Tech Page: Your Brain on Music

MODELING MUSICAL COGNITION
Author:
Publish date:
Image placeholder title
Image placeholder title

FIG. 1: Neurons in the brain can be modeled as cortical columns, which comprise minicolumns of about 100 neurons each. The team at UCI uses this model as the basis for its research into higher brain function.

No one knows exactly how the brain works, but many scientists are trying to figure it out. Among them is a group of researchers at the Center for the Neurobiology of Learning and Memory at the University of California, Irvine. The team is studying higher brain function (HBF), which includes creative and analytical activities such as music, mathematics, and chess.

HBF occurs in the cortex, which forms the outer surface of the brain. Most of the brain''s nerve cells, called neurons, are located in this surface layer. The neurons receive, process, and send tiny electro-chemical signals. Each neuron has a single output and thousands of inputs from other neurons. When the combination of input voltages exceeds a certain threshold, the neuron fires a pulse to another neuron.

Amazingly, there are 10 to 100 billion neurons in the brain with at least 100 trillion connections between them. Many researchers believe these connections are used for data storage, which corresponds to more than 10TB in a 3-pound package you can hold in your hand. However, the science of neurology is far from being able to unravel the workings of the brain at the level of individual cells, so many researchers use models of brain activity in which groups of neurons work together as a unit.

In one such model, the cortex is organized into basic processing networks called cortical columns (see Fig. 1), each of which is further divided into mini-columns of approximately 100 neurons. Each mini-column is strongly connected to neighboring minicolumns, and some are also connected to distant minicolumns.

The electrical state of each minicolumn varies over time according to the activity of each neuron. Within a column, this behavior is called the firing pattern, which depends on the strength of the connections between minicolumns and the timing of the patterns. During HBFs, such as composing music, these firing patterns are relatively stable and somewhat periodic, evolving over seconds or minutes.

The UCI team has designed a computer simulation based on the cortical-column model. They have limited each minicolumn to three electrical states, corresponding to high, average, and low levels of neuron activity. Because of this, the minicolumns are called trions. Six trions constitute a cortical column in the UCI model, which gives rise to more than half-a-million firing patterns.

Some of these patterns are more probable than others in the model and in the brain itself. The most probable patterns correspond to the inherent firing patterns with which we are all born. As we are exposed to stimuli such as different types of music, these inherent patterns are enhanced, contributing to our abilities and preferences in all areas of creative and analytic thought.

The UCI team has written software that steps through firing patterns in one or more cortical columns every 100ms, calculating each pattern from certain probabilities and the strength of the connections between trions. The software then maps these firing patterns to MIDI note numbers and channels to play notes with different instrumental timbres.

This research has yielded many interesting results. For example, the music generated by the model actually exhibits recognizable cadences and modulations. Different mappings of the same firing patterns produce different flavors of music, including waltzes, minuets, certain types of folk music, and different periods of Western art music. It is also possible to teach the model a musical theme and let it develop variations on that theme.

Although it is the most subjective of all HBFs, music was selected to study because it is highly structured and universally appreciated, even at birth. It is also very time-dependent, which is one hallmark of the trion model. The team hopes to discover common traits between music and other HBFs, leading to a deeper understanding of the brain and how it works in all its myriad tasks.