cs.CVSep 22, 2026
SaveMeasuring the evolution of camera distance across a century of film
Organizations: School of Information, University of California, Berkeley, CA 94704, USA · Cinema Studies Program, Bowdoin College, Brunswick, ME 04011, USA
Abstract
The rise of computer vision and artificial intelligence has made possible new forms of large-scale computational measurement. We apply these techniques to a deep collection of 5,205 digitized films viewed in theaters between 1922-2025 (covering popular, prestigious, and independent movies) to trace the development of one of the most fundamental ways through which film communicates: by manipulating the space between the camera and its subject. This work finds abrupt changes with the rise of new technologies in sound and television, and allows us to shed an empirical light on gender disparity (women, despite having substantially less screentime than men, are disproportionately the subject of closer shots), and illustrate how animated films both inherit and break free from the norms of live-action filmmaking.
Figures & tables
Figure 1: Examples of distance categories.
Figure 2: Shot distances over time (popular films), along with 95% confidence intervals (resampling complete movies).
Figure 3: Gender rate, popular films; women have 20% (+5pp) more medium closeups than men ( under permutation test).
Figure 4: Distribution of shots in paired popular live-action and animated movies.
Figure S1: Total number of feature films over time. Source: IMDb.
Figure S2: Total number of digitized movies in our popular collection.
Figure S3: Total number of digitized movies in our prestige collection.
Figure S4: Total number of digitized movies in our American independent collection.
Figure S5: Fraction of movies that are sound, AFI.
Figure S6: TV penetration in U.S. households. Plotted from data in Statistical Abstract of the United States: 1980 , table 997 (Households With Television Sets: 1950 to 1980) [ 29 ] and Historical Statistics of the United States, Colonial Times to 1957 [ 48 ] .
| model | accuracy [95% CI] | within-one accuracy [95% CI] | training time (min) |
|---|---|---|---|
| SigLIP 2 (base 224px) | 0.6680 [0.6332, 0.7038] | 0.9497 [0.9381, 0.9619] | 25.3 |
| SigLIP 2 (base 384px) | 0.6893 [0.6599, 0.7223] | 0.9561 [0.9461, 0.9661] | 102.4 |
| SigLIP 2 (So400m 384px) | 0.6807 [0.6476, 0.7108] | 0.9382 [0.9237, 0.9514] | 447.9 |
| ConvNeXt-Large | 0.6400 [0.6050, 0.6750] | 0.9358 [0.9193, 0.9502] | 62.5 |
| Swin-B | 0.6516 [0.6161, 0.6857] | 0.9420 [0.9269, 0.9548] | 68.7 |
| ResNet-50 | 0.5984 [0.5629, 0.6339] | 0.9149 [0.8962, 0.9325] | 17.3 |
Table S1: Model comparison summary
| Genre | Medium close-up (%) | Movies |
|---|---|---|
| Sci-Fi | 25.3 (24.1, 26.5) | 292 |
| Thriller | 24.9 (23.9, 26.0) | 512 |
| Horror | 23.4 (22.1, 24.6) | 287 |
| Action | 22.9 (22.2, 23.6) | 1003 |
| Mystery | 22.0 (20.8, 23.2) | 347 |
| Documentary | 21.2 (19.0, 23.6) | 121 |
Table S2: Share of screen time at medium close-up by genre, with 95% bootstrap confidence intervals.
Figure S7: Prevalence (popular) within genres (Drama, Comedy, Adventure, Romance, Action).
Figure S8: Top 5 most common aspect ratios used in popular films.
Figure S9: Camera distance over time within each aspect ratio.
Figure S10: Prevalence of distance categories over time, popular movies.
Figure S11: Prevalence of distance categories over time, popular movies (detail).
Figure S12: Prevalence of distance categories over time, prestige movies.
Figure S13: Prevalence of distance categories over time, prestige movies (detail).
Figure S14: Raw vs. PPI corrected (middle categories).
Figure S15: Raw vs. PPI corrected (edge categories).
Figure S16: Prevalence of distance categories over time by collection (popular, prestige, American independent).
Figure S17: Difference in prevalence between shots focused on women vs. men over time, popular collection.
Figure S18: Difference in prevalence between shots focused on women vs. men over time, prestige collection.
Figure S19: Difference in prevalence between shots focused on women vs. men over time, all collections.
| Shot distance | M (%) | W (%) | W M (%) |
|---|---|---|---|
| Extreme close-up | 0.2 (0.1, 0.3) | 0.1 (0.1, 0.1) | 0.1 ( 0.2, 0.0) |
| Close-up | 0.7 (0.5, 0.9) | 0.8 (0.7, 1.1) | 0.2 ( 0.1, 0.4) |
| Medium close-up | 18.0 (16.2, 19.9) | 17.5 (15.6, 19.5) | 0.5 ( 2.3, 1.5) |
| Medium | 48.6 (46.9, 50.2) | 49.0 (47.1, 50.9) | 0.4 ( 1.5, 2.6) |
| Medium long | 9.0 (8.1, 9.9) | 10.3 (9.3, 11.4) | 1.2 (0.2, 2.4) |
| Long | 23.1 (20.9, 25.4) | 21.8 (19.9, 23.9) | 1.3 ( 3.6, 0.6) |
Table S3: Share of a character’s on-screen frames at each shot distance, for men (M) and women (W) among animated human characters in the year-matched animated/live-action sample, with 95% bootstrap confidence intervals (resampling at the level of movies).
| Shot distance | M (%) | W (%) | W M (%) |
|---|---|---|---|
| Extreme close-up | 0.3 (0.2, 0.5) | 0.2 (0.1, 0.3) | 0.1 ( 0.3, 0.1) |
| Close-up | 1.2 (0.9, 1.6) | 1.3 (0.8, 1.9) | 0.1 ( 0.4, 0.7) |
| Medium close-up | 16.0 (14.0, 17.9) | 16.2 (13.6, 18.7) | 0.2 ( 1.9, 2.4) |
| Medium | 32.0 (29.8, 34.3) | 32.5 (29.4, 35.9) | 0.5 ( 2.1, 3.7) |
| Medium long | 7.9 (6.7, 9.0) | 7.4 (6.3, 8.7) | 0.4 ( 1.8, 0.8) |
| Long | 41.5 (38.6, 44.9) | 40.8 (36.9, 45.0) | 0.8 ( 3.9, 2.5) |
Table S4: Share of a character’s on-screen frames at each shot distance, for men (M) and women (W) among animated non-human (animal or other) characters in the year-matched sample, with 95% bootstrap confidence intervals (resampling at the level of movies).
| Shot distance | M (%) | W (%) | W M (%) |
|---|---|---|---|
| Extreme close-up | 0.0 (0.0, 0.0) | 0.0 (0.0, 0.0) | 0.0 ( 0.0, 0.0) |
| Close-up | 0.3 (0.2, 0.3) | 0.4 (0.2, 0.6) | 0.1 ( 0.0, 0.3) |
| Medium close-up | 36.7 (34.4, 38.9) | 39.7 (37.3, 42.2) | 3.0 (1.7, 4.4) |
| Medium | 48.9 (47.1, 50.5) | 47.5 (45.6, 49.2) | 1.4 ( 2.6, 0.2) |
| Medium long | 8.4 (7.7, 9.2) | 7.9 (7.1, 8.8) | 0.6 ( 1.3, 0.2) |
| Long | 5.7 (5.2, 6.2) | 4.5 (4.0, 5.2) | 1.1 ( 1.7, 0.5) |
Table S5: Share of a character’s on-screen frames at each shot distance, for men (M) and women (W) among characters in the live-action half of the year-matched sample, with 95% bootstrap confidence intervals (resampling at the level of movies).
| Shot distance | M (%) | W (%) | W M (%) |
|---|---|---|---|
| Extreme close-up | 0.0 (0.0, 0.0) | 0.0 (0.0, 0.0) | 0.0 ( 0.0, 0.0) |
| Close-up | 0.2 (0.2, 0.3) | 0.3 (0.3, 0.3) | 0.0 (0.0, 0.1) |
| Medium close-up | 25.6 (25.0, 26.1) | 30.6 (30.0, 31.2) | 5.0 (4.7, 5.3) |
| Medium | 51.0 (50.6, 51.3) | 49.7 (49.2, 50.1) | 1.3 ( 1.6, 1.0) |
| Medium long | 14.6 (14.3, 14.9) | 12.4 (12.0, 12.7) | 2.2 ( 2.4, 1.9) |
| Long | 8.5 (8.3, 8.8) | 7.0 (6.7, 7.3) | 1.5 ( 1.7, 1.3) |
Table S6: Share of a character’s on-screen frames at each shot distance, for men (M) and women (W) among characters in popular films, across all years, with 95% bootstrap confidence intervals (resampling at the level of movies).
Figure S20: Paired test of animated (human, non-human) characters vs. live-action ones.