Pitch Anticipation: a study of glide tone perception

(1)

PITCH ANTICIPATION

A study of glide tone perception

July 15th 2015

Jurre Thuijs

10067469

Master thesis

Arts & Culture: Musicology

Universiteit van Amsterdam

Supervisors: M. Sadakata & J.A. Burgoyne

(2)

Abstract

In this thesis, a research on the influence of musical training on pitch anticipation has been conducted. This has been done by conducting one main task testing the anticipation of a pitch, and four

supplementary tasks to test all the components of pitch anticipation: an auditory reaction time task, pitch discrimination task, pitch memory task and glide imagery task. Musicians outperformed non-musicians in the main pitch anticipation task, the pitch discrimination task and the pitch memory task. Also, complex interactions between the results of all five experiments have been found. These results show that musical training enhances pitch anticipation, mainly due to musicians' enhanced ability to resist interference, most probably due to a heightened state of arousal and attention. The lower pitch discrimination thresholds of musicians also positively influenced pitch anticipation. No significant difference between the auditory reaction times or glide imagery ability of musicians and non-musicians was found. A final model on pitch anticipation was conceived, which includes all fixed effects of the main pitch anticipation task, and simplified factors of the abilities tested in the supplementary experiments.

(4)

Introduction

As we all know, music can be highly pleasurable. A chill down your spine, goose bumps on your arms, or a sudden feeling of joy. But apart from these physiological phenomena, what is this pleasure? Why should music be pleasurable at all? To find out, we should start in the brain. There is evidence that music activates structures in the striatal dopaminergic reward system (Salimpoor et al.; 2011, 2013). Salimpoor et al. carried out two separate studies, one on familiar music and one on unfamiliar music. In the 2011 study, the researchers had subjects bring in their own pleasurable music, and measured with many kinds of scans and other methods if dopamine was released during a 'chill' one gets from listening to music they enjoy. They found that the intense pleasure experienced when listening to music is associated with dopamine activity in the mesolimbic reward system, in both ventral and dorsal striatum. However, appreciation of music is complex and depends on sociocultural factors, experience and memory, so what is the reason for this dopamine release exactly? Salimpoor argues that the main reason for dopamine release when listening to familiar music is incentive salience or 'wanting', which results in expectation and anticipation of a certain desirable note, chord, timbre or other musical characteristic. This tension results in arousal.

In their 2013 study, Salimpoor et al. researched pleasure inducement in music that was heard for the first time. Subjects were presented with 60 30-second long musical pieces, and were asked to bid their own money on the pieces, indicating how much they were willing to offer to buy the piece and listen to it another time. This amount correlated with the amount of pleasure they experienced when listening to the piece of music. Through this experiment, Salimpoor et al. found out that, once again, the dorsal and ventral striatum showed activity proportional to the reward value of the stimulus. This proves that direct familiarity is not necessary for activity in these dopamine target regions. Salimpoor et al. suggest three explanations for the appreciation of unfamiliar music: "(i) highly individualized accumulation of auditory cortical stores based on previous listening experiences, (ii) the corresponding temporal expectations that stem from implicit understanding of the rules of music structure and probabilities of the occurrence of temporal and tonal events, and (iii) the positive prediction errors that result from these expectations." Therefore, in both studies Salimpoor et al. ascribe the pleasure inducement of music to expectation.

By conducting the abovementioned studies, Salimpoor et al. have proven that music, familiar or unfamiliar, induces pleasure via the release of dopamine. Dopamine release is known for

reinforcing certain basic biological behaviors with high adaptive value. This raises an important question: why is dopamine released while listening to music, when music has no clear adaptive value? Dopamine release causes a pleasurable feeling, but why? The dopaminergic reward system is there to reinforce evolutionary adaptations; to direct the organism to a positive evolutionary behavior (Previc, 1999). It is widely believed that dopaminergic systems are one of the, if not the most important systems for human evolution. For example, eating can be pleasurable because eating is obviously

(5)

beneficial to survival, and therefore the dopaminergic system evolved to trigger when an organism eats beneficial food. Yet this same system is activated when listening to music. Where the adaptive value of music lies is not that obvious, and there are many different theories that try to find it. Darwin himself argued that the role of music is in the process of sexual selection. Some ascribe the adaptive value of music to its positive effect on social cohesion. Others try to find adaptive value in benefit for survival, including this thesis. In my bachelor thesis Music as an evolutionary aid, I have sketched a theory of music being an aid (a helpful byproduct) to other evolutionary adaptations, such as language, emotion, auditory scene analysis, motor control and the anticipation of future events. All of these adaptations are extra-musical, but music can activate brain regions associated with them, and music can and influence and improve (parts of) them (Thuijs, 2012). Music can also be labeled as an exaptation: an existing trait that existed that is put to new use (Honing, 2009).

To understand how this works, first we must look at the distinction between music and musicality. Henkjan Honing defined musicality as "a natural, inborn quality that is an outcome of our biology. It is largely based on qualities present in everyone from birth." (Honing, 2009, p. 55) The basic abilities that are needed to develop musicality and music are innate, but they formed by learning. David Huron explains that the ability to learn is essential in a rapidly changing environment, and innate instincts only work when the environment, or a certain part of the environment stays the same for long periods of time (Huron, 2006). Music is a learned phenomenon, so it cannot be passed down in our genes and influence evolution in a direct way. While the core abilities that are needed to make music are passed down in our genes, music itself is not. So how can a learned phenomenon put its stamp on evolution? James Baldwin came up with a solution: an evolutionary theorem called the Baldwin Effect: an evolved capacity to learn (Baldwin, 1896). This effect stands at the core of the theory of music being an evolutionary aid or exaptation. American paleontologist George Simpson describes the effect as the following: 'characters individually acquired by members of a group of organisms may eventually, under the influence of selection, be reinforced or replaced by similar hereditary character.' (Simpson, 1952) What this means, is that abilities that are learned and trained can be of influence to the evolutionary process of an organism. Simpson divides the Baldwin effect into three steps:

'(1) Individual organisms interact with the environment in such a way as systematically to produce in them behavioral, physiological, or structural modifications that are not hereditary as such but that are advantageous for survival, i.e., are adaptive for the individuals having them. (2) There occur in the population genetic factors producing hereditary characteristics similar to the individual modifications referred to in (1), or having the same sorts of adaptive advantages. (3) The genetic factors of (2) are favored by natural selection and tend to spread in the population over the course of generations. The net result is that adaptation originally individual and non-hereditary becomes hereditary.' (Simpson, 1952)

(6)

The ability to learn is very adaptive in a changing environment. Because our ability to make music relies so heavily on learning, it is entirely plausible that our musicality has an evolved capacity to learn, therefore being an example of the Baldwin Effect. To summarize the theory of music as an evolutionary aid: music and musicality first originated as a byproduct of other adaptations (Pinker, 1997) such as language, motor control, auditory scene analysis, emotion and emotional calls, and anticipation of future events, but it was then put to new use as an exaptation, aiding and training the adaptations it originated from, while evolving the capacity for auditory learning and taking on new roles.

For the current study, we will closely examine one of the adaptations music has close ties to: anticipation of future events. This is the adaptation Salimpoor et al. ascribed their findings on dopamine release to, so it could be at the core of the evolutionary purpose of music. Anticipation of future events is an evolutionary adaptation, because it provides faster and more efficient processing (Bastiaansen, Bocker, Cluitmans, & Brunia, 1999). It is pleasurable, because it provides a learning opportunity (Gebauer, Kringleback, & Vuust, 2012). It can be dissected into two parts: anticipatory attention and motor preparation. The first is the attention one gives to a stimulus, while trying to predict the next stimulus. This can result in the preparation of a motor skill. In music, the sense of rhythm provides a nice example of this effect. For example, clapping along to a beat is such a prepared motor skill. It is an example of anticipation: the listener fixes his attention on the stimulus and

prepares a movement, which in this case results in an anticipating motor skill that is called entrainment. It is not a reaction to a stimulus, but an anticipating 'guess' of where the next beat is predicted to be.

There are researches on beat anticipation (e.g. Repp, 2012), but not many on pitch

anticipation. Pitch anticipation is the predicting activity of expecting a pitch in time. It is an important part of music perception that seems to be somewhat overlooked in music cognition. Therefore, in this research, an experiment on the anticipation of pitch will be conducted. A comparison between musicians and non-musicians will be made on this matter, to answer the question whether musical training improves this extra-musical skill. This will be done by taking the music out of the equation, and looking at the most basic anticipation of pitch using sine tones and glide tones.

It is also not clear what precisely entails anticipation of pitch. It is a very complex ability that requires basic abilities like pitch discrimination and auditory imagery, as well as certain developed memory and prediction systems. Starting with David Hurons ITPRA theory (Huron, 2006), a model for this ability will be constructed by conducting five different experiments: a reaction time task, the main pitch anticipation task, a pitch discrimination task, a pitch memory task, and an auditory imagery task. In this way, all components of pitch anticipation will be examined, after which a comprehensive model can be made to understand this complex ability and the influence of musical training on it better.

(7)

General Information

Methods

The following methods apply to all five experiments: Participants

14 musicians and 18 non-musicians were tested, aged 18-29 (mean = 21.7), including 19 females and 13 males. Musicians had to have more than six years of formal musical training and had to still be actively practicing an instrument at the point of the experiment. Musicians had an average number of years of musical training of 11.9 years. Non-musicians must have had no or less than 3 years of formal musical training and had not to actively practice any instrument at the point of the experiment. The mother language of all participants was Dutch. There were 28 right-handed participants, and 4 left-handed participants. No participants reported having absolute pitch.

Materials

All sine tones and glide tones were generated using SuperCollider (v.3.6.6), using a logarithmic gliding scale for the glide tones. All sounds were digitized at a sample rate of 4.41 kHz into a WAV format, with normalized loudness. All sounds were presented through the same headphones (AKG K540). Inquisit 4 was used to construct and run the experiments, RStudio and Microsoft Excel were used for the statistical analysis. Participants responded using a button box.

Task and procedure

The experimental session consisted of five tasks, consisting of a simple auditory reaction time task, a pitch discrimination task, the main pitch anticipation task, a pitch memory task and a glide rate perception task (presented in that order). The entire session took around 50 minutes, with breaks in between the experiments and sometimes in between blocks. Before partaking in the experiments, participants signed an informed consent form after being told the general outline of the research. Any further questions were answered, although no specific information about the tasks or the purpose of the experiment was given. Afterwards, participants filled in a questionnaire. Participants took part voluntarily and were not reimbursed for their participation.

Ethics Committee

This research was approved by the Ethics Committee of the University of Amsterdam's Faculty of Humanities.

(8)

Experiment I

:

Reaction Time

The first component of anticipation we will look at, is auditory processing time at the most basic level. Reaction time tests are a good way of recording the time a person needs to process and act on a certain stimulus. Experiments using auditory reaction times have been conducted by Bret Aarden of The Ohio State University in 2003, in which he asked his participants to indicate whether the pitch contour of the melody they heard had ascended, descended or remained the same after each note (Aarden, 2003). The most significant influence on reaction time he found was 'process': the tendency for small steps in pitch to be continued in the same direction. People were significantly faster when responding to process-like configurations than they were in other situation (59 ms on average).

Another reaction time experiment has been conducted by Rebecca Woelfle and Jessica A. Grahn, in which they compared interhemispheric transmission times (ITTs, meaning the time it takes for a signal to travel between hemispheres) between musicians and non-musicians (Woelfle & Grahn, 2013). They found no significant difference between the two groups, but they did find a significant difference between the crossed-uncrossed difference (CUD, meaning the difference between processing the stimulus and the movement of the hand in the same hemisphere, or in crossed hemispheres) of auditory and visual reaction times for musicians. Musicians had faster auditory reaction times than visual reaction times. A possible explanation for this, is that musicians develop a greater number of corpus callosum fibers (which connect the left and right hemispheres) for auditory processing at the cost of visual processing. This hypothesis does need to be tested further, but it is a viable one.

However, Charmayne Hughes and Elizabeth Franz did find a highly significant difference between the visual reaction times of non-musicians and musicians (Hughes & Franz, 2007). Although their focus was on finding a difference between bimanual and unimanual responses between musicians and musicians, they found that musicians had faster visual reaction times than

non-musicians in a simple visual reaction time test. This finding somewhat contradicts with the findings of Woelfle and Grahn, but they ascribe the effect to the greater efficacy of interhemispheric connections as well.

In this experiment, auditory reaction times of musicians and non-musicians will be compared. The main purpose of this first experiment is to subtract the reaction time from the response time of the main experiment (II: Pitch Anticipation), , to control for individual differences in reaction timing and get the actual response time participants aimed for. Therefore, only a basic analysis of the results will be reported here. Nevertheless, it would be interesting to see a difference in the reaction times between the two groups, because that would suggest an effect of musical training on auditory processing speed. In connection to anticipation, a correct expectation does facilitate a faster response time (Huron, 2006, p. 51). However, it will be impossible to form a correct expectation because the silent intervals will be

(9)

randomized. Therefore, also following the findings mentioned above, we do not expect to find any significant differences between the general auditory reaction times of musicians and non-musicians.

Methods

Experimental Design

Participants were instructed to press a button on a button box when they heard a sine tone (710 Hz, at a comfortable level), as fast as they could. They were presented with 20 trials in total, at 2 to 8 second random intervals. Participants were instructed to only use their dominant hand, and to keep their gaze at the black fixation cross in the middle of the screen. The sine tone kept playing until they pressed the button. The response time was recorded from the onset of the sine tone to the moment they pressed the button.

Statistical Analysis

The latest 5% of the responses were excluded from the analysis, to reduce the influence of outliers that could result from lapses in attention or slowness from adjusting to the task. No anticipatory responses under 100 ms, meaning responses that were faster than humanly possible when purely reacting, were found. This left a total of 609 responses (335 for non-musicians, 274 for musicians). All data analyses were performed using RStudio and Microsoft Excel software. An α level of 0.05 was used for all statistical tests.

Results

Using a linear mixed-effects model and ANOVA tests, no significant effect of musical training on auditory reaction times was found (p = 0.432). The mean reaction times for non-musicians and non-musicians were 290 ms and 285 ms respectively. A Levene's Test for testing

homogeneity of variance came out to be insignificant (p=0.124). Figure 1.1 shows the plot for the average (median) reaction times for non-musicians and non-musicians.

Figure 1.1 Boxplot of average auditory reaction

(10)

Discussion

As was expected, no significant effect of musical training on auditory reaction time was found. This finding is similar to the findings of Woelfle & Grahn (2013) mentioned in the introduction. It must be said that this was a very small and short experiment with only 20 trials per participant, but 20 trials do provide a solid average auditory reaction time for each participant.

This finding shows us that musical training does not necessarily enhance the speed of basic auditory processing and motor control. The mean reaction times for each participant will be used in the model for Experiment 2 as an offset for the final value, to control for individual differences in reaction timing. This is the main purpose of this experiment.

(11)

Experiment II

:

Pitch Anticipation

As was said before, anticipation is a very complex ability that is constructed of many components. In his book Sweet Anticipation, David Huron devised an elaborate theory of anticipation, which he dubbed the ITPRA theory (Huron, 2006). In this theory, Huron explains that any event causes five consecutive responses: 1) imagination response (pre-outcome): this response is future-oriented, focused on predicting what might happen and the way the subject feels about that prospect, 2) tension response (pre-outcome): this response comes after the imagination, and is characterized by optimum arousal in preparation for the outcome of the anticipated events, 3) prediction response (post-outcome): the first post-outcome response, which is characterized by a pleasing or disappointing outcome regarding the accuracy of the anticipation the subject has made, 4) reaction response (post-outcome): this response is characterized by a worst-case assessment of the outcome, and the subject's immediate reaction to it, 5) appraisal response (post-outcome): this response is characterized by reflecting on the final outcome that results in negative or positive reinforcements.

These responses also happen while listening to music, as Huron said. We imagine, become aroused, predict, react to and reflect on music constantly while making or listening to music. The imagination response in music is represented by the auditory imagery we form in our heads, which we will discuss extensively in Experiment 5. As we have seen in the research done by Janata and Paroo, musicians do have an enhanced pitch acuity and precision in their auditory images. They also found an effect of musical training on the timing of a note after an imagined part of a scale, although it was a smaller effect than for pitch (Janata & Paroo, 2006). Therefore, the imagination response should be more precise for musicians when doing a musical task. However, we did not find a significant effect of musical training for the task in Experiment 4. As was explained, the imagery system constructed from long-term memory is influenced by exposure and experience with certain tones and relations. Since the musicians that were tested were all highly exposed to and experienced in Western music and its corresponding scales, they did perform better in the task by Janata and Paroo, but not in our task. The imagination response Huron stated, should therefore not be more precise for a pitch anticipation task using glide tones as stimuli.

The next response is the tension response. This response happens just before the onset of the event, preparing the body and mind for the event. Huron splits this response into two components: motor preparation and perceptual preparation (attention). The notion of motor preparation is an important one when using timing as a response method. When a certain group can prepare their limbs better, they would certainly perform better at a timing task. Although we did not find a difference between reaction times of musicians and non-musicians in Experiment 1, a more complex task might address other specialized systems. Therefore, if a task using timing can be designed that incorporates the anticipation for a pitch, the responses could tell us something about the anticipatory system of the

(12)

participants. It could also tell us something about their perceptual preparation or attention, where a more attentive person would perform better at such a task than a non-attentive person. The attentive person would be able to make a better prediction causing an enhanced preparation of motor skills. Huron (2006) explains this by saying that "the goal is to match arousal and attention to the expected outcome and to synchronize the appropriate arousal and attention levels so that they are reached just in time for the onset of the event. [...] (this arousal) helps us to react more quickly and to perceive more accurately" (p.9). These last two remarks are the components of the tension response. A heightened arousal would therefore make participants perform better at a pitch anticipation task. Arousal can entail an increased heart rate and blood pressure, deeper and more rapid breathing, increased perspiration, faster muscle response, dilated pupils, orienting the head to or away from the stimulus, and purging distracting thoughts.

The other three responses are all post-outcome. The reaction response and appraisal response are similar, the reaction response being a quick assessment of the situation, and the appraisal response being a slow assessment. For this experiment, they are of less interest, because they happen after the anticipation. The prediction response, however, is of great importance to forming our expectations. Huron describes it as a reward system for expected and unexpected events, where expected events are valued positively, and unexpected events are valued negatively. This is a direct consequence of

evolution, as unexpected events are not desirable for survival. The prediction response is a result of the exposure effect, and misattribution. The exposure effect is constructed by an unconscious stream of stimuli, from which the familiar arises. Stimuli that we were exposed to a lot become familiar, and research dictates that people prefer the familiar over the unfamiliar (Huron, 2006, pp. 131-132). Two theories on why we prefer the familiar have been posed by Robert Zajonc, and Robert Bornstein & Paul D'Agostino. Zajonc's explanation for the exposure effect is that the familiar causes a more relaxed state, because familiar stimuli reduce the likelihood of orienting responses (Zajonc, 1968). Bornstein & D'Agostino ascribe the preference of familiarity to a misinterpretation of the ease of processing as a good stimulus (Bornstein & D'Agostino, 1994). Huron himself takes the latter theory a step further, combining misattribution (connecting a certain emotion to an unrelated stimulus) with the exposure effect. The most frequently occurring past event is the most likely future event, and when this accurate prediction is rewarded (because unexpected events are undesirable), it is misattributed to the stimulus. He calls this the prediction effect. This is the reason why predicted stimuli are pleasurable, keeping in mind that this is only one part of the ITPRA-theory, and that the other responses can also produce (dis)pleasure. If we apply the prediction effect to music, we can say that music can be pleasurable because of this effect. In the chapter Auditory Learning, Huron provides an overview of evidence suggesting that auditory learning is shaped by the frequency of occurrence of stimuli (Huron, 2006, pp. 59-72). Because musicians have been exposed to music and its broad scope of tonal material a lot more than non-musicians, the familiarity is much higher, causing a faster processing of expected stimuli.

(13)

In this experiment, this theory of the faster processing of expected stimuli will be applied to the anticipation of pitch. This will be done using glide tones and sine tones with non-musical

frequencies, because these rule out the possibility of learned musical systems such as key perception, that would give musicians a clear advantage in exposure and their musical training. People are relatively unexposed to glide tones, so this could provide evidence for the possibility of transposing the predictions formed by music to the anticipation of a new, unfamiliar and a-musical stimulus. This will be measured by using a timing task, because the faster processing caused by the heightened exposure effect for musicians could provide a more precisely timed response. In short, we expect to see a more precisely timed response for musicians compared to non-musicians, because of their ability to process auditory stimuli faster.

Methods

Experimental Design

First, a single non-western tone was presented for 2 seconds. Then, after a silence of 2 seconds, a gliding tone was presented. The gliding tone matched the first single tone in pitch after 4 to 5 seconds with 200ms differences and outliers of 2 and 8 seconds. Two different gliding rates were used: 'fast' and 'slow', for which 'slow' (mean rate of 14 Hz/s) was twice as slow as the 'fast' (mean rate of 28 Hz/s) rate. Also, two different non-musical single tones were used (710 Hz and 510 Hz). The gliding tones were either linearly ascending or descending in pitch. This resulted in 60 different combinations in total. The participants were instructed to press a button on a button box on the precise moment they expected the first tone to reappear in the gliding tone. They were instructed to try to precisely time their reaction on the frequency heard before, not earlier or later. The time from the onset of the stimulus to the pressing of the button was recorded (referred to as the anticipatory latency). The difference between this anticipatory latency and the actual time the tone reappeared in the gliding tone resulted in an early, precise or late response record. This tells us whether the participant was early (imprecise prediction) , precisely on time (precise prediction) or too late (reactive).

For this analysis, the responses to the 2 and 8 second outliers were excluded, because their purpose was just to introduce more variance in the intervals. Also, any responses that were 3000 ms too late or too early were excluded, because these response times were most likely due to a drop in concentration or attention. This decision was made based on observations that were made during the testing. The reaction times that were recorded in Experiment 1 were used as an offset in the random intercept model, to eliminate personal differences in auditory reaction times between participants, and to get the actual time the participants aimed for. An ANOVA test showed a preference to the model with the reaction time offset, rather than without. All data analysis was performed using RStudio and Microsoft Excel software. An α level of 0.05 was used for all statistical tests.

(14)

Results

For this task, a logistic regression with a random intercept model was used, with the following fixed effects: musicianship (musician or non-musician), the direction of the glide (up or down), the speed of the glide (slow or fast), and the duration from the onset of the glide to the matching tone (4, 4.2, 4.4, 4.6, 4.8 or 5 seconds). ANOVA tests showed a significant effect for the duration to the tone

(p<0.005), the speed of the glide (p<0.001) and the interaction between musicianship and the direction of the glide (p<0.05). No significant difference between musicians and non-musicians was found in an ANOVA test, but a Levene's test for homogeneity of variance came out highly significant for

musicianship (p<0.001).

The overall response times of musicians and non-musicians tended to be before the occurence of the matching tone (early). The mean response time of musicians was 250 ms before the occurence of the matching tone, the mean response time of non-musicians was 549 ms early. A drop-off in response times was found the longer participants had to wait for the occurence of the matching tone. When the matching tone occured after 4 seconds, the mean response time was 257 ms before the occurence of the matching tone. When the matching tone occured after 5 seconds, the mean response time was 525 ms early. Participants also tended to respond earlier when the glide rate was slow, in comparison to fast. When the glide rate was fast, the mean response time was 280 ms early, and when the glide rate was slow, the mean response time was 548 ms early. Also, non-musicians tended to respond earlier when the glide moved upwards (629 ms early) rather than downwards (468 ms early). No such difference in response times for the direction of the glide was found for musicians. A graphical overview of the results for each condition is shown in Figure 2.1 and 2.2.

(15)

Figure 2.1 Boxplot of predicted response times of musicians and non-musicians for each condition (duration to the matching tone is treated separately in Figure 5.2) The matching tone in the glide occurred at the 0 ms mark.

Figure 2.2 Boxplot of predicted response times of all participants for the duration between the onset of the glide tone and the occurrence of the matching tone.

(16)

Discussion

Four separate significant effects on anticipatory timing were found in this experiment: the duration to the matching tone, the glide rate, the interaction between the direction of the glide and musicianship, and musical training itself. However, the effect of musical training did not show up in an ANOVA test, but did show up in a Levene's test. This case and the other effects will be discussed here. Duration to matching tone

The participants showed a slight tendency to respond earlier the longer the duration to the matching tone was. This can be explained by two things: the fading of short-term pitch memory (Rakowski, 1994), and the 'fear' of being too late. The fading of pitch memory is probably the best explanation for this effect. As we will see in Experiment 4 and the research done by Diana Deutsch, the influence of tonal distracters on pitch memory is highly significant (Deutsch; 1970a, 1970b, 1972, 1974). The longer the participants have to retain pitch information while they are presented with a distracter, the more difficult it will be to precisely store this information in the short-term memory. This fade in the precision of pitch information is very likely of influence in this experiment, and can explain why participants responded less precisely the longer they had to retain the pitch information. The effect of distracters will be examined and discussed extensively in the chapter on Experiment 4.

Having said that, the 'fear' of being too late must not be overlooked. Some participants did report that they did not want to be too late in their responses. This could urge them to press faster the longer it took to reach the matching tone. This is an interesting strategy that is quite characteristic for anticipatory behavior. Following the quote by Huron (2006): "the goal is to match arousal and attention to the expected outcome and to synchronize the appropriate arousal and attention levels so that they are reached just in time for the onset of the event. [...] (this arousal) helps us to react more quickly and to perceive more accurately", it is very likely the optimum arousal tends to happen sooner the longer the arousal can 'build'. This phenomenon can explain why participants tended to respond earlier when the duration to the matching tone was longer.

Glide rate

The participants showed another tendency to respond sooner: when the glide rate was slow, in comparison to fast. This effect shares some characteristics with the duration to the matching tone effect: the pitch memory precision and arousal/attention levels were likely to also have an effect here.

The reason that participants responded early rather than late, can again be ascribed to the 'build-up' of arousal, resulting in a faster response. However, the precision of short-term pitch memory was key for the slow glide rate, because participants had a smaller window of frequencies to base their expectation on. The pitch of the glide moved more slowly over time, so the more precise the pitch stored in their memory was, the more precise their response time was. Precisely storing pitch information while being distracted by other tones is a difficult task. For example: when the first tone was 710 Hz, a difference of responding 270 ms earlier (the average difference between a slow and a

(17)

fast glide rate response) comes down to an 8.5 Hz difference in the glide, which is about a 20 cent difference for this frequency range. This difference is quite small, therefore it is clear that pitch memory precision is most likely the reason this difference between the fast and slow rate occurred. In Experiment 4, we will test the participants' precision of pitch memory while being distracted by a glide tone, to examine what effect distracters have on being able to retain a pitch in your memory. In

Experiment 5, we will test what the effect of good auditory imagery has on anticipation. This could also explain why responding to the slow rate is harder than responding to the fast rate, if we find a difference in precision between these two in an auditory imagery test.

Musicianship & Direction

A strange third effect showed up in the results: non-musicians tended to respond sooner when listening to an upwards glide, rather than a downwards glide. The explanation for this effect is speculative, because there is no apparent reason for why this should be the case. It could have something to do with a difference in arousal: an upwards moving glide might be more associated with tensing/building up, while a downwards moving glide is more closely associated with release or closure. The most likely explanation for this effect, is that non-musicians are more susceptible to this feeling of rising and are therefore more highly aroused, causing them to time their response sooner. Musicians can rely on their more precise pitch memory, resulting in a difference in strategies between the two groups, where musicians are not as susceptible to other associations with direction.

Musicianship

The most important factor in this experiment is the difference between non-musicians and musicians. Although an ANOVA test did not come out to be significant for this factor, a Levene's test did. When we look at Figure 5.1, a very clear difference in variance can be seen. In this case, this is actually not a bad thing, because it shows us that musicians as a group are much more precise than non-musicians. The reason this effect is not significant in an ANOVA test, is because the early and late responses somewhat level each other out, resulting in similar mean response times between the two groups, which skews the results of the ANOVA tests. It is very clear that the responses of the non-musician are much more spread out (proven by the Levene's test), which means that non-musicians are much less precise in anticipating the first pitch in follow-up glide tone than musicians. This confirms the hypothesis that was stated in the introduction: musicians can process tonal material faster and more precisely than non-musicians.

Now, what does this processing entail precisely? And what do musicians learn during their musical training that is enhancing these abilities? To answer this question, all the different components of anticipation in this experiment must be evaluated. The components we found that are important to the kind of anticipation in this experiment are: imagination (auditory imagery), tension (arousal and motor preparation), prediction, and memory (short-term pitch memory). Overall pitch discrimination

(18)

abilities are also of influence, because one must be able to distinguish tones to do this task. To test the influence of these components, Experiment 1, 3, 4, and 5 have been devised and conducted.

(19)

Experiment III

:

Pitch Discrimination

There has been a fair amount of research comparing the pitch discrimination ability of non-musicians and musicians. An early research was conducted by M. F. Spiegel and C. S. Watson in 1984, where they tested the pitch discrimination thresholds for non-musicians and musicians (Spiegel & Watson, 1984). They found that half of the non-musicians had similar thresholds as the musicians, while the other half had thresholds of up to five times as large. They divided the non-musicians into two subpopulations in their discussion: those who were significantly more likely to have had either a high degree of musical or psychoacoustic experience and those who were not likely to have had those experiences. Those who were in the first subpopulation had a similar pitch discrimination ability as the musicians. The fact that half or their non-musicians were likely to have had musical or psychoacoustic experience might have blurred the difference between the two groups.

This difference between non-musicians and musicians has also been found by L. Kishon-Rabin et al. in 2001 ( (Kishon-Kishon-Rabin, Amir, Vexler, & Zaltz, 2001). They also found that the thresholds of the non-musicians were around twice as large as those of the musicians. This was also significantly related to the years of musical experience. In 2006, C. Micheyl et al. replicated this effect even more strongly, finding a threshold difference between non-musicians and musicians by a factor of six, meaning non-musicians' thresholds were six times larger (Micheyl, Delhommeau, & X. Perrot, 2006). Micheyl ascribes this large effect to their more stringent selection procedure, in which he only selected classical musicians and non-musicians with absolutely no musical or psychoacoustic

experience.

Following these findings, we expect to find a significant difference between non-musicians and musicians for this experiment, where musicians will have a lower pitch discrimination threshold than non-musicians (up to six times lower). The main purpose for this experiment was, again, for later application as a factor in the final model and as a criterion for excluding (nearly) tone deaf

participants. Also, a new factor that was not present in previous researches will be introduced in the current pitch discrimination experiment, namely the direction the second tone moves in comparison to the first tone (up or down). This new factor was introduced because of the results of a pilot study of the main pitch anticipation task, in which a difference was found between the response times for an upwards moving glide and a downwards moving glide. This inspired us to try to pinpoint where this effect might come from, and therefore the directional properties of the tones used in this experiment were treated separately as a factor. The interaction between musicianship and the direction of the glide found in Experiment 2 provides further purpose for including this factor.

(20)

Methods

Experimental Design

Participants were presented with two 1 second sine tones, separated by a 0.5 second silent interval. Participants had to respond using a button box with two buttons labeled 'up' and 'down'. They had to press 'up' when the second sine tone was higher than the first, and 'down' when the second sine tone was lower than the first. The frequency difference between the two tones got increasingly smaller, using a three up, one down staircase procedure (Levitt, 1971). A total of 12 steps were included, which ranged from a 100 cent to a 2 cent difference (for details, check Table 1 in the Appendix). The first tone was always 710 Hz (non-musical pitch), the second tone was always higher or lower than the first tone. The direction the second tone moved in after the first tone (up or down) was treated separately. If a participant made a mistake in the upwards direction, but not in the downwards direction, the upwards strand moved a step down, while the downwards strand moved a step up, and vice versa. The two strands were presented together in groups of six (three down, three up), and were randomized. The trial ran for a total of 18 times, resulting in 108 (18*6) responses per participant. There were no pauses or practice runs. Participants were instructed to only use one (dominant) hand, and keep their gaze at the black fixation cross in the middle of the screen. They were also instructed not to sing or hum along with the tones.

For the analysis, the most frequent step of the best five steps the participant reached was calculated, so that we ended up with two numbers: one for the upwards direction and one for the downwards

direction. These step numbers were then transformed to their corresponding number of difference in cents. We used cents rather than the commonly used Δf/f measure, because it is a clear scale which is transposable across the entire domain of frequencies. Also, it transposes the logarithmic scale of frequencies into a linear scale, which is much easier to interpret in numbers and graphs.

Two participants were excluded from the analysis, because they were not able to discriminate the first step of 100 cents (one semitone). Both participants were non-musicians, which resulted in 14 musicians and 16 non-musicians. All data analyses were performed using RStudio and Microsoft Excel software. An α level of 0.05 was used for all statistical tests.

(21)

Results

Using a generalized linear model and two-way ANOVA tests, a significant difference was found between non-musicians and musicians for their pitch discrimination ability (p<0.005), but a Tukey test showed a significant difference only for the upwards direction ( p<0.05, mean difference = 10.8 cents) between the two groups, in which musicians had a lower pitch discrimination threshold. No significant difference between the downwards direction responses of non-musicians and musicians was found. Also no significant difference between the downwards and upwards direction responses within the non-musician group was found.

Figure 3.1 Boxplot of the amount of cents musicians and non-musicians were able to discriminate for

the upwards and downwards direction.

However, a Levene's test for homogeneity of variance came out to be significant for the difference between non-musicians and musicians (p < 0,001), which will be discussed extensively below. The mean number of cents non-musicians were able to discriminate was 15.6 (upwards) and 9.4

(22)

4.6 (downwards). This equates to an average difference between non-musicians and musicians of 10.7 cents for the upwards direction and a 4.5 cent difference for the downwards direction. The mean thresholds of non-musicians were therefore roughly 2 to 3 times larger than those of the musicians. A graphical overview of the results is shown in Figure 3.1.

Discussion

Musician - non-musician difference

As expected, the effect found by Spiegel & Watson, Kishon-Rabin et al., and Micheyl et al. was replicated here, as we found that the average thresholds of musicians were 2 to 3 times smaller than those of non-musicians. The effect was similar to that of the first two, possibly due to the selection procedure, which was not as strict as the one in Micheyls research. The procedural difference between this research and that of the others might also play a role in the differing results. Because we treated the direction between the two pitches separately, the experimental procedure was quite different from a standard pitch discrimination experiment. All kinds of musicians were admitted to this research (classical, jazz, pop) and some non-musicians had a very small amount of musical experience. The difference between the pitch discrimination thresholds of non-musicians and musicians has been ascribed to the experience musicians had learning to 'listen carefully' (Spiegel & Watson, 1984), training their ability for many hours.

Homogeneity of variance

Before any further conclusions can be made, however, we need to look at the difference in

homogeneity of variance between the two groups. A Levene's test came out to be significant in testing this property. Simply put, this means that the thresholds of the non-musicians were spread out much more than the thresholds of the musicians. When looking into the thresholds of non-musicians, a split can be seen between musicians, dividing them into two groups. One group (75%) of

non-musicians had average thresholds of under 12 cents, and the other 25% were way above that (16-50 cents). It seems that the split between non-musicians that Spiegel & Watson found, was also apparent in this experiment. When those 25% of non-musicians were excluded from the analysis, all significant effects disappeared. This tells us that the population of non-musicians is much more diverse than musicians when it comes to pitch discrimination abilities, and although these abilities do improve with musical training, mere exposure to music and language can yield similar results for a large amount (in this case 75%) of the population.

Direction

Furthermore, a new independent factor in pitch discrimination research has been introduced: the direction between the first and second tone. This required a new kind of experimental design, treating the upwards and downwards 'strands' of stimuli separately. This factor was introduced to find out if there was a difference in pitch discrimination for an upwards moving pitch and a downwards moving

(23)

pitch. As we found out, there was a significant difference for the upwards direction between musicians and non-musicians, but not for the downwards direction. There was no significant difference between the upwards and downwards direction within the non-musician group, but the variance between the upwards and downwards responses is much higher for the upwards direction.

Non-musicians have similar pitch discrimination abilities to musicians for the downwards direction, but have thresholds that are around 3 times larger on average for the upwards direction. This asymmetry is quite peculiar, because this would mean that people learn to discriminate downwards moving pitches outside of musical experience better than upwards moving pitches. However, keep in mind that these effects were caused by the 25% of non-musicians who had much larger thresholds than the other 75%. To summarize all of these findings, a speculative conclusion can be made that non-musicians who did not learn to 'listen carefully' (Spiegel & Watson, 1984) have better pitch

discrimination abilities for the downwards direction than the upwards direction. More research with bigger groups of participants is needed to solidify this finding. It can be said with certainty that musicians show no difference in their pitch discrimination ability between the upwards and downwards moving pitch.

(24)

Experiment IV

:

Pitch Memory

Before we can tackle pitch memory, we have to look at memory in general. There is a clear dichotomy in memory: the short-term memory (STM) and the long-term memory (LTM) (James, 1890). For this experiment, we will be looking into the STM. The STM ranges from 4-30 seconds, but it is usually on the order of 4-8 seconds (Snyder, 2009). Aside of this temporal limit, the STM may also have a capacity limit for the number of items that can persist in the STM at one time. Miller proposed that this number was 7 ± 2 (Miller, 1956), but it has since been reduced to a maximum of 4 or less (Cowen, 2005). For this experiment, pitch memory is the main focus. Simply put, it is the memory of a single pitch over time. This ability is an essential part of building musical expectations and forming scales and keys. It is also an essential part of pitch discrimination, so it has close ties to the previous experiment. You could say that pitch memory forms the ability to discriminate pitches over time. Fewer than 1 in 1000 people possess the ability to identify a pitch outside of any context (Snyder, 2009). This ability is called absolute pitch (AP). Most other people have a relative pitch discrimination ability. This means that pitches can only be identified in relation to other pitches. Pitch memory is a part of the STM for those who possess relative pitch discrimination abilities. For people with AP, the pitch memory has crossed over into their LTM. People who possess AP can link a pitch to a

conceptual pitch name effortlessly, making their pitch memory much more resistant to interference. Interference is the negative effect that other, distracting pitches can have on your memory of a certain pitch. Diana Deutsch has done extensive research on this interference effect in the 1970's, as a part of constructing a model for pitch memory. For the pitch memory system, she phrased

abovementioned two limitations of memory as hypotheses: time and capacity (Deutsch, 1999). She phrased a third hypothesis for the memory of pitch as "the function of an organized system whose elements interact in specific ways." In her book The Psychology of Music, she mentions an extensive array of researches on pitch memory. She starts by stating that when listeners are presented with two pitches directly after each other, most find it very easy to determine whether the two pitches are the same or different (Deutsch, 1970). The response was the same when a 6 second silent interval was introduced. Although it has been shown that pitch memory fades gradually over time (Rakowski, 1994), the effect over a 6 second time span is very small. Interference took effect when eight other tones were interpolated in between the two tones, as participants made 40% more errors with the interpolated tones in comparison to silence. From these findings, Deutsch refines her hypotheses and tries to find an explanation for this interference effect. She comes up with three sub-hypotheses: 1. the interference effect is caused by attention distraction, 2. the interference is caused by a limited capacity of pitch information or 3. the interference is caused by interactions that occur within the system that is specialized for pitch information.

(25)

compare the pitches of two tones that were separated by an interval of 5 seconds (Deutsch, 1970b). The two tones were the same, or differed by a semitone. In these 5 second intervals, either six tones or six spoken numbers were interpolated. There were four conditions: for the first and second, they were instructed to simply ignore the interpolated tones or spoken numbers, and judge whether the two pitches were the same or different. For the third, they were instructed to recall the numbers and compare the two pitches. For the fourth, they were instructed to simply recall the numbers. The results were striking: as expected, the interpolated tones greatly impaired the participants' pitch memory, but the interpolated numbers produced only a minimal impairment, even when the participants were asked to recall them (so they had to direct their attention to them). She also found that these decrements are larger when presented with interpolated tones that are within the octave of the test tones, compared to an octave higher or lower (Deutsch, 1974), and that pitches with close relations to the first or last pitch also have an effect on pitch memory (Deutsch, 1972). These findings caused Deutsch to indicate that "decrements in pitch memory resulting from interpolated tones are due to interactions that take place within a specialized system."

Following this conclusion of Deutsch, one can expect that musicians have more refined specialized systems for pitch memory because of their extensive learning and experience. Surprisingly little behavioral research on simple pitch memory differences between non-musicians and musicians has been conducted. A lot of research has gone into effect of musical training on tone sequence memory, but not the memory for isolated tones. In the researches of tone sequence memory, they found that the memory for these sequences is enhanced by musical training (Minsel, 1995; Williamson, Bradeley & Hitch, 2010).

The aim of this experiment is to find the effect of musical training on the memory for isolated pitches, and the interference effect of interpolated glide tones on that memory. The cleanest way to research this is using sine waves, because these sounds are unnatural and relatively unexposed, so musicians cannot derive a big advantage on that front. To keep in line with Experiment 2, instead of using interpolated tone sequences as distracters, linear (in tone, logarithmic in frequency) glide tones will be used for this experiment. These distracting glide tones will interfere with the pitch memory of the participants, indicating if there is any effect of musical training on resisting this interference. Also, the directional property of the glide tones is an interesting condition in connection to pitch memory. It can provide evidence for the use of a specialized system for direction, in combination with an absolute pitch height system. We expect to see a positive effect of musical training on pitch memory (resisting the interference effect), because musicians might have a more precise or prolonged storage system for pitches due to their musical experience that caused learning. This hypothesis is similar to the

hypothesis Deutsch put forward about people with AP being more resistant to interference because of their ability to link pitch names to pitches. In this case, no musicians will have AP so they would have to have another kind of learned ability, and the most viable hypothesis here seems to be an enhanced specialized STM storage system for pitches, instead of the LTM storage system AP possessors have.

(26)

Also, the influence of an interpolated directional glide can tell us whether directional information is a part of this system or not.

Methods

Experimental Design

Participants were presented with a single sine tone, non-musical in pitch, after which a distracting gliding tone was presented. After the distracter, the same or a different tone was presented. Participants had to respond whether the first tone was the same as, higher than or lower than the second tone using a button box. All tones were simple sine tones. The single tones were 2 seconds long, the distracters were 5 seconds long. There were 1 second silent intervals between the first single tone and the glide and between the glide and the second single tone. This meant that participants had to keep the first tone in their memory for seven seconds in total, but with 2 seconds of silence. There were three different first single tones: 510 Hz, 610 Hz and 710 Hz. This was done to minimize the possibility of participants getting used to the stimuli, which would make the task much easier. The distracters were always the same upward glides and downward glides, which moved from 700 to 560 Hz in 5 seconds (downwards), or vice versa (upwards). The second single tones were presented in three different categories: the same (+/- 0 cents), slightly higher or lower (+/- 45 cents), and a lot higher or lower (+/- 90 cents). Non-musical pitches and pitch intervals were used to rule out any advantage for musicians. A practice block of 4 trials was completed before the start of the actual test.

Results

For this task, a logistic regression with a random intercept model was used, with the following fixed effects: musicianship (musician or non-musician), the direction of the distracting glide (up or down), the difference between the single tones in cents (+/- 0 cents, +/- 45 cents, +/- 90 cents) and the starting frequency (510, 610 and 710 Hz). Musicians were significantly more correct in their responses than non-musicians (p < 0,001). A significant interaction was found between the direction of the glide and the difference between the single tones (p < 0,001), as can be seen in figure 4.1 and 4.2. A third significant effect was found for the starting frequency (F1) (p < 0.01), for which a Tukey test showed that the difference in correct responses between 510 Hz and 610 Hz was highly significant (p = 0.001), and the difference in correct responses between 710 and 610 Hz was nearly significant (p = 0.059). There was no significant difference in correct responses between 710 and 510 Hz. An overview of the mean correct percentages can be found in figure 4.1. An alternative result can be seen in figure 4.2, in which a prediction of the preferred model is shown. This corrects for the random intercept model prerequisites, and is the preferred visualization of the results.

(27)

Figure 4.1 Mean correct percentages for non-musicians and musicians for each condition.

Figure 4.2 Predicted responses using a logistic regression with a random intercept model for the

(28)

For the direction and difference interaction, a Tukey test showed significant differences for the conditions below (see table 4.1).

Table 4.1 P values for all conditions of the direction-difference interaction. Important significant

differences are highlighted.

Discussion

Musicianship

Many significant effects on pitch memory have been found in this experiment. The first and most important one is the effect of musical training. Musicians outperformed non-musicians in every condition (as can be seen in figure 4.2). This provides strong evidence for an enhanced specialized STM pitch storage system as an effect of long-term musical training. Another possibility could be that musicians have better strategies to do the task correctly, but these strategies are likely to be based on a good pitch memory resisting the interference effect of the distracters.

F1- F2 Difference & Distracter Direction

An interesting interaction was found between the distracter direction and the difference between the first test tone (F1) and the last test tone (F2). As is presented in the plots, the direction F2 moved in comparison to F1 is countered by the direction of the distracter glide. On average, more mistakes were made when the glide and the F2 direction moved in the same direction than when the glide and F1-F2 direction moved in the opposite direction. It seems that the distracting glide created a kind of false sense of pitch information for relative pitch perception. For example: when a participant was presented with a downwards glide and a - 90 cent F1-F2 difference, the directional information of the glide interfered with the absolute pitch information stored of F1. The glide seems to 'pull' the absolute pitch information down with it, resulting in participants making more errors in their relational

discrimination, thinking that F1 was lower than it actually was, and labeling F2 as 'same' or even 'higher'. The opposite applies to the upwards glide direction.

Another significant difference was found for the +/- 45 cent condition (see table 4.1). This proves that difficulty increases when the F1-F2 difference gets smaller. However, this F1-F2

Direction, Difference Down, - 90 cents Up, - 90 cents Down, - 45 cents Up, - 45 cents Down, + 45 cents Up, + 45 cents Down, + 90 cents Up, + 90 cents Down, +/- 0 cents Up, +/- 0 cents Up, - 90 cents < 0.001 *** x x x x x x x x x Down, - 45 cents 0.03149 * < 0.001 *** x x x x x x x x Up, - 45 cents 0.99997 < 0.001 *** 0.10494 x x x x x x x Down, + 45 cents 0.99988 < 0.001 *** 0.00471 ** 0.98452 x x x x x x Up, + 45 cents 0.99081 < 0.001 *** 0.33407 0.99994 0.84364 x x x x x Down, + 90 cents < 0.001 *** 0.99999 < 0.001 *** < 0.001 *** < 0.001 *** < 0.001 *** x x x x Up, + 90 cents 0.45188 0.15270 < 0.001 *** 0.16094 0.84747 0.04103 * 0.04291 * x x x Down, +/- 0 cents 0.25692 0.20456 < 0.001 *** 0.06375 . 0.67543 0.01210 * 0.05846 . 1 x x Up, +/- 0 cents < 0.001 *** 1 < 0.001 *** < 0.001 *** 0.00220 ** < 0.001 *** 0.99984 0.25572 0.33359 x

(29)

difference had a much bigger effect with the distracters in comparison to a normal pitch discrimination test. As we know, around 95% of participants can hear differences of under 12 cents, even going up to 2 cents. However, if we compare the +/- 90 and +/- 45 cent conditions, a significant drop-off in accuracy can be seen, so the distracter heavily interferes with the precision of pitch memory. No significant effect is found between the +/- 90 cent difference and the 0 cent difference. These findings prove that the addition of an interpolated distracter greatly affects pitch discrimination abilities over time, for both non-musicians and musicians. More precisely, it can be concluded that the precision of the pitch memory system that is needed for pitch discrimination is harmed by interpolated distracters, caused by interactions inside the specialized system for pitch perception (in the sense of Deutsch, 1999). The results show us that directional information of pitches is a critical part of this system.

Frequency 1 & Endpoint effect

A third significant effect was found: the initial frequency (F1). This effect can be ascribed to two things: whether F1 was in the range of the glide distracter, and an endpoint effect. When F1 was 610 Hz, it was inside the frequency range of the glide, and when it was 510 or 710 Hz it was respectively below and above the frequency range of the glide. For this effect, we can refer to the research done by Deutsch mentioned above, in which she found that the interference effect is larger when the

interpolated tones were within the octave of the test tones (Deutsch, 1974). This effect is replicated here.

Another complicating factor is the endpoint effect. The endpoint here, is the frequency on which the glide stops. This could confuse listeners, although the task was explained carefully. This effect accounts for some anomalies in the percent correct data, but when corrected for the random intercept model, these anomalies disappear. This endpoint effect is not included in the model, because these mistakes were probably made because participants were confused and started comparing the endpoint and F2, instead of F1 and F2, so these are not responses that are representative of the task they were instructed to do. However, it is important to note that this is an experimental design flaw, and it should be corrected in any further research, so that all frequencies are either within or outside of the frequency range of the glide distracter.

(30)

Experiment V

:

Glide Imagery

The ability to hear music or non-musical sounds silently in our heads is a peculiar phenomenon. Just like we have visual imagery in our heads, we also form auditory imagery. In Auditory Imagery, M.J. Intons-Peterson defines it as "the introspective persistence of an auditory experience, including one constructed from components drawn from long-term memory (Intons-Peterson, 1999)." Auditory imagery has a strong connection to memory, but also has a creative quality to it. A famous claim made by many composers is that they can hear their symphonies in their heads before they have even played them or written them down. Auditory imagery is subjective, and can therefore not be directly

measured by an observer. It can however be inferred on the basis of indirect measures (Hubbard, 2010).

For example, pitch acuity in auditory imagery has been researched by P. Janata and K. Paroo of the Darthmouth College in New Hampshire in 2006. In one of their experiments, participants were presented with an ascending sequence of tones following a major diatonic scale, with two conditions: a) all tones leading up to the final tone in the scale were presented, or b) the first three or five tones were presented, and participants had to imagine the next few tones before the final tone was presented (Janata & Paroo, 2006). Participants then had to respond whether the final tone was in either in pitch or in time. They found no difference in participants' pitch judgments between the two conditions, but they did find a severe difference in their timing judgments. Participants were much worse at correctly judging the timing of the final note when they had to imagine the tones. Interestingly, they also found that musical training increased the ability to precisely imagine pitches by comparing non-musicians and musicians performing this task.

This experiment will be similar to that of Janata and Paroo, but instead of musical tones, non-musical glide tones will be used. It is interesting to see if such tones still produce the effect Janata and Paroo found, because glide tones are not necessarily musical in itself. They do have some musical quality, but gliding tones are much more rare in music than diatonic tones, so the effect of exposure and musical training should be much lower. Removing this musical quality from the stimuli might alter the results found by Janata and Paroo. Still, we expect to see a better performance by musicians, because of their enhanced pitch discrimination and pitch memory abilities (see Experiment 3 and 4). Again, the main focus of this experiment was to create a factor for the ability to follow a glide for use in the model of the main experiment.

Methods

Experimental Design

Participants were presented with different gliding tones of 4 seconds, then a silence of 2 seconds, after which the gliding tone was followed up with a correct or incorrect second gliding tone. The correct

(31)

gliding tone is a follow-up of the first tone as if two seconds have been cut out of it, so it continues as it would have after 2 seconds. The incorrect gliding tone is a second glide tone that does not match up with the continued pitch of the first glide tone. The three different conditions were a correct follow-up glide tone, a clearly incorrect follow-up glide tone, which was the same as the first glide tone, and a mildly incorrect glide tone, which was a tone that started higher or lower than it should have if the first tone was uninterrupted. The participants had to respond whether the second glide tone was a correct or incorrect follow-up of the first glide tone by pressing the corresponding button on a button box. Also, two different gliding rates were used: 'fast' and 'slow', for which 'slow' (mean rate of 14 Hz/s) was twice as slow as the 'fast' (mean rate of 28 Hz/s) rate. A third variable was the direction of the glide, which moved up or down.

Results

A logistic regression with a random intercept model was used, with the following fixed effects: musicianship (musician or non-musician), the direction of the glide (up or down), the degree of

Figure 5.1 Predicted responses using a random intercept model, for the probability of being correct

for each significant condition, 0 being incorrect and 1 being correct.

correctness of the second glide (correct, clearly incorrect and mildly incorrect), and the speed of the glide (fast or slow). ANOVA tests showed a significant effect for the follow-up correctness (p < 0.001), where all participants performed best at the clearly incorrect condition, slightly worse at the correct condition, and much worse for the mildly incorrect condition. A significant interaction

(32)

between the speed of the glide and the follow-up correctness (p < 0.001) was found, for which subjects performed significantly worse at the mildly incorrect condition in combination with a slow rate in comparison to a fast rate. No significant effect was found of musical training or the direction of the glide. A graphical overview of the results is shown in Figure 5.1.

Discussion

We found significant differences for the speed of the glide and the correctness of the follow-up glide. For the follow-up glide that was mildly incorrect, participants made much more mistakes for a slow glide rate than for a fast glide rate. This is probably due to the smaller difference between the endpoint of the first glide and the starting point of the second, or the smaller amount of prediction information included in a slow glide rate (less frequencies to make a good estimation). For the 'clearly incorrect' condition, a ceiling effect is apparent. The 'clearly incorrect' condition consisted of the glides that were exactly the same, which was really easy for all participants to hear. For the 'correct' condition, a small difference in favor of the slow glide rate can be seen, but this effect was not significant by a small margin (p = 0.06).

The effect of musical training on pitch acuity in auditory imagery found by Janata and Paroo was not replicated here. The biggest difference between these two researches are that Janata and Paroo used musical tones, for which the last tone was in tune or slightly out of tune, so a sense of key was important in this task. For this experiment, we used glide tones which provided no sense of key. Also, musicians have had much less exposure to and training with glide tones than with separated, diatonic tones. This can explain why there was no difference in acuity between non-musicians and musicians for this task. It tells us that the auditory imagery system is indeed constructed from components drawn from long-term memory, which must include the ability to sense a key from a scale and a training-effect caused by exposure and experience.

Pitch Anticipation: a study of glide tone perception

PITCH ANTICIPATION

A study of glide tone perception

July 15th 2015

Jurre Thuijs

10067469

Master thesis

Arts & Culture: Musicology

Universiteit van Amsterdam

Supervisors: M. Sadakata & J.A. Burgoyne

Table of contents

Abstract

Introduction

General Information

Methods

Experiment I

:

Reaction Time

Methods

Results

Discussion

Experiment II

:

Pitch Anticipation

Methods

Results

Discussion

Experiment III

:

Pitch Discrimination

Methods

Results

Discussion

Experiment IV

:

Pitch Memory

Methods

Results

Discussion

Experiment V

:

Glide Imagery

Methods

Results

Discussion