Why do languages change? Well, there's been many theories about why languages change. This has intrigued people since time immemorial and it seems that almost everybody has an idea. One early example can be found in Bible in the form of the Tower of Babel, where God decided humans got a little too much hubris (oops...wrong mythology) and so made their lives miserable by giving everybody different languages.

As science became a more dominant force in society, scientific explanations to language change were proposed. Here's a few through the years:

Language Decay?

The 18th century view of language is one of decay and decadence. Their reasoning was that the old Indo-European languages like Sanskrit, Greek and Latin all have complex declension and conjugation schemes, where as the modern Indo-European languages have far fewer cases for declension and conjugation. This "loss" of declension and conjugation cases was a result of speakers of the language getting increasingly careless about their speech (read "lazy"), so the modern speakers are "decadent" as they have allowed the once complex language to decay into such a "simple" language.

Obviously, this "decadence" argument has one major flaw. Modern languages are not less powerful means of expression than the ancient ones.  Even though the number of declensions and conjugations has dwindled, other parts of speech such as particles and auxiliary verbs have evolved to take their place. Anything that can be expressed in the ancient tongue can still be expressed today. Ultimately, this theory is highly subjective, as it relies on personal opinions, not scientific facts, of what is "highly evolved" and what is "decadent". Therefore this is not science.

Side note here: Even though linguistics has moved beyond this 18th century theory of language decay, many self-appointed pundits are still using this excuse to stamp out dialectal variations throughout the world by justifying the dialects as "decadent". This is, of course, complete nonsense, as even the most weird sounding dialect has regular grammatical structure and works perfectly to express ideas as well as the standard language.

Natural Law?

The next theory, proposed by the group of language scientists who called themselves `Neogrammarians' (Junggrammatiker) in the late 19th century, is one of natural process. The Neogrammarians stated that changes are automatic and mechanical, and therefore cannot be observed or controlled by the speakers of the language. They found that what sounds like a single "sound" to a human ear is actually a collection of very similar sounds (usually called allophones). They call these similar sounds "low-level deviation" from an "idealized form". They argue that language change is simply a slow shift of the "idealized form" by small deviations.

The obvious problem here is that without some kind of reinforcement, the deviation might go back and forth and cancel out any change. Then the Neogrammarians patched this theory by adding reasons for reinforcing the deviation such as simplification of sounds, or children imperfectly learning the speech of their parents.

The simplification of sounds basically states that certain sounds are easier to pronounce than others, so the natural tendency of the speakers is to modify the hard-to-say sounds to easier ones. An example of this would be the proto-Romance word /camera/ "room" changing into early French /camra/. It is hard to say /m/ and /r/ one after another, so it was "simplified" by adding /b/ in between, to /cambra/ (hence leading to modern French "chambre"). A more recent example is the English word "nuclear", which many people pronounce as "nucular". The problem with this theoretical patch is that, since not everything in a language is hard-to-pronounce (unless you're speaking Klingon), the process would only work for a small part of the language, and could not be responsible for a majority of sound changes.  Secondly, it is highly questionable to determine whether "nucular" or "nuclear" is easier to pronounce. You'll get different answers from different people. Simplification no doubt exists, but using it as a reason (not a symptom) of language change is applicable in too few cases and too subjective to be scientific.

The next patch, that of children incorrectly learning the language of their parents, doesn't work either. Let's take an extreme case in the form of immigrants. What is observed is that children of immigrants almost always learn the language of their friends at school regardless of the parents' dialect or original language. (And yes, the children become multilingual, but that's another story...) In fact, children of British immigrants in the United States nearly always speak with one of the many regional American accents. So in this case, the parents' linguistic contribution becomes less important than the social group the child is in. Which leads to...

It's Social Bonding

The last theory advanced during this century is a social one, advocated by the American linguist William Labov to account for the spread of pronunciation changes. What he found was that at the beginning of a spreading sound change, a small part of a population pronounces certain words that have, for example, the same vowel, differently than the rest of the population. This occurs naturally since humans don't all reproduce exactly the same sounds. However, at some later point in time, for some reason this difference in pronunciation starts to become a signal for social and cultural identity. Others of the population who wish to be identified with the group either consciously or (more likely) unknowingly adopt this difference, exaggerate it, and apply it to change the pronunciation of other words. If given enough time, the change ends up affecting all words that possess the same vowel, and so that this becomes a regular linguistic sound change.

We can argue that similar phenomena apply to the grammar and to the lexicon of languages. An interesting example is that of computer-related words creeping into standard American language, like "bug", "crash", "net", "email", etc. This would conform to the theory in that these words originally were used by a small group (i.e. computer scientists), but with the boom in the Internet everybody wants to become technology-savvy. And so these computer science words start to filter into the mainstream language. We are currently at the exaggeration phase, where people are coining weird terms like "cyberpad" and "dotcom" which didn't exist before recent developments in computer technology.

To me the social theory of language change sounds much more plausible than other previous theories. Humans are, after all, social animals, and rarely we do things without a social factor.