Empirical and Critical Methods in Musicology

Types of Behaviors

Recall that empirical knowledge is knowledge gained through observation. When we carry out research involving people, what we observe are various kinds of behaviors.

For convenience, it’s helpful to conceptually distinguish six types of behaviors: gross behaviors, social behaviors, topical behaviors, metabolic behaviors, self-report behaviors, and artifactual behaviors. As we’ll see, some behaviors straddle more than one of these six categories, so these aren’t hard-and-fast distinctions. However, they’re still helpful ways of organizing observations.

1. Gross Behaviors

Gross behaviors are those behaviors that involve externally observable actions. Examples of gross behaviors include bobbing your head, swaying, blinking, humming, singing, talking, shouting, and laughing. Two of the most common gross behaviors evident in Western culture including clapping hands and tapping feet. Gross behaviors can also involve aspects of posture, such as slumping, or leaning, lowering your head, or sitting upright. Gross behaviors also include facial expressions, such as frowning, raising or lowering your eyebrows, sneering, pouting, and so on.

An interesting example of the use of gross behaviors in a music study is the work of Olaf Post. The famous Concertgebouw concert hall in Amsterdam maintains a library of video-footage of all the concerts that take place. One of the unique features of the Concertgebouw is that there are several rows of seats behind the stage. So the video recordings capture — not just what’s going on on-stage — but also the behavior of (at least) some of the audience members.

Now when people attend a symphony concert, they tend to just sit and occasionally applaud. But in fact, people (from time-to-time) cross their legs, or scratch their noses — adjust their glasses, turn their heads, fold their arms — in short, people in the audience figet.

Using the video recordings, Olaf Post counted the number of figetting events over the course of different performances. He found that the amount of figetting parallels the musical structure in a variety of ways. What’s especially remarkable is that Post found that audiences were significantly less figety when listening to music by Gustav Mahler, compared with a composer like Anton Bruckner.

The important point is that — even the systematic observation of figetting carries information that can be potentially useful in music-related research.

3. Topical Behaviors

In science, the word topical means “pertaining to the surface.” A “topical medicine” is a medicine that is applied to the skin. Hence, topical behaviors are behaviors that relate in some way to the skin. There are four classic topical measures: skin temperature, skin conductance, pilomotor activity, and electromyographic responses.

Skin temperature changes slowly and can be measured with a topical thermometer. Some parts of the body tend to be warmer than others, and the distribution of heat can change over time. Skin temperature can also be imaged using an infrared camera. Skin temperature is determined principally by the amount of blood flowing near the surface of the skin. Some changes of blood flow can be seen directly — such as when a person blushes or when they go pale. But most changes are more subtle — although they can still be measured. When a performer is suffering from stage-fright, their hands and feet will tend to become cold. This occurs because of the constriction of blood vessels in the body’s extremeties. This phenomenon is referred to as peripheral vaso-constriction.

A second topical measure relates to how much we sweat. The skin is covered with sweat glands, and the amount of sweating can change rapidly. Sweat is a fluid with a high salt content, which means that it’s a relatively good conductor of electricity. Two electrodes are placed on the skin, and a small (unnoticeable) voltage is applied continuously. When sweating increases, the resistance of the skin drops and so the electrical current increases.

These changes can be quite rapid (on the order of half a second or so) and are easily (and cheaply) recorded. Measures of skin conductance are variously referred to as galvanic skin response (which is abbreviated GSR) or as skin conductance response (which is abbreviated SCR). Skin conductance is widely used in polygraphs — that is, in lie detectors. Sweating increases rapidly when the sympathetic nervous system is active. This includes experiences of fear, the startle response, the defense reflex, the orienting response, feelings of anger, and sexual arousal.

A third topical response involves body hair. hair follicles can flex or relax. When the follicles flex the skin takes on a distinctive goose-bump texture. The technical name for this is the pilomotor response. This response is musically important in the phenomenon of frisson where a listener experiences “chills” or “thrills.” With the help of a video camera and a close-up lense, the pilo-motor response can be observed visually. We simply point the camera at the back of the neck or the arm of a participant. Looking at the camera output, an observer can then visually characterize the degree to which the skin takes on a gooseflesh appearance.

A fourth topical measure arises from the muscles underneath the skin. Muscles produce electrical potentials that can be measured. These measures are referred to as electromyography (abbreviated EMG). EMG potentials can be measured either at the surface of the skin, or more directly by inserting a small needle or electrode into the muscle. Muscles generate large electrical potentials when they’re flexed, but they also produce spontaneous electrical activity when at rest. The resulting muscle activity can be highly informative depending on the muscle being monitored. For example, facial muscles echo various emotional states. By way of example, EMG activity in the zygomatic muscles (involved in smiling) provides a good indication of positive feelings — even if there’s no visible evidence of smiling. An example of the use of EMG in music research is in the work of Ulf Dimberg who has demonstrated that EMG activity provides a useful measure of the perceived pleasantness or unpleasantness of a sound.

A particularly illuminating set of muscles are those related to the eyes. We can identify three main aspects to eye behavior. The simplest is blinking — which is one of the most reliable indicators of the startle response. When you hear a loud unexpected sound, it’s very likely that you’ll blink. A second aspect is the size of your pupils. The dark region at the center of the eye responds very quickly and continuously to changes of stimulation, including sounds. A third aspect is the movement of the eyes themselves — known as saccadic movements.

John Sloboda has carried out studies examining how the eyes move when musicians read musical notation. It turns out that the movements of the eyes differ depending on whether the music is predominantly homophonic or predominantly polyphonic. There are also observable differences between how the eyes move for highly trained musicians versus less experienced musicians reading musical scores.

Notice that eye behaviors form a special category that might be regarded as either gross behaviors or as topical behaviors.

4. Metabolic Behaviors

Metabolic behaviors relate to body conditions under the skin’s surface. Classic examples of metabolic behaviors include heart-rate, blood pressure, (core) body temperature, and respiration.

Of course many metabolic observations can be measured from the body’s surface. For example, we can measure a person’s pulse simply by pressing a couple of fingers against their wrist. Or we can measure blood pressure using a cuff wrapped around a person’s arm — and so on. There’s certainly room to debate whether these should be categorized as metabolic or topical measures. Nevertheless, traditionally, heart rate and blood pressure are considered metabolic measures.

Pulse can be measured with wrist-watch like devices that can collect data continuously over many hours or even days. Two classic heart responses are of particular interest: the tachycardic and bradycardic responses. A tachycardic response occurs when the heart-rate increases briefly and then returns to normal. This is associated with fear or alarm. A bradycardic response occurs when the heart-rate decreases briefly, rebounds above normal briefly, and then returns to normal. This is associated with interest or attentiveness.

Another measure of interest is the heart-rate variability (or HRV). This has been found to be especially informative in a variety of different ways.

Another metabolic behavior is breathing or respiration. Respiration can be measured using a string-like or band-like device that’s placed snugly around a participant’s chest. It can also be measured using an electronic device taped to a person’s chest. Breathing influences the amount of oxygen in the body. Rather than measuring respiration, there exists a simple device (an oximeter) that clips onto the finger — and can be used measure more directly the oxygen level in the blood.

One of the most useful physiological measures is the electrical activity generated by the brain. Eletroencephalography (or EEG) measures tiny changes of voltage that reflect the electrical activity of large groups of neurons. Although these voltages are very small, they can still be measured by placing electrodes on a person’s scalp. EEG measures are rather crude — representing the average activity of assemblies of millions of neurons. When enough neurons fire at roughly the same time, they create a big-enough electrical potential to be measured at the surface of the head.

In music-related research, it’s most common to look at the electrical activity that’s evoked in response to a particular sound. These are so-called event-related potentials (abbreviated ERPs). ERPs have characteristic signatures that can indicate, for example, whether a person noticed or paid attention to a given sound. These signatures have proved particularly useful in answering questions about what infants are able to hear or distinguish. For example, research by Laurel Trainor and her colleagues has determined when infants are able to recognize atonal violations of tonal melodies. Infants can’t talk about what they hear, but their EEG responses can give important clues about how they’re experiencing the sounds.

An important class of metabolic measures is to be found in hormone levels. Hormones are chemical messengers, commonly transported by the blood that influence cell behavior in many ways. Examples of hormones include epinephrine and norepinephrine (also known as adrenaline and noradrenaine), dopamine, serotonin, oxytocin, prolactin, cortisol, histamine, estrogen, testosterone, insulin, as as others. Some of these hormones (like testosterone) can be measured in saliva. So we can have a participant simply spit into a small cup, or use a mouth swab. Another common technique has the participant chew on a small absorptive plastic device.

Other hormones can be measured only by taking a blood sample. Depending on the sensitivity of the assay method, the researcher may need only a small dab of blood (that can be collected from a pin-prick), or a larger sample of blood that would require a professional nurse to do a blood-draw.

Some hormones are present only in the brain. Unfortunately, these hormones can be measured only by examining the cerebro-spinal fluid — which is gathered through a spinal tap. This procedure is much too invasive to be used for casual research purposes, like studying music.

An example of observing changing hormone levels in response to music is the work of Hajime Fukui who found that testosterone levels are lower when people listener to their preferred music.

In the past decade, it’s become increasingly common to use brain scanners in music research. At least four types of scanners can be distinguished. Functional Magnetic Resonance Imaging (or fMRI) can be used to measure how much oxygen is being used in different parts of the brain. Areas of the brain that are the most active use more oxygen. The localized blood volume in that area changes rapidly, and fMRI methods can be used to pin-point these places when a person is engaged in different tasks. MRI machines are noisy — producing a loud intermittent banging sound that limits their use in auditory tasks. The machines are bulky and expensive.

Positron Emission Tomography (or PET) involves the injection of mildly radio-active substances into the blood-stream. Depending on the substance that is injected, PET can provide very useful indications of where in the brain a particular neurochemical is released or congregates. Using this technique, for example, Varlery Salipoor and her colleagues were able to show that listening to chill-inducing music released dopamine in regions of the brain associated with the experience of pleasure. PET is quite invasive since participants are injected with radio-active substances. The preparation of these materials by an experienced chemist makes PET especially expensive.

A third technique is MagnetoEncephalography (or MEG). Like EEG, this method traces the electrical activity of the brain. However, it monitors magnetic information rather than voltage — giving much better resolution and is able to provide better information about activity going-on deeper in the brain. Compared with fMRI and PET, MEG has an especially good temporal resolution, resolving electrical events in the brain with a precision around 10 milliseconds. Since MEG scanners measure magnetic fields, the scanning must take place in a magnetically shielded room. In addition, liquid helium is used to cool the machine, which raises the cost significantly.

A fourth method is functional Near-Infrared Spectroscopy (or fNIRS). Infrared light is able to penetrate through the scalp and skull, and for some distance into the cortex. The reflected light is useful for detecting changes in blood hemoglobin. Like fMRI, fNIRS can be used to infer oxygen uptake in regions of the brain associated with neural activity. In addition, NIRS is much more portable than fMRI, PET, or MEG machines. Some manufacturers provide wireless instruments that allow researchers to study freely moving people. However, due to the absorption of light, NIRS can only be used to scan the outer cortical tissues of the brain. It is unable to measure subcortical structures. NIRS is especially effective for imaging the brains of infants. The technique is non-invasive, and the thin skulls of infants allows the light to penetrate more deeply into the neural tissue.

5. Self-Report

Without question, the most common behaviors used in research is the self-report. Self-report is the label given to any behavior in which we simply ask participants something. An obvious example of a self-report is when we ask someone to indicate their age on a questionnaire. In these cases, we rely on special knowledge possessed by the participant.

Self-report observations can be collected in many ways, including both formal and informal interviews. For example, we might ask a child what songs she knows, how long she’s been studying an instrument, or what she thinks of her music teacher.

Questionnaires and surveys all involve self-report. These can be done with pencil and paper, through brief interviews, or using electronic media such as having people answer questions via the web.

Many self-report behaviors involve asking someone to offer an opinion. IQ tests and personality tests — all rely on self-report. Often responses are structured, so that the participant simply selects one of a set of predefined answers. For example, a statement may be presented, and the participant is asked to indicate the degree to which they agree with it: strongly-agree, agree, agree somewhat, undecided, disagree somewhat, disagree, or strongly disagree. In music, it’s common to ask which of two sounds is more dissonant, or more memorable, or more whatever.

These are referred to as forced-choice responses. If only two choices are provided, the mode of responses is commonly known as a two alternative forced choice — which is frequently abbreviated 2AFC. Responses might involve typing on a computer, pointing-and-clicking, or pressing a button of some sort.

Self-reports include elicited statements and introspective reports — such as asking a performer to describe (for example) how he or she begins practising a new musical work.

Apart from interviews, conversations, surveys or questionnaires, responding by pressing a button is also typically regarded as a form of self-report.

Self-report is commonly used in research, even when there are more objective methods available. For example, rather than asking a person their age, we might follow-up by using government records to obtain a copy of their birth certificate. But in the majority of studies, the researcher has little reason to doubt the accuracy of the information provided by the participant.

Nevertheless, there are common situations where participants do not accurately self-report. For example, in self-report, people often claim to be taller than they are, and it’s very common for people to mis-report their weight. When accurate data is needed, the researcher may choose to perform objective measurements.

The biggest advantage of self-reports is that they are often the easiest kind of data to collect. However, they are also easily confounded by the views, the ideas or the beliefs of the participant. These beliefs may or may not be accurate. For example, a musician might intellectually conclude that there’s nothing inherently sad about the minor chord. Our musician might even claim that he or she doesn’t hear the minor chord as having any sad connotations. In many cases, we should simply believe what a person says. But sometimes people deceive themselves — and so what is said doesn’t necessarily accurately reflect what they’re experiencing.

Implicit Measures

In these sorts of cases it’s useful to use a method in which the person’s beliefs are sidelined. One of the best ways to do this, is by employing so-called implicit measures. A simple example is provided by the affective priming method. Frank Ragozzine carried out a simple experiment where he flashed words on a screen and asked participants to respond as quickly as possible to whether the word was a Happy word or a Sad word.

First of all, this task is easy. The words are very clearly happy or sad. For example, sad words might include glum, down, sorrow, blue and blah. Happy words might include smile, jolly, pleased, jumping and sunny. The participant simply had to press one of two buttons as quickly as possible — either the happy-word button or the sad-word button.

Now consider what happens when we play either a major or minor chord immediately before the word appears on the screen. What Ragozzine found was that playing a major chord improved the reaction speed for happy words but reduced the speed for sad words. At the same time, he found that playing a minor chord reduced the speed of reaction for happy words and improved the reaction speed for sad words. In other words, the chords facilitated performance when they conformed to the major-happy or minor-sad conventions — but they interferred with performance when the didn’t match the major-happy or minor-sad conventions. What’s nice about this method is that the task is simply too fast for people to give any thought. We can test people from different cultures and different backgrounds and see how they respond — without having to rely exclusively on what they say or claim.

In general, researchers prefer these implicit observations over explicit measures. Implicit methods reduce the impact of what a person believes — and instead focuses on how they behave. The most revealing kinds of behaviors are those that are spontaneous and unconscious. When a person taps their foot or smiles, it’s likely that they’re enjoying the music — whatever they might say. Spontaneous and implicit measures have been used, for example, to determine how prejudiced a person is. Few people would ever say that they are prejudiced against people of African descent, or women or foreigners. However, various implicit association tasks can be quite revealing about a person’s unspoken dispositions. Our actions can betray attitudes that our conscious selves might find quite uncomfortable.

As researchers, then, we’re always on the look-out for spontaneous, easily observable behaviors that are not strongly regulated by conscious thought.

I recall one day, casually listening to a recording by the musical humorist Peter Schickele — better known as P.D.Q. Bach. It was a live recording — and of course there was laughter. Now music-induced laughter isn’t a common behavior. But I remember being so struck by the behavior itself: laughter is an easily observable behavior that’s quite spontaneous and not strongly regulated by conscious thought.

That observation led to a major study of musically-evoked laughter. With Joy Ollen, for example, we carefully studied 640 instances of audience laughter in response to musical jokes.

Although it might seem tangential to more commonplace musical experiences, we learned a lot from that research.

So once again, although self-report is perhaps the most important source of behavioral observations for studying music, there are times when the researcher needs to be suspicious of the capacity of participants to accurately introspect and honestly report what they’re experiencing.

A useful check on observations from self-report, is to look for converging evidence using implicit methods where the conscious thoughts and beliefs of the participant are sidelined.

6. Artifactual Behaviors

A final category of behaviors might be called artifactual behaviors. People create things: in cultures all over the world we find all kinds of different musical instruments; musicians also make recordings — in lots of different formats; people make music videos, they take photographs, they write in diaries and blogs, they write program notes, they even write lyrics down on slips of paper.

In the case of music, some of the most important artifacts include sound recordings and musical scores. Scores and recordings are the familiar starting points for many forms of traditional music scholarship — especially the work of music theorists and scholars doing music analysis.

When approached systematically, empirical observations rely on the usual quantitative measures. A scholar might count the number of instances of a certain event — like the number of Neapolitan chords in some corpus. Even simple counts — like the number of notes or the number of measures in a work can sometimes prove useful.

Historical musicologists make use of a variety of techniques for making sense of manuscripts. For example, paying close attention to the hand-writing in a manuscript might help to resolve whether two manuscripts were written by the same person, or by two different people.

It’s often helpful to measure or count features as a way of helping to resolve a question or conjecture. What is the average angle of the stems compared with the horizontal staff. Is there a slight leaning to the left or to the right? How long are the stems? What kind of cursive style is used for the textual underlay — and so on. Tallying-up a series of physical measures from a manuscript can help provide converging evidence for a particular historical interpretation.

Conventionally, analysis might involve interpreting the harmonies in a score and perhaps relating certain harmonic progressions to different styles, genres or periods.

In recent decades, many musical scores have become available online, and so many measurements can be automated with the assistance of computer software.

A simple example here might be the so-called melodic arch. Musicians long ago observed an apparent tendency for melodic phrases to rise upward and then fall downward forming a sort of arch. Think of “My Bonnie Lies Over the Ocean” or “Somewhere Over the Rainbow.” In each case, the phrases tend to ascend at the beginning and descend towrd the cadence. But no sooner do you think of this — than all kinds of exceptions come to mind. So, for example, the song “Joy to the World” and the American national anthem both begin high — dropping downward — and then rise upward toward the end. So is there any merit to the notion of a melodic arch?

The three accompanying graphs are from Huron (1996). The graphs show what happens when the pitches for a large number of melodic phrases are averaged together. The three graphs illustrate phrases of 7-, 8- and 9-notes in length. In each graph, the first plotted point represents the average pitch height (in semitones above middle C) for all the first notes in the phrases. The second point represents the average pitch height for the seconds notes in the phrases, and so on.

Although we can identify plenty of individual exceptions, on average, it is indeed the case that for many genres of Western music, the melodic arch is more than a figment of our imaginations. In general, there is a tendency for melodies to ascend and then descend over the course of individual phrases.

As we’ve seen, there are innumerable artifacts arising from human behavior that can be measured or tabulated. These include scores, sound recordings, and even from musical instruments. All of these can provide useful observations — either for exploratory research or for hypothesis-testing.

References:

Ulf Dimberg (1987). Facial reactions and autonomic activity to auditory stimuli with high and low intensity. Psychophysiology, Vol. 24, p. 586.

Ulf Dimberg (1989). Perceived unpleasantness and facial reactions to auditory stimuli. Uppsala, Sweden: Uppsala Psychological Reports, No. 414.

Hajime Fukui (2001). Music and testosterone: A new hypothesis for the origin and function of music. Annals of the New York Academy of Sciences, Volume 930, pp. 448-451.

David Huron (1996). The melodic arch in Western folksongs. Computing in Musicology, Vol. 10, pp. 3-23.

Olaf Post (2011). “The way these people can just listen!”: Inquiries about the Mahler tradition in the Concertgebouw. PhD Dissertation, Columbia University Department of Music.

Frank Ragozzine (2011). Cross-modal affective priming with musical stimuli: Effect of major and minor triads on word-valence categorization Journal of ITC Sangeet Research Academy, Vol. 25, pp. 8-24.

Matthew Salganik, Peter Dodds & Duncan Watts (2006). Experimental study of inequality and unpredictability in an artifical cultural market. Science, Vol. 311 (February 10, 2006), pp. 854-856.

Valorie Salimpoor, Mitchel Benovoy, Kevin Larcher, Alain Dagher & Robert Zatorre (2011). Anatomically distinct dopamine release during anticipation and experience of peak emotion to music. Nature Neuroscience, Vol. 14, pp. 257-262.

John Sloboda (1985). The Musical Mind. Oxford: Oxford University Press.

Weaver (1943). [On eye movements while reading music.]