Organizing a comparative film study in StoryFolder

StoryFolder — 2026-09-13

Part of: How to prepare a university film class with scene-study materials

Last updated: 11 September 2026.

Write the research question first, then decide on the unit you'll count (shot, scene, or a recurring motif) and the rule for which films and scenes go in. Build one board per film, all with the same fields, and pilot those fields on one short scene before coding the rest. Keep what you saw and what you think it means in separate fields. Keep the bibliography in a reference manager, and write down every boundary decision you make. StoryFolder holds the shot-level evidence, and the paper, the citations and the video essay are made elsewhere.

That split follows the way university film research projects are staged. Lehigh's Reel American History project guidelines move students from "a scene-by-scene breakdown with time marks for easy reference to any part of the film", through reviews, historical and filmic context and comparison films, to works cited (Lehigh University). The shot evidence is one component among fourteen.

What is the research question and unit of analysis?

A question that the evidence could answer either way, and a unit small enough to count consistently.

"How does the director use close-ups?" is a topic. "Do reaction close-ups arrive earlier in the scene in the director's later films?" is a question, and it tells you the unit (the reaction close-up), the field you need (its position in the scene) and the comparison (early films against late).

StoryFolder's board works at shot level. Detection gives you shots, and Merge combines a run of shots into one section when your unit is larger (Splitting & merging). Decide the unit before you start. Merging halfway through a corpus makes the early and late films incomparable.

Which films and scenes belong in the corpus?

The ones your written inclusion rule lets in, and you need the rule before you choose. For example: every scene in which the two leads first speak alone, in each of the director's features from 1990 to 2010.

Record three things for each film, because a reader will ask:

• Edition. Theatrical cut, director's cut, restoration, or the streaming version. Name the one you watched, because a different cut is a different set of shots.
• Excerpt points. The in and out times of every scene you coded.
• Access. Where the copy came from and on what terms.

In StoryFolder these go on the video, not the shot. Create video-level fields (Film, Year, Edition, Source, Included because) under Notes & Metadata → Video Metadata, and each film's board carries its own provenance (Add metadata to a video). Library search reads those fields, so typing an edition name finds the boards coded from it.

How should the codebook be designed?

Fixed categories for anything you'll compare, free text for anything you'll quote, and a written definition for every category.

1. Draft the fields from your question. Only the fields the question needs. Every field is a decision you'll make hundreds of times.
2. Write a one-line definition for each value. When does MCU become CU? Is a slow push-in a "track" or a "zoom"? Your definitions are where your method lives, and they go in your methods section.
3. Pilot on one short scene. Code it, then look at every place you hesitated. Rewrite the definitions and add a value if you kept wanting one.
4. Freeze the codebook before coding the corpus. If you change a definition later, re-code what came before.

In StoryFolder, a fixed category is a Dropdown (one answer per shot) or a Tag (several). A count is a Number, and a note is Text (Field types). Fields belong to your account, so every board in the corpus automatically has the same codebook, which is the property a comparison needs. The board mechanics are covered in the scene-study preparation guide.

A codebook for the reaction close-up question might look like this:

Part of the codebook in Preferences, with the option values visible under each dropdown.

AI Autofill can make a first pass on visible categories such as Scale, and it's included on every plan (the free plan covers an account's first 100 shots). Treat its values as a second coder whose work you check. A Dropdown can only receive one of its own options, but the model has no access to your written definitions. It also leaves any field you've switched off alone, so keep Shot function and Interpretation switched off.

How are observations kept separate from interpretations?

In different fields, and with a record of doubt.

Observation holds what another researcher would see on the same frame. Interpretation holds what you think it does. When a later reader challenges a claim, you can show that the evidence stands even if they reject the reading.

Two more habits protect the study:

• Tick Uncertain whenever you hesitate over a code or a boundary. Report how many shots were uncertain. A filter on that field lists them for a second look.
• Keep the counterexamples. A shot that doesn't fit gets coded like any other. Filtering it out makes your pattern look cleaner than the film is.

Where do secondary sources and citations live?

In your reference manager and library databases. StoryFolder holds a pointer.

Put each source's citation key (the one Zotero, EndNote or your library tool gives it) into Source ref on the shots that source discusses. The bibliography, the PDFs and the citation formatting stay in the reference manager. StoryFolder doesn't format citations or store papers.

How do I compare patterns without overclaiming?

Report the sample, the boundary rules, the missing values and both the mean and the median, and describe the shots that go against the pattern.

Within one film, filtering does most of the work. Filter to Shot function is Reaction and the board shows only those shots, in order, with their thumbnails. The Analyze view's duration numbers (total shots, average, median, longest, shortest) are free and follow the active filter (Shot duration metrics). The per-field distribution charts and the duration-over-time chart are Pro.

Across films, each board is analysed on its own, so the comparison is assembled outside. On Pro, export each film's board as a spreadsheet: one row per shot with start, end, duration and a column per field. Stack them in one workbook with a Film column added, and pivot. That workbook is also your dataset appendix.

One board's export, cropped to four rows. Timings are the detected shots'; the Scale and Shot function values are examples.

On the numbers themselves:

• Mean and median shot length diverge. One long take pulls the average up a long way. Report both.
• Counts depend on boundaries. If you merged dissolves into one shot, say so, because another researcher's count will differ.
• Proportions beat raw counts across scenes of different length.
• Say what the sample can't show. Six scenes from four films is evidence about those scenes.

How does the evidence move into the final project?

Into three kinds of output, each made outside StoryFolder:

Video essay courses put the composing in the editor. Illinois's MACS 485 has students do "weekly parameter-based videographic exercises" in "non-linear editing software (Adobe Premiere Pro)" (Illinois). Middlebury's Videographic Film & Media Studies adds peer feedback, where students "provide more detailed feedback on each others' videos" (Middlebury). Peer review and assessment happen in the course's own tools.

StoryFolder's editorial export builds that timeline from whatever shots are currently filtered, so filter to the evidence shots and export. It's Pro. The timeline points at the one file you imported for that board, and the selects reel is free. Your field values don't travel as markers, so bring the PDF or spreadsheet alongside. Rights for the final essay are part of the course: Sussex's Video Essays module includes "how to approach issues of copyright law and fair use" (Sussex, P3099).

What goes in the limitations statement?

Four sentences a methods section needs: the corpus rule and its size, the boundary decisions, who coded and whether anything was double-coded, and how many shots were marked uncertain. Write it as you go. Rebuilding it from memory at the end is how corpus rules quietly change.

FAQ

What is a unit of analysis in film studies? The thing you count or compare, such as a shot, sequence, scene or motif. It has to be fixed before coding, because merging shots partway through makes early and late films incomparable.

How many films should a comparative study include? As many as your written inclusion rule admits. Report the number, because a claim about six scenes is evidence about those six scenes.

Should I report average or median shot length? Both. One long take can pull the mean far above the median, and StoryFolder's Analyze view shows both for free.

Can StoryFolder compare several films at once? Each board is analysed separately. On Pro, export each board's spreadsheet and combine them in one workbook to compare.

Can StoryFolder manage my citations? No. Keep the bibliography in a reference manager, and store each source's citation key in a text field on the shots it discusses.

Can I send my coded shots to Premiere or Resolve for a video essay? Yes, on Pro. Editorial export writes an EDL, Final Cut Pro 7 XML and FCPXML of the filtered shots against the file you imported. Your field values don't become markers.

Should AI Autofill code my corpus? Only as a first pass on visible categories like shot scale, and every value it writes should be checked. It never sees your written definitions, and it leaves fields switched off untouched.

← All articles