Durations of bass figures

[johentsch]

To my knowledge, we have not addressed this question systematically in opp. 1, 3, and 4, but now that we're getting op. 2, I suggest we come up with a guideline together.

Processing figured bass computationally is tricky because we use a lot of implicit knowledge to know for which notes the absence of a figure signifies a 53 label and for which it doesn't. Even though we will not fix this problem here, thinking about the durations of the figures that do exist may alleviate it.

For starters, here is a condensed version of the MuseScore 3 documentation:

Each figured bass group has a duration, which is indicated by a light gray line above it [...].
Initially, a group has the same duration of the note to which it is attached. [...]
Setting the exact figured bass group duration is only mandatory in two cases:

1. When several groups are fit under a single staff note (there is no other way).
2. When continuation lines are used, as line length depends on the group duration.

However, it is a good practice to always set the duration to the intended value for the purposes of plugins and MusicXML.

Keeping note durations for bass figures is a reasonable default in many cases. Extending the duration of a label (in absence of extension lines in the original) is an act of interpretation. In some places it may me uncontroversial and desired, in others it could be an interpretive stretch. Which is why we should aim for consistency.

Please consider this excerpt from #260:

A few thoughts, please chime in:

• The 5 in m. 5.3 prevents playing a 6 chord and I think it is uncontroversial that the E-minor chord is held over the transitory D. I would vote for extending the duration of 5 to a quarter.
• The 65 that follows is ornamented with its root position chord (in modern terms). I could see this as an argument both for keeping the length of an 8th or to turning it into a quarter.
• mm. 6 & 7: I can see 9 and 6 extended to quarter duration which would reflect the harmonic rhythm and highlight E and G# (m. 7: C# and E) as passing notes.
• Same for 4, 54, and #.

I think these thoughts boil down to me suggesting that durations should be used to encode an interpretation of the harmonic rhythm that is based on the other voices and on how the average player would realize the thorough bass.

[johentsch]

I would also welcome the insights from you, @kigerh, because you have thought quite a bit about this question in the context of https://doi.org/10.5334/tismir.191 I think.

[davor-k]

Hi @johentsch ! Could you clarify for me what are the consequences of extending the duration of the figured bass labels? Is that information being extracted and processed in some way?

If it isn't, I frankly don't see benefits of arbitrarily (meaning: based on interpretation) adding or extending duration lines. The way I see it, figured bass is a prescriptive tool first, rather than descriptive.

Unless there's something I'm missing, trying to optimize the durations of all figures would be a tedious process, it would impose our interpretation on the score itself (and there really is a lot to interpret), and it would eventually describe the same thing that our usual harmony annotations do with far more detail.
It feels like interpreting how long should the extension line last is the same interpretation one has to do when actually annotating harmonies, so it would be kind of redundant.

That's my opinion but, I repeat, I was under the impression that figured bass in the score didn't really do anything computationally; I thought it only informed us annotators in deciding on the right label.
I might be wrong so please let me know and I'll gladly reconsider the issue.

[johentsch]

It feels like interpreting how long should the extension line last is the same interpretation one has to do when actually annotating harmonies, so it would be kind of redundant.

Thanks @davor-k, that's a good point. At the same time, it brings up the idea that the update of the thorough-bass durations could be done together with the harmonic analysis, would that make sense to you?

and there really is a lot to interpret

Then again, don't you think the example above is rather unambiguous? I actually doubt that including these decisions when writing the thorough bass during score creation would add a considerable amount of time.

Is that information being extracted and processed in some way?

Yes, the information shows up for each extract figure in the chords TSVs in the column thoroughbass_duration. In this discussion we need to consider also the consquence of not caring about the lines because that results in default durations, basically giving away a feature that may be useful to some.

TL;DR: I don't think it's a lot of extra work nor that there will be so much ambiguity (for Corelli), but maybe someone has an example where the decisions would be extremely difficult or haphazard?

[kigerh]

Unless there's something I'm missing, trying to optimize the durations of all figures would be a tedious process, it would impose our interpretation on the score itself (and there really is a lot to interpret), and it would eventually describe the same thing that our usual harmony annotations do with far more detail. It feels like interpreting how long should the extension line last is the same interpretation one has to do when actually annotating harmonies, so it would be kind of redundant.

Hello and thanks for highlighting this issue, @johentsch. Since we are analyzing the figures of our scores in our research, this is a problem that has occured before. We're currently thinking about adding some general rules to our program code, with which we can recognize and skip passages, but this needs to be completely accurate. It's also not limited to passages (at least not in our corpus); we encountered this problem with octave leaps in the bass as well.
Defining these rules or in your case entering the duration of figures should be done with a fair degree of certainty. I agree with you on this, @davor-k. For our corpus I think this should be possible, but I can't say for Corelli. Maybe @mskirsch has some insights on this, the question of the scope for interpretation in general and the possible difference between figures and the annotations?

[ameliabrey]

I'm happy to continue engraving precise durations of thoroughbass symbols -- it's worth a note that annotators should probably also be checking this, as their interpretations might well conflict and we don't want such incongruencies in the published data. So we do pass on that chore all the way down the pipeline -- it's just something to be mindful of from a management standpoint.

[vzhyu]

Was there a resolution to this? I too am happy to engrave with interpreted precise durations. I originally just kept it default to the same length as whatever note it was attached to, but especially in cases where we can see the exact voice leading that influence the throughbass, it feels incomplete to ignore the precise durations. I don't see this being particularly more complicated than the annotation/reviewing procedure we have with Roman numerals.

[johentsch]

Thanks for your thoughts, everyone! Having considered the question and the different arguments for a while now, I propose the following:

• we deviate from the default only in unambiguous cases (see below);
• we defer deviations requiring interpretation to the harmonic annotation and make it part of the guidelines to adapt thorough bass figures according to one's analysis.

By unambiguous cases I mean repetitions of a bass note such as the 5 in m. 7 from op02n06a:

This is a rhythmization of a half note and has been adapted accordingly (shortcut [C]+6 in edit mode). The same argument would apply if the dotted quarter was instead a quarter followed by an eighth rest. There may be rare cases where the second of two repeated notes is understood to mean "35" rather than prolonging the previous note's figure but this should be highlighted as a special case (maybe by actually adding the editorial figure [53]) lest someone thinks it's oversight. In any case, such a decision may be updated by an annotator later.

On a sidenote, this discussion highlights the fascinatingly ambiguous ontological status of bass figures somewhere between annotation labels expressing a harmonic interpretation/realization and an integral part of the score, the thing to be interpreted and annotated. So far, we have treated the figures purely as the latter, implicit in the guideline to take the figure into account for harmonic analysis. What we might try out when annotating op. 2, adapting the durations of labels according to one's harmonic analysis, is not actually to infer with the original score: The original figures do not come with engrained durations. Rather, it is adding information to the unfigured space, reducing its uncertainty. Looking forward to this experiment!

[johentsch]

Here comes a specification of our guideline to show that it is to be applied in the strict sense, i.e. only for tone repetitions.
Those who play thorough bass might find themselves tempted -- and need to resist that temptation -- to prolong durations in the following examples:

neither b. 1 nor b. 3 are prolonged	b. 3 is not prolonged Edit: the # resolving the 4 should be aligned, see next post	b. 4 is prolonged

In the first two examples, prolonging the label to a dotted quarter's length would effectively mean prolonging the bass note's duration as well. In that sense, it is closer to the idea of thorough bass figures as harmonic interpretation, which is one we decided to eschew. In the third example, the prolongation on b. 4 -- although an edge case -- can be viewed as unambiguous because it follows the one on b. 3, based on immediate tone repetition, and can be viewed as an ornamented variation of it.

As always, comments are welcome.

[johentsch]

In the Pepusch print, thirds resolving fourths are consistently placed on the same horizon. In order to stay true to the original edition, we need to use continuation lines, e.g.

This requires prolonging the 54 figure to a quarter. By extension of this logic, we should include other suspensions in our list of "obvious cases" and prolong them accordingly, e.g.