I've got a statement in one of my courses that just says "In large samples, log-likelihoods are approximately quadratic near the maximum." Presumably this has something to do with second derivatives, but can someone very quickly please just explain what is actually meant? Thanks.