>>I just tested this a few times and switched the order around. It appears you are correct. While the time changes depending on whether it is first, 2nd, 3rd, etc in the order (OCCURS() versus $ etc), $ does seem to be consistently the fastest.
>
>In Boris tests OCCURS() was the winner over $.

:-)
No way!

? [$], SECONDS()-m1                   && 0.045
? [OCCURS], SECONDS()-m1       && 0.284

I even did this:

CLEAR
CREATE CURSOR crsTest (MyMemo M)
APPEND BLANK
REPLACE MyMemo WITH REPLICATE([A],16000000)+[ My String]
DIMENSION atests[5]
atests[1] = [aaa = AT("My String", MyMemo)]
atests[2] = [aaa = RAT("My String", MyMemo)]
atests[3] = [aaa = "My String" $ MyMemo]
atests[4] = [aaa = OCCURS("My String", MyMemo)]
atests[5] = [SELECT * FROM crsTest WHERE MyMemo LIKE "%My String%" INTO CURSOR aaa]
RAND(-1)
CREATE CURSOR Passed (Num I)
INDEX ON Num TAG Passed
SELECT crsTest
DO WHILE .t.
   lnRandom = INT(RAND()*10000000 % 5 + 1)
   IF NOT SEEK(lnRandom,[Passed])
      DoTest(lnRandom)
      INSERT INTO Passed VALUES (lnRandom)
      IF RECCOUNT([Passed]) > 4
         EXIT
      ENDIF
   ENDIF
ENDDO

FUNCTION DoTest(lnNumber)
    lcTest = atests[lnNumber]
    m1  = SECONDS()
    &lcTest
    ? lcTest, SECONDS()-m1
RETURN

When I start it several times the [$] almost always has the same value 0.045-0.046, the rest vary from 0.045 to 0.980.