Following up a 3-way interaction in a mixed ANOVA using LMATRIX and MMATRIX in SPSS

#1
Hi everybody,
I have conducted a memory study, where participants were either exposed to a stress task or a non-stressful control task in either an immediate or a delayed condition and then had to recall three different types of items. So I have a mixed design with one within-subjects factor "Item type" (a1, a2, a3) and two between-subjects factors "Delay" (b1, b2) and "Stress" (c1, c2) and recall performance as the dependent variable. I’m investigating a certain memory effect, that is constituted by the recall performance for item type a3 being significantly lower than the recall performance for item type a2. My research question is whether this memory effect is differently affected by stress depending on whether stress followed immediately after learning or after a delay.
Using GLM I've calculated a mixed ANOVA that yielded a significant main effect of Item type on recall performance, a significant main effect of Delay, a significant Delay x Stress interaction and a significant three-way interaction of Item type x Delay x Stress. Planned contrasts showed that this three-way interaction was significant for the contrast “Itemtype a2 versus a3” (my effect of interest) and not for the contrast a1 versus a2.
I now want to further decompose this three-way interaction. From the plots it looks as if in the immediate stress condition the memory effect is gone (recall of a3 not lower than recall of a2), while it is present in the immediate non-stress condition. In the delayed condition the pattern seems somewhat reversed with a stronger memory effect in the stress condition than in the no-stress condition, though the memory effect does not seem to be all gone in the delayed no-stress condition.
In an article (Decomposing interactions using GLM in combination with the COMPARE, LMATRIX and MMATRIX subcommands in SPSS) by Howell & Lacroix (2012) I read that the next step would be to look if the simple interaction effects are significant. So, I now want to calculate the Item type x Stress interaction at the 2nd and 3rd level of the Item type factor and at each level of the factor Delay. I want to use the LMATRIX and MMATRIX subcommands in the SPSS Syntax, but as I’m not familiar with the SPSS Syntax and with defining contrasts and I could not find an example corresponding to my design, I got stuck. I know, I need two LMATRIX subcommands (one for each level of “Delay”) and have to define one contrast in each of those subcommands and then I need one MMATRIX subcommand, but I don’t know what I have to write after the subcommands.

Here’s what I tried so far:

GLM a1 a2 a3 BY Delay Stress
/WSFACTOR=Itemtype 3 Repeated
/MEASURE=recall
/METHOD=SSTYPE(3)
/PLOT=PROFILE(Itemtype*Stress*Delay Itemtype*Delay*Stress)
/PRINT=DESCRIPTIVE
/CRITERIA=ALPHA(.05)
/WSDESIGN=Itemtype
/DESIGN=Delay Stress Delay*Stress
/LMATRIX =”no-stress versus stress in immediate condition”
Stress 1 -1
Delay*Stress 1 -1 0 0
/LMATRIX="no-stress versus stress in delayed condition"
Stress 1 -1
Delay*Stress 0 0 -1 1
/MMATRIX="a2 versus a3"ALL 0 1 -1.

I must have done something wrong, for SPSS calculates the first hypothesis test (it’s significant), but then it tells me that the 2nd LMATRIX cannot be estimated and the second hypothesis test cannot be calculated.
Can somebody please help me and tell me, what I should write in the Syntax editor to follow-up my three-way interaction. Thanks a lot in advance. Best,
Laura