A total of 175 subjects were recruited, including 133 individuals between 18 and 85 years as controls, and 42 subjects with symptomatic age-related hearing loss (presbycusis) between 64 and 90 years. Symptomatic presbycusis patients (hearing loss between 0.5 and 4 kHz >35 dB HL) were prospectively recruited from patients above 60 years that consulted for hearing loss complaints at the Otolaryngology Department of the Clinical Hospital of the University of Chile as part of the Chilean government program¹ to fit hearing aids in presbycusis patients. Controls that gave a self-report of no feeling of hearing loss (Sindhusake et al., 2001), and had no history of otological and audiological diseases were recruited from relatives of consulting presbycusis patients and from University staff and students. The screening hearing handicap inventory for the elderly (HHIE-S) was applied in controls and presbycusis subjects older than 59 years. The HHIE-S has been validated for Spanish speaking population (Lichtenstein and Hazuda, 1998) and measures hearing complaints in daily life. Following suggestions given by Lichtenstein and Hazuda (1998) for Spanish speaking population, possible control subjects with more than 10 points in the HHIE-S were excluded from this study. Presbycusis was confirmed using audiometric thresholds that evidenced bilateral and symmetric hearing loss greater than 35 dB HL in pure tone averages (PTA) between 0.5 and 4 kHz. Patients with middle ear pathology, evidenced by examination of the tympanic membrane or by audiological tests (e.g., conductive hearing loss or flat middle ear compliance) were excluded from this study. All volunteers were Chileans, used Spanish as their native language, and had no clinical history of neurological and psychiatric disorders. All procedures were approved by the scientific ethics committee of the Clinical Hospital of the University of Chile. All subjects gave written informed consent in accordance with the Declaration of Helsinki.

First, we analyzed MBEA global and tasks (T1–T6) performances (the number of correct responses and accuracy [100^∗number of correct responses/(correct + incorrect responses)]) in the three studied groups: (1) controls between 18 and 60 years (n = 84), (2) controls aged > 60 years (n = 49) and (3) presbycusis subjects > 60 years (n = 42). The analysis of the two groups of control subjects aged between 18 and 85 years (n = 133) allowed us to calculate reference values, including means, and cut-off values for global and tasks MBEA performances. Normal distribution of data was evaluated using Shapiro–Wilk tests, and differences between groups were evaluated with Kruskal–Wallis and Dunn post hoc tests. Differences between frequency counts of the histogram distributions were evaluated with X²-tests. Fisher exact test was used to evaluate the numbers of amusic subjects in the three different groups. Descriptive statistics were performed using the Systat software Sigmaplot v12.5. Within all statistical tests p-values < 0.05 were considered as significant.

Next, to study possible relations of MBEA performance with demographic and audiological data in the aged population, we built generalized linear models (GLMs) with R programming language (R version 3.2.1, R Core Team, 2015) using two datasets and excluding controls <61 years. The first dataset included the 49 control individuals aged more than 60 years and the 42 patients with symptomatic presbycusis (n = 91). This dataset was used to evaluate the effects of presbycusis, age, sex, years of formal education and musical training. The second dataset included only presbycusis individuals (n = 42). In this case, we evaluated the effects of education, musical training and different audiological measures, including PTA thresholds, the percentage of word discrimination and HHIE-S scores.

Models were fitted using a binomial family and a logit link, however, a quasibinomial family was used in the presence of overdispersion. We evaluated the effect of each factor by separate and also controlling for the effect of the other factors. In this last case, we first fitted a full model that was simplified to obtain the minimal adequate model. The full model included all the main effects and their paired interactions. Model simplification was performed using X²-tests or F-tests depending whether a binomial or a quasibinomial family was used, respectively. The less significant factor was removed each time. We first removed the interactions and then the main effects. If an interaction was significant but not one of the main effects included in the interaction, the factor was not removed from the model. Model simplification stopped when a significant difference between the tested models occurred (p-value < 0.05), allowing obtaining the minimal adequate model. We checked if the model showed overdispersion, and in that case the model was refitted using a quasibinomial family. The significance of each factor was obtained with a Type-III analysis of variance (Wald-test for binomial family and F-test for quasibinomial family) using the R library “car” (Fox and Weisberg, 2010). While presbycusis, sex and musical training were included as categorical predictors, the other variables where included as continuous predictors. In the case of meter-task analyses (T5) an influential outlier having a low score was removed from the model.

Figure 2 shows box-plots of correct responses in the global MBEA and in the six tasks (T1–T6) of the MBEA in three studied groups. Correct responses in the global scores and in the six tasks of the MBEA were not normally distributed (Shapiro–Wilk, p < 0.05). A Kruskal–Wallis analysis followed by a Dunn post hoc test showed significant differences in global MBEA correct responses [H₍₂₎ = 87.987, p = 0.001] between the three studied groups (Figure 2A). Similarly, correct responses during task T1 were significantly different between the three studied groups [Kruskal–Wallis, H₍₂₎ = 82.793, p < 0.001, Dunn post hoc test]. MBEA tasks T2, T3, T4, and T6 have significant differences between the groups of controls ≤60 years compared to aged controls and presbycusis subjects, but no differences between aged controls and presbycusis [T2: Kruskal–Wallis, H₍₂₎ = 72.606, p < 0.001; T3: H₍₂₎ = 73.739, p < 0.001; T4: H₍₂₎ = 61.236, p < 0.001; T6: H₍₂₎ = 86.829, p < 0.001, Dunn post hoc tests]. Regarding task T5, the only significant difference was obtained between controls ≤60 years and presbycusis subjects [Kruskal–Wallis, H(2) = 8.652, p = 0.013, Dunn post hoc test], while there was no significant difference between controls ≤60 years and aged controls.

We studied possible factors contributing to individual and group MBEA performance. Figure 3 shows individual global MBEA accuracy data plotted with corresponding age, years of education, and group differences between subjects with or without music training. GLMs were used to evaluate a possible dependence of global MBEA accuracy on age, education, sex, music training, and symptomatic presbycusis in subjects aged 60 or more years (n = 91). GLMs fitted to test the effect of each factor separately showed a significant dependence of global MBEA performance on age [F_(1,89) = 11.69, p = 0.0009502], education [F_(1,89) = 24.745, p = 3.165^∗10^-6], music training [F_(1,89) = 8.796, p = 0.003875], and presbycusis [F_(1,89) = 16.678, p = 9.66^∗10^-5] while no significant effect was found for sex [F_(1,89) = 1.829, p = 0.1797]. The minimal adequate models fitted to evaluate the effect of each factor while controlling for the effects of the other factors on global MBEA accuracy suggest that the most relevant factors were presbycusis [F_(1,85) = 26.9284, p = 1.416^∗10^-6], education [F_(1,85) = 26.4913, p = 1.682^∗10^-6] and the interaction between musical education and sex [F_(1,85) = 6.2886, p = 0.01405]. The results of the minimal adequate models of the effects of the six tasks of the MBEA indicated that presbycusis had a significant effect on tasks T1, T2, T4, T5, and T6, while education was significant in tasks T1, T2, T3, T4, and T5, and music training in T1, T2, T3, T5, and T6. Table 3 shows a summary of the statistics values of the minimal adequate models for the performances in the global and six tasks of the MBEA in aged (>60 years) and presbycusis subjects.

Figure 4 shows HHIE-S scores and mean audiometric thresholds in frequencies between 0.25 and 8 kHz in presbycusis subjects. We found a significant correlation between hearing complaints (HHIE-S) and PTA audiometric thresholds [Figure 4C, Spearman, R₍₄₂₎ = 0.392, p = 0.01]. To determine possible audiological factors contributing to the MBEA performance in presbycusis subjects (n = 42), GLMs were fitted considering audiometric thresholds, speech discriminations and HHIE-S scores in addition to years of education and musical training as potential explanatory variables. The minimal adequate model for global MBEA accuracy in presbycusis patients included education, musical training and the interaction between musical training and HHIE-S scores; however, this last factor was not significant as a main effect (Table 4). Regarding the six tasks of the MBEA, musical training was included in the minimal adequate models of the three melodic tasks, one temporal task and in the memory task (T1, T2, T3, T5, and T6), while education was included in both temporal tasks (T4 and T5). HHIE-S was a significant main effect factor in T5 and was included in significant interactions in T2, T3, T5, and T6. Audiological measures (PTA thresholds and speech discrimination) were significant factors included in the minimal adequate model for the accuracy of MBEA meter and memory task (T5 and T6), and PTA appears in a significant interaction in T2. A summary of factors included in minimal adequate models for MBEA performance in presbycusis subjects is shown in Table 4. When tested separately, audiological variables showed non-significant effects on global and specific tasks (T1–T6) of MBEA performance (data not shown).

The MBEA scores obtained in this work allowed us to compute normative data for the Chilean adult population. Table 2 shows that, except for the rhythm task (T4), in average, our Chilean sample of controls aged between 18 and 85 years had two less correct responses than the original Canadian population used by Peretz et al. (2003). Similarly, our own cut-off scores (with two SD) were lower than those reported in Canadians, United States, and Chinese population (Peretz et al., 2003; Cuddy et al., 2005; Nan et al., 2010; Pfeifer and Hamann, 2015). One important factor to explain these differences is the age of the evaluated subjects, as the Chinese and United States groups were young volunteers (<40 years), while our normative sample included people between 18 and 85 years. Differences with the Canadian group, which included subjects between 14 and 79 years, could be attributed to cultural differences between Canadian and Latin-American cultures. For instance, Paraskevopoulos et al. (2010) found cultural differences, mainly in the rhythm task between the original and a Greek version of the MBEA. In our case, the Chilean population is influenced by western music, but also by music from different indigenous cultures such as Andean cultures in the northern region of Chile, and by Mapuche culture from the southern region, being the latter predominantly monotonic and rhythmic music (Robertson, 2007). Therefore, we propose that MBEA performance differences can be explained by different cultural and educational background of Chilean subjects.

Aged subjects have lower performance when completing tests aimed to evaluate the perception of different features of sounds. For instance, older individuals have difficulties discriminating different frequencies (Clinard et al., 2010), frequency modulations (He et al., 2007), distinguishing consonant/dissonant two-note cords (Bones and Plack, 2015), determining pitch variations (Russo et al., 2012), recognizing changes in sound sequence presentation (Fitzgibbons and Gordon-Salant, 2001; Fitzgibbons et al., 2006) and the occurrence of gaps between tones (Schneider and Hamstra, 1999). Here, we found that meter perception (T5) is preserved in aged subjects with no hearing complaints (Figure 2), while pitch and rhythm perception, and musical memory skills are affected by the normal aging processes. A speculative explanation could arise from the case of a musician with a brain tumor that had a right posterior temporal lesion (Baird et al., 2014). After surgery, he was evaluated with the MBEA, showing that meter perception (T5) was preserved, while melodic, rhythm and memory skills were impaired (Baird et al., 2014). In agreement with Liégeois-Chauvel et al. (1998), Baird et al. (2014) proposed that the cortical region located in the right posterior STG is important for melodic, rhythm, and memory abilities, while the anterior STG is critical for meter perception. On the other hand, Sihvonen et al. (2016) studied 77 patients with post-stroke acquired amusia and found that right hemisphere lesions were the most commonly affected brain regions, including the STG, insula and striatum. In addition, they found that temporal anterior lesions were more frequent in rhythm amusia, while posterior temporal and parietal lesions in pitch amusia. Together, neuroanatomical studies show that the brain regions involved in music perception are lateralized to the right STG, however, more studies are needed to define more precisely specific brain regions involved in pitch and temporal dimensions.

Whether there are specific music perception impairments in presbycusis patients is relatively unknown, as only a few studies have evaluated music perception in symptomatic presbycusis patients. For instance, Fitzgibbons and Gordon-Salant (2015) studied the discrimination of intervals within rhythmic tone sequences in aged subjects and in presbycusis patients. These authors found no differences between normal-hearing aged controls (n = 13) and presbycusis patients (n = 15). However, probably the sample sizes of the evaluated groups were not enough to demonstrate significant differences between aged subjects with normal hearing compared to presbycusis patients. In the present work, we found that presbycusis was the most important factor determining global MBEA accuracy. Specifically, presbycusis was an important factor for melodic (T1 and T2), temporal (T4 and T5), and memory tasks (T6) (Table 3).

Because music discrimination of melodic and temporal dimensions can be considered as a cognitive task (Peretz and Zatorre, 2005; Zatorre and Salimpoor, 2013), background experience is an important factor determining performance, constituting an auditory or cognitive reserve (Middleton and Yaffe, 2009; Skoe and Kraus, 2014). For instance, the educational level is a known factor that reduces cognitive decline (Qiu et al., 2001), and a positive relationship between years of education and the performance on auditory processing tests has been found (Murphy et al., 2016). Moreover, although musicians and non-musicians may show similar decays in auditory thresholds with age, musical trained individuals show a better performance during auditory tasks (Zendel and Alain, 2012).

Here, we found that the educational level and music training diminished the alterations observed in music discrimination in presbycusis patients (Figure 3 and Table 4). The mechanisms of the enhancement of the cognitive reserve by music training and by the educational level can be related to previous works showing that the psychoacoustical improvements produced by music training are accompanied by neural plasticity changes in the auditory system (Herholz and Zatorre, 2012). Music related brain plasticity can be observed during early childhood, as evidenced by improved speech in noise perception and larger brainstem responses in children with music training (Strait et al., 2012). In addition, the effects of 8 weeks of auditory training are manifest in behavioral and in electrophysiological subcortical responses in older adults (Anderson et al., 2013). Notably, White-Schwoch et al. (2013) showed that the latencies of brainstem responses in aged individuals are faster in those with music training performed decades before, during childhood or adolescence. In the present work, we confirmed the protective consequences of previous music training in elderly and presbycusis subjects using a perceptual task. Importantly, these protective effects appears with 1 year of musical training, showing that it is not necessary to be a professional musician to get these protective effects. However, whether there is a greater protective effect with more years of music training or in professional musicians is a question that should be addressed in future studies. In addition, the lack of objective measurements of music training might be another limitation of the present results.

Regarding educational level, we found that in presbycusis patients, the years of education were important for MBEA performance in temporal tasks (T4 and T5), while music training for pitch tasks (T1, T2, and T3) and for meter task (T5), showing that different cognitive reserve factors can have consequences on different dimensions of music perception.