An alternative definition of AFNI can be found at this page.

AFNI refers to both the entire software package (comprising dozens of program and plugins), and also to the specific program of that name. The situation is confusing, but that's the way it is.

The name starts with the letter "A", which means that AFNI almost always appears first on any list of FMRI software packages. This is as it should be, of course, but was not intentional in any way.

Why AFNI is Free
I am often asked why I chose to give AFNI away, rather than sell it, or keep it as in-house software for MCW use alone. Frankly, I don't want the support hassles that would be incurred with commercialization - right now, your only support option is to hope that I will answer your e-mail (or message board posting) before the innermost circle of hell thaws out. As to keeping it proprietary, I've benefited greatly from code and other things that people have put on the Internet, and so felt like giving something back.

Asking For Help
If you do send me e-mail about a problem, please realize that there are well over 100,000 lines of code in the package - not counting stuff taken off the Internet, such as the cdf and volpack packages - and that I don't keep all the details in my head (I have to save most of my synapses for Star Trek and Tolkien info). This means that you should send me a detailed description of what you did, including a copy of the command line, if you were running a batch program, any error messages from the program, and some information about the datasets you input (e.g., the output of 3dinfo). Just saying "I tried this and it didn't work. What's wrong?" won't usually be enough.

With the AFNI package, I provide you with some of the tools needed to analyze FMRI data. This does not mean that I am an expert on analysis of your particular data. I probably can't advise you on what is the proper set of tools to apply. Questions that there are no authoritative answers to include

• Should I blur my results before intersubject averaging (or other analysis)? How much?
• What are the best values for the clustering parameters rmm and vmul (cf. Q24)?
• Should I average time series and then analyze, or analyze first and average later?
• I don't see any subject motion; should I register images anyway?
• Which is better: correlation analysis, ICA, or fuzzy clustering?
• Which is better: AFNI, MedX, or SPM?
• Is there an infinite set with cardinality between that of the natural numbers and that of the continuum?

FMRI Courses
The Medical College of Wisconsin offers a 3-day course on "How to do FMRI", including a 1 day introduction to the use of AFNI. More information can be found here. This course can get you going in FMRI data gathering and analysis - it is not really appropriate for those who already have extensive FMRI experience.

The MGH NMR center also offers a short course on FMRI; information about this course can be found here.

Some educational materials about AFNI and FMRI in general can be found on this web site here.

AFNI Version Numbers
Version numbers are assigned by whim. The first version distributed outside of MCW was 1.10, released in Feb 1995. Version 2.00 was the upgrade from 3D to 3D+time datasets (which involved a lot of internal changes). Since then, the version number has only crept upwards, since no vast changes to the package have been made - only incremental additions, mainly in the form of new "3d*" command-line programs and new plugins. You can check your version of AFNI by running the command line program afni_vcheck, or by using the Define Datamode->Misc->Version Check menu item in the interactive AFNI program.

Referencing AFNI:
If you wish to refer to AFNI in a publication, then the proper reference is

RW Cox.
AFNI: Software for analysis and visualization of functional magnetic resonance neuroimages.
Computers and Biomedical Research, 29: 162-173, 1996.

[Answer last changed 12 Nov 2001]