I fit three binary multilevel logistic regression models. Schematically they look like this:

```
model1 <- DEP.VAR ~ IND.VAR1 + (1|WORD)
model1 <- DEP.VAR ~ IND.VAR1 + IND.VAR2 + (1|WORD)
model1 <- DEP.VAR ~ IND.VAR1 + IND.VAR2 + IND.VAR2 + (1|WORD)
```

I calculated the ICC for each model. With each additional covariate, the ICC increases. Why? I don't understand why adding predictor variables should increase the amount of within-group homogeneity.