Technical report EOG correction of blinks with saccade coef®cients: a test and revision of the aligned-artefact average solution Rodney J. Croft a, * , Robert J. Barry b a Department of Behavioural & Cognitive Sciences, Imperial College of Science, Technology & Medicine, Charing Cross Hospital, St. Dunstan's Road, London, W6 8RP, UK b Department of Psychology, University of Wollongong, Wollongong, Australia Accepted 2 November 1999 Abstract Objective: The `aligned-artefact average' (AAA) procedure was advanced by the authors as a technique suitable for removing eye movement-related artefacts from the EEG. It was proposed that this method would correct both blink and non-blink artefact from the EEG, using the same set of correction coef®cients (Bs). However, recent evidence suggests that this simpli®cation is not always accurate. Thus, we test here a revision of the AAA, including an appropriate allowance for the radial EOG (REOG) component, that does allow the use of the same Bs for the correction of blink and non-blink artefact. Methods: Blink (and saccade) ERP data from 15 subjects were corrected using the AAA method, with Bs calculated from the same blink (and saccade) data set (referent waveforms), or a different set of blink (and saccade) data, or using the new revised AAA procedure (RAAA). Results: AAA Bs calculated from saccades corrected blinks poorly (and vice versa). However, the RAAA Bs corrected blink ERPs better than blink-derived Bs, and saccade ERPs better than saccade-derived Bs. It was also found that irrespective of correction type, inclusion of REOG improved correction. Conclusion: EOG correction is more accurate when the radial channel is included, but inclusion of REOG (and/or HEOG) is not suf®cient to resolve the discrepancy between blink and saccade correction. Using the RAAA procedure, both blink and non-blink data can be corrected using the same set of Bs. q 2000 Elsevier Science Ireland Ltd. All rights reserved. Keywords: EOG; Artefact-aligned average; AAA; Saccades; Blinks; RAAA 1. Introduction Eye movement-related artefact in the EEG causes serious dif®culties in the interpretation of EEG (as is described more fully in Croft and Barry, 1998a). One method of redu- cing this problem is to remove the ocular voltage from the EEG using EOG correction techniques. For example in Eq. (1) the true EEG (estTEEG) at time point (i) is estimated as the measured EEG (MEEG) minus a proportion (B) of the voltage measured about the eyes (Gratton et al., 1983). B is typically calculated separately for each subject and scalp site using Eq. (2) (the least squares method), where X and Y are the EOG and EEG voltages, respectively, at time i. With the use of such a procedure, the experimenter does not have to reject data contaminated by ocular artefact. estTEEG i  MEEG i 2 B:EOG i  1 B  X X i 2 X i Y i 2 Y i = X X i 2 X i  2 2 A dif®culty with such methods is that B has been reported to vary across eye movement type (Corby and Kopell, 1972) and frequency (Gasser et al., 1985), making implementation problematic. However, it has been argued that this B varia- bility may be a function of varying signal (eye movement potential) to noise (neural potential and external artefact) ratio at scalp sites (Croft and Barry, 1998a). The `aligned- artefact average' (AAA) method was thus advanced as a means of overcoming this dif®culty, and it was found that B variation across eye movement types (spontaneous blinks, cued blinks, `up' and `down' saccades) was removed with this method (Croft and Barry, 1998b). The AAA method differs from traditional techniques (e.g. Gratton et al., 1983) in that instead of calculating Bs from raw EOG and EEG data, it calculates Bs from EOG and EEG averages aligned on eye movements. This method is similar to that of Semlitsch et al. (1986), who corrected Clinical Neurophysiology 111 (2000) 444±451 CLINPH 99684 1388-2457/00/$ - see front matter q 2000 Elsevier Science Ireland Ltd. All rights reserved. PII: S1388-2457(99)00296-5 www.elsevier.com/locate/clinph * Corresponding author. Tel.: 144-181-846-7343; fax: 144-181-846- 1670. E-mail address: r.croft@ic.ac.uk (R.J. Croft)