How to read versions of a Kubrick screenplay

Not every text called a screenplay describes the same stage. Some set out what will be filmed; others revise a proposal or record dialogue that is already part of a film being prepared for distribution.

Identify the document before comparing its words

Reading screenplay versions means starting with their material identity. The film’s title is not enough: you need the date, provenance, stated authorship and relationship to other documents. Two copies may contain the same text but have different annotations. A bound version is not necessarily the latest, and a loose sheet may belong to a localised revision, not a complete screenplay.

The Kubrick Archive record SK/10/1/21 describes a Lolita screenplay dated 10 May 1960 and attributed to Vladimir Nabokov. It distinguishes a bound copy from an unbound one with deletions. This is a description of materials, not a transcription of their differences. It supports the claim that physical variants exist; it does not allow us to reconstruct the content of the deletions without examining the pages.

The example shows why the word complete needs context. A text may be complete as a writing proposal without matching the finished film. The document’s completeness does not mean that all its scenes were shot or that they kept the same place. The film adaptation of Lolita requires us to distinguish the written work from the audiovisual result, so that one version is not treated as the sole measure of fidelity.

The collaboration with screenwriters introduces another note of caution. The final writing credit and the authorship of an annotation do not answer the same question. Someone may be credited for the screenplay without having written every change visible in a copy. Conversely, a handwritten contribution does not prove that its author was solely responsible for the entire scene.

Partial revisions and pages that should not be assembled arbitrarily

Another Lolita record, SK/10/1/8, catalogues typed inserts, probably prepared during production, with changes in Kubrick’s hand on some pages. The word probably is part of the scope of the description and must be retained. Replacing it with an exact filming date would add a certainty that the record does not offer.

These inserts raise a different problem from that of a complete draft. To understand their function, we would need to establish which pages they replace, whether they alter the numbering or continuity, and which copy they should be linked to. It is not correct to attach any loose fragment to any screenplay for the same film and present the result as a historical version that actually circulated among the participants.

The sequence of revisions cannot be inferred from the number of corrections alone, either. A clean copy may postdate an annotated one because it incorporates its changes; it may also be a parallel copy intended for another purpose. The text, dates and provenance need to be compared. The appearance of greater polish does not, by itself, establish a chronology.

The written development of A.I. is a reminder that a project can pass through different authors and stages before its final realisation. Comparing proposals requires keeping those attributions intact, rather than constructing a supposed single vision of the director from fragments produced at different times. The project’s name remaining the same does not erase the differences between its materials.

The archival description helps locate these distinctions, but it has limits. A record may group several items without detailing every change. The analysis should state when it relies on the catalogue and when on direct examination of the original. Claiming that a particular person crossed out a phrase requires more specific evidence than the mere presence of their name in the collection.

The text produced after filming is not a promise of what was going to be filmed

Eyes Wide Shut offers an example of documentation classified as post-production material: series SK/17/4/4 includes annotated transcripts for editing, records of dialogue timings, and a combined list of dialogue, continuity and subtitles. The record describes tools related to material that had already been worked on. They should not automatically be read as instructions that preceded the filming of each scene.

The temporal direction matters. A preparation screenplay can help us ask what had been planned; a transcript can help identify what was said in a working version. The fact that both contain similar words does not prove they are interchangeable. To study a change, we need to know whether we are comparing a proposal with a result, or two later records of different results.

The continuity and shot-identification documents can establish connections, but they do not justify skipping the film itself. A written line may not have been performed, may have been filmed in several ways, or may have been left out of the edit. The correspondence must be checked against identified scenes and versions, without treating the absence of a line as automatic proof of a reason for removing it.

Dialogue and staging also need to be separated. A film may preserve the words but change their effect through pauses, framing, performance or sound. Textual comparison detects certain differences, but does not exhaust the screenplay’s film adaptation. An analysis that only lists added and deleted lines overlooks decisions that transform the same conversation without changing its verbal content.

The discarded material ultimately requires precise vocabulary. A written alternative is not a filmed scene, and a filmed scene is not necessarily part of an exhibited edit. Screenplay versions let us trace possibilities and revisions when their relationships are known. On their own, they do not offer a complete alternative film or a certain explanation of every change between the writing and the release.

Sources: Stanley Kubrick Archive, University of the Arts London, SK/10/1/21, Lolita, 10 May 1960; SK/10/1/8, Script inserts, circa 1960; SK/17/4/4, Dialogue and Film Editing, 1998–1999. Analysis of catalogue descriptions, without attributing readings of unconsulted manuscripts.

View all articles in Kubrick’s working method, documents and archive