Shot length distributions in German cinema, 1929 to 1933
This post compares the shot median shot lengths and the dispersion of shot lengths in German films from 1929 to 1933, inclusive. The films are grouped by year, and can also be divided into the silent films of 1929 and the sound films of the other years.
Shot length data was collected from the Cinemetrics database for 67 films released from 1929 to 1933, inclusive.
As the distribution of shot lengths in a motion picture are typically asymmetric with a number of outliers, the median shot length is used as a robust measure of location because it is not dependent on an underlying probability distribution and has a high breakdown point. The estimator Qn is used as a robust measure of scale, and calculates the distance of each data point from every other . Qn has a breakdown point of 50% and a bounded influence function, and is therefore robust. As this estimator is not dependent upon an underlying probability distribution or a measure of location, it is appropriate for the asymmetric distributions typically encountered in the cinema. For details on how to calculate Qn see here.
Kruskal-Wallis analysis of variance (corrected for ties) was used as an omnibus test of the difference between the films grouped by year, at a significance level of 0.05. If this test returned a significant result Dunn’s post-hoc test (corrected for ties) was employed for the pairwise comparison of groups, using a critical z-value of 2.3263 at a significance level of p = 0.01.
Effect sizes of difference between groups were estimated using the Hodges-Lehmann median difference of pairwise comparisons (HLΔ), and this result is reported with a distribution free (Moses) confidence interval.
All calculations were performed using Microsoft Excel 2007.
The statistical data for each film is given in Tables 1 through 5. Shot length data for these films is presented in Figure 1 for the median shot lengths and Figure 2 for Qn.
For the median shot lengths, the results show that there is a statistically significant difference (KW-ANOVA: Hc = 14.0359, p = 0.0064). Group comparisons were carried out using a Dunn post-hoc test, which provided significant results for the silent 1929 films with the sound films in 1930 (Tc = 3.5482), 1931 (Tc = 2.4476), 1932 (Tc = 2.5739), and 1933 (Tc = 2.8444). There are no significant differences in the distribution of the median shot lengths for any other pairwise comparisons.
Turning to Qn, the same patterns we see for the median shot lengths are evident. There is a statistically significant difference (KW-ANOVA: Hc = 19.4967, p = 0.0006); and that this difference occurs in the pairwise comparisons between 1929 and 1930 (Tc = 4.1611), 1929 and 1931 (Tc = 2.9438), 1929 and 1932 (Tc = 2.9669), and 1929 and 1933 (Tc = 3.2416), while there are no significant differences for any other pairwise comparisons.
Table 1 Median shot length and Qn for German films released in 1929 (n = 12)
Table 2 Median shot length and Qn for German films released in 1930 (n = 11)
Table 3 Median shot length and Qn for German films released in 1931 (n = 14)
Table 4 Median shot length and Qn for German films released in 1932 (n = 17)
Table 5 Median shot length and Qn for German films released in 1933 (n = 13)
There is clearly a difference in the style of the silent films of 1929 (n = 12) when compared with the sound films from 1930 to 1933 (n = 55). The sample median of the median shot lengths for films released in 1929 is 3.8s (95% CI: 2.8, 4.7) with an interquartile range of 1.6s, and for Qn is 2.6s (95% CI: 1.6, 3.5) and IQR = 1.5s. The sample median of the median shot lengths for films released between 1930 and 1933 is 6.1s (95% CI: 5.5, 6.7) with IQR = 2.9, and for Qn is 5.5s (95% CI: 4.9, 6.1) and IQR = 2.4s. Dividing the sample into silent and sound films, the change in the median shot lengths is estimated to be an increase of HLΔ = 2.2s (95% CI: 1.0, 3.4) and the change in the dispersion of shot lengths is estimated to be an increase of HLΔ = 2.6s (95% CI: 1.5, 3.5). From these results we can say that the stylistic changes that occur in German cinema with the coming of sound is (1) a slowing in the rate at which films are cut and (2) an increase in the dispersion of shot lengths in German cinema. This difference can be clearly seen in the box plots of these samples in Figures 1 and 2.
Figure 1 The distribution of median shot lengths for films produced in Germany 1929 to 1933, inclusive
Figure 2 The distribution of Qn for films produced in Germany 1929 to 1933, inclusive
Comparing these results to earlier results posted on this blog for Hollywood and German cinema (see here and here), we can that the change in film style that occurred in Hollywood with the introduction of sound technology occur in Germany, only after they have already occurred in Hollywood.
Posted on September 23, 2010, in Cinemetrics, Film Analysis, Film History, Film Studies, Film Style, Film Technology, German Cinema, Silent cinema and tagged Cinemetrics, Film Analysis, Film History, Film Studies, Film Style, Film Technology, German Cinema, Silent cinema. Bookmark the permalink. 4 Comments.