it at. It was released primarily on a streaming service to televisions and laptops, and the resolution was specified for a theatrical presentation most of its audience never saw.
Black and white is the other formal decision and it is doing a specific job. The period is 1970, the memory is Cuaron’s own, and monochrome removes the documentary specificity of color while keeping the detail, which places the film at one remove from the record.
A camera that refuses to select forces the reader to. In prose the equivalent is describing a room without ranking its contents, and letting the significant detail sit in a list with everything else. The cost is that some readers will miss things. That is the trade and the film accepts it.
Constant-speed movement equalizes attention. Anything that varies its emphasis is telling the reader what to care about.
Put your important events at the edge. A reader who finds something themselves holds it differently from one who was shown it.
Yalitza Aparicio
She was a trainee teacher with no acting experience, was cast from an open call, and carries the film in two languages.
The performance is almost entirely watchful. Cleo listens, waits, works, and registers things without responding to them, which is both the correct characterization for a domestic employee and the hardest thing to ask of somebody untrained.
Aparicio also speaks Mixtec in the film and the language is not translated in several scenes. Cleo and another worker talk to each other in a house where nobody else understands them, which is a piece of privacy the film gives her and the employers never notice.
She was also cast from an open call in Tlaxiaco and had not intended to act at all, having qualified as a teacher, and the film’s central performance belongs to somebody who was doing something else entirely the year before.
What she does that a trained actor might not have is refuse to perform emotion at the available moments. The scene where she tells Fermin she is pregnant is delivered flatly, in a cinema, and her face does not change.
The hospital sequence is the most difficult material in the film and is filmed in a single long take with real medical staff, and Aparicio plays it without any of the escalation the sequence invites.
Sound
The film’s most technically ambitious element is not the image, it is the sound design, which is a full spatial mix used with unusual discipline.
Everything is placed. Street vendors, dogs, a radio in another room, a train, and the mix establishes the geography of a house and a city continuously, so an audience always knows what is happening outside the frame.
That is doing the same job the camera is refusing to do. The image declines to direct attention and the sound tells you the world extends past the edges, and together they produce a scene the viewer has to navigate.
The Corpus Christi sequence is where it is most effective. Gunfire and shouting arrive from off screen, the camera stays inside a shop, and the audience assembles an event from sound while looking at furniture.
The beach sequence is the other one and it works differently. A long lateral take, two children in the water, and a woman who cannot swim walking in after them, and the camera stays at a distance and does not cut for the duration of it.
The dog is the film’s most repeated detail and it is doing real work. There is dog mess in the driveway in almost every shot of it, somebody is always about to clean it, and nobody ever does, which is the household’s condition in a single recurring object.
Sound establishes what a frame excludes. In prose, the equivalent is reporting what a character hears from elsewhere, which extends a scene beyond the room at almost no cost. A radio in another room, a conversation upstairs, traffic. These are the cheapest worldbuilding available and most writers use none of them.
The knife-sharpener’s whistle recurs across the film and is the clearest example. It arrives from off screen at intervals, means nothing to the plot, and establishes a city that continues at its own pace regardless of what is happening in the house.
Deliver a large event from off-stage while your scene continues. The reader builds it and the contrast between the small foreground and the large background is the effect.
Let the two channels do opposite jobs. The image narrows and the sound widens, or the reverse, and the reader has to hold both.
The Politics Are at the Edges
The film covers the period of the Corpus Christi massacre of 1971, in which a student demonstration was attacked by a state-backed paramilitary group, and it is in the film for about four minutes.
That proportion is the argument. Cleo is buying a crib when it happens, the event arrives through a window, and one of the men involved turns out to be somebody she knows.
The film’s position is that history happens next to people who are doing something else. Nobody in it discusses politics, nobody has a position, and the state violence is a thing that occurs while a woman is shopping.
Class is handled the same way. The employers are kind, are genuinely attached to Cleo, and the film establishes without comment that she sleeps in a small room, eats separately, and is told she is family by people who pay her.
The garage is the sharpest single image of it. The father’s car barely fits, the manoeuvring takes a full minute of screen time, and the film films a man’s status symbol as a daily inconvenience for the household that has to work around it.
Marina de Tavira plays the employer and was nominated, and the part is written with real difficulty. Sofia is being abandoned by her husband, is frightened, takes it out on Cleo, apologizes, and does it again, and the film treats all of that as ordinary.
Green Book
Best Picture went to Green Book. This film won Director, Cinematography and Foreign Language, and became the first film from a streaming service to be a serious Best Picture contender.
That last fact is most of the explanation. A substantial portion of the membership was hostile to Netflix in 2019, the theatrical release was limited by design, and a vote against the film was available as a vote about distribution.
Cuaron has also said he shot the film in the order of the events and largely without giving the cast the script, which is an approach that produces the reactions the film relies on and is only available to somebody with his standing.
The comparison itself is stark. Green Book is a film in which a white driver’s education is the arc and a Black musician is the instrument of it, released the same year as a film about a domestic worker told from her position, and the Academy chose the first.
That pairing is the third instance of the same pattern on this page, after Driving Miss Daisy over Do the Right Thing and Crash over Brokeback Mountain, and the consistency across thirty years is the point of the list.
When an institution makes the same choice three times across thirty years, it is a preference instead of an accident. The comfortable version of a difficult subject wins. That is worth knowing before you decide which version to write, and it is not an argument for writing the comfortable one.
It is also the only entry of the three where the losing film was not in English, which is a fourth variable and a real one. No film has ever won Best Picture in a language other than English except Parasite, which managed it the very next year the following year.
Novelty in distribution can cost you on merit. A vote about the platform is indistinguishable from a vote about the work and both are counted the same way.
A subject told from the position of the person it happens to is harder to sell than the same subject told from the position of somebody learning about it.
Should you watch Roma?
Yes, and watch it with proper sound. A camera that pans at constant speed past everything and refuses to select, a sound mix that puts a massacre outside a shop window, a non-professional lead who declines to perform emotion, and a film about the woman the form usually keeps in the background.
Frequently Asked Questions
This film appears on one of my film lists