Analysis of Musical Interpretation: Herbert von Karajan

Analyses of musical interpretations play an increasingly important role in recent musicology. Until now, research has been focused on soloistic instrumentation (piano music especially), while larger scale works (such as orchestral music) have hardly been considered. The main reason for this is the difficulty in the sufficient differentiation of individual instrumental groups, which is indispensably vital in the highly complex orchestral works of the 19th and 20th centuries. Using the latest computer- assisted methods, which are to be further developed and refined within the scope of this project, the option of an empirical analysis has been unlocked. The application of these methods will initially focus on the works of Herbert von Karajan, but will subsequently open up a general foundation for research into the interpretation of orchestral music (including that of other conductors). There are several reasons for focusing on Karajan. On the one hand, he is a conductor who made use of the at the time newest recording techniques and has achieved more than almost any other pioneer in the field of audio documentation. On the other hand, there exists in this case a very large collection of (in large part still unpublished) music-related and non-musical resources: approximately 1,800 audio recordings from 1939 to 1989, about 100 rehearsal recordings, 700 studio productions, roughly 1000 live concert recordings, concert films, rehearsal documentations, letters, and recorded conversations concerning the details of vinyl and CD recordings of works conducted by Karajan himself. The following methods have been chosen in order to approach a specific profile of the conductor Karajan: the comparative analysis of Karajans interpretations of one and the same orchestral work at different times the comparative analysis of tempi, articulation, and instrumental balance in selected recordings conducted by Karajan as well as by other conductors of his time the music-aesthetical profiling of Karajan as a conductor and media artist the development of computerized methods for the relationship between practical performances and scores (performance and sheet music) as well as for the comparative analyses of the recordings of individual works the development of possibilities to automatically (computerized) discover and categorise sections that are characterised by marked differences in articulation, tonal balance, tempi, treatment of the individual instrument groups, etc. in comparative recordings in order to demonstrate Karajans specific interpretation profile. This decidedly interdisciplinary and data-based research approach is carried out in close cooperation between musicology and computer science and will lend new and ground-breaking impulses to the still relatively young musicological discipline of interpretation research.

The project set out to accomplish a data-driven study in musical interpretation, connecting information from both manually and algorithmically derived annotations with approaches from traditional musicological approaches (consideration of literature, theoretical analyses, source studies), performance studies, and film studies. Although the project was conceived as a case study on Austrian conductor Herbert von Karajan (1908-1989), we wish to demonstrate the promise of such data-intensive, interdisciplinary approaches to musicological studies of expressive performance, and help to establish this as a new open standard approach in musicology. Our approach went beyond already existing projects (e.g., "Mazurka", "PETAL", "Gleitende Tonhöhen") in that we focus exclusively on orchestral music conducted by a specific performance artist for the first time. In addition to musicological and computational research results, the project had access to hitherto ignored sources from the "Eliette and Herbert von Karajan Institute" and the Archive of the Salzburger Festpiele (Salzburg), and the Berlin Philharmonic archives. The project is based on four cornerstones: First, we determined a methodological framework for selecting, conceptualizing, and annotating orchestral music from different time periods, including the development of an onset-detecting algorithm and an extensive study to compare the accuracy of manual and automatic annotation of orchestral pieces. The corpus of the pieces itself developed over time, though the basic criteria - namely the focus on Karajan's core repertoire, i.e., canonic composers from the 18th-20th centuries, but excluding works with vocal components (esp. music theater) -, have also been decided on at the earliest stage of the project. Second, we carried out comparative analyses of performances of these orchestral works by Karajan and other contemporary conductors, focusing on various parameters and aspects such as tempo, dynamics, timbre, and articulation, deploying different methods and tools ranging from close listening to visualization (e.g., spectrograms) to manual (Sonic Visualiser) and automatic annotation (Algorithm). Third, we analyzed several of Karajan's music films with both quantitative and qualitative methods to determine the respective visual and dramaturgical concepts, and to deepen our understanding of Karajan's production as a media artist. Lastly, the inclusion of various secondary sources - writings and (voice) recordings by Karajan, archive sources, orchestra parts of the Berlin Philharmonic, and interviews with contemporary witnesses, among others - allowed for a historically and socially sensitive contextualization of our findings and for deepening our understanding of Karajan's artistic concepts. So far, results of the project have been published in three books and (currently) 16 papers, and have been presented at several conferences, in poster sessions, newspaper articles, and two radio broadcasts.