For general purpose, mainstream programming in neuroscience I'd say MATLAB rates pretty high. In addition to its own Neural Network Toolbox and textbooks such as MATLAB for Neuroscientists, freeware software like SPM8 (the current version of a widely-used neuroimaging data/signal analysis tool) are written in MATLAB and some require it. Also, learning the NEURON environment (or similar environments/languages like GENESIS) are great only if it happens that the lab you apply to uses it, otherwise you have to learn to use whatever language or software they do. Lots of psychology departments with neuroscience labs use SAS or (unfortunately) SPSS, both of which are designed to avoid programming and the latter especially makes customizing difficult. Also, while MATLAB is nowhere near as good a language to learn for scientific computing if you really wish to learn to program as is e.g., Python or C/C++, it is vastly easier to use for scientific computing, generalizable to other languages in many ways, and easier to learn thanks to both an highly developped command line as well as script-writing/debugging environment.

Unfortunately, it's also expensive. Basically, if your university or lab isn't paying for it, it's probably unaffordable. The problem is the same is true for SPSS and SAS. Luckily, this is not true of Octave, which is quite similar to the language MATLAB (as opposed to the environment) and there are decent programming environments out there. It's also not true of R, which is great but lacks certain features MATLAB has for modelling as it is through and through a statistical software package.

C/C++ & Java are both less widely used than Python and harder to learn. If you can't go with MATLAB and don't find an Octave interface you like, being familiar with Python and R will go along way. Also, while various institutions use different R IDEs/GUIs (RStudio, R-Commander, and RExcel are all popular and have free versions), none are so different that learning to use R with one will make much of a difference. I would also check out Sage Mathematics which incorporates both R as well as various scientific computing packages which use Python like NumPy & SciPy.

RA positions tend to be pretty flexible even if you are being paid (I use that word loosely). If you can show you have a decent understanding of programming fundamentals, particularly in languages that, while less powerful, are more like MATLAB, Octave, R, and even Python (which in the sciences is typically used as a scripting language in my experience and perhaps is the closest thing to a widely used programming language in general and one widely used in in the sciences), then you're probably good. I've never come across another researcher or lab that uses assembler, I've found too many that rely on button-pushing not programming, and the minority that use highly specialized packages like NEURON tend not to expect an RA to be familiar with them (also, there aren't that many and of the labs that do use so highly specialized software environments many are custom anyway).

For general purpose, mainstream programming in neuroscience I'd say MATLAB rates pretty high. In addition to its own Neural Network Toolbox and textbooks such as MATLAB for Neuroscientists, freeware software like SPM8 (the current version of a widely-used neuroimaging data/signal analysis tool) are written in MATLAB and some require it. Also, learning the NEURON environment (or similar environments/languages like GENESIS) are great only if it happens that the lab you apply to uses it, otherwise you have to learn to use whatever language or software they do. Lots of psychology departments with neuroscience labs use SAS or (unfortunately) SPSS, both of which are designed to avoid programming and the latter especially makes customizing difficult. Also, while MATLAB is nowhere near as good a language to learn for scientific computing if you really wish to learn to program as is e.g., Python or C/C++, it is vastly easier to use for scientific computing, generalizable to other languages in many ways, and easier to learn thanks to both an highly developped command line as well as script-writing/debugging environment.

Unfortunately, it's also expensive. Basically, if your university or lab isn't paying for it, it's probably unaffordable. The problem is the same is true for SPSS and SAS. Luckily, this is not true of Octave, which is quite similar to the language MATLAB (as opposed to the environment) and there are decent programming environments out there. It's also not true of R, which is great but lacks certain features MATLAB has for modelling as it is through and through a statistical software package.

C/C++ & Java are both less widely used than Python and harder to learn. If you can't go with MATLAB and don't find an Octave interface you like, being familiar with Python and R will go along way. Also, while various institutions use different R IDEs/GUIs (RStudio, R-Commander, and RExcel are all popular and have free versions), none are so different that learning to use R with one will make much of a difference. I would also check out Sage Mathematics which incorporates both R as well as various scientific computing packages which use Python like NumPy & SciPy.

RA positions tend to be pretty flexible even if you are being paid (I use that word loosely). If you can show you have a decent understanding of programming fundamentals, particularly in languages that, while less powerful, are more like MATLAB, Octave, R, and even Python (which in the sciences is typically used as a scripting language in my experience and perhaps is the closest thing to a widely used programming language in general and one widely used in in the sciences), then you're probably good. I've never come across another researcher or lab that uses assembler, I've found too many that rely on button-pushing not programming, and the minority that use highly specialized packages like NEURON tend not to expect an RA to be familiar with them (also, there aren't that many and of the labs that do use so highly specialized software environments many are custom anyway).

My friend does computational neuroscience using mostly R and Python.

In my old lab, Python was used a lot for next-gen sequencing data.

One of my classmates uses R mainly to look at genomic changes after drug administration.

It really depends what you want to do. An RA arriving in a neuroscience department would probably be provided with all the software needed to analyse datasets and even explore methodology. However, if you were looking to adapt methods, or write your own analysis packages, you need to consider what the data are and what kind of methods you will use.

Probably the most universally useful language for the high-level maths and manipulation you might need is MATLAB. Useful and wide-ranging tools like SPM are written using MATLAB, and although it has a GUI, knowing the basics will help you manipulate the code, program batch analysis and so on.

For PET, much of the analysis requires image manipulation and curve fitting. MATLAB has very good optimisation and image toolboxes that make these processes fast and easy, using simple programming. Although software does exist to analyse PET, for research most institutions will employ some kind of in-house MATLAB code to investigate methodology. It's not difficult to learn and will open areas of research previously unavailable.

From your tags, it seems you are more interested in fMRI and EEG. Well, again, EEG tools are often written in MATLAB, either in-house or internationally developed, such as FieldTrip. It is not necessary to have a deep understanding of MATLAB to use these toolboxes, but if you do, then you can not only have a better idea of what happens when you press a button, but you can change it and maybe contribute, or write your own, better adapted to your needs.

fMRI is slightly different - little of the fMRI analysis in our department runs on MATLAB. Python and R scripts are generally used to call optimised tools, usually written in C (eg AFNI library). Much of this C code is manipulated locally to suit individual studies. There are now AFNI Matlab libraries too, which can be used for general fMRI analysis (eg FMRISTAT). But if you speak to our fMRI people, they would recommend some level of C or shell scripting especially if you work in a NeuroDebian type environment, where you might want to play with FSL, AFNI and so on. Again, you can get away without knowing much language, but an understanding of many of the commands will speed things up and let you get away with less human interaction.

A lot of this depends on personal preference. I have been programming in MATLAB for 15 years, so have some attachment to it, but it has limitations when it comes to intense parallel repetitive tasks, and despite the GUIDE, its GUI manipulation can be laborious. So if it's a pretty user interface you're after, you may want to also learn Java.

If you don't want to do any mathematics, and it turns out your work is going to be opening a data file, moving data around and closing it again, and you want to automate it, I'd recommend using something else, Python, C++ or C#. So make your choice depending on what you will spend your time doing. Or just learn one and live with the fact it might be a bit trickier to program what you want, but at least you don't have to learn two!

I agree with the answers above.  Being proficient in MATLAB and some kind of stats coding (SPSS/SAS/R) would be incredibly beneficial.  Different schools have different budgets - some can afford the more expensive licenses while others are forced to stick with free or cheaper options.  I would suggest that it doesn't hurt to be familiar with using Linux and a Mac, as many need these platforms for some of the more advanced and large batch fMRI analyses.

Just to reiterate what Andrew mentions, Octave is essentially a free version of Matlab, with less functionality (but the same basic core). Octave would therefore be a big plus on a CV, even if the lab used Matlab. However, it is worth noting that not absolutely everybody has heard of it, and you might have to explain that it is basically the same as Matlab.

A lot of people are moving to Python, but the centre-of-mass is still probably with Matlab. Both are good, and if you have specific labs in mind, you could try to find out which, if either, they used. But frankly, I wouldn't worry too much about the specific language. Experience with *any* kind of programming (be it C++, Java, Python, Matlab, Flash, or whatever) would be a huge plus, since fundamentally the ideas are all pretty interchangeable.

It's different based on what are you exactly studying. for instance, for simulation and also forecasting future results, we need the ready to use class libraries such as what we got in MATLAB and Python. but sometimes making something fast in calculations and expectancy, needs some open source libraries and also a programming language with best compatibility with mathematical functions and using hardware resources. now it's better to go for R language or python and maybe c++. and etc.

In addition to the great recommendations, I would also like to share the Oracle AI Products which can provide a complete AI Ecosystem with the support for several machine learning models like Keras and TensorFlow which can prove both cost effective and efficient based on the size of the workload and the parallel compute power required . Please have a look at https://www.oracle.com/uk/artificial-intelligence/products.html for your respective use cases. In addition to that I would also strongly recommend investigating Oracle Java Streams and/or Scala-Spark , both of which work off the GraalVM to support Polyglot development potentially enabling combination of prewritten models in the likes of Python, Java, JavaScript, Scala, C, C++ and R to build Neuroscience data analytics (Ref https://www.graalvm.org/ )