Enhanced audiovisual associative pair learning in migraine without aura in adult patients: An unexpected finding

Abstract

Background

Altered sensory processing in migraine has been demonstrated by several studies in unimodal, and especially visual, tasks. While there is some limited evidence hinting at potential alterations in multisensory processing among migraine sufferers, this aspect remains relatively unexplored. This study investigated the interictal cognitive performance of migraine patients without aura compared to matched controls, focusing on associative learning, recall, and transfer abilities through the Sound-Face Test, an audiovisual test based on the principles of the Rutgers Acquired Equivalence Test.

Materials and Methods

The performance of 42 volunteering migraine patients was compared to the data of 42 matched controls, selected from a database of healthy volunteers who had taken the test earlier. The study aimed to compare the groups’ performance in learning, recall, and the ability to transfer learned associations.

Results

Migraine patients demonstrated significantly superior associative learning as compared to controls, requiring fewer trials, and making fewer errors during the acquisition phase. However, no significant differences were observed in retrieval error ratios, generalization error ratios, or reaction times between migraine patients and controls in later stages of the test.

Conclusion

The results of our study support those of previous investigations, which concluded that multisensory processing exhibits a unique pattern in migraine. The specific finding that associative audiovisual pair learning is more effective in adult migraine patients than in matched controls is unexpected. If the phenomenon is not an artifact, it may be assumed to be a combined result of the hypersensitivity present in migraine and the sensory threshold-lowering effect of multisensory integration.

Introduction

It is generally accepted in the literature that altered sensory processing is a characteristic cognitive feature in migraine, whether during migraine attacks or in between them (1). The alteration in visual processing, in particular, has been extensively documented. Our research group has shown modifications in visual processing between migraine attacks, in both adult and pediatric patients. These alterations span from basic visual functions like contrast sensitivity to more intricate tasks such as visually guided associative learning (2–6). In our investigations concerning visually guided associative learning, retrieval, and transfer, we have utilized a modified version of the Rutgers Acquired Equivalence Test (RAET), often referred to as the ‘face-fish test’ after the visual stimuli employed (7). In this test, subjects are initially taught pairs of visual stimuli, where cartoon faces serve as antecedents and colored fish as consequents, through trial-and-error learning. They receive feedback on the accuracy of their guesses during this phase. Subsequently, when assessing the subjects’ ability to retrieve the learned information and generalize it by applying the pairing rule to previously unlearned stimulus pairs, no feedback is provided. Thus, the paradigm involves three different cognitive tasks: learning, retrieval, and transfer. Our research has revealed distinct alterations in various phases of this test among both adult and pediatric migraine patients (5,6).

RAET is an acquired equivalence paradigm. Acquired equivalence (AE) is a form of learning where generalization is increased between two superficially dissimilar stimuli that have previously been associated with similar outcomes. In other words, the subject learns that two or more stimuli are equivalent in terms of being mapped onto the same outcomes or responses (8).

The neural correlates of AE are rather well documented. Initially, it was demonstrated that patients with damage to the basal ganglia fail to learn the stimulus pairs, while patients with damage to the hippocampi fail to generalize (7). Later, Shohamy and Wagner provided an elaborate interpretation of AE as integrative encoding (9), in which the substantia nigra-striatum loop (SN-S) and the ventral tegmental area-hippocampus loop (VTA-H) play the central role. These structures process not only visual but audiovisual multisensory information (10,11), and are frequently highlighted in the literature as being affected by migraine (12–17).

Following the principles of the RAET, we have also developed an audiovisual counterpart known as the sound-face test (SFT). Instead of cartoon faces, this test employs sounds as antecedent stimuli and cartoon faces as consequents, thereby creating audiovisual stimulus pairs. We have applied this test to diverse groups, including adults and children, both healthy and afflicted by various conditions (18,19).

Importantly, in a recent study conducted on healthy children and adolescents (aged 5 to 17 years) with RAET and SFT, we demonstrated significantly superior multisensory associative pair learning in migraine-free children and adolescents as compared to purely visual pair learning (20). To put it simply, in migraine-free children and adolescents, audiovisual pair learning appears to be more efficient than simple, unimodal visual pair learning. However, in a more recent study, we found that this advantage is absent in pediatric migraine (21). The next logical question was whether this deficit can also be observed in adult migraine sufferers, or if it is somehow compensated for as the central nervous system matures. This study sought to answer this question.

Material and methods

Certain elements of this section may be similar to our earlier publications (5,18–20,22,23), as we consistently employ the same cognitive test paradigm, across various populations, and under the same testing conditions. This approach allows for a reliable comparison of data among different populations.

Participants

The patients were recruited from the Department of Neurology, Faculty of Medicine, University of Szeged, Hungary on a voluntary basis. Inclusion criterion was a diagnosis of migraine without aura as determined by the same neurologist according to ICHD-3 (24). Patients with other neurological, psychiatric, ophthalmological, or otologic disorder were not eligible for this study and a negative history was verified from the patients’ files. A further exclusion criterion was color vision deficiency, which was tested using Ishihara plates before testing. For all patients, at least five days had passed since the last attack at the time of testing, and they were not taking any anti-migraine medication or other drugs that could potentially influence central nervous system function.

The sample size was determined by the rigorous application of the diagnostic and inclusion/exclusion criteria. In total, 61 newly diagnosed patients with migraine were approached during the study period. Of them, 42 patients met the participation criteria and were willing to volunteer for the study. These 42 patients all finished the study per protocol and their data were used for the analyses.

The study was carried out according to a one case-one control design. Controls were matched from a control database that we have been developing since the first application of the audiovisual test variant (18), specifically for later comparisons. This database contains data from healthy adult volunteers of both sexes and all ages, who at the time of testing were free of any neurological, psychiatric, ophthalmological, or otological disorder or history thereof, and did not take any medication that could potentially interfere with central nervous system function. This database is continuously updated, with new volunteers typically recruited directly through the research group's network of contacts or from among the students at the University of Szeged. The matching was done according to age and level of education. Regarding age matching, we strived to ensure that the age of the controls exactly matched that of the patients. The tolerance limit was ± 2 years. Sex-matching was deemed unnecessary, as we previously demonstrated in a larger sample that performance in the RAET paradigm is not influenced by sex (22), and there was no reason to believe that the audiovisual nature of the task would introduce sex-dependent variations. To avoid matching bias, the selection of controls was done by a laboratory staff member who was not familiar with the measurement results. They received a list indicating the number, gender, and age of patients for whom controls are needed, along with the age tolerance range. This colleague selected controls from the control database according to the specified parameters, copied these with the results into an Excel workbook, and sent this to the person conducting the statistical analysis. Therefore, the person selecting the controls had no information about the exact data to which the control data would be compared.

The study protocol conformed to the ethical principles of the Declaration of Helsinki in all aspects. Before testing, the volunteers were informed about the background, aims, and procedures of the study both orally and in written form. None of the subjects received any compensation for their involvement, and they were informed that the study had merely scientific purposes without direct diagnostic or therapeutic use, and they were free to quit at any time. All volunteers verified their voluntary participation by signing an informed consent form. The study protocol was approved by the Regional Research Ethics Committee for Medical Research at the University of Szeged, Hungary (Reg. No. 27/2020-SZTE).

The testing procedure and the applied audiovisual test

The test was administered in a quiet room during the morning hours by one of the seven members of our team who were trained to conduct this test. The tests were executed on a personal computer, with a cathode-ray tube (CRT) screen (refresh rate: 100 Hz). The participants were sitting at a standard distance (114 cm) from the computer screen. The auditory stimuli were administered through Sennheiser HD439 over-ear headphones (Sennheiser, Germany) at SPL = 60 dB, to both ears at the same time. On the keyboard, the M and the X keys were labeled as ‘right’ and ‘left’, respectively. The participants were tested individually, one after the other. There was no time limit or forced responses. There was no time constraint for the responses to avoid performance anxiety. Participants in both groups were tested under the same conditions.

The applied audiovisual test, the Sound-Face test (SFT), was developed in our laboratory (18) and is based on the unimodal visual Rutgers Acquired Equivalence Test, an associative learning paradigm developed by Myers et al.(7) Initially, the task of the subject in this test is to make associations between antecedent and consequent stimuli through trial-and-error learning based on feedback. Later, the recall of these associations is tested without feedback, along with a hitherto unseen association. This way, the test assesses associative learning, recall and transfer.

In general terms, the test paradigm consists of two phases: the acquisition and the test phases. The test phase is further broken down into two parts: retrieval and generalization/transfer.

In the acquisition phase, the subject learns the association between the antecedents and the consequents through trial-and-error learning. The computer provides immediate visual feedback on the correctness of the subject's guess. The subject must achieve a certain number of consecutive correct answers after the presentation of each new association to be allowed to proceed. This number is four when the first association is presented and is increased by two upon the presentation of each new association that follows (up to a maximum of 12). Thus, the length of the acquisition phase varies among subjects, depending upon how efficiently they learn. There are altogether four antecedents and four consequents. Of the eight possible associations, six are taught in the acquisition phase. During the acquisition phase, the subjects also learn implicitly that certain antecedents are equivalent in terms of their relation to the consequents. Such information is crucial for the correct identification of the new associations in the transfer part of the test phase.

In the test phase, no further feedback is given about the correctness of the responses (guesses), and the subject must recall the already acquired six associations. This is the retrieval part of the test phase. Then two new, hitherto unknown associations are presented, which are derivable from the six known associations. The task of the subject is to correctly identify these new associations based on previous knowledge about the antecedents. This is the generalization/transfer part of the test phase. The subjects are not informed that new associations are introduced during the test phase, only that their task is the same, but without feedback. The test phase consists invariably of 48 trials (36 recall trials and 12 transfer trials in a random sequence).

In SFT, the version of the paradigm applied in this study, the antecedents are sounds (a cat's meow, the sound of an engine starting, a guitar chord, and a woman saying a word - A1, A2, B1, B2) and the consequents are visual stimuli (four cartoon faces- X1, X2, Y1, Y2). In each trial, the subject simultaneously hears one of the antecedents, and sees two different consequents on the right and left sides of the screen. The task is always to choose the correct consequent by pressing the ‘left’ or the ‘right’ button. In the acquisition phase, the computer provides immediate visual feedback on the correctness of the guess: a green checkmark for a correct guess and a red X for an incorrect guess (Figure 1). The duration of the antecedent sound stimulus is always 1500 ms, and the visual consequents remain on the screen until the volunteer responds.

Figure 1.

A summary of the audiovisual test. In the acquisition phase, the task of the subject is to associatively learn audiovisual stimulus pairs by trial-and-error learning supported with feedback. In the test phase, the retrieval of the previously learned stimulus pairs is tested (retrieval) and new, hitherto unknown associations are also presented, which are derivable from the previously learned (known) associations (generalization). No feedback is given in the test phase.

Data analysis

In this study, we compared the performance of patients and controls based on four key parameters: the number of trials required to learn associations during the acquisition phase (NAT), the learning error ratios in the acquisition phase (ALER), the retrieval error ratios in the test phase (RER), and the generalization error ratios (GER). Error ratios were calculated by dividing the number of incorrect answers by the total number of trials in each respective phase. Additionally, reaction times (RT) were measured with millisecond accuracy during the acquisition, retrieval, and generalization phases. RT was defined as the time elapsed from the appearance of stimuli to the participants’ responses. Only the correct answer's RTs were analyzed. Any values exceeding three standard deviations (SD) from the mean were excluded from the analysis.

Statistical analysis was conducted using Statistica 14.0.0.15 (TIBCO Software Inc., USA). Due to the non-normal distribution of the data (Shapiro-Wilk test), hypothesis testing was performed using the Mann–Whitney U test. Effect sizes (d) and post-hoc power calculations were computed for significant differences using G*Power 3.1.9.4 (Universität Düsseldorf, Germany). Graphs were generated in Jamovi (version 2.3.21).

Results

All volunteers from both groups managed to complete the test. We present data from 42 patients (comprising seven males and 35 females, with a mean age of 34.4 ± 10.8 and a range of 18–55 years) and 42 healthy controls (12 males and 30 females, with a mean age of 35.1 ± 11.3 years and a range of 19–54 years). In both groups, 40.5% of the volunteers (N = 17) held higher education certificates, while 59.5% (N = 25) had university degrees. Patients had been experiencing attacks for a median of 10 years, with a range of 2–40 years. The estimated median number of attacks was 300, with a range of eight to 3000 attacks. The estimated median attack frequency was two attacks/month, with a range of 0.2–12 attacks/month. Please note that these values are estimates based on patient-reported data.

As part of our preliminary analysis to assess patient group homogeneity, we divided the participants into two subgroups using the median attack number of 300 as the cutoff point. To conduct this analysis, we employed the Mann-Whitney U test, comparing the four parameters previously mentioned (NAT, ALER, RER, and GER). The results showed no significant difference (p > 0.05) between the two subgroups, indicating that our patient group is indeed homogenous. Furthermore, this suggests that the average number of attacks does not exert a significant impact on learning performance.

Performance in the acquisition phase

The patients needed significantly fewer trials to learn the associations than the matched healthy controls (U = 566.5, p = 0.048, d = 0.57, 1-β= 0.81). The median NAT was 46 (Q1: 44 – Q3: 52) in the patient group and it was 53.5 (Q1: 46 – Q3: 65) in the control group. This is shown in Figure 2.

Figure 2.

Boxplot comparison of the number of trials needed for the completion of the acquisition phase (NAT). The upper and lower margin of the boxes indicate the upper and lower quartile, respectively. The line within the boxes marks the median. The upper whiskers indicate the 90th percentile, and the lower the 10th percentiles. The dots represent individual data points.

The acquisition error ratio (ALER) was also significantly lower in the patient group (U = 585, p = 0.008, d = 0.59, 1-β = 0.84). The median ALER was 0.023 (Q1: 0.000 – Q3: 0.057) in the patient group and it was 0.051 (Q1: 0.023 – Q3: 0.091) in the control group. This is shown in Figure 3.

Figure 3.

Boxplot comparison of the ratios of erroneous responses in the acquisition phase (ALER).

Finally, reaction times in the acquisition phase were significantly different between the two groups (U = 660, p = 0.048, d = 0.53, 1-β = 0.65). The median acquisition RT was 1369.110 ms (Q1: 1197.145 ms – Q3: 1501.463 ms) in the patient group and 1567.073 ms (Q1: 1220.881 ms – Q3: 1846.152 ms) in the control group. This is shown in Figure 4.

Figure 4.

Boxplot comparison of the reaction times in the acquisition phase.

Performance in the test phase

There was no significant difference between the RERs of the two groups (U = 730, p = 0.175). The median RER was 0.000 (Q1: 0.000 – Q3: 0.028) in the patient as well as in the control group (Q1: 0.000 – Q3: 0.028). There was no significant difference in retrieval RTs either (U = 684, p = 0.077). The median retrieval RT was 1253.937 ms (Q1: 1112.486 ms – Q3: 1464.361) in the patient group and 1390.177 ms (Q1: 1198.000 ms – Q3: 1760.000 ms) in the control group.

The comparison of GERs brought similarly insignificant results. There was no statistically significant difference between the patients and controls (U = 739, p = 0.202). The median GER was 0.000 (Q1: 0.000 – Q3: 0.083) in the patient group and 0.000 (Q1: 0.000 – Q3: 0.167) in the control group. Generalization RTs did not differ significantly either between of the two groups (U = 688, p = 0.083). The median generalization RT was 1580.375 ms (Q1: 1352.583 ms – Q3: 1927.917) in the patient group and 1791.761 ms (Q1: 1324.417 ms – Q3: 2394.300 ms) in the control group.

Table 1 summarizes the descriptive statistics for all parameters in both groups.

Table 1.

Descriptive statistics of the studied parameters. RT (A, R, G): reaction times for acquisition, retrieval and generalization phases, respectively. Raw data are provided in a supplementary table

	Group	N	Median	Minimum	Maximum
NAT	control	42	53.50	41	93.00
	migraine	42	46.00	41	86.00
ALER	control	42	0.051	0	0.22
	migraine	42	0.023	0	0.15
RT(A)	control	42	1567.07	855.63	3538.90
	migraine	42	1369.11	957.23	1952.95
RER	control	42	0.00	0	0.25
	migraine	42	0.00	0	0.25
RT(R)	control	42	1390.18	874.35	2692.63
	migraine	42	1253.94	923.94	1896.19
GER	control	42	0.00	0	0.67
	migraine	42	0.00	0	0.50
RT(G)	control	42	1791.76	936.75	4803.14
	migraine	42	1580.38	997.50	2742.80

Discussion

In our previous research, we established that adult migraine patients significantly underperform compared to healthy controls in the purely visual RAET test, regarding both acquisition and transfer (6). In this study, we observed that migraine patients exhibited a significantly enhanced acquisition of audiovisual stimulus pairs compared to the control group of healthy individuals, without any notable difference in the transfer of learned information. This distinction suggests that when employing audiovisual stimuli, the performance of migraine patients diverges notably from the pattern observed with purely visual stimuli. Specifically, the process of acquiring stimulus pairs among migraine patients can be characterized as supernormal, indicating a superior performance relative to healthy controls. This is indeed a surprising and unexpected result. It is generally accepted in the literature that altered multisensory processing is a characteristic feature of migraine (25–29), but, to our knowledge, no study before has reported enhanced multisensory performance in migraine patients. Furthermore, improved associative learning performance was accompanied by significantly shorter reaction times, which is exactly what one expects in the context of multisensory facilitation (30). Finally, the effect sizes are not negligible, and the comparisons were not underpowered, so it is not likely that the effect is merely statistical.

While these observations appear to support the authenticity of the effect, we initially harbored concerns that it could be an artifact, potentially due to what we suspected might be anomalously low performance in the control group. These concerns were prompted by an analysis of results from a prior study we conducted with a healthy adult cohort (N = 55, mean age: 31.36 years) using SFT (31). In that study, the performance of the subjects closely paralleled that of the patient group of the current study. Conversely, in our initial study validating the SFT with 141 healthy participants (mean age: 31.21 years), the performance was more aligned with that of the control group in the present investigation (18). This discrepancy necessitated an evaluation to identify which control group's performance was aberrant. To this end, we undertook hypothesis testing on a consolidated dataset—comprising results from the study populations in both the preceding SFT studies, alongside the control group in the current study—provided as Online Supplementary Material. For the hypothesis testing, the Kruskal-Wallis test was used, with a significance limit reduced to p = 0.017 because of the multiple comparisons. The test returned significant results for NAT and ALER, which prompted us to perform pairwise comparisons. The pairwise comparisons clearly indicated significant difference in both parameters between the performance of the study populations of the studies by Tót et al. (31) and Eördegh et al. (18). At the same time, there was no significant difference in the said parameters between the study population of Eördegh et al. (18) and the control group of the current study. This, especially considering that the study by Eördegh and colleagues was conducted on a larger sample, suggests that, contrary to our initial suspicion, it was not the control group of the current study whose performance was anomalously low. Instead, it was the study population of the study by Tót and co-workers that exhibited supernormal performance. This underscores the validity of our study's findings regarding the patient population's enhanced performance.

We know that in the purely visual version of the RAET paradigm, migraine patients perform significantly worse than healthy controls, both in terms of acquisition and generalization (6). The results of the present study show that when we use the audiovisual paradigm, migraine patients perform significantly better than healthy controls in terms of acquisition, and perform at the same level, nearly flawlessly, in terms of transfer. We also know that the control group in the present study did not perform significantly worse than healthy controls previously tested with the same audiovisual test. From this, it follows that the performance of migraine patients in the audiovisual version of the paradigm is supernormal in terms of acquisition and normalizes in terms of transfer. In other words, the application of audiovisual stimuli appears to have exerted a generally enhancing effect on the migraine group. How could this be explained?

It must be emphasized that at our current level of knowledge, we can only provide speculative answers, or more accurately, we can propose hypotheses. This is primarily because the literature on the effects of migraine is quite scarce, essentially limited to our own published research, despite the fact that the general literature on the cognitive effects of migraine is significant (32–34). Similarly, while multisensory processing in migraine is a topic in the literature (25,25,28), no one has specifically explored it in the context of AE. Thus, our explanations are based directly on our previous work only, and other that this, we can only argue indirectly. Furthermore, it must not be forgotten that our study is just one study, and since neither we nor others have previously examined adult migraine patients with this audiovisual test, it cannot be categorically stated that we did not recruit a patient group with above-average learning performance. Although we consider it unlikely, we cannot entirely rule out this possibility. In compliance with ethical principles, we only included volunteers, which, like in many similar studies, introduces a self-selection bias. Finally, a limiting factor is that although we knew the time elapsed since the last attack, presumably excluding postictal effects, we could not precisely predict when the next attack of a patient would occur, which means some of our measurements could have been influenced by preictal effects. It is evident, therefore, that there are multiple reasons why further investigations would be necessary before we can claim that the observed phenomenon is real.

Still, if we accept the reality of this phenomenon as a working hypothesis, it is logical to ask if its function is to offset the deficit shown in the unimodal task. It certainly plays such a role, but this is no explanation for why the effect is not limited to mere compensation in the acquisition phase, allowing patients to achieve performance on par with healthy controls. It is quite difficult to explain why, if this phenomenon indicates the operation of a mechanism specifically serving to compensate for a deficit, there is a need for overcompensation.

A much simpler, but in our opinion, more plausible explanation could be that this phenomenon is simply a byproduct of the general sensory hypersensitivity observed in migraine. Schwedt (25) points out that hypersensitivity across various sensory modalities can be observed in migraine and that these individual hypersensitivities may also lead to cross-sensitization, meaning that the hypersensitivity of one modality sensitizes another modality, and these ultimately exacerbate the headache itself. The benefit of multisensory integration is that it makes the processing of sensory information from the world faster and more efficient, presumably by enhancing the initial subthreshold component of a typical response to unimodal sensory stimulation (35,36). Accordingly, although specific conclusions cannot be drawn from the data of our study on this matter, it does not seem exaggerated to assume that the threshold-lowering effect of multisensory integration could create a hypersensitization in the already sensitized environment of migraine, resulting in an outcome similar to what we observed in the sensory learning task.

Conclusion

The results of our study support those of previous investigations, which concluded that multisensory processing exhibits a unique pattern in migraine. The specific finding that associative audiovisual pair learning is more effective in adult migraine patients than in matched controls is surprising and difficult to explain with our current knowledge. If this phenomenon is not an artifact, it may be assumed to be a combined result of the hypersensitivity present in migraine and the sensory threshold-lowering effect of multisensory integration. Further research is necessary to clarify these issues, particularly important would be to investigate whether the result can be reproduced in other adult patient populations, or possibly in a different type of bimodal associative learning task.

Article highlights

The study compared cognitive functions between migraine sufferers and controls in an audiovisual test.

Migraine patients outperformed controls in associative learning.

No significant differences in memory retrieval or generalization were found.

Enhanced audiovisual associative learning in migraine patients may link to their sensory hypersensitivity.

Supplemental Material

sj-xlsx-1-cep-10.1177_03331024241258722 - Supplemental material for Enhanced audiovisual associative pair learning in migraine without aura in adult patients: An unexpected finding

Supplemental material, sj-xlsx-1-cep-10.1177_03331024241258722 for Enhanced audiovisual associative pair learning in migraine without aura in adult patients: An unexpected finding by Kálmán Tót, Gábor Braunitzer, Noémi Harcsa-Pintér, Ádám Kiss, Balázs Bodosi, János Tajti, Anett Csáti, Gabriella Eördegh and Attila Nagy in Cephalalgia

Supplemental Material

sj-xlsx-2-cep-10.1177_03331024241258722 - Supplemental material for Enhanced audiovisual associative pair learning in migraine without aura in adult patients: An unexpected finding

Supplemental material, sj-xlsx-2-cep-10.1177_03331024241258722 for Enhanced audiovisual associative pair learning in migraine without aura in adult patients: An unexpected finding by Kálmán Tót, Gábor Braunitzer, Noémi Harcsa-Pintér, Ádám Kiss, Balázs Bodosi, János Tajti, Anett Csáti, Gabriella Eördegh and Attila Nagy in Cephalalgia

Supplemental Material

sj-xlsx-3-cep-10.1177_03331024241258722 - Supplemental material for Enhanced audiovisual associative pair learning in migraine without aura in adult patients: An unexpected finding

Supplemental material, sj-xlsx-3-cep-10.1177_03331024241258722 for Enhanced audiovisual associative pair learning in migraine without aura in adult patients: An unexpected finding by Kálmán Tót, Gábor Braunitzer, Noémi Harcsa-Pintér, Ádám Kiss, Balázs Bodosi, János Tajti, Anett Csáti, Gabriella Eördegh and Attila Nagy in Cephalalgia