Conceived and designed the experiments: AB JM VN DD. Performed the experiments: JM JJ VN CZ CB. Analyzed the data: JM. Wrote the paper: AB JM DD.
The authors have declared that no competing interests exist.
Tune deafness (TD) is a central auditory processing disorder characterized by the inability to discriminate pitch, reproduce melodies or to recognize deviations in melodic structure, in spite of normal hearing. The cause of the disorder is unknown. To identify a pathophysiological marker, we ascertained a group of severely affected TD patients using the Distorted Tunes Test, an ecologically valid task with a longstanding history, and used electrophysiological methods to characterize the brain's responses to correct and incorrect melodic sequences. As expected, we identified a neural correlate of patients' unawareness of melodic distortions: deviant notes modulated long-latency auditory evoked potentials and elicited a mismatch negativity in controls but not in affected subjects. However a robust P300 was elicited by deviant notes, suggesting that, as in blindsight, TD subjects process stimuli that they cannot consciously perceive. Given the high heritability of TD, these patients may make it possible to use genetic methods to study cellular and molecular mechanisms underlying conscious awareness.
The appreciation of music requires that the brain process and decode a complex stream of acoustic signals in order to extract and consciously perceive salient features such as pitch, harmony, and melody. If this is done successfully, the qualities of music that are consciously experienced can evoke a wide range of emotions, memories, or images
Tune deafness (TD) is one such disorder
The TD phenotype is assigned using the Distorted Tunes Test (DTT), a convenient, ecologically valid, and reliable instrument that has been in use for more than four decades
TD is a common disorder, affecting approximately 2% of the population
In an effort to identify a pathophysiological marker in TD, we ascertained a group of severely affected subjects. We then used electrophysiological methods (electroencephalography and event related potentials, EEG/ERP) to characterize the brain's responses to a series of familiar melodies that contained correct and deviant terminal notes. This paradigm, a modification of the DTT, allowed us to directly investigate the central, clinically relevant features of the disorder–TD subjects' inability to recognize melodic deviations (see
EEG/ERP methods are ideal for such studies because they offer excellent temporal resolution and provide quantitative information about both perception and higher order processing of acoustic information. These methods have been widely used for years to study music processing in the human brain
There are also distinct, well-described ERP components that are signatures of the brain's response to deviant (mismatched, “oddball” or otherwise distorted) auditory stimuli. These include the so-called mismatch negativity (MMN)
Since by definition, tune deaf subjects are unaware of deviations in melodic structure, we predicted that both the MMN and the P300 response to such deviations would be absent in these subjects. We have found that this is not the case. While our results provide a potential electrophysiological substrate for auditory unawareness, they also suggest that tune deaf subjects are processing musical abnormalities, but are doing so outside of conscious awareness.
EEG data were acquired while subjects heard different versions of familiar tunes that either contained a correct (standard) or an aberrant (deviant) note at the end of the melodic sequence (
(A) Bingo and (B) Happy Birthday are illustrated with correct versions on the bottom and incorrect versions at the top. Deviant terminal notes are indicated by arrows.
Characteristic long latency auditory evoked potentials (AEPs) containing clear P1, N1 and P2 components were elicited in response to standard tones in both TD and control subjects. These components displayed typical latencies and waveform morphology (
(A) Grand averaged ERP waveforms from the control group at electrode coordinates FCz and POz from correct notes (indicated in blue) and deviant notes (red). (B) ERPs from the same electrode coordinates and conditions from the TD group. For both groups, correct notes elicited a clear P2 component. Following deviant notes however, the P2 was strongly diminished for the control group but showed no significant attenuation for TD participants. A 2-way ANOVA revealed an interaction of Group x Note: F(1,16) = 5.6, p = .0309; and main effects of Note: F(1,16) = 7.99, p = .012; and Group: F(1,16) = 5.55, p = .0315. Planned comparisons indicated a significant difference between correct and deviant notes for controls (correct 2.62 µV, deviant 0.66 µV, t (16) = 3.72, p = .001) but not TD (correct 3.16 µV, deviant 2.99 µV, NS). In contrast, deviant notes elicited components in the P300 latency range for both control and TD participants. Two-way ANOVA showed a significant main effect of Note for P3a (F(1,16) = 13.59, p = .002), but no effect, of Group or Group x Note interaction. Planned comparisons revealed that the P3a was significantly greater in response to deviant than correct notes in both groups (controls, deviant 3.97 µV, correct 1.02 µV, p = .0069; TD, deviant 2.50 µV, correct 1.00 µV, p = .0149). Two-way ANOVA showed a significant main effect of note for P3b (Note: F(1,16) = 8.73, p = .009), an effect of group (F(1,16) = 10.61, p = .005) but no group x note interaction. Planned comparisons showed that the P3b was significantly greater in response to deviant than correct notes in both groups (controls, deviant 4.43 µV, correct 2.02 µV, p = .0293; TD, deviant 2.54 µV, correct 0.99 µV, p = .0192). (C) Topographic maps of the P2, P3a, and P3b ERP distributions from deviant and correct final notes and mean group latencies for each component. (D) Corresponding topographic maps from the TD group.
The later components of the AEP, including P2, may be modulated when a MMN is elicited at the same latencies by deviant auditory stimuli. Consistent with this, a MMN with characteristic morphology and a latency overlapping that of the P2 was evoked by the abnormal melodic sequences in controls, but not in TD subjects (
(A) Depicts the MMN waveform component (left) and topographic distribution (right) from the control group. (B) Depicts the comparable waveform and topographic map from TD participants. A 1-way ANOVA indicated a significant main effect of group (F(1,16) = 5.46, p = .0328).
In contrast, a typical P300 response was evoked by deviant notes in both controls and TD subjects. This included P3a and P3b components, both showing characteristic latencies and waveform morphology. The P3a was more robust in anterior channels; the waveform from channel FCz is illustrated in
Beyond their characteristic latencies and waveform morphologies, components of the P300 response have been specifically linked in human subjects with evoked oscillations in the delta band
Waveforms in blue (both ERP and evoked power) represent responses to correct notes, while waveforms in red represent responses to deviant notes. (A) Comparison for controls and (B) tune-deaf participants. Both groups produced a peak evoked delta response for deviant tunes with peak latencies approximating those of the P3b. Peak delta responses were markedly lower for correct notes. A 2-way ANOVA showed a significant main effect of note for evoked delta power (Note: F(1,16) = 12.33, p = .0029; controls, deviant 12.52 µV, correct 5.85 µV; TD, deviant 13.47 µV, correct 3.25 µV), but no group effect, and no group x note interaction. (C) Topographic maps of the evoked delta response (left) and the P3b component (right) for controls. (D) Comparable maps for TD participants. Note: In this figure, evoked delta waveforms and topographic maps have been scaled for display purposes so that they can be directly compared to the ERP data.
Tune deafness has been characterized in a number of ways. However, the pathophysiological mechanisms that underlie the defining feature of the disorder–subjects' inability to recognize distortions in melodic structure–have rarely been investigated in context. Here we have attempted to do so in a group of severely affected TD subjects. We used a modification of the Distorted Tunes Test, a clinically relevant, ecologically valid paradigm–that is, one that makes it possible to investigate, in context, the essential behavioral features of this disorder.
We used EEG/ERP to measure the brain's responses to correct and distorted melodies. (
The long-latency components of the AEP reflect the earliest cortical responses to auditory stimuli. Of these, the P1 and N1 components were evoked by deviant as well as standard notes–that is, they were unmodulated by melodic abnormalities–in controls and TD subjects. Significant group differences were instead related to the later, P2 component, which was selectively attenuated in response to deviant tones in controls. This was not unexpected. Nittono and coworkers, using a similar paradigm, reported an attenuation of the P2 component in response melodic abnormalities in normal individuals
Because the role of the P2 is less well understood than that of the other long-latency AEP components, it is not clear what its selective attenuation may denote in physiological terms. However, the P2 may be obscured when a MMN, typically generated within same time window, is present, and this is the difference between TD and controls that may be most critical.
The MMN, is a well-established marker of change detection, most commonly elicited by deviant auditory stimuli
Such a MMN, with a characteristic latency and morphology, was elicited by melodic abnormalities in controls (
In stark contrast, a robust P300 was evoked by melodic abnormalities in both TD subjects and normal controls. The P300 is another well-established signature of change detection; it is frequently differentiated from MMN in that while it can be generated automatically
Although the P300 response was originally described as single entity, it is now clear that it consists of two distinct components - P3a and b
Our results therefore represent an apparent paradox. The absence of the MMN provides the predicted marker for auditory unawareness that is the hallmark of tune deafness. But the unexpected presence of the P300 indicates that TD subjects are at the same time processing the very abnormalities that they cannot consciously perceive. A recent paper, demonstrating a dissociation between conscious perception and behavioral performance in this patient population
Such a phenomenon, sometimes referred to as knowledge without awareness, has been recognized for some time
In TD, the symptoms may be related to anatomical distinctions between the MMN and P300. Although both the MMN and P300 are indices of change detection, they are structurally dissociable: the neural assemblies that generate these waveforms are situated in different regions of the brain. The sources of the MMN are located for the most part in unimodal auditory areas of the superior temporal gyrus
These anatomical differences suggest a model that may account for knowledge without awareness in tune deafness. Since it is the MMN that distinguishes TD and controls, our results suggest that conscious perception in TD subjects is likely disrupted at the level of the unimodal auditory cortex. This supports a more general notion that neural computations in early auditory areas are necessary for determining whether deviant auditory information is consciously perceived. Previous studies have indeed suggested that feature-specific neurons within these auditory regions regulate the access to auditory awareness in a bottom up fashion, and that responses underlying the MMN itself may operate as a gateway to consciousness
At the same time, the robust generation of the P300 suggests that deviant stimuli are nonetheless being discriminated and selectively processed in higher order frontoparietal cortices, having bypassed earlier mechanisms that regulate conscious perception.
A plausible explanation for this might be that acoustic stimuli are reaching the cortical sources of the MMN and P300 via independent, parallel pathways. This in turn, suggests a pathophysiological mechanism for TD similar to the one that accounts for the symptoms of blindsight,. In blindsight, perceptual processes mediated by direct projections from thalamus to primary visual cortex are disturbed, while alternate pathways to association areas remain intact
Independent activation of the different cortical areas is clearly not the rule in the normal brain. The natural activation pattern may be consistent with a model
In light of this, TD may provide a novel opportunity to study conscious perception. Unlike lesion-based disorders such as blindsight it should be possible to study the process in a brain that is structurally intact, and in a patient population that is far larger. In addition, TD is a highly heritable condition, and investigations currently underway may identify the genetic variants that underlie this condition. The tune deaf population may thus constitute a group in which the problem of consciousness might be approached at the cellular and molecular level using the tools of genetic research.
Tune deaf subjects were ascertained by random screening of 1218 individuals in two metropolitan areas. Among those scoring in the lowest 10th percentile on the DTT, 20 subjects were identified who had normal hearing, were medically and neurologically normal and were free of other confounding factors. Of these, eight subjects consented to participate in this protocol; one subject was excluded due to the presence of excessive EEG artifact, so that seven TD subjects (4 females, 3 males, ages 18–33,
All participants were right-handed
The EEG paradigm employed a modification of the DTT
Participants were seated facing a 34 cm LCD monitor and were asked to fixate on a 1.5 cm cross at a 5 degree visual angle from 1.5 m and listen to a series of melodies. Melodies were presented at 90 Db through a single speaker, located 1.5 m in front of the participant. The interstimulus interval between melodies was 3s. Of the melodies prepared, 180 (87 correct, 93 containing deviant notes) were presented in random order, using Neuroscan STIM software
All electrophysiological signals were recorded using 9 mm sintered silver silver-chloride electrodes. EEG was recorded from a 60 channel electrode cap, conforming to the extended 10–20 electrode placement system
Individual EEG trials were visually inspected and those that contained artifacts or exceeded 100µV of EOG were excluded from the analysis. ERP averaging was time-locked to the onset of the final notes in the remaining trials. Waveform peak amplitudes and latencies were derived from a 1000 ms ERP with a 200 ms baseline interval. For the P1, N1, P2, N2, P3a, and P3b peak amplitudes and latencies were derived for both correct and deviant terminal notes. The P1 peak was operationally defined as the maximum positive amplitude between 30 and 110 ms at electrode FCz, the N2 as the maximum negative amplitude between 70 and 160 ms at FCz, the P2 as the maximum positive amplitude between 100 and 260 ms at FCz, the N2 as the maximum negative amplitude between 200 and 350 ms at electrode POz, the P3a as the maximum positive amplitude between 270ms and 500 ms at electrode FCz, and the P3b as the maximum positive amplitude between 350 and 500 ms at electrode POz. In order to obtain peak amplitudes and latencies for the MMN, ERP waveforms from deviant notes were subtracted from those of correct notes. The MMN peak was defined as the maximum negative amplitude between 150 and 250 ms at electrode FCz.
Separate 2-way repeated measures ANOVA were conducted in order to determine the effects of note type (correct or deviant) and group (control or TD) on the amplitudes and latencies of the P1, N1, P2, N2, P3a, and P3b. Due to potential covariation between experimental conditions, introduced by the repeated measures design, the Huynh and Feldt Epsilon correction was applied to each calculated F-statistic. All tests were held to a family-wise α of .05. Hypotheses specific mean comparisons were performed using paired t-tests, with Bonferroni corrections in order to maintain the specified experiment-wise type I error rate.
Induced delta power was calculated using the Event-Related Bandpower function of Neuroscan's Edit software
Supplementary Notes.
(0.07 MB DOC)
Correct Note. Comparison of ERP responses to correct notes from the control and tune-deaf groups. Electrophysiological findings are represented topographically on 3D head models: Left (control) and right (tune-deaf) panels also display the ERP waveform as recorded from electrode coordinate FCz as a frame of reference, timers, and color scales.
(5.83 MB MOV)
Deviant Note. Comparison of ERP responses to deviant notes from the control and tune-deaf groups. Electrophysiological findings are represented topographically on 3D head models: Left (control) and right (tune-deaf) panels also display the ERP waveform as recorded from electrode coordinate FCz as a frame of reference, timers, and color scales.
(6.24 MB MOV)
The authors would like to thank Drs. David Poeppel, Alex Martin, and Jonathan Fritz for critical and insightful reviews of this manuscript.