Interactions between learning and non-learning plasticity in the beadlet sea anenome Actinia equina: A multidimensional reaction norm approach.

Learning is widespread in animals and is thought to have evolved because it enables them to change the way they behave on the basis of new information. For example, when an animal encounters a new event it will often respond by hiding. But if that event is experienced multiple times and turns out not to be dangerous, the animal will gradually reduce the amount of time it hides for, each time the same event is encountered. This simple form of learning is called habituation and the benefit of habituating is that it reduces the amount of time spent hiding needlessly. Instead, that time can be used for other important activities such as searching for food or mates. While habituation has been well documented in an array of species, important gaps in our knowledge of this process, and its evolutionary causes, remain.

First, animals also adjust their behaviour in the absence of learning, for example by hiding for longer if a threat like a predator is present. These types of changes are known as behavioural plasticity. At present, little is known about how learning to habituate could be affected by non-learned plasticity. We would expect plasticity to affect learning because both are dependent on an animal's ability to process information about external conditions. For instance, habituation might occur more slowly (a slower decline in hiding times across repeated exposures to a stimulus that turns out to be non-harmful) if additional risks such as a predator are also present. Equally, habituation might occur more rapidly when the rewards of emerging from hiding, such as the presence of food, are higher. Additionally, habituation might be less likely when the intervals between repeated stimulation are longer, due to the process of 'forgetting'. Interestingly, forgetting might be as beneficial as learning because it would prevent habituation in the absence of relevant information. Our first objective is to understand how habituation is affected by risk, reward and the timing of learning opportunities. Second, it is known that individuals differ in their behaviour including in their pattern of learning (e.g. how quickly they habituate) but we do not know if they differ in how habituation is affected by additional conditions such as those discussed above.

Here, we will test for such differences and, where present, determine what proportion of those differences can be explained by genetic variation. If differences in learning pattern (and how it is affected by additional conditions) are present, between individuals and genotypes, this would show that learning can be subject to evolution by natural selection. We will investigate these questions in a sea anemone, a member of the cnidarian phylum, which also includes jellyfish and corals. Hiding behaviour is easy to study in these animals because they retract their feeding tentacles when physically disturbed (in this case using a water jet), and the duration of this hiding response can be repeatedly stimulated and automatically measured using an artificial intelligence system to analyse video footage. Cnidarians are also an important group of animals as they occupy the same level in the evolutionary tree of life as more complex 'bilateral' species such as vertebrates, insects and molluscs. Therefore, any patterns of learning present in cnidarians could be ancient, having evolved independently of the central nervous systems seen in more complex animals. Furthermore, the insights about learning in these simple animals could then be applied to studies of more complex and captive species where we want to devise conditions that optimise habituation for reasons of animal welfare and production. In summary, we will study the conditions that affect habituation in sea anemones to gain insights into the evolution of learning and provide practical knowledge that can then be adapted to study and manipulate learning in more complex animals.

Unifying learning with the wider concept of phenotypic plasticity opens up novel questions about its evolution, which we will address using habituation in Actinia equina, a proven model system for collecting longitudinal behaviour data. We will focus on three broad questions: (1) Does learning interact with non-learned behavioural plasticity? (2) Do these interactions differ between individuals? (3) How much of this variation can be explained by genotype? Additionally we will ask whether non-learned plasticity affects residual variance around learning trajectories. To address these questions we will use mechanical stimulation to induce startle responses, where the feeding tentacles are rapidly retracted before slowly reopening. Pilot data indicates that A. equina habituate to this stimulation, reopening more quickly on average over repeated exposures. We will evoke repeated startle responses across three categories of manipulated extrinsic conditions: chronology of exposure, rewards (flow rate - equivalent to the availability of prey) and risks associated with reopening (potential for injury from a conspecific). Each individual from three experiments will be genotyped in order to produce a relatedness matrix. This will allow genotype, as well as individual identity, to be a predictor in our models. Furthermore, we will build on previous experiments by implementing a high throughput system for behavioural phenotyping, combining automated stimulation with data extraction via machine learning and computer vision. These studies will allow new insights of the evolution of learning by revealing its potential exposure to selection and by addressing knowledge gaps about learning in the absence of a central nervous system. Fundamental principles demonstrated in A. equina could then be adapted and applied to more complex situations where optimising the rate of habituation in captive species could yield welfare and productivity benefits.