Table 6.

Multiple linear regression models predicting item difficulty parameters from lexical properties. The linear coefficient associated with each predictor in each of the regression models is given in the corresponding cell. The strongest simple linear predictor is shaded. Item difficulty parameters are measured on a logit scale; the Word-T parameter is measured as a probability.

		Intercept	Lexical Property			Model fit
			LexFreq	PhonLeng	PhonDens	df	rmse	simple R2
Item Difficulty	Attempt	−0.04	−0.15		−0.24	172	.59	.29
	LexSem	−1.03	−0.14			173	.74	.05
	LexPhon	−0.49	−0.24			173	.47	.29
	LexSel	−2.04	−0.17	+1.03		172	.79	.28
	Phon	−0.26	−0.19	+0.42	−0.32	171	.41	.66
	Word−T	0.01			+0.10	173	.13	.47