Conditioned Taste Aversion

From MagnetoWiki
Jump to: navigation, search


This article describes some of the fundamental behavioral properties of conditioned taste aversion learning. The article is a work in progress -- in particular, no specific citations are provided yet.

As a form of associative learning by which an animal avoids a novel taste or food that has been previously paired with a toxin, CTA has been widely used as a marker of aversive effects caused by drugs and treatments. CTA learning is remarkably sensitive, and often reveals a treatment effect even when no other behavioral effect is detectable. Purely exteroceptive sensory stimuli, such as light or auditory cues produced by the MF or the apparatus, are likely to be ignored because CTAs overwhelmingly favor a novel taste as the conditioned stimulus [53]. Likewise, stressful effects of the exposure procedure, such as restraint or cutaneous discomfort, are not sufficient to act as the unconditioned stimulus in CTA learning [39, 53]. Aversive effects that activate interoreceptors or induce nausea in humans, however, are favored to induce CTAs [53].


Conditioned taste aversion (CTA) is a form of associative learning in which an animal learns to avoid and reject a food after the taste or flavor of the food is paired with a toxic consequence. For example, an animal encounters a new food with a distinctive (but not necessarily aversive) taste. The animal eats the food, but subsequently gets sick. The animal will associate the taste with the sickness, so that the next time it encounters the food, it will avoid it, or eat less of it. This form of learning has obvious advantages for survival, and animals are extremely good at learning what foods are safe to eat, and what foods are toxic and should be avoided.

The animal can learn a CTA if food is actually toxic, or under experimental conditions in which the food is not toxic, but we administer a separate toxin to the rat after it eats the food. We typically use a sweet tasting solution (e.g. 5% sucrose or 0.1% saccharain) paired with a mildly toxic dose of lithium chloride (0.15M LiCl; 12 ml/kg i.p.). Rats, like most animals, will readily consume a sweet solution the first time they taste it. After letting the rats drink sucrose for a few minutes, we inject the LiCl which makes the rat mildly sick for a couple of hours. When we give the rats sucrose to drink the next day, they drink little if any of the solution. (Control rats get sucrose paired with a non-toxic injection of 0.15 M NaCl; because the rats don’t get sick after the sucrose, they will readily drink it the next day.)

CTA is a form of associative learning, meaning that it involves a learned change in response to a stimulus after the association of the stimulus with another effect. (This is opposed to 1) non-associative forms of learning, such as habituation or sensitization, that rely on repetition of the stimulus to induce a change in response, or 2) declarative or episodic memory (loosely called cognitive or conscious learning.) As a form of classical (Pavlovian) conditioning, CTA learning is almost reflexive in its acqusition and expression, without the requirement for conscious thought.

The components of CTA learning parallel the components of Pavlovian conditioning: The unconditioned stimulus (US) produces an unconditioned response, but after pairing with an unrealted or neutral conditioned stimulus (CS), the CS will come to elicit a conditioned response (CR) that more or less resembles the UR.

In Pavlovian conditioning:

US -> UR
meat -> salivation
CS -> no response
bell -> no response
CS + US -> UR
bell + meat -> salivation
CS -> CR
bell -> salivation


US -> UR
LiCl -> toxic effect
CS -> ingestion
sucrose -> ingestion
CS + US -> UR
sucrose + LiCl -> toxic effect
CS -> rejection
sucrose -> rejection

CTA differs from classical pavlovian conditioning in 2 important ways:

1. the CS + US must occur almost simultaneously in pavlovian conditioning, with very little delay between the two stimuli (milliseconds to seconds.) In CTA learning, an interval of up to 12 h can be occur between taste and toxin. Thus, CTA learning is “long-delay” learning.
2. the CR is qualitatively similar to the UR in pavlovian conditioning: the dog will salivate to the bell, just as it salivated to the meat (although it may salivate less to the bell than to the original meat.) In CTA learning, the CR and UR are different: LiCl does not induce a rejection response, it just makes the rat sick: However, when LiCl is paired with sucrose, the subsequent behavioral response of the rat to sucrose is reversed and the rat expresses a CR unrelated to the UR it expresses to LiCl. This second distinction is subtle, but it may turn out to be important.

Some people have proposed finer classification systems, such as conditioned taste avoidance in which the animal only passively avoids or rejects the taste but does not actively reject the food, as measured with active rejection responses during an intraoral infusion (see the work of L. Parker and K.P. Ossenkopf) . However, I am more comfortable refering generically to CTAs in which the animal avoids ingestion of the food, or decreases the amount of food ingested, or actively rejects the food. In all three cases there is a decrease in the rat’s acceptance or preference for the food: it is not clear that the apparent quantitative differences in CTA strength reflect qualitatively different processes.

Properties of a CTA

CTA learning has a number of properties that distinguish it from other forms of associative learning. Animals may reject a food for many reasons (e.g. satiety, innate aversion, or non-CTA forms of learning). In order for an animal’s rejection of a food to be labeled a CTA, the following properites of acquisition should be demonstrated:

1. The CTA is acquired best with novel tastes.

A rat or other animal learns a CTA best when the toxic effect is paired with a novel taste the rat has never before encountered before. Put another way, if the toxin is paired with a familiar taste the rat has tasted many times before without getting sick, the rat is less likely to learn a CTA (or, the magnitude of the CTA will be attenuated when a familiar taste is paired with a toxin). This is called “learned safety”: the rat has learned that a taste is not associated with illness by non-contingent pre-exposure (i.e. the experience of the taste was not contingent on getting sick afterwards) that a taste is not associated with illness: this makes the rat resistant to CTA against that taste. Thus, If learned safety does not attenuate the animal’s rejection of a food, the animal is not rejecting the food because of a CTA.

2. The conditioned stimulus is an orosensory stimulus.

A CTA is the pairing of a toxic effect with a taste, flavor, or oral somatosensory stimulus. Historically, it has been very difficult to associate the effects of a interoceptive toxin with a conditioned stimulus that works through non-oral modalities (e.g. pairing a visual, auditory, or somatic stimulus with a toxic injection of LiCl).

An important consequence of this property is that CTA learning is relatively resistant to contextual conditioning or modulation. Context is a nebulous term refering to multimodal sensory input derived from the environment of the experiment as a whole. Thus the spatial properties, visual characteristics, ambient noise, etc. of the testing situation make up the context; but for CTA learning, the context is far less important than the orosenory stimulation. Thus, if the animal's rejection of a food can be attributed to a learned response to contextual, non-orosensory stimuli, then the behavior is not due to CTA.

(Note that other species may be more sensitive than rats to sensory modalities outside the mouth and nose. For example, birds and primates can easily associate visual cues with intereoceptive stimuli to form visual aversions. Presumably the ability to learn food aversions would evolve to take advantage of the primary senses involved in food detection.)

3. The toxic unconditioned stimulus is an intereoceptive, not exteroceptive, stimulus.

Parallel to property number 2, CTA requires that the toxic, negative or aversive stimulus paired with the taste be an interoceptive stimulus. An interoceptive stimulus acts on internal receptors (such as nerves in the gut), while exteroceptive stimuli are detected by receptors on the external surface of the animal (like somatosensation) or that can detect stimuli at a distance (like vision or audition). Thus, a rat finds cutaneous shock, intense light flash, or a loud bang to be aversive stimuli – but the rat finds it very difficult to associate taste with any of these exteroceptive stimuli. Conversely, the rat associates very easily a taste with toxic postingestive effects (such as intragastric copper sulfate that irritates the gut and the sensory vagus.) A convenient consequence of this is that rats do not form CTAs because of the prick of the injection needle!

Although this definition by receptor location sounds clear cut, the interoceptive stimulus is is an ill-defined concept. The essential characteristics of intereoreceptive stimuli that induce CTA have not been rigorously delineated. In addition to literal toxins, internal stimuli as diverse as gastric distention, rotation, or depletion of metabolic fuels can induce CTAs. Other treatments that clearly affect internal state are not sufficient, however, such as restraint stress (which elevates heart rate, glucocorticoid secretion, and gastric acid secretion). It would be interesting to test whether other forms of homestatic stress (e.g. cold stress) or internal afferent stimulation (e.g. intraintestinal capsaicin) are sufficient to induce CTA.

A general rule of thumb is that anything that makes a human feel sick will be sufficient to induce a CTA in a rat. The converse, however, is not true: rats have acquired CTAs to tastes paired with saline, and with drugs of abuse that are usually self-administered, and with other compounds that have little or no negative effect in humans. Jim Smith has an excellent defnition of this property: anything that makes a rat feel “different “ after a meal will be sufficient to induce a CTA.

Thus, if the rat’s rejection of the food is due to learning due to exteroceptive stimulation, the behavior is not due to CTA

[Properties 2 and 3 are probably true but they are based on only a few early papers: they need to be confirmed in a rigourous (modern?) fashion!]

4. The CTA can be acquired in a single trial (with sufficiently strong stimuli)

There are other forms of single-trial learning, such as fear conditioning (pairing a tone with foot-shock.) However, most forms of classical conditioning (such as pavlovian salivation or eyeblink conditioning) require dozens or hundreds of pairings across many days to achieve a robust conditioned response. This makes CTA a convenient model system to study, because the changes in the brain during learning must be relatively large and rapid to persist after only a single trial.

[Note that although CTA can be acquired in a single trial with a strong toxic US, a CTA can also be acquired after multiple pairings of the taste with a weaker stimulus.]

5. The CTA tolerates a long delay between taste and toxin

CTA can be aquired even with an interstimulus interval of minutes to hours (as opposed to the upper limit of milliseconds to seconds required by other forms of classical conditioning ). There is at least one report that CTA learning is optimized for a long-delay: in other words, rats learn a CTA better with if there is a delay of 15-30 min (vs a delay of seconds) between the taste and the toxic effect.

This property is quite unique to CTA learning, and has important implications. It makes functional sense: after eating a toxin, the postingestive effects might no be apparent for minutes to hours later, so the animal needs to associate the taste with a delayed effect. However, it might also imply that CTA learning is mediated by unique neural and molecular mechanisms that tolerate long delays (or at least that familiar synaptic mechanisms have been adapted to this particular temporal niche.)

[Note that we may be able to take advantage of the long-delay characteristics: for example, we should be able to intervene between the taste and toxin, e.g. with protein synthesis inhibitors to block gene expression induced by the taste independent of the expression induced by the toxin. At the very least, we can pursue the underlying molecular mechanisms at a leasurely pace, knowing that the taste-induced changes they must persist for a couple of hours…]

These 5 properties define CTA learning as single-trial, long-delay associative learning between taste and toxin. Therefore, in order to establish that a new phenomenon is CTA learning, it is necessary to demonstrate that it possesses these 5 properties. [Of course, additional properties might also be required, for example taste specificity or the ability to extinguish the CTA slowly with unpaired exposures to the taste.]

Importantly for future research, the same properties should be established for non-behavioral correlates of CTA learning, such as physiological or neural responses that parallel the behavioral learning. For example, the taste of sucrose induces c-Fos expression in the NTS of conditioned rats (but not unconditioned rats). Because this neural correlate of CTA expression has met the above 5 properties, we believe it is a true correlate of the associative processes of CTA learning.

Related articles include:

Preferences and Aversions
Measuring CTA Expression
Conditioned and Unconditioned Stimuli.