I'm afraid we've both wandered out of the territory of what DirectRT can do, and into something only MediaLab could handle.

I was incorrect when I suggested DirectRT could do a response time dependent skip, that itself is something you could only do with MediaLab. That is doubly true for your wish to have a complex skip of reaction time and response correctness.

Using DirectRt embedded within MediaLab could accomplish what you want, but it would be a VERY laborious set-up. I'm thinking you would have to have have each individual trial of the IAT be a separate DirectRT input file, with a brief MediaLab screen inbetween to allow a complex skip (utilizing responses.xls) to determine if the participant needs to see a message about hitting the wrong key or taking too long. Of course, this more complex program juggling could add delays itself, meaning you will not achieve the precision you are looking for.

So I think we're back where you started, you want to delay the appearance of the "too slow" screen. I don't see a practical way to do that. You can adjust what's on the too slow screen, you can adjust how long the too slow screen displays for, but I don't think you can insert a delay between the maximum time and the display of the too slow screen.