When right becomes less right: Neural dedifferentiation during suprasegmental speech processing in the aging brain

Abstract

This study combines event-related sparse-temporal acquisition fMRI with structural MRI to investigate older participants (n = 26, mean age = 70.64) with age-typical peripheral hearing. While participants were presented with locally time-reversed sentences of varying temporal integrity, they performed an auditory pattern-matching task. The major aim of the study was to find out whether functional lateralization for speech perception according to the ’Asymmetric Sampling in Time’ (AST) hypothesis shows a similar pattern in elderly individuals as has been previously observed in younger adults. Our findings indicate that, unlike results previously obtained from younger adults, older individuals did not demonstrate the same pattern of rightward lateralization in response to suprasegmental speech cues in the three auditory regions of interest (ROI), namely Heschl's gyrus (HG), planum temporale (PT) and posterior lateral superior temporal gyrus (pSTG). A frequentist statistical approach did not yield evidence for functional lateralization in the aging brain, and this was corroborated by Bayesian evidence which supported the absence of lateralization in older adults in response to the suprasegmental manipulation. However, a relationship between structural measurements and functional responses demonstrated that individuals with thicker right PT showed less variance in lateralization. Hence, this study extends the division of labour between the left and the right auditory cortex during the processing of continuous spoken language as proposed by the 'AST'-hypothesis in younger adults to a lifespan context.

Introduction

Confluent evidence from neuroscientific studies of recent years resulted in the development of frameworks that strongly suggest a bilateral organization of the language-related network. While these models still delineate language as a left dominant faculty, they also acknowledge the contributions of the right perisylvian territory as being essential for the processing of language (Friederici, 2011; Hickok and Poeppel, 2007; Sammler et al., 2015; Specht, 2014). The ’Asymmetric Sampling in Time’ hypothesis (Poeppel, 2001, 2003) provides an elegant proposal for the division of labour between the left and the right auditory-related cortex with respect to the processing of continuous spoken language. According to this hypothesis, an initial stage of speech perception relies chiefly on subsegmental, temporal fine-structure (lower γ-band: about 40 Hz) and suprasegmental, temporal envelope information (θ-band: about 4 Hz). Whereas fine-structure information characterizes phoneme representation, the envelope information represents the more global units that constitute prosodic modulation. Furthermore, the ’Asymmetric Sampling in Time’ hypothesis postulates a hemispheric preference in temporal information processing: while the left hemisphere, namely the auditory-related cortex, processes primarily fine-structure information, the right auditory-related cortex appears to be preferentially driven by envelope information, that is intonation contour. A number of studies in the last decade have supported this hypothesis and have identified the posterior auditory-related cortex as the principal processor of temporally-changing speech cues (Boemio et al., 2005; Geiser et al., 2008; Hesling et al., 2005a, b; Hurschler et al., 2013, 2015; Liem et al., 2014; Meyer et al., 2002; Plante et al., 2002; Zaehle et al., 2004; Zhang et al., 2010). Hence, it appears that spoken language has evolved as a bilateral function even though this view does not challenge the privileged role of the left peri-sylvian cortex for language functions as higher linguistic processes, namely syntax and grammar, are still thought to be strongly left lateralized (Friederici, 2011, 2012). However, during the initial stages of pre-lexical processing, there is a lateralized specialization. The result of this computation includes information about the phonemic and prosodic pattern of a continuous spoken utterance. Based on this information, incoming acoustic signals are recognized as a known language and distinguished from unknown languages, music or environmental sounds. In addition, this computation lays the foundation for the subsequent higher-order linguistic processes that integrate syntax, grammatical structure and meaning. Previous studies have reported differences in lateralization between age groups for neural responses to CV-syllables (Bellis et al., 2000), auditory steady-state responses (Goossens et al., 2016) and semantic decision and verb generating tasks (Nenert et al., 2017; Szaflarski et al., 2006). All these studies agree in that language lateralization that can be observed in children and young adults decreases in older adults. However, the predictions of the ’Asymmetric Sampling in Time’ model have so far mainly been tested with younger adults and hence not much is known about the stability of the lateralization patterns in response to temporally-modulating speech sounds across the life-span.

Sensory degradation is especially evident in the realm of speech perception, since the likelihood of experiencing sensori-neural hearing loss increases with increasing age (Brant and Fozard, 1990; Cruickshanks et al., 1998; Wiley et al., 2008; Roth et al., 2011). However, loss of hearing sensitivity cannot account for the entirety of speech perception differences between younger and older adults and thus, investigating the ‘central’ factors of speech processing is also crucial for the understanding of difficulties in speech comprehension typically encountered by older individuals (Peelle and Wingfield, 2016). Taken together, the aging brain is subject to neurofunctional changes in response to alterations of peripherally-driven (e.g. degradation of sensory perception) (Lin et al., 2014; Profant et al., 2014) or morphologically-triggered (i.e. cortical atrophy) changes (Reuter-Lorenz and Park, 2014; Sowell et al., 2003). The exploration of spoken language processing in middle-aged adults and seniors is therefore of particular interest because peripheral and central functions are required in order to successfully compute and integrate speech signals and, thus far, little is known about the central contribution to the comprehension of spoken language regarding this tension between age-related hearing and cortical atrophy (Humes et al., 2012). The acquisition of knowledge about various factors of speech perception is vital for the development of diagnostic methods and treatment approaches for age-related difficulties in spoken language perception that can better account for the multifactorial circumstances of this phenomenon. Ultimately, the inability to partake in spoken communication can have a detrimental effect on well-being (Arlinger, 2003; Vannson et al., 2015). Thus, ameliorating the effects of age-related declines in speech perception might potentially contribute to a better quality of life in older adulthood. Furthermore the knowledge about the relationship between plastic structural and functional changes in older adulthood may help contribute to a better understanding how the brain adapts to hearing aids (Giroud et al., 2017).

The main aim of this study was to shed light on how older adults process spectro-temporal speech information. For this purpose, we applied the paradigm of a previous fMRI study that investigated suprasegmental speech perception in young (M_age = 25, SD = 3) adults (Liem et al., 2014). The integrity of slowly-changing acoustic cues was parametrically manipulated according to a procedure introduced by Saberi and Perrott (1999). The results of our previous fMRI study showed that lateralization to the right auditory-related cortex increased with parametrically increasing temporal integration window length, supporting the idea of a preference of the right auditory-related cortex for suprasegmental speech information. We tested the predictions of the ’Asymmetric Sampling in Time’ hypothesis for suprasegmental speech processing in a sample of older adults without considerable peripheral hearing loss primarily to evaluate the effect of central losses on spoken language comprehension.

If this pattern of lateralization persists across the lifespan, one would expect to see the same kind of lateralization pattern in older adults as seen in younger adults. If this specific kind of processing, however, is subject to lifespan changes, we would expect to see a different pattern which could indicate either a switch of lateralization towards the other hemisphere or a loss of asymmetric processing. In light of results from other work investigating lateralization in speech processing (Bellis et al., 2000) and from other cognitive domains (Reuter-Lorenz and Park, 2014; Cabeza, 2002), it seems, however, more likely that lateralization decreases.

Regarding structural plasticity research so far, cortical thickness of the human brain is subject to age-related changes (Thambisetty et al., 2010; Salat et al., 2004) and there is initial evidence that the structure of the right auditory cortex plays an important role in the maintenance of speech processing abilities at an older age (Giroud et al., 2018). In particular, cortical thickness of right auditory-related cortex regions has been positively related to performance in various auditory functions, meaning that participants with a thicker cortex in those regions scored higher in the tasks. This observation suggests that, in the context of age-related central hearing loss, neuroplastic processes which affect the right auditory-related cortex are associated with behavioural performance. Interestingly, findings regarding cross-domain plasticity also suggest that the right auditory-related cortex, specifically the PT, is involved in neuroplastic reorganization processes (Shiell et al., 2016). On a behavioural level, older adults, compared to younger individuals, appear to rely more strongly on suprasegmental speech information (Wingfield et al., 2000) which, according to the ’Asymmetric Sampling in Time’ framework, is preferentially processed by right hemispheric non-primary auditory regions. Taken together, these findings all point towards a crucial role of the right auditory-related cortex for central speech processing in older adulthood.

Currently, there is a debate around what the specific age-related changes in patterns of neural activation may mean for the individual's performance, or their ability in a given task or task domain at hand. In general, there are two competing hypotheses which make different predictions for the relation between lateralization patterns and behavioural performance. On one hand, changes in lateralization in old age could represent the brain's attempt to compensate for deficits in the domain of the task by recruiting additional areas of the brain (Cabeza, 2002). Along these lines, less lateralized processing would be expected to lead to a better performance, since an increased recruitment of contralateral homologue areas would indicate that more neural resources are available. It should be noted that this hypothesis has been formulated with respect to prefrontal activation; however, to the best of our knowledge, there is no evidence yet that would rule out the possibility of a similar mechanism for auditory-related areas. Thus, we consider compensation as one candidate mechanism that may account for changes in lateralization during elementary speech processing following age-related central hearing loss. An alterative hypothesis states that changes in lateralization are a consequence of neural dedifferentiation (Baltes and Lindenberger, 1997; Li et al., 2004), meaning that neural responses to stimuli become less specific in older adulthood. In other words, the human aging brain buys robustness at the expense of specialization. With respect to spoken language comprehension, this would manifest in a less specified lateralization pattern which could be characterized by a larger variation in functional lateralization. Increases in dedifferentiation in aging have been linked to decreases in cognitive abilities (Li et al., 2001; Park et al., 2010). Thus, if increased symmetric processing is a means of compensation, one would expect participants with a less lateralized activation pattern to perform better in a behavioural task. Alternatively, if reduced lateralization accompanies neural dedifferentiation, one would expect individuals who show less variation in their neural response pattern to perform better in a behavioural task. Hence, a secondary aim of the study was to investigate how the lateralization pattern in older adults fitted to current alternative frameworks of neurocognitive aging, namely compensation versus dedifferentiation. In order to test the predictions made by each hypothesis, mean lateralization across conditions and inter-condition standard deviation were calculated and used to predict mean behavioural performance in the in-scanner task.

Lastly, it was of interest to find out how age-related cortical atrophy affected temporal lobe neural functioning, given that gray matter is subject to lifespan cortical thinning (Sowell et al., 2003; Fjell et al., 2009; Fjell and Walhovd, 2010), in particular with respect to central hearing loss (Lin et al., 2014). Despite the fact that only little is known about this relationship so far, there is at least initial evidence that demonstrates a direct link between brain morphology and auditory brain function (Greve et al., 2013; Liem et al., 2012b) as well as performance (Giroud et al., 2018). Thus, another aim was to investigate whether measures of cortical anatomy, namely cortical thickness and cortical surface area, were related to brain function by correlating them with mean lateralization and inter-condition standard deviation. Investigating the interplay between brain structure, function and behavioural outcome in relation to aging thus seems to be a promising approach to a better understanding of the complex nature of the aging brain.

In summary, given that older adults tend to show less lateralization in general, we expected that older adults would show a bilateral pattern of neural responses to suprasegmental speech information, and that this could best be explained by a loss of specificity of neural responses in older adulthood and not by compensation. Furthermore, we assumed a relationship between structure of the brain and brain function. In particular, we expected that cortical thickness especially would be a relevant anatomical parameter related to neural function in older adulthood.