Authorship and attribution challenges in early music research

Authorship and Attribution in Early Music Research

Scholars of early music face a problem that is one of an important nature and one that endures and that is the question of who is actually the composer of music surviving today and in the form of various written manifestations from early period of music. The work of Bruce Haynes entitled: "The End of Early Music: A Period Performer's History of Music for the Twenty-First Century" states that the "true Canonic musical experience requires knowing who wrote the piece one is hearing, knowing when they lived and knowing where they fit in the hierarchy of the Pantheon." (Haynes, 2007)

Haynes writes that John Spitzer "…for his dissertation…traced the histories of works attributed to great composers that were later proved spurious; they disappeared from the repertoire when they lost their pedigrees." (2007) Spitzer made the observation how "…when in 1964 they were re-ascribed to Hofstetter, Haydn's Opus 3 string quartets began to disappear from the repertoire." (in Haynes, 2007) Haynes states the opinion that it appears that "the label…is more important than the product." (2007)

I. Challenges of Attribution and Authorship of Musical Compositions

Haynes emphasizes the importance of the authorship of music by stating "When the listener suddenly exclaims to himself 'Beethoven, 'he does not than just hang a label on the music. With one word he calls up an entire context of Beethoven's biography, Beethoven's other works, Beethoven's patrons, early 19th century Vienna, what critics have said about Beethoven, and so on. The listener hears the remainder of the Fourth Symphony in a different frame of mind, because knowing its authorship has greatly enriched the context in which the work is perceived and appreciated." (Haynes, 2007) Knowing the authorship of a piece of music serves to satisfy the need "to put art into its historical context (since it is no longer in one). It also contributes to our sense of a piece's identity, since each performance is a little different (and in some cases so different as to render the music unrecognizable)." (Haynes, 2007)

According to Haynes 'musicology' came into being during the 'Romantic' period and resulted in a plethora of biographies and other collected works of various composers being produced and at the same time "it gave the Romantics an unholy obsession with attributing works, and who influenced whom." (Haynes, 2007) Another factor driving such need for authorship identification was that of printed and typed programs at various performances in which the name of the composers were published along with the works to be performed. Programs are stated to serve the function "of announcing their (musical compositions) historical credentials especially for those who might recognize an otherwise unintelligible sound mass as a rendition of Purcell or Beethoven." (Haynes, 2007)

Haynes relates that it is just as one might suppose in that authorship was less of an issue in the days prior to "…the 'genius personality' became a dominant issue." (2007) the first four centuries in which music in England was notated there was not ascription whatsoever to the composer and even after that "attribution of songs often seems not to have been for the music but for the texts." (2007) it was not until the ending of the fourteenth century that "manuscripts start regularly giving composers names…[and]…attributions were sporadic as late as the sixteenth century." (Haynes, 2007) Haynes writes that even in today's world "attribution is not an important issue except in Canonic music." (2007)

II. Kranenburg -- Attribution by Quantifying Compositional Strategies

The work of Peter van Kranenburg entitled: "Composer Attribution by Quantifying Compositional Strategies" states that in order to 'describe a musical style, or differences between styles, or the historical development of certain styles, a theory of style is necessary. This applies to "traditional" descriptions of musical style as well as studies in which tools and algorithms from information technology are used." (2006)

Kranenburg reports a pilot experiment in which a data set is assembled with "16 fugues for organ that are listed in the catalog of compositions of Johann Sebastian Bach. Of six of these fugues the authorship has been disputed. Also five fugues of his eldest son, Wilhelm Friedemann Bach, and eight of his most important student, Johann Ludwig Krebs, are incorporated. So we have a three-class data set." (2006)

Kranenburg states that each of the compositions are "segmented using a segmenting method…so each composition is represented by a "cloud" of points. The Fisher-transformation…can be used to project the data points onto a two-dimensional space in such a way that the classes are optimally separated.

Figure 1

Source Kranenburg (2006)

This projection is stated by Kranenburg to be depicted in the background of the figure labeled Figure 1 in this study which demonstrates that the "compositions of each composer do form a cluster." (2006) in figure one it is indicated "…where the data points of the disputed fugues projected. Some interesting observations can be made. The F minor fugue BWV 534.ii, is projected among the fugues of J.L. Krebs. This fugue has been ascribed to W.F. Bach. With the current result, that ascription can be rejected. An ascription to J.L. Krebs seems more likely. A suggested composer for BWV 536.ii is J.P. Kellner. If this is true, Keller's style resembles more the style of J.S. Bach than that of the other two composers. BWV 537.ii is said to be composed partly by J.S. Bach (bar 1 -- 40) and partly by J.L. Krebs. The first part is projected among the works of J.S. Bach indeed. The second part however, is outside of both the Bach-region and the Krebs-region. The ending of the fugue is in the region between J.S. Bach and Krebs. This does not fully support the hypothesis, but it shows that a large part of the fugue is not Bach-like. Also Bach's authorship of the fugue in C minor, BWV 546.ii, has been doubted. The current evaluation shows us that, with respect to the styles of W.F. Bach and J.L. Krebs, this fugue has the characteristics of the style of J.S. Bach. The fugue in D minor, BWV 565.ii, the second part of the most famous organ work in existence, is not projected among the other compositions of Bach. This confirms the doubts expressed…" (Kranenburg, 2006)

Kranenburg concludes by stating that since not all candidate composers are represented in the data set that the current results "…don't offer enough evidence to draw conclusions about the authorship of the involved compositions. It is, however, clear that the proposed method is very helpful in finding hypotheses about differences in personal styles and thus for studying authorship problems." (Kranenburg, 2006)

IV. Composer Identification Based on Typical Pattern Detection

The work of Salas and Gelbukh (nd) entitled: "Musical Composer Based on Detection of Typical Patterns in a Human Composer's Style" states that the advance in informational technology and the use of computational tools "make it possible to develop models of music in order to automate composition processes and eventually build systems that automatically generate music." (nd) the following are stated to be possible applications of such systems:

(1) Creation of new styles of music by finding patterns of different styles and mixing them, helping people to compose music.

(2) Providing tools that allow the user to edit the compositions generated by the system, resulting in a user composition, but which contains patterns that great musicians create and thus, people without musical knowledge can make good music.

(3) Enabling computers have the capacity to carry out a process until now reserved for humans. Making this, machines will get human characteristics creating another way of human-machine communication.

(4) Breaking the monopoly of large music companies that currently decide what music is to be listened to.

(5) Having machinery for the generation of live music for restaurants, offices, shops, etc. with compositions created in real time by indefatigable musicians.

(6) Providing tools that allow children from a very young age to have direct contact with the process of musical composition, which according to studies stimulates the mind for better performance in all other human activities. (Salas and Gelbukh, nd)

Salas and Gelbukh state that it is very important to gain and understanding that will result in the generation "as much resemblance to the examples with which the system is trained 'when using supervised learning techniques." (Salas and Gelbukh, nd) One of the primary aspects of a melody are characteristics relating to the "pattern of the sequences of notes used for each author. These patterns of transition can be measured statistically to determine the probability of moving from one state to another in a Markov process, what can be used to generate music which reflects these patterns of transition making it similar." (Salas and Gelbukh, nd)

Salas and Gelbukh report that the evolving computerized composition systems have "proved to be a versatile tool that can be used to model various phenomena of reality. In the case of our system we have obtained satisfactory results and the opportunity to hear compositions forever different. The development of this work is open as part of an overall process of composition and will be more complete if it manages to inspire a person to engage in development of computer music." (Salas and Gelbukh, nd)

Salas and Gelbukh state that they are currently in the process of "…developing systems to improve the different sounds and effects that can be obtained, in addition to using matrices of 3, 4 or n dimensions, which may reflect the many variables involved in a musical work." (nd) Additionally stated is that Salas and Gelbukh are extremely "…interested in the development of models that can generate musical compositions that not only reflect the patterns of typical sequences of sounds, but also incorporate the emotional content of music."( Salas and Gelbukh, nd)

The work of Backer and Kranenburg (2004) entitled: "On Music Stylometry -- a Pattern Recognition Approach" reports a study of various experiments in which methods of statistical pattern recognition are applied for music style recognition and disputed musical authorship attribution" Specifically stated is that as follows:

"Values of a set of 20 features (also called "style markers") are measured in the scores of a set of compositions, mainly describing the different sonorities in the compositions. For a first study over 300 different compositions of Bach, Handel, Telemann, Mozart and Haydn were used and from this data set it was shown that even with a few features, the styles of the various composers could be separated with leave-one-out-error rates varying from 4% to 9% with the exception of the confusion between Mozart and Haydn which yielded a leave-one-out-error rate of 24%. A second experiment included 30 fugues from J.S. Bach, W.F. Bach and J.L. Krebs, all of different style and character. With this data set of compositions of undisputed authorship, the F minor fugue for organ, BWV 534 (of which Bach's authorship is disputed) then was confronted. It could be concluded that there is experimental evidence that J.L. Krebs should be considered in all probability as the composer of the fugue in question." (Backer and Kranenburg, 2004)

The work of Liu (2002) entitled: "Modeling music as Markov chains - composer identification" reports a research study in which it is posited that music "…can be thought of as random processes, and music style recognition can be thought of as a system identification problem. Then, a general framework for modeling music using Markov chains is described, and based on this framework, a two-way composer identification scheme is demonstrated." (Liu, 2002) This work reports a study in which the statistical reliability of human listening skills ensures the correct identification of the composer of a work will be examined.

Liu (2002) relate that this scheme is one that "…utilizes the Kullback-Leibler distance as the metric between distribution functions, and it is shown that under the condition when the marginals are identical, the scheme gives maximum likelihood identification." (2002)

Computational models of music styles are addressed and it is related that the steps of building a Markov chain model for a music style include the following steps:

Step 1: Define the repertoire of that style.

Step 2: Encode all the works of that repertoire.

Step 3: Define the state space that consists of musical events.

Step 4: From all the encoded works available in that repertoire, calculate the matrix of conditional event-transition probabilities.

Step 5: Let be the Markov chain corresponding to {P, S}. Call Ca a Markov chain model for that particular style of music. (2002)

According to Liu, Step 1 can be difficult because "musical styles are often of a vague notion to human beings. Is Beethoven's Moonlight Sonata classical or romantic? Here, it is supposed that the human researcher somehow clearly define the repertoire, anyway." (2002) as well, Liu state that the second step "…is not quite straightforward, either. It all depends on what musical features are relevant. Data-Entry is another practical difficulty in this step, because it takes forever if the repertoire is huge." (2002)

The third step is referred to be Liu as "huge" and one that involves "…defining an interesting prominent feature associated with that style, and quantifying that feature in a clever way. The state space consists of all the possible values to which the feature is quantified." (2002) Finally Liu state that the fourth step or the 'transmission matrix' "…is obtained by normalizing the histogram of state-transition instances in all the works of that repertoire." (2002)

Selfridge -- Field and Sapp relate that the chain models of Markov "..are built up for Mozart's and Haydn's scale degree class transitions" and that the repertoire is comprised by the following:

(1) all of their string quartets,

(2) 100 movements from Mozart and (3) 212 movements from Haydn.

(4) the scale degree is defined relative to the tonics, i.e., all works are transposed into the same key.

(5) Also, for the sake of reducing the dimensionality of transition matrix, the system does not distinguish between the same degree class at different octaves. The system does not distinguish between major keys and minor keys, either.

(6) Then, the two-way identification tests based on Kullback-Leibler distances were conducted for each of the 4 voices. The tests were conducted in a bootstrap manner. In other words, it is always assumed that the computer has been exposed to the all the works in the repertoire except the one to be identified. (2002)

The following figure lists the results to the Mozart vs. Haydn identification tests reported by Selfridge -- Field and Sapp (2002)

Figure 2

Mozart vs. Haydn identification test

Part

Mozart

Haydn

ViolinI

68.0%

64.2%

ViolinII

58.0%

64.2%

Viola

61.0%

53.8%

Cello

57.0%

52.8%

Source: Liu (2002)

Repertoires, in a controlled experiment were defined as "two randomly assigned and mutually exclusive lists of quartets. Results of this test are shown in the following figure 'random repertoires identification tests'.

Figure 3

Random repertoires identification tests

Part

Random a

Random B

ViolinI

42.6%

56.1%

ViolinII

44.5%

53.4%

Viola

47.4%

51.9%

Cello

43.1%

53.4%

Source: Liu (2002)

It is reported that human listening tests conducted via the Internet and through the instrument of a survey involves a user being played the "MIDI piano version of, a quartet from either composer by equal chance, and is asked to identify the composer. As of June 11, 2002, web users average 59.0% of accuracy in 1865 attempts. How experienced the users are with the repertoire ranges from novices of classical music to string players that have played some of the works." Source: Liu (2002)

Summary and Conclusion

It is clear that labels do carry with them meaning that is and should well be applied in listening to musical composition for to fail to integrate the historical information and the time period information as well as the composer with the listening experience will leave the listener a little short of the pure enjoyment of the music for what it is as written in the time of the composer who composed the musical work. This study has shown that the accuracy of the human listener is not adequate although the human listener appears to be correct more than half of the time. This leaves this researcher wondering which of the musical compositions that have yet to be tested for accuracy have attributed wrongly the authorship to individuals who indeed were not the composers of those works and wondering then indeed who the real composer of the work might actually be. There is much more in the way of study in this area of attribution of authorship in musical composition prior to the late 18th and early 19th century, when it became important to know who the composer of music played in public performances indeed were.