I'm not sure these are cleanly separable or observable. Reasoning is assisted by recall, and if there is some distinct reasoning ability then one can use it to make up for poorer recall by deriving knowledge.
In computational terms: you can gain speed up computation via memoization (trade off memory for more computation), but you can also trade off computation speed for memory with suitable compression. If you only had a black box, how would you distinguish these different internal mechanisms? I don't think you can, increasing either recall or reasoning ability will both improve performance on any conceivable intelligence metrics.