This is a _fantastic_ example of sampling bias. You simply have no idea when small children are reading silently. You assume they aren't reading until they vocalize, and then when they stop vocalizing you change your assumption to assume they are still reading even though you can't hear it.