Gravar-mail: Does the mean adequately represent reading performance? Evidence from a cross-linguistic study