Age-Related Changes in the Attentional Modulation of Temporal Binding

Age-Related Changes in the Attentional Modulation of Temporal Binding

Jessica Pepper

8th September 2021

In multisensory integration, the time range within which visual and auditory information can be perceived as synchronous and bound together is known as the temporal binding window (TBW). With increasing age, the TBW becomes wider, such that older adults erroneously, and often dangerously, integrate sensory inputs that are asynchronous. Recent research suggests that attentional cues can narrow the width of the TBW in younger adults, sharpening temporal perception and increasing the accuracy of integration. However, due to their age-related declines in attentional control, it is not yet known whether older adults can deploy attentional resources to narrow the TBW in the same way as younger adults.
This study investigated the age-related changes to the attentional modulation of the TBW. 30 younger and 30 older adults completed a cued-spatial-attention version of the stream-bounce illusion, assessing the extent to which the visual and auditory stimuli were integrated when presented at three different stimulus onset asynchronies, and when attending to a validly-cued or invalidly-cued location.
A 2x2x3 mixed ANOVA revealed that when participants attended to the validly-cued location (i.e. when attention was present), susceptibility to the stream-bounce illusion decreased. However, crucially, this attentional manipulation affected audiovisual integration in younger adults but not in older adults. Whilst no definitive conclusions could be drawn about the width of the TBW, the findings suggest that older adults have multisensory integration-related attentional deficits. Directions for future research and practical applications surrounding treatments to improve the safety of older adults’ perception and navigation through the environment are discussed.

Ageing, attention, TBW, multisensory integration, stream-bounce illusion

Participants
This study used a total of 60 participants; 30 younger adults (15 males, 15 females) between 18-35 years old (M = 21.37, SD = 1.30) and 30 older adults (11 males, 19 females) between 60-80 years old (M = 67.91, SD = 4.71). This sample size was determined via an a-priori power analysis using the data of Donohue et al. (2015) and Chen et al. (2021), who conducted similar experiments (see pre-registration on www.aspredicted.com, project ID #65513). All participants were fluent English speakers. Participants were required to have normal or corrected-to-normal vision. Participants were ineligible to proceed with the experiment if they had a history or current diagnosis of neurological conditions (e.g. epilepsy, mild cognitive impairment, dementia, Parkinson’s Disease) or learning impairments (e.g. dyslexia), or had severe hearing loss resulting in the wearing of hearing aids.
Participants were recruited via opportunity sampling; the majority of younger participants were students at Lancaster University and were known to the researcher, whilst the majority of older participants were members of the Centre for Ageing Research at Lancaster University. All participants were able to provide informed consent.

Pre-screening tools
Participants were asked to complete two pre-screening questionnaires using Qualtrics survey software (www.qualtrics.com), to assess their eligibility for the study.
Speech, Spatial and Quality of Hearing Questionnaire (SSQ; Appendix A; Gatehouse & Noble, 2004). Participants rated their hearing ability in different acoustic scenarios using a sliding scale from 0-10 (0=“Not at all”, 10=“Perfectly”). Whilst, at present, no defined cut-off score on the SSQ is available as a parameter to inform decision-making, previous studies have indicated that a mean score of 5.5 is indicative of moderate hearing loss (Gatehouse & Noble, 2004). As a result, people whose average score on the SSQ was lower than 5.5 were not eligible to participate in the experiment.
Informant Questionnaire on Cognitive Decline in the Elderly (IQ-CODE; Appendix B; Jorm, 2004). Participants rated how their performance in certain tasks now has changed compared to 10 years ago, answering on a 5-point Likert scale (1=“Much Improved”, 5=“Much worse”). An average score of approximately 3.3 is the usual cut-off point when evaluating cognitive impairment and dementia (Jorm, 2004), therefore people whose average score was higher than 3.3 were not eligible to participate in the experiment.
The mean scores of each pre-screening questionnaire are displayed in Table 1. An independent t-test revealed that there was no significant difference between age groups on the SSQ questionnaire [t(58) = -1.15, p=.253]; however, there was a significant difference between age groups on the IQ-CODE questionnaire [t(58) = -13.29, p<.001].
Table 1
Mean scores on the SSQ and IQ-CODE pre-screening questionnaires, for both younger and older adults. Standard deviations displayed in parentheses.
Age group SSQ IQ-CODE
Younger 8.34
(1.10) 1.74
(0.51)
Older 8.67
(1.13) 3.03
(0.09)


Experimental Design
This research implemented a 2(Age: Younger vs Older) x 2(Cue: Valid vs Invalid) x 4(Stimulus Onset Asynchrony [SOA]: Visual Only [VO] vs 0 milliseconds vs 150 milliseconds vs 300 milliseconds) mixed design, with Age as a between-subjects factor and Cue and SOA as within-subjects factors.
The experiment consisted of 16 different trial conditions (Table 2), randomised across all participants. Replicating the paradigm used by Donohue et al. (2015), the experimental block contained 72 validly-cued trials and 24 invalidly-cued trials, which were equally distributed between each side of the screen (left/right) and SOA conditions; this means that each participant completed 144 valid trials and 48 invalid trials for each SOA.


Table 2
Number of trials within each Cue and SOA condition. All participants completed a total of 768 trials.
SOA (ms) Cue
Valid (Left)
N Valid (Right)
N Invalid (Left)
N Invalid (Right)
N
0 72 72 24 24
150 72 72 24 24
300 72 72 24 24
VO 72 72 24 24


Stimuli and Materials
Participants completed the experiment remotely, in a quiet room on a desktop or laptop computer with a standard keyboard. All participants were asked to wear headphones/earphones. A volume check was conducted at the beginning of the experiment; participants were presented with a constant tone and asked to adjust the volume of this tone to a clear and comfortable level.
The stimuli used in the task were replicated from Donohue et al. (2015). Each trial started with an attentional cue in the centre of the screen – a letter “L” or a letter “R” instructing participants to focus on the left or the right side of the screen. In addition to this, 2 pairs of circles were positioned at the top of the screen, one pair in the left hemifield and one pair in the right hemifield. The attentional cue lasted for 1 second, and 650 milliseconds after this cue disappeared, the circles in each pair started to move towards each other downwards diagonally (i.e. the two left circles moving towards each other and the two right circles moving towards each other).
In the trials, one pair of circles moved towards each other, intersected, and continued on the same trajectory (fully overlapping and moving away from each other). This full motion of the circles formed an “X” shape, with the circles appearing to “stream” or “pass through” each other. On the opposite side of the screen, the other pair of circles stopped moving before they intersected, forming half of this “X” motion. On 75% of the trials, the full “X”-shaped motion appeared on the side of the screen that the cue directed participants towards (validly-cued trials); on the other 25% of trials, the full motion occurred on opposite side of the screen to where the cue indicated, and the stopped motion occurred at the cued location (invalidly-cued trials).
In addition to these visual stimuli, on 75% of the trials, an auditory stimulus was played binaurally (500Hz, 17 milliseconds), either at the same time as the circles intersected (0ms delay), 150ms after the intersection or 300ms after the intersection. The remaining 25% of the trials were visual-only (i.e. no sound was played). Participants were told that regardless of whether a sound was played, they must make their pass/bounce judgements based on the full motion of the circles (the “X” shape), even if the full motion occurred at the opposite side of the screen that they were attending to.
The experiment ended after all 768 trials – participation lasted approximately 1 hour. The experiment was built in PsychoPy2 (Pierce et al., 2019) and hosted by Pavlovia (www.pavlovia.org).

Procedure
Prior to the experiment, a brief meeting was organised between the participant and the researcher via Microsoft Teams, to explain the task and answer any questions. Participants were emailed a link to a Qualtrics survey, which included the participant information sheet, consent form, demographic questions and pre-screening questionnaires. If the person was deemed eligible to take part in the experiment, Qualtrics redirected participants to the experiment in Pavlovia.
Participants were then presented with instructions detailing the attentional cue elements of the task and asking them to base their judgements on the full X-shaped motion of the stimuli. Participants were asked to press M on the keyboard if they perceived the circles to “pass through” each other or press Z if they perceived the circles to “bounce off” each other, answering as quickly and as accurately as possible.
Participants completed a practice block of 10 trials, then the test session commenced. After each set of 10 random trials, participants had the opportunity to take a break. Participants were provided with a full debrief upon completion of the experiment, and all participants could enter a prize draw to win one of two £50 Amazon vouchers.

Statistical Analyses
This study required two separate mixed ANOVAs to analyse main effects and interactions, investigating significant differences between groups and conditions.
Reaction Times.
For the first dependent variable of reaction times (RT), mean RTs were calculated for each participant in each Cue x SOA condition, representing the time taken, in milliseconds, for each participant to press M or Z on the keyboard at the end of each trial. A 2(Age: Younger vs Older) x 2(Cue: Valid vs Invalid) x 4(SOA: 0ms vs 150ms vs 300ms x Visual-Only) mixed ANOVA was then conducted on these mean RTs.
Bounce/Pass Judgements.
For the second dependent variable of the bounce/pass judgements, the percentage of “Bounce” responses provided in each Cue x SOA condition was calculated for each participant. A 2(Age: Younger vs Older) x 2(Cue: Valid vs Invalid) x 3(SOA: 0ms vs 150ms vs 300ms) mixed ANOVA was then conducted on these percentage data. Visual-Only (VO) trials were compared separately for valid and invalid conditions using a paired samples t-test. Post-hoc paired samples t-tests were also used to investigate significant differences between the 0ms, 150ms and 300ms SOA conditions.
Bounce/Pass Judgements: Pairwise comparisons. To analyse pairwise comparisons in the significant interaction of Age and Cue, responses in each SOA condition were collapsed – that is, a grand mean percentage of “Bounce” responses was calculated by averaging the percentage of “Bounce” responses in the 0ms, 150ms and 300ms trials in the Valid condition and in the Invalid condition. This produced an overall Valid and an overall Invalid mean percentage of “Bounce” responses for each participant. A 2(Age: Younger vs Older) x 2(Collapsed Cue: Valid vs Invalid) mixed ANOVA was conducted on this collapsed data to investigate differences between the proportion of “Bounce” responses in the Valid and Invalid condition for younger adults, and in the Valid and Invalid condition for older adults. In addition, 2 separate one-way ANOVAs were conducted on this collapsed data (Age as the between-subjects factor, and Valid or Invalid as the within-subjects factor) to investigate differences between younger and older adults in the Valid condition, and differences between younger and older adults in the Invalid condition (Laerd, 2015).
Significance.
An alpha level of .05 was used for all statistical tests. Any responses (judgements or RTs) that were ±3 standard deviations from the mean were considered anomalous and were removed from the analyses. Mauchly’s test of sphericity was violated for the main effect of SOA, therefore Greenhouse Geisser adjusted p-values were used where appropriate. As an a-priori power analysis determined the desired sample size for this study, and this sample size was achieved, non-significant results will not be due to the study being underpowered. Statistical analyses were conducted using SPSS (version 25, IBM).

Lancaster University

Data/SPSS.sav; Data/Excel.xlsx

Pepper2021

Robert Taylor

Open

None

English

Data

LA1 4YF

Dr Helen Nuttall

MSC

Cognitive, Perception

60 participants - 30 younger adults and 30 older adults

ANOVA