Elsevier

Neuropsychologia

Volume 141, April 2020, 107379
Neuropsychologia

Effects of aging on encoding of walking direction in the human brain

https://doi.org/10.1016/j.neuropsychologia.2020.107379Get rights and content

Highlights

  • We developed a novel approach to study neural tuning functions in humans using fMRI.

  • fMRI decoder confusion functions used as proxy measure for population-based tuning.

  • Walking direction could be decoded from retrosplenial complex and early visual cortex.

  • Age-related broadening of directional tuning found only in visual cortex.

  • Approach offers a novel window into age-related dedifferentiation.

Abstract

Human aging is characterized by impaired spatial cognition and reductions in the distinctiveness of category-specific fMRI activation patterns. Yet, little is known about age-related decline in neural distinctiveness of information that humans use when navigating spatial environments. Here, we asked whether neural tuning functions of walking direction are broadened in older versus younger adults. To test this idea, we developed a novel method that allowed us to investigate changes in fMRI-measured pattern similarity while participants navigated in different directions in a virtual spatial navigation task. We expected that directional tuning functions would be broader in older adults, and thus activation patterns that reflect neighboring directions would be less distinct as compared to non-adjacent directions. Because loss of distinctiveness leads to more confusions when information is read out by downstream areas, we analyzed predictions of a decoder trained on directional fMRI patterns and asked (1) whether decoder confusions between two directions increase proportionally to their angular similarity, (2) and how this effect may differ between age groups. Evidence for tuning-function-like signals was found in the retrosplenial complex and early visual cortex, reflecting the primarily visual nature of directional information in our task. Significant age differences in tuning width, however, were only found in early visual cortex, suggesting that less precise visual information could lead to worse directional signals in older adults. At the same time, only directional information encoded in RSC, but not visual cortex, correlated with memory on task. These results shed new light on neural mechanisms underlying age-related spatial navigation impairments and introduce a novel approach to measure tuning specificity using fMRI.

Introduction

A central goal of aging research is to understand how aging-related neurobiological changes affect computational functions of the brain. One important approach has been to investigate how aging changes the representation of sensory information in the brain (Voss et al., 2008; Carp et al., 2011; Schmolesky et al., 2000), which in turn might affect cognitive operations that rely on these representations (Baltes and Lindenberger, 1997; Li et al., 2001). A prominent finding in this regard is that neural patterns are less specific to the category of sensory information in older adults, a phenomenon commonly referred to as neural dedifferentiation (e.g. D. C. Park et al., 2004; Koen and Rugg, 2019, for recent reviews). Here, we studied age-related neural dedifferentiation in the domain of spatial navigation.

In particular, in this study we asked if aging changes how brain areas sensitive to visual and spatial information encode angular walking direction during navigation (Cullen and Taube, 2017; Blair and Sharp, 1996). In young animals, electrophysiological recordings of visually sensitive neurons in primary visual cortex (V1) (De Valois and De Valois, 1980) and direction-sensitive neurons in the thalamus (Taube et al., 1990a, 1990b) have revealed that although most neurons have a preferred stimulus, they are not firing in an all-or-none fashion. Rather, cells tend to fire proportionally to the similarity between the observed stimulus and their preferred stimulus, exhibiting response properties that are well approximated by a so-called Gaussian ‘tuning function’ centered around the preferred stimulus. Modelling work has also shown that a population of cells with those tuning properties will optimally encode an approximately Gaussian likelihood function of the stimulus given the population response; and suggested that this likelihood function is read out, or decoded, by downstream populations that compute optimal behavior based on sensory input (Jazayeri and Movshon, 2006; Averbeck et al., 2006). The focus of the present paper was therefore to understand age-related differences in the properties of population-based tuning functions that encode directional information, which we here broadly define as information used to derive walking direction.

Understanding age-effects on population-level tuning properties is important given the large number of previous investigations that have suggested a loss of specificity of neural representations in older animals and humans. This originated from reports of fMRI activation patterns in inferior temporal cortex losing categorical specificity with increasing age, i.e. activity patterns evoked by face-, place- or word-stimuli are more similar in older versus younger adults (e.g. D. C. Park et al., 2004; Voss et al., 2008; Burianová et al., 2013; Carp et al., 2011). Neural dedifferentiation has also been linked to memory impairment with older age (Zheng et al., 2018; Koen et al., 2019) and related changes to similarity of neural representations might play a crucial role in the encoding and retrieval of memory content (Koen et al., 2019; Sommer et al., 2019). Moreover, electrophysiological recordings in V1 of senescent Rhesus monkeys have found that tuning curves of visual orientation responsive neural populations broaden with age, effectively widening the spectrum of orientation angles a single neuron responds to (Leventhal et al., 2003; Schmolesky et al., 2000). According to the neural broadening hypothesis these changes in firing properties of neural populations are a potential mechanism behind neural dedifferentiation, a notion which found support in a recent fMRI study (J. Park et al., 2012).

However, while electrophysiological recordings showed broadening within a single, continuous domain (e.g. visual orientation), the fMRI evidence is based on increased pattern similarity across distinct domains processed in anatomically separate brain areas (e.g., faces vs. houses). This is an important difference because the broader tuning functions over a continuous domain found in animals likely relate to changes in local inhibitory control (Leventhal et al., 2003). The mechanisms underlying cross-category dedifferentiation across areas as found in humans, on the other hand, must be non-local and are generally much less well understood. Thus, our focus on age-related changes in tuning properties of areas sensitive to walking direction, a continuous variable, would allow us to build a closer link to animal studies. Moreover, the investigation of visual and directional representations that support spatial navigation might lead to insights into why age-related memory impairments are particularly pronounced in the spatial domain (Moffat, 2009; Lester et al., 2017), since spatial memory relies on a sense of direction, for instance during path integration (McNaughton et al., 2006; Seelig and Jayaraman, 2015).

We investigated age-related changes in how information underlying walking direction is represented in the human brain. As is the case for most natural circumstances, walking direction in our task was directly related to visual input, and thus visually-independent directional signals are difficult to discern from visually-dependent directional signals. Therefore, our focus was to understand how any information that could be used to determine one's walking direction was processed in the brain. We analyzed fMRI data from a previous study that used a spatial virtual reality (VR) navigation paradigm (Schuck et al., 2015). This work has shown that the neural underpinnings of different spatial navigation strategies are changed, and partly dedifferentiated in older adults (see also Schuck et al., 2013). In the present paper we went beyond this work by investigating the encoding of walking directional information that is involved in any spatial strategy. Our hypotheses were threefold: first, we expected that fMRI signals stemming from directionally- and visually-tuned neural populations will allow us to decode walking direction above chance (directional and visual similarity were linked in the present data, as they are in daily life). Second, the similarity of two representations arising from different walking directions should be inversely proportional to the angular difference between these directions. Because our focus was on representational structure from the perspective of downstream areas which read out population level tuning functions (Jazayeri and Movshon, 2006; Averbeck et al., 2006), we investigated the probability of a decoder in confusing similar patterns, rather than the similarity directly. A tuning function-like signal should lead to systematically more confusions between neighboring directions, effectively taking the shape of a Gaussian tuning function as seen in the analysis of electrophysiological recordings in animals (Mazurek et al., 2014). Using backwards walking periods and computational analyses of input similarity, we also explored to what extent directional signals were visually dependent. Finally, our most central hypothesis was that older adults should show decreased specificity of walking directional representations, which we tested by comparing the width of the fMRI-derived tuning functions.

Section snippets

Participants

This study is a re-analysis of data from 26 younger (21–34) and 22 older (56–74) male participants, as reported in Schuck et al. (2015). In addition to the exclusion criteria used in the original study (insufficient task performance, signal loss), we excluded participants with an unsuitable distribution of walking direction events that resulted in too little data for at least one direction to train the classifier (three participants, one younger, two older; for details see supplementary

Classification of walking direction

Classification accuracies for each ROI can be found in Fig. 3. One-sided permutation tests (104 iterations) indicated above-chance classification accuracy in the EVC, RSC, and Subiculum (all padj.≤.006) but none of the other ROIs (padj. ≥ .054). Only decoding accuracy in the RSC- and EVC masks, however, exceeded classification level in M1 (both t(42) < 2.58, padj. ≤ .033t42<2.58padj..033). While EVC classification can be expected to be based on visual signals, MRI sensitivity to directional

Conclusions

In this study we used fMRI to investigate age-related changes in the specificity of walking direction-sensitive neural signals. More specifically, we asked a set of three hierarchically structured questions: whether it is possible to decode angular walking direction during free movement, if the similarity of neural patterns associated with these directions declines gradually with larger angular differences, as predicted by directional tuning functions, and whether older adults show broadened

Code availability

Code is available at https://github.com/koch-means-cook/direction_decoding.

Declaration of competing interest

The authors declare no conflicts of interest.

Acknowledgement

This work was funded by a research group grant awarded to NWS by the Max Planck Society (M.TN.A.BILD0004).

We thank Douglas Garrett and Ulman Lindenberger for their helpful comments on the manuscript. Furthermore, we like to thank Lennart Wittkuhn, Nir Moneta, Samson Chien, Anika Löwe, and Ondřej Zíka for their remarks over the course of this project.

References (53)

  • N.W. Schuck et al.

    Human aging alters the neural computation and representation of space

    Neuroimage

    (2015)
  • Nicolas Schuck et al.

    Effects of aging and dopamine genotypes on the emergence of explicit memory during sequence learning

    Neuropsychologia

    (2013)
  • N.W. Schuck et al.

    Aging and a genetic KIBRA polymorphism interactively affect feedback- and observation-based probabilistic classification learning

    Neurobiol. Aging

    (2018)
  • F. Thurm et al.

    Dopamine modulation of spatial navigation memory in Parkinson's disease

    Neurobiol. Aging

    (2016)
  • M.W. Voss et al.

    Dedifferentiation in the visual cortex: an fMRI investigation of individual differences in older adults

    Brain Res.

    (2008)
  • Y. Yang et al.

    Aging affects contrast response functions and adaptation of middle temporal visual area neurons in rhesus monkeys

    Neuroscience

    (2008)
  • A. Abraham et al.

    Machine learning for neuroimaging with scikit-learn

    Front. Neuroinf.

    (2014)
  • D.E. Adamo et al.

    Age differences in virtual environment and real world path integration

    Front. Aging Neurosci.

    (2012)
  • J. Ashburner et al.

    Computing average shaped tissue probability templates

    Neuroimage

    (2009)
  • B.B. Averbeck et al.
    (2006)
  • P.B. Baltes et al.

    Emergence of a powerful connection between sensory and cognitive functions across the adult life span: a new window to the study of cognitive aging?

    Psychol. Aging

    (1997)
  • H.T. Blair et al.

    Visual and vestibular influences on head-direction cells in the anterior thalamus of the rat

    Behav. Neurosci.

    (1996)
  • M. Brett et al.

    Nibabel: Access a Cacophony of Neuro-Imaging File Formats

    (2018)
  • M.B. Cai et al.

    Representational structure or task structure? Bias in neural representational similarity analysis and a Bayesian method for reducing bias

    PLoS Comput. Biol.

    (2019)
  • K.E. Cullen et al.

    Our sense of direction: progress, controversies and challenges

    Nat. Neurosci.

    (2017)
  • S.W. Davis et al.

    Qué PASA? the posterior-anterior shift in aging

    Cerebr. Cortex

    (2008)
  • Cited by (0)

    1

    During the work on his dissertation, Christoph Koch was a pre-doctoral fellow of the International Max Planck Research School on the Life Course (LIFE, www.imprs-life.mpg.de; participating institutions: Max Planck Institute for Human Development, Freie Universität Berlin, Humboldt-Universität zu Berlin, University of Michigan, University of Virginia, University of Zurich).

    View full text