Prompt (P)			Error correction (E)			Reinforcement (R)
	Trial t			Trial t			Trial t
	R	Not R		R	Not R		R	Not R
Trial t - 1
P	a	b	E	a	b	R	a	b
Not P	c	d	Not E	c	d	Not R	c	d

where a, b, c and d represent the lag 1 tallies of the two-event sequences. Odds ratios for each contingency table were then computed using the standard formula: Odds ratio = ad/bc. 95 % confidence intervals were also computed (Bakeman and Quera 2011). We thus had three contingency indices a) the odds of receiving reinforcement on trial t given a prompt on the previous trial t - 1, (b), the odds of receiving reinforcement on trial t given error correction on the previous trial t - 1, and (c), the odds of receiving reinforcement on trial t given reinforcement on the previous trial t - 1. For each contingency, an odds ratio significantly >1.0 would indicate that participants would be more likely to make a correct response on a trial given that a particular contingency had occurred on the previous trial than by chance. Similarly, for each contingency, an odds ratio significantly lower than 1.0 would indicate that participants would be less likely to make a correct response on a trial given that a particular contingency had occurred than by chance. All other analyses were conducted using SPSS version 19 (SPSS, Inc.)