Overfeeding, i.e., providing animals more feed than they consume, is a common practice in the aquaculture industry. The main reason is the concern of underfeeding fish, causing slow growth, low feed efficiency and increased stress levels and aggression (Cañon Jones et al., 2010; Martins et al., 2012). However, overfeeding results in release of uneaten feed into the environment, representing a significant source of pollution (Taranger et al., 2014; Ahmed et al., 2019). Furthermore, overfeeding wastes valuable resources and undermines the sustainability of the aquaculture industry. Because a substantial amount of the total production cost (40-50%) in salmon farming is related to feed (Asche & Oglend, 2016; Iversen et al., 2020) overfeeding also reduces profitability of aquaculture companies. Optimizing the feeding levels and thereby increasing feed efficiency and reducing nutrient waste has been a priority of the aquaculture industry, prompting the development of feeding support technologies such as echo sounders (Juell et al., 1993) and pellet detection-sensors to address this issue (Brijs et al., 2021). However, basic research and data collection is still needed to develop suitable models that can predict fish appetite and feed intake related to each meal based on a deep understanding of these physiological processes.

Feed intake is a complex process, driven by several external factors such as light, season, temperature, feed availability, feed composition, stress, predators, etc., and internal factors such as life stage, gut filling, stored energy, activity levels, etc. (Volkoff et al., 2010; Rønnestad et al., 2017). Therefore, it is very challenging to predict the best feed delivery rate and amount, which ideally correspond to the number of pellets that the fish can capture before they go to waste (Bailey & Alanärä, 2001). For farmed fish, the amount eaten during a meal (defined as “any of the regular occasions in a day when a reasonably large amount of feed is eaten” (Blundell et al., 1987)) depends on the feeding schedule combined with the appetite or motivation to eat at the time (Volkoff, 2016). Additionally, predictable feeding practices have been shown to have a positive impact on fish welfare compared to unpredictable feeding times (Cañon Jones et al., 2012). Normally, a meal may be defined by the amount of feed offered and its duration. Here, meal consumption is defined as number of eaten feed (pellets).

Defining the factors which determine meal initiation (process that leads the fish to start the meal) and meal termination (satiation; a process that leads the fish to stop the meal) are crucial in order to understand and predict feed intake (Chudtong and De Gaetano, 2021). Meal initiation is affected by appetite, defined as the desire to eat and/or hunger, that combines hedonistic signalling pathways with metabolic need to ingest feed to acquire the energy and nutrients necessary to satisfy the demands of homeostasis and growth (Berthoud, 2011; Soengas et al., 2018). On the other hand, meal termination may be mostly dictated by satiation, which is the feeling of fullness after ingesting feed. The next meal ingestion is then constrained by satiety, a multi-factorial and time related process that inhibits further feed intake and represents decline in hunger after the end of an eating episode.

For a better understanding of voluntary feed intake in Atlantic salmon, we have collected data from three short-term independent experiments, in which individual feed intake was monitored for 6 to 12 days. Two main hypotheses were generated: 1) the amount of feed consumed in the previous meal, interacts with time between meals (feed deprivation) to affect the amount of feed intake in the next meal and 2) the average feed intake rate increases with fish size. To minimise variability in stress and energy requirements outside routine metabolic demands, all analysis was conducted on individual fish without social interactions with conspecifics, excluding competition and aggression.

Return of appetite in salmonid species has been found to be correlated with the duration of feed deprivation and emptying of the stomach (Brett, 1971). Our study confirmed this by demonstrating that the feed consumed by individual Atlantic salmon handfed to satiety was positively correlated with feed deprivation time, i.e., the longer the time between the meals, the higher number of pellets were ingested ( Figure 2B ). These results are interconnected with how much the fish ate in the previous meal: we show that feed intake is significantly reduced by the numbers of pellets consumed in the previous meal after a longer feed deprivation period ( Figure 2A ). For the shorter periods between meals (2 h feed restriction), the negative effect of previous meals was not observed, which suports the hypothesis that the gastric emptying (or stomach fullness) is related with the return of appetite (Brett, 1971). Previous studies have correlated the return of appetite with gastric filling and evacuation time (Grove et al., 1978), and it is estimated that 48 h are needed to completely evacuate the Atlantic salmon stomach content (Aas et al., 2017). Also should be taken into consideration that the time to evacuate the stomach is correlated with the size of the meal ingested (Gómez-Pearanda & Clavijo-Restrepo, 2012), as well as to the physical properties of the feed (Aas et al., 2021; Bogevik et al., 2021). In this context, the use of multiple meals a day, such as in this case study, can potentially affect the gastric evacuation time and delay the return of appetite. In addition, a gastric overload reduces absorption efficiency (Jobling, 1986). Nevertheless, it has been observed that different feeding regimes (feed offered in excess for 3 h per day, during the light period of the day only, and continuously) have no effect on salmon growth (Jørgensen & Jobling, 1992).

Larger fish generally will empty a meal of a given absolute size from the stomach at a faster rate than small fish (Flowerdew & Grove, 1979). This can also be deduced from the present data if we assume that gastric evacuation time, and thus filling is correlated with a higher feed intake. It has been hypothesised that the amount consumed in the second meal is equivalent to the amount of feed already transferred to midgut (stomach content evacuated) (Huebner & Langton, 1982). Additionally, the maximum stomach volume is proportional to fish weight, and thus the positive correlation found in the present study between feed consumption and fish growth is expected. However, the interrelationship of feed intake with appetite, hunger and satiety factors is complex (reviewed in (Volkoff, 2016; Rønnestad et al., 2017)), and one factor alone such as the stomach feed content is not enough to explain the return of appetite and consequently feed intake. A range of other factors, such as dietary composition (Forsman & Ruohonen, 2009), the status of energy homeostasis in the body where also the metabolic rate including the activity are important. Also, conditioning and hedonistic factors signalled through neural and hormonal pathways (Soengas et al., 2018) affect the motivation to eat. Additionally, many external factors, such as water temperature (Volkoff & Rønnestad, 2020; Navarro-Guillén et al., 2023), hypoxia (and its combined effect with temperature (Hevrøy et al., 2012)), photoperiod and stress play important roles in this complex process (reviewed in (Volkoff et al., 2010; Conde-Sieira et al., 2018; Soengas et al., 2018; Volkoff & Rønnestad, 2020). Feeding is an important factor for the welfare of farmed fish and understanding the physiological and behavioural mechanisms controlling appetite and feeding behaviour in fish can help predict the quantity of feed that will be ingested at each feeding opportunity.

An important basic premise concerns the timescale of the experimental trials. In the present work we have restricted ourselves to the dynamics of feed intake taking place over a few days. However, it would be of great interest to extend these analyses over several months and at different life stages. Nevertheless, our results validate that feeding should be dynamically adjusted to fish size (growth) and clearly also to the feed deprivation time and feed consumption in previous meals. Additionally, the fact that the fish were handfed in these trials adds additional uncertainty because of potential differences between the feeding staff, i.e., in the human behaviour (“patient” vs “rush” feeder). This random factor was not accounted for in the present analysis. However, this is also relevant to the current feeding practice in salmon culture pens, where individual feeders (with guidelines but also with their own personality) assess appetite via cameras and make their own decision of meal initiation, feeding intensity and meal termination. This work shows the importance of precision and prediction of rations for each meal and supports equipment and models for autonomous feeding systems in the salmon industry. Furthermore, our study assessed only the feed intake of individual fish, removing therefore the social interaction from the equation. It has been shown that social interaction can affects the dynamic of fish feeding (Kadri et al., 1996; Cañon Jones et al., 2010), particularly during feed restriction situations (Cañon Jones et al., 2017). Thus, it would be of interest to do a similar analyses in a fish group-based trial.

Our study demonstrated that length of feed deprivation and meal size have a clear impact on feed intake and growth. It also suggests that feed waste can be reduced by optimising feed intake of fish by allowing for gastric emptying, partial digestion, and absorption of previously ingested feed, thus avoiding overfeeding. Furthermore, our results quantify the link between growth and feed intake and provide statistical modelling parameters that can be used to explore further the basic mechanisms regulating voluntary feed intake of Atlantic salmon.

The raw data supporting the conclusions of this article will be made available by the authors upon request, without undue reservation.

The animal study was reviewed and approved by a local representative at the Institute of Marine Research in compliance with the Norwegian Animal Research Authority regulations.

The experimental trials were planned and conducted by MS, TN, EH, TH and IR. The statistical study was designed by AG, FZ and IR, while AG and FZ conducted the statistical analysis and produced Figure 2 . All authors contributed to the article and approved the submitted version.