Identical mixed models in SPSS and R nlme, with different degrees of freedom. Which to trust and why?

Question

I am analyzing a multilevel dataset with an AR(1) error structure and random intercept and slope. I fit what I believe is the exact same model in SPSS and R- my coefficients and standard errors are identical to the third decimal point, but my degrees of freedom are dramatically different. I have been able to find some information on how nlme calculates df, but not how SPSS does it. Given that these analyses are for a scientific paper and thus p-values matter (yes I know), I need to figure out what is going on and which DF to trust.

Here is my code for SPSS:

MIXED BF_lnLP_incr BY   moduleJustDone WITH 
    BF_RSPcl_incr timePoint
  /CRITERIA=CIN(95) MXITER(1000) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, 
    ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
  /FIXED=BF_RSPcl_incr moduleJustDone timePoint | SSTYPE(3)
  /METHOD=REML
  /PRINT=CORB COVB G  LMATRIX R SOLUTION TESTCOV
  /RANDOM=INTERCEPT timePoint  | SUBJECT(subjID) COVTYPE(VC)
  /REPEATED=timePoint | SUBJECT(subjID) COVTYPE(AR1).

Results for SPSS:

Type III Tests of Fixed Effects             
Source          Numerator df    Denominator df    F      Sig.
Intercept       1               725.420         15.262  .000
BF_RSPcl_incr   1               989.998         490.641 .000
moduleJustDone  3               876.414         1.334   .262
timePoint       1               413.480          .013   .909

Code for R:

lme(BF_lnLP_incr~as.factor(moduleJustDone)+timePoint+BF_RSPcl_incr,random=~timePoint|subjID,
          correlation = corAR1(,form=~timePoint|subjID),
          data=data,na.action=na.omit,method="REML")

Results for R:

                          numDF denDF  F-value p-value
(Intercept)                   1   712 299.8204  <.0001
as.factor(moduleJustDone)     3   712   6.9441  0.0001
timePoint                     1   712   5.2299  0.0225
BF_RSPcl_incr                 1   712 490.5761  <.0001