Also, there is Stacking involved. So, the pure Sum might not work. If the result can be grouped for all the previous months from start of the year, e.g. Mar 2012 (Jan 2012, Feb 2012 and Mar 2012)

Simple file is attached to illustrate cumulative adding scenario - imagine a bank account, where daily transactions are going in and out. You wish to aggregate values per month, then show them as cumulative, and also see the monthly breakdown.

RUNNINGTOTAL([Payment])
Formula will add every new value to the existing total.

SUBSET_SUM([Payment], SUBSET([Month])) will add the values for each month.

Different nature of the values in these fields will affect choice of functions for aggregation and visualisation :
Sum - for individual Payments;
Last non-empty value for [Total] (you want to see the last balance figure in each month);
Singleton value for [Total per Month], considering they are all the same for records in one month.
Bar/Line View works well for this scenario, especially the cascading option (pictured), giving you both the cumulative value and monthly splits.

Thank you very much Paola!
The formula does sum up the numbers the way required; and it works fine in the graph too.
But, there is also a tiny flaw in applying the CLICK filter. Because the data is still grouped by single Month Year, when user clicks a bar for example Feb 2012 which shows data summary of Jan and Feb 2012; the other graph on the same tab will only show Feb 2012.
I think maybe the only way to achieve this is to hard code the data in the database.

Not sure what kind of behaviour you wish to achieve, but keep in mind that in all Views you can choose the data set - All data/ Filtered/ Selected/ Query etc. so you can choose whether the graph will respond to filtering.