The [[syntax:arith_expr | arithmetic expression]] ''<EXPRESSION>'' is evaluated and expands to the result. The output of the arithmetic expansion is guaranteed to be one word and a digit in Bash.
**Note** that in Bash you don't need the arithmetic expansion to check for the boolean value of an arithmetic expression. This can be done using the [[syntax:ccmd:arithmetic_eval | arithmetic evaluation compound command]]:
echo $(($x)) # Ok. Avoid expansions within arithmetic. Use variables directly.
echo $(("$x")) # Error. There is no quote-removal in arithmetic contexts. It expands to $(("1")), which is an invalid arithmetic expression.
echo $((x[0])) # Good.
echo $((${x[0]})) # Ok. Nested expansion again.
echo $((${x[$((${x[!$x]}-$x))]})) # Same as above but more ridiculous.
echo $(($x[0])) # Error. This expands to $((1[0])), an invalid expression.
* The original Bourne shell doesn't have arithmetic expansions. You have to use something like ''expr(1)'' within backticks instead. Since ''expr'' is horrible (as are backticks), and arithmetic expansion is required by POSIX, you should not worry about this, and preferably fix any code you find that's still using ''expr''.