Let's do it as a scalar query to see how it pans out

SQL> select empno,
  2         sal,
  3         ( select count(distinct sal)
  4           from   emp e2
  5           where  e1.sal<=e2.sal)
  6  from emp e1;

     EMPNO        SAL (SELECTCOUNT(DISTINCTSAL)FROMEMPE2WHEREE1.SAL<=E2.SAL)
---------- ---------- ------------------------------------------------------
      7369        800                                                     12
      7499       1600                                                      6
      7521       1250                                                      9
      7566       2975                                                      3
      7654       1250                                                      9
      7698       2850                                                      4
      7782       2450                                                      5
      7788       3000                                                      2
      7839       5000                                                      1
      7844       1500                                                      7
      7876       1100                                                     10
      7900        950                                                     11
      7902       3000                                                      2
      7934       1300                                                      8

So you can see for each row, we are counting all the distinct salaries in the table that are greater than or equal to the salary we are currently on.

Thus when that count is (say) 2, we know that the row were are on, has a sal that is either matched or exceeded by at most one other salary in the table.

Hope this helps.