Table 4.

Contingency tables comparing inference methods for identifying significant improvement in response to 3 weeks of aphasia therapy, with thresholds for significance shown in parentheses.

Threshold 1	Status	SCANT (4 items)
		Unimproved	Improved
PNT (p < .07)	Unimproved	41	2
	Improved	5	4
		SCANT (3 items)
		Unimproved	Improved
PNT (p < .17)	Unimproved	40	4
	Improved	6	6
		SCANT (2 items)
		Unimproved	Improved
PNT (p < .40)	Unimproved	36	7
	Improved	9	11

Note. Participants were randomly assigned to semantically oriented or phonologically oriented therapy; the test items were not treated for naming. The inferences based on SCANT scores relied on a fixed threshold of change out of 20 items. The inferences based on PNT scores relied on a Fisher's exact test comparing frequencies of correct responses out of 174 items, with alpha criteria for significance matched to the corresponding SCANT threshold's test–retest percentile score to approximate the same purported Type I error rate. SCANT = Severity-Calibrated Aphasia Naming Test; PNT = Philadelphia Naming Test.