This question is ill-designed. There is no "best language" because the language is just a tool. That's like asking what is the best tool a carpenter can have on their belt. There may be a common tool they frequently use or one that can be used in a variety of settings, but rarely there is just one best tool. Instead, it's a question of the variety of things they can build with the tools they are capable of using, and what they are good at building depends on their interests. Some people want to do visual data exploration (R) and others want to work on parallel processing machine learning algorithms (Mahout). Some tools can be used in conjunction (I'm looking at you iPython notebooks! But we can even throw in SAS/IML calling R here).
The question would be better posed as "what tools should someone learning statistics today be comfortable with?" However, even this leaves it open to the analyst's interests. If you want to just do analysis, then a mix of SAS, R, Python (Numpy/Scipy/Scikit), Stata, or Matlab are good, but what if you want to be able to deploy an application that somebody can use? Then you need to learn something like Node.js, C#, Java, and so on. But today "Big Data" is both challenging and new, and there's a whole host of emerging technologies around it (namely, Hadoop and its ecosystem of Mahout, Hive, Pig, Tez).