Sergey,

Yes they do, that's why the final result is controlled by the least accurate value involved in the calculation.

>It was vary long time ago I stadied this stuff so I may be wrong but doesn't implicit errors compound when you do arithmetichs with such numbers?
>
><snip>
>>And none of this changes the fact that when these numbers are stored in a binary computer with a fixed number of bits that additional errors are going to be introduced.
>>
>>3.1 ± 0.05 * 2.5567 ± 0.00005 is 7.9 ± 0.05.
>>3.1000 ± 0.00005 * 2.5567 ± 0.00005 is 7.9258 ± 0.00005
>>
><snip>