11

Using dots, ties and tuplets of standard note values (whole, half, quarter, etc.), are there values that cannot be represented?

Examples:

  • 1/8 is an eighth
  • 7/16 is a two-dotted-quarter
  • 15/8 is a three-dotted whole
  • 1/3 results from a triplet of a whole
  • 11/24 is a tie of 1/3 and 1/8
    ...

what rational numbers are not in this list?

Update:
Said in a general form: what n/m cannot be represented?

nightcod3r
  • 655
  • 5
  • 14
  • Using tuplets and ties, anything can be represented, because any fraction of a beat can be represented using tied notes within a tuplet, and then this can be tied to any number of beats. You could end up with some pretty ugly looking rhythms though. – Bob Broadley Apr 04 '16 at 09:16
  • 4
    Irrational portions of a beat or measure cannot be represented, but I don't see how they could reasonably be accurately played. I.e., the cardinality of the set of possible note values is countably infinite. – Todd Wilcox Apr 04 '16 at 10:14
  • 4
    Is this really some kind of math question in disguise? I believe the answer depends on how general [tuplets](https://en.wikipedia.org/wiki/Tuplet) you allow. For example, if you want a note with the mathematical duration 1/97 (one ninety-seventh), is it allowed to create a tuplet embracing 97 notes, and write the number "97" above it? – Jeppe Stig Nielsen Apr 04 '16 at 12:49
  • See related [question](http://music.stackexchange.com/questions/43113/are-ties-necessary). – nightcod3r Apr 04 '16 at 16:45
  • @theonlygusti sure you can. I am not sure how one would *write* such a note, but one option would be to use the Greek letter as a notehead. – phoog Jun 07 '20 at 07:49

3 Answers3

27

Math Alert! Also, I will be very much discussing what is theoretically possible, not necessarily what is convenient for the poor musician.

The notation for musical rhythm is more or less equivalent to writing a fractional number in binary (e.g. using a radix point). Each note type represents a different place value. For example:

  • Whole note = 1.02
  • Half note = 0.12
  • Quarter note = 0.012
  • Eighth note = 0.0012
  • and so forth...

Since you can (theoretically) add an arbitrary number of flags to a note head, you can carry this out as long as you please.

Dotted notes (single or double) represent adding two (or three) adjacent places:

  • Dotted half note = 0.102 + 0.012 = 0.112

Ties allow you to add any place values, regardless of if they are adjacent:

  • Half note tied to eighth note = 0.1002 + 0.0012 = 0.1012

Since, given enough place values, any numerical value can be represented in binary (i.e. a sum of the value of each place), the answer to your question is technically, no, there are no values that cannot be represented. However, this is not practical, since many numbers would require an infinite series of digits to represent. These can be broken into two classes of numbers (as you allude to in your question) -- rational and irrational.

Tuplets can be used to handle the rational case. If ties represent addition, then an n-tuplet represents division of a time unit into n equally-sized portions. Any rational number can be written as a fraction, m/n, which has the same value as 1/n added to itself m times. This number can be represented musically by tying m copies of an n-tuplet. For example:

  • 31/87 can be represented by tying together 31 notes that are 87-lets.

However, in the irrational case, the value cannot be represented as a ratio or fraction, so tuplets don't work. This leaves you back to writing out an infinite series of tied notes, which cannot be improved upon because irrational numbers require an infinite, non-repeating representation.


The final question that must be considered is the precision required. While theoretically, you would require an infinite number of notes, with an increasing number of flags on the staff, the human ear is limited in how precisely it can detect rhythms. At a certain point, rounding off becomes inevitable. For example, even NASA requires only 15 digits of pi (in base ten) in order to calculate the positions of interplanetary probes. Digits beyond this are insignificant enough as to not matter. Similarly, any number inside a computer must be rounded off to some finite approximation, simply because of limited memory.

So if you limit yourself to a basic unit of precision (perhaps a 128th note?), then you can represent any multiple of that basic unit (using ties), as well as any rational number (using tuplets).

Caleb Hines
  • 20,376
  • 2
  • 50
  • 106
  • Looks like your question could be reduced to the second section, more precisely to n/m = 1/m + ... + 1/m (n times), where 1/m is either a standard note value or an m-let, and '+' is a tie. – nightcod3r Apr 04 '16 at 16:53
  • Brief note on naming convention: I deliberately placed n on the bottom (m/n) so that you could use the generic term "n-tuplet". It's clear that any integer *m* (0 < *m* < *n*) is valid, since you can tie any number of notes. So the question boils down to whether you can have an arbitrary *n* in an n-tuplet. – Caleb Hines Apr 04 '16 at 17:08
  • via Wikipedia (Tuplet): Besides "triplet", the terms "duplet", "quadruplet", "quintuplet", "sextuplet", "septuplet", and "octuplet" are used frequently. The terms "nonuplet", "decuplet", "undecuplet", "dodecuplet", and "tredecuplet" had been suggested but up until 1925 had not caught on. By 1964 the terms "nonuplet" and "decuplet" were usual, while subdivisions by greater numbers were more commonly described as "group of eleven notes", "group of twelve notes", and so on. – Caleb Hines Apr 04 '16 at 17:10
  • ...and I just saw your comment on the question that you're allowing any number of tuplet. So then, any rational number can be represented. – Caleb Hines Apr 04 '16 at 17:16
  • @CalebHines As long as they are allowed in conventional notation, sure! So, may be your answer could be rewritten as to first present the proof that any duration in Q has an expression with tuplets and/or ties, and then may be comment on the convenience of representing some of them with dots, just for the sake of readability (I assume this is the reason, since they don't seem to be necessary). – nightcod3r Apr 04 '16 at 18:39
  • @nightcod3r I'll probably leave the answer as-is. To my mind, it makes more sense to start from the *most common yet specific* case (standard note values being binary place values), then to progress through successively *less common and more general cases* (dots and ties being addition, the former being more common) until you get to the *least common yet most generic* case (tuplets being any rational number), before ending with the *impossible* case (using infinite notes to write irrational numbers). Your suggested approach seems (IMHO) akin to teaching division before learning place values. – Caleb Hines Apr 04 '16 at 20:29
  • 1
    @BarryTheHatchet Indeed. Or, for the non-integer portion, "fractional part": *"For a positive number written in a conventional positional numeral system (such as binary or decimal), the fractional part equals the digits appearing after the radix point."* -- https://en.wikipedia.org/wiki/Fractional_part – Caleb Hines Apr 04 '16 at 20:40
  • 1
    @CalebHines You're probably right, I was seeing more from a formal point of view, mathematically speaking, but it's musicians (and the like) what we're targeting here, so specific-to-general it is. – nightcod3r Apr 04 '16 at 20:40
4

Notation as we know it has been around for many centuries, and although it was quite vague initially, it's sorted. Using ties, dotted notes and tuplets, as you say, means that any note duration in any time signature can be written. O.k. as Bob says, some could end up ugly and difficult to read, but if something had been deemed unwriteable, a new configuration would surely have been introduced by now.

Tim
  • 183,051
  • 16
  • 181
  • 444
1

One can approximate irrational note values rather easily in standard notation. For example, a note with the value Sqrt(2)/2 can be approximated by a fraction like 12/17 (larger numbers in the fraction make the approximation better and smaller ones make it less accurate.) The theory of continued fractions explains how it works. The error for this approximation is less than 1/289. It's always less than 1/denominator squared.

One then writes a 17-plet (17 eighths in the space of a half note for example) then groups 12 of them together (dotted half would do). Of course, another irrational value would be necessary to fill out the measure (should the piece be written with barlines.)

ttw
  • 24,207
  • 1
  • 31
  • 76