Pilot study: Comparison of between and within-polarity differ-

Prior to exploring variability in polarity sensitivity across multiple conditions in Experiment II, we conducted a pilot study in a group of 16 young adults with normal hearing to evaluate whether changes associated with stimulus polarity could be differentiated from changes due to test-retest variability. That is, we compared between-polarity differences with within-polarity test-retest differences for vowels with h1 (similar to Aiken & Purcell, 2013). Results of the pilot study indicated that absolute differences in EFR amplitude due to polarity i.e., between-polarity

differences, could be significantly greater than absolute differences in test-retest variability for certain carriers and hence are unlikely due to test-retest variations. Please refer to Appendix B for further details on the pilot study. As will be discussed below, results of the pilot study were consistent with the response patterns seen in Experiment II. Stimulus artifact checks were completed for the Experiment II stimulus. The false positive rate obtained was close to the assumed α

for response analysis (see Appendix C for further details).

5.3.3

Results

Effect of noise variability on polarity-sensitive amplitude differences in EFRs

To assess if electrophysiological noise contributed substantially to polarity-sensitive differences in EFR amplitude, a correlation analysis was completed between

estimates between polarities A and B. The correlation between the two variables was found to be non-significant for both F1 (r(110) = 0.052; p = 0.590) and F2+ bands (r(108) = -0.035; p = 0.718). This suggests that variations in

electrophysiological noise did not contribute substantially to the observed polarity-sensitive differences in EFR amplitude.

Individual variability in polarity-sensitive amplitude differences of EFRs

Like in Experiment I, individuals varied in their degree of change in EFR amplitude due to polarity. Figure 5-3 illustrates histograms of amplitude differences due to polarity for all three vowels, with and without h1, and for both F1 and F2+ carriers. The histogram peak of the F1 carrier of /u/ without h1 (Figure 5-3A) is shifted leftward towards the bin of zero differences compared to the with h1 condition. This suggests a decrease in polarity-sensitive amplitude differences with removal of h1. The F1 carrier of /i/ (Figure 5-3C) shows a modest trend of leftward shift in polarity-sensitive amplitude differences with removal of h1. The F2+ carrier of /i/ (Figure 5-3F) shows a trend of rightward shift in the histogram peak with removal of h1. The carriers /u/ F2+ and /a/ F1 and F2+, do not demonstrate a clear shift in histogram peak with and without h1. These results indicate that the effect of stimulus polarity on the EFR may be specific to certain vowels and formant carriers.

/u/ F1 with h1 /u/ F1 without h1 /a/ F1 with h1 /a/ F1 without h1 /i/ F1 with h1 /i/ F1 without h1

<-70 -69 to -50 -49 to -30 -29 to -10 -9 to 10 11 to 30 31 to 50 51 to 70 71 to 90 91 to 110 >110

/u/ F2+ with h1 /u/ F2+ without h1

<-70 -69 to -50 -49 to -30 -29 to -10 -9 to 10 11 to 30 31 to 50 51 to 70 71 to 90 91 to 110 >110

/a/ F2+with h1 /a/ F2+ without h1

<-70 -69 to -50 -49 to -30 -29 to -10 -9 to 10 11 to 30 31 to 50 51 to 70 71 to 90 91 to 110 >110

/i/ F2+ with h1 /i/ F2+ without h1

Frequency of occurrence

Amplitude difference (nV, Polarity A-Polarity B)

2 4 6 8 10 2 4 6 8 10 A B C D E F

Figure 5-3: Individual variability in polarity-sensitive amplitude differences across carriers. The vertical dashed line indicates the bin containing zero/no difference in response amplitude between the two polarities.

Effect of polarity on F1 carriers

Of the 20 participants, 16 had a significant EFR detected in at least one polarity (Bonferroni corrected). Figure 5-4 illustrates mean EFR amplitudes for the two polarities and Polarity AB average for all stimuli along with the average noise estimates in each condition. RM-ANOVA indicated a significant three-way interaction between polarity, vowel and h1, F(1.53, 23) = 13.71, p < 0.001; η2

partial

= 0.48. In post-hoc analyses, statistically significant differences in EFR amplitude were found between Polarity A, B and Polarity AB average conditions for /u/ with h1. Polarity A amplitude (M = 147.29 nV; SD = 60.02) was significantly greater than Polarity B amplitude (M = 87.26 nV; SD = 45.91) by a mean difference of 60.04 nV,t(15) = 7.99, p < 0.001, 95% CI [44.03, 76.04] and significantly greater than Polarity AB average amplitude (M = 116.47 nV;SD = 51.41) by a mean difference of 30.83 nV, t(15) = 8.33,p < 0.001, 95% CI [22.94, 38.72]. Also, Polarity AB average amplitude was significantly larger than Polarity B amplitude by a mean difference of 29.21 nV, t(15) = 7.44, p < 0.001, 95% CI [20.84, 37.59]. The difference between Polarity A and Polarity AB average approached significance for /i/ with h1 (t(15) = 2.59,p = 0.020; critical p value = 0.011). EFR amplitudes did not vary significantly between polarity conditions for /a/ with h1 and all vowels without h1. Of the 16 participants who had a significant detection in at least one polarity, one participant did not have a significant detection in the Polarity AB average for the stimulus /u/ F1 without h1.

0 50 100 150 200

/u/ with h1 /a/ with h1 /i/ with h1 /u/ without h1 /a/ without h1 /i/ without h1

EFR amplitude (nV) Vowel stimulus Polarity A Polarity B Polarity AB average 0 50 100 150 200

/u/ with h1 /a/ with h1 /i/ with h1 /u/ without h1 /a/ without h1 /i/ without h1

EFR amplitude (nV)

Vowel stimulus

Figure 5-4: Comparison of EFR amplitudes between polarities across F1 carriers (Error bars represent ±1SD; n = 16). * indicates a significant difference. Solid grey bars represent the average noise estimate in each condition.

Effect of polarity on F2+ carriers

Of the 20 participants, 14 had a significant F2+ EFR detected in at least one polarity (Bonferroni corrected). Figure 5-5 illustrates mean EFR amplitudes for the two polarities and Polarity AB average for all stimuli along with the average noise estimates in each condition. RM-ANOVA indicated a significant main effect of vowel, F(1.41, 18.28) = 4.82,p = 0.031, η2

partial = 0.27. Averaged across polarity

conditions and vowels with and without h1, post-hoc pairwise comparisons indicated that EFR amplitudes were significantly higher for /a/ (M = 117.72 nV; SD = 38.37) relative to /i/ (M = 98.93 nV; SD = 51.06) by a mean difference of 18.71 nV, t(13) = 2.88, p = 0.013, 95% CI [4.66, 32.76]. A main effect of polarity was not significant,F(1.01, 13.08) = 0.65, p = 0.434, η_partial2 = 0.05. Of the 14 participants who had a significant detection in at least one polarity, one participant did not have

a significant detection in the Polarity AB average for the stimulus /u/ F2+ without h1 (this was not the same individual as for /u/ without h1, above).

0 50 100 150 200

/u/ with h1 /a/ with h1 /i/ with h1 /u/ without h1 /a/ without h1 /i/ without h1

EFR amplitude (nV) Vowel stimulus Polarity A Polarity B Polarity AB average 0 50 100 150 200

/u/ with h1 /a/ with h1 /i/ with h1 /u/ without h1 /a/ without h1 /i/ without h1

EFR amplitude (nV)

Vowel stimulus

Figure 5-5: Comparison of EFR amplitudes between polarities across F2+ carriers (Error bars represent ±1 SD; n = 14). Solid grey bars represent the average noise estimate in each condition.

5.4

Discussion

5.4.1

Experiment I: Incidence of polarity-sensitive

differences in EFR amplitude

Results of Experiment I illustrate that EFR amplitudes to vowel stimuli of opposite polarities vary across individuals. A few individuals exhibit differences over and above the expected test-retest variability and these differences are unlikely due to variations in noise. When response amplitude estimates are corrected for the

influence of noise, the estimated absolute test-retest variation is 29 nV or 40% of the mean amplitude (Wilding, McKay, Baker, & Kluk, 2012). In the present experiment,

the mean amplitude of Polarity A and B across 24 participants was 100.13 and 92.71 nV respectively. Using the more conservative estimate, the estimated test-retest variation in EFR amplitude would be expected to be within∼39 nV (40% of [100.13+92.71]/2). Therefore, of the 24 participants with a significant detection in at least one polarity, 29% exhibited differences greater than that attributable to test-retest variation. As well, the polarity-sensitive differences obtained here were unexplained by variations in electrophysiological noise. Electrophysiological noise is less likely to have caused large changes in EFR amplitude across polarities partly due to the interleaved method of stimulus presentation where every sweep consisted of both the Polarity A and Polarity B stimulus. The interleaved method shares the variability in extraneous variables such as a change in participant state, electrode placement and change in electrode and inter-electrode impedances between the two test conditions being compared. The test-retest variability estimate chosen here (∼39 nV) was based on studies that evaluated variability across test sessions and, therefore may be higher than the variability experienced here, where the test paradigm was designed to reduce differences in extraneous variables between conditions. However, in summary, it is evident that EFRs elicited to opposite polarities of vowels vary significantly in almost 30% of individuals.