TABLE 2.

Model comparison for models predicting pen-movement disfluency (velocity peak count) in the character copying task.

	Fixed factor(s) added	X2, df, p
Model 1	Feature malformed (vs. correct)	100, 1, <0.001
Model 2	Target feature is a curve (vs. straight line)	420, 1, <0.001
Model 3	Character is a letter (vs. symbol)	2.2, 1, 0.136
Model 3a	(Child age)	0.09, 1, 0.77
Model 4	Pen-control measures	52, 4, <0.001
Model 5	Interactions between pen-control measures and whether the target feature is a curve	35, 4, <0.001
Model 6	Interactions between pen-control measures and whether the character is a letter	13, 4, 0.013
Model 7	Letter-knowledge measures	10, 3, 0.017
Model 8	Interactions between literacy-ability measures and whether the character is a letter	4.2, 4, 0.24

Models were nested, with all fixed factors added at a particular stage carried forward to subsequent model, with the exception of age (Model 3a). Model 1 was compared with an intercept-only model. Model 7 is the best fit model.