Accompanying this, there's a feature by Laurel Hart in praise of Mel. Plus there's the regular "Sufficient Data" infographic by me and Ben Morris, this time detailing Mel's myriad screams.
A few people have asked about methodology and what these things actually involve. So...
Much as I wish it had all been my idea, editor Marcus Hearn proposed an infographic based on Mel's screams to tie in with the rest of the issue, and suggested I look at waveforms and decibels. After some time trying to dig into that myself, I called in the technical genius of Nimbos who ran a spectral analysis of Mel's first three screams from the closing moments of Part Nine of The Trial of a Time Lord.
In the graph above, the vertical Y axis shows frequency in Hertz and the colour represents volume in decibels. The first scream, circled on the left, begins at about 3,000 Hz and then drops to about 2,000 Hz - roughly moving from a C note to a B. More or less the same thing happens in the second scream, circled to the right of the first scream and in the centre of the image. A third scream, circled to the right of this, seems to be a partial repeat of the second scream, presumably to extend it and added in the edit.
Below this, also circled on the right of the graph, is the start of the closing theme. At 988 Hz, this is a B, so Bonnie pitched her screams an octave above this, starting at C and then descending to B to segue neatly into the theme music, as per the instruction from director Chris Clough.
In each scream, the volume peaks at about -28 dB, though obviously volume is dependent on the setting of your speakers. But how loud a scream sounds to us can also be affected by context, and note how much other sound - or orange - there is going on at the same time: incidental music, explosions, and yet the screams are distinct.
When we compare screams from episodes, the context can be different - for example, without incidental music or explosions. That means a scream might sound very different, and yet turns out to be at a similar pitch and volume. Other factors that might affect what we're measuring include how sound was originally recorded, how that's been adjusted for the Blu-ray releases we worked from, and how the sound may be compressed or otherwise affected in undertaking this kind of analysis.
Accepting the potential margins of error, we at least had a starting point for comparison with other episodes. This, for example, is the spectrum for Mel's screams while trapped in the Rani's bubble trap in Part Two of Time and the Rani:
It comprises more but shorter screams, and the context of music and sound effects is different, so this graph is very different from the last one. Yet the screams (outlined in the green box) are all at about 2,200 Hz, and many feature the same kind of pitch shift from C to B as we saw before. The context is different in these two episodes but Mel's screams are very similar.
In fact, across her tenure in Doctor Who, Mel's screams are impressively consistent. Unfortunately for me, that means the data is not sufficiently varied to make for an interesting infographic.
While Nimbos worked diligently on this spectral analysis, I watched every one of Mel's episodes, skipping forward to the scenes in which she appeared to note the time and duration of all screams. This raised some issues over the precise criteria for defining a scream. Screams often occur right at the end of an episode and so are included in the recap at the start of the next episode - should those be counted once or twice? There are occasions when Mel cries out warnings or shouts things like "Nooo!" - but do those count as screams? There are also instances when, because of music, sound effects or other activity on screen, I couldn't be certain if Mel was screaming or not. I could identify the start and end of some screams to the accuracy of a tenth of a second but others are clouded by surrounding noise and I needed a consistent approach - ultimately deciding to round up each scream to the nearest full second.
These factors then had to be relayed in the infographic, in as few words as possible.
Having gone through the episodes twice to ensure the data was right, I then had to come up with a way to present the result visually in an engaging way, led by a new and striking illustration. It's difficult to do this until you have the data in front of you and a sense of its scale, range, peaks and troughs.
I usually sketch some rough ideas before settling on something to include with my written-up brief for the illustrator, so that queries or objections can be raised before it gets laid out. Sometimes this takes a while - and the solution then springs to mind as I'm doing the washing up. But the way to convey this particular data set presented itself during my watch-through and I mocked up the following:
As Ben Morris worked on that, it became evident that showing both time of screams and duration made things too cluttered. That meant further revising the data set to trim down the text presented. Then assistant editor Peter Ware had to point out, with heroic tact, that I'd cut one small bit we needed. Eek!
The result seems to have gone down well. If you like this sort of thing, there's a new "Sufficient Data" infographic every four weeks in DWM. You can still buy Whographica, a book of Doctor Who infographics, by me, Steve O'Brien and Ben Morris. Steve and I also wrote a book of infographics about Buffy the Vampire Slayer.
Oh, and you might like to read my own interview with Bonnie Langford for DWM in 2014.
