Language contains two subsystems, one of sounds and the other of meanings. If you are given the four English speech sounds [p] [l] [i] [d] and asked to combine them into sequences that sound like English words, you will find [plid] [pild] are permissible, while *[pdli] *[dpli] *[lipd] *[idlp], etc. are not. The permissible sequences sound like English words. But they are not, because they do not stand for anything. On the other hand, meanings are conveyed by certain speech sounds or sequences of speech sounds. In English, [DEAD BODY OF A PERSON] is expressed by the word corpse. In this case, we say the concept or the meaning is lexicalized. In English there is no word to stand for the concept [DEAD PLANT]. When certain speech sounds correspond to a certain meaning, a unit of language arises. The same sounds can be recombined to mean something else. In some cases, the same sequence of sounds can mean different things (such as homophones and polysemes). This shows that meanings and sounds make up two subsystems of language. No systems of animal communication possess this feature. The barks of a dog are not analyzable. Animal communication systems cannot be cut into segments and then be reorganized into meaningful sequences. In other words, human languages are discrete while animal communication systems are non-discrete.