Normal brain and cognitive aging are challenging to study, as advancing age is often associated with various pathologies that compromise cognition. However, there are accumulating evidences suggesting that, even in pathology-free conditions, normal aging is associated with changes in the neural basis of cognition (Hedden and Gabrieli, 2004; Kalpouzos et al., 2009). Cognitive functions are disproportionally impacted by aging, the earliest and most concerned are processing speed, working memory, spatial ability, reasoning, and long-term memory (Hedden and Gabrieli, 2004; Salthouse, 2010). By contrast, measures of semantic knowledge and verbal abilities are relatively stable across the lifespan (Park et al., 2002).

Studies have reported that the brain retains its dynamic properties in aging, with structural and functional rearrangements allowing for new learning and skills improvements (Draganski and May, 2008). It has been suggested that all lifelong environmental features and demands contribute to the establishment of a cognitive reserve, thereby partly counteracting the age-related cognitive decline (Foubert-Samier et al., 2010). Therefore, besides genetic dispositions, both former and current individuals' experiences may influence the quality of cognitive aging (Hultsch et al., 1999; Wilson et al., 2005; Stern, 2012), raising hopes for therapeutic interventions and/or daily lifestyle recommendations (Green and Bavelier, 2008). Epidemiological works have shown that, in addition to education (Evans et al., 1993) and occupational activity (Adam et al., 2013), the degree of individuals' engagement in leisure activities covaries with cognitive functioning in old age (Wang et al., 2012). For instance, Christensen and Mackinnon (1993) found a positive statistical link between mental, social, and physical activities and the cognitive performances of a large sample of participants aged 70–89 years. Longitudinal studies have confirmed that practicing leisure activities in later life is linked to increases in cognitive reserve (Schooler and Mulatu, 2001).

Regarding musical practice, cross-sectional works in children and young adults have reported an association with better performances across a wide range of cognitive domains. Individuals who practice music in their spare time have been found to significantly outperform matched non-musicians on verbal memory (Chan et al., 1998; Brandler and Rammsayer, 2003; Ho et al., 2003; Franklin et al., 2008; Jakobson et al., 2008), vocabulary (Forgeard et al., 2008), spatial ability (Brochard et al., 2004), non-verbal reasoning (Forgeard et al., 2008), short-term memory (Huntsinger and Jose, 1991; Tierney et al., 2008), and working memory (Franklin et al., 2008).

It has been hypothesized that music training has transformative effects on specific cognitive functions because they share mechanisms (e.g., auditory processes involved both in music and in language; Patel and Iversen, 2007; Fauvel et al., 2013; Dawson, 2014). Another claim is that music lessons, being scholar-like activities, instead of leading to specific cognitive improvements, result in a small general gain in Intelligent Quotient (IQ; Schellenberg, 2004), possibly through the potentiation of executive functions (Hannon and Trainor, 2007; Schellenberg and Peretz, 2008; but see Schellenberg, 2011).

However, environmental factors influencing human cognition are challenging to study as well, and previous positive results from quasi-experimental studies should be interpreted with caution regarding the direction of the causality (e.g., high-functioning people are more likely to take music lessons; Schellenberg, 2011). Moreover, a recent well-controlled (random assignment) longitudinal study observed that visual art or music preschool instruction did not influence children's cognitive development (Mehr et al., 2013).

In seeking for potential factors that could promote successful cognitive aging, cross-sectional studies have revealed that both past and recent musical training are associated with better cognitive performances in later life (Hanna-Pladdy and MacKay, 2011; Hanna-Pladdy and Gajewski, 2012). A study conducted in older musicians (60–83 years) with two levels of expertise (high activity (>10 years) vs. low activity (<10 years) found that they scored higher than non-musicians on the delayed recall of a geometric shape (visual long-term memory), a verbal naming task, and Parts A and B of the Trail Making Test (TMT A and B), which measures processing speed (Hanna-Pladdy and MacKay, 2011). Although the performances of the low-activity musicians were halfway between those of the controls and high-activity musicians, suggesting a causal explanation, the differences were not significant. Moreover, those musicians who were still actively engaged in music at the time of the study did not perform better than those who had ceased.

In another study (Hanna-Pladdy and Gajewski, 2012) where general lifestyle activities were controlled for, findings were replicated for visuospatial judgment, verbal memory, verbal working memory, planning functions (non-musicians made more rule violations in an executive function-related task), and phonemic fluency. Statistical analyses revealed that the musicians' visuospatial ability was predicted by recent musical engagement, whereas auditory-verbal memory seemed to be influenced by the early age of musical acquisition. The authors suggested that it could reflect differences in use-dependent adaptation periods depending on cognitive domain.

In their investigation of the association between musical practice and late-life cognition, already published works have focused on musicians who started their training around 10 years old, while quasi-experimentally manipulating the current engagement (active vs. inactive), and by implication, the total duration of training (more or less than 10 years; Hanna-Pladdy and MacKay, 2011; Hanna-Pladdy and Gajewski, 2012). Therefore, the question of whether old musicians who started practicing music in adulthood also display better cognitive performances remains unanswered.

Moreover, any differences uncovered by comparing musicians and non-musicians at a single point in time are of a quantitative nature, and do not inform about the cognitive aging quality of these two populations (i.e., is musical practice associated with better performances in old age and weaker cognitive decline).

The first aim of the present work was to confirm that individuals practicing music since childhood display better cognitive performances than non-musicians. Then, we wanted to explore whether these two populations differ in terms of aging-related performances evolution. Finally, we looked at whether two different characteristics of practice (age at training onset and current frequency of training) account for older musicians.

Indeed, in the perspective that musical training influences the aging of cognitive abilities by stimulating them, we hypothesized that:

Therefore, in Study 1, we assessed several cognitive functions in middle-aged and older musicians (all having started in childhood), and compared their cognitive performances, together with the age-dependent differences, with those of middle-aged and older non-musicians (cf. Figure 1).

In Study 2, we wanted to explore whether two variables of practice (i.e., lifetime period at training onset and current frequency of training) could explain the older musicians' higher scores. We therefore included a sample of older individuals who had started music later in life (mean age at training onset = 42.7 years, SD = 11), and measured their cognitive performances on tests having revealed significant interactions in Study 1. In addition, we ran a correlation analysis between these test scores and all the older musicians' current frequency of training.

In order to explore whether regular engagement in musical activities is associated with a decrease in the age effect on some cognitive functions, we studied the cognitive performances of middle-aged and older musicians in comparison with control participants. In line with the literature, results were as follows:

To seek for a link between practice characteristics and older musicians' superior performances, we conducted a second study (Study 2), in which we included a sample of different older musicians (they had begun practicing music 30 years later on average than the musicians in Study 1). Results indicated that there was no link between frequency of training and verbal fluency performances. Moreover, because the two musician groups (long-term and short-term) had identical performances on the semantic version, we concluded that the older musicians' better ability in this test was not explained by their practice, and probably constituted a confounding factor.

For phonemic fluency, results showed another pattern, indicating that musicians who had started their musical training during adulthood did not differ from controls. This could be in line with works showing that practice-related brain or cognitive differences are more pronounced when training starts at an early age (Elbert et al., 1995; Schlaug et al., 1995; Amunts et al., 1997; Steele et al., 2013). As verbal fluency assesses mainly verbal functions, our results also corroborated studies reporting better performances across a wide variety of language-related tests for children who have learned to play music (e.g., vocabulary; Forgeard et al., 2008; Moreno et al., 2009). In line with a previous investigation (Hanna-Pladdy and Gajewski, 2012), the present study showed that musical training in childhood is also associated with better literacy in old age, regardless of current musical engagement. The long-term influence of early environmental factors on brain and cognition has already been attested in animal studies (Fernández-Teruel et al., 1997). Moreover, in humans, educational attainment is known to be a major factor in determining individuals' cognitive reserve later in life (Stern, 2012), even when occupation is controlled for. In the field of music, Gooding et al. (2013) found that the amount of musical engagement from childhood until adulthood explained older individuals' verbal fluency and long-term memory performances. Therefore, it could be that the cognitive advantage we observed in the older musicians was linked to the musical education they had received during the sensitive period of childhood. From this perspective, music lessons would act as an environmental enrichment which endows people who benefit it with better verbal skills in later life. Then again, given that our two studies had a cross-sectional design, our results may also indicate that individuals whose verbal skills are particularly good in old age are also those who were most likely to come to music early, and to continue their musical training lately.

To sum up, our cognitive investigation of musicians yielded little evidence of reduced age-related changes owing to musical training. In line with this work's hypotheses, phonemic fluency was the only cognitive variable that potentially exemplified a positive use-dependent effect in aging. Hanna-Pladdy and Gajewski (2012) had previously reported better phonemic fluency performances in older individuals who had practiced music in childhood, and the current status of training (active vs. inactive) was not linked with performances (Hanna-Pladdy and MacKay, 2011). This is in agreement with the absence of an association with frequency of training in the present study. In addition, we showed that the period at training onset seems to matter, as older musicians who had started musical practice in adulthood were no different from controls. This could be because there is a sensitive period for brain and cognitive plasticity (Penhune, 2011). The possibility that musical instruction in childhood could have a long-lasting influence on cognition in later life fits in with the definition of reserve, and the notion that “sensitive periods are epochs in development where specific experiences have enhanced and long-lasting effects on behavior and the brain” (Knudsen, 2004).

We found that only a verbal domain-related task fulfilled our criteria to be qualified as “stimulated by musical practice in aging.” This could be viewed as an additional argument in favor of the existence of a special link between musical practice and verbal skills (Chan et al., 1998; Ho et al., 2003; Forgeard et al., 2008). Explanations for this presumed use-dependent enhancement of speech fluency include the processing of common auditory features in music and language (Patel and Iversen, 2007), and the strengthening of the auditory–motor neural coupling that serves both in music and in language production and perception (Bangert et al., 2006).

However, as already stated, phonemic fluency was not better performed by musicians compared with non-musicians for the middle-aged period of life. This is not really what we would expect in the case of actual stimulation. Moreover, in this work, many factors that are known to be associated with cognitive functioning and musical engagement were not controlled for, including the early sociocultural environment (family income, parental first language, parental education, etc.; Elpus, 2013). The first step of the present study investigated older people who had an ongoing musical training since childhood until late-life. Ecologically, such individuals are scarce, and they probably have particular personality traits that prompt them not only to play music, but also to engage in other behaviors that may help them to develop a cognitive reserve (healthy diet, moderate smoking/drinking, and other physical, social, and intellectual activities). Even more important, in this study, none of the IQ scores (verbal or performance) was assessed, but they are important determinants of individuals' cognitive reserve and the likelihood of their taking up music lessons (Schellenberg, 2011). Indeed, if music lessons do lead to improvements in IQ (Schellenberg, 2004), a comprehensive work has rigorously highlighted that children with a high IQ are more inclined than children with a lower IQ to attend music lessons (Schellenberg, 2011).

In the present study, a proxy for performance IQ (Raven's matrices test) showed no between-groups difference, but we found that the semantic fluency task was performed better by musicians, regardless the life period at training onset and the current frequency of training. This result could reflect a tendency of high verbal IQ people (or with a large vocabulary) to come up with—and pursue—music training later in life. However, only the musicians since childhood displayed better performances on phonemic fluency. Therefore, as in Hanna-Pladdy and Gajewski's study (2012), it could suggest that old musician individuals who started practicing music in their childhood display better verbal functions beyond what could be explained by their initially higher verbal IQ.

To conclude, the quality of people's cognitive reserve depends on many intricate factors. It is highly conceivable that people who keep good cognitive functioning in old age are also those who are likely to pursue or expand their daily-life activities. In turn, maintaining these activities may certainly help to keep their cognition young. However, in our work (N = 79/5), which did not control for such selection bias-related confounding factors, we found that highly trained and active musicians (mean training duration = 38 years; mean frequency of training = 10.86 h/week) did not perform any better than control participants on long-term memory (verbal and visual), non-verbal reasoning, and visual scanning (digit–symbol coding). For processing speed and auditory-verbal short-term memory, musicians scored higher but displayed no weaker age-related difference, thus ruling out a protective effect of practice (although both processing speed and auditory short-term memory are highly solicited in musicians). Playing music, as speaking, are auditory-motor behaviors, and there are reports that some individuals own predisposing functional brain organization to learn faster in these abilities (Zatorre, 2013). It is therefore conceivable that some people keep disposing neural substrates that prompt them to pursue music later in life.

Only for phonemic fluency was musical practice associated with a different age-related change in performances, depending on the period at training onset (musicians who started in adulthood performed identically to controls). Therefore, according to us, this test was the solely potential candidate for a use-dependent adaptation interacting with musicians' cognitive aging.

Acquiring an expertise through deliberate practice leads to automation and to strategic processes that allow performances to be improved in a cost-effective way (neural efficiency; Kelly and Garavan, 2005). Other styles of behavior, such as exposure to novelty, may be more efficient in increasing old individuals' cognitive reserve.

Baptiste Fauvel was actively involved in this study from design to drafting. He greatly contributed to participant recruitment and data acquisition. Finally, he conducted all the statistical analyses, produced the figures, participated in the interpretation of the results, and played a central role in revising the article in the wake of the reviewing process. Justine Mutlu made a substantial contribution to this study at every stage. Her assistance was particularly invaluable during revisions, when she participated in the supplementary statistical data analyses, the interpretation of the new results, the rational organization of the ideas, and the design of the figures. Moreover, she revised the manuscript many times in order to eliminate English grammatical mistakes and improve wording. At the outset, Eider M. Arenaza-Urquijo had provided invaluable ideas for interpreting the results. These ideas more or less vanished from the first version of the paper, but became central in the revised one. Moreover, she proved to be a huge help with the statistical data analyses. Finally, she was always generous with her advice, and shared many of her ideas during the revision process. Mathilde Groussard played an active part in this study from beginning to end. She helped in the design and the data acquisition coordination, and brought her theoretical knowledge to bear in the interpretation of the results. Finally, she revised the manuscript several times, allowing essential improvements to be made. As director of the laboratory, Francis Eustache ensured that we had all the material resources we needed to bring this study to a successful conclusion. His experience and knowledge in neuropsychology were essential for conceiving and designing the study, setting out the issues, and interpreting the results. Finally, he revised the manuscript, and provided relevant advice regarding the rational organization of the ideas. Béatrice Desgranges contributed her extensive experience to this study. Her knowledge in the field of cognitive aging proved essential to the study design, the interpretation of the results, and the drafting of the manuscript. Several of her ideas lie at the very core of the final version. Even more important, she highlighted several methodological problems–and how to solve them. Hervé Platel was the person behind this whole research project, and he coordinated the teamwork from start to finish. On many occasions during the design of the study and the interpretation of the results, his theoretical knowledge about music cognition saved us a huge amount of time. Moreover, his knowledge regarding the state of the art helped us to exploit the data in the most relevant way.