The FUSES Model of Musical Preferences
"The FUSES Model of Musical Preferences," by Daniel J. Levitin, PhD, Signal Patterns Scientific Advisory Board
Most of us know what kind of music we like – we have favorite radio stations, CDs, mix tapes, and mp3 files. But describing our preferences and tastes to another person can be difficult. We naturally turn to describing our tastes in terms of genres such as "rock" or sub-genres such as "heavy metal" or "speed metal." But there can be vast differences in the ways that people from different backgrounds and geographical regions use these terms. In one study I conducted, a 22-year old said that he liked AC/DC because it "pumps me up and gets me energized," while another said he liked the same group because "it is soothing after listening to Slayer."
Because different people don't use terms like "metal" or even "soothing" the same way when describing the music they like, relying on the descriptions and reports of listeners can be imprecise. What we've done at Signal Patterns is to remove this level of subjectivity and replace it with an objective question: Do you like this particular song?
Our research team has carefully chosen a small number of musical prototypes or "probes" that research has shown can help to better understand a given individual's musical tastes. The prototypes are the result of more than 10 years of research using tens of thousands of respondents, both on the internet and in laboratory settings. The prototypes can't be songs that everyone likes because they have to distinguish one set of tastes from another. In other words, a song that everyone likes or that everyone hates doesn't discriminate one listener from another. The prototypes can't be songs that one already knows because then any memory associations with that particular song could color the judgments (you broke up with your boyfriend the last time you heard it; you partied with your friends in high school to that song, etc.).
Your responses to our set of songs give you a distinct "musical personality," one that brings us closer to predicting other music that you might like. It also brings us closer to predicting what personality attributes you likely possess, and what attitudes and opinions you hold. Although the science behind this is rigorous, the research is still very much in-progress. We strive to make more accurate predictions than our competitors, but we still have a way to go. We know, for example, that people who like country music tend to hold more traditional social values. While this is not the case for everyone, it is a tendency that is borne out by the statistics. And we are discovering new correlations and refining these models all the time.
Underlying our experiment are statistical techniques that help us to notice patterns in the data obtained from many users. For example, we have found that some songs "co-vary" meaning that people who like one tend to like the other. Similar to the Big Five test of personality, the result of research using those statistical techniques is a tentative answer to the important scientific (and practical) question: "How many different relatively independent ways are there to characterize music?" In other words, are there key elements underlying all music that allow us to characterize music as combinations of these terms?
Across thousands of responses to a diverse array of musical examples, it turns out that there are not thousands or even hundreds of independent ways. As with personality, the magical number is around five. In studying the ways in which musical tastes cluster, we were able to discover these five factors, which we have labeled: Forcefulness, Urban, Smoothness, Earthy, and Sophisticated. Any given song can be seen as a distinct combination of these factors, and any given listener can be assigned a distinct score along each dimension, yielding a "musical personality profile." Then, we can match your scores on these five dimensions to songs in our database. We can also find correlations between musical personalities and conventional (psychological) personality, values, opinions, and attitudes.
