[ACT-R-users] EMMA question (bug?)

[Dario Salvucci]

Dear Marc,

I'm not sure what constitutes "short" encoding times for you, but  
certainly the shorter they are, the more skipping you should observe.

At first glance, the code change below may have adverse effects -- if  
recog-time is between fprep-time and total-time, it's not clear to me  
whether this would do the right thing (specifically whether the  
encoding would complete at the correct time).  But the code as is  
doesn't look quite right either, since as you suggest, the encoding- 
complete might blast the pending movement after preparation.  Part of  
the difficulty here is that the current eye module serves a dual role  
as both a vision module and an eye-movement-motor module; this problem  
you're having seems to reflect the tension between the two roles.

I'll run some further tests with this and we'll be in touch.  (I'll  
also communicate with Mike Schoelles and Mike Byrne, who have spent  
much precious time keeping EMMA up-to-date -- and a sincere thank-you  
to them for their work by the way. :) )

Best,
Dario


On Jul 9, 2008, at 4:39 PM, Marc Halbrügge wrote:

> Hi,
>
> while modeling some visual search behavior, I ran across the  
> following problem:
>
> Salvucci (2001) states that "EMMA can produce skipped gazes by  
> canceling
> saccades when encoding occurs quickly (i.e., before eye-movement  
> preparation is complete)"
>
> but the ACT-R module skips gazes much too often when the encoding  
> time is short. I tracked that down to line 391 in method initiate- 
> eye-move:
>
> (when (<= recog-time (total-time sacc-mvmt))
>  (schedule-event-relative (max 0 recog-time)  'ENCODING- 
> COMPLETE   ;;mjs
>
>
> I my opinion, this should be (changes in bold)
>
> (when (<= recog-time (FPREP-TIME sacc-mvmt))
>  (schedule-event-relative (max 0 recog-time)  'ENCODING- 
> COMPLETE   ;;mjs
>
>
> What do you think?
>
> Greetings
> Marc

_____________________________________
Dario Salvucci, Ph.D.
Associate Professor
Department of Computer Science
Drexel University
http://www.cs.drexel.edu/~salvucci/