Formatted Sound
Johan Malmstedt · M/C Journal · 2024
Locating the Format What is format radio? At a glance, the answer might seem simple. In common parlance, the concept is often presented as a pejorative counterpart to non-commercial broadcasting. However, to render the concept synonymous with commercial broadcasting neglects its historical specificity. Previous research has demonstrated the nuanced factors at stake in the selection of music at specific broadcasting stations (Ahlkvist and Fisher 301-325). Beyond economic structures and conditions, however, remains the matter of whether we can posit an aesthetic expression that epitomises format radio. Previous research has focussed predominantly on semantic source materials and theoretical propositions to hone in on the question. However, my wager is that the signal content itself can help us reveal something about the nature of formats. To pursue such a task, the Swedish case offers a promising possibility. Swedish media archives provide the opportunity to study how, and if, formatting tendencies can be detected beyond the realm of commercial broadcasting. Unlike many of its global counterparts, Sweden maintained a public service radio monopoly until the late twentieth century. Throughout the 1980s, experiments were done with regional radio, and by 1993 full commercial licencing was permitted. The result is a rare situation where the entire flow of daily Public Service Broadcasting (PSB) can be studied from the 1980s onwards, which in turn allows for large-scale analysis of the relationship between organisation structure and content stylistics before and after the introduction of commercial broadcasting. Broadcasting in Sweden was during this time maintained by the company Sveriges Radio (SR), and had been in more or less the same form since the early 1920s. The organisation was not directly owned by the state, but financed through the so-called broadcasting licence, which in turn depended upon governmentally decided goals. The overall ambition was similar to the usual PSB objectives: the guiding documents for broadcasting would thus be inclined to emphasise values like objectivity and diversity – yet little is said about aesthetics (Banerjee and Seneviratne 10). This arrangement was the setting for regular conflicts throughout Swedish radio history, in which the demand of the audience squared off with the ideals of broadcasters. Yet in the last decade of the century, the media environment was about to change in a precedented way. During these years, conceptual and economic tension between commercial and public service radio reached new heights, in turn forcing the matter of formats on the agenda. Research by Stjernstedt and Forsman, while not exclusively focussed on radio formats, addresses the theme within the broader framework of commercial Swedish radio. Their findings, along with those of Hedman and Jauert, suggest the influence of commercial formats on Swedish radio prior to the formal introduction of commercial broadcasting. This allows for an interesting epistemic possibly: if their propositions are correct, it would allow us to study the character of radio formats, without being intermixed with the scope of commercial broadcasting. The following analysis thus attempts to track the actual changes in the content, which could reveal the influence of format radio on PSB. For critics like David Hendy and Wolfgang Hagen, the format comes down to a question of self-similarity, “Programmierung von Selbstverständlichkeit” as Hagen critically dubs it (333), supposedly induces a certain superficial uniqueness of music stations, despite a fundamental sameness in their content. For this reason, the analysis is focussed on repeated similarities in the musical content. Methodological Approach Given the significant role ascribed to music in the theorisation of radio formatting, this aspect appears to be a potentially apt focus for constructing an experimental analysis. This is also practical as it allows for the translation and application of established spectral analysis methods from the realm of music analysis. In the context of audio data, spectral analysis refers to a process where the frequency components of a sound signal are decomposed and examined as numerical values. This method is crucial in audio engineering to understand the underlying structure and composition of sound and can assist in identifying specific patterns or anomalies in audio data. The analysis uses a sampling approach, concentrating on full-day broadcasts from music channels P2 and P3 in order to capture the general channel characteristics. These were the two formal music channels of SR during the time, with P3, the popular music channel, being directed at a younger audience, and P2 offering a more diverse mixture of classic and world music. The data, selected randomly from five weekdays across each of the years 1988, 1991, 1994, and 1999, are examined for indicators of self-similarity and format structuring. The hypothesis guiding this study suggests that format radio, shaped by technical standards and theoretical principles, exhibits a degree of radiophonic self-similarity. This proposition is explored quantitatively, applying statistical methods to the spectral data to assess similarities. The analysis itself is executed in combining a set of Python libraries. Initially, a segmentation model from The National Audiovisual Institute of France is used to isolate the musical content from the broadcasts. Then, the audio analysis library librosa converts these segments into spectral data, providing insights into timbre, key, and tempo. The results are presented using Principal Component Analysis (PCA) visualisations. These visualisations employ dimension reduction to represent the relationships between audio segments in a Euclidean space, with the proximity of nodes indicating similarity between data points. Even though the PCA method has limitations, well-discussed in the methodological literature, it grants certain insight into the structure of the dataset (Jolliffe and Cadima). In essence the method uses two-dimensional space to plot the relationship between n-dimensional data objects. The implication is that the distance between each entry entails something about the general similarity between the features of these data points. Critical scholarship, like that of Johanna Drucker, has brought attention to the epistemological tendencies inherent in these standardised statistical methods. The PCA method was also at the centre of discussions concerning the general credibility of computational literary studies in 2019, and I here follow Andrew Piper’s conclusion that the limitations are less a reason to abandon ship, and more an encouragement to remain “skeptical” (11). Thus, it is a call for complementary and comparative approaches. In order to do this, my analysis also employs audio recognition tools to compare the feature elements of the data set. By comparing both these different levels of analysis, across yearly samples from the decade, the study traces the evolution of musical styles and formats, potentially relating them to wider technological, cultural, or sociopolitical shifts. Results Fig. 1: The distribution of observations from the years 1988, 1991, 1994, and 1999, arranged from the top right to the bottom left. For each year, around 1,000 songs from each channel were studied. Data from 1988 indicate that the popular music channel P3 had a somewhat broader distribution spectrum in its content than its more eclectic counterpart. Initially, this may seem counterintuitive, as one might expect the expanding range of classic and world music of P2 to exhibit greater sonic variety than a channel playing mainstream and popular music. A possible explanation for these results can be found in prior research. Musicologist Alf Björnberg has provided a detailed account of the musical content of Sveriges Radio. His narrative emphasises that P3 underwent a significant content shift during the 1980s. Throughout this decade, the channel began to accommodate artists and genres that had been marginalised in the expanding landscape of broadcast media for decades (320). While P2 was not without variation during this era, P3 experienced a more notable change, moving towards avant-garde rock and experimental electronic music. This shift reflects a broader trend in the radio landscape, where traditional boundaries of genre and style were increasingly blurred, allowing for more diverse and experimental sounds to emerge in mainstream channels. The visualisation of these data not only highlights these historical shifts but also provides a quantitative basis for understanding the evolution of musical trends and preferences within the radio broadcasting domain. Björnberg’s study, predominantly centred on playlists and textual documents, mainly focusses on the period up to the end of the 1980s. Shortly summarising the 1990s, Björnberg concludes that 'the effects of commercial competition were most noticeable for P3' (326), especially during the early years of this competition. The data reveal a surprising trend in this regard too. At first glance, it might appear that P3 is reducing its musical variety, which could be interpreted as a response to the new, more rigidly formatted radio landscape. However, this perception is an illusion created by the rescaling of graphs to accommodate P2's possibly drastic content expansion. When calculating the average distance between points in P3's data, it becomes evident that the variation remains within the same range, with a fluctuation of only 0.4. This figure contrasts with P2's distance, which increases from 34.2 to 46.7 between 1988 and 1994. Instead of P3 reducing its musical variety, it is P2 that broadens its musical offerings. This can be seen as a response to the identified issue. Sveriges Radio addressed the new competitive situation with two seemingly contradictory initiatives: seeking a unique sonic iden