During decision−making, animals consider not only the current but also the past quality of options. For example, when humans evaluate performance (e.g. sales) of employees, they do not only consider the average performance but also the trend of performance　ascending performance is often viewed as more favorable than descending performance. In our study, we test if non-human animals have a similar bias when they are evaluating options using house-hunting by the acorn ant, Temnothorax curvispinosus, as our model system. Our data show that when nest-site quality is static over time, ant colonies tend to prefer the nest site which was better (i.e. darker) between two nest options. However, when the nest quality changes—one improves and the other worsens—over time, more colonies choose the low-quality, but improving, nest than the high-quality, but worsening, nest. These results suggest that a continuous change of option quality may influence evaluation. We discuss alternative explanations for our results, possible mechanisms and potential ecological benefits for keeping track of the nest-site quality.

Experimental procedure

Prior to the experiment, we induced a colony emigration to an empty home nest placed in the middle of the experimental arena (19 x 27 cm) (see Sasaki and Pratt 2018 for the detail of the emigration procedure). After 24 hours, we introduced two kinds of potential target nests, namely an “improving” nest and a “worsening” nest, one on each side (Figure 1a). These target nests were identical to the home nest except the interior light level, which was controlled by putting light filters on the roof (see Supplementary Information for the detail of the nest design; Figure S1). The home nest was always darker than target nests to prevent early migration. The improving nest initially had a very bright interior light level (approx. 1600 lux) but became darker, and more preferable (Franks et al. 2003; Sasaki et al. 2019), over time (Figure 1b). The interior light level of the worsening nest, on the other hand, was initially very dark (approx. 3 lux) but became brighter, and less preferable, over time (Figure 1b). Note that the interior light level of the worsening nest was always darker, thus more preferable, than that of the improving nest until day 4. Temnothorax ants have low sensitivity to light within a bright range, and thus light levels were chosen based on the previous finding (Sasaki et al. 2013). Every morning until day 5, we adjusted the interior light level by replacing the light filters of the secondary roof. To minimize changing the environmental cues, we used the same glass slides for the secondary roof throughout the experiment. To test whether decisions were influenced by the trend in quality, both potential nests had the same interior light level (approx. 400 lux) on day 5, and, on the following day (day 6), the roof of the home nest was removed to induce an emigration. We chose five days as the duration of the experiment based on the previous studies, which showed that Temnothorax ants retain nest information for at least six days (Langridge et al. 2004; Santos et al. 2022). We assayed nest-site preference by recording the site occupied by the colony 12 hours later. If one site contained more than 90% of colony members including all queens and brood items, we designated that as the choice. If no site achieved this criterion, we did not record a preference. This occurred only once out of a total of 30 tests.

Until the roof of the home nest was removed on day 5, the interior light level of the home nest (approx. 1 lux) was always darker than those of the target nests, so the colony stayed at the home nest although some scouts might visit and assess these potential nests throughout the experiment. To confirm these visits and measure the number of scouts, we recorded the number of ants in the target nests at around 9 am each day until the secondary roof was replaced for each of the 10 colonies. The food (the agar-based diet (Bhatkar and Whitcomb 1970) and spam meat) and water tube were placed next to the home nest (Figure 1a) throughout the experiment. The arenas had three photography LED lights above them and received 1540-1690 lux light intensity. Note that the interior light levels reported above are estimations based on the light transmittance (i.e. f-stop) of the light filters.

In addition to the “trend” condition described above, we also conducted two additional conditions as control conditions. In both the control conditions, the procedure was identical to the one used for the trend condition except that the interior light levels of the target nests were constant over time until day 4. In the “average” control condition and the “last-point” control condition, the interior light levels of the target nests were the average of the light levels for each nest type (1024 lux for the improving nest and 88 lux for the worsening nest) and the last level for each nest type (565 lux for the improving nest and 200 lux for the worsening nest), respectively (Figure 1c). On day 5, the light level for the target nests became identical (400 lux), and the roof of the home nest was removed to induce emigration. Out of the 30 tests conducted for each of the two control conditions, six tests in the average condition and two tests in the last-point condition did not achieve the consensus criterion.

We used 30 colonies (see Supplementary Material for the detail of the subjects), all of which were tested for each condition, and the order of the conditions was randomized for each colony. The interval between the tests was at least two weeks. Before each test, all glass slides were washed using a commercial dishwasher, and the experimental arena was cleaned with ethanol. Cardboard plates were made fresh for each test and never reused.

Colonies in the trend condition chose the improving nest over the worsening nest more frequently than the ones in the average condition did, suggesting that the nest-quality change influenced the nest preference (i.e. the nest with improving quality became more preferable than the one with worsening quality). However, because there was not a significant preference for the poor nest site in the trend condition (see Results for details), one possible alternative hypothesis was that the nest-quality change impaired the ability to assess nest quality (Burns et al. 2016), leading to no nest-site preference. To test this hypothesis, we collected additional 29 colonies (see Supplementary Material for the detail of the subjects) and ran another experiment similar to the trend condition. In this “follow-up trend” condition, we also changed the qualities of two nests—one improved and the other worsened over time—but, unlike the trend condition, the quality of the improving nest was better than that of the worsening nest in later days (days 3 and 4; Figure 1d). Note that the means of these over-time qualities were the same (approx. 700 lux). All colonies were tested only once, and four colonies did not achieve the consensus criterion. In the follow-up trend condition, we did not record the number of ants in the target nests each day as we did for the trend and control conditions.    

Analysis

Preferences in the binary choice were assayed with a χ² goodness of fit test. The effect of the nest-quality change was tested with a χ² test of independence. We have removed the choices that did not achieve the criterion for consensus. We ran a generalized linear mixed-effects model (GLMER) and tested colony size (the numbers of workers and brood) and order of the test as fixed effects and colony ID as a random effect . The statistical software R (v. 4.1.2) was used for all analyses.  The data and code can be accessed online (Tyler et al. 2023).