The possibility that tympanic membrane temperature (TMT) may be a quick and inexpensive objective indicator of lateralized emotional processes offers the tantalizing possibility that TMT may ultimately prove useful in research, diagnosis, and treatment of disorders that involve emotional dysregulation.1,2 For example, individual differences in hemispheric lateralization of emotion have been shown to interact with hemisphere of application of rapid transcranial magnetic stimulation (rTMS) to influence the effectiveness of rTMS in the treatment of depression;3 to the extent TMT could be used to determine lateralization of negative emotion, response to rTMS might be optimized. At the micro level, neurotransmitter activity is lateralized.4 Given that psychotropic drugs are specialized to act on particular neurotransmitter systems, individual differences in hemispheric lateralization of emotion, measured via TMT, could be used to guide appropriate pharmacological treatment of mental disorders. In contrast with electroencephalographic (EEG) measures or imaging techniques, which are relatively difficult to use, require specialized knowledge, can be costly, and are time-consuming, TMT may provide simple, inexpensive information regarding hemispheric activity and emotional state, with minimal training.
The absolute mechanisms by which TMT may be related to hemispheric activity are not completely known. It has been argued that the relationship between TMT and lateralized hemispheric activity is dependent on whether TMT measures are compared within versus between subjects,2 with increased TMT being associated with 1) decreased ipsilateral hemispheric activity, in within-subjects designs; and 2) increased ipsilateral hemispheric activity, in between-subjects designs. Thus, an absolute difference between left and right TMT is indicative of relative hemispheric activation within an individual at a given point in time, although precisely which hemisphere is more active, given a left–right TMT difference, may vary.5–8
Studies measuring baseline TMT report a positive relationship between warmer left TMT and positive affect, and between warmer right TMT and negative affect in children, interpreting these findings as support for traditional left-hemisphere/approach, right-hemisphere/withdrawal models of lateralization of affective valence.1,9 That is, their results were interpreted to indicate that unilaterally increased TMT was associated with increased ipsilateral hemispheric activity.
However, work from our lab has reported no simple relationship between unilaterally increased or decreased TMT and positive or negative affect (e.g., happiness, sadness, calmness, or anxiety) at baseline.10 In addition to methodological differences between studies, the lack of relationship may stem, at least in part, from weak baseline differences in emotion. That is, it may be that in order to detect relationships between affect and TMT, emotional arousal must reach a particular threshold, which did not exist in our previous samples.
To our knowledge, no studies have attempted to induce multiple affects in order to detect effects on TMT. Given the suggestions that TMT-hemispheric activity-affect relationships may in part be modulated by the strength of the emotion being experienced, in the present study we used a mood induction protocol to induce emotions in participants in order to examine affect–TMT relationships.11
The present experiment sought to disentangle some of the factors that influence TMT–affect–hemisphere relationships. Here, we examined TMT at baseline and after a mood-induction protocol. Given our design, we were able to formulate several hypotheses. First, at baseline, we predict replication of previous findings of an association between increased absolute difference between the left and right TMT and increased anger.10,11 It is not clear whether we will replicate previous findings of increased left versus right TMT being associated with increased positive versus negative affect, respectively. Because mood-induction, by definition, is predicted to increase emotional affect relative to baseline, any TMT–affect relationships may be enhanced, relative to baseline, after mood-induction.
As part of a larger protocol, 74 right-handed undergraduates (as determined by an inventory12), with normal or corrected-to-normal hearing, participated for $20.00 remuneration and/or extra credit in a Psychology course. Four individuals were not included in analyses because of failure to follow instructions (N=2) or TMT measured incorrectly (N=2); final N=70 (40 women). All read and signed an informed-consent form, and the experimental protocol received IRB approval.
Immediately after reading and signing the consent form, left and right TMT were taken, using a Braun ThermoScan Pro 4000 (WelchAllyn) tympanic membrane thermometer, with probe covers (two per subject), while the participant was seated (TMT: Time 1). Participants then completed the Brief Mood Inventory Scale (BMIS)13 and an affect grid (to be reported elsewhere) and were randomly assigned to Mood-Induction Condition (MIC) or to a Neutral Condition (NC). There were four MICs: Calm, Anxious, Happy, and Sad. For each MIC (but not NC), participants were instructed to think of an MIC-congruent memory while listening to mood-congruent music for 5 minutes.14 Individuals in the NC were instructed to sit quietly and to think about an event from their past. No information regarding the emotional valence of the event about which they were to think was given in NC, and no music was played. Immediately after MIC (or NC), left and right TMT were again measured (TMT: Time 2), and participants again completed the BMIS (BMIS: Time 2) and an affect grid (to be reported elsewhere). After an additional task (to be reported elsewhere), participants viewed a funny video (Animal Planet’s Einstein, The Talking Parrot; see http://www.youtube.com/watch?v=7rfGEtALHYs), were debriefed, and were given $20.00 for their participation.
Three affect subscale measures of the BMIS were calculated for each Time (BMIS Time 1 versus BMIS Time 2). The three subscales were: Happy/Sad (happy, lively / sad, gloomy), Calm/Anxious (calm, content / nervous, jittery), and Warmth/Anger (loving, caring / grouchy, fed-up). Subscale measures for the three BMIS subscales at each time were computed via the formula: (sum of the two positive valence scores) minus (sum of the two negative valence scores). Thus, higher scores at each time indicate increased positive affect.
Four TMT measures were examined, including temperature at each tympanic membrane (Left versus Right), the difference between the Right and Left TMT (r–lTMT), and absolute difference between right and left TMT (ar–lTMT).
At Time 1, correlations between TMT measures and each of the three BMIS subscales were calculated. At Time 2, we also examined correlations between BMIS scores and the TMT-dependent measures as a function of NC and MIC, with the Happy and Sad MICs being examined on the Happy/Sad BMIS subscale, and the Calm and Anxious MICs being examined on the Calm/Anxious subscale. All three BMIS subscales and TMT measures were examined at Time 2 in the NC condition.
Changes in affect and TMT as a function of Time (Time 1 [T1] versus Time 2 [T2]) in the Neutral Condition (NC) were analyzed in order to examine effects of Time on either mood or TMT measures in the absence of any mood-induction intervention.
Examining only those individuals who experienced mood-induction–congruent changes in affect from T1 to T2 in the MIC conditions (calculated via T1 minus T2, with a negative number indicating increased positive affect at T2 and a positive number indicating increased negative affect at T2, on the appropriate BMIS subscale; see Table 1), we also conducted four 5 (MIC/NC: Happy, Sad, Calm, Anxious, and NC) × 2 (T1 versus T2) mixed ANOVAs, examining the TMT measures. Finally, because research suggests that positive versus negative affect, regardless of specific mood, might be related to TMT measures,1,9 we collapsed across MIC, such that individuals who experienced positive mood as a function of the mood-congruent MIC (individuals who became happy in the Happy MIC and who became calm in the Calm MIC; N=15) or negative mood (individuals who became sad in the Sad MIC and anxious in the Anxious MIC; N=18) were grouped. We then conducted four 3 (Mood: Positive, Negative, Neutral) × 2 (Time: T1 versus T2) mixed ANOVAS examining TMT measures.
TABLE 1.Demographic Information as a Function of NC and MIC Condition
| Add to My POL
|Total N||Experiencing Mood-Congruent Emotion|
|Condition||N||N Gender||Age, years||N||Gender||Age, years|
|Neutral||16||7 women||19.13 (0.29)||N/A||N/A||N/A|
|Calm||15||11 women||19.27 (0.33)||9||6 women||19.44 (0.44)|
|Anxious||11||8 women||19.27 (0.30)||5||4 women||19.14 (0.40)|
|Happy||13||5 women||20.15 (0.48)||7||4 women||20.43 (0.68)|
|Sad||15||9 women||19.60 (0.35)||13||7 women||19.43 (0.36)|
Time 1 Overall Affect and TMT Relationships
Overall, correlations between scores on the three BMIS subscales and all TMT measures did not reach significance for all comparisons (N=70). Because men and women may differ in their lateralization of cerebral functions, we examined these relationships as a function of gender.15 No relationship between the variables was found in women for all comparisons (N=40). However, in men (N=30), there was a positive correlation between Right TMT and Happy/Sad scores (r=0.40; p <0.05) and a negative correlation between ar-lTMT and Warmth/Anger (r = −0.43; p <0.05). No other TMT–Affect relationships reached significance.
Time 2 Affect and TMT Relationships as a Function of NC and MIC
Correlations between the three BMIS subscales and TMT measures in the NC condition (N=16) at T2 revealed a trend for a negative correlation between Left TMT and the Calm/Anxious subscale of the BMIS (N=16; r = −0.46; p=0.07), and a negative relationship between Right TMT and the Calm/Anxious subscale of the BMIS (r = −0.73; p <0.01). A negative correlation between T2 ar-lTMT and Warmth/Anger (r = −0.62; p <0.01) was also revealed. A negative relationship between T2 r-lTMT and Warmth/Anger (r = −0.58; p <0.05) also occurred. No other relationships obtained significance.
Correlations between the BMIS Happy/Sad scale and TMT measures in the Happy MIC (N=13) at T2 revealed a positive correlation between Happy/Sad score and ar-lTMT (r=0.61; p <0.05). Examination of correlations among only those individuals who became happy after mood-induction (negative scoring in the formula [Happy/Sad T1 minus Happy/Sad T2]; N=7), maintained this relationship, albeit without significance (r=0.64).
Correlations between the BMIS Happy/Sad scale and TMT measures in the Sad MIC (N=15) at T2 revealed no significant relationships. However, examination of only those individuals who became sad after mood-induction condition (positive scoring in the formula (Happy/Sad T1 minus Happy/Sad T2; N=13) revealed a positive correlation between Left TMT and Happy/Sad score (r=0.60; p <0.05). No other relationships were significant.
Correlations between the BMIS Calm/Anxious scale and TMT measures in the Calm MIC (N=15) at T2 revealed no significant relationships. Examination of only those individuals who became calmer (negative scoring in the formula [Calm/Anxious T1 minus Calm/Anxious T2]; N=9), supported this lack of significance.
Correlations between the BMIS Calm/Anxious scale and TMT measures in the Anxious MIC (N=11) at T2 revealed no significant relationships. Examination of only those individuals who became anxious (positive scoring in the formula [Calm/Anxious T1 minus Calm/Anxious T2; N=5]), supported the lack of significance, albeit with the small N of 5.
Note that, given the small number of subjects per group, analyses examining gender effects were not possible (see Table 1 for demographic information as a function of MIC, NC, and effectiveness of mood-induction procedure).
TMT-Dependent Measures as a Function of MIC/NC Group and Time
Only participants who demonstrated increased mood-induction–congruent feelings at Time 2 relative to Time 1 were included in analyses examining TMT-dependent measures. All NC participants were included in analyses.
A 5 (NC/MIC: Happy, Sad, Calm, Anxious, Neutral) × 2 (Time: T1 versus T2) mixed ANOVA examining Left TMT revealed no main effects or interactions.
The same analysis examining Right TMT revealed a main effect of Time, with increased Right TMT at T2 (F[1,4]=18.73; p <0.01) compared with T1, regardless of NC/MIC group.
Two 5 (NC/MIC: Happy, Sad, Calm, Anxious, Neutral) × 2 (Time: T1 versus T2) mixed ANOVAs, examining r–lTMT and ar–lTMT, revealed a marginal main effect of Time on r–lTMT, with increased r–lTMT at T2 relative to T1 (F[1,4]=3.34; p=0.07; see Table 2 for means and SE as a function of MIC and Time for each of the dependent variables.)
TABLE 2.TMT Variable (mean, standard error; °C) as a Function of MIC and Time
| Add to My POL
|Left TMT||Right TMT||r–lTMT||ar–lTMT|
|Condition||Time l||Time 2||Time l||Time 2||Time l||Time 2||Time l||Time 2|
|Neutral||36.36 (0.14)||36.38 (0.11)||36.42 (0.13)||36.62 (0.10)||0.06 (0.10)||0.24 (0.06)||0.25 (0.08)||0.27 (0.05)|
|Calm||36.20 (0.18)||36.33 (0.12)||36.24 (0.13)||36.53 (0.14)||0.03 (0.10)||0.20 (0.08)||0.23 (0.06)||0.24 (0.07)|
|Anxious||36.43 (0.20)||36.50 (0.23)||36.70 (0.22)||36.87 (0.17)||0.27 (0.12)||0.37 (0.16)||0.33 (0.07)||0.37 (0.16)|
|Happy||36.36 (0.16)||36.42 (0.10)||36.64 (0.18)||36.67 (0.19)||0.28 (0.09)||0.25 (0.10)||0.29 (0.08)||0.34 (0.03)|
|Sad||36.48 (0.12)||36.54 (0.11)||36.54 (0.14)||36.67 (0.09)||0.06 (0.06)||0.13 (0.05)||0.13 (0.05)||0.17 (0.03)|
Last, we collapsed across MIC, comparing individuals who experienced mood-induction condition-congruent positive (N=16, Happy, Calm) versus negative (N=18, Sad, Anxious) affect and the NC in four 3 (Mood: Positive, Negative, Neutral) × 2 (Time: T1 versus T2) mixed ANOVAS examining TMT measures.
Results mimicked those above, with a main effect of Time on Right TMT (increased Right T2 relative to T1 TMT (F[1,2]=23.02, p <0.01) and a main effect of Time on r–lTMT (F[1,2]=5.44; p <0.05), with r–lTMT increasing at T2 relative to T1.
Unlike previous studies that attempted to induce changes in hemispheric activity via sustained unilateral gaze in order to detect concurrent changes in TMT and emotion,8,11 or that examined resting affective state and TMT measures,1,9,10 we directly altered emotion in order to examine change in TMT measures in this case. Such a manipulation, by increasing affective arousal, should allow for detection of TMT–affect relationships.
First, we again replicate a relationship between increased absolute difference in activity between the left and right TMT and decreased warmth/increased anger, finding this relationship among men at baseline, and in the neutral condition (who experienced no mood-induction) at Time 2. These results support and strengthen TMT as an indicator of feelings of warmth/anger.
Second, at baseline, we found a positive relationship between right TMT and increased happiness/decreased sadness (among men), suggesting an association between increased TMT and decreased ipsilateral hemispheric activity at baseline. Given that left-hemisphere activity is typically associated with positive/approach emotional valences such as happiness, increased right TMT at baseline here should be cautiously taken to indicate decreased right-hemisphere activity. Thus, at baseline, we find a tentative relationship between increased unilateral TMT and decreased ipsilateral hemispheric activity. In the Time 2, post–mood-induction condition, individuals who became sad after the Sad MIC demonstrated a positive relationship between Left TMT and Happiness/Sadness, such that increased Left TMT was associated with increased happiness/decreased sadness. Similarly, individuals in the Neutral condition showed a negative relationship between Right TMT and the Calm/Anxious subscale, such that increased Right TMT was associated with increased anxiety.
Moreover, in two related studies, task related TMT effects were linked to ipsilateral hemispheric activation.6,7 These findings suggest that, after a unilateral task (e.g., mood-induction in the Sad MIC), increased TMT is associated with ipsilaterally increased hemispheric activity.
In any case, however, questions remain about the relation between TMT and baseline resting versus post-manipulation active states. The possibility that the relationship between increased TMT and increased hemispheric activation changes from contralateral, during resting, to ipsilateral, during active states, raises the possibility that there will be intermediate stages where TMT is uncorrelated with hemispheric activation as the relationship transitions from contralateral to ipsilateral. This possibility may account for some negative results of links between TMT and affect. Also, the question of whether these findings are specific to measures of affect (as opposed to cognition), or due to differences in measurement techniques here versus those in other studies (e.g., small sample sizes, examination of only strongly right-handed individuals here, use of single measures of TMT per ear here, possible interactions between experimenter and participant gender on TMT), deserves further investigation.
We would like to point out that there were no changes in TMT measures as a function of mood-induction condition here. This may be due to the small n per MIC who experienced mood-congruent changes in affect, although the lack of significance when collapsing across positive versus negative affect (increasing ns available for analyses) argues against this possibility. Another possibility is that stable individual differences in hemispheric activation are overlayed upon temporary increases in hemispheric activity because of the mood-induction procedure; thus, changes in left versus right hemispheric activity as measured via TMT could be obscured by individual differences. In light of the other findings presented here and in previous research; clearly, there is some relationship between TMT, hemispheric activity, and affect, although the absolute mechanisms of action and the particulars of these relationships need further investigation.
There are several unexpected findings that should be addressed. First, decreased left and right TMT were both associated with increased calmness/decreased anxiety. Given that these results were found in the NC condition, these findings suggest that bilaterally increased hemispheric activity is associated with decreased calm/increased anxiety—although this is only tentative. Second, a relationship between increased absolute difference between left and right TMT and increased happiness/decreased sadness in mood-congruent individuals at Time 2 in the Happy MIC was found. It is not clear what this relationship might indicate, particularly in the context of findings demonstrating an association between increased ar-lTMT and decreased warmth/increased anger.
The finding that, regardless of condition, right TMT increased from Time 1 to Time 2 warrants discussion. It is interesting to note that our control condition involved memory retrieval, in the absence of any mood-induction, whereas all the MIC conditions also required memory recall, as well. Given evidence that right hemisphere activity is associated with episodic recall,16 in conjunction with the hypothesis that post-task TMT is associated with ipsilaterally increased hemispheric activity, it may be that we inadvertently increased right-hemisphere activity (measured via increased right TMT post-task), associated with memory retrieval. Certainly future investigation should examine memory–TMT relationships.
Another possible explanation for the increase in RTMT from Time 1 to Time 2 involves a phenomenon called emotional hyperthermia, in which induced emotional states, particularly negative ones, leads to increases in bodily temperatures.17 Most of this research has been performed with non-humans and has looked only at whole-brain temperature; the current results suggest that emotional hyperthermia in humans may be more pronounced in the right hemisphere.
There are several weaknesses of the present work that limit the conclusions that can be drawn here, including small sample sizes, particularly among individuals experiencing mood-congruent MIC affect. Future work should increase the number of participants experiencing particular moods. Also, although we find gender effects in the baseline condition, small ns at Time 2 preclude analyzing gender effects. Future work should include gender as a measure.
Regardless of any of the above methodological issues, we again demonstrated a relationship between increased difference in activity between the cerebral hemisphere—regardless of the direction of that difference—and increased anger. Given that this is the third report wherein this relationship has been found, and given the conflicts in the literature concerning other TMT–affect–hemispheric activity relationships, TMT–anger effects may be the most consistent finding in the burgeoning field of TMT–affective valence investigations. The variables that mediate this relationship (for example, gender) deserve further investigation.
Data collection was performed at Merrimack College, North Andover, MA.
The authors thank William S. Helton for helpful comments on an earlier version of this manuscript.