Neural noise is associated with age-related neural dedifferentiation

Abstract

AbstractAge-related neural dedifferentiation – reductions in the selectivity and precision of neural representations – contributes to cognitive aging and is thought to result from age increases in neural noise. This research has primarily used fMRI to examine age-related reductions in neural selectivity for different categories of visual stimuli. The present experiment used EEG to examine the link between neural noise and age-related neural dedifferentiation indexed by the scene-selective (P200) and face-selective (N170) ERP components. Young and older adults viewed images of scenes, objects, and faces during a 1-back task. Whereas both the P200 and N170 showed age-related slowing of peak latency, only the P200 showed age-related reductions in amplitude that were independent of visual and contrast acuity. We also examined the relationship between the ERP peak measures and an index of neural noise, namely the 1/f exponent of the frequency power spectrum. For the P200 amplitude, higher levels of neural noise were associated with smaller P200 amplitudes in young, but not older adults. In contrast, there was an age-invariant relationship between neural noise and N170 amplitude in the left hemisphere with higher levels of neural noise being associated reduced N170 amplitudes. While the present findings provide novel empirical evidence broadly consistent with predictions from computational models of neural dedifferentiation, the results also highlight potential limitations of the computational model that necessitate revision. The results also suggest that, at least for the P200, maintaining levels of neural noise similar to young adults might preserve levels of neural selectivity.Significance StatementA prominent theory of cognitive aging proposes that age-related cognitive decline results from increases in neural noise that reduce the selectivity of neural representations. We examined this predicted link between neural selectivity and neural noise with ERP components that show selectivity for scenes (P200) and faces (N170) and the 1/f aperiodic exponent measure of neural noise. The amplitude for the scene-selective, but not face-selective, ERP component was reduced in older adults, with both components showing age-related slowing. Critically, older adults with higher levels of neural noise showed lower levels of neural selectivity for scenes, but not faces. While these results provide some evidence supporting computational models of neural dedifferentiation, they also highlight important limitations of the model that require revision.