1.7.3.3 ‘Magnosensitive’ tasks
3.7 Methodological considerations
For the larger older adult data set (Study IV and preliminary cross-modal results), several different control analyses were conduct-ed to ensure that the group differences in the temporal acuity tasks did not result from differences in other participant features.
The results of this group, presented here, were quite well com-parable to those of the smaller young adult data set (Studies II and III). Unfortunately, these control analyses could not be con-ducted for children, because of the noncomputerized experimen-tal setting used with them. (Additional age-effect controls were presented in section ‘effect of adult age’.)
3.7.1 Standard deviation of responses in the temporal acuity tasks
The temporal acuity threshold of each task was the mean of six elementary thresholds [elementary threshold = (upper reversal + lower reversal) / 2)]. The variation in these elementary thresh-olds was expected to be wider if the participant was careless or had difficulties in sustaining attention. Therefore, a coefficient of variation was computed (variation between elementary thresh-olds / temporal acuity threshold). Similar mixed ANOVAs (read-ing group x method x modality/modality combination) as for the temporal acuity thresholds were conducted independently for the unimodal and crossmodal tasks. No significant effect of read-ing group could be detected [unimodal: F(1, 77)=0.10, p < 0.76;
crossmodal: F(1, 77)=0.74, p < 0.40], nor was any of the interac-tions with reading group statistically significant. Further, in sim-ple regression analyses, the variation between elementary thresh-olds in a specific threshold search was used as an independent variable and the specific temporal acuity threshold value as a dependent variable. The resulting residual was interpreted to represent a corrected threshold value and one-way planned ANOVAs (reading group x task) were computed on this varia-ble. The overall reading group differences did not diminish with these corrected thresholds.
3.7.2 Response probability curves
Every single response in the temporal acuity tasks was recorded for later off-line analyses, and response probability curves were plotted based on these data (unimodal 1185 + crossmodal 1185
= 2370 plots, probability of answering correctly/SOA, separately for each presentation type). Despite the identical instructions and order balancing in every task, the response biases differed be-tween the tasks, but not so much bebe-tween the reading groups.
The responses were classified as unbiased, when the participant made incorrect answers both in ‘x first’ and ‘y first’ or in ‘simul-taneous’ (in-phase) and ‘nonsimul‘simul-taneous’ (out-of-phase) pres-entations. There were no biased participants in the auditory, vis-uotactile, and audiovisual TOJ tasks, and only 1 to 3 biased par-ticipants in the other TOJ tasks. In the TPA tasks, somewhat more participants exhibited biased performance, more so for the si-multaneous presentation (with the exception of the tactile TPA).
There was no statistically significant difference between the read-ing groups in the number of biased participants, except in the audiovisual TPA [χ2(1)=5.64, p < 0.02], where more fluent read-ers were simultaneity biased compared to the dyslexic readread-ers (24 and 13 participants, respectively).
3.7.3 Response reliability
In every task, a cut-off SOA value that was double the temporal acuity threshold was calculated for each individual. That is, if the threshold SOA was 100 ms, the cut-off point was 200 ms. Af-ter this, the probability of responding correctly when the SOA was longer (easier) than the cut-off point was calculated. This was used as an index of responding correctly at easy temporal rates throughout a single task. The reading groups did not differ from each other in their indices, except for the auditory TOJ [t(77)=2.06, p < 0.05], in which the fluent readers actually per-formed more poorly than the dyslexic readers. This indicates that both reading groups could sustain a similar level of response accuracy throughout the temporal acuity tasks. Further,
control-ling the response reliability index via regression analyses did not affect the group differences in temporal acuity tasks.
3.7.4 Threshold reliability
A Guttman split-half reliability was calculated for each task from the first six and the last six reversal points of the threshold search.
The split-half reliabilities calculated for the two reading groups were quite comparable, ranging from 0.81 to 0.92 in the TOJ tasks, and from 0.69 to 0.92 in the TPA tasks. One exception was the visual TOJ task, in which the split-half reliability in the case of fluent readers was only 0.17 as compared to 0.76 in the case of dyslexic readers. Interestingly, this was the only task in which the adult groups did not differ from each other statistically sig-nificantly.
3.7.5 Response latency
It was then investigated whether or not the response latencies in the temporal acuity tasks between the reading groups differed.
This could have affected, for instance, the memory load in a giv-en task, and therefore possibly the threshold values. For each participant, the total time in seconds required for obtaining a specific temporal acuity threshold was calculated in every task.
The groups’ total time differences were compared with ANCOV-As: reading group was a between-subjects factor, and the specif-ic temporal acuity threshold and the number of responses in the given threshold search (as the time depended directly on these variables) were used as covariates. The group response latencies differed both in the tactile TOJ [F(1,75)=4.50, p < 0.04] and TPA [F(1,75)=5.49, p < 0.03], as well as in the visuotactile TPA [F(1,75]=4.96, p < 0.03] and audiovisual TOJ [F(1,75)=6.59, p < 0.02]. The dyslexic readers required a longer time to respond in these tasks. However, these differences did not alter the groups’
temporal acuity threshold differences, when controlled via re-gression analysis.
3.7.6 Effect of memory
The possibility that the temporal acuity threshold differences between reading groups had resulted from differences in short-term/working memory capacity was ruled out next. In regres-sion analyses, WAIS-R digit span forward or backward was used as an independent variable and each temporal acuity threshold was used separately as a dependent variable. The resulting re-siduals indicated a memory-corrected temporal acuity thresholds and were used in the same single task analyses as the original temporal acuity variables. Generally, this did not have an effect on the group differences, except in the case of the auditory TPA and audiovisual TOJ tasks when span forwards was the inde-pendent variable. As for span backwards, the results were al-tered in the case of the auditory TPA, audiotactile TOJ and TPA and audiovisual TOJ and TPA tasks. In the young adults’ data (Studies II and III), the group differences in the auditory TPA and audiotactile TOJ tasks were similarly diminished by control-ling either span task.
3.7.7 Effect of intelligence
The groups’ temporal acuity threshold differences did not essen-tially change when a similar regression-based correction was used for controlling WAIS-R PIQ and VIQ. The only exceptions were the auditory TPA, audiovisual TOJ and TPA (the latter only for PIQ) tasks. In the young adults’ data (Studies II and III), the au-ditory TPA and TOJ tasks, and the audiotactile TOJ task, the dif-ferences were similarly diminished by controlling the PIQ or VIQ.