Intake of food and avoidance of poison are crucial to an organism’s survival in a dynamic environment. The process of deciding whether food is “safe” or “hazardous” relies heavily on the gustatory and somatosensory systems, which are involved in evaluating diverse properties of putative foods, such as: chemosensory, e.g., modality, intensity; orosensory, e.g., texture, temperature, pungency; and gratification capacity (Rosenblum, 2008; Carleton et al., 2010).

The sense of taste differs from the other senses in two major characteristics: (1) Each taste input has dual labeling, one related to its physical (texture and temperature) and chemical features, and the other to its hedonic value and safety; (2) Deciding on the safety aspects of a novel sensation, such as the taste of an unfamiliar food, involves a different temporal scale from other senses: taste association with unconditioned stimulus (US) takes 1–12 h (Bures et al., 1998; Merhav and Rosenblum, 2008), compared with a few seconds for responses to other sensations.

Two main strategies are used by various species, including humans, in responding to various tastes: genetic programming, i.e., affinity to sweet tastes, aversion to bitter ones, and complex learning mechanisms that involve several forebrain structures (Rosenblum, 2008). Modification of the basic genetic programming and memories of new tastes and their values are expected to be mediated, at least in part, by the gustatory cortex (GC; Yamamoto et al., 1984, 1985; Rosenblum, 2008; Doron and Rosenblum, 2010), which is defined according to its cytoarchitectonic boundaries as the dysgranular part of the insular cortex (IC; Burwell, 2001). This is in line with its unique anatomical connections (Figure 1), through which it receives multimodal sensory inputs, including visceral, gustatory, and somatosensory information from sensory thalamic nuclei (Fujita et al., 2010). However, other brain regions, such as the basolateral amygdala (Bla) and dorsal hippocampus, have been shown to be activated during novel taste processing (Yefet et al., 2006; Doron and Rosenblum, 2010), and also to be necessary for safe taste–memory consolidation (De la Cruz et al., 2008). Most of the studies of the molecular mechanisms underlying taste memory were performed in mice and rats, therefore, the present review will focus on these studies.

Taste recognition on the receptor level, as well as taste reactivity, as defined genetically, are beyond the scope of this review; they have been extensively covered elsewhere (Frank et al., 2008; Carleton et al., 2010). The first objective of this review is to describe taste-related behavior paradigms, in order to elucidate the molecular and cellular mechanisms underlying learning and memory. We then provide a detailed analysis of the currently known molecular and cellular mechanisms of taste learning in the GC, which resides in the IC. We aim to focus mainly on studies during the last decade; previous studies have been reviewed elsewhere (Bures et al., 1998; Rosenblum, 2008).

The molecular mechanisms underlying learning and memory are the subject of ongoing research. Although learning and memory are considered to be two different processes, they involve a continuum of events. Following a discrete event, physical changes underlying memory encoding and processing of the information to be stored takes place. Ethologically, this is described as the acquisition phase; it involves creation of an internal representation of the novel information. This representation remains labile for some time, while the process of consolidation takes place. During this process the new memory becomes increasingly resistant to disruption (Alberini, 2011), which can involve several types of intervention: behavioral (e.g., Merhav et al., 2006), pharmacological (e.g., Rosenblum et al., 1993), or others (Bures et al., 1998). It has been shown that the consolidation of new memory can be disrupted by many events, including: blocking synthesis of new RNA or protein, e.g., by actimomycin D or anisomycin, respectively; disruption of the expression or function of specific proteins; new learning; brain cooling; seizure, e.g., through electric shock; brain trauma; and regional brain lesions (Alberini, 2011). During the consolidation phase, the memory is transformed from short-term memory (STM), which may last from minutes to hours, to long-term memory (LTM), which may last from days to a lifetime (McGaugh, 2000; Kandel, 2001; Dudai, 2004). The time frame of the “consolidation phase” can vary within a given learning paradigm; it depends on the manipulation, and may reflect several different cellular and molecular processes (Figure 2). One may ask whether the difference between consolidation and maintenance of memories is an artificial one, since the processes can also be regarded as continuous processing of the information by varied molecular and cellular mechanisms within a certain cortical area. They can also be considered in terms of temporally differential involvement of several brain structures.

The next phase of memory processing is use or retrieval of the memory, during which it is susceptible to further changes, associated with a process of reconsolidation, which mediates its restabilization. The stability of a memory depends on its age: new memories are sensitive to post-reactivation disruption, but older ones are more resistant (Alberini, 2011).

Ethologically, learning is usually classified into non-associative (habituation and sensitization), associative (relationships between amounts and events), and incidental learning (learning in the absence of explicit external reinforcement; Gibb et al., 2006; Morris, 2006; Miranda et al., 2008; Rosenblum, 2008; Lindquist et al., 2009). Over the years, different behavioral paradigms have been developed in order to study the different types of learning, and also to address the abovementioned temporal phases of memory formation.

One of the most widespread taste-learning paradigms is the negative-learning, conditioned taste aversion (CTA) paradigm, in which an association is formed between a novel taste (serving as a conditioned stimulus – CS) and malaise (serving as an unconditioned stimulus – US), resulting in the animal’s subsequent avoidance of the novel food associated with delayed food poisoning (conditioned response – CR; Garcia et al., 1955; Bures et al., 1998). The acquisition of CTA is subserved by specific brain regions, including the IC and the amygdala, although their precise role in CTA is still unclear (Yamamoto et al., 1994; Lamprecht and Dudai, 1996), and it has been demonstrated that induction of neurotoxic lesions in the IC disrupts acquisition of CTA (Roman et al., 2006; Roman and Reilly, 2007). Similarly, lesions of the basolateral region (Bla), but not of the central nucleus of the amygdala, selectively disrupt CTA performance (Nachman and Ashe, 1974; Morris et al., 1999; St Andre and Reilly, 2007). CTA is commonly used when a hippocampus-independent form of learning is addressed.

A widely used paradigm of positive-learning is latent inhibition of CTA (LI-CTA), in which an animal learns that a novel food is safe, and displays less aversion following CTA than animals subjected to CTA alone (Bures et al., 1998). The LI-CTA paradigm involves presentation of a novel taste to an animal prior to CTA. Since this early experience elicits no negative consequences, the animal displays reduced aversion to the same taste following CTA (Rosenblum et al., 1993). This modulation of behavior can be attributed either to reduced strength of the association at the time of the CTA or to competition during the retrieval phase (Lubow, 1989). LI-CTA is a form of incidental learning that depends on both the quantity of novel taste consumed and the functionality of the GC (Rosenblum et al., 1993; Merhav and Rosenblum, 2008).

Under certain conditions, CTA can be extinguished (Berman et al., 2003). Extinction reflects a decrease in the CR in the absence of reinforcement of a conditioned stimulus. Behavioral evidence indicates that extinction involves an inhibitory learning mechanism in which the extinguished CR reappears following presentation of an unconditioned stimulus. However, studies have shown that the memory was not erased in rats and humans (Lin et al., 2010).

Taste learning and CTA have several advantages as paradigms for studying molecular mechanisms underlying learning and memory. These include: one-trial learning; strong incidental learning; clear and short learning time; minimal behavioral manipulation, since the animals can learn in their home cage with very little interference from other modalities; the sensory input is clearly defined, and therefore can be quantified in molecular terms; clearly defined cortical area(s) subserve the learning; and high reproducibility (Bures et al., 1998; Rosenblum, 2008).

The most studied form of neuronal adaptations is Hebbian plasticity, which includes long-term potentiation (LTP), and its reciprocal, long-term depression (LTD; Collingridge et al., 2004; Malenka and Bear, 2004; Feldman, 2009; Pozo and Goda, 2010). Hallmark features of LTP are that synaptic changes are associative, rapidly induced, and input specific. These features facilitate reinforcement of active synaptic connections with a given set of sensory stimuli, thus eliciting an activity-induced increase in synaptic efficacy, which is widely expressed in several pleo- and neocortical areas. Taken together, these characteristics render LTP an attractive model for a cellular basis for learning and memory (Bliss and Collingridge, 1993; Neves et al., 2008; Rosenblum, 2008; Sjostrom et al., 2008).

Long-term potentiation has been described in several brain areas, including the hippocampus and the cortex, and in conjunction with taste learning, (Escobar et al., 1998, 2002; Ramirez-Lugo et al., 2003; Chen et al., 2006; Alme et al., 2007; Bramham, 2007; McGeachie et al., 2011; Rodriguez-Duran et al., 2011). It was demonstrated that tetanic stimulation of the basolateral amygdaloid nucleus (Bla) induced N-methyl-d-aspartate (NMDA)-dependent but metabotropic glutamate receptor (mGluR)-independent LTP in the IC (Escobar et al., 1998, 2002; Jones et al., 1999). It is important to note that IC–LTP and CTA both involve similar molecular mechanisms in the IC, such as NMDA receptor (NMDAR) dependence, activation of ERK1/2, and induction of immediate early genes (IEGs), including Zif268, Fos, Arc, and Homer (Jones et al., 1999; Rodriguez-Duran et al., 2011).

However, it remains to be more robustly proven that LTP-like mechanisms in the IC subserve taste learning. It was shown that induction of LTP in the Bla–IC projection prior to CTA enhanced CTA retention (Escobar and Bermudez-Rattoni, 2000). In addition, it was shown that on the one hand, intracortical microinfusion of brain-derived neurotrophic factor (BDNF) induced LTP in the Bla–IC projection of adult rats (Escobar et al., 2003), and on the other hand, that intracortical microinfusion of BDNF prior to CTA training enhanced retention of this task (Castillo et al., 2006).

In a follow-up study, Rodriguez-Duran et al. (2011) have shown that when the paradigm was reversed, i.e., CTA was performed prior to LTP induction in the Bla–IC projection, CTA training prevented the subsequent induction of LTP in the Bla–IC projection for at least 120 h after CTA training. In addition, they showed that CTA training produced a persistent change in the possibility of inducing subsequent LTP in the Bla–IC projection in a protein-synthesis-dependent manner, and inferred that changes in the possibility of inducing subsequent synaptic plasticity contributed to the formation and persistence of aversive memories (Rodriguez-Duran et al., 2011).

Long-term potentiation consists of two phases on the molecular level: induction, which triggers potentiation, and maintenance, which sustains the potentiation over time. Many molecules have been shown to be involved in the induction phase of LTP, but very few have been implicated in the maintenance phase. Since the working hypothesis is that LTP is the cellular mechanism underlying LTM storage, the molecular mechanisms relevant to the maintenance phase are highly important. To the best of our knowledge, to date, only a single molecule, protein kinase Mζ (PKMζ), has been found to be necessary for maintenance in both the hippocampus (spatial learning) and the IC (CTA): local injection of ZIP, a PKMζ selective inhibitor, into the hippocampus resulted in reversal of LTP maintenance in vivo and loss of 1 day old spatial memory (Pastalkova et al., 2006). Similarly, its injection into the GC resulted in reversal of long-term CTA memory in a dose-dependent manner, whereas other serine/threonine protein kinase inhibitors are capable of interference with long-term memory formation, but are ineffective once the memory has been established (Shema et al., 2009, 2011).

Protein kinase Mζ is the brain-specific atypical protein kinase C (PKC) isoform, which unlike full-length PKC isoforms, is a cleaved form comprising the independent catalytic domain of PKCζ, and is a second messenger-independent kinase. PKMζ is constitutively active in sustaining LTP maintenance, and studies have shown that it mediates synaptic potentiation specifically during the late phase of LTP. LTP induction increases new PKMζ synthesis, leading to enhanced synaptic transmission. The mechanism underlying L-LTP and spatial memory maintenance by PKMζ is thought to involve AMPA receptor phosphorylation and trafficking, with subsequent changes in the amplitude of excitatory postsynaptic potential (EPSP) concomitant with dendritic translation regulated by PKMζ phosphorylation (and inhibition) of Pin1 (Sacktor, 2008; Vlachos et al., 2008; Navakkode et al., 2010; Parvez et al., 2010; von Kraus et al., 2010; Westmark et al., 2010; Mei et al., 2011; Sajikumar and Korte, 2011).

The working hypothesis is that an organism relies on its sensory system, specifically, in the present context, the sense of taste, to create an internal representation of a given physical or chemical stimulus. The sensory information is converted into neuronal activity that subserves the various phases of learning, i.e., acquisition, consolidation, and retrieval. Information regarding physical/chemical properties and the significance of a given taste reaches the GC via several different neurotransmitters, and elicits the release of several neurotransmitter systems in the GC, where relevant receptors are expressed as well. Several molecular changes in the GC have been found to be correlated with novel taste learning at various time points after exposure to a novel taste. The molecules involved include acetylcholine (ACh), dopamine, noradrenaline, gamma-aminobutyric acid (GABA), glutamate, and various neuropeptides (Figure 3; Rosenblum et al., 1995, 1996, 1997; Berman et al., 2000; Belelovsky et al., 2005, 2009; Koh and Bernstein, 2005; Merhav et al., 2006; Banko et al., 2007; Costa-Mattioli et al., 2007; Elkobi et al., 2008; Merhav and Rosenblum, 2008; Barki-Harrington et al., 2009b; Doron and Rosenblum, 2010). However, only the muscarinic-cholinergic and NMDARs have been extensively studied with regard to their roles in taste–memory acquisition, consolidation, and retention (Jones et al., 1999; Gutierrez et al., 2003; Nunez-Jaramillo et al., 2008; Rosenblum, 2008).

Mitogen-activated protein kinases are a family of serine/threonine kinases implicated in regulation of cell proliferation and differentiation (Seger and Krebs, 1995; Belelovsky et al., 2007). Three major groups of MAPKs have been identified in mammalian cells: extracellular signal-regulated kinase (ERK), c-jun N-terminal kinase (JNK), and p38 MAP kinase. High levels of the ERK isoforms, ERK1 (p44 MAPK) and ERK2 (p42 MAPK), have been detected in neurons in the mature CNS (Fiore et al., 1993). ERK activity has been shown to be crucial for several forms of learning and memory, including fear conditioning, CTA memory, spatial memory, step-down inhibitory avoidance, and object recognition memory. In a study in which several pharmacological agents were locally injected into the rat IC, it was found that NMDARs, mGluRs, muscarinic, beta-adrenergic, and dopaminergic receptors were all implicated in acquisition of the new taste memory, but not in its retrieval, although these neurotransmitter/neuromodulator systems differed in their role in acquisition. In addition, it was demonstrated that of all the receptors studied, only NMDA and muscarinic receptors specifically mediated taste-dependent activation of ERK1–2, whereas dopaminergic receptors regulated the basal level of ERK1–2 activation (Berman et al., 2000). Several studies have demonstrated the differential role of ERK1 and 2 in cell growth and proliferation and synaptic plasticity. For example, ERK1 knockout mice displayed enhancement of striatum-dependent long-term memory, in conjunction with facilitation of LTP in the nucleus accumbens, indicating the importance of ERK2 (Mazzucchelli et al., 2002; Vantaggiato et al., 2006).

ERK activation occurs downstream to neurotransmitter release and activation of the forebrain cholinergic neurons during and immediately after acquisition of an inhibitory avoidance response. ERK plays a major role in learning, by promoting cellular integration of divergent downstream effectors that may trigger differing responses, depending upon which subsets of scaffolding anchors, target proteins, and regulatory phosphatases are involved (Giovannini, 2006). MEK–ERK, the upstream kinase of ERK, is a crucial signal transduction cascade in synaptic plasticity and memory consolidation (Sweatt, 2001), and its inhibition affected both early and late phases of LTP in the hippocampus (Rosenblum et al., 2002).

Extracellular signal-regulated kinase activation was shown to be correlated with novel taste learning, although the amount of protein remained unchanged (Berman et al., 1998; Belelovsky et al., 2005). In addition, the expression of LTP in the GC was ERK-dependent, and operated in a positive feedback mode (Jones et al., 1999). In a recent study, it was shown that bilateral injection of U0126, a specific MEK inhibitor, to the GC prior to learning resulted in inhibition of ERK1/2, as well as attenuation of CTA (Rosenblum, 2008) and blockade of learning-induced elevation of PSD-95 3 h following taste learning (Elkobi et al., 2008).

The various MAPKs are activated within differing time periods after novel taste learning; ERK activation appears to begin a few minutes up to 1 h following novel taste consumption (Rosenblum, 2008). It has been shown that novel taste consolidation requires the elevation of ERK1/2 activity in the IC 20 min after taste consumption. However, JNK1/2 was activated 1 h after novel taste learning, whereas p38 was not modified at any of the examined time points (Berman et al., 1998). Interestingly, the time scale of ERK activation was species-specific; it was shown to differ between mice and rats (Swank and Sweatt, 2001). ERK expression following taste learning is not only time-restricted, but also space-restricted; it was activated in the GC, but not in the hippocampus, after the same length of time following learning (Yefet et al., 2006). The various MAPKs also differed in BDNF-induced activation: intrahippocampal microinfusion of BDNF that aimed to induce LTP resulted in rapid phosphorylation of ERK and p38, but not of JNK (Ying et al., 2002). These effects were observed in the dentate gyrus, but not in other examined hippocampal regions, and were shown to be MEK–ERK-dependent.

Whereas the upstream regulation of ERK in the GC is well studied, our knowledge regarding its downstream targets remains fragmentary. MAPK substrate, ELK-1, a transcription factor regulating immediate early expression of genes via the serum response element (SRE) DNA consensus site (Besnard et al., 2011), has been shown to be phosphorylated in a time frame similar to that of ERK activation after novel taste learning, and neurotransmitters required for induction of LTM are also required for ELK-1 phosphorylation in the GC in response to taste learning (Berman et al., 2003). Further studies are needed to identify other possible targets of ERK and their mechanisms of action underlying taste processing in the GC.

The distinctive biochemical characteristic of memory consolidation is its dependence on synthesis of functional proteins in the relevant brain regions (Davis and Squire, 1984). Indeed, local application of anisomycin, a protein-synthesis inhibitor, to specific brain regions affected CTA in a dose-, site-, and time-dependent manner (Rosenblum et al., 1993; Meiri and Rosenblum, 1998). For instance, local application of anisomycin to the GC attenuated CTA and taste learning under the LI paradigm (Rosenblum et al., 1993); however, the same treatment had no effect on STM (Houpt and Berlin, 1999), which suggests that short-term taste memory is independent of protein synthesis. Temporally, taste learning is sensitive to protein-synthesis inhibitor(s) from just before learning until up to 100 min afterward (Rosenblum, 2008). In addition, extinction of CTA is dependent on functional protein synthesis in the prefrontal cortex (Akirav et al., 2006). It is generally accepted that protein translation affects LTM consolidation by modulation of synaptic strength, since protein-synthesis inhibitors prevented transformation of early LTP to late LTP. However, other modifications of intrinsic neuronal properties also subserve learning-related behavioral changes. For example, learning-induced reduction in the postburst after hyperpolarization (AHP) lasts for days after completion of training, and is implicated in maintenance of learned skills. It has been recently demonstrated that synaptic activation-induced short-term postburst AHP reduction in hippocampal and pyramidal neurons could be transformed to a protein-synthesis-dependent long-term AHP reduction that persisted for long time periods (Cohen-Matsliah et al., 2010). Such learning-induced AHP reduction has been shown to be maintained by persistent activation of PKC and ERK in piriform cortex neurons (Seroussi et al., 2002; Cohen-Matsliah et al., 2007), but has been found to be CaMKII-independent (Liraz et al., 2009).

Translation consists of three phases: initiation, elongation, and termination. Of these, the initiation phase is the most tightly regulated, and is affected by phosphorylation of initiation factors (IFs) and ribosomal proteins (Proud, 2000). Initiation in eukaryotes serves as the rate-limiting step in protein synthesis, and therefore serves as an important target for translational control (Costa-Mattioli et al., 2009). There is considerable evidence in the literature that increased initiation results in enhanced learning, and vice versa. In knockout mice lacking the translation repressor eukaryotic initiation factor 4E-binding protein (4EBP2; Banko et al., 2007) or with reduced phosphorylation of eukaryotic initiation factor 2α (eIF2α; Costa-Mattioli et al., 2007), and therefore with enhanced initiation, no differences in taste recognition in parallel to enhanced CTA learning were observed; furthermore, other types of learning and plasticity in the consolidation phase were enhanced as well. Conversely, knockout mice lacking either S6K1 or S6K2, characterized by reduced initiation rates, exhibited impaired taste learning (Antion et al., 2008). In a model analogous to CTA in the chick, eukaryotic translation initiation factor 2B (eIF2B) was found to be both correlated with and necessary for taste–memory consolidation (Tirosh et al., 2007).

The second phase of translation, elongation, requires activity of elongation factors (EFs). Eukaryotic elongation factor 2 (eEF2) mediated ribosomal translocation (Ryazanov and Davydova, 1989) and was phosphorylated on Thr56 by a specific Ca²⁺/calmodulin-dependent kinase. Phosphorylation of this kinase inhibited its activity and led to general inhibition of protein synthesis (Nairn and Palfrey, 1987). It has been shown that following novel taste learning, eEF2 phosphorylation was increased in the GC, indicating attenuation of translation elongation (Belelovsky et al., 2005). This finding, which is counter-intuitive, implies that the situation is more complex than implied by the simple model of “the more IFs, the better the taste learning” (Rosenblum, 2008). Nevertheless, there was increased initiation in the same samples, manifested as increased phosphorylation levels of ERK and S6K1 (Belelovsky et al., 2005).

Therefore, we have proposed a putative mechanism, by which increased initiation concomitant with decreased elongation in the same neurons in the GC might increase expression levels of mRNAs that are poorly initiated. This was reflected, for example, in the case of phosphorylation of the α subunit of eIF2 (eIF2 α), which, in turn, led to suppression of general translation (Hinnebusch et al., 2000), concomitant with stimulation of translation of ATF4 (Lu et al., 2004; Vattem and Wek, 2004), which is required for late phase LTP and LTM (Bartsch et al., 1995; Chen et al., 2003). The suggested mechanism could perform a switch-like function in expressing a specific set of mRNAs within a restricted time window in a cellular microdomain/s such as the synapse.

Immediate early genes, activated transiently and rapidly by various cellular stimuli, are regulated at the transcription level. They provide a first response to these stimuli, in advance of protein synthesis. Many of the IEGs are transcription factors or other DNA-binding proteins (Plath et al., 2006). Expression levels of IEGs have been shown to increase in neuronal populations that subserve stimulus encoding in response to various kinds of behaviors (Campeau et al., 1991; Guzowski et al., 2005; Mattson et al., 2008; Koya et al., 2009). Several such IEGs have been implicated in taste memory, including cFOS, Activity-regulated cytoskeleton-associated protein (Arc/Arg3.1), Homer, and BDNF (Saddoris et al., 2009; Doron and Rosenblum, 2010), and have been shown to be elevated following other forms of learning as well. The aforementioned IEGs belong to a subclass termed effector neuronal IEGs, which mediate NMDAR-regulated phenotypic changes in the brain (Kaufmann and Worley, 1999). Many of these have been immunohistochemically detected in dendrites (Lyford et al., 1995; Tsui et al., 1996).

In all, a more thorough understanding of the molecular mechanisms underlying taste memory and of the functioning of the corresponding regulatory processes in relevant neurons that subserve taste-learning processing circuits should help to elucidate many important basic aspects of neuronal function. The main future questions and the directions of future research might be more general in their nature, reflecting the potential development of knowledge and understanding in the field of biological mechanisms of learning and memory, or they might address questions specific to the taste systems in the mammalian/human brain.

For example, dissection of the temporal dynamics of any memory – acquisition/consolidation and the establishment of remote memory retrieval, relearning, and reconsolidation – which was outlined in Figure 2 can be misleading. It is clear that acquisition and retrieval are well defined in time. However, the consolidation phase is defined mainly in negative rather than positive terms; it is sensitive to various disruptions at different time points following acquisition, and the connections between the various molecular entities involved in the process of taste learning are not well described. In Figure 3 we outline some of the processes known to be taking place in the GC following novel taste learning. This off-line processing of information in the brain may represent a more continuous rather than a phasic process. However, the mechanisms and molecular/cellular participants in these processes are yet to be identified (Figure 3). It is clear that in order to better understand the consolidation or off-line processes, it is necessary to enhance the measurements of the biochemical factors that are correlated with and necessary for taste learning, and to improve their resolution to a few cubic micrometers or to cellular/subcellular levels. The tools for this kind of in vivo single-cell analysis have been dramatically improved recently, and will be used to obtain better descriptions of the molecular and cellular mechanisms underlying learning and memorizing processes. Moreover, we and others will aim to identify the specific circuit or neuronal ensemble involved in creation and maintenance of any given memory and its value.

Other aims, more specific to the taste system, will be to understand the neuronal mechanisms underlying the taste–memory condition of waiting “on hold” for many hours, with its physical and chemical information as a conditioned stimulus, pending arrival of the unconditioned stimulus or the digestive information that will enable the taste to be tagged as safe or dangerous. This specific and unique ability of taste-learning beautifully represents the flexibility of the neuronal system underlying learning and memory processes, and can also teach us about the limits of the neuronal systems abilities to create simple and associative memories.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.