Table 2.

Pearson correlation and confidence intervals for correlation differences Δr between unimodal and multimodal visual models. The top part of the table shows the results for the distributional linguistic model, whereas the bottom part shows the results for the word association model

Dataset	$n$	va = Distributional linguistic, vb = Visual
		Unimodal				Multimodal
		$r_{v_a}$	CI95	$r_{v_b}$	CI95	β	$r_{v_{\mathit{ab}}}$	CI95	Δr	CI95
MEN	942	.79	[0.77, 0.82]	.66	[0.62, .70]	0.38	.82	[0.80, 0.84]	.03	[0.02, 0.04]
MTURK‐771	260	.67	[0.59, 0.73]	.49	[0.39, 0.58]	0.38	.71	[0.64, 0.76]	.04	[0.01, 0.08]
SimLex‐999	300	.43	[0.33, 0.52]	.54	[0.45, 0.61]	0.55	.56	[0.48, 0.64]	.14	[0.07, 0.21]
Silberer (Sem.)	5,799	.73	[0.71, 0.74]	.78	[0.77, 0.79]	0.53	.82	[0.82, 0.83]	.10	[0.09, 0.10]
Silberer (Vis.)	5,777	.59	[0.57, 0.61]	.74	[0.73, 0.75]	0.65	.75	[0.74, 0.76]	.16	[0.15, 0.17]
Average		.55		.55			.64		.08

Dataset	$n$	va = Word association, vb = Visual
		Unimodal				Multimodal
		$r_{v_a}$	CI95	$r_{v_b}$	CI95	β	$r_{v_{\mathit{ab}}}$	CI95	Δr	CI95
MEN	942	.79	[0.76, 0.81]	.66	[0.62, 0.70]	0.45	.81	[0.78, 0.83]	.02	[0.01, 0.02]
MTURK‐771	260	.74	[0.68, 0.79]	.49	[0.39, 0.58]	0.33	.75	[0.69, 0.79]	.00	[−0.01, 0.02]
SimLex‐999	300	.72	[0.66, 0.77]	.54	[0.45, 0.61]	0.43	.73	[0.68, 0.79]	.01	[−0.01, 0.03]
Silberer (Sem.)	5,799	.84	[0.83, 0.85]	.78	[0.77, 0.79]	0.58	.87	[0.86, 0.88]	.03	[0.03, 0.04]
Silberer (Vis.)	5,777	.73	[0.72, 0.74]	.74	[0.73, 0.75]	0.65	.79	[0.78, 0.80]	.06	[0.05, 0.07]
Average		.71		.55			.73		.02