We introduce a steganalytic method which takes advantage of statistics that were preserved to prevent the chi-square attack. We show that preserving statistics by skipping certain groups of pixels—apart from reducing the maximum payload—does not diminish the ability to recognise steganographic modifications. The effect is quite reverse: The new detection method works more reliably than the chi-square attack, if the same message was embedded by overwriting least significant bits and straddled over the whole image.