Table 1.

Overview of critical issues, key issues, alternative viewpoints, working hypotheses, and simulated effects/model explorations

Critical issues	Key issues	Alternative viewpoints	Working hypotheses	Simulated effects/explorations
Lateralization assumptions	How can the language network be strongly left lateralized in patients but be bilateral, albeit asymmetric, in healthy participants?	1) Impaired language function after left but not right hemisphere damage (14–16)	Leftward hemispheric asymmetry generates bilateral yet asymmetric lateralization in simulated BOLD and also greater likelihood of chronic impairment after left than right damage	1) Functional asymmetry followed computational capacity 2) A function-structure pattern (42) 3) Impaired performance after left but not right damage to the model (17)
		2) Bilateral and asymmetric brain activations in healthy individuals during language tasks (21–29)
The computational bases of language recovery	A lack of an implemented model for the computational bases of language recovery	1) Degeneracy (13, 30)	The mechanisms are not mutually exclusive and they can be utilized as a part of the recovery process	Various analyses on model behavior and explorations of the underpinning computations in both damage and undamaged conditions
		2) Variable neurodisplacement (13)
Theories of aphasia recovery	1) What are the dynamic activation shifts in poststroke aphasia and recovery?	1) Perilesional up-regulated activation (5, 49–54) 2) The “regional hierarchy framework” (5–9)	Patterns of recovery are related to the differential capacity available in left and right hemisphere systems and lesion severity	1) Dynamic patterns of activation shifts during recovery (50) 2) Lesion severity as a determiner for recovered performance and brain activation patterns 3) Different types of interconnectivity
	2) What is the effect of transcallosal connectivity on healthy and impaired function?	3) Right hemisphere activation (2, 8, 10–12, 50, 55, 56)
Multiple measures	What is the relationship between multiple brain measures and recovered function in patients?	1) Behavioral measures compared with fMRI activation (8, 10–12, 52)	Different measures provide different types of information	1) Model accuracy better tracked by the RSA than unit activation (cf. BOLD) 2) A conceptually similar RSA pattern to data in ref. 71
		2) Potential applications of multiple voxel pattern analysis (71, 72) in the patient studies