Collaborative process when composing for dance

My collaborative process has developed over a number of years through working with choreographers in a range of contexts, both professional and educational. I have found that the main factor separating collaboration with choreographers from that with musicians is the necessity to adapt and translate musical vernacular into more general terms that can facilitate effective communication. This is most crucial in the early stages of a collaboration, when most discussions are entirely abstract, and aimed towards establishing a rough outline of structure and content. Once some material is in place, the choreographer can indicate specific aspects or elements of it and refer to them when discussing the composition.
My collaborative process can be roughly divided into four stages: an initial discussion stage with the choreographer, discussing the plan and concept of the work, both choreographic and musical; an initial creative stage, in which musical ideas will be formulated and tested with the dance; the main stage of musical composition, when the bulk of the material is written, often concurrently to the dance being choreographed; and the final stage when both score and choreography are being refined to ensure cohesive and effective interaction. As part of my process, I try and attend as many of the dance rehearsals as possible while the choreography is being created, so that I am not relying on verbal descriptions of the choreography to create the score, but can instead observe directly as the movement is created. 
One of the methods I use to facilitate communication in the early stages of a collaboration is the use of reference tracks. In essence, rather than the choreographer trying to describe abstract sonic features, they can choose music samples that match what they want, and use these to demonstrate aspects of music that they want to include in their work. The focus on aspects of music is crucial, as a choreographer may show a track that they like the rhythm of, and the composer needs to know that this is the factor in focus. As an example, in my first dance collaboration, the choreographer showed me a track that featured violin with an electronic, dupstep-style backing. I then began writing demos in a similar genre, not realising that she had shown me the track to demonstrate the violin and melody, rather than the genre. Once the misunderstanding had been identified, she showed me another track to indicate the overall genre she wanted for the work. 
In the initial ideas phase, I find that it is important not to work too specifically to the requested specifications. For a start, it is unlikely the choreographer communicated their exact intentions, or that they were interpreted completely correctly. As such, I find a quick demo approximating the brief works best to allow the choreographer to identify the aspects of the music which they like, and those which need adjusting. In addition to this, as the work develops and evolves, the choreographer may find themselves drawn to music different from that which they originally intended. 
Once initial material for the work has been decided upon, it must be expanded and developed. It is at this stage that I find it most useful to observe rehearsals, and to immerse myself in the creative process of the choreography. This allows a composer to be aware of changes to the structure as they occur, and to gain an in-depth understanding of the purposes behind the movement, whether narrative, metaphorically, or abstract. It can also be helpful to adopt techniques being used to generate choreographic material, and apply them to musical material. Such parallels in methodology can contribute to cohesion in the work overall.
The final stage involves tweaking elements in the material that may have developed independently of the choreography, in a way that does not suit it. In my experience, this most often involves adjusting the length of sections in the music to align them with corresponding sections in the choreography. In can also involve the addition of cues into the music, to signal the dancers where they may have difficulty distinguishing when a particular movement or change is to take place. 
This is a summary of my own practice, but it is by no means the only advisable method. While advantageous for many reasons, constraints of time and budget make it less common in the professional realm, where choreographers may wish for music to be finalised before choreography to streamline their process, and composers may prefer to compose to a brief rather than attending rehearsals and developing a piece over weeks or months. However, if time and money are not considerations, and creative partners wish to work closely in the development of a work, there is much to be gained from close, ongoing cooperation throughout the collaborative process.