Table 4.

RMSE (and 95% CIs) between reconstructions from the reduced dataset (only using voxels with RFs) or from different versions of the layered IEM using the same voxels^{^a}

	Reduced data	p/s/a/b	p/a/b	p/s/b	s/a/b	p/a	s/a	p/s	p	a	s	None
Combined occipital V1–V4	0.133 (0.109, 0.170)	0.146 (0.146, 0.146)	0.148 (0.147, 0.148)	0.142 (0.141, 0.143)	0.194 (0.194, 0.194)	0.148 (0.148, 0.148)	0.194 (0.194, 0.194)	0.141 (0.141, 0.141)	0.143 (0.143, 0.143)	0.194 (0.194, 0.194)	0.194 (0.194, 0.194)	0.193 (0.193, 0.193)
Combined parietal V3A/B and IPS0	0.834 (0.744, 0.959)	0.415 (0.410, 0.419)	0.426 (0.422, 0.431)	0.411 (0.407, 0.416)	0.443 (0.439, 0.447)	0.425 (0.421, 0.430)	0.451 (0.447, 0.456)	0.410 (0.405, 0.416)	0.410 (0.406, 0.414)	0.441 (0.437, 0.445)	0.447 (0.442, 0.452)	0.834 (0.744, 0.959)

↵^aTo generate CIs, the resampling of the real data is performed at the level of the fits to the reconstructions, whereas resampling layered IEM RMSEs is described in Materials and Methods. Abbreviations for vRF changes included in each model: None, vRF parameters identical to fixation condition; a, amplitude; s, size; p, position; b.