An association of food sources with odors prominently guides foraging behavior in animals. To understand the interaction of olfactory memory and food preferences, we used glucose-averse (GA) German cockroaches. Multiple populations of cockroaches evolved a gustatory polymor-phism where glucose is perceived as a deterrent and enables GA cockroaches to avoid eating glucose-containing toxic baits. Comparative behavioral analysis using an operant conditioning paradigm revealed that learning and memory guide foraging decisions. Cockroaches learned to associate specific food odors with fructose (phagostimulant, reward) within only 1 hr of condi-tioning session, and with caffeine (deterrent, punishment) after only three 1 hr conditioning ses-sions. Glucose acted as reward in wild type (WT) cockroaches, but GA cockroaches learned to avoid an innately attractive odor that was associated with glucose. Olfactory memory was retained for at least 3 days after three 1 hr conditioning sessions. Our results reveal that specific tastants can serve as potent reward or punishment in olfactory associative learning, which reinforces gustatory food preferences. Olfactory learning therefore reinforces behavioral resistance of GA cockroaches to sugar-containing toxic baits. Cockroaches may also generalize their olfactory learning to baits that contain the same or similar attractive odors even if they do not contain glucose.

2.1. Insects
The wild-type (WT) B. germanica colony (Orlando Normal) was collected in Florida in 1947 and has served as a standard insecticide-susceptible strain in many studies. The glucose-averse (GA) colony (T-164) was collected in Florida in 1989 and shown to be aver-sive to glucose; continued artificial selection with a glucose-containing toxic bait fixed the homozygous GA allele(s) in this population (approximately 150 generations as of 2020) resulting in maximal sensitivity to glucose as a deterrent. All cockroaches were main-tained on rodent diet (Purina 5001, PMI Nutrition International, St. Louis, MO) and dis-tilled water at 27°C, ~40% RH and a 12:12 h L:D cycle (8 pm to 8 am photophase, 8 am to 8 pm scotophase). We tested 10–18 days-old males during scotophase in this study. 

2.2. Olfactometers
Each linear two-choice olfactometer was composed of two connected tubes. The first was an acclimation chamber (acrylic tube, 15 cm long, 3.2 cm i.d.) with a swivel metal screen gate. The upwind end of the acclimation chamber (near the gate) was connected to a bioassay tube (acrylic tube, 54.5 cm long, 3.2 cm i.d.) with a 15-cm-long acrylic divider sealed vertically in the upwind end. The divided bioassay tube formed a linear two-choice assay tube (Fig. 1A). Each olfactometer was connected to a vacuum pump that provided a linear air velocity of 25 cm/s through the tube; the tubes were exhausted outside the building. Fluorescent lights covered with red photographic safety filters placed 60 cm be-low and above the olfactometers facilitated observation in the dark. In each bioassay, an individual cockroach was introduced into the acclimation chamber and allowed to accli-mate to the airflow for 5 min. Only quiescent insects were used in the assays.  

2.3. Lures Containing Odors
Each red rubber septum lure (#1780 J07, Thomas Scientific, Philadelphia, PA, USA) was placed in a 1.5 ml microcentrifuge tube. Undiluted odor compounds or compounds dissolved in solvents were applied to each lure (total volume 100 µl) and placed in the end of the divided bioassay tube. Following traditional associative learning terminology, we termed the training odor associated with a tastant during the conditioning session as the conditioned stimulus (CS) (Fig. 1B).

2.4. Feeding Tubes Containing Tastants 
An aqueous tastant solution (total liquid volume 1.5 ml) was loaded into a feeding tube (1.5 ml microcentrifuge tube) and plugged with a small cotton ball. It was placed downwind of the lure of the divided bioassay tube, so that cockroaches could be exposed to specific odors as they fed. However, to prevent cockroaches from contacting the lure, a metal screen was placed between the feeding tube and lure. In the conditioning session, cockroaches were allowed to freely drink each of the two tastant solutions. Based on our previous findings [33,34], we used the 90% Effective Concentration (EC90) for each tastant, obtained from dose-feeding response curves: 300 mM fructose (EC90 for appetitive re-sponses in both strains), 1000 mM glucose (EC90 for appetitive responses in WT and aver-sive responses in GA cockroaches) and 10 mM caffeine (EC90 for aversive responses in both strains). In WT cockroaches, fructose and glucose were rewards or appetitive uncon-ditioned stimuli (US+) for positive association with odors, and caffeine was punishment or aversive unconditioned stimulus (US-) for negative association with odors (Fig. 1B). In GA cockroaches, fructose acted as a reward, whereas caffeine and glucose were deterrent and served as punishment. When cockroaches associated a CS with US+ by conditioning (training), they were expected to prefer the CS in preference assays. When they associated CS with US-, they were expected to avoid or ignore the CS. 

2.5. Chemicals
All chemicals, except vanilla and chocolate extracts, were obtained from Sigma Al-drich (St. Louis, MO, USA). Vanilla (All Natural Pure Vanilla Extract, McCormick, Hunt Valley, MD, USA) and Chocolate (Chocolate Extract, OliveNation, Avon, MA, USA) were obtained from local grocery stores. 

2.6. Conditioning Session
The two-choice olfactometers were also used for the conditioning session, defined as the training of insects before bioassays. One day before any bioassays or conditioning, 20 adult males (10–12 days-old) were placed in a plastic cage (14.3 × 10.5 × 9.5 cm; T-79, Alt-hor Products, Windsor Locks, CT, USA) containing distilled water, rodent diet and egg carton shelter. The cage had two ports in line with each other to accept the bioassay tube upwind (air inlet) and a downwind exhaust tube (Fig. 1B). These ports were kept closed during the 1-day-acclimation phase. In the Unconditioned odor preference assay, indi-vidual males were obtained directly from this cage. In the conditioning session for the other three bioassays, the downwind port was connected to a vacuum pump, as described in section 2.2 (Olfactometer). The upwind port was connected to the bioassay tube. After removing the rodent diet from the arena, feeding tubes and lures were placed upwind in the divided portion of the bioassay tube. Cockroaches were allowed to freely visit the feeding tubes and thus trained themselves to associate tastants and odors for 1 hr from 12 to 1 pm (mid-scotophase), when cockroaches actively forage [52]. All insects were attract-ed to odor sources and contacted the feeding tubes within 5 min after starting the condi-tioning, and all of them returned to the cage (and shelter) within 1 hr. After the condition-ing session the ports were closed, rodent diet was returned to the cage, and males were kept in the cage until the next conditioning session or bioassays (Fig. 1B). 
Traditional classical and operant conditioning rely on the assumption that changes in conditioned response probability observed during training adequately represent neu-ronal plasticity, and commonly, behavioral plasticity is quantified by averaging over a population of identically treated animals. On the other hand, recent studies using honeybees and American cockroach suggest that, even if insects were trained by well controlled methods, the average behavioral score of the group does not represent in-dividual behavior, which is driven by unique personality traits. The individual learning of such species may be influenced by various factors including their sensory sensitivity, their ability to learn a task, the speed of learning and their asymptotic performance. In this study, however, we did not identify individuals, because the group of insects was self-trained. Therefore, learning curves of individuals were not evaluated during the con-ditioning session, and in Bioassays 2–4 we tested the retained olfactory memory as the average behavioral score of groups after training.

2.7. Bioassay Procedures 
Bioassays were conducted between noon and 4 pm, in mid-scotophase. Unless stated otherwise, a single male was tested only once in a clean olfactometer. When the male was quiescent in the acclimation chamber, the gate was opened carefully, the lures containing odor stimuli were introduced at the upwind end of the olfactometer, and the insect’s re-sponse was noted by direct observation. A positive response was scored when the male entered the divided bioassay tube within 2 min and remained there for 15 sec. After each bioassay, each olfactometer was flushed out with fresh air for 2 min. The positions of the two lures were randomly switched between the left and right sides of the divided section of the bioassay tube. Every five bioassays, the olfactometers were washed with distilled water and ethanol. The percentage of males responding was calculated by the formula: % responding = # of insects making a choice / total # of tested insects. The percentage choice for each lure was calculated as: % choice = # of insects choosing the lure either on the right side or left side / total # of insects making a choice. Chi-square tests (α = 0.05) and Tukey’s Wholly Significant Difference (WSD) tests (α = 0.05) were used to compare treatments, the two lures, and cockroach strains.

2.8. Bioassay 1: Unconditioned Odor Preference
Considering that in their natural environment, cockroaches approach certain innate-ly preferred odors and avoid innately repellent odors, we screened for attractive food odors in various food sources, including plant materials and human food for use in Bioassays 2 through 4 (Fig. 1C). Distilled water, ethanol and mineral oil were used as solvents. As general odors contained in sweet snacks and chocolate drink, we used 3-methyl-1,2-cyclopentanedione (coffee and caramel flavor, 10-4, 10-3, 0.01, 0.1, 1, 10 µg in 100 µl mineral oil per lure), 4,5-dimethyl-2-ethylthiazole (burnt hazelnut odor, 10-8, 10-7, 10-6, 10-5, 10-4, 10-3 µg in 100 µl mineral oil per lure) and 2,4,5-trimethylthiazole (musty, nutty, and brown cocoa and coffee odor, 10-6, 10-5, 10-4, 10-3, 0.01, 0.1 µg in 100 µl mineral oil per lure). As general plant terpenes, we used beta-caryophyllene (sweet woody spicy and peppery odor) and farnesene (sweet, woody, herbal and green aroma) at 10-4, 10-3, 0.01, 0.1, 1, and 10 µg in 100 µl mineral oil in each lure. As aromatic aldehydes, we used benzaldehyde (almond odor, 10-4, 10-3, 0.01, 0.1, 1, 10 µg in 100 µl mineral oil per lure) and vanillin (vanilla odor, 4-hydroxy-3-methoxybenzaldehyde, 10-4, 10-3, 0.01, 0.1, 1, 10 µg in 100 µl ethanol per lure). As a blend of sweet odors, commercial vanilla extract (All Natural Pure Vanilla Extract, McCormick, Hunt Valley, MD, USA) and chocolate extract (OliveNa-tion, Avon, MA, USA) were dissolved in distilled water at 0.01, 0.1, 1 equivalents of the original product to find the optimal concentrations for our bioassays. Each odor was ap-plied to a single red rubber septum lure and placed in one side of the divided bioassay tube. A solvent-only lure was placed in the other side of the divided tube. We tested 20–30 insects at each concentration of each odor source. Additionally, a two-choice test using vanilla and chocolate was conducted with the undiluted original products to assess the unconditioned (innate) odor preferences for these two attractive odors (Table S1 and S2).

2.9. Bioassay 2: Conditioned Odor Preference After Conditioning with a Single Odor
To test whether the insects associated odor with either rewarding or punishing tastants, cockroaches self-trained (operantly conditioned) with a combination of a single odor and a single tastant in the conditioning session, then olfactometer bioassays were carried out (Fig. 1C). During conditioning, one side of the divided bioassay tube contained a single combination of ‘odor (lure) + tastant (feeding tube)’. Six types of combinations were prepared: ‘Vanilla (CS) + Frucotse, Glucose or Caffeine (US+ or US-)’ and ‘Chocolate (CS) + Fructose, Glucose or Caffeine (US+ or US-)’. The other side of the tube contained a lure and the feeding tube contained distilled water. To test the impact of the training, we tested two types of conditioning sessions. The first was a single conditioning session of 1 hr, after which the insects were tested for their odor preference in the two-choice preference assay using both vanilla and chocolate odors, namely, the CS (either vanilla or chocolate) and US (either vanilla or chocolate) approximately 24 hrs later (Fig. 1C). The second training paradigm was three successive 1 hr conditioning sessions (1 hr at 12–1 pm each day for three days), followed by odor preference assays approximately 24 hrs later. In a comparison of odor preference among the treatments, we used the results of the innate odor preferences from Bioassay 1 as control. If trained cockroaches chose the CS more than untrained cockroaches do, we considered that they associated the CS with the US+. If trained cockroaches preferred the CS less than the untrained cockroaches do, we consid-ered that they associated CS with the US-. We tested 30–40 males in each treatment.

2.10. Bioassay 3: Conditioned Odor Preference After Conditioning with Two Odors 
To test whether insects associated two combinations of tastants and odors, males were trained with both vanilla and chocolate odors using two types of tastants. The di-vided tubes contained different combinations of ‘lure + tastant’: ‘Either Vanilla or Choco-late + either Fructose or Caffeine’, ‘Either Vanilla or Chocolate + either Fructose or Glucose’ and ‘Either Vanilla or Chocolate + either Glucose or Caffeine’. Males received either one or three successive 1 hr conditioning sessions. In this paradigm, both vanilla and chocolate used in the two-choice olfactometer bioassays acted as CS associated with either US+ or US- (Fig. 1C). Data analysis was by the same methods described in section 2.9 (Bioassay 2). 

2.11. Bioassay 4: Retention of Olfactory Memory
To test the retention of olfactory memory, we exposed insects to three successive 1 hr conditioning sessions (1 hr at 12–1 pm each day for three days). The combinations of ‘lure + tastant’ were ‘Vanilla + Fructose and Chocolate + Caffeine’, ‘Vanilla + Fructose and Chocolate + Glucose’ and ‘Vanilla + Glucose and Chocolate + Caffeine’. Conditioned pref-erence bioassays with vanilla and chocolate were conducted 2, 3 and 5 days later (Fig. 1C). Data analysis was by the same methods described in section 2.9 (Bioassay 2).