1

TINNITUS HUB

EVERYTHING YOU
NEED TO KNOW

TINNITUS
MODELS
Included in this
document:
Tinnitus models

Hyperactivity models

Filling-in models

Predictive coding models

w w w. t i n n i t u s h u b . c o m

w w w. t i n n i t u s h u b . c o m
Contents
Tinnitus models 2
An introduction to tinnitus models 3

Overview of all models 5

Hyperactivity models 8
Central gain 9

Frontostriatal gating 10

Thalamocortical dysrhythmia 11

Filling-in modles 13

Predictive Coding Models 15

Auditory Attention model 17

Sensory Precision Integrative Model 17

Perception-update (PU) model 18

Neural synchrony 20

Attention in Tinnitus 21

Final thoughts 23

Appendix 24

Bibliography 25

Towards a better
understanding of tinnitus
Many people with tinnitus want to understand
where the tinnitus sound comes from - how
does your body generate a sound that’s not re-
ally there? This resource will help you unders-
tand the different ways in which tinnitus could
be generated. We will provide an overview of
the tinnitus models that are currently being di-
scussed in scientific literature, as well as a fra-
mework to assess the explanatory power of
these models.
You will learn that tinnitus is now considered a
condition of the brain. However, this was not
always the case. It was long thought that tin-
nitus was an ear problem. Various conditions
of the ear can trigger tinnitus, but for tinnitus
to become persistent we need to look beyond
the ear as well.
Chapter 1 presents an introduction into tinnitus
models research and an overview. Then, we
move on to the three main models that have
1
been theorised to explain tinnitus: hyperacti-
vity models, filling-in models, and predictive
coding models. Hyperactivity models (Chap-
ter 2) state that tinnitus results from excessive
spontaneous activity in the auditory pathway,
which the higher brain regions then interpret as
sound. Filling-in models (Chapter 3) are based
on the brain “filling in” any missing sound input
following auditory deprivation such as hearing
loss. Finally, predictive coding models (Chap-
ter 4) think of the brain as a “prediction machi-
ne” that constantly generates and updates a
mental model of the environment, and views
tinnitus as the result of prediction errors. Then
we address auditory attention (Chapter 5), as
attention is an important factor in tinnitus re-
search. We conclude with some final thoughts
and questions for the future.
w w w. t i n n i t u s h u b . c o m
01
2

TINNITUS MODELS
INTRODUCTION
To understand tinnitus, we first have to understand more about
hearing in general. How does a sound normally travel from the
environment into our consciousness?

“The auditory pathway” describes how sound, when it reaches our

ears, makes its way to the brain.

w w w. t i n n i t u s h u b . c o m

Tinnitus models
It all starts in the outer ear, consisting of the
pinna or auricle (the visible part of the ear), the
ear canal, and the ear drum, on the left side of
the image. Sound can be described as a wave
which causes vibrations in the air, and when it
reaches the ear drum, sound waves make the
ear drum vibrate. The vibration of the ear drum
in turn transfers the sound to the middle ear,
which is an air-filled cavity containing three
small bones: the malleus, the incus, and the
stapes. These three bones, collectively called
the ossicles, move because of the vibrations
of the ear drum. The stapes is attached to the
Medical geniculate
Inferior colliculus
Superior olive
Cochlear nucleus
reached the auditory nerve, we enter the realm
of the central hearing system. This difference
in peripheral and central pathway is important,
as different theories place the origin of tinnitus
in different parts along the hearing pathway. All
the models in this leaflet place their emphasis
on central mechanisms, as they are receiving
the most attention in current literature.
The auditory nerves from each ear take the
sound signal to the cochlear nuclei, from whe-
re the signal is relayed across different structu-
res in the brain. From the cochlear nuclei, part
3
inner ear through a membrane, and its vibrat-
ions make the fluid in the cochlea move in a
wave-like manner. The cochlea is a coil-sha-
ped bone in the inner ear which contains hair
cells which are stimulated by the movement of
the inner-ear fluid. The hair cells are microsco-
pically small, and they change the sound wa-
ves into a signal that can be transmitted to the
brain through the vestibulocochlear nerve (the
auditory nerve). At this point, we have arrived
at the end of what is called the peripheral part
of the auditory pathway. Once the sound has
of the sound signal also crosses over to the
Auditory receiving
centers
Medical geniculate body
Inferior quadrigeminal body
Lateral lemniscus
Olivary nucleus
Dorsal cochlear nucleus
Ventral cochlear nucleus
Vestibulocochlear nerve
other side of the brain, which is called to de-
cussate. Therefore, both the left and right he-
mispheres or sides of the brain under normal
circumstances receive sound information from
both ears. After several processing steps, the
signal arrives in the medial geniculate body in
the thalamus, which also processes sensory in-
formation arriving from for example the visual
pathway. The thalamus is often thought of as a
“hub”; it is an important gray matter structure in
the brain responsible for information exchange
but also regulation of consciousness and sleep.
w w w. t i n n i t u s h u b . c o m
4

We actually have two thalami: one on each side
of the brain, and in the majority of people they
are connected. From the thalami, there are ner-
ve fibres projecting out to the cerebral cortex,
and the fibres relevant for hearing project out
to the auditory cortex. The auditory cortex can
be seen as the “final destination” of the sound
signal and marks the end of what we call the
primary pathway. However, even after all the-
se steps, the processing of sound is still very
basic, and it is only when the auditory cortex
communicates with other regions of the brain
that we make sense of the sounds coming into
our ears.
At present, most researchers look at the central
part of the auditory pathway to find an expla-
nation. However, there is not one model that
explains it all, but each has its own explanatory
power. It could be that different models explain
different subtypes of tinnitus.
Another important question is: if there is no
sound actually coming into the ear, like is nee-
ded to activate the primary auditory pathway,
then where is that sound generated? We will
now turn to a discussion of the most prominent
models and theories of tinnitus.

ONE KEY QUESTION IS: AT WHAT STAGE IN ALL OF THIS,

FROM RECEIVING A SOUND IN THE EAR TO BECOMING
CONSCIOUS OF THAT SOUND, SHOULD WE LOOK TO FIND
AN EXPLANATION FOR TINNITUS?

w w w. t i n n i t u s h u b . c o m

Overview of all models
We will start by presenting a table with the diffe-
rent models we will discuss in this leaflet. The-
se models were chosen because they all tackle
what we have dubbed the “hard problem” of
tinnitus, i.e. how can we explain the existence
of tinnitus and how it originates? There will be
other models in the literature as this list is not
intended to be exhaustive, but we believe this
is a good representation of the models current-
ly used and talked about by researchers. Mo-
dels that focus more on reactions to tinnitus rat-
her than its origins will not be discussed in this
leaflet in detail. We will only address attention
in tinnitus in a final paragraph as attention is a
key factor of investigation in the field of tinnitus.
The table on the next page is adapted from
Sedley et al. (2016) and is a tool of assessing
the explanatory power of the different models
in this leaflet. We ask: how well does this model
explain common observations in tinnitus? For
example, when people undergo surgical cut-
5
ting of the auditory nerve (as part of medical
treatment for a tumor for example), there are
accounts of their tinnitus improving, but there
are also accounts of the tinnitus being exacer-
bated. If the model can explain these pheno-
mena, the table will list a “yes”, if it cannot, it will
list a “no”. Another key phenomenon in tinnitus
is that some people with hearing loss get tinni-
tus, whereas other people do not. There is no
one answer to why this is the case, but we can
ask about these models if they can account for
it. Finally, one might ask why the tinnitus often
begins later than the onset of hearing loss rat-
her than at the same time.
In addition to the overview of tinnitus models,
in the appendix you will find an explanation of
EEG (electroencephalography) as an example
research method for the interested reader, as
some of the models will draw upon knowledge
derived from EEG, which is a tool to measure
brain activity.
w w w. t i n n i t u s h u b . c o m
6

TINNITUS MODELS
OVERVIEW 1/2

Improvement by Exacerbation by
auditory nerve auditory nerve
section section

Central Gain No Yes

Frontostriatal Gating No Yes

Thalamocortical No Yes
Dysrhythmia

Filling-in Models No Yes

Auditory Attention No Yes

Sensory Precision Potentially Yes

Integrative Model

Sensory Precision Potentially Yes

Integrative Model

Perception Potentially Yes

Update Model

Neural Synchrony No Yes

w w w. t i n n i t u s h u b . c o m
 7

TINNITUS MODELS
OVERVIEW 2/2

Occurence of Occurence of tinnitus

hearing loss later than onset of
without tinnitus hearing loss

Central Gain Potentially Potentially

Frontostriatal Gating Yes Potentially

Thalamocortical Potentially Potentially

Dysrhythmia

Filling-in Models No Potentially

Auditory Attention Yes Potentially

Sensory Precision Yes Yes

Integrative Model

Sensory Precision Yes Yes

Integrative Model

Perception Yes Potentially

Update Model

Neural Synchrony Yes Yes

w w w. t i n n i t u s h u b . c o m
02
8

HYPERACTIVITY
MODELS
Hyperactivity models state that tinnitus is caused by excessive
spontaneous neural activity relayed from “subcortical” regions
to auditory cortex. As discussed in the introduction, the cortical
part of the auditory pathway (the auditory cortex) could be seen
as the “final destination”. Therefore, “subcortical” models place
their emphasis on the parts of the auditory pathway that come
before the cortex. We have discussed there are quite a few
steps that happen before sound reaches the auditory cortex.
The sound is first changed into a signal that can be interpreted
by the brain and triggers neural activity.

However, a key feature of our brains is that it is always active so

long as you are alive. This activity is also called “spontaneous
activity”. Imagine this as being in total darkness, but still seeing
random colours and shapes to some extent. Spontaneous acti-
vity is also observed along the auditory nerve, even in silence.
One theory says it could be that this spontaneous activity beco-
mes excessive, and is then interpreted as a meaningful sound
signal when it reaches auditory cortex.

w w w. t i n n i t u s h u b . c o m
 9

Three most prominent

HYPERACTIVITY MODELS
1. Central Gain 2. Frontostriatal 3. Thalamocortical
Gating Dysrhythmia

These three models explain how the spontaneous activity can become excessive and could
potentially explain tinnitus.

Central gain
“Central gain” refers to a compensatory increa-
se in the central auditory activity in response
to the loss of sensory input (Auerbach et al.,
2014). By central, we mean activity in the cen- AUDITORY DEPRIVATION
tral nervous system, in this case starting from & CENTRAL GAIN
the auditory nerve. A loss of sensory input can
occur because of hearing loss for example, as +
less sound signal is being generated, which
can lead to decreased activity along the audi- ALTERED SPONTANEOUS
tory pathway. However, a key feature of neural NEURONAL ACTIVITY
activity is that it tries to remain constant. This
happens because of homeostatic mechanisms
which try to stabilise the average neural activity
after decreased input from the auditory nerve.
This stabilisation can be achieved in two ways: TINNITUS
by increasing excitatory gain and decreasing
inhibition. The result is that the neurons be-
come more excitable, amplifying the sponta-
neous activity, leading to hyperactivity (Scha-
ette and McAlpine, 2011). This hyperactivation
could then be interpreted by the brain as an
actual sound. The model can explain how hea-
ring loss could lead to tinnitus, but it does not
address why not everyone with hearing loss
gets tinnitus. It can also explain how ear plug-
ging or cutting of the auditory nerve can cause
tinnitus as these are also scenarios in which a
loss of input occurs.

w w w. t i n n i t u s h u b . c o m
10
Frontostriatal gating
Frontostriatal circuits are neural pathways that
connect frontal regions (behind the forehead)
of the brain to the basal ganglia (or “striatum”).
These circuits act as central gatekeepers,
which evaluate the relevance and affective
meaning of sensory stimuli such as sounds.
They modulate information flows via descen-
ding and cortico-cortical pathways. In the intro-
duction, we discussed the primary auditory pa-
thway with the auditory cortex as the end-point.
In this case, information is going up to the brain,
called ascending. Information can also travel
“away” from the brain, such as when we move
our limbs, and this is made possible through
descending pathways. Different regions of the
brain’s cortical structures also communicate
with each other, through cortico-cortical pa-
thways. It is suggested that if the frontostriatal
circuit is faulty, there is a lack of suppression of
irrelevant sensory information, resulting in tinni-
tus (Rauschecker et al., 2015).
Tinnitus Talk interviewed Dr. Rauschecker in
one of their podcasts in which he describes his
model in two steps: the first step is central gain
Default prediction
of silence
Based on Sedley et al. (2016),
“An Integrative Tinnitus Model
Based on Sensory Precision“
Spontaneous
auditory
activity
Deficient Central
noise Gain
cancelling
Delta/theta
synchrony
GABA
deficiency
w w w. t i n n i t u s h u b . c o m
as discussed above, and the second step is a
failure of the gating mechanism which allows
the hyperactivity to get through to conscious-
ness in the form of tinnitus sensation. You can
listen to the podcast here. Dr. Rauschecker also
thinks about the gating theory as a “switch”:
the gating system might turn on and off, which
could explain why a lot of people do not notice
their tinnitus every awake minute, but rather in
specific situations. The model can also explain
why not everyone with hearing loss gets tin-
nitus, because some people’s gating systems
might be intact, and this gating system could
be seen as a “second line of defense”. Howe-
ver, this model as well as central gain relies on
a loss of sensory input to explain tinnitus. This
might be hard to reconcile with observations of
people that have tinnitus but no visible hearing
loss. These observations are controversial and
there are researchers that suggest that in these
cases a “hidden” hearing loss could exist, as
this has been found in animals (Furman et al.,
2013).
Tinnitus
Cross-modal
inputs
Cholinergic
excess
Attention
Maladaptive Additional
plasticity factors?

Thalamocortical dysrhythmia
Another hyperactivity model is called thala-
mocortical dysrhythmia (Vanneste et al., 2018).
This model is about the coupling of the thala-
mus to cortical regions (such as auditory cor-
tex). As we can recall from the introduction and
Figure 1, the last step in the auditory pathway
before sound reaches the auditory cortex, is
the medial geniculate body. This structure is
part of the thalamus. We saw the thalamus is
a very important part of the brain, which can
be seen as a “hub” where all sorts of sensory
information is exchanged. It plays an important
role in relaying sensory signals to the cortex.
How does this link to tinnitus?
As discussed in the two previous hyperactivity
models, tinnitus can be attributed to a decrease
in activity coming into the auditory nerve, trig-
gering hyperactivity through homeostatic me-
chanisms, as well as a failure in cancelling out
excessive spontaneous activity, or a failure in
the “gating system” as Dr. Rauschecker descri-
bes it. It is proposed that these two processes
alter the auditory thalamocortical signal trans-
mission, leading to thalamocortical dysrhyth-
mia (TCD). Neural activity can be categorised
according to different frequency waves, and it
gets quite technical here. Brain waves are typi-
cally divided into delta, theta, alpha, beta, and
gamma bands (ranked in increasing frequency,
for further reading see down below).
γ Auditory
Severe auditory deafferentation
Based on De Ridder et al. (2015),
“Thalamocortical Dysrhythmia:
A Theoretical Update in Tinnitus“
11
In the introduction of this document, we spo-
ke about EEG (electroencephalography), which
can be used to measure these brain waves.
Thalamocortical dysrhythmia can be observed
using EEG, and is characterised by a slowing
down of the resting state alpha activity to theta
activity, whereas the gamma activity increases.
Gamma waves are the fastest brain waves and
are thought to modulate perception and con-
sciousness. The idea is that in TCD, theta and
gamma activity are coupled, and this coupling
allows for theta burst firing to integrate the tin-
nitus-related gamma activity into conscious-
ness networks (De Ridder et al., 2015). TCD is a
theory that is not exclusive to tinnitus but also
used in research into for example neuropathic
pain and depression. It also relies on the as-
sumption of decreased sensory input arriving
at the thalamus, and therefore just like the
hyperactivity models above it cannot explain
why people without hearing loss might suffer
from tinnitus. However, as said before, there
is discussion in the scientific community about
whether tinnitus patients who do not present a
hearing loss in standard diagnostic tests might
have a “hidden” hearing loss.
Consciousness
Enabling
Network
Cortex
γ
Thalamus Parahippocampus
Ear
w w w. t i n n i t u s h u b . c o m
12

Conclusion on
Hyperactivity models
Hyperactivity theories can explain a lot of different phenomena observed in tinnitus. They rely on
the premise that there is a loss of sensory input, resulting in hyperactivity which is not cancelled
out properly by the brain.

Therefore, it can explain different scenarios in which a loss of input occurs, such as the emergen-
ce of tinnitus through hearing loss, ear plugging, and exacerbation of tinnitus by cutting the audi-
tory nerve. However, they cannot explain observations of patients in whom alleviation occurred
after cutting the auditory nerve.

w w w. t i n n i t u s h u b . c o m
03
13

FILLING-IN
MODELS

w w w. t i n n i t u s h u b . c o m
14
Filling-in models
Filling-in models state that the tinnitus percept
is a consequence of the brain “filling in” the in-
put that is missing at parts of auditory cortex
that are “deafferented”, meaning there are mis-
sing nerve cells leading up to those parts.
Going back to what was discussed in the intro-
duction, hair cells in the cochlea transform the
physical sound waves into a sound signal that
can ascend the auditory pathway and reach the
auditory cortex. The cochlea is a coil-shaped or
shell-shaped bone in the inner ear filled with
fluid and lined with microscopic hair cells. A nu-
ance can be added here: the specific location
of the hair cell in the shell determines its sound
frequency sensitivity (See Figure 2 below). The
hair cells in outer parts of the cochlea, or the
“base”, are sensitive to high-frequency or high-
pitched sounds, and if you move closer to the
“apex” or middle, they become sensitive to
low-frequency sounds. This organisation is cal-
led a tonotopic organisation. This organisation
is also reflected in the auditory cortex, where
specific parts of the auditory cortex respond to
frequency-specific information.
2,000 Hz
1,500 Hz
600 Hz
400 Hz
apex
cochlear
duct 200 Hz
800 Hz
20,000 Hz
1,000 Hz
7,000 Hz
5,000 Hz
w w w. t i n n i t u s h u b . c o m
Now, if there is damage to the hair cells at a
specific frequency, this results in a reduced sig-
nal arriving in the corresponding part of the au-
ditory cortex. We can call this deafferentation of
that part, and the filling-in model proposes that
the brain fills in the missing input at that speci-
fic part of the auditory cortex. According to the
theory, these deafferented parts could receive
their input from adjacent normally-functioning
cortex in the case of minor hearing loss, or via
auditory memory retrieval in the case of seve-
re hearing loss (De Ridder et al., 2014). In the
first case, one would expect the tinnitus pitch
to be adjacent to the frequency of the hearing
loss, and in the second case of auditory memo-
ry retrieval, the tinnitus pitch should precisely
match the hearing loss. Hard clinical evidence
for either of these is lacking, however it might
be that this is the case for a subtype of tinnitus.
The theory would be difficult to uphold for peo-
ple who report their tinnitus pitch fluctuates for
example.
high-frequency waves
(1,500 - 20,000 Hz)
3,000 Hz basilar membrane
base apex
medieum-frequency waves
(600 - 1,500 Hz)
basilar membrane
base apex
4,000 Hz low-frequency waves
(200 - 600 Hz)
basilar membrane
basilar
membrane
base apex
04
15

PREDICTIVE
CODING
MODELS
Predictive coding models see the auditory pathway (described in
the Introduction) as hierarchical. Lowest levels of the hierarchy are
in the inner ear, where the sound is first received. As the sound
travels through the auditory pathway to the auditory cortex, it is
received by higher levels of the hierarchy. Our brain builds up our
perception of the world at each level by making inferences about
the environment (Hullfish et al, 2019).

w w w. t i n n i t u s h u b . c o m
16
Predictive coding models
At each level of the hierarchy, the brain uses
an “error unit” to compare the evidence (the
sounds you are receiving from the real world)
from the lower levels to the prior expectations
about the sound held in memory of that par-
ticular level. This comparison allows the brain
to adjust its prediction through an “error sig-
nal” created by the error unit. The error signal
increases in strength to reflect how much the
sounds from the real world differ from your ex-
pectations.
The error signal can be modified by precision
(how confident the brain is about its own ex-
pectations), as well as emotions, attention
and memory (Eggermont & Kral, 2016; Sedley,
EVIDENCE
outside sound info
1 2
4 3
ESTIMATE
BECOMES
PRIOR EXPERIENCE
w w w. t i n n i t u s h u b . c o m
2019). The information that was corrected by
the error signal is then sent up to the next level
in the hierarchy as the updated evidence, whe-
re the same process will occur. The updated
information at the current level also becomes
the new, more precise prior expectation for the
next time such sounds are presented at this le-
vel.
Predictive coding occurs as a cyclical process.
Environmental sounds and prior expectations
are compared in the error unit. Then , an error
signal is created through this comparison. The
error signal updates the prior expectation,, so
next time a similar sound is perceived, it will be
compared to the updated expectation.
ERROR
comparison between
prior expectations &
real world sounds
ESTIMATE
error updates the brain‘s estimate
of what the sound should be,
passes this on as evidence to to
the next error unit
based on Hullfish et al., 2019

Three most prominent
PREDICTIVE CODING MODELS
1. Auditory Attention 2. Sensory Precision
Model Integrative Model
Auditory attention model
The proposed model of auditory attention re-
lies on a person learning the contexts in which
certain sounds are expected to be heard (Ro-
berts et al., 2013). When a person with normal
hearing hears a sound, expected sounds ba-
sed on memory of a specific context are com-
pared to sounds from the environment (like in
the diagram above). If the memory of the con-
text and sounds coming from the real-world
match, they cancel each other out and existing
sound representations continue to guide cog-
nitive processing and behaviour.
However, when a person develops tinnitus as
a result of hearing loss, this process is chan-
Sensory precision integrative model
Neurons in the auditory systems can fire spon-
taneously, generating signals, without any in-
formation from the real world underlying those
signals. This spontaneous neural activity is nor-
mally much weaker than signals based on true
sensory information, so the brain ignores it.
Even an error signal can be spontaneous, and
so it is not based on any real comparisons oc-
curring in the error unit. In the graph above, the
error signal is the signal that is used to adjust
the existing expectation based on the com-
17
3. Perception-Update
Model
ged. As the sounds coming from outside are
now limited, the expected sounds are not ful-
ly matched by the sound coming from the en-
vironment. Auditory attention then attempts
to present a more accurate representation of
the environment, incorporating irregular neural
activity from regions where hearing has been
lost. Auditory attention becomes constantly
used to make up for the lack of input from the
environment, and the neural activity it picks up
is consciously perceived as tinnitus in the hea-
ring-impaired ear.
parison between prior expectations and real
sounds from the environment.
The Sensory Precision Integrative Model focu-
ses on spontaneous errors signals, called “tin-
nitus precursors”. Tinnitus precursors have low
precision (e.g., are not supported by memory
or attention) and so these error signals are ig-
nored because the real sensory information is
seen as more convincing during the compari-
son. However, the tinnitus precursor could be
given too much precision and be unusually
w w w. t i n n i t u s h u b . c o m
18
strong, in which case the brain does perceive
a phantom sound instead of silence. The rea-
son for the change in strength or precision of
the signal may include other models of tinni-
tus, for example insufficient sensory gating me-
chanisms, attentional differences, or a memory
of a sound in people with hearing loss. If the
tinnitus precursor continues to be perceived,
it may start to gain precision, and could even-
tually replace the old default prior expectation
of silence with a new default prior expectation
of the tinnitus sound.
Importantly, the new prior expectation of tin-
nitus does not need the factor that allowed
it to become the default to remain present
(for example, when tinnitus occurs as a result
of noise exposure or ear plugging). In other
words, even if the original instigator of tinnitus
is resolved, the prior expectation of tinnitus
could remain, as it has become the new norm.
The change in prediction can be measured
with Mismatch Negativity evoked potentials
via EEG (Sedley et al, 2019). Mismatch Negati-
vity occurs at around 150-250 ms after detec-
tion of a change in the environmental sounds.
This signal is usually studied through oddball
Perception-update (PU) model
This model suggests that tinnitus occurs as a
result of “data compression error”. Changes in
the characteristics of a sound (e.g., pitch) are
constantly updated in the inner ear by compa-
ring the newest sound to the sound just befo-
re. A prediction error occurs when there is a
change in sound characteristics, which is sig-
nalled from the inner ear up to the auditory cor-
tex. This prediction error is represented by the
“On” and “Off” N1 evoked EEG potentials in this
model (see Appendix for more information). N1
detects change by comparing information of a
previous stimulus with the next stimulus (Noda
et al, 2018). Two subtypes of N1 response are:
w w w. t i n n i t u s h u b . c o m
paradigms (e.g., Sedley et al, 2019), where an
irregular stimulus occasionally appears amid
identical repeating stimuli.
For example, Mismatch Negativity was used
to explore the differences in the perception
of volume change at a frequency of tinnitus in
people who experience tinnitus compared to
those who do not (Sedley et al, 2019). It was
found that people with tinnitus had larger Mis-
match Negativity responses (and thus respon-
ded more strongly) to an event where the vo-
lume of a tone that sounded like their tinnitus
became unexpectedly louder, compared to the
responses in people who did not experience
tinnitus. In the same groups, smaller Mismatch
Negativity responses were noticed in people
with tinnitus compared to those without tinnitus
when the volume became unexpectedly quie-
ter.
A working interpretation of these findings was
that, when the volume became quieter, it soun-
ded more like tinnitus in participants who ex-
perienced it, so the sound was expected, un-
like for those who did not experience tinnitus.
1) the On-N1, which is seen when a new sound
begins after silence, and 2) the Off-N1, which
can then be seen when that particular sound
stops. The bigger the change between the
previous sound characteristics and the new
sound characteristics, the larger N1 responses
are. As the On and Off N1 responses show the
relative magnitude of the change between pre-
vious sound and the current sound, rather than
their absolute magnitude, these responses can
regulate how strong the prediction error that
updates expectations of the error unit is.
Higher levels of the hierarchy generate their
expectations of what the sounds will be, and

send these to lower levels. The lowest level of
the hierarchy is the inner ear, where real life
sounds are compared to expectations from
the higher-level regions. If there is a change
in sound, a prediction error is created to up-
date the expectations of the higher levels.
This occurs at each level of the hierarchy, so
the process is bidirectional, which helps the
brain make more accurate expectations about
the sounds. The next level of the hierarchy is
called the representative higher level, becau-
se the predictive values are created there and
are sent down to the inner ear. At the highest
levels, the predicted value is compared to the
“perceptual value” (sound volume of a specific
frequency). Usually, the predictive and percep-
tual values are similar. Comparisons have to
reach the highest level for sounds to be con-
sciously perceived.
In the case of hearing loss, when there is no
change in sensory input, the auditory system
becomes uncertain of the input it is meant to
the perceptual value
prediction
“the predicted value“
prediction
prediction
(current DATA)
Inner Ear
19
perceive. The predictive and perceptual values
may then drift from the sensory information
that they would otherwise receive, potentially
due to amplified error signals between diffe-
rent levels. When the drift is large, a phantom
noise may be perceived as the predicted value
becomes very different from the environmental
input.
The difference between the PU model and the
Sensory Precision Integrative model is that, in
the latter, the error signals can occur randomly,
independent of hearing loss, whereas the PU
model requires auditory system damage, such
as hearing loss, to upset the balance of the
comparisons across the different levels of the
hierarchy.
the perception level
the highest level
prediction error
the representative
higher level
prediction error
prediction error
(Audtiory N1 response)
w w w. t i n n i t u s h u b . c o m
20

Neural synchrony
Synchronous neural firing occurs when indivi-
dual neurons are driven to fire simultaneously
- not to be confused with central gain, where
the overall firing rate increases rather than how
the firing of different signals link together. Such
simultaneous firing leads to oscillatory brain
waves that can be recorded with EEG.
There is constant spontaneous activity in the
auditory system, as mentioned in the Sensory
Precision Integrative Model. Humans usually
have spontaneous activity when there are no
external sounds. External sounds suppress the
spontaneous activity in the auditory system,
which allows us to detect outside noises better
(Galazyuk et al, 2019). This spontaneous activi-
ty can be affected (e.g., increased) by physio-
logical state, certain drugs, or exposure to high
levels of noise.
One of the explanations behind the sponta-
neous activity not impeding our ability to per-
ceive silence is that the usual neuronal syn-
chrony is below the threshold at which a sound
becomes consciously perceived. An increase in
the synchronous spontaneous firing has been
previously attributed to tinnitus (Eggermont &
Tass, 2015).Changes in the specific types of
EEG waves have also been noticed in peop-
le with tinnitus. The slower brain waves called
theta and alpha, which occur over the temporal
lobe, are called tau rhythm. (Eggermont & Tass,
2015). Howe- ver, it is not clear whether the
reduction in tau rhythm is indicative of tinnitus
itself, because reduced tau rhythm and increa-
sed delta oscillations have also been related to
our reaction to tinnitus (such as distress from
its annoyance or loudness) (Weisz et al, 2005).

Conclusion on
Predictive coding models
The three described models, which are based on the principles of predictive coding, could po-
tentially explain the onset of tinnitus in people with a variety of characteristics, including those wit-
hout evident hearing loss in the case of Sensory Precision Integrative Model. Auditory attention
and perception-update models do require some form of hearing loss. Essentially, these theories
propose that if there are differences between what the brain is expecting to hear, and what you
actually hear, the brain may try to compensate for this - the models propose different mechanisms
of how you brain does this. If the difference is either large enough (in case of ear plugging, ear in-
fections, or other types of temporary disruption that does not necessarily lead to hearing loss), or
carries on for some time (e.g. hearing damage), it can instigate tinnitus to make up for the differing
auditory information that comes from the outer compared to the inner worlds.

w w w. t i n n i t u s h u b . c o m
05
21

ATTENTION
IN TINNITUS
The attention-switching model cannot explain tinnitus presence,
however it attempts to explain why tinnitus affects some people
more than others.

w w w. t i n n i t u s h u b . c o m
22
Attention in tinnitus
Ability to switch attention from one sound to
another may depend on 1) cognitive control
processes such as inhibitory control and wor-
king memory, and 2) on the ability to regulate
emotional response to certain stimuli. Atten-
tion switching does not necessarily need to be
consciously executed. Our brain does a lot of
these tasks without us intentionally controlling
its decisions.
In people with chronic tinnitus, the ability to
switch attention and cognitive control are im-
paired, which can be seen through behavioural
tests and evoked potentials such as P3a (Man-
narelli et al, 2017; Trevis et al, 2016). For exam-
ple, a slowing was seen in attentional switching
to an unexpected event (a sound that had an
unexpected pitch)) in participants with tinnitus
compared to those without tinnitus (Mannarelli
et al, 2017). This is potentially caused by cogni-
tive load (the number of tasks the brain has to
w w w. t i n n i t u s h u b . c o m
work on at once) in people with chronic tinnitus
being increased as the phantom sound uses
up needed resources, so it is harder to direct
attentional resources towards other attentional
tasks. This may be the reason behind the at-
tention switching impairments.
More severe tinnitus may be related to more
stress in a person’s life, which in turn may de-
crease the person’s ability to regulate emoti-
ons. Indeed, people who focus more atten-
tion internally (those who pay more attention
to their thoughts and the feelings inside their
body compared to others) are more stressed
as a result of tinnitus (Newman et al, 1997). As
such, differences and changes in the function
of attention are likely to come as the result of
tinnitus and may be involved in its maintenance
and the emotional response to it, rather than
its emergence (Newman et al, 1997; Sedley et
al, 2016).
 23

Final thoughts
We have presented a variety of tinnitus mo- As always in science, there remain a lot of
dels. Quite understandably, you will be left questions to be answered. Scientists around
with the question: So which one is true? Un- the world are working hard to discover more
fortunately, there is no one-size-fits-all model about tinnitus and new findings will likely point
of tinnitus. It is generally thought that there are us in new directions again. Hopefully, at a mi-
different subtypes of tinnitus. Therefore, the nimum, we have been able to provide you with
models presented here might very well all be more understanding of where your tinnitus
correct in their own way, just for different peo- might come from.
ple. Not everyone’s tinnitus is the same. Scien-
tists around the world are working very hard on
the “subtyping” problem, and hopefully that will
lead to novel insights into the neurophysiology
of tinnitus and inform tinnitus models research
further.

Other people might think after reading this: So

how does this help us find a cure for tinnitus?
The more we know about how tinnitus is ge-
nerated and perceived, the better equipped
we are to find the source. For example, the
idea that tinnitus is a problem of networks in
the brain opens up the opportunity for different
therapy options than the idea that tinnitus was
an ear problem. One therapy option that is cur-
rently being researched is called “non-invasive
brain stimulation” and it is thought it could dis-
rupt some of the pathological activity seen in
conditions like depression but also tinnitus.

w w w. t i n n i t u s h u b . c o m
06
24

APPENDIX

w w w. t i n n i t u s h u b . c o m

EEG signals
EEG records your brain activity using sensors
placed onto a special cap that you wear. The
cap is tight fitting and looks a bit like a swim-
ming cap. The cap holds the sensors on your
head. The sensors have to be covered with gel
to make sure they make a good connection.
There are two ways to look at EEG signals:
analysis through time and through frequency
domains. Time analysis gives you the changes
that occur during a specific period of time, whe-
reas frequency analysis allows you to see how
much of the brain signal is in particular frequen-
cies of brain waves.
There are a number of brain waves: delta (1-4
Hz), theta (4-7 Hz), alpha (8-14 Hz), beta (15-30)
and gamma (30+ Hz). Lower frequency theta
and alpha oscillations occurring over the tem-
poral lobe are called tau rhythm. (Eggermont &
Tass, 2015). These have been used in tinnitus
research, which will be mentioned below. The-
re are a number of event-related signals that
have been used to research tinnitus. These si-
gnals are time-locked to a specific stimulus in
an experiment.
25
N1 is a brain signal that can be recorded using
EEG, which is evoked at around 100 ms after
a sound change occurs. N1 detects change by
comparing information of a previous stimulus
with the next stimulus (Noda et al, 2018). Two
examples of N1 response are: 1) the ‘On-N1’,
which is seen when a new sound begins after
silence, and 2) the Off-N1, which can then be
seen when that particular sound stops. The big-
ger the change between the previous sound
characteristics and the new sound characteris-
tics, the larger N1 responses are.
Mismatch Negativity is a later brain signal that
occurs at around 150-250 ms after detection
of a change in the environmental sounds. This
signal is usually studied through oddball para-
digms (e.g., Sedley et al, 2019), where an irre-
gular stimulus occasionally appears amid iden-
tical repeating stimuli.
w w w. t i n n i t u s h u b . c o m
26

Contributors
Content creators
• Tori Kok - PhD student, University College London, Ear Institute, funded by
National Institute for Health Research, Biomedical Research Centre (NIHR BRC)
• Kate Yukhnovich - PhD student, Newcastle University, funded by RNID and the
Masonic Charitable Foundation

Patient reviews
• Hazel Goedhart - Director, Tinnitus Hub
• Markku Vesala - Director, Tinnitus Hub
• Linda Gendreau - Tinnitus Hub Patient Expert Panel
• Frédéric Tacussel - Tinnitus Hub Patient Expert Panel

Content validators
• Professor Raj Shekhawat, Flinders University
• Dr Will Sedley, Newcastle University

Graphic Design
• Tom E.

w w w. t i n n i t u s h u b . c o m
 27

BIBLIOGRAPHY
1 Auerbach, B. D., Rodrigues, P. V. & Salvi, R. J. (2014). Central gain control in
tinnitus and hyperacusis. Frontiers in neurology, 5, 206-206.

2 Eggermont, J.J. & Kral, A. (2016). Somatic memory and gain increase as
preconditions for tinnitus: Insights from congenital deafness. Hearing Re-
search, 333, 7-48.

3 Eggermont, J. J., & Tass, P. A. (2015). Maladaptive neural synchrony in

tinnitus: origin and restoration. Frontiers in neurology, 6, 29. https://doi.
org/10.3389/fneur.2015.00029

4 Galazyuk, A.V., Longenecker, R.J., Voytenko, S.V., Kristaponyte, I., Nelson,

G.L. (2019). Residual inhibition: from the putative mechanisms to potential
tinnitus treatment. Elsevier Hearing Research, 375, 1-13.

5 Hasselmo, M.E. & Sarter, M. (2010). Modes and models of forebrain choli-
nergic neuromodulation of cognition. Neuropsychophamacology, 36, 52-
73.

6 Hullfish, J., W. Sedley, & S. Vanneste. (2019). Prediction and perception: In-
sights for (and from) tinnitus. Neuroscience & Biobehavioral Reviews, 102,
1-12.

Mannarelli, D., Pauletti, C., Mancini, P., Fioretti, A., Greco, A., De Vincentiis,
7 M. & Fattapposta, F. (2017). Selective attentional impairment in chronic tin-
nitus: evidence from an event-related potentials study. Clinical Neurophy-
siology, 128(3), 411-417.

8 Newman, C.W., Wharton, J.A. & Jacobson, G.P. (1997). Self-focused and so-
matic attention in patients with tinnitus. Journal of the American Academy
of Audiology, 8, 143-149.

9 Noda, K., Kitahara, T. & Doi, K. (2018). Sound Change Integration Error: An
Explanatory Model of Tinnitus. Frontiers in Neuroscience, 12(831).

10 Rauschecker, J. P., May, E. S., Maudoux, A. & Ploner, M. (2015). Frontostria-

tal Gating of Tinnitus and Chronic Pain. Trends in Cognitive Sciences, 19,
567-578.

w w w. t i n n i t u s h u b . c o m
28

11 De Ridder, D., Vanneste, S. & Freeman, W. (2014). The Bayesian brain:

Phantom percepts resolve sensory uncertainty. Neuroscience & Biobeha-
vioral Reviews, 44, 4-15.

12 De Ridder, D., Vanneste, S., Langguth, B. & Llinas, R. 2015. Thalamocortical

Dysrhythmia: A Theoretical Update in Tinnitus. Frontiers in Neurology, 6,
124.

13 Roberts, L.E., Husain, F.T., Eggermont, J.J. (2013). Role of attention in the
generation and modulation of tinnitus. Neuroscience & Behavioural Re-
views, 37(8), 1754-1773.

14 Schaette, R. & Mcalpine, D. (2011). Tinnitus with a normal audiogram: phy-

siological evidence for hidden hearing loss and computational model. The
Journal of Neuroscience, 31, 13452-13457.

15 Schlee W, Hartmann T, Langguth B, Weisz N. (2009). Abnormal resting-sta-

te cortical coupling in chronic tinnitus. BMC Neuroscience, 19, 10-11.

16 Sedley, W. (2019). Tinnitus: Does Gain Explain? Neuroscience, 407, 213-

228.

Sedley, W., et al. (2019). Exposing pathological sensory predictions in tin-

17 nitus using auditory intensity deviant evoked responses. The Journal of
Neuroscience, 39(50), 1308-19.

18 Sedley, W., et al. (2016). An Integrative Tinnitus Model Based on Sensory

Precision. Trends in neurosciences, 39(12), 799-812.

19 Trevis, K.J., McLachlan, N.M. & Wilson, S.J. (2016). Cognitive Mechanisms
in Chronic Tinnitus: Psychological Markers of a Failure to Switch Attention.
Frontiers in Psychology, 7, 1262.

20 Vanneste, S., Song, J.J. & De Ridder, D. (2018). Thalamocortical dysrhyth-

mia detected by machine learning. Nature Communications, 9, 1103-1103.

21 Weisz ,N., Moratti, S., Meinzer, M., Dohrmann, K., & Elbert, T. (2005). Tin-
nitus perception and distress is related to abnormal spontaneous brain
activity as measured by MEG. PLoS Med, 2(6), 153.

w w w. t i n n i t u s h u b . c o m