Age-related sensory decline mediates the Sound-Induced Flash Illusion: Evidence for reliability weighting models of multisensory perception

Abstract

Perception of our world is proposed to arise from combining multiple sensory inputs according to their relative reliability. We tested multisensory processes in a large sample of 2920 older adults to assess whether sensory ability mediates age-related changes in perception. Participants completed a test of audio-visual integration, the Sound Induced Flash Illusion (SIFI), alongside measures of visual (acuity, contrast sensitivity, self-reported vision and visual temporal discrimination (VTD)) and auditory (self-reported hearing and auditory temporal discrimination (ATD)) function. Structural equation modelling showed that SIFI susceptibility increased with age. This was mediated by visual acuity and self-reported hearing: better scores on these measures predicted reduced and stronger SIFI susceptibility, respectively. Unexpectedly, VTD improved with age and predicted increased SIFI susceptibility. Importantly, the relationship between age and SIFI susceptibility remained significant, even when considering mediators. A second model showed that, with age, visual 'gain' (the benefit of congruent auditory information on visual judgements) was predicted by ATD: better ATD predicted stronger visual gain. However, neither age nor SIFI susceptibility were directly associated with visual gain. Our findings illustrate, in the largest sample of older adults to date, how multisensory perception is influenced, but not fully accounted for, by age-related changes in unisensory abilities.

Figure 1

(a) Structural equation model (SEM) for model 1a and (b) SEM for model 1b. In 1b the mediator is the difference score between self-reported vision and self-reported hearing. ***p = < 0.001, **p < = 0.01, *p < = 0.05. Paths are labelled with standardised coefficients. Red dashed lines indicate negative coefficients, green continuous lines indicate positive coefficients, grey path indicates constrained path for confirmatory factor analysis. SIFI = accuracy for illusory 2B1F trials. The thickness of each line corresponds to the significance of the path. Contrast sensitivity reflects area under the contrast sensitivity curve. Visual acuity reflects “visual acuity score” – see methods.

Figure 2

(a) Structural equation model (SEM) for model 2a and (b) SEM for model 2b. In 2b the mediator is the difference score between self-reported vision and self-reported hearing. ***p = < 0.001, **p < = 0.01, *p < = 0.05. Red dashed lines indicate negative coefficients green continuous lines indicate positive coefficients, grey path indicates constrained path for confirmatory factor analysis. Thickness of each line corresponds to its significance level. Contrast sensitivity reflects area under the contrast sensitivity curve. Visual acuity reflects “visual acuity score” – see methods.

Figure 3

Structural equation model (SEM) for model 3; ***p = < 0.001, **p < = 0.01, *p < = 0.05. Red dashed paths indicate negative coefficients, green continuous paths indicate positive coefficients, grey path indicates constrained path for confirmatory factor analysis. Variance p values all significant < 0.001 and not shown. SIFI = accuracy for illusory 2B1F trials. The thickness of each line corresponds to its significance.

Figure 4

(a) Selection criteria for participants available from wave 3 of TILDA (b) distribution of ages in the sampled cohort.