Table 3.

Summary of the SP and MPT-Naming models’ naming response type predictions and parameter associations with independent behavioral data. The “predictive parameters” column lists the model parameters that were found to have the strongest association with the relevant data. All parameters were used when generating cross-validation predictions. For the multi-word repetition tests, the reported R² is from a multiple regression model using both parameters as predictors. The reported t values are standardized β coefficients from logistic regression models. LexForm, calculated from the MPT-Naming abilities, is the relative likelihood that a Formal error on a given naming trial originated from a failure at the lexical versus the phonological selection stage.

Language Measure	Prediction / Association Measure	SP Model	Predictive Parameters	MPT-Naming Model	Predictive Parameters
Picture Naming Response Types
Scale-level (n = 50, 7 categories)	Mean RMSE	2.03 items	All	1.89 items	All
Item-level (n = 18,200, 7-8 categories)	Accuracy	42.40%	All	67.51%	All
Semantic Association Test Scores
Synonymy Triplets (n = 127)	R2	.47	S	.51	LexSel
Peabody Picture Vocabulary Test (n = 127)	R2	.38	S	.47	LexSel
Camels and Cactus Test (n = 127)	R2	.42	S	.47	LexSel
Pyramids and Palm Trees Test (n = 75)	R2	.06	P	.19	LexSem
Speech Production Test Scores
Diadochokinetic Rate (n = 66)	R2	.18	P	.22	Phon
Nonword Repetition Test (n = 127)	R2	.35	P	.44	Phon
Philadelphia Repetition Test (n = 127)	R2	.26	P	.41	Phon
Immediate Serial Recall Span (n = 127)	R2	.49	S, P	.53	LexPhon, Phon
WAB Repetition Subtest (n = 88)	R2	.54	S, P	.70	LexPhon, Phon
Formal Error Grammatical Categories
Test-level rate of non-nouns (n = 83)	R	.32	S-P	.43	LexPhon
Item-level risk of non-noun (n = 1,504)	t	2.13	S-P	2.94	LexPhon