Musical Memory

Musical memory is about how we remember musical elements: that could be songs or melodies or rhythms. It’s about whether there are any rules to this, whether we remember all songs equally well, whether there are some songs that are easier to remember than others. It’s also about how we remember them, what strategies we have, how that’s related to musical training, for example, and whether there are differences between people in their ability to remember musical material.

It’s important from a general psychology perspective because if we study musical memory we study how the mind deals with nonverbal auditory information. Music is a great domain for that because we can’t attach meaning easily to melodic elements or rhythms, they are what they are, you can’t easily recode them in verbal terms, for example, or associate visual memories with that. So it’s really good domain a study of the human mind and memory in general. From a musical perspective it’s interesting to understand for music teachers and musical learners how we approach learning music and is there a good strategy to learn a new song, commit that to memory, play from memory at a performance, how should we do that and why are certain structures easier to learn than others. That’s the general idea of what this topic is about.

The study of musical memory, if we leave out the really early history, started maybe in the 1960s and 70s when people started within the cognitive paradigm, within psychology to run auditory memory experiments. In addition to verbal materials some auditory psychologists started using individual tones, for example, Diana Deutsch was an important researcher who started in the late 60s researching memory for individual tones or short tone sequences. Most of the musical stimuli were melodies or short melodies of some sort but they were specially constructed so they sound very artificial. They might have been sine tone generated from sine tone generator and with no rhythm: very boring music really, a psychological experimental music that you wouldn’t choose to listen to. But the great advantage of having these stimuli was that you could manipulate them and you would have full control over what people have to study in this experiment. That gives you an angle to compare different conditions between simple and complex melodies, melodies that have different types of contours and different patterns of ups and downs, whether adding rhythm would help or would make memory worse, for example, whether there are interference effects between if you have to remember one tone here and compare it to one tone there does it matter what series of tones are in the middle, for example.

So there were dozens of experiments ran and studies published on musical memory of these very simple musical stimuli that ordinary musicians wouldn’t even consider music at all, but for a psychologist it is still a different thing to remember speech, for example.

Later in the 80s and 90s people started using more natural musical materials, real melodies taken from songs or from classical music, rhythms as they would occur in real music. They also started to compare music from different cultures. Does it make a difference whether I as a Western listener need to remember an African song or a Finnish song, for example? Are they easily equal for me? How much does the experience with the particular musical culture help to remember that particular material? This got increasingly complex and there was more input from musicians and musical practitioners in this line of research as well because the pressing questions were not about artificial short sine tone series, but more about real melodies.

One of the results that has been replicated a lot of time is that musical contour is something that is really important in remembering novel stimuli, the pattern of ups and downs in pitch. That’s something that we extract very easily and readily from new melodies that we don’t know in advance. In addition we seem to abstract tonal informations, which key does a melodic series belong to. Then, if we have these two types of information, contour and tonality, we are able to reconstruct a melody a fairly faithfully if we pay close attention. That’s a theory that Jay Dowling published in 1978, the theory of contour and tonality for musical memory that these would be the two primary constituents in how we remember melodies.

Obviously, if you look at more complex music, a full song or a full symphony there’s also the question where the attention goes, what elements we focus on, whether we have an analytical ear and really focus on individual musical elements or whether we are just in a mode of listening, we rather extract emotions, gestures and kind of intentional meaning that the performer put into music? That might limit our capacity to process individual elements very faithfully. But in a real world musical listening processes it’s often not important to commit everything that we listen to to memory.

The important thing anyway is to remember that most of these processes work unconsciously. Even if you’re not a musician but you’re a normal music listener you will be able to remember melodies and compare one melody that you heard a minute ago to another one that you hear sometime later. So there’s a lot of implicit knowledge that is important in music and that’s a very interesting domain. You can learn music or understand music and remember music and process music just by being exposed to music culture. That’s different from their many other domains where you need to be taught explicitly, for example, what the meaning of individual elements of are. That’s not the case with music and that distinguishes music from, say, learning another language which you just couldn’t do by listening to the radio in an unknown language.

These days research has evolved to look at more and more ecologically valid stimuli, more music as it occurs in the real world. One of the lines of research that we’ve been conducting here at Goldsmith over recent years is to take melodies from a large corpus of popular melodies, Western commercial pop music as it was released from the 1950s to today and try to identify features in these melodies to see whether there’s anything in the melodies that makes them more or less memorable. We are not manipulating the melodies but we’re taking them as they have been published and then trying to find associations between these features. The features could be complexity or the range of the intervals, the distribution of duration values, anything to see whether these features are associated with how listeners process them cognitively. Then we try to explain why some melodies are easier to remember than others.

This line of research has been successful to a degree. We did find a set of features that in combination are responsible for a better memory or better memorability but there’s still a lot of unexplained variance in the data. So there’s a lot that we don’t know. We don’t know why some melodies are remembered better than others. A lot might have to do with an interaction between people, that person listening history and the stimulus. For example, if you’ve never been exposed to blues music and for the first time you hear a blues melody then this new structure that you’re presented with might not fall into your template. You learn templates and therefore you might find this more difficult to remember than other melodies. This interaction between individuals, listeners and stimuli is something that I think is worth exploring that many people try to get at at the moment.

One thing that’s also helping right now is that there are computational tools from music information retrieval and from other areas that help us to describe these melodies or music in general in computational terms so we have very clear and precise definitions of musical features that could be relevant for human cognition. That also gives the possibility to search large corpora of music and melodies for these features and that again helps in psychological research to select stimuli that have much of a certain feature or very little of a certain feature and we can compare the two sets of stimuli and see how people deal with that and see whether this feature makes a particular difference.

So in summary computational tools, larger corporal of melodies and larger and more varied samples of participants regarding their expertise are the three factors that are driving this field at the moment. They could deliver exciting results and might bring us closer to the question to understand how we remember musical melodies.