Predictive and motivational factors influencing anticipatory contrast: A comparison of contextual and gustatory predictors in food restricted and free-fed rats

Highlights • Using an anticipatory negative contrast (ANC) paradigm, food restricted animals can act selectively in their eating behavior.• Contextual and gustatory predictors in a within-subject design are sufficient for anticipatory negative contrast development.• Changes in reward palatability may underlie contextually-driven anticipatory negative contrast.• An increase in premature port entries to the unavailable sipper – a second measure of ANC – in all groups reveals a direct influence of response competition on ANC development.


Introduction
In modern society, eating has a time and a place but when faced with strong sensory cues linked to food the temptation to eat can be overwhelming. The sights, sounds and smells of a familiar kitchen while dinner is being prepared can evoke a strong drive to eat but in refraining from doing so one can reserve their appetite for the main course. Such consummatory choice behavior is also at play in foraging animals where decisions based on prior knowledge of territorial food sources can result in animals passing up a nutritionally-insufficient option for one of greater value that is likely available ( [23,47], and reviewed in [27,43,49]). Such behavior highlights the ability of animals to integrate past information regarding food availability for implementing an effective foraging strategy.
The behavioral principles that underlie intertemporal decisionmaking regarding food have been well studied experimentally across various scientific disciplines [39,44,51]. In one such recent investigation, Billard et al. [2] observed that when a preferred prey was available at night cuttlefish acted selectively in their food choices, choosing to forgo a less preferred option (crab) in the day and consuming more of their preferred option (shrimp) at night. When shrimp were subsequently made unavailable the same cuttlefish adapted their behavior and became more opportunistic in their feeding, choosing to eat more crab during the day. Ultimately, these animals were able to base their prey choices on prior experiences of availability to optimize their feeding behavior.
A related intertemporal choice paradigm has been frequently used for investigating food-related contrast effects in rodents [6,13,31]. Here, in anticipation of a palatable food source rats learn to restrict the consumption of a less rewarding food type, resulting in an increased consumption of the preferred food later in the session when it is made available. These sessions are compared with other in which the preferred food is not made available and the less rewarding food is provided throughout. Over multiple sessions rats reduce their intake of the less preferred food source selectively in sessions where the preferred food follows. This predictive restriction of food intake is called anticipatory negative contrast (ANC) and has been shown to develop with different types of food sources as well as drugs of abuse of different hedonic value (Lucas et aI., 1990; [6,18,20,42]) -with the disparity between the two comparative rewards being of utmost importance in developing a contrast effect [14,15,37]. Experimentally, whether or not the session is one where the less preferred food will be followed by a more preferred option is often signposted by sensory information such as discrete sensory stimuli, contextual cues and/or gustatory sensations where solutions are flavored while keeping the nutritional content equivalent [17,32]. These modulating cues or "occasion setters" are more or less effective in ANC development [13,32,52], indicating that an accurate memory of the predicted reward in a given environment is required for contrast effects to occur.
While a robust reduction of consumption in anticipation of a future reward is well established using ANC paradigms, the psychological processes underlying this phenomenon are less understood. In their 1994 study, Flaherty et al. proposed three mechanisms for explaining ANC behavioral manifestations: 1) a progressive devaluation of the first food source when anticipating a preferred option; 2) competing behavioral responses, such as spatial competition with the unavailable port, reducing the amount of time dedicated to licking/eating; and 3) active inhibition of the urge to consume the less preferred option despite animals, in some experiments, being motivated by food restriction. While interpretations based on data from variations of the ANC paradigm mostly support ANC resulting from a reward devaluation of the first food source (Flaherty and Rowan, 1995;[52]; see also discussion [41]) there have been some conflicting findings as to how the motivational state of the animal (i.e. food restriction) [15,50] and the nature of the predictors (contextual vs flavors) [17,32] impact ANC development. Food restriction has been shown to decrease [52] but also increase consumption of the less preferred food option [15,50], as increasing food intake in an opportunistic fashion would help meet the animals' metabolic requirements. Moreover, gustatory cuesas when a flavor is added to the solutions -have resulted in no significant contrast effects [17,32], a result that has been interpreted as the flavor acting as a secondary reinforcer for the preferred food source thereby facilitating its consumption [11,17,32].
In the current study we aimed to shed light on these conflicting data by comparing contextual and gustatory ANC predictors in both food restricted and free-fed rats. To do so we have used a novel carbohydrate (solution 1; maltodextrin) and condensed milk (solution 2) sequence for further establishing a contrast effect with nutritionally-relevant food sources. Interestingly, different parameters of licking behavior have been related to distinct motivational processes ( [8]; reviewed in [10,25] and [40] where changes in the solution's palatability and/or incentive value have been shown behaviorally to correspond to how much the reward is liked and wanted, respectively [1]. As such we have analyzed lick microstructure to relate changes in lick parameters to changes in the hedonic value of the solutions, as this could shed light on the underlying psychological mechanisms of ANC.
Our results indicate that, despite being hungry and potentially benefiting from opportunistic feeding, rats that are food restricted are selective in their eating behavior and show strong contextually-driven ANC. In addition, they do so after fewer training sessions than their free-fed counterparts. These differences mirrored changes in palatability for the less preferred solution across the different sessions as gauged by lick microstructure analysis. Moreover, in contrast to previous research, gustatory predictive cues in both food restricted and free-fed rats were sufficient for ANC development, an effect that could not be explained by hedonic changes as determined by lick measurements of palatability and a pre/post-conditioning flavor preference test.

Animals
For all experiments 40 male Sprague Dawley rats, of approximately 300 g at the start of testing, were purchased from Charles River (Cambridge, UK) and housed in pairs. The room was kept at 21 • C and humidity of between 40% and 70% under a 12 hour light-dark cycle (lights on at 7:00). Rats had free access to standard lab chow (Teklad Global Diet, Envigo) and water. Prior to the experiment, half of the rats were put on a mild food restriction, and given 12.5 g of lab chow per day. This food restriction continued on experimental days but rats were put on a free-access diet during the weekends. For food restricted animals, daily food ration was given immediately following the day's behavioural session. All rats were weighed daily Monday through Friday and a percentage of baseline weights calculated. All testing was conducted in accordance to the Animals (Scientific Procedures) Act 1986 (PPL# PFACC16E2).

Behavioral protocol
Animals were trained and tested in two identical operant chambers (30.5 × 24.1 × 21.0 cm; Med Associates), each located inside a soundand light-attenuated aluminum outer chamber (1200 × 700 × 700 cm). The behavioral chambers were equipped with a house light located on the left wall and 2 retractable sippers located on the right wall, which when extended fully were located approximately 1 cm behind the chamber wall. Sippers were accessed via ports (oval shaped H: 1.5 cm W: 1.1 cm) in the wall through which rats needed to poke their noses. This arrangement ensured that only the tongue could contact the sipper and prevented the formation of fluid bridges meaning that individual licks were recorded with high fidelity. Contact lickometers (Med Associates) were used to detect licks and ports were fitted with infrared beam breaks for detecting port entries. The house light was turned on at the beginning of each daily session and turned off at the end of it. Equipment was controlled by a computer running Med-PC IV Software Suite (Med Associates). Webcams were used to monitor the animals' behavior.
Animals were pre-trained for 10 min per day to lick a palatable 10% sucrose solution in all experimental boxes (Context A and B) for 6 days prior to the experiment. For experiments using a contextual predictor, a standard Med Associates box with clear plexiglass walls and a barred floor was used for Context A and a modified version with a striped wall covering, a fine wire mesh floor and continuous white noise (75 dB level) played on a speaker was used for Context B. Both sippers were extended simultaneously for 5 min, retracted for 20 s and presented again for a further 5 min. Licks were recorded from both sippers at this time to check that there was no location bias within the contextual behavioral paradigm.
The ANC conditioning protocol ( Fig. 1) consisted of one session run across two days, a control and an experimental day. On Day 1 (control day) of context predictor experiments, sipper 1 containing a 2% maltodextrin + 0.2% sodium saccharin (Malt) solution was extended (phase 1; 5 min) followed by an inter-phase interval (IPI; 20 s) during which both sippers were retracted and finally extension of sipper 2 containing an identical solution as phase 1 (phase 2; 5 min). On Day 2 of each session (experimental day) of context predictor experiments, animals were placed in the second context and sipper 1 was extended containing Malt solution, identical to control day (phase 1; 5 min). This was followed by an IPI (20 s) with no sippers extended and finally, sipper 2 extended containing a 50% (weight per volume) Malt (2% maltodextrin + 0.2% sodium saccharin) and 50% condensed milk (CM) solution (phase 2; 5 min). For flavor predictor experiments, the context was always identical and flavor was used as a predictor by adding either grape or cherry Kool-Aid (0.05%) to solutions. Specifically, flavor A was added to both phase 1 and phase 2 solutions on control days (phase 1, Malt / phase 2, Malt) whereas flavor B was added to both solutions on experimental days (phase 1, Malt / phase 2, CM). For all ANC conditioning experiments the first sipper location (left or right) and the predictor (i.e. flavor and context) were counter-balanced. For each of the 16 sessions (32 test days), the timing of detected licks at each spout and the number of head entries in each port was recorded.
For the flavor preference test performed before and after ANC conditioning, two sippers each containing one of two flavors (grape or cherry) + 0.2% sodium saccharin were extended one at a time for a maximum time of 30 s, or 10 s after first sipper contact. Overall, 10 "forced choice trials" where only 1 sipper was presented at a time was followed by 30 "free choice trials" where both sippers were presented together. In the flavor preference tests, we recorded licks at each spout and analyzed the total licks for each solution in each type of trial (freechoice vs. forced-choice). At the end of the experiments animals were humanely culled via a Schedule 1 method.

Statistical analysis and code availability
Behavioral data (lick and port entry timestamps) were extracted from data files and analyzed using custom Python scripts that measured numbers of licks for each solution. All raw data files and the analysis scripts are available at the following link: DOI: 10.5281/zenodo.4772860. From these measurements we obtained and analyzed the following variables: Total licks (at each spout); Premature port entries; Licks per cluster (solution palatability); and Total clusters (incentive value of solution). From these, we determined total licks in each phase, total head entries in both unavailable ports, number of lick clusters and licks per cluster were recorded for each of the 32 days (16 sessions). Lick microstructure was analyzed by using interlick intervals to divide licks into clusters. Clusters were defined as runs of licks with no interlick intervals > 500 ms [8]. Experimental and Control day data was normalized by dividing total licks on experimental days by total licks on control days. For statistical analysis of within session behavioral variables, two-way repeated measures ANOVA were used with Condition (control vs. experimental) and Session as within subject variables. For two-way statistical analyses Condition (control vs experimental) and Session (ANC conditioning days) were compared and for three-way ANOVA, Diet (FF and FR) was included as a variable.

Total licks: Contextual predictors drive ANC in free-fed and food restricted rats
We used a modified ANC paradigm with a novel sequence of rewarding solutions to gauge the effectiveness of contextual or gustatory predictors for enhancing contrast effects in free fed (FF) and food restricted (FR) Sprague-Dawley rats (Fig. 1). When using contextual cues as predictors, we saw clear ANC develop in FF rats (n = 10) as evidenced by reduced consumption during phase 1 specifically on days when a more preferred solution was expected. Statistically, total lick measurements revealed a significant effect by ANC conditioning day (session) (2way ANOVA: F(1,9) = 6.20, P = 0.03) and interaction between session and Malt-Malt and Malt-CM days (condition) (F(7,63) = 4.62, P = 0.0003) ( Fig. 2A1-left). Post-hoc Bonferroni analysis showed significant differences in total phase 1 licks between Malt-Malt and Malt-CM days on sessions 6, 7 and 8. No significant difference was seen for condition (F (7, 63) = 1.75, P = 0.11). As expected in phase 2, FF rats ( Fig. 2A1-right) increased their lick rates to the highly palatable CM solution on Malt-CM days compared to the less preferred phase 2 Malt solution on Malt-Malt days (FF, 2-way ANOVA, Condition: F(7,63) = 8.46, P < 0.0001; Session: F(1,9) = 36.70, P = 0.0002; Interaction: F(7,63) = 9.17, P < 0.0001; Bonferroni post-hoc differences -Sessions 2 through 8). Using a novel sequence of rewards in the current study, these results extend previous findings that FF rats can develop robust ANC [6,15,22].
FR rats (n = 10) also showed strong ANC ( Fig. 2B1-left). Statistically, Overall, these results indicate that contextual cues can act as effective signals for decreasing animals' intake of a less palatable Malt solution in anticipation of a more preferred CM option whose intake increases when it is made available. Moreover, FR animals developed ANC after fewer conditioning trials (session 3) than FF animals (session 6) suggesting that despite being hungry FR rats can act selectively in their feeding  behavior.

Premature port entries: contextual predictors drive ANC in free-fed and food restricted rats
A commonly used ANC measure is based on the amount of sipper contact during licking. In addition to quantifying contrast effects using total licks, we measured entries to the phase 2 sipper port as a second measure of anticipation (Fig. 3). In addition, this analysis aims to shed light on the contribution of competing behavioral responses (i.e. spatial competition with the unavailable sipper) in reducing the amount of time spent licking/eating that may contribute to a decrease in total licks on Malt-CM days. Differences in premature port entries between Malt-Malt and Malt-CM days were seen in both FF and FR rats (FF: 2-way ANOVA, Condition: F(1 9) = 19.91, P = 0.002; Session: F(7,63) = 6.54, P < 0.0001; Interaction: F(7,63) = 10.05, P < 0.0001; Bonferroni post-hoc differences -Sessions 3,4,6-8) (  (Fig. 3B2-left). Overall, using premature port 2 entries as a second measure of anticipation these results further show that contextual cues are sufficient predictors for robust ANC development.
In summary, when using contextual predictors, ANC was seen to develop in both FF and FR rats and this may be associated with a change in palatability of the non-preferred maltodextrin solution on experimental vs. control sessions.

Premature port entries: Gustatory predictors drive ANC in free-fed and food restricted rats
To further support the impression that gustatory cues can act as effective predictors for ANC to develop we next looked at premature phase 2 port entries. In line with the total lick data, both FF and FR rats on Malt-CM days made significantly more premature phase 2 port en-  (Fig. 6B2-right). There was, however, no significant difference between IPI phase 1 port entries  (Fig. 6B2left). These results add support to the data on total licks showing that gustatory cues are sufficient predictors for ANC to develop.

Lick microstructure analysis for ANC driven by gustatory cues reveals no change in lick patterning
To determine the underlying hedonic processes behind the total lick changes resulting from using flavored cues we next analyzed the lick microstructure of lick responses influenced by gustatory predictors. In  (Fig. 7B1-right). Notably, these results show that in either motivational state (FF or FR) licks per cluster in phase 2 shifted from high to low levels with conditioning trials (Fig. 7A1,B1-right).
The amount of total clusters in phase 1 for either FF or FR rats did not differ in Malt (Fig. 7B2-right) which is in contrast to the changes in lick microstructure for contextually-driven ANC. These differences in the lick microstructure between animals using contextual (Fig. 4) and gustatory predictors (Fig. 7) suggest that the factors underlying ANC for each are distinct.

Flavor preference tests show an increased lick frequency to both flavors following conditioning
To shed light on the hedonic changes in reward properties that underlie ANC we next tested whether the flavor cues without maltodextrin gain appetitive value as a result of contrast conditioning using a flavor + saccharin preference test . Grape and cherry flavored 0.2% saccharin solution without Malt or CM were made available before and after ANC conditioning to determine baseline flavor preference and changes in preference as a result of conditioning (Fig. 7C1). For FF rats, a 2-way repeated measures ANOVA revealed a significant effect by test (pre vs post conditioning) but no main effect by Condition or Interaction between the two (Pre/Post: F(1,9) = 58.04, P < 0.0001; Condition: F (1,9)=2.12, P = 0.18; Interaction: F(1,9) = 1.83, P = 0.21) indicating that there was no initial preference before conditioning and that licking to both flavors, irrespective of which flavor predicted CM, increased after conditioning (Fig. 7C2). Similar results were seen for FR rats, with the exception that the main effect by Pre/Post was mostly driven by the flavor given on Malt-Malt days (2-way ANOVA, Pre/Post: F(1,9) = 24.21, P = 0.0008; Condition: F(1,9) = 0.34, P = 0.56; Interaction: F (1,9) = 0.35, P = 0.57) (Fig. 7C3). These results suggest that pairing flavors with Malt enhances intake of both gustatory cues, regardless of which one predicts CM. The current experiments however cannot rule out that at least part of this increase is due to a decrease in flavor neophobia associated with repeated flavor presentation.

Contextual predictors are more effective ANC cues than gustatory predictors
In the current study, four unique contrast conditions were tested (Context: FF and FR; Flavor: FF and FR). To formally compare each condition, we next analyzed the normalized phase 1 total lick data (= total licks: Malt-CM/Malt-Malt) to determine the effectiveness of each condition compared to one another in promoting ANC (Fig. 8). A 3-way ANOVA used to compare the results by Diet, Predictor, and Session showed a main effect by Session and an Interaction between Session and Predictor (Session: F(3.46,124.6) = 6.57, P = 0.0002; Session and Predictor: F (7, 252) = 2.07, P = 0.04; No effect of Diet: F(1, 36) = 0.3132, P = 0.58). Post-hoc Bonferroni tests, however, did not show a significant different on any session between any of the groups. The results here, however, need to be carefully considered as the experiments using flavor and contextual predictors were not performed simultaneously.

Discussion
Using a novel sequence of nutritionally-relevant rewards, in the current study we tested the predictive ability of contextual and gustatory cues on ANC and the motivational processes that underlie this behavior.
Our results indicate that both contextual and gustatory information are effective predictors for ANC to develop. The selective reduction of Malt consumption may be at least partially due to spatial competition between the two sipper ports as increases in premature port entries in phase 1 were significantly increased on experimental days. In contextually-driven anticipatory sessions, our lick microstructure analyses suggest that ANC may be determined by a change in the rewarding properties of the Malt solution that is in all other respects equivalent to the one presented on control days -the only difference being the context  signaling the availability of CM in phase 2. In contrast, ANC driven by gustatory predictors was not mirrored by changes in the lick microstructure indicating that the contrasting effects seen here are perhaps driven by competing factors. Due to their biological relevance, gustatory cues can be robustly associated with food sources, potentially acting as strong secondary reinforcers that can compete with negative contrast effects [17,32]. Since the Malt solution is itself rewarding [46], animals may develop a strong liking for both flavors, irrespective of whether or not one also predicts CM in phase 2. In line with this idea, our flavor preference tests demonstrate a statistically equivalent increase in licking for both flavors after ANC conditioning.
Previous research, however, has shown that within-subject ANC can develop without physical predictors but rather continuous alternating days alone can signal the different conditions ( [17], but see [48] for a failure to find such an effect). It can be argued that this may explain why, unlike previous research, our study shows significant ANC using gustatory predictors. One argument against this alternative interpretation is that in current study every five days animals were given a 2-day break and by doing so may have disrupted the alternating day contingency. Instead, it is plausible that ANC driven by flavor predictors seen here is due to the fact that, unlike previous research [17], similar flavoring was added to the phase 2 Malt-CM solution. By doing so our design may have helped strengthen the relationship between the flavored Malt solution in phase 1 with the similarly flavored Malt-CM solution in phase 2 thereby resulting in robust flavor-driven ANC.
Stereotypical patterns of rodent licking behavior are thought to reflect different underlying motivational processes where the amount of licks in a cluster of continuous licks has been shown to relate to reward palatability (reward liking) while the amount of total clusters changes with incentive value (reward wanting) as well as signal its post-ingestive feedback properties ( [8]; reviewed in [10,25] and Naneix et al., 2020). Support for these parameters as useful metrics of reward properties has come from studies investigating changes in the hedonic characteristics of rewarding solutions where alterations in lick dynamics have been shown to correspond to changes in reward magnitude [9] or the motivational state of the animal [7], [36]. In the present study, only contextually-driven ANC showed a corresponding change in the lick microstructure where a change in licks per cluster was seen to develop between control and experimental days as ANC developed. These results are consistent with recent work by Wright et al., [52] showing differences in licks per cluster that were related to contextually-driven ANC using sucrose rewards of different magnitude. In phase 2, CM lick patterning was also different in the two predictor groups despite both groups showing similar total CM lick rates. In both free-fed and food restricted groups, licks per cluster and total clusters for CM increased in sessions predicted by contextual cues while for animals relying on gustatory cues there was a shift from a licks per cluster strategy to more licking bouts as ANC developed. Interestingly, compared to contextual cues, gustatory predictors for food restricted rats resulted in more total licks in phase 1 (Total licks: Context, 11,567 ± 874; Flavor, 14,968 ± 915 Unpaired T-Test, P = 0.02) suggesting an increase in the incentive value of the flavored Malt solutionsan interpretation which is consistent with the results of the flavor preference test. Thus, it is possible that the addition of a flavor to the maltodextrin solution in phase 1 enhances its palatability vs. a non-flavored maltodextrin solution. This change in Malt palatability may in effect be impacting the hedonic disparity between the two rewards by altering the incentive value of CM. Such changes are reminiscent of those seen in temporal discounting paradigms where a comparative analysis of two rewards separated by time can impact the incentive value of each [4,30,38]. Despite showing comparable total CM lick rates, our data show that, with experience, differences in CM lick patterning appear between gustatory and context groups suggesting that the predictive and hedonic value of the phase 1 solutions are differentially affecting the motivational processes underlying CM consumption.
Occasion-setters have been shown to be at play when features are trained so that they disambiguate the relationship between another stimulus with an outcome [24]. As such, the within-subjects nature of the current design using different predictors warrants a description of the paradigm in terms of occasion setting mechanisms. The target Malt is followed by the rich CM on half of the occasions, and therefore is an ambiguous predictor of CM. It is under such ambiguity that the context (Experiment 1) and the flavor (Experiment 2) stimuli disambiguate the meaning of Malt. In other words, Malt is followed by CM only when a feature is present. This design is thus reminiscent of an occasion setting design. One characteristic of occasion setting is that it occurs best when the occasion setter is presented serially with the target stimulus that it disambiguates. Although occasion setting has been observed with both simultaneous and serial compounds, it is much stronger with serial compounds [19,24]. This parallels the present results in that better ANC was observed when contextual cues were used relative to when flavor cues were used. Because contextual cues were experienced before Malt, whereas flavor cues were experienced simultaneously with Malt, the advantage of serial over simultaneous presentation of stimuli seen in occasion setting experiments can explain the difference between contextual and flavor cues observed in the current experiments. The fact that contrasting effects are based on relative rather than the absolute value of the rewards make them sensitive to various factors. For instance, their value-based as well as temporal disparity can strongly impact contrast [6,16,18,37]. The motivational state of the animal is another key factor and experiments have shown that food deprivation can result in animals not developing ANC or even showing positive induction where animals increase consumption to the less rewarding option [15,50]. In the current study, all food-restricted animals developed ANC. For those using contextual cues as predictors, these differences occurred after fewer conditioning trials than free-fed animals. Moreover, these animals consumed more CM than free-fed animals demonstrating a preferential food seeking approach for maximizing CM intake (CM Total licks: FR, 20,817 ± 963.0; FF, 17,500 ± 859 Unpaired T-Test, P = 0.02). These results may be attributed to the novel food choice sequence (Malt followed by CM solution) where the disparity in the hedonic value and nutritional content between the two was different than past investigations [39,6,18,42]. Condensed milk has been shown to be a strong reinforcer as reward seeking studies have shown similar effort-based responding for condensed milk and cocaine [5], [33], [45]). This highly palatable food choice may thus be effective in promoting a selective feeding strategy even when animals are food restricted. The current findings thus indicate that, despite the benefit in taking an opportunistic approach to maximally consume in both phases, food restricted animals can act selectively in their feeding choices so as to potentially maximize their intake of a nutritionally-rich food option, and that learning mechanisms underlie these choices.
A number of different neural processes must be at play when animals learn to predict rewards of different hedonic value. The lick microstructure analysis performed in the current study suggests that gustatory and contextual predictors may be mediating ANC through different motivational or cognitive processes. Studies investigating the neurobiological mechanisms that underlie ANC have mostly focused on the changes in reward value that might contribute to ANC development. While dopamine activity encodes information predicting a future reward as well as the reward itself [3] the role it plays in ANC development is unclear. Using systemic administration of the monoamine stabilizer (− )-OSU6162, Feltmann et al., [12] showed that it had no impact on anticipatory contrast. Moreover, lesions of the nucleus accumbens (NAc) -a region important for integrating dopamine-based reward processing have no effect on ANC [29] suggesting that alternative regions may be involved. One candidate circuit might be prefrontal cortex (PFC) where reward based and sensory signals converge for supporting memory [21]. In addition, PFC functioning has been heavily implicated in cognitive control mechanisms for selecting appropriate actions [34] often with delays imposed between stimulus and response [26,35,28]. As ANC requires the integration and working memory representation of sensory content predicting reward the PFC may play an important role in ANC development.
In summary, our results indicate that despite the potential benefit in taking an opportunistic feeding approach, food-restricted animals can use contextual predictors to act selectively in their feeding choice and that these changes may stem from learned changes in the hedonic properties of the readily available food source. Gustatory predictors can also be used to optimize intake of a preferred food option in both free-fed and food restricted rats but less effectively than contextual cues -a results that may be due to the high predictive strength of flavors linked to a food source that differentially impacts the underlying processes that drive ANC. Future investigation of the neural activity contributing to these motivational and cognitive changes will help elucidate the neurobiology that allows animals to optimize their foraging strategies.

Declaration of Competing Interest
The authors declare no competing financial interests.