New IMSLP Music Typesetting Guidelines

Grandados - Goyescas

IMSLP has developed a set of minimum quality guidelines for newly typeset works posted to IMSLP.

Why bother? There are plenty of manuscript editions on IMSLP that make for very bad reading, and there are also some scans that are difficult to read. The answer is that while IMSLP acknowledges that there are scores on the site that are not usable in rehearsal or performance, those scores may have a particular musicological value, or they may be the only score in existence for a given work. The manuscripts in particular have a very valuable place in the collection, especially autograph manuscripts from which one gleans the composer’s original intentions. First editions have similar high musicological value.

Ideally, IMSLP aims to have a combination of musicologically valuable editions, even if they are not very usable in rehearsals or performances, combined with high-quality modern typesets that make rehearsals and performances easy – in other words, we want the best of both worlds.

Any professional musician can tell many stories about their worst-nightmare rehearsals and performances with badly copied music. Unfortunately, in some cases, the lack of alternative editions makes it necessary to use bad music regardless of the problems. I will give three examples.

My first example was a near-disaster in a concert. It was a work where the composer had typeset the two separate movements in separate files. When he printed, there were two page 1s, two page 2s, etc. On the day of the concert we had a sound check, after which the second violinist left her music on a table while we took a small break. While she was out, an enthusiastic student decided to look at the music. He must have shuffled things around, because we got on stage with the second violin pages in the wrong order. Luckily, the violinist saw that something was not quite right before we started playing, but there was a lot of embarrassment trying to find the page 2 that started with the correct consecutive bar number for the page 1 from the same movement – all of this in front of the audience, and in the end there was still one page that was not in the right order. Needless to say, it was not exactly a note-perfect performance, which was really unfortunate, since I really think it was a good piece.

I can understand that the composer was working under deadline. I also know that getting a file to print with a first page number that is not “1” can be a very obscure thing to figure out how to do. I had a graduate student spend hours trying to get her multiple-file thesis to do exactly that. In the case of my concert, it ended with less than acceptable results. And although the violinist perhaps should have bound the music, there were also page turn issues that made the binding a bit impractical. She did say she will never leave music lying around again.

My first suggestion to typesetters is to keep notes on all the unusual things that you will not need often, but cannot afford to spend hours looking for each time. I have a document I keep on my laptop that is a list of exactly those things. It might take a long time to find a solution the first time, but after that I know where to go quickly if I cannot remember how to do something.

My second example also involved a contemporary work, but the sponsor would not pay for more than two rehearsals, and we had three works to perform. One of the works was rhythmically complex, in 12/8 metre. The composer had written the simplest rhythms for him to enter into the computer, but they obscured the pulses, for example minims instead of a dotted crotchet tied to a quaver. At the beginning of a bar, it is relatively easy to work out that the next note is 1/3 of the way through the beat. But further along in the bar, the problem compounds itself, and all the musicians at various places in the piece had to think “is the next note 1/3 or 2/3 of the way through the beat?”

That kind of hesitation while a musician calculates something is fatal, especially at fast tempos. Needless to say, the rehearsals and concert were torturous. At one point, the first violinist was banging her bow on the stand, complaining about an effective syncopation that was not written like one. She also kept exclaiming “Mein Gott!” The composer, who had been asked not to attend the first rehearsal, insisted on being there while we waded through all of this for the first time. And if he was unhappy with the first rehearsal, he was even more unhappy with the performance.

I do not have much sympathy. It is fairly easy to work out that three hours * two rehearsals = six hours of rehearsals / three works = two hours rehearsal per work (less if you consider that his was not the longest work on the programme).

For performing musicians, the solution we want is for you to give us visual anchors that match the aural ones. Our attitude is simple; the audience cannot tell whether we are playing a minim or a dotted crotchet tied to a quaver anyway. They hear the equivalent of four quavers no matter what we are looking at, so just give us the version that makes the music easy to put together. We do not really care if it takes you more time to put in two notes with a tie. It is your work, and your performance, so if you want it to be good, take the extra time.

I do have to mention one exception to this – sometimes music involves things that are meant to be “off the pulse” and they need to visually look like that. Many years ago, a colleague of mine decided to re-write Mozart’s G minor symphony. His attitude was that musicians do not emphasize things that are not on barlines (to start with, this is condescending), so he wanted phrases starting part way through bars changed so they started on barlines. Solution? Throw in a 2/4 or 3/4 bar, re-bar the next section, and throw in the equivalent at the end of the section to get back to where things were. (I have a musicological problem with that anyway – an upbeat is frequently not supposed to be emphasized – the main stress for many of this type of phrase would be on the barline note anyway, but musicological arguments are not the point here.)

It was an out of town concert with a rehearsal in the afternoon. There was an uproar in the rehearsal, much indignation, and a general lack of ability to read something we had been reading for many years another way. Notes we knew as upbeats were now downbeats. The only person who had a mistake-free rehearsal was the ancient, nearly-blind principal flautist who simply shut the music and played from memory. In between the rehearsal and concert, someone made the manager drive back to pick up the “real” parts. Problem solved, and as one musician put it: “Don’t mess with Mozart.” (He was not that polite.)

The moral of this story is that if a piece has been around long enough to be relatively well known, musicians will hear it they way they have been seeing it for years. Changing the barring is visually jarring, and off-putting in the extreme. (And of course, if it is a new piece and you want the first few notes of a phrase to lead into a stressed note, then they probably need to be ahead of a barline anyway.)

Actually, though, the top of my story list happened many times. It would be any number of performances of Russian music – Shostakovich and Prokofiev to name two composers, with the infamous Russian manuscript parts. Many of the fast passages had leger lines that were not lined up properly, resulting in 3rd-leger-line Es that were higher visually than 4th-leger-line Gs. Try reading that in semiquavers at 144 to the crotchet!

Since those were old hand-copied parts, one would think everything would be solved with modern computer typesetting programs. Certainly the leger lines are all neatly lined up. So why is it that some modern typesets are just as bad, if not worse, than sloppily handwritten manuscript?

The reason is that computer typesetting does not automatically solve all problems. In fact, to get the best out of your typesetting program, you still have to proofread, and do quite a number of things manually. You have to spend time learning more than the basics, and you definitely cannot rely solely on midi dumps (which recognize rhythms and notes from a sound file) or recognition software (using scanned music).

Producing good typesets

There are two aspects to producing a good typeset: one is learning what constitutes good notation, and the other is learning the finer points of whatever music notation package you use, so that you can do the more complicated typesetting.

The guidelines are posted here, along with a list of some good resources, but in order to understand the seeming pickiness of the guidelines, it might help to understand what sorts of problems badly typeset music cause for the performing musician. Inaccurate notes are by far not the only problem. Herewith (and helped by forum members who also have problems with bad typesets) my litany:

  • The general size of staves and noteheads needs to be large enough to read easily, but not so large as to waste space, thereby adding to the page turn problems. Harp players cannot bend very far forward to read tiny notes, but if they bring the stand close, they will bang their left hand on it when they reach for low notes.
  • Any elements (accidentals, dynamics, etc.) that collide with each other on the page cause misreading. If one item obscures another, neither can be read quickly or accurately.
  • Staves that are not adequately spaced can cause collisions of high and low notes, or dynamics, between the staves. Again, if one item obscures another, neither can be read quickly or accurately. Spacing that is too wide can waste space and add page turn problems.
  • Rhythms that are not visually spaced proportionally to those rhythms can cause musicians to misread. The faster the tempo, the more likely they will make mistakes.
  • Parts without adequate rests at the page turns can cause missed entries, especially for instruments that are “one-on-a-part” (for example, woodwinds, brass, percussion, harp). Even string players might have page turns where it is not a good idea to lose half the section sound because players on the inner desks have stopped playing to turn the page.
  • Badly planned repeats cause extra page turns, especially backward page turns.
  • Musicians do know that accidentals revert to the key signature after the bar line, but there is a reason for courtesy accidentals – we frequently have a fraction of a second to make a decision, and hesitating to double-check something causes stumbles and mistakes. If you want an accurate reading, give us the courtesy accidentals. Typesetting programs do not automatically do this, especially if the music is entered by playing it in. These must be added manually.
  • Using incorrect symbols can cause hesitation, for example using tildes (“~ ~ ~ ~ ~”) instead of the traditional trill squiggle. And the midi-dump equivalent of a trill is a measured set of fast notes that not only is not really a trill, it takes up way too much space on the page and makes it difficult to see at a glance how long the trill is.
  • Repetitive patterns – it is easy to get lost in these if the typesetter has not used the correct symbols to signify the repetitions, or indicated how many times the pattern repeats. If your players are thinking “was that the 7th or 8th repetition?” you are asking for the next phrase to be in the wrong place.
  • Similar to the previous item, leaving many bars of rest as individual bars wastes paper, gives us bad page turns, and it is also very easy to get lost in them. If a musician gets lost in these, the next entry will be wrong. Group multiple-bar rests in the traditional way – according to phrases, e.g. –4– –8– –12–, where each of those groups is an audible phrase.
  • Musicians with many bars rest need adequate cues toward the end of the rests, or they cannot find their entry.
  • Badly chosen cues can cause missed entries. I was once given a viola cue in a loud passage. I have no idea how the copyist thought I would hear it!
  • Unlabelled cues can also cause missed entries.
  • Elements that should not be full size (cues and grace notes or other ornaments) can cause misreading if they are full size. A musician might actually play a cue if it is the same size as the notes he is supposed to play, and grace notes that are too large cause the base rhythm of the bar to look like it has too many beats.
  • Unclear articulations can cause tripping. Even if they do not cause an actual misreading of notes, you still want musicians to do your choice of articulation, not theirs. Slurs that start or stop in between notes are the biggest culprit here. Be clear as to where a slur starts and stops.
  • Clef changes that are in the middle of a run, or some other awkward place, can cause wrong notes. However, it can be impossible to write a very long run without either leger lines or a clef change, and many performing musicians would choose leger lines over a clef change in this case. If you absolutely have to change clef in the middle of a passage, at least do not put it in front of a note with an accidental. Choose a diatonic note that is at the beginning of a beat, and preferably at an octave or some other landmark place within the run.
  • With all clef changes, check the part extraction, because in some cases, the computer program puts the part entirely in the first clef you chose, with no changes afterward. Proofreading is essential, and you might have to change clefs manually in the parts.
  • Pages that are not numbered can cause confusion if they are loose and unbound. If you are saving separate movements as separate files, make sure movement 2 is consecutively numbered after the last page number of movement 1, otherwise you will have multiple page 1s, page 2s, etc.
  • Elements from a score that are printed once at the top of the score should, but do not always, come through to individual parts in an automated part extraction. This and many other things can be checked by simple proofreading, but you can imagine what happens in rehearsal when some parts have a tempo change and others do not . . .
  • From time to time, we have all had “with” indications without the “without” indications. For example, violinists who are still using mutes when everyone else has changed to fortissimo, or seeing that they were supposed to take a mute off that they never had on in the first place. The same applies to pianists who cannot work out where the composer intended to stop the una corda pedal. Brass and woodwind players really do not like keeping their necks in strained positions longer than absolutely necessary for a schalltrichter auf passage. The same applies to solo or soli passages that need to go back to tutti. All of these indications need to be clearly marked. Conductors do not like wasting rehearsal time trying to figure out what you intended.
  • None of the following instruments – horn, trumpet, or instruments in the clarinet and saxophone families – read their parts in concert pitch, and while there are some historical parts for clarinet, horn, and trumpet in C, those musicians would have been playing instruments pitched in C, not modern instruments pitched in B-flat, F, or E-flat. Traditionally, a score has these instruments written in the pitch they read, not in concert pitch, and professionally trained conductors know how to read them. Some modern scores put everything in concert pitch, which helps conductors who have not had this training. However, regardless of which system your score uses, the part extraction needs to be in the pitch those instruments read. It is never acceptable to make a saxophonist read concert pitch, and you are asking for bucketfuls of note mistakes if you do not transpose the part for him. The common transpositions are B-flat or A for clarinet (though A clarinet is never used in wind band settings), B-flat for trumpet, and F for horn. Use them.
  • Computer automated transpositions can cause misreading. We all know that everyone should be able to read 7 sharps, but if there is less likelihood of mistakes with 5 flats, then use that instead. Concert B major, for a clarinetist playing a B-flat instrument should be D-flat major, not C-sharp major. This has to be done manually, by making the key signature on that line independent, and manually changing the key to D-flat.
  • Parts with no bar numbers or rehearsal numbers/letters are very difficult to rehearse from, and this problem expands radically with the length of the piece – who wants to count to bar 200?
  • (This is a personal bugbear that not everyone might agree with.) It became the fashion at one point to number bars in increments of 10, but phrases rarely start every 10 bars, and this type of numbering is a nuisance when there is a multi-bar rest. As a performing musician, I would rather see a barline where the phrase starts, because it gives me a visual landmark on the page for an aural landmark that is easy to hear. Take the trouble to go through the score and mark where the multi-bar rests should divide, rather than letting the computer do an arbitrary division based on a number that has never been a common phrase length.

If you have professional musicians playing your music, you need to give them the chance to give you a professional product in a minimum amount of time. Rehearsals are usually in short supply anyway, and all of us (this should include you) would prefer to spend most of this time on interpretation. If you have amateurs playing your music, bear in mind that they deal with bad parts less well than professionals do. They, if anything, need even better parts.

The fact is that if music is extremely complex – but accurate, clear, large enough, adequately spaced, with proportional rhythmic spacing, good page turns, adequate cues, and all of the rest of the above list, musicians can sightread fairly accurately, which means less rehearsal time wasted on correcting and adding things to the part that should have been there to start with.

This actually happens all the time in some areas of the industry. Part of my training was in Los Angeles, where there is a well-established sound track recording industry. They work on tight deadlines and no rehearsal time. What is put on the stand in any session has to be put down on the track in that session. Musicians will not get a chance to come back the next day after they have practiced the part. They have to be good readers, but the composers and copyists have to provide clear parts. And they do – in fact long before Finale and Sibelius they produced clear and accurate parts by hand. On top of that, many of the sound tracks are incredibly complex pieces of orchestration (think Star Wars, for one).

It is simple – give me what I want to see and I will give you what you want to hear.

Becky L. Steltzner
Head of Woodwind & Chamber Music
SA College of Music
University of Cape Town

6 thoughts on “New IMSLP Music Typesetting Guidelines

  1. Kevin

    Great post, Becky! I recently had a performance of a piece of mine that I did not have time to copy into a notation software program. My hand writing is pretty good, but the result: okay, considering the circumstances, if you didn’t mind a few misleading coffee stains and a few snide remarks by the players.

  2. nikolaus

    hey Becky, thanks for your post!
    I totally agree with your points. As an arranger/conductor I spend many hours on good sheets, and sometimes I wonder if I am too picky. Now I found quite every rule that I made up for myself in your post. And now I know why people come to me and ask for my service :-). Which I also was wondering about, because good typesetting seems to have become second nature to me…

    I myself had some bad moments in reading bad sheets. For instance accompanying in a casting with a poor midi-transcription. No key signature, very awkward chords appearing. Who can instantly tell out of g# – c – eflat – f# that Aflat7 is intended? :-)

  3. Dr. Heinz Anderle ("punch")

    Literature on music engraving rules is scarce – in German, we still have the book by Herbert Chlapik “Die Praxis des Notengraphikers” (the practice of the notation graphics designer), in English, it is the book by Ted Ross “Teach yourself the art of music engraving and processing”.

    High quality computer engraving requires high quality, high-speed professional software which is not available for free, elegant computer fonts, and a good visual judgement of proportion and balance aside from the knowledge of essential musical basics.

    The highest standard of classical engraving had been reached by major German publishers around 1900 – one might look at the old Schubert complete edition. For music of the classical and romantic era such an appearance of conductor’s score and performance parts is highly desirable.

  4. Thomas

    For improving quality, I think in the long run it’s essential to agree on a standardized way of providing the original sources as well as MusicXML files (if possible) together with the PDF. If there is an edition with quality flaws that could be ironed out, there is no point in making IMSLP dependant on the original uploader to do so. He might not be active any longer on IMSLP, the appropriately licensed score might be from a different source and the original author might not be trackable at all or he might not have the time, will or skills to polish the edition (that might be why there are flaws in the first place).

    I totally agree that having scores e.g. in Finale format *only* is unacceptable because of their inaccessibility. PDF to this date is the most accessible format for sheet music and is a must. But not keeping the sources in my eyes is a sin. The value that’s lost could even be monetarily estimated. Ask a professional music engraver what he charges for typesetting some music. It’s not much to make a living from, but still totally unaffordable for a non-profit project like IMSLP. So, to re-typeset a whole piece because of quality issues in another version that could rather easily be solved is an unjustifiable waste of resources. And I’m not even speaking of other possibilities of reusing the sources like preparing a concert pitch score from a transposing one and vice versa, re-instrumentation, arranging, adjusting for other paper sizes or visually impaired musicians. Those applications might be out of scope for IMSLP’s original aims.

    Yes, there are problems with proprietary source files. You’d most of the time need the appropriate software package, you’d also need the used fonts, maybe even the version of the software package matters. Yes, there are problems with MusicXML. The export and import paths often are lossy and sometimes cause funny results. Yes, people sometimes use the software in strange ways that makes it unnecessarily difficult for others to work with the files. But can this justify throwing away data that in the business world would be worth maybe tens of thousands of Dollars/Euros/whatever in total?

    I know that some people attach source files and/or MusicXML files to the PDFs and I’ve seen Lilpyond sources on discussion pages. Lacking other possibilities right now, I consider this a good idea. But I think there should be a proper way of archiving the sources. Pasting code on the discussion page strictly speaking is misuse and quickly gets unmanageable, especially if there is some actual discussion going on, the piece is rather long or there are multiple versions of the same piece. Attaching something to a PDF is certainly not rocket science, but most people would have to obtain the tools for it and learn how to use them (speaking for myself, I’ve never done it so far). It bloats the PDF files, not all PDF viewers support opening attachments and I’d consider it bad data management practice.

    Right now, I have the impression that IMSLP discourages providing source and MusicXML files. Encouraging it makes much more sense to me.

    1. Becky L. Steltzner

      This last comment, being a plea for approval to upload a specific proprietary type of source files to the wiki, is not appropriate to the article, because it does not specifically deal with typesetting issues (non-collisions, placement of text, etc. etc.). This has been brought up at the forums, and the people who would have to do the work of quality control would follow those rather than continually checking this article for comments, so I would suggest you post there instead.


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Twitter Users
Enter your personal information in the form or sign in with your Twitter account by clicking the button below.

Twitter Tweet This