Proceedings of the 9th Conference on Interdisciplinary Musicology – CIM14.

Berlin, Germany 2014

T HE G EOMETRY OF THE P ERUVIAN L AND Ó AND ITS D IVERSE

R HYTHMIC PATTERNS
Juan F. Miranda Medina1 , Jan Tro1
1 Department of Music, Norwegian University of Science and Technology (NTNU). Trondheim, Norway.
Correspondence should be addressed to: [email protected]

Abstract: The landó in Perú is considered to be sensual, musically also historical because it presented the first landó ever recorded:
interesting and complex to perform. This work presents a new insight Samba Malató. The only thing Nicomedes and his company had to
into the landó using different metrics taken from the fields of computer start with where some fragments of the melody and text of the song.
science and mathematics. These measures attempt to quantify musical Therefore he and his sister Victoria resorted to memories both of
concepts such as rhythmic complexity, syncopation, and regularity. In elder witnesses and of their own, as well as to ancestral memory [6,
addition, a measure based on the Fourier transform is proposed as a ch. 2][7], thus “challenging the primacy of historical documents
rhythmic regularity measure. over collective memory” [6].
During the Afro-Peruvian revival the landó suffered a period of styliza- The landó did not undergo the same fast stylization process as
tion that started in the 1950s and culminated in the 1970s. The three the festejo [8], but it remained largely ambiguous in its musical
drum patterns that became representative for the landó are compared
to the one used in the first landó ever recorded, Samba Malató. The
form, a feature that gave it a mysterious and attractive character for
use of phylogenetic trees supports the uncertain history behind the younger performers [1]. Eventually some rhythmic patterns in the
landó, since none of these four rhythms seems to be the ancestor or cajon 2 began to be used consistently, which interestingly enough
progenitor of all the others. Nevertheless, this might open possibilities differed very much from those used in Samba Malató. One of these
for further research including musical transcriptions of old recordings, patterns, as pointed out by Feldman, is rhythmically identical to
which could also be analyzed using the methods discussed in this article. a bell timeline that is often used in West Africa and the “Black
Atlantic” [4, p. 215]. It is said to have been a contribution of
Caitro Soto [1], but there is no consensus on the matter. In a
1. I NTRODUCTION personal conversation of the main author with José “Lalo” Izquierdo
(co-founder of Perú Negro, one of the most influential music and
The landó is a rhythm in the Afro-Peruvian répertoire that un- dance companies of Afro-Peruvian music), he asserted that Ronaldo
avoidably evokes mystery and uncertainty, both because of its Campos (founder of Peru Negro) together with other members of
obscure origins as well as its multiple possible readings [1]. This the company gathered many rhythms and conducted field work
work begins by discussing the origins of its revival and posterior during several trips that included not only the vicinities of Lima
stylization, and then presents the most representative drumming (such as Chincha and San Luis de Cañete) but also the Peruvian
patterns for the landó, which will be subsequently analyzed using north coast (Zaña, Lambayeque, Chiclayo), which were relevant
methods borrowed from computer science [2] and mathematics [3] because they used different instrument than in the capital, such
to gain a better understanding of these complementary rhythms that as the checo and the angara (pers. comm., October 26, 2014).
together constitute the basis of the landó. Izquierdo emphasized on the landó as an example of “popular
manifestation” (manifestación popular). Rosa Elena Vásquez, an
2. L AND Ó : ORIGINS AND RE - PRESENTATION active Peruvian musicologist, stated that Caitro Soto and Pititi,
There is no tracing back the roots of the landó. Nicomedes two of the most prominent cajon players of the time, together
Santa Cruz stated rather convincingly that the landó stemmed with composer and singer Chabuca Granda and her guitarist Felix
from the Brazilian lundu, a rhythm he said was first danced Casaverde, may be responsible for the evolution of the rhythmic
by Afro-Brazilian slaves with Angolan roots, that then became patterns of the landó in a sort of collective creation. The musical
popular among the ballrooms in Brazil [4]. Though the evidence piece Cardo o Ceniza can be regarded as a culminating point in the
that supports Nicomedes’ hypothesis is questionable, much of the evolution of the cajon patterns for the landó (pers. comm, October
Afro-Peruvian population—and among them many musicians and 26, 2014)[1].
performers—embraced this as a truth. Furthermore, Nicomedes
established a genealogy of rhythms according to which the landó 3. P RESENTING THE RHYTHMIC PATTERNS FOR THE
was the ancestor of the zamacueca, from which the marinera LAND Ó
derives. 1 Thus, the landó became an essential symbol of double
consciousness linking African roots to Peruvian identity [4]. There are a number of different ways in which a rhythm can be
represented [11]. Binary representation divides the time it takes for
The Afro-Peruvian revival was a movement that started in the a rhythm to repeat itself in smaller time units of equal duration.
1950s, inspired partially because of migration from people from Each of these units, from now on referred to as pulses, is numbered
the mountains to the capital, thus creating a nostalgic remembrance starting the count from zero. When a musical attack occurs, the
of the colonial period. Following the example of José Durand who corresponding pulse is represented with an x or a ‘1’, and it is
founded the first dance and music theater company based on Afro- referred to as an onset. Those pulses that are not onsets are
Peruvian traditions, Nicomedes founded Cumanana in 1958 [6, p. designated with a o or a ‘0’. All onsets are considered to last only
55], and recorded the album with the same name in 1964, which one pulse, which as an important difference compared to traditional
was to become perhaps the most influential piece of work for the Western staff notation. If two consecutive onsets are ones, it is
perception and diffusion of Afro-Peruvian music [4]. It was in a because the instrument has been struck two times in consecutive
booklet sold together with the album that Nicomedes presented his time units or pulses.
hypothesis for the origins of the landó. The album Cumanana was
Take the clave son as an example. The binary representation
of the clave son timeline would be xooxooxoooxoxooo, or
1 Both the marinera and the zamacueca are representative rhythms in [1001001000101000], using Arabic numbers. The minimum time
Peruvian coastal music (also known in Peru as música criolla or creole
music) which is mostly practiced in Lima and its surroundings. The 2 The cajon—in Spanish the correct spelling would be cajón—is an Afro-

marinera, nevertheless, has different variations in the north coast of Peru Peruvian unpitched percusion instrument commonly referred to in English
such as the tondero and the marinera norteña. Both are related to the as “drum box”. Rafael Santa Cruz [9] and Denegri [10] discuss from
Chilenian cueca [5]. different perspectives the role of the instrument in Peruvian coastal music.
Proceedings of the 9th Conference on Interdisciplinary Musicology – CIM14. Berlin, Germany 2014

6 6
00 00 5 5
11 01 11 01 4 4
10 02 10 02 3 3
2 2
1 1
09 03 09 03 0 0
1 2 3 4 5 6 1 2 3 4 5 6
08 04 08 04 Samba Malató Variation 1
07 05 07 05 6 4
06 06 5 3
4
Samba Malató Variation 1 3 2
2 1
1
00 00 0 0
11 01 11 01 1 2 3 4 5 6 1 2 3 4 5 6
10 02 10 02 Variation 2 Zamacueca

09 03 09 03
Figure 2: Histogram of the inter-onset distances for the landó
08 04 08 04 cajon patterns.
07 05 07 05
06 06
Variation 2 Zamacueca 4. R HYTHMIC COMPLEXITY RELATED TO INTER - ONSET
AND ADJACENT ONSET DISTANCES
Quantifying rhythmic complexity is not a trivial endeavor. African
Figure 1: Four distinguished cajón patterns for the landó. and Indian rhythms are reputed by their performative complexity [2,
p. 107][13] (i.e., how difficult they are to play even by a trained
musician), while complexity can also be related to syncopation (a
Samba Malató x x x x x x o x o o x o concept further discussed in Section 7. Another way of understand-
Variation 1 x o x o x x o x o x o x ing a rhythm’s complexity is by its unpredictability. This section
Variation 2 x o x o x x o x x o x x relates the notion of unpredictability to a rhythm’s inter-onset and
adjacent distances.
Zamacueca o x o x x o o x o x x o
The inter-onset distance is the minimum number of pulses that must
be traversed from one of the pulses to reach the other one (either
Table 1: Binary representation of the most representative clockwise or anticlockwise). For the Samba Malató rhythm in
cajon landó patterns. Figure 1, the distance between the onset at position 04 and the onset
at position 07 is three. However between the onsets at positions 10
and 05 the distance would be 5 (moving counterclockwise), not 7.
Define the distance matrix D as a matrix containing every possible
unit (or time between pulses) in Western staff notation would be a inter-onset distance. Since the distance from onset i to onset j is the
16th note, since there are 16 onsets in the clave son timeline. A same as that from j to i, D is a symmetric matrix of K × K elements
rhythm R with N pulses and K onsets will hereafter be denoted by with zeros in its main diagonal—the distance from an onset to itself
R(N, K); the clave son would then be represented as R(16, 5). is zero. Figure 2 shows a histogram representation of all the inter-
Table 1 shows the main landó cajon patterns in binary notation. The onset distances contained in D—distances from i to j and from j
labels are identical to those used by Feldman [4, ex. 1 p. 215]. to i are counted only once—versus the frequency of occurrence
Since the cajon can played as to produce both a low-pitch bass of each particular distance for each landó pattern. An equivalent
sound and a high-pitched slap sound, both of these sounds have representation can be used for adjacent onset distances, but it is not
been represented with an x for all rhythms except for the zamacueca. included here for the sake of brevity. The following subsections
In the basic zamacueca pattern all of the pulses are onsets that are discuss different measures based on the histogram representation.
either high- or low-pitch sounds [9]. Because the high-pitch sounds 4.1. Entropy and the number of distinct inter-onset and
are more accentuated in the interpretation of the zamacueca, these adjacent onset durations
are represented with a x, and the bass sounds with a o.
The concept of entropy was first proposed by Shannon in the context
Circular representation is a natural extension of binary repre- of digital communication systems [14]. In the binary representation
sentation, since the binary sequence is supposed to repeat itself of a rhythm there are two symbols that can be “transmitted”, a x or a
indefinitely. Figure 1 shows the circular representation for the o. Consider two simple rhythms, the first one given by xxx and the
landó cajon patterns. The outer large circle represents time, and second one by xox. In the first rhythm the symbol o is never used,
each pulse is numbered starting from zero. The small black hence, we know what to expect after the first x, another two xx.
circles represent the onsets. Adjacent onsets (AO) are joined with This is not the case, however, for the second rhythm in which we
the thick black line forming a unique polygon for each rhythm have one o every three pulses. This adds an element of “surprise”,
(i.e., for each of the 2N possible binary combinations). The unpredictability, or complexity to the rhythm.
light gray lines show all the possible inter-onset (IO) connections.
Circular representation opens many possibilities for analysis based From the histograms in Figure 2, a new set of values equivalent to a
on geometry and computer science techniques. discrete probability density function can be defined as
In what follows, many of these techniques, comprehensively de-
frequency of occurrence of the i-th distance
scribed by Toussaint [2], are reviewed and reflected upon based pi = , (1)
on their application to the landó patterns. First the technique is sum of the frequency of occurrences of all distances
explained or defined, while the numerical results of its application
to the landó are summarized in Table 2. The Fourier transform N−1
so that it is true that ∑i=0 pi = 1. The entropy H of a given rhythm
is proposed as an alternative for the task of quantifying rhythmic can then be defined as
regularity, which to the knowledge of the authors is a novel appli-
cation of a mathematical tool extensively used in acoustics [12], H = − ∑ pi log2 (pi ). (2)
electronics, and telecommunications [3]. i
Proceedings of the 9th Conference on Interdisciplinary Musicology – CIM14. Berlin, Germany 2014

The entropy can be computed both for inter-onset and adjacent Direct Swap Many-to-many Chronotonic
onset histograms. Since it is an indirect measure of the number
SM V1 V2 Z SM V1 V2 Z SM V1 V2 Z
of distinct inter-onset/adjacent onset distances, these have also been
SM 0 8 15 7 0 4 4 6 0 8 10 12
counted. From Table 2 we notice that all four landó patterns have
V1 8 0 2 3 4 0 2 3 8 0 2 8
six different IO distances, while only the Samba Malató and the
V2 15 2 0 5 4 2 0 5 10 2 0 10
zamacueca reach an AO distance of 3. This is interesting since these
Z 7 3 5 0 6 3 5 0 12 8 10 0
two rhythms could be intuitively perceived as simpler compared to
variations 1 and 2. Though counting the number of IO distances did
not differentiate the rhythms between each other, the IO distance Table 3: Distance matrix using the directed swap, many-to-
entropy of variation 2 was the highest with a score of 2.55. As a many assignment distance, and chronotonic distance applied
reference, the entropy of a rhythm that spans six distances and with
a perfectly flat IO histogram is 2.59.
to the four landó rhythms: SM: Samba Malató, V1:
variation 1, V2: variation 2, Z: zamacueca.
4.2. Onset complexity and distinct distances
The complexity with which an onset contributes respect to all the
others can be measured by counting the number of distinct IO 5. D ISTANCE METRICS AND PHYLOGENETIC ANALYSIS
distances with which the onset contributes. This calculation can
be carried out for all of the onsets and added up, which will If we were to listen to a rock song, and next a salsa, we would
yield a different approach towards rhythmic complexity. Table 2 immediately notice that the two are based on very different rhythms,
shows that it is variation two that has the highest score. Though though both might have the same time signature. Can this perceived
musical perception is inherently subjective, it is worth mentioning musical difference between two rhythms be quantified? Computer
that variation two is often perceived as difficult to play and sing at science has several methods that are used on binary strings, and that
the same time for amateur Peruvian cajon players, compared for can therefore be applied to rhythms in binary/circular representa-
example to the zamacueca. It is left to future work to carry out a tion [2, ch. 33].
perceptual test on the complexity of interpretation of these rhythms. One of these methods is the directed swap distance. Directed swap
maps each element of the rhythm with the most onsets (i.e., the
4.3. Deepness dense rhythm) to the one with the least number of onsets—the
Deepness is the property by which the frequency of occurrence of sparse rhythm, with the condition that every onset in the sparse
each IO distance is different. The rhythmic pattern of variation rhythm must be connected to at least one element of the dense
1, for example, is a deep rhythm. The degree of shallowness of rhythm. Then, it calculates the distance of each pair of mapped
a rhythm can be measured by comparing its sorted frequencies of elements, and the total rhythmic dissimilarity results from the
occurrences sorted in ascending order to those of a deep rhythm, addition of these pair-distances. The mapping process is illustrated
also sorted in ascending order [2, p. 285]. From Table 2 we see that in Figure 3 for the Samba Malató and variation 1 patterns. While
the Samba Malató and the variation 2 patterns are tied at seven, for some cases directed swap may yield satisfactory results, it not
while the zamacueca is the least deep of all rhythms with a score of always matches the judgment of musicologists [2, p. 252], and from
eight. Figure 3 it is not difficult to see why. In some cases, two onsets that
are at the same time position are not mapped to one another, because
mapping can only happen in one direction. The alternative is the
S. Mal. Var. 1 Var. 2 Zamac. many-to-many assignment distance, in which mapping is done both
ways with the condition that no element of any of the two rhythms
Nr. of pulses 12 12 12 12 should remain unassigned (“non-mapped”).
Nr. of onsets 8 7 8 6 A third alternative is the chronotonic distance which is very clearly
IO and AO Distances explained for the case of Flamenco timelines [15, 16]. In order
Nr. disctint IO dur. 6 6 6 6 to calculate the chronotonic distance, a rhythm R = xooxxox, for
Nr disctint AO dur. 3 2 2 3 example, would be represented by the sequence of values of its
adjacent onset distances, s = [3, 1, 2]. By plotting the sequence s
Entropy – IO 2.51 2.40 2.55 2.52
against itself (a technique known as TEDAS notation [2, ch. 33]) we
Entropy – AO 1.15 0.65 0.99 1.52 obtain a continuous-time representation of R which we can call r(t).
Unique nr. of IO dist. 40 33 41 26 The chronotonic distance dchron between two rhythms R1 and R2 is
Shallowness 7 0 7 8 the sum of the areas between the two curves, or mathematically
Tallness 6 6 6 6 speaking Z
Syncopation dchron = (r1 (t) − r2 (t)) dt (3)
Off-beatness 3 3 3 2
Syncopation – score 18 16 19 16 Table 3 shows the distance matrices using all three distance metrics.
Syncopation – worst 24 22 24 20 Note that the distance matrix from the many-to-many assignment
Syncopation – best 16 14 16 12 distance is significantly different than that of the directed swap
Syncopation – norm. (%) 25 25 37.5 50 distance. In the following subsection, the matrices of the many-
Dir. swap 14 14 17 12 to-many assignment and chronotonic distances are used to generate
phylogenetic trees to better understand the relation of the four landó
Many-to-many 14 14 17 12
patterns to each other.
Chronotonic 50 50 52 44
Regularity 5.1. Phylogenetic analysis
Reg. - Fourier 1.82 1.83 2.08 1.58 The use of phylogenetic trees not only allows the graphic repre-
Reg. - dir. swap 8 8 9.25 6.75 sentation of the distance matrices in Table 3; it is also used to
Reg. - Many-to-many 8 8 9.25 6.75 identify clustering relationships and information that can be used
to reconstruct ancestral rhythms [2, ch. 35][15]. A technique
Reg. - Chronotonic 24.5 23 25 20 called SplitsTree has been used to generate a graph from the
Oddity 4 5 2 0 distance matrices which in this cases perfectly matches the distance
matrices [17]. In the SplitsTree technique new nodes are inserted
Table 2: Summary of different measures applied to the four if necessary to improve the distance matching. These nodes may
cajon patterns for the landó. suggest the presence of ancestral rhythms. In the case of the landó,
as discussed in Section 1, there are very little traces of what could
have been the original ancient rhythm that was once played, and
Proceedings of the 9th Conference on Interdisciplinary Musicology – CIM14. Berlin, Germany 2014

Directed swap ρ1 1 1 1 1 1 1 1 1 1 1 1 1
Zamba Mal. ρ2 1 0 1 0 1 0 1 0 1 0 1 0
ρ3 1 0 0 1 0 0 1 0 0 1 0 0
ρ4 1 0 0 0 1 0 0 0 1 0 0 0
Var. 1
ρ6 1 0 0 0 0 0 1 0 0 0 0 0
Many-to-many ρ12 1 0 0 0 0 0 0 0 0 0 0 0

Zamba Mal. ∑ ρi 6 1 2 2 3 1 4 1 3 2 2 1
∩i6=1 ρi 0 1 0 0 0 1 0 1 0 0 0 1
Var. 1
Table 4: Matrix containing all regular rhythms for N = 12
Figure 3: Comparing the directed swap and the many-to- pulses. The two last rows are used for the syncopation and
many distance measurements for the Zamba Malató and Var. off-beatness measures in Section 7.
1 cajon patterns. The direct swap distance measurement
yields 8, while the distance is only 4 for using many-to-many 6. R HYTHMIC R EGULARITY
assignment.
Using circular representation, regular rhythms may be simply
described as regular polygons. Let a regular polygon (or rhythm)
Zamacueca with a side length of l be denoted as ρl . For ρl to be considered a
regular rhythm, the number of pulses N divided by l must yield an
integer number. Hence the number of regular rhythms in N pulses is
equal to the number of dividers of N. Table 4 shows all the regular
Many-to-many rhythms for the landó, where N = 12. Two additional rows are
2.5
included at the end that will be used to quantify syncopation and
off-beatness in Section 7.
0.5 6.1. Average distance to regular rhythms as a measure of
Var. 1 1
2.5 Samba regularity
1 A proposed measure for rhythmic regularity is to average the
0.5 Mal.
directed swap distance of rhythm R to all regular rhythms except
1 ρ1 and ρN . Despite the limitations of the directed swap distance
already mentioned in Section 5, there is an interesting fact that
Var. 2 might not have been taken into consideration before. Calculating
the regularity of ρ2 , ρ3 , ρ4 , ρ6 individually based on the principle
of averaging the directed swap distance yields 5, 3.5, 3.5 and 5,
respectively. But should not every regular rhythm have the same
Zamacueca value of regularity according to a rhythmic regularity measurement?
Despite this inconsistency the metric was computed for the four
landó patterns using the directed swap distance, many-to-many
assignment distance and the chronotonic distance. As Table 2
shows, the three yield the same main result: the most complex
Chronotonic 6 pattern is variation 2, while the Samba Malató pattern comes in
second place, followed by variation one and finally the zamacueca,
which is the most regular of all according to the average directed
6 Samba swap distance method.
2 Mal.
2 6.2. The Discrete Fourier Transform and regularity
Var. 2
Var. 1 The Fourier Transform is certainly not new to the field of music
and acoustics. Its usefulness resides in its ability of decomposing
Figure 4: SplitTrees corresponding to the many-to-many and a signal—a discrete or continuous sequence of values ordered in
chronotonic distance metrics for the cajon landó patterns time—in to its periodic components [3].
Every signal used in music and acoustics can be decomposed into
(100% fit accuracy). a sum of periodic functions (i.e., functions that repeat themselves
identically after a certain period of time) which in the case of the
Fourier transform are cosine functions (considering real signals
nothing indicates that this rhythm would be similar to any of the only). Hence, an audio signal can be reconstructed from the sum
four patterns that this work examines. In the case of the many- of cosine signals after there frequency, amplitude, and phase is
to-many SplitsTree, the three inserted nodes (i.e., the white nodes) determined. From this explanation it would seem that the Fourier
have the same total distance to all other nodes. The chronotonic tree transform is the perfect tool to find out what regular cycles added
seems to suggest that there is a rhythm, located exactly in the middle together build a rhythm, and hence determine its regularity.
of the tree (the only white node) which can be the ancestor of all The Discrete Fourier Transform (DFT) of a discrete-time signal (a
the other rhythms, in the sense that its distance to all four rhythms signal that has been sampled in time and that four our purposes can
would be minimum. It is compelling that neither the Samba Landó considered to be a vector) yields also a discrete set of values with a
(the first landó ever recorded) nor the zamacueca are anywhere length equal to the length of the discrete-time signal, which in our
near the center in neither of the two trees. The zamacueca, in case is N, the number of pulses. Its equation is given by
particular, is considered to be the ancestor of the marinera. It may
be that one can find a connection between the zamacueca and the 1 N−1
other three patterns not by comparing them to the basic zamacueca Γ [k] =
N ∑ R[n]e−i2πkn/N (4)
pattern, but by listening to all recordings and notate the variations n=0
that where commonly used when the Afro-Peruvian revival started
and continued. where Γ [k] is the k-th sample of the DFT, e corresponds to the
Proceedings of the 9th Conference on Interdisciplinary Musicology – CIM14. Berlin, Germany 2014

√
exponential function and i = −1. Figure 5 shows the amplitude 1
and phase of the DFT, where the amplitude is given by [3] ρ1 ρ2 ρ3
0.8 ρ4 ρ6 ρ12

|DFT{R}|
q
0.6
|Γ [k] | = ℜ {Γ [k]}2 + ℑ {Γ [k]}2 (5)
0.4
and the phase by
0.2
∠Γ [k] = arctan (ℑ {Γ [k]} /ℜ {Γ [k]}) (6) 0
0 2 4 6 8 10 12
where ℜ{•} and ℑ{•} denote the real and imaginary operators, Frequency index
respectively.
Figure 5 shows the DFT of the regular rhythms for N = 12. For 1
the regular polygon of side l, ρl , the only non-zero values of the
0.5

∠DFT{R}
amplitude of the DFT are at multiples of N/l, and these values are
identical to each other. Furthermore, note that for all of the regular
rhythms in Figure 5, the sum of the non-zero values adds up to one. 0
The phase of the DFT is zero for all the regular rhythms because
none of them is delayed in time. −0.5
Applying the DFT to the four landó patterns we observe a very −1
different response. Figure 6 shows that for none of the patterns do 0 2 4 6 8 10 12
the amplitudes of the DFT add up to one. Note that the zamacueca Frequency index
has zero-valued amplitude DFT coefficients every other frequency
index. This is consistent with the fact that the zamacueca is six
pulses long, so it is present two times in the 12-pulse sequence. Figure 5: Discrete Fourier transform of the regular rhythms
Figure 1 also shows that there is an axes of odd-symmetry by tracing
a straight line between pulses 04 and 10. We see that the DFT that fit in N = 12 pulses.
is useful to identify regular patterns and axes of symmetry for a
rhythm.
Based on the preceding discussion, the authors propose the follow-
ing regularity measure g based on the discrete Fourier transform,
7. S YNCOPATION AND OFF - BEATNESS
Syncopation is certainly a basic concept in music that is not easy
N−1 to define in an absolutely precise (i.e., quantitative) manner [2,
g= ∑ |Γ [k] |. (7) p. 67]. Lerdahl and Jackendoff presented a metrical hierarchy for
k=0 rhythms with sixteen pulses. The method can be applied to the
landó twelve-pulse patterns by generating the syncopation matrix
The proposed measure is elegantly simple, and an upper bound for shown in Table 4. The rows of the syncopation matrix S contain
g is not difficult to derive taking into account Parseval’s theorem [3] all the regular rhythms for a 12-pulse sequence (or equivalently,
all regular polygons in 12-pulse circular cycle; see Table 4). By
N−1 N−1 adding the rows of this matrix we obtain the weight vector ws =
∑ R[n] = N ∑ |Γ [k] |2 = K (8) [6, 1, 2, 2, 3, 1, 4, 1, 3, 2, 2, 1] (see the row labeled ∑ ρi in Table 4). A
n=0 k=0 measure SC for syncopation can then be defined as
which bounds the energy of the DFT, and since binary representa-
tion is used, this bound is equal to the number of onsets. N−1
Then, to find the upper bound the general problem can be posed as SC = ∑ ws [i]R[i]. (9)
follows: given that x12 + x22 + · · · + xN
2 = a2 , where a is a real and i=0
positive constant, and xi for all i ∈ {1, 2, . . . , N} is real and positive,
what values of xi maximize the sum x1 + x2 + · · · + xN ? It can √ be The higher SC, the less syncopated the rhythm is. By sorting the
shown that this sum is maximized when x1 = x2 =√ · · · = a/ N. elements in ws ascendent/descendent order, and adding only the
Hence, the maximum value the sum can attain is a N. For the first K values we can find the lowest/higher values of SC for a
case of the DFT, from (8), we find that the upper bound equals
√ rhythm with N pulses and K onsets. These values are also included
K. Then, if a rhythm R(N, K) is regular, its regularity measure in Table 2. An additional value (expressed as a percentage) is
g equals one. If it is not, then the computed by normalizing SC respect to the highest and lowest
√ maximum value of irregularity values. This indicates how unsyncopated a rhythm is considering
that it can reach is bounded by K. The reader may have noticed
that this measure does not take into account the phase of the DFT that it has N pulses and K onsets, as opposed to only considering
coefficients, which means that even if a rhythm is delayed in time SC, which is not a relative measure.
(i.e., rotated) the value of g will remain the same. We can say It is rather surprising that variation 2 has a slightly higher SC
that g quantifies the regularity of the necklace of R (the set of all score than the Samba Malató pattern. But when considering the
the rhythms that can be obtained by rotating R one pulse to N − 1 normalized syncopation in percentage, the zamacueca turns out
pulses). Computing the regularity of a necklace using an average to be the least syncopated with a score of 50. The authors are
distance as in Section 6.1 could become computationally expensive aware that this measure must be taken with caution, in the sense
if the number of pulses N is large, since several comparisons would that thorough perceptual tests would have to be carried out with a
have to be carried out for each shift of the necklace. Using the representative population of subjects both with and without musical
regularity measure based on the DFT, however, this is trivial since training in order to validate the measure.
very efficient algorithms exist to compute the Fourier Transform [3], Another measure that Toussaint proposes that is indirectly related to
and the measures must only be calculated once for any shift of the syncopation is the off-beatness measure. The method be understood
necklace. as counting the number of onsets that fall in an unaccented position.
From Table 2, we see that the Fourier regularity measure also It can be defined as multiplying every element in rhythm R with
ranks the zamacueca as the most regular of all the four patterns, a vector resulting from the column-wise intersection of rows two
while variation 2 is ranked as the most irregular. According to to six of the matrix presented in Table 4. As shown in Table 4,
this measure, variation 1 is slightly more irregular than the Samba in the row labeled ∩i6==1 ρi , for N = 12 this off-beatness vector
Malató pattern, with a score of 1.83 compared to 1.82. is voffb = [0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1]. The off-beatness measure is
Proceedings of the 9th Conference on Interdisciplinary Musicology – CIM14. Berlin, Germany 2014

0.8 between rhythms.

S. Mal. Var. 1
0.6 Var. 2 Zamac. ACKNOWLEDGMENTS
|DFT{R}|

0.4 The authors would like to acknowledge the kind contribution of

Peruvian musicians such as Victor Garcı́a V., Lalo Izquierdo,
0.2 Maria del Carmen and Mano Madera, Pancho Vallejos, “Pepe”
Villalobos and his family, Juan Cotito Medrano, Manongo Mujica,
0 and guitarrist Adolfo Zelada. Our special gratitude to Rosa Elena
0 2 4 6 8 10 12 Vásquez for the interesting discussions, as well as to Carl H.
Frequency index Waadeland for his insight. Our gratitude also goes to Per Jørgen
and Steffen Kirkness for encouraging this adventurous path into
200 Peruvian music.
100
∠DFT{R}

R EFERENCES
0 [1] J. F. León Quirós: The aestheticization of tradition:
professional Afroperuvian musicians, cultural reclamation,
−100 and artistic interpretation. Ph.D. thesis, The University of
Texas at Austin, 2003.
−200 [2] G. T. Toussaint: The geometry of musical rhythm: what makes
0 2 4 6 8 10 12
a ”good” rhythm good? CRC Press, Taylor & Francis Group,
Frequency index Boca Raton, FL, 2013.
[3] J. G. Proakis and D. G. Manolakis: Digital signal processing:
principles, algorithms, and applications. Prentice Hall, Upper
Figure 6: Discrete Fourier transform of the landó cajón Saddle River, N.J., 1996.
patterns. [4] H. C. Feldman: The Black Pacific: Cuban and Brazilian
Echoes in the Afro-Peruvian Revival. In Ethnomusicology,
volume 49(2):206–231, 2005.
hence defined as [5] W. Tompkins: Las tradiciones musicalesde los negros de
N−1
la costa del Perú. Centro de Música y Danza de la
Pontificia Universidad Católica del Perú, Centro Universitario
OFFB = ∑ voffb [i]R[i]. (10)
de Folklore de la Universidad Nacional Mayor de San Marcos,
i=0
2011.
This measure is of questionable usefulness because it does not [6] H. Feldman: Black Rhythms of Peru: Reviving African
consider the number of onsets. A rhythm with N = K = 12 will Musical Heritage in the Black Pacific. Wesleyan University
score the maximum value for the off-beatness measure, but it can Press, 2006.
hardly be considered to be off-beat. The shortcomings of the [7] M. M. Rojas: Positioning the Past in the Present: Feldman’s
measure are also notorious for the landó, since from Table 2 we “Memory Projects.” A Review of Heidi C. Feldman’s Black
obtain an off-beatness value of three for all rhythms except the Rhythms of Peru: Reviving African Musical Heritage in the
zamacueca, which scored a two. Black Pacific (Middletown, CT: Wesleyan UP, 20. In A
Contracorriente, volume 5(1):274–284, 2007.
8. C ONCLUSION [8] J. F. León: Mass Culture, Commodification, and the
Consolidation of the Afro-Peruvian ”Festejo”. In Black Music
This work has studied four drumming patterns for the Afro- Research Journal, volume 26(2):213–247, 2006.
Peruvian rhythm landó using computational and mathematical
methods related to rhythmic complexity, dissimilarity, syncopation, [9] R. S. Cruz: El cajon afroperuano. RSANTACRUZ E.I.R.L.
regularity, and number of perceived durations. Phylogenetic Ediciones, Lima, 2004.
analysis corroborates the little connection that these patterns have [10] M. A. Denegri: Cajonistica y Vallejistica. Editorial San
to each other, in the sense that none of them can be considered to Marcos E.I.R.L., Lima, 2009.
be an ancestor of the other three. Possible future work would be to [11] Y. Liu and G. T. Toussaint: Mathematical Notation,
transcribe drumming patterns (and especially the variations of these Representation, and Visualization of Musical Rhythm: A
patterns) from early recordings of the Afro-Peruvian revival. It Comparative Perspective. In International Journal of
might for example be, that the vast array of variations surrounding Machine Learning and Computing, volume 2(3):261–265,
the zamacueca provides the ancestral rhythm that is missing in the 2012.
chain. [12] G. Loy: Musimathics: The Mathematical Foundations of
The result from the computational measures suggest that the second Music. MIT Press, 2011.
landó variation is the most complex and irregular, while the first [13] K. Agawu: Representing African Music: Postcolonial Notes,
variation is the most syncopated. Despite its deceivingly simple Queries, Positions. Routledge, 2014.
structure, the Samba Malató pattern has a rather high syncopation
[14] R. Togneri and J. Christopher: Fundamentals of information
value. The zamacueca is not prominent in any of the measures,
theory and coding design. CRC Press, 2003.
possibly because it is a supporting rhythm in the ensemble. Only the
first landó variation is deep, which means that its onsets encompass [15] J. M. Dı́az-Báñez, G. Farigu, F. Gómez, D. Rappaport, and
a wide variety of distances that may be perceivable by the listener. G. T. Toussaint: El Compás Flamenco: A Phylogenetic
Analysis. Bridges Conference, 2004.
A new measure for rhythmic regularity based on the discrete Fourier
transform has been presented. The measure is easy to calculate, and [16] C. Guastavino, F. G. Mez, G. Toussaint, F. Mar, and
all regular rhythms score one, while the measure’s value increases E. G. Mez: Measuring similarity between flamenco rhythmic
with irregularity. The measure has an upper bound that depends patterns. In Journal of New Music Research, pages 129–138,
only on the squared root of the number of onsets. 2009.
The field of computational ethnomusicology is relatively new, [17] D. H. Huson and D. Bryant: Application of phylogenetic
and it seems promising to continue using computer science and networks in evolutionary studies. In Molecular Biology and
mathematical techniques to gain a better understanding of the Evolution, volume 23(2):254–267, 2006.
mysteries surrounding rhythm in general as well as the genealogy