Bass amplification impacts emotional, neural and physiological responses to music

[thewas]

Something must of us knew intuitively is researched and seems to be proven now:

https://www.sciencedirect.com/science/article/pii/S0003682X25004657

Highlights​

• Bass amplification significantly increases music-evoked arousal, particularly in live music settings.
• Electrodermal activity provides a more sensitive and practical measure of emotional engagement than electroencephalography.
• Our results suggest that bass affects emotional experience through mechanisms beyond low-level auditory processing.

Abstract​

Live music is highly appreciated for its emotional impact, often enhanced by louder sound levels to boost audience arousal and engagement. As high sound levels cause hearing damage and disturb nearby residents, focusing on audio quality offers a safer way to enhance emotional responses to music. However, how quality parameters, such as the balance between low and high frequencies, impact and link emotional, neural and physiological responses is unclear.
This study examines how low-frequency amplification affects listeners' arousal and its connection to neural and physiological responses during music listening. Two experiments were conducted: (i) in controlled laboratory conditions and (ii) in more ecological, live settings.
Subjective reports indicate that amplified bass significantly increases arousal, with a lesser but noticeable effect on valence. Electroencephalography (EEG) recordings show that early auditory components are unaffected by bass amplification, but the arousing effect is linked to enhanced oscillatory features in the low delta (2-5 Hz) frequency range, suggesting active, predictive tracking of music.
In natural music-listening settings, portable electrodermal activity (EDA) sensors were used to measure emotional and physiological responses. Results confirm that bass amplification increases arousal and that EDA better captures emotional integration in response to bass amplification than EEG. This suggests that low frequencies engage additional sensory or emotional circuits beyond traditional auditory pathways, and that EDA provides a more objective and practical measure of emotional responses in naturalistic environments.
Overall, bass amplification effectively enhances the emotional music experience, and EDA is a valuable tool for objectively capturing emotional responses in live settings.

 

[Sokel]

That was really interesting (and dense, I'll have to go back and read it again) .

What's evident is that we are still a little far from understanding lows and the mechanisms we perceive it.

 

[Chr1]

Vive la bassheads!

 

[Keith_W]

Thanks for posting the study. This is a summary and criticism of the study for those who don't want to spend time reading it.

Experiment. This study aimed to look at the emotional impact of three different bass settings on listeners. Music was selected for bass content, and three filters were applied:
- LF-Low: a LPF at 80Hz was applied to simulate small speakers with no subs
- LF-Mid: a LPF at 40Hz was applied to simulate larger speakers
- LF-High: no LPF was applied

The frequency response with the 3 filter settings were verified with a mannikin with in-ear microphones as pictured above (fig 4A in the study). The first thing to note is that the "LF-High" has substantially more bass than a typical HRTF curve, or the Harman headphone target curve. Look closely: up to +15dB at 20Hz. As a reminder, this is what a standard HRTF and Harman curve look like:

On the left, some HRTF's measured with a B&K dummy taken from here. On the right, variance of the Harman curve across 350 headphones from rtings.com. I saw this and went "oookaaaaayyyy" and noted that the authors did not say anything about how they are pumping far more bass energy into their test subjects than normal. Anyway, I decided to read on to see what the rest of the study said.

Data collection: via EEG, EDA, and subjective response. Some explanation is required.

EEG (Electroencephalogram). A grid of electrodes is placed on the head to monitor electrical activity of the brain. The study mentions in passing that EEG's rely on the test subject to remain still. The electrodes are sensitive, and motion can result in spurious electrical activity. The study justifies its use of EEG to monitor "pleasure and activation", which they referred to as "valence and arousal". They justified this method on previous studies, specifically Russell's circumplex model of affect.

EDA (Electrodermal activity). Typically, the test subject wears two rings, and the electrical resistance between the rings is measured. The idea is that sweat gland activity changes the electrical resistance, thus providing a measure of emotional state. Four of their 20 subjects did not show any EDA activity, so they were excluded from the study - meaning that 20 were recruited, but 16 participated.

Subjective response: the study was run in two phases. In the lab, test subjects pointed at a circle on a screen and the x, y coordinates were recorded. The circle refers to Russell's circumplex model of affect which looks something like what I show above. They were also handed a button which they pressed when they felt emotionally aroused - in the paper, they referred to this as "FBP" (Feedback Button Press).

Results. This was the result for two tracks of music:

Subjective appraisal - read this with reference to the circumplex model I showed before. We can see that track 3 produced a cluster of subjective emotional responses varying from sad to angry (with some respondents finding the same track "happy" and "relaxed"!), and track 1 produced a much stronger cluster of "happy". The authors chose to show all 3 filters (low, mid, high bass) in the same graph which makes it difficult to see whether one particular filter setting produced a tighter cluster or not. Regardless, they did some statistical analysis and concluded "we observed a significant impact of the lower end of the playback system's frequency response on both the arousal and valence reported by the participants, but this impact is 20 to 30 times smaller than that of the music track."

EEG appraisal - I don't have the skills to evaluate this section of the study, but the authors concluded that EEG appraisal for emotional response was rather weak with very few p-values reaching statistical significance.

EDA appraisal - study the equalisation gains graph shown above very carefully. Notice something odd? The LPF kicks in from 1kHz for both "low bass" conditions. So the "lowest bass" condition removes all bass, all midbass, and nearly all the midrange as well. The result would be unbelievable tinny. This can provoke an emotional reaction too - one of "annoyance" instead of arousal. But anyway, let's look at the violin graphs.

Think of this as a normal distribution curve flipped 90 degrees and mirrored. The fatter the graph, the more the respondents. A graph like the lower left (condition K) where all the violins are equal for all four bass conditions, indicates that the skin resistance did not change across all four filters. The different violin graphs represent the data collected over indoors and outdoors phases of the experiment and different tracks. This time, analysis of the EDA's did show a correlation between emotional arousal and bass volume, with a few of the graphs reaching statistical significance.

Comment. Unless I am missing something, it seems as if the test subjects listened to equalisation with either much more bass than normal, or much less bass than normal, even removing substantial portions of the midrange. From the study, I agree that that EDA is better than EEG at detecting differences in emotional arousal, although it should be said that EEG is known to produce unreliable results if the subject moves. I am not sure if the author's conclusion that EDA is better than subjective appraisal is valid, my understanding of statistics is too poor to properly analyse that result. Visually studying the clusters was not helpful because of the way the graph was presented.

Whilst I want to believe that more bass = better emotional response, I am not sure if this is a representative study. We need some comment from some of ASR's other academic types. @youngho? @Thomas Lund? @j_j ?

 

[j_j]

Whilst I want to believe that more bass = better emotional response, I am not sure if this is a representative study. We need some comment from some of ASR's other academic types. @youngho? @Thomas Lund? @j_j ?

So many variables.
So few controls.

 

[pablolie]

Thanks for posting the study. This is a summary and criticism of the study for those who don't want to spend time reading it.

Experiment. This study aimed to look at the emotional impact of three different bass settings on listeners. Music was selected for bass content, and three filters were applied:
- LF-Low: a LPF at 80Hz was applied to simulate small speakers with no subs
- LF-Mid: a LPF at 40Hz was applied to simulate larger speakers
- LF-High: no LPF was applied

View attachment 474404
The frequency response with the 3 filter settings were verified with a mannikin with in-ear microphones as pictured above (fig 4A in the study). The first thing to note is that the "LF-High" has substantially more bass than a typical HRTF curve, or the Harman headphone target curve. Look closely: up to +15dB at 20Hz. As a reminder, this is what a standard HRTF and Harman curve look like:

View attachment 474406View attachment 474405

On the left, some HRTF's measured with a B&K dummy taken from here. On the right, variance of the Harman curve across 350 headphones from rtings.com. I saw this and went "oookaaaaayyyy" and noted that the authors did not say anything about how they are pumping far more bass energy into their test subjects than normal. Anyway, I decided to read on to see what the rest of the study said.

Data collection: via EEG, EDA, and subjective response. Some explanation is required.

EEG (Electroencephalogram). A grid of electrodes is placed on the head to monitor electrical activity of the brain. The study mentions in passing that EEG's rely on the test subject to remain still. The electrodes are sensitive, and motion can result in spurious electrical activity. The study justifies its use of EEG to monitor "pleasure and activation", which they referred to as "valence and arousal". They justified this method on previous studies, specifically Russell's circumplex model of affect.

EDA (Electrodermal activity). Typically, the test subject wears two rings, and the electrical resistance between the rings is measured. The idea is that sweat gland activity changes the electrical resistance, thus providing a measure of emotional state. Four of their 20 subjects did not show any EDA activity, so they were excluded from the study - meaning that 20 were recruited, but 16 participated.

View attachment 474408

Subjective response: the study was run in two phases. In the lab, test subjects pointed at a circle on a screen and the x, y coordinates were recorded. The circle refers to Russell's circumplex model of affect which looks something like what I show above. They were also handed a button which they pressed when they felt emotionally aroused - in the paper, they referred to this as "FBP" (Feedback Button Press).

Results. This was the result for two tracks of music:

View attachment 474409

Subjective appraisal - read this with reference to the circumplex model I showed before. We can see that track 3 produced a cluster of subjective emotional responses varying from sad to angry (with some respondents finding the same track "happy" and "relaxed"!), and track 1 produced a much stronger cluster of "happy". The authors chose to show all 3 filters (low, mid, high bass) in the same graph which makes it difficult to see whether one particular filter setting produced a tighter cluster or not. Regardless, they did some statistical analysis and concluded "we observed a significant impact of the lower end of the playback system's frequency response on both the arousal and valence reported by the participants, but this impact is 20 to 30 times smaller than that of the music track."

EEG appraisal - I don't have the skills to evaluate this section of the study, but the authors concluded that EEG appraisal for emotional response was rather weak with very few p-values reaching statistical significance.

View attachment 474411

EDA appraisal - study the equalisation gains graph shown above very carefully. Notice something odd? The LPF kicks in from 1kHz for both "low bass" conditions. So the "lowest bass" condition removes all bass, all midbass, and nearly all the midrange as well. The result would be unbelievable tinny. This can provoke an emotional reaction too - one of "annoyance" instead of arousal. But anyway, let's look at the violin graphs.

View attachment 474414

Think of this as a normal distribution curve flipped 90 degrees and mirrored. The fatter the graph, the more the respondents. A graph like the lower left (condition K) where all the violins are equal for all four bass conditions, indicates that the skin resistance did not change across all four filters. The different violin graphs represent the data collected over indoors and outdoors phases of the experiment and different tracks. This time, analysis of the EDA's did show a correlation between emotional arousal and bass volume, with a few of the graphs reaching statistical significance.

Comment. Unless I am missing something, it seems as if the test subjects listened to equalisation with either much more bass than normal, or much less bass than normal, even removing substantial portions of the midrange. From the study, I agree that that EDA is better than EEG at detecting differences in emotional arousal, although it should be said that EEG is known to produce unreliable results if the subject moves. I am not sure if the author's conclusion that EDA is better than subjective appraisal is valid, my understanding of statistics is too poor to properly analyse that result. Visually studying the clusters was not helpful because of the way the graph was presented.

Whilst I want to believe that more bass = better emotional response, I am not sure if this is a representative study. We need some comment from some of ASR's other academic types. @youngho? @Thomas Lund? @j_j ?

Thanks for the excellent summary! My reactions:

(1) I am not sure I want to have a measurable *physical* response to music and especially loud bass. In fact, I hate *feeling* excessive bass, which we have probably all experienced in nightclubs or concerts. Not my personal preference even remotely.

(2) The more subdued bass delivery curves seem to be extremes (especially the one that tanks and basically eliminates bass - that would be indeed sound terribly tinny) and it does not take a study to establish critical info is missing there. Of course people will gravitate towards the presentation that approximates a linear 20-20k response the most, we'd hope. It's a bit like asking people to state their preference on cheeseburgers with one having no cheese and the other one being flowing with melting cheese goodness.

I am no bass-head, but of course agree bass extension is a key aspect to authentic music sound. We can argue about how much is too much and finer aspects of it (which we have done often in other threads). To me the key is it needs to sound *balanced*. I abhor lazy bathtub frequency delivery.

 

[JSmith]

This indicated very low frequencies made people dance more;

https://www.cell.com/current-biology/fulltext/S0960-9822(22)01535-4

We tested whether non-auditory low-frequency stimulation would increase audience dancing by turning very-low frequency (VLF) speakers on and off during a live electronic music concert and measuring audience members’ movements using motion-capture. Movement increased when VLFs were present, and because the VLFs were below or near auditory thresholds (and a subsequent experiment suggested they were undetectable), we believe this represents an unconscious effect on behaviour, possibly via vestibular and/or tactile processing.

https://www.cell.com/action/showPdf?pii=S0960-9822%2822%2901535-4

JSmith

 

[pablolie]

This indicated very low frequencies made people dance more;

https://www.cell.com/current-biology/fulltext/S0960-9822(22)01535-4

...

When I saw Jurassic Park, people weren't dancing to the beat of the Tyrannosaur walking up the the Jeep. Bass marks the beat, so of course total lack thereof will not invite happy feet. Doesn't mean an excess of it is needed to get people to dance. "If the presence of bass makes people dance, then more and more bass will make more people dance" has all the elements of a logical fallacy.

 

[j_j]

When I saw Jurassic Park, people weren't dancing to the beat of the Tyrannosaur walking up the the Jeep. Bass marks the beat, so of course total lack thereof will not invite happy feet. Doesn't mean an excess of it is needed to get people to dance. "If the presence of bass makes people dance, then more and more bass will make more people dance" has all the elements of a logical fallacy.

I can tell you that every dance place I know of uses the same old "boner" curve (yes, yes, they call it that, sorry) that looks like a smile. 20dB of bass boost, and 15 at 15kHz, and zero at 1k.

It sounds awful, but it's pervasive, and has caused literal fights in the mixing booth when something was a show instead of a dance. Some people do not want to adapt. Ever. I've had somebody push me off the chair and adjust the EQ (which was proper set for the act) because "the house guy does it that way". I had to get a roadie help out.

Handy things, roadies.

 

[Thomas Lund]

Whilst I want to believe that more bass = better emotional response, I am not sure if this is a representative study.

There are several studies of music’s emotional influence on listeners, for instance J. Panksepp & G. Bernatzky (2002) "Emotional sounds and the brain: the neuro-affective foundations of musical appreciation”, R. Zatorre & V. Salimpoor (2013) "From perception to pleasure: music and its neural substrates”, N. Kathios et al. (2023) "Musical anhedonia, timbre, and the rewards of music listening”; and also studies of low frequency sound (vibroacoustics) in the treatment of pain.

In the auditory envelopment (AE) studies published, we used abstract LF sound, i.e. noise. Humans could possibly be more sensitive to AE stimulation in association with music. With regards to live music and AE, there is a difference in the modulation of AE at each listener, depending on amplification or not. Discrete instruments in a hall or a club generate more such contrasts than a PA.

A large PA, however, is able to convey high LF and VLF sound levels, which seems what was investigated here. From ISO226 curves ("Fletcher-Munson”) it can also be noted how a small drop in VLF level is highly noticeable. Maybe we don’t like this when dancing or being otherwised sparked by enjoyable music. Anyway, thanks for the prod.