ABA Research - Crunching the Numbers

[Guest guest]

If we compare the different ABA research studies in terms of change

in IQ (dIQ) we get the following results:

Study Duration Total Hours Change in IQ (dIQ)

------------ ------------- of ABA ABA Control

----------- ------- -------

Lovaas 1987 24 months 4,160 29.0 10.4

2005 14 months 1,970 31.3 8.7

Sallows 2005 35 months 5,650 24.5 N/A

Cohen 2006 36 months 5,850 25.0 14.0

Normalizing for the total hours of ABA therapy in each study we can

derive the following numbers:

Study dIQ-ABA per dIQ-ABA minus dIQ-Control

------------ 1,000 hours per 1,000 hours of ABA

of ABA -------------------------

-----------

Lovaas 1987 6.97 4.47

2005 15.89 11.47

Sallows 2005 4.34 N/A

Cohen 2006 4.27 1.88

The last three studies sought to duplicate the results of the first

study which was done by O. Ivar Lovaas in 1987. But numerically they

are way off. In terms of delta-IQ for ABA per 1,000 hours of

treatment the 2005 study yielded a value 3.7 times greater

than the Cohen 2006 study. If we look at delta-IQ-ABA minus delta-IQ-

Control the difference is even more pronounced - 2005 is 6.1

times the value of Cohen 2006.

We would not expect such large differences if the ABA therapy in all

four studies was being conducted in the same way. I have yet to read

any explanation on why these numbers are so far apart. The two

closest studies are Sallows 2005 and Cohen 2006 which differ by only

1.6 percent in delta-IQ-ABA per 1,000 hours, but they are 38 percent

less than what Lovaas reported earlier.

If Cohen 2006 and Sallows 2005 are more accurate than the other

studies then this is pretty depressing news since if we take the

second column as a measure of ABA effectiveness versus the control

then Cohen 2006 is 2.4 times smaller than what Lovaas reported.

Just some food for thought.

--- Tom Marking