Cheating in secret sharing has been considered in several papers. Recently cheating in visual cryptography has been considered in [10], where (2,n)-threshold visual cryptography schemes are provided. In this paper we provide new (2,n)-threshold visual cryptography schemes. Our model is different from the one considered in [10]; in particular we aim at constructing cheating immune schemes without the use of extra information, like additional shares or images as done in [10]. We have provided a formal definition of cheating which requires that a group of cheaters be able to deterministically force a honest participant to reconstruct a wrong secret. The (2,n)-threshold schemes that we provide do not allow such cheating, regardless of the number of cheaters.