Table 1:

A small number of studies and real-world scenarios where each type of variability depicted in Figure 1 is observed.

Within-category noise variability on contrastive dimension	Non-contrastive variability is noise-like	Non-contrastive variability is patterned but not correlated	Non-contrastive variability is patterned + correlated
√ Galle, Apfelbaum, & McMurray, 2015	√ Rost & McMurray, 2009, 2010; √ Quam et al., 2017, Exp. 1	√ Barcroft & Sommers 2005, Exp. 2, moderate variability condition (3 speakers)*	∅ Quam et al. 2017, Exp. 2 √ Creel 2014a
Language input predominantly from one caregiver	Learning from a large number of talkers with similar speech cues (e.g., VOTs)	Learning from small number of caregivers with distinct voices but similar speech cues (e.g., VOTs)	Certain talkers use certain speech forms (e.g., Foulkes, Docherty, & Watt, 1999)

Note. Top row: Experiments where: √ variability helps; ∅ no effect of variability. Bottom row: possible situations where such variability might occur.

^*This study was conducted with adults learning second-language vocabulary. Three speakers facilitated learning relative to one speaker.