0 0.01 0.02 mean kde 60 80 100 120 140 mean Onset deciles 0 0.01 0.02 tmean 60 80 100 120 140 tmean Onset deciles 0 0.01 0.02 mean lp 60 80 100 120 140 mean lp Onset deciles 0 0.01 0.02 tmean lp 60 80 100 120 140 5 tmean lp 0 0.01 0.02 mean 60 80 100 120 140 mean Onset deciles 0 0.01 0.02 tmean 60 80 100 120 140 tmean Onset deciles 0 0.01 0.02 mean lp 60 80 100 120 140 mean lp Onset deciles 0 0.01 0.02 tmean lp 60 80 100 120 140 tmean lp 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 Onset (ms) Onset (ms) Session 1 Session 2 Onset deciles Onset deciles The onset of ERP face sensitivity in a sample of 120 subjects = 87 ms [81, 94] 50 100 150 200 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 delta mean lp, n = 120 onset (ms) 0.14 [0.05, 0.22] + 0 [−0.0004, 0.0015] 50 100 150 200 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 Q mean lp, n = 120 onset (ms) 0.51 [0.48, 0.53] + 0 [0, 0.0006] 20 40 60 80 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 delta mean lp, n = 120 age (years) 0.21 [0.16, 0.27] + 0 [−0.0016, 0.0003] 20 40 60 80 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 Q mean lp, n = 120 age (years) 0.54 [0.53, 0.56] + 0 [−0.0005, 0.0001] −100 −50 0 50 100 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 Distribution of test − retest differences Onset differences (ms) −20 0 20 40 2 4 6 8 Distribution of test − retest differences Deciles Onset differences (ms) −30 −20 −10 0 10 20 30 60 70 80 90 100 110 120 130 Onset differences (ms) Onset deciles in session 1 Test−retest comparison 0.4 0.45 0.5 0.55 0.6 0.65 0 5 10 15 Effect size 0 0.1 0.2 0.3 0.4 0.5 0.6 0 1 2 3 4 5 6 delta Effect size 0.7 Q Difference (ms) 0 50 100 150 200 250 0 0.01 0.02 0.03 mean vs. mean lp mean mean lp 60 80 100 120 −10 0 10 0 50 100 150 200 250 0 0.01 0.02 0.03 mean lp vs. tmean lp mean lp tmean lp 60 80 100 120 −10 0 10 Onset (ms) Onset (ms) (1) Institute of Neuroscience and Psychology, University of Glasgow, UK. Corresponding author: Magdalena.Bieniek@glasgow.ac.uk (2) Department of Psychology, Behaviour and Neuroscience, McMaster University, Hamilton, ON, Canada Magdalena M. Bieniek, Allison B. Sekuler, Patrick J. Bennett and Guillaume A. Rousselet 1 1 2 2 Previous studies: ERP sensitivity to faces starts to be visible 50-130 ms. However, these results could be compromised by several problems : (1) high-pass filtering at 1 Hz and above which can smear the onsets back in time (Rousselet, 2012, Frontiers in Psychology); (2) lack of control for multiple comparisons ; (3) group statistics , which ignore individual differences; (4) small sample sizes ; (5) no reliability assessment. When does the human visual system detect faces? METHODS: Stimuli : Data analyses: single subject ERP analyses using Matlab 2012 + EEGLAB + LIMO toolbox (Pernet et al., 2011) Design : group 1 – 120 trials/condition; (Rousselet et al., 2009) group 2 – 128 trials/condition; (Rousselet et al., 2010) group 3 – 150 trials/condition. (Bieniek et al., 2013) Participants : 120 healthy participants (60 females), age 18-81, 74 subjects tested twice. REFERENCES: Bieniek, M. M., Pernet, C. R., & Rousselet, G. A. (2012). Early ERPs to faces and objects are driven by phase, not amplitude spectrum information: evidence from para-metric, test-retest, single-subject analyses. J. Vis. 12, 1–24. Bieniek, M. M., Frei, L. S., & Rousselet, G. A. (2013). Early ERPs to faces: aging, luminance, and individual differences, Frontiers in Psychology, doi: 10.3389/fpsyg.2013.00268. Cliff, N. (1996) Ordinal methods for behavioral data analysis, Mahwah, N. J., Erlbaum. Pernet C. R., Chauveau N., Gaspar C. & Rousselet G. A. (2011) LIMO EEG: a toolbox for hierarchical LInear MOdeling of ElectroEncephaloGraphic data, Computational Intelligence and Neuroscience, Vol.2011, http://www.hindawi.com/journals/cin/2011/831409/. Wilcox, R. R. (2011) Comparing Two Dependent Groups: Dealing with Missing Values, Journal of Data Science, 9, 471–486. Wilcox & Muska (1999) Measuring effect size: A non-parametric analogue of omega(2), The British journal of mathematical and statistical psychology, 52, 93–110. Rousselet, G. A., Husk, J. S., Pernet, C. R., Gaspar, C. M., Bennett, P. J., & Sekuler, A. B. (2009). Age-related delay in information accrual for faces:evidence from a parametric, single-trial EEG approach, BMC Neuroscience, 10:114. doi:10.1186/1471-2202-10-114. Rousselet, G. A., Gaspar, C. M., Pernet, C. R., Husk, J. S., Bennett, P. J. & Sekuler, A. B. (2010). Healthy aging delays scalp EEG sensitivity to noise in a face discrimination task, Froniers in. Psychology, 1:19. doi:10.3389/fpsyg.2010.00019. Rousselet, G. A., Gaspar, C., Wieczorek, K., & Pernet, C. R. (2011). Modeling single-trial ERP reveals modulation of bottom-up face visual processing by top-down task contraints (in some subjects). Frontiers in Psychology, 2 (137), doi:10.3389/fpsyg. 2011.00137. Rousselet, G. A. (2012). Does filtering preclude us from studying ERP time-courses? Frontiers in Psychology 3:131. doi:10.3389/fpsyg.2012.00131. Our study: ERP face sensitivity starts at ~ 87 ms [81, 94] post-stimulus. We solved previous problems by: (1) applying a causal Butterworth high-pass filter which does not distort onsets; (2) using a spatial-temporal bootstrap correction for multiple comparisons; (Pernet et al., 2011, Rousselet et al., 2011, Bieniek et al., 2012); (3) single subject and group level analyses; (4) sample of 120 subjects , with a wide age spectrum of 18-81 years old; (5) test-retest reliability assesment. RESULTS: (A) ERP face sensitivity onset distributions: kernel density estimates + deciles Onset (ms) Onset (ms) Neither low-pass filtering, nor using 20% trimmed means instead of means across ERP trials affects the onset distributions. (B) Comparison of onset distributions (C) Face sensitivty ERP onsets did not change with age Effect size distributions (D) No relationship between individual onsets and the effect sizes at the onset times (A) (B) (C) Session 1 Measures of effect sizes: - delta (range [-1 : 1]): Cliff’s (1996) d statistic estimates the probability that a randomly selected observation from one group is larger than a randomly selected observation from another group (Cliff, 1996); - Q (range [0.5 : 1]): probability of correctly deciding whether a randomly selected observation belongs to the first of two groups, reflects the degree of overlap between two distributions (Wilcox & Muska, 1999). (D) (E) Face sensitivity ERP onsets were reliable across two sessions in 74 retested subjects. (E) 20 40 60 80 0 50 100 150 200 250 mean lp, n = 120 Onset (ms) Age 90 [80, 100] + 0 [−0.2, 0.2] median = 87 ms [81, 94] 20 40 60 80 0 50 100 150 200 250 mean lp, n = 74 Onset (ms) Age 92 [78, 105] + −0.09[−0.4, 0.2] median = 85 ms [79, 92] Session 1 Session 2 EEG recording: kde Onset deciles kde kde 1 9 5 1 9 5 1 9 5 1 9 5 1 9 5 1 9 5 1 9 5 1 9 Onset deciles 9 1 7 5 3 0.7 median onset: 90 ms [82, 98] median onset: 87 ms [81, 94] median onset: 90 ms [82, 98] median onset: 87 ms [81, 94] median onset: 89 ms [81, 96] median onset: 85 ms [78, 92] median onset: 89 ms [83, 96] median onset: 86 ms [79, 93]