SENSORY PRECONDITIONING Avoidance but not aversion following sensory preconditioning with flavors: A challenge to stimulus substitution Supplementary Data: From pre-exposure, aversion training, and test expressed as a function of solution type. D.M. Dwyer, K.V. Burgess, & R.C. Honey Cardiff University Short Title: SENSORY PRECONDITIONING Submitted in May 2012 to: Journal of Experimental Psychology: Animal Behavior Processes Address for correspondence: School of Psychology Cardiff University Tower Building, Park Place Cardiff CF10 3AT, UK Tel: +44 (0)29 2087 6285 Email: DwyerDM@cardiff.ac.uk, BurgessK1@cardiff.ac.uk, Honey@cardiff.ac.uk 1 SENSORY PRECONDITIONING Experiment 1 Data from exposure (AB and CD pairing trials) and aversion training (A+ and Ctrials) are shown in Supplementary Table 1. The exposure data (consumption and lick cluster size separately) were subjected to mixed ANOVAs with a within-subject factor of trial type (AB or CD) and a between-subjects factor of solution assignment (AC = Suc/Malt or AC = Salt/Lem). For consumption, there were no significant main effects or interactions (all Fs < 1). For lick cluster size, there were also no significant main effects or interactions (largest F(1,22) = 1.75, p = .199, MSE = 84.61, for the interaction). The aversion training data were also subjected to mixed ANOVAs with a withinsubject factor of trial type (A+ or C-) and a between subjects factor of solution assignment (AC = Suc/Malt or AC = Salt/Lem). For consumption, there was a significant effect of solution assigment, F(1,22) = 17.57, p < 0.001, MSE = 5.40, reflecting lower consumption when the A/C solutions came from the salt and lemon pair rather than the sucrose and maltodextrin pair. There was no main effect of trial type, or interaction (largest F(1,22) = 1.03, p = .321, MSE = 13.68, for main effect of trial type). For lick cluster size, were no significant main effects or interactions (largest F(1,22) = 1.49, p = .235, MSE = 214.51, for the interaction). Data from the test trials with stimuli A-D, split as factor of whether A and C were sucrose and maltodextrin or salt and lemon, are shown in Supplementary Table 1. These data were subjected to mixed ANOVAs with within-subject factors of aversion condition (i.e. whether the stimuli should be subject to an aversion or not: A & B versus C & D) and training type (direct conditioning or sensory preconditioning: A & C versus B & D), plus a between-subjects factor of solution assignment (AC = Suc/Malt or AC = Salt/Lem). The main manuscript reports the analysis of this data collapsed over the factor of solution assignment (see data in Figure 1 and inferential analyses in the result section of Experiment 2 SENSORY PRECONDITIONING 1), thus only the effects of solution assignment and its interactions will be considered here. For consumption, the the interaction between solution assignment and training type was significant, F(1,22) = 25.57, p < .001, MSE = 0.91, but no other main effects or interactions involving the solution assignment factor were significant (all Fs < 1). The significant interaction reflects that fact that less of the A and C solutions was consumed than of the B and D solutions, but that this difference was smaller when A and C were sucrose and maltodextrin. There were no significant effects or interactions involving the solution assignment factor for the analysis of lick cluster size (Fs < 1). In summary, the counterbalancing factor of whether the solutions sucrose and maltodextrin, or lemon and salt, were assigned to be stimuli A and C or stimuli B and D influenced the amount of A and C consumed during aversion training, and the amount of A and C consumed relative to B and D during test. Critically, it did not interact with aversion conditioning in any way and so there is no suggestion that it could have influenced the analysis of aversions produced by direct or sensory preconditioning as described in the main manuscript. The absence of any effects of solution assigment on the lick cluster measure suggests that, possibly due to due to water deprivation enhancing the value of all fluids, unconditioned differences between solution types had relatively little impact on this factor. Experiment 2 Data from exposure (AB and CD pairing trials) and aversion training (A+ and C- trials) are shown in Supplementary Table 2. The exposure data (consumption and lick cluster size separately) were subjected to mixed ANOVAs with a within-subject factor of trial type (AB or CD) and between-subjects factors of LiCl strength (Low or High) and of solution assignment (AC = Suc/Malt or AC = Salt/Lem). For consumption, the main effect of trial type approached standard levels of significance, F(1,44) = 2.93, p = .094, MSE = 6.22, 3 SENSORY PRECONDITIONING reflecting the numerically greater consumption during AB than CD trials. No other main effects or interactions were significant (largest F(1,44) = 1.40, p = .243, MSE = 6.22, for the trial type by solution assignment interaction). For lick cluster size, there was a significant effect of LiCl strength, F(1,44) = 4.24, p = .045, MSE = 112.16, reflecting the fact that rats in the group to be injected with high doses of LiCl had lower lick cluster sizes than rats in the group to be injected with low doses of LiCl. The effect of solution assignment approached significance, F(1,44) = 2.84, p = .090, MSE = 112.16, reflecting the fact that rats in the AC = Suc/Malt sub-group showed numerically lower lick cluster sizes. No other main effects or interactions were significant (largest F(1,44) = 1.80, p = .186, MSE = 88.19, for the trial type by solution assignment interaction). The aversion training data were also subjected to mixed ANOVAs with a within-subject factor of trial type (A+ or C-) and between-subjects factors of LiCl strength (Low or High) and of solution assignment (AC = Suc/Malt or AC = Salt/Lem). For consumption, there was a significant effect of solution assigment, F(1,44) = 27.96, p < 0.001, MSE = 6.71, reflecting lower consumption when the A/C solutions came from the salt and lemon pair rather than the sucrose and maltodextrin pair. There were no other significant main effects or interactions (largest F(1,44) = 1.74, p = .194, MSE = 6.71, for the LiCl strength by solution assignment interaction). For lick cluster size, there were no significant main effects or interactions (largest F(1,44) = 2.49, p = .122, MSE = 118.06, for the main effect of LiCl strength). Data from the test trials with stimuli A-D, split as factor of whether A and C were sucrose and maltodextrin or salt and lemon, is shown in Supplementary Table 2. These data were subjected to mixed ANOVAs with within-subject factors of aversion condition (i.e. whether the stimuli should be subject to an aversion or not: A & B versus C & D) and training type (direct conditioning or sensory preconditioning: A & C versus B & D), plus between subjects factors of LiCl strength (Low or High) and solution assignment (AC = 4 SENSORY PRECONDITIONING Suc/Malt or AC = Salt/Lem). Again, the main manuscript reports the analysis of these data collapsed over the factor of solution assignment (see data in Figure 2 and inferential analyses in the result section of Experiment 2), thus only the effects of solution assignment and its interactions will be considered here. As in Experiment 1, for consumption, the interaction between solution assignment and training type was significant, F(1,44) = 37.60, p < .001, MSE = 3.60, but no other main effects or interactions involving the solution assignment factor were significant (largest F(1,44) = 1.69, p = .201, MSE = 10.40, for the main effect of solution assignment). Again, the significant interaction reflects that fact that less of the A and C solutions was consumed than of the B and D solutions, but that this difference was smaller when A and C were sucrose and maltodextrin. For the analysis of lick cluster size data, there was a signifcant interaction between LiCl strength and solution assignment F(1,44) = 6.66, p = .013, MSE = 350.85 (reflecting the fact that for the Low LiCl group the AC = Suc/Malt group had numerically lower lick cluster sizes than the AC = Salt/Lem group and that this was reversed for the High LiCl group), and also a significant 3-way interaction between solution assignment, aversion condition and training type, F(1,44) = 5.06, p = .030, MSE = 187.12. This reflected a modulation of the 2-way interaction between aversion condition and training type, whereby difference between cluster sizes for A and C was smaller for the AC = Suc/Malt than the AC = Salt/Lem group (though still significant in both cases, F(1,44) = 11.79, p < .001, MSE = 16.13 and F(1,44) = 48.77, p < .001, MSE = 16.13, respectively) and the differences between cluster sizes for B and D was larger for the AC = Suc/Malt than the AC = Salt/Lem group (though not approaching standard levels of significance in either case, largest F(1,44) = 1.08, p = .303, MSE = 13.82). In summary, the counterbalancing factor of whether the solutions sucrose and maltodextrin, or lemon and salt, were assigned to be stimuli A and C or stimuli B and D influenced the amount of A and C consumed during aversion training, and the amount of A 5 SENSORY PRECONDITIONING and C consumed relative to B and D during test. Critically, it did not interact with aversion conditioning in any way. Although there were effects of solution assignment on lick cluster size in this experiment, the critical result whereby there was an effect of this measure for direct conditioning, but not sensory preconditioning, was present in both of the counterbalancing subgroups. Again, there is no suggestion that the assignment of stimulus conditions to flavour stimuli could have influenced the analysis of aversions produced by direct or sensory preconditioning, as described in the main manuscript. Experiment 3 Data from exposure (AB and CD pairing trials) and aversion training (A+ and C- trials) are shown in Supplementary Table 3. The exposure data (consumption and lick cluster size separately) were subjected to mixed ANOVAs with a within-subject factor of trial type (AB or CD) and between-subjects factors of LiCl delivery (immediate or trace), and of solution assignment (AC = Suc/Malt or AC = Salt/Lem). For consumption, there were no significant main effects or interactions (largest F(1,28) = 1.05, p = .315, MSE = 11.69, for the trial type by LiCl delivery interaction). For lick cluster size, no main effects or interactions were significant (largest F(1,28) = 2.67, p = .113, MSE = 108.72, for the trial type by solution assignment interaction). The aversion training data were also subjected to mixed ANOVAs with a within-subject factor of trial type (A+ or C-) and between-subjects factors of LiCl delivery (immediate or trace) and of solution assignment (AC = Suc/Malt or AC = Salt/Lem). For consumption, the effect of solution assigment approached conventional levels of statisitcal significance, F(1,28) = 3.96, p = 0.057, MSE = 8.33, and there was a signifcant interaction between solution assignment and LiCl deivery, F(1,28) = 7.20, p = 0.012, MSE = 8.33. This reflected lower consumption when the A/C solutions came from the salt and lemon pair rather than the 6 SENSORY PRECONDITIONING sucrose and maltodextrin pair in the trace group but not the immediate group. There were no other significant main effects or interactions (Fs < 1). For lick cluster size, there were no significant main effects or interactions (largest F(1,28) = 2.87, p = .101, MSE = 452.00, for the three way interaction between trial type, LiCl delivery, and solution assignment). Data from the test trials with stimuli A-D, split as factor of whether A and C were sucrose and maltodextrin, or salt and lemon, are shown in Supplementary Table 3. These data were subjected to mixed ANOVAs with within-subject factors of aversion condition (i.e. whether the stimuli should be subject to an aversion or not: A & B versus C & D) and training type (direct conditioning or sensory preconditioning: A & C versus B & D), plus the between-subjects factors of LiCl delivery (immediate or trace) and solution assignment (AC = Suc/Malt or AC = Salt/Lem). Again, the main manuscript reports the analysis of this data collapsed over the factor of solution assignment (see data in Figure 3 and inferential analyses in the result section of Experiment 3), thus only the effects of solution assignment and its interactions will be considered here. For consumption, the interaction between solution assignment and training type was significant, F(1,28) = 13.62, p = .001, MSE = 3.66, but no other main effects or interactions involving the solution assignment factor were significant (largest F(1,28) = 2.23, p = .146, MSE = 13.52, for the interaction between LiCl delivery and solution assignment). As in the previous experiments, the significant interaction reflects that fact that less of the A and C solutions was consumed than of the B and D solutions, but that this difference was smaller when A and C were sucrose and maltodextrin. There were no significant main effects or an interaction involving the factor of solution assignment for the analysis of lick cluster size (Fs < 1). In summary, the counterbalancing factor of whether the solutions sucrose and maltodextrin, or lemon and salt, were assigned to be stimuli A and C or stimuli B and D influenced the amount of A and C consumed during aversion training, and the amount of A 7 SENSORY PRECONDITIONING and C consumed relative to B and D during test but did not interact with aversion conditioning in any way. Again, there is no suggestion that the assignment of stimulus conditions to flavour stimuli could have influenced the analysis of aversions produced by direct or sensory preconditioning that were described in the main manuscript. 8 SENSORY PRECONDITIONING 9 SENSORY PRECONDITIONING Supplementary Table 1: Experiment 1 AB CD A+ C- Exposure & Aversion Training Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size AC = Suc/Malt 9.3 (0.8) 28.6 (1.9) 8.7 (0.6) 29.5 (2.8) 11.6 (0.6) 33.1 (3.0) 12.7 (0.4) 41.4 (3.3) AC = Salt/Lem 8.9 (0.8) 31.2 (3.9) 8.5 (0.6) 25.1 (2.3) 8.8 (1.1) 33.5 (3.6) 9.9 (1.2) 35.5 (5.5) A B C D Test Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size AC = Suc/Malt 2.5 (0.4) 16.6 (1.8) 7.8 (1.1) 36.4 (5.4) 12.4 (0.7) 44.8 (3.7) 10.5 (1.4) 41.5 (5.9) AC = Salt/Lem 2.5 (0.7) 16.2 (3.8) 9.4 (0.5) 31.0 (3.3) 11.3 (1.0) 42.6 (5.6) 11.6 (0.5) 38.9 (4.3) Note: A-D refer to the test solutions (see Table 1 of the main manuscript for the exact design). AB and CD refer to the data averaged across the three AB and CD exposure sessions. A+ and C- refer to the data from the aversion training sessions (where LiCl was injected following consumption of A but not C). A, B, C, and D refer to data from the test sessions. AC = Suc/Malt refers to the solution counterbalancing where animals received sucrose and maltodextrin as either A or C (and thus received salt and lemon as B or D). AC = Salt/Lem refers to the solution counterbalancing where animals received salt and lemon as either A or C (and thus received sucrose and maltodextrin as B or D). 10 SENSORY PRECONDITIONING Supplementary Table 2: Experiment 2 AB CD A+ C- Exposure & Aversion Training Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size AC = Suc/Malt 10.6 (0.4) 35.9 (3.3) 9.3 (0.6) 32.5 (2.9) 11.8 (1.1) 34.6 (3.9) 11.5 (0.5) 34.2 (2.7) AC = Salt/Lem 9.7 (0.4) 36.8 (3.9) 9.3 (0.7) 34.0 (3.2) 8.3 (1.2) 29.6 (3.8) 7.8 (0.9) 28.4 (3.1) AC = Suc/Malt 10.4 (0.6) 27.3 (2.3) 8.8 (0.9) 27.3 (2.1) 11.2 (0.5) 39.5 (3.7) 10.4 (0.5) 29.8 (2.0) AC = Salt/Lem 9.9 (0.6) 31.4 (1.6) 9.7 (0.7) 35.4 (3.1) 8.8 (1.3) 35.0 (3.0) 8.7 (1.0) 36.5 (5.4) Low LiCl High LiCl A B C D Test Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size AC = Suc/Malt 5.6 (1.0) 28.6 (2.9) 8.0 (0.9) 38.1 (5.2) 10.6 (0.9) 39.7 (4.2) 9.1 (1.0) 42.6 (4.0) AC = Salt/Lem 3.8 (1.1) 17.8 (3.9) 10.1 (0.8) 29.4 (1.8) 10.5 (1.0) 36.7 (2.9) 10.9 (0.6) 31.5 (2.1) AC = Suc/Malt 4.5 (1.0) 20.0 (4.0) 7.0 (1.2) 33.7 (4.9) 10.1 (0.5) 36.5 (3.0) 8.3 (0.7) 36.9 (4.1) AC = Salt/Lem 2.6 (0.8) 13.2 (2.5) 8.4 (1.1) 43.5 (5.6) 9.7 (0.8) 50.5 (7.5) 12.1 (0.6) 42.1 (4.8) Low LiCl High LiCl 11 SENSORY PRECONDITIONING Note: A-D refer to the test solutions while Low and High refer to the LiCl doses used (see Table 1 of the main manuscript for the exact design). AB and CD refer to the data averaged across the three AB and CD exposure sessions. A+ and C- refer to the data from the aversion training sessions (where LiCl was injected following consumption of A but not C). A, B, C, and D refer to data from the test sessions. AC = Suc/Malt refers to the solution counterbalancing where animals received sucrose and maltodextrin as either A or C (and thus received salt and lemon as B or D). AC = Salt/Lem refers to the solution counterbalancing where animals received salt and lemon as either A or C (and thus received sucrose and maltodextrin as B or D). 12 SENSORY PRECONDITIONING Supplementary Table 3: Experiment 3 AB CD A+ C- Exposure & Aversion Training Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size AC = Suc/Malt 9.7 (1.0) 28.2 (4.6) 9.1 (0.7) 30.1 (3.8) 11.1 (1.9) 49.5 (6.1) 11.6 (1.2) 37.7 (6.1) AC = Salt/Lem 10.9 (1.2) 24.7 (4.0) 9.6 (1.3) 33.5 (3.9) 11.8 (2.0) 38.0 (4.5) 11.8 (2.1) 55.2 (14.4) AC = Suc/Malt 9.4 (0.9) 28.9 (8.0) 10.9 (1.0) 25.4 (3.4) 14.1 (0.8) 42.1 (5.7) 13.5 (0.6) 37.0 (8.2) AC = Salt/Lem 9.6 (0.9) 27.7 (3.3) 9.9 (1.4) 34.5 (6.1) 10.1 (1.8) 47.4 (9.0) 10.7 (1.7) 35.3 (6.8) Immediate Trace A B C D Test Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size Consumption Cluster Size AC = Suc/Malt 2.3 (0.5) 10.1 (2.1) 7.5 (1.4) 40.4 (7.5) 12.5 (0.9) 45.9 (5.6) 9.6 (1.4) 38.4 (4.8) AC = Salt/Lem 2.9 (0.9) 15.0 (3.3) 9.7 (0.9) 39.6 (4.7) 12.4 (2.2) 51.0 (8.3) 12.8 (1.2) 37.7 (5.2) AC = Suc/Malt 4.9 (1.0) 14.9 (5.2) 8.5 (1.9) 44.1 (12.7) 14.7 (1.5) 40.5 (6.5) 12.1 (1.9) 33.9 (5.3) AC = Salt/Lem 5.1 (1.7) 15.5 (5.5) 10.2 (0.6) 38.7 (8.9) 11.0 (1.1) 38.0 (5.2) 12.0 (0.7) 32.3 (6.1) Immediate Trace 13 SENSORY PRECONDITIONING Note: A-D refer to the test solutions while Immediate and Trace refer to the interval between the presentation of A and injection with LiCl (see Table 1 of the main manuscript for the exact design). AB and CD refer to the data averaged across the three AB and CD exposure sessions. A+ and C- refer to the data from the aversion training sessions (where LiCl was injected following consumption of A but not C). A, B, C, and D refer to data from the test sessions. AC = Suc/Malt refers to the solution counterbalancing where animals received sucrose and maltodextrin as either A or C (and thus received salt and lemon as B or D). AC = Salt/Lem refers to the solution counterbalancing where animals received salt and lemon as either A or C (and thus received sucrose and maltodextrin as B or D). 14