Language is much older than writing. But audio and visual cues from sounds and signs don’t leave physical traces the way writing does. So when linguists want to figure out how people talked before history started being recorded, we need to engage in some careful detective work, by comparing two or more similar, known languages to (potentially!) reconstruct a hypothetical common ancestor.  

There is the common assumption, that all languages have the same origin language (proto-language), from which all other languages developed. In Europe, this language is considered to be Proto-Indo-European (PIE), which is considered the ‘ancestor language’ of all European language families. Some assume that this language was spoken from around 4500 BC to 2500 BC. 

Scientists are trying to reconstruct this Proto-Indo-European language using methods of linguistic reconstruction. This reconstruction is based on our current knowledge of how the different languages developed and changed over time. Scientists then try to apply this developent “backwards” in order to reconstruct a past language. 

First, a crash course in "proto-langauges", also known as common ancestors, or parent languages, are grandmums and granddads of current, known languages, which are descended from them and have evolved from them. One proto-language can result in many languages that form a language family, depending on how promiscuous that proto-language was. (Sorry.) Latin is a proto-language, an exception in that it has been very well documented, unlike others.

The earliest known proto-languages include Proto-Indo-European, Proto-Afroasiatic and Proto-Austronesian, which resulted in modern-day languages in modern-day Southeast Asia, parts of continental Asia, Australasia, and the Pacific. Knowledge of proto-languages isn't, generally, very detailed - these languages are old, and reconstructing them takes time, even for the best linguists. The good folk over at UC Berkeley have gone and done one better, by creating a time machine - or, if you prefer, a "computer program" - that can

reconstruct more than 600 Proto-Austronesian languages from an existing database of more than 140,000 words, replicating with 85 percent accuracy what linguists had done manually.

Those numbers are insane, and this is great, big news for everyone, because it saves the time linguists would have otherwise spent to painstakingly recreate and reconstruction proto-languages - hence, the time machine. (Time and Reconstructing Discourse in Speech? It's probably bigger on the inside.)

The model the researchers at UC Berkeley made use of the established theory in linguistics that the evolution of words follows along the branches of the family tree of that language. The method linguists use to find and research proto-languages is called the "comparative method", which is exactly what it sounds like: finding links and connections between languages, between sounds and their evolution, and if they share similarities to one mother tongue.

"To understand how language changes -- which sounds are more likely to change and what they will become -- requires reconstructing and analyzing massive amounts of ancestral word forms, which is where automatic reconstructions play an important role."