Film Orchestration - Part 6: String Section 2 - Divisi

(this is an updated and extended post that I published in 2020 on this Patreon)

After talking about general string line-up questions in the last part, we will have a closer look at specific properties of the string section. They are the only instruments in the orchestra where by default several players share the same part. For people who come from a DAW background and who are transitioning into the "real orchestra" world there are a few things that need to be taken care of which normally doesn't play a huge role when you work with samples.

One of these things that really do work considerably different with real players than with samples is divisi. Generally speaking, divisi means to split one or several sections into several voices. The most striking reason to do something like this would be the need to have more notes played at the same time than you would get with every section playing unison. But there are also timbral reasons for divisi which we will cover later.

The execution of a divisi usually is done "per stand". Usually, two string players of the same group share one music stand. Dividing the section in two means that the outer player would play the higher note and the inner player would play the lower note. Splits into three or more sections are usually handled individually by the section and are being decided by the section leader. Also, if as an orchestrator you know the exact string line-up you can specify the exact divisi. Generally, you would try to even out the numbers on all voices as much as possible but there might be instances where for example one line carries a more important melody than the other where you might want to put more players on the more important line. In such cases indications in the score use specific numbers (e.g. 6 players/4 players) or desks (3 desks, 2 desks). More specifics on this a little later on.

While the specific technical execution of a divisi often doesn't have a large influence on the sound, the actual divisi itself does and has some striking implications on how you should write.

The traditional way to indicate a divisi on the score sheet would be to write "divisi" or "div." at the beginning of the split and sometimes "unis."or "unison" when the section joins back together (and repeat accordingly on every new divisi). The reasoning is that otherwise players might attempt to play the two voices as double stops (which often create intonation issues and don't really help to keep up the substance). In recent times, and especially in the session world, the default would be to play any split as divisi unless marked otherwise. For a few years now I have stopped marking divisi in my score sheets and haven't had a single question or problem with that, even with traditional concert orchestras. So, while it is technically not perfectly correct, it really helps to reduce unneccesary clutter on the score and parts.

The general goal should be to only divide when really necessary. A divisi always means that you only have at max half of the section per note than with a unison.

Interestingly, this has no big influence on the volume (of course it will decrease but not by half) but more on the substance of sound. Especially when your string section is quite small to begin with, every divisi should have a striking reason.

The sound of a string section lives from the many players that create that ensemble sound which only at quite high numbers create the substance in sound that we expect from an orchestral string section. When writing for samples, of course we usually don't have that loss of substance when splitting the section so you might not be aware of the consequences of a divisi.

Here are a few things to pay attention to when writing divisi:

1. In general, the higher you go with violins, the more instruments you need in order to sustain the substance. In general, in spite of violins usually being present at the highest numbers in a string sections they are the most sensitive regarding divisi. Splitting a small violin section of for instance only 8 players into two groups of 4 each really high up in their register will create a very thin chamber sound but not an orchestral sound. But even exposed splits of larger sections will lead to a considerable decrease in substance. So the general strategy if you need to split violins would be to do it on the second violins and leave the first violins together to provide enough substance on the highest line. In smaller orchestras, this sometimes might even lead to the decision to double the first violins at pitch with the second violins in cases where you really need substance in sound. (While in larger line-ups you would rather go for something like doubling the first violins an octave lower in the seconds). On the other hand, it is quite common to split the celli, even though you might only have 6 or 4 of them. However, they are quite capable of providing enough volume and their diminishing substance will be hidden by the higher sections on top. As a little side story, John Williams score for ET called for 20 first violins, which even for a large symphony orchestra is quite a lot. But of course the main theme soaring at the highest registers of the violins needed enough players to provide the substance and scale that was needed.

2. Dynamics have a strong influence on how a divisi sounds. Splitting string sections into several divisis at high dynamic degrees will very likely cause the section to sound quite thin and lose substance. However, this problem is much less prominent at low dynamics. Even quite small string sections will create a satisfying divisi at low dynamics without sounding overly thin. So a decision about a divisi should also be taken based upon the dynamics.

3. When dividing sections you will get a better result if you divide them in simpler intervals. The more tricky to intonate the note you're playing is based on what's going on around you as a string player, the less secure the execution of the pitch will be. If your desk neighbor is playing a perfect fifth higher than you it is considerably easier for you to intonate compared to them playing something like a minor second or another highly dissonant interval with you. So if you need to involve complex harmony or dissonant intervals into your string divisi, try to place these intervals not within the same sections but rather between different sections.

4. A divisi can be a timbral decision to create a really fragile string sound. This of course works much better at softer dynamics. A great example is Venus by Gustav Holst where the first and second violins are split into 4 notes each. But as you can see, this decision is executed very consequential here by exposing that divisi sound and really splitting into many voices, leaving no doubt that the sound is desired and not "by accident".

Having said that, there are several special cases for divisi of which I want to show you a few to put the points above into perspective.

First let's have a look at this section from the Jurassic Park Main Theme.

The strings for this passage look like this in the score sheet:

Notice that the violins and violas are split into two voices each. The first violins are split into octaves enclosing the second violins. The Violas fill up harmony in the mid register. In this case, the high strings provide a harmonic backdrop of sustaining chords while the horns play the theme. They provide high sustained harmonic frequencies that almost feel like an organ. So in spite of the relatively high dynamics and the high register, the divisi work well as they don't need substance for the task that they're assigned for. The enclosing of the seconds in the first might be to create a more homogenous sound but I would argue that superimposing would probably not make a huge difference in sound.

A similar example that leads this concept even further is Aase's Death by Grieg. Particularly at the linked time you can see that all sections but the basses are split into octaves creating reinforcing harmonics and an even more extreme "organ effect".

Another divisi strategy can be seen and heard in Anakin's Theme by Williams.

The score for that passage looks like this:

If you look for instance at bar 7 beat 3, you can see that 2nd Violins, Violas and Celli are all split in two voices with the higher voice in Celli doubling the lower voice of the Violas and the higher voice of the Violas doubling the lower voice of the 2nd Violins. You see this strategy continues loosely over the next bars. We don't know for sure why this decision was made but a strong reason could be to homogenize the string sections even more so that a really homogenous chord pad is being created. By the way, also notice that in spite of the heavy divisi in the middle registers, the leading melody in the first Violins is being presented in unison.

And just to show you the complete opposite of this strategy, let's skip ahead a little later into that piece

Here you can see a typical Williams orchestration strategy for sweepy climaxes to concentrate almost the entire string section on the leading melody in octaves. In some cases, he would even couple the Celli into that melody yet another octave lower effectively leaving only melody and bass line to the string section. In this passage, there are some horns and low woodwinds providing sustaining harmonies, otherwise it would sound quite empty.

The reinforcing harmonics of all these octaves create as much substance as possible on this melody line giving it this sweepy quality.

With all these strategies above we are basically moving in the field of "traditional orchestration". In the scoring world there are a few additional things to discuss regarding divisi which as stated many times before are a result of the specific circumstance of the music being written for recording and not for stage performance:

Divisi can also be used to specifically make use of the mic setup or the room itself. I mentioned earlier that the standard procedure for a divisi into two voices would be to split on the stand which consequentially means that both voices have an identical presence and spatial distribution in the room with the maximum level of homogenization. For dramatic effects however, you can consciously decide against this strategy. In this case the decca tree (which usually is the main microphone set consisting of three microphones above the conductor's pointing left, center and right) plays a decisive role. As the signal from this microphone set is used as the fundament of most mixes, it usually is the most present signal in the final mix. With the strings covering a large surface area and for instance the first violins spanning a space between "right below the tree on the left side" to "farthest away from the tree on the left side" you can use divisi strategies for distance effects.

A relatively common use would be to orchestrate the crescendo on a long note by having it "creep from the back to the front", e.g. starting with the last few desks and adding the front players bit by bit to the note. In this case a divisi on the stand would not really support this effect but you would need to divisi from back to front, eg. specifying "last 3 desks" and "first 4 desks" or even split into 3 groups or more. You can form extremely elaborate schemes to play with distance and panning in a natural space with such divisi strategies. However, I would advise to not overestimate the effect. In an isolated context it will be audible to hear these distance differences but it is not as extreme as you might imagine at first. With several close and other mics being part of the mix, the differences get equalized and additionally the distance differences are not great enough to create clear "far away" and "close" extremes. Especially with the realtively non-directional sound projection of the strings and the reflections in the room, the effect of such strategies very rarely justifies the effort.

However, effects of "big vs small" are relatively pronounced in a recording stage. If you decide to have 3 violins playing a line versus the entire 16 player 1st violin section can indeed create a quite dramatic effect, especially when it is supported in the mix by focussing more on close mics on passages with the 3 violins and going more for ambient mics with the section. One effect I like a lot is to open a piece with for instance a string quintet just featuring the first players and on a pivotal moment let the entire section join. This creates a quite spectacular "cinemascope" effect as suddenly you gain substance, stereo spread, volume and silkiness all at once. Again, this strategy supported by appropriate mix automations will create really impressive musical effects. Theoretically the possibilities are endless with divisi.

Another strategy that is very common in film scoring (but not exclusive to it) is to use divisi to mix different playing techniques on the same pitches. For instance having half the players play regular sustains while the other half plays the same note tremolo creates a nice mixture. Even more spectacular are such divisi where you have dynamic waves interlocking (e.g. while the sustains decrescendo the tremolos crescendo and vice versa). A great mixture of such kinds are sustains and harmonics (sounding at pitch or an octave higher). Such strategies are extremely common in more modern scoring approaches where you generally try to avoid sounding too "traditional" and where it is desirable to have a constant change of timbre and hue on sustaining notes. Doing such splits in unequal proportions adds even more variety (e.g. just one solo cello doing harmonics in dynamic waves over the rest of the section playing a soft regular sustain). So the possibilities are virtually endless and putting some effort into divisi can be a great playground for innovative sounds and textures. These are very often strategies that you barely think about when working with samples because either such elaborate divisi splits are simply not possible or you are using patches that have such divisis already "baked in" without really thinking too much about how they got to the sound. However understanding the strategies behind this adds a much deeper understanding of orchestration and texture and being able to construct these on your own without relying on "sample overdubs" of course gives you way more freedom.

In any case it is a vital information for any orchestrator to know the exact string line-up so you can adjust your writing accordingly, especially when using divisi in a more traditional manner. In a small string line-up you would prefer to avoid divisi and solve issues differently than in a large string line-up where you can relatively freely divide sections without the danger of losing considerable substance.

The bottom line should be that the decision to divisi or not should always be made based on context or musical intention. In general and traditional writing, it is advisable to strive for as little divisi as possible. Usually, this isn't really super problematic as often chord voicings need fewer voices than you might imagine in order to be effective. Trying to combine sustain notes and staccato notes into an entire string section as often seen in epic music might be another special case where in the recording world (unless London or LA) it might be the wiser choice to do string overdubs rather than divisi a lot.