Supplementary Material Materials & Methods and Results

1Supplementary Material – Materials & Methods and Results 2 3 4 5

6Individual ant workers show self-control

7Stephanie Wendt1,2, Tomer J. Czaczkes1

81Zoology/ Evolutionary Biology, University of Regensburg, D-93053 Regensburg

112 Corresponding author, email: [email protected] 12

13Materials and Methods

14Study species

15Eight stock colonies of the black garden ant Lasius niger were collected on the University of Regensburg 16campus. The colonies were kept in 30x30x10cm foraging boxes with a layer of plaster covering the bottom. 17Each box contained a circular plaster nest box (14 cm diameter, 2 cm height). The colonies were queenless 18with around 1000-2000 workers and small amounts of brood. The colonies were fed three times a week with 19Bhatkar mix [1] and Drosophila fruit flies.

20One sub-colony of 500 individuals was formed from each stock colony, and these eight fixed-size sub- 21colonies were used for our experiments. Sub-colonies were maintained identically to the stock colonies. The 22sub-colonies were starved four days prior to the experiments in order to achieve a uniform and high 23motivation for foraging. Any ants which died or were removed from the sub-colonies were replaced from the 24original stock colonies. Water was available ad libitum.

25General Methods

26The aim of this series of experiments was to test whether hungry Lasius niger foragers can ignore a food 27source they would normally exploit, if they are aware of a higher quality food source elsewhere. To begin the 28experiment, a sub-colony was connected to an 80 x 1 cm long paper-covered runway via a 40 cm drawbridge 29(figure 1). A 5mm diameter drop of 1M sucrose (Sigma-Aldrich) was placed on an acetate feeder either at the 30end of the runway (120cm from the nest), or at 70cm (see below for treatment description). 2-4 ants were 31allowed onto the runway, and the first ant to reach the feeder was marked with a dot of acrylic paint on its 32gaster. The other ants were then returned to the nest. The marked ant was allowed to fill its crop and return 33to the nest (training run). In the nest, the ant unloaded her crop to her nestmates and was then allowed 34back onto the runway for a second visit. On the second visit, an additional 5mm sucrose droplet of varying 35molarity (1M, 0.75M or 0.25M depending on treatment) was placed in the middle of the runway at 60cm 36from the nest. The ant could not pass the new droplet without contacting it as the drop filled almost the 37entire width of the runway (about 0.5cm in diameter). However, the ant could bypass the droplet with ease 38if it chose to. Both food sources were marked with landmarks located beside the runway which the ants 39could use to aid location memorisation. Since Camponotus rufipes workers can travel from 0.5 to 9 km with 40the energy gained in a single foraging trip only, the costs to move between 120 and 60 cm are considered to 41be very minimal [2]. We then observed whether the ant drank at the near feeder, chose to skirt around it to 42reach the original feeder, or drank at both feeders. The ant was considered to have fed at a feeder if her 43mandibles were dipped into a drop for more than 1 second. Ants which fed on both feeders usually fed on 44the close feeder first and then moved on to the far feeder. Some ants again returned to the close feeder 45after that, but this proportion was very minimal. Again, the ant was allowed to drink and return to the nest 46when it had finished foraging. Shortly before the ant reached the nest, it was removed, to prevent pseudo 47replication. The average time between the training and testing visits was about three to seven minutes.

48Treatments

49Each ant experienced one of 4 possible treatments, which varied in the location of the first feeder and the 50molarity of the second feeder.

511M far vs 0.25M close – In this treatment the initial feeder (1M) was placed at the end of the runway (120 52cm from the nest). On its second visit, the ant encountered a 0.25M feeder 60 cm from the nest; that is, 60 53cm before the location of the original feeder. The aim of this treatment was to test whether ants can forego 54foraging on a relatively poor quality food source when they are aware of a higher quality food source further 55away. L. niger workers from colonies deprived of food for 4 days will readily drink and deposit pheromone to 560.25M sucrose [3], Oberhauser, Koch and Czaczkes in prep).

571M far vs 1M close – This treatment was identical to the first treatment, except that the new food source 58had an identical quality (1M) to the original food source. The aim of this treatment was to test whether ants 59that forego a close feeder in favour of a farther one do so because they are comparing the values of the two 60feeders, or whether ants, once they memorise a food location, ignore other food sources, for example due 61to behavioural momentum [4].

621M far vs 0.75M close – This treatment was identical to the first treatment, except that the second feeder 63provided 0.75M sucrose. The aim of this treatment was to test whether ants would forego a lower quality 64feeder if the quality difference was not very large.

651M 70cm close vs 0.75M 60cm close – After collecting data on treatments 1-3, we found that ants mostly 66chose the far, 1M quality feeder when the new feeder offered 0.25M, but mostly chose the new feeder 67when it offered 0.75M. To ascertain whether this was due to the ants not being able to reliably distinguish 680.75M from 1M, or whether it was due to an explicit choice of the nearer feeder to save energy and walking 69time, we carried out treatment 4. Here, the first (1M) feeder was placed only 70cm from the nest and the 70second feeder, offering 0.75M, was placed 60cm from the nest, as in treatment 3. This maintained the 71quality relationships used in treatment 3, but greatly reduced the distance difference.

72We tested 10 individuals from each colony on each treatment, for a total of 80 individuals.

73In an additional test, we repeated the ‘1M far vs 0.25M close’ and ‘1M far vs 1M close’ treatments, but 74allowed the ants to visit the far feeder twice before presenting the near feeder. Sample sizes in these 75experiments were lower (29 and 28 ants for the two treatments, respectively). 76Statistical Analysis

77Statistical analyses were carried out in R v. 3.3.2 [5] using Cumulative Link Mixed Models (CLMMs) in the 78“Ordinal” package [6]. We used CLMMs, because the response variable contains three ordered factors. Thus, 79an ordered linear regression was necessary. As multiple ants were tested per colony, colony identity was 80added as a random effect. The three decision codes (1 = far-away feeder, 0.5 = both feeders, 0 = close 81feeders) were brought into the model as ordered factors.

82We used the following model formulae:

83 Choice = DecisionCode

84 ~ treatment

85 + (random effects: colony)

86With treatment as a fixed effect.

87In addition to that, to explicitly disentangle the effects of food quality and distance to the nest, we 88performed post hoc pairwise comparisons between the four different treatments.

89CLMM Summary for decision code ~ treatment + (1 | colony)

90This model compares the choices made in the four different treatments 1M vs. 0.25M, 1M vs. 1M, 1M vs. 910.75M and 1M at 70cm vs. 0.75M when ants were allowed to visit the far-away feeder once before being 92confronted with a second, closer feeder.

treatment 1M vs. 0.25M 1M vs. 1M 1M vs. 0.75M 1M [70cm] vs. 0.75M 1M vs. 0.25M Estimate = 3.39 Estimate = 3.77 Estimate = 2.64 z = 8.55 z = 8.55 z = 7.21 p < 0.001 p < 0.001 p < 0.001 1M vs. 1M Estimate = Estimate = 1.06 Estimate = -0.75 -3.39 z = 1.53 z = -1.96 z = -8.14 p = 0.13 p = 0.0503 p < 0.001 1M vs. 0.75M Estimate = Estimate = Estimate = 1.13 -3.77 -1.06 z = 2.78 z = -8.55 z = -1.53 p < 0.01 p < 0.001 p = 0.17 1M [70cm] vs. Estimate = Estimate = 0.75 Estimate = 1.13 0.75M -2.64 z = 1.96 z = 2.78 z = -7.21 p = 0.0503 p < 0.01 p < 0.001 93Table S1: Estimates, z-values and p-values for the pairwise comparisons of all four treatments with one training visit before adding a 94second feeder.

95CLMM outputs of the pairwise comparisons for all four treatments with one training visit before testing

96Coefficients: 97 Estimate Std. Error z value Pr(>|z|) 98treatmenttest1-0.75/1 -3.7672 0.4406 -8.550 < 2e-16 *** 99treatmentt1-0.75/closer1 -2.6431 0.3666 -7.210 5.58e-13 *** 100treatmentnewgood1 -3.3895 0.4166 -8.137 4.06e-16 *** 101--- 102Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 103 104 105Coefficients: 106 Estimate Std. Error z value Pr(>|z|) 107treatmentt1-0.75/closer1 0.7464 0.3814 1.957 0.0503 . 108treatmenttest1-0.25/1 3.3895 0.4166 8.137 4.06e-16 *** 109treatmenttest1-0.75/1 -0.3777 0.4447 -0.849 0.3957 110--- 111Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 112 113 114Coefficients: 115 Estimate Std. Error z value Pr(>|z|) 116treatmentt1-0.75/closer1 1.1241 0.4048 2.777 0.00548 ** 117treatmentnewgood1 0.3777 0.4447 0.849 0.39572 118treatmenttest1-0.25/1 3.7672 0.4406 8.550 < 2e-16 *** 119--- 120Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 121 122 123Coefficients: 124 Estimate Std. Error z value Pr(>|z|) 125treatmentnewgood1 -0.7464 0.3814 -1.957 0.05035 . 126treatmenttest1-0.25/1 2.6431 0.3666 7.210 5.58e-13 *** 127treatmenttest1-0.75/1 -1.1241 0.4048 -2.777 0.00549 ** 128--- 129Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

131GLMM summary for decisionCode1 ~ treatment + (1 | colony)

132This model compares the choices for the far feeder made in the four different treatments 1M vs. 0.25M, 1M 133vs. 1M, 1M vs. 0.75M and 1M at 70cm vs. 0.75M when ants were allowed to visit the far-away feeder once 134before being confronted with a second, closer feeder.

treatment 1M vs. 0.25M 1M vs. 1M 1M vs. 0.75M 1M [70cm] vs. 0.75M 1M vs. 0.25M Estimate = 2.99 Estimate = 4.06 Estimate = 3.01 z = 6.59 z = 6.30 z = 6.6 p < 0.001 p < 0.001 p < 0.001 1M vs. 1M Estimate = Estimate = 1.06 Estimate = 0.02 -2.99 z = 1.53 z = 0.04 z = -6.59 p = 0.13 p = 0.971 p < 0.001 1M vs. 0.75M Estimate = Estimate = Estimate = -1.05 -4.06 -1.06 z = -1.5 z = -6.30 z = -1.53 p = 0.178 p < 0.001 p = 0.17 1M [70cm] vs. Estimate = Estimate = Estimate = 1.05 0.75M -3.01 -0.02 z = 1.5 z = -6.60 z = -0.04 p = 0.13 p < 0.001 p = 0.971 135Table S2: Estimates, z-values and p-values for the pairwise comparisons of all four treatments with one training visit before adding a 136second feeder.

137GLMM summary for decisionCode0.5 ~ treatment + (1 | colony)

138This model compares the choices for both feeders made in the four different treatments 1M vs. 0.25M, 1M 139vs. 1M, 1M vs. 0.75M and 1M at 70cm vs. 0.75M when ants were allowed to visit the far-away feeder once 140before being confronted with a second, closer feeder.

treatment 1M vs. 0.25M 1M vs. 1M 1M vs. 0.75M 1M [70cm] vs. 0.75M 1M vs. 0.25M Estimate = 0.48 Estimate = 0.17 Estimate = -0.87 z = 0.87 z = 0.34 z = -1.95 p = 0.51 p = 0.73 p = 0.051 1M vs. 1M Estimate = Estimate = Estimate = -1.35 -0.48 -0.31 z = -2.69 z = -0.87 z = -0.55 p < 0.05 p = 0.51 p = 0.73 1M vs. 0.75M Estimate = Estimate = 0.31 Estimate = -1.05 -0.17 z = 0.55 z = -2.27 z = -0.34 p = 0.58 p < 0.05 p = 0.73 1M [70cm] vs. Estimate = 0.87 Estimate = 1.35 Estimate = 1.05 0.75M z = 1.95 z = 2.69 z = 2.27 p = 0.1 p < 0.05 p < 0.05 141Table S3: Estimates, z-values and p-values for the pairwise comparisons of all four treatments with one training visit before adding a 142second feeder. 143GLMM summary for decisionCode0 ~ treatment + (1 | colony)

144This model compares the choices for the close feeder made in the four different treatments 1M vs. 0.25M, 1451M vs. 1M, 1M vs. 0.75M and 1M at 70cm vs. 0.75M when ants were allowed to visit the far-away feeder 146once before being confronted with a second, closer feeder.

treatment 1M vs. 0.25M 1M vs. 1M 1M vs. 0.75M 1M [70cm] vs. 0.75M 1M vs. 0.25M Estimate = 3.24 Estimate = 3.56 Estimate = 2.35 z = 7.16 z = 7.51 z = 5.71 p < 0.001 p < 0.001 p < 0.001 1M vs. 1M Estimate = Estimate = 0.32 Estimate = -0.89 -3.24 z = 0.71 z = -2.27 z = -7.16 p = 0.48 p < 0.05 p < 0.001 1M vs. 0.75M Estimate = Estimate = Estimate = -1.20 -3.56 -0.32 z = -2.91 z = -7.51 z =- 0.71 p < 0.01 p < 0.001 p = 0.48 1M [70cm] vs. Estimate = Estimate = 0.89 Estimate = 1.20 0.75M -2.35 z = 2.27 z = 2.91 z = -5.71 p < 0.05 p < 0.01 p < 0.001 147Table S4: Estimates, z-values and p-values for the pairwise comparisons of all four treatments with one training visit before adding a 148second feeder.

150 151 152Data for two training visits before testing self-control (Figure S2)

153Methods for this data were the same as those provided in the main text. The difference here is that ants 154were allowed to visit the far feeder twice before they were confronted with a second, closer feeder. In the 155data shown in figure 2 in the main text, ants were allowed to visit the far-away feeder only once. 156

157Figure S2: The proportions of decisions made for the different feeders for all treatments with two training visits before testing self- 158control instead of one training visit. The close feeder was located at a distance of 60 cm from the nest while the far-away feeder was 159located at 120 cm distance to the nest. Treatment is named as molarity of the far-away feeder vs. molarity of the close feeder. 160Sample sizes are as follows: 1M vs. 0.25M: n= 29; 1M vs. 1M: n= 28. Decisions in 1M vs. 0.25M were significantly different from the 161other treatment. 162Comparison between one training visit and two training visits

163This model compares the decisions made by ants which were allowed to visit the far feeder once (1M vs. 1M 164= treatment newgood1 and 0.25M vs. 1M = treatment test1-0.25/1) and ants which were allowed to visit the 165far feeder twice (1M vs. 1M = treatment newgood 2 and 0.25M vs. 1M = treatment test 2) before 166confronting them with a second, closer feeder.

167Coefficients: 168 Estimate Std. Error z value Pr(>|z|) 169treatmentnewgood1 -0.8563 0.5184 -1.652 0.0986 . 170treatmenttest 2 4.8248 1.1419 4.225 2.39e-05 *** 171treatmenttest1-0.25/1 2.2756 0.5542 4.106 4.03e-05 *** 172--- 173Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 174 175 176Coefficients: 177 Estimate Std. Error z value Pr(>|z|) 178treatmenttest1-0.25/1 -2.549 1.062 -2.399 0.0164 * 179treatmentnewgood1 -5.681 1.086 -5.231 1.68e-07 *** 180treatmentnewgood 2 -4.825 1.142 -4.225 2.39e-05 *** 181--- 182Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

183When ants received two training visits before testing self-control, there are significantly more choices for the 184far feeder in the 0.25M vs. 1M treatment than in the 1M vs. 1M treatment (Z = 4.225, P < 0.001). 185There is no significant difference between the 1M vs. 1M treatment with one visit before testing and with 186two visits before testing. However, in treatment 0.25M vs. 1M, there are significantly more choices for the 187far feeder when ants were allowed to feed on the far feeder twice before testing self-control (Z = 2.399, P < 1880.05). This shows that ants show even better self-control when they are more experienced with a food 189source due to more visits before testing.

190References

1911. Bhatkar A, Whitcomb WH. 1970 Artificial Diet for Rearing Various Species of Ants. Fla. Entomol. 53, 229– 192 232. (doi:10.2307/3493193)

1932. Schilman PE, Roces F. 2006 Foraging energetics of a nectar-feeding ant: metabolic expenditure as a 194 function of food-source profitability. J. Exp. Biol. 209, 4091–4101. (doi:10.1242/jeb.02478)

1953. Detrain C, Prieur J. 2014 Sensitivity and feeding efficiency of the black garden ant Lasius niger to sugar 196 resources. J. Insect Physiol. 64, 74–80. (doi:10.1016/j.jinsphys.2014.03.010)

1974. Podlesnik CA, Jimenez-Gomez C. 2016 Contrafreeloading, reinforcement rate, and behavioral momentum. 198 Behav. Processes 128, 24–28. (doi:10.1016/j.beproc.2016.03.022)

1995. R Core Team. 2016 R: A language and environment for statistical computing. Vienna, Austria: R 200 Foundation for Statistical Computing. See https://www.R-project.org.

2016. Christensen RHB. 2015 ordinal: Regression Models for Ordinal Data. R package version 2015.6-28.