NewScientist suggests using bad grammar is a good way to make your passwords more secure. As computing power increases exponentially, the safety of passwords is going down as well. The article mentions that $3,000 worth of hardware is enough to guess 33 billion passwords per second.
An algorithm developed by Ashwini Rao and colleagues at Carnegie Mellon University in Pittsburgh, Pennsylvania, makes light work of cracking long passwords which make grammatical sense as a whole phrase, even if they are interspersed with numbers and symbols. Rao's algorithm makes guesses by combining words and phrases from password-cracking databases into grammatically correct phrases. While other cracking programs make multiple guesses based on each word in a database, putting in "catscats" and "catsstac" as well as just the word "cats", none of the programs make the jump to combine multiple words or phrases in a way that makes grammatical sense, like "Ihave3cats", for instance.
Ten per cent of the long passwords that Rao and her team tested were cracked exclusively using their grammar-sensitive methods, unyielding in the face of other well-known cracking algorithms such as John the Ripper and Hashcat.