Using studies that include deviations in simple sound features, one might identify changes in MMN associated with tinnitus in two possible ways: the first is that the presence of tinnitus causes changes in MMN as a result of either the sound properties itself, altered auditory predictions, or secondary to distress or altered attention; and the second possibility is that MMN could reflect altered sound processing that might affect one’s propensity to tinnitus (e.g. determination of regularities in auditory signals). These possibilities are not mutually exclusive. A number of studies have been carried out to investigate alterations in deviance detection in chronic subjective tinnitus compared to controls, using different types of oddball paradigm (Tables 1 and 2). Studies that did not feature a control group were not included because tinnitus mechanisms and biomarkers were of particular interest, rather than correlates of tinnitus distress.

Table 1 Published MMN studies on tinnitus that used stimuli frequencies < 4 kHz that included a control group Table 2 Published MMN studies on tinnitus that used stimuli frequencies > 4 kHzthat included a control group

There are some confounds that have affected MMN research generally, as well as tinnitus research specifically, including age, hearing and sex differences as well as comorbid hyperacusis (Tables 1 and 2). There remains a need to disentangle the effects of each factor on MMN responses, in order to be able to interpret which of the observed changes truly relate to tinnitus directly, and what these changes signify about the mechanisms of tinnitus and/or the tinnitus status of the individual(s).

Types of MMN (oddball) Paradigm Used in Tinnitus Research

In the classical oddball paradigm, a standard stimulus is repeated, with an unexpected stimulus (deviating in a particular sound feature), randomly replacing up to approximately 20% of stimuli (Fig. 3) (e.g. [42].). Typically, a minimum number of consecutive standards are required prior to each deviant. Other types of oddball paradigm include the roving paradigm, in which there are two types of standard stimuli, and deviants are the pseudo-random transitions from one standard tone to the other (Fig. 4) (e.g. [43]). Another type of oddball paradigm is a multi-feature paradigm (Fig. 5) (e.g. [33]). Every second stimulus is a deviant of some kind (interspersed with standards), but each deviant only differs in one of the following characteristics: frequency (usually ± 10%), intensity (usually ± 10 dB), duration (25 ms rather than 75 ms), perceived location (800 us latency difference between left and right stimuli sources) or there may be a silent gap (7 ms silent gap in the middle of the 75 ms stimulus tone).

Fig. 3

Examples of classical oddball paradigms. a A classicalfrequency paradigm with tones at 1 kHz. The rare 1.5 kHz deviant tone is displayed as the 4.th stimulus. b Another type ofclassical oddball paradigm that has been used in tinnitus research, where one of the tones is matched to the tinnitus pitch of a participant. The standard tones are usually slightly different frequency to tinnitus (e.g. in this example they are set 0.5 kHz higher than the tinnitus pitch). These paradigms can also have a deviant in the opposite direction to the direction of the tinnitus pitch deviant (in this example, the tinnitus pitch is a downward deviant and the second deviant is an upward deviant)

Fig. 4

Example of a roving oddball paradigm. In this example, the frequency of the tones is the roving factor. The first tone at each frequency is the deviant, but this frequency becomes the new standard once it is repeated

Fig. 5

Example of a multi-feature oddball paradigm. (Reproduced from [44]) The size of the tone marker represents the intensity (typically, 2nd and 3rd tone being 3 and 6 dB lower than the 1st, respectively). Every second stimulus is a deviantin one factor, but has been preceded by numerous consecutive “standards” with respect to that factor

Frequency Deviants

So far in MMN and tinnitus literature, frequency has been the most explored sound feature, which seems logical based on the nature of tinnitus, and a similar focus in the wider MMN literature. Notably, the majority of these oddball studies, whether classical or multi-feature, focused on lower frequencies (commonly 1000 Hz). While this follows MMN research into other phenomena such as ageing, cognitive function and even Parkinson’s disease, this choice of frequency to study tinnitus is in other respects counter-intuitive, considering tinnitus pitch tends to be fairly high frequency, likely due to its high comorbidity with sensorineural hearing loss [29, 45,46,47,48]. Here, we separately consider MMN findings at frequencies remote from, and close to, tinnitus frequencies, using 4 kHz as an arbitrary initial cut-off, but also sub-categorising higher frequency stimulus studies based on whether stimuli were matched to participants’ specific tinnitus frequencies.

Frequencies Below 4 kHz

At a lower frequency (usually 1 or 2 kHz standard), three studies found a weaker MMN response amplitude to upward frequency deviants using a classical oddball paradigm [49,50,51], while two other studies found a similar response between participants with tinnitus and controls both in classical and multi-feature paradigms [42, 52]. The aforementioned multi-feature study also saw a reduced MMN amplitude in tinnitus in response to a downward frequency deviant [52]. However, another study found no differences in response to downward deviants [53]. Unfortunately, [53] did not report the exact frequencies used in their control condition, only mentioning that the standard tones were one octave below the edge frequency of each tinnitus participant, with three downward deviant conditions (1, 2 or 4% lower than standard frequency). The edge frequency is calculated by finding the largest difference in hearing threshold between one tested frequency and the next (e.g. 2000 Hz and 4000 Hz in standard Pure Tone Audiometry). Finally, there were also two multi-feature studies that did not differentiate between the directionality of the frequency deviant, who also found an overall weaker amplitude to these deviants [33, 54].

Some of the studies had limitations in terms of samples sizes and control matching, either for age or in hearing thresholds. Previous studies found reduced MMN amplitudes, associated with increasing age, in response to both upward and downward frequency deviants (as well as duration, but not intensity, deviants) [47, 55,56,57]. Unfortunately, the two studies that formally matched on both age and hearing did not differentiate between upward and downward frequency deviants; therefore, it is unclear whether the directionality of the deviant played a role in this set of findings. However, the only two studies that found similar MMN amplitudes in response to upward frequency deviants between the two groups were matched in hearing thresholds but had tinnitus participants who were significantly older than controls. Other factors that may have influenced those findings include the higher anxiety score in the tinnitus group, which has previously been related to larger MMN amplitudes, and the additional presence of distant higher frequencies within the paradigm [53, 58]. This may be an important factor for the MMN responses, in control participants in particular, as previous studies have shown different patterns, in an intensity oddball paradigm run on healthy participants, simply on account of blocks of a different frequency being present elsewhere in the experiment, even though the current and recently preceding stimuli were identical [35, 41, 59, 60]). These differences could occur through adaptive processes to a particular environment due to mechanisms such as frequency-specific adaptation [61, 62].

Frequencies Above 4 kHz

Six studies investigated frequency deviant effects at higher frequencies, three of which matched the stimulus frequencies to tinnitus pitch of their participants [63,64,65].

Tinnitus Pitch-Matched Stimuli

Two of the studies with tinnitus pitch-matched stimuli had age-matched control samples, with the tinnitus group having THI scores > 38 [63, 64]. Another study had age-matched control participants but did not report THI scores [65]. The latter study had the largest difference between the standard (500 Hz) and deviant (tinnitus match or 8 kHz) frequencies. The authors found that during a passive paradigm (with no active auditory task but with a distractor, i.e. a silent movie), there were no differences in MMN amplitude in response to an upward frequency deviant towards the tinnitus frequency between tinnitus and control groups [65]. In [64], tinnitus participants were recruited in a clinical setting, based on matching their tinnitus pitch to 8 kHz. Eight kilohertz was used as a downward deviant (D1) from the standard (8.5 kHz). There was also an upward deviant set at 9 kHz (D2). The control group did not show a significant MMN response to either of the deviants (though there was a trend towards a larger response for D1). The tinnitus group showed a significantly stronger MMN amplitude to D2 which was further away from their tinnitus frequency but not D1, which was the tinnitus pitch match [64]. The tested ear of the tinnitus group on average had some hearing loss (up to 60 dB at 8 kHz) while controls had normal hearing, which may have affected responses both to D1 and the standard. Additionally, the hearing thresholds in the tinnitus group could have been lower for 9 kHz than 8 kHz, thus leading to a combined upward intensity and upward frequency deviant. The second study that used matched tinnitus frequency as a deviant used a standard tone that was 100 Hz “different” from the matched tinnitus pitch [63]. They did not specify the direction in which the deviant tone changed, but this stimulus set-up was probably most like D1 in [64]. This study formally matched both age and hearing thresholds of their participant groups [63]. Here, MMN amplitudes were larger in the tinnitus group than the controls in response to the frequency change towards the tinnitus match, which was incongruent with the D1 finding in [64]. Differences between this and the other two studies might lie in better hearing matching or exclusion of participants with low pure-tone loudness discomfort level (LDL) scores [63]. LDLs have been associated with presence of hyperacusis, which remains a significant confounding factor in tinnitus research [37, 66,67,68]. However, [63] did not report the LDL cut-off values they used in their exclusion criteria so it is not possible to assess their reliability [66].

Non-Tinnitus Pitch-Matched Stimuli

One study used the audiometric lesion edge (LE) of each tinnitus participant as the tinnitus pitch match [53]. LE was defined as the point just below the frequency at which hearing ability began to deteriorate according to pure-tone audiometric testing. However, a recent study showed that LE is much lower than the tinnitus frequency in people with sensorineural hearing loss, and frequency of maximal hearing loss tended to be just below the tinnitus frequency regardless of the exact configuration of the hearing loss (steep/notched/gradual/inverted “U”) [69]. This homeostatic theory of tinnitus pitch arising from compensatory hyperactivity within the hearing loss area has been supported by other research over the tonotopic model (which posits that the activity occurs in the tonotopic area just below the hearing loss region) (e.g. [45, 46, 70]). Regarding the [53] experiment, the LE frequency was used as a standard tone along with a control standard tone which was one octave below LE. Deviants were 1, 2 or 4% below either of the standard conditions. There was a deviant type dependency in the LE condition, where the 1% deviant elicited a stronger response in the tinnitus group compared to controls while the 2% deviant elicited a weaker response. Furthermore, there was an inverse relationship between tinnitus distress and the 2% deviant MMN amplitudes. Notably, as there was only one other study that looked into downward frequency deviants, which used a multi-feature paradigm rather than the classical frequency oddball paradigm, there could be a plethora of reasons for the discrepancy including the differences in sample-matching limitations as well as paradigm contexts (including both presence of other deviant types and far-away frequencies).

Two further studies explored MMN response amplitudes to frequency deviants but did not make an explicit pitch-matching attempt. Both of these used multi-feature paradigms but different frequencies, and one included fewer deviant types than the other. One study, which also included lower frequencies, used frequencies around 5 kHz and found a reduction in MMN amplitudes in response to both downward and upward frequency deviants in tinnitus [52]. This, however, was the study that had significant differences in age between the two samples, which may have led to the reduction of MMN amplitudes independently of tinnitus (e.g. [47, 57]). Another study investigated potential effects of high tinnitus distress on MMN amplitudes compared to a group with low tinnitus distress and a control group, all of which were hearing and age-matched [71]. The frequencies used in this study were around 8 kHz. All groups showed similar MMN amplitudes in response to downward frequency deviants. In response to the upward frequency deviants, the low tinnitus distress group had similar MMN amplitudes to controls. However, the high distress tinnitus group had significantly weaker MMN than either of the other two groups. While there is a possibility that the lack of tinnitus pitch-matching also affected this result, it is interesting to compare it with [63], which saw larger MMN amplitudes in the tinnitus group than the controls in response to a deviant tone that was 100 Hz closer in frequency to the tinnitus pitch-match than the standard tones. Both studies matched participants in age and hearing, but [63] excluded participants with suspected hyperacusis by excluding those with low LDLs. LDL scores do sometimes have low sensitivity however, so these tests may miss some cases of the condition [72]. Neither of the tinnitus distress groups in [63] had a similar alteration in response pattern to [71] compared to controls, and the most clear differences between the two comparisons were the paradigm and the explicit lack of hyperacusis (as [71] did not use a hyperacusis measure). However, there is also the possibility that the smaller control group sample size in [63] led to an unrepresentative result.

Frequency Deviants: Conclusion

Overall, while a number of studies have explored the effect of tinnitus presence on response to frequency deviation (particularly in lower frequencies), there are still many gaps in the literature. For example, there is a notable lack of downward frequency deviant studies in lower frequencies using classical oddball paradigms, and a lack of differentiation between deviant directionality in multi-feature studies. Furthermore, out of 11 published papers that included a control group and investigated MMN responses to frequency deviants, six did not fully match the groups on hearing and age, and only one explicitly measured and excluded hyperacusis. While the multi-feature paradigm studies have generally been more successful in group matching, these studies did not attempt to look at responses to deviants near the tinnitus pitch (though two authors did use frequencies that possibly were close to tinnitus frequencies [52, 71]). Uniting the positive points from the different studies, such as controlling for hyperacusis, fully matching groups and looking at frequencies both near and away from tinnitus, into one research project may help to clarify discrepancies in the findings so far.

However, allowing for these limitations and caveats, the prevalent finding from most studies has been that MMN response amplitudes are smaller to frequency deviants where the deviant and standard frequencies are distant from the tinnitus frequency.

Intensity Deviants

Seven studies investigated effects of tinnitus presence on intensity deviant responses. Two of these, both multi-feature paradigm studies, focused only on intensity changes using lower frequency tones distant from the tinnitus frequency [33, 54]. One multi-feature paradigm study utilised both lower and higher frequency tones but did not match the higher tone to tinnitus frequency [52]. Four other studies used a roving oddball paradigm, where a louder tone was pseudo-randomly alternated with a quieter tone and vice versa [35, 43]. Two of these, however, are at the pre-print stage of publication [60, 73]. The quieter tones in these experiments were set at − 6 dB relative to the louder tones.

Multi-Feature Paradigm Studies

A study in which the tinnitus group was significantly older on average than the control group showed that there was a reduction in MMN amplitude to both upward and downward intensity changes both when tones played were around 1 kHz and 5 kHz [52]. These results were supported by [54], which did not differentiate between the two deviant directions but found that overall, tinnitus was associated with reduced MMN amplitude to intensity deviants to tones around 1 kHz. These authors also accounted for cognitive function, including only participants who scored within normal range on the Montreal Cognitive Assessment Test [54]. This is important as some previous research suggested that decreased MMN responses to both upward and downward intensity deviants are related to impaired working memory (e.g. [48, 74]), and decreased MMN responses to upward intensity deviants were related to higher age [75], though others found no differences between younger and older participants (with and without depression) or older healthy participants compared to participants of similar age but with mild cognitive impairment [76, 77]. However, another tinnitus study with a similar design to [54] found no differences in MMN amplitudes to intensity deviants in the tinnitus group compared to controls, despite also having hearing and age matched groups [33]. There is a possibility that these differences could be due to varying levels of hyperacusis presence among the different tinnitus samples within these multi-feature studies, because [60] showed that at 1 kHz tones, a group of participants with tinnitus and hyperacusis (T + H +) had weaker responses to intensity deviants than controls, but a group with tinnitus but without hyperacusis (T + H-) showed no such difference.

Roving Paradigm Studies Tinnitus-Like Frequencies Only

In one study [7], two near-tinnitus frequencies were used (centre frequency of tinnitus of each participant and edge frequency of tinnitus, which was just noticeably below the tinnitus range of each participant) [43]. Participants with tinnitus had larger MMN responses to upward deviants, but smaller MMN responses to downward deviants, compared to the control group [43]. During this experiment, and in keeping with most previous studies, a passive task was utilised in which each participant chose a subtitled, silent film to watch during the EEG recording. Another study contrasted results obtained during this passive task with those during an active auditory attention task, and an active visual attention [73]. This study only used the edge frequency of the tinnitus, measured using the method from [43]. The MMN response patterns, both in the tinnitus and the control groups, recorded during the passive task were consistent with [