A variety of factors have been shown to be important in the formation of an association in Pavlovian learning. Here we review some of those important factors. In general, investigators have been interested in finding the condition thought to be both necessary (i.e., the given factor is required for learning to occur) and sufficient (i.e., given the presence of the factor in question, learning will inevitably follow) for learning to occur. The research, however, has more frequently taken the form of a result pointing to the insufficiency or non-necessity of a given factor. Below is the list of important factors.
1. Stimulus Novelty - Learning is faster when the CS & US are novel from the outset of conditioning.
a. CS Preexposure Effect (sometimes known as latent inhibition) Ð This illustrates the importance of CS novelty in Pavlovian learning. Learning a CS-US association is shown to be slow in a group exposed to the CS (without the US) prior to the conditioning phase, relative to a control group not pre-exposed to the CS. Since CS preexposures slow subsequent excitatory and inhibitory conditioning, it appears as though conditioned inhibition is not the source of the interference that is observed with nonreinforced CS preexposures.
b. US Preexposure Effect Ð This illustrates the importance of US novelty in Pavlovian learning. Learning a CS-US association is shown to be slow in a group exposed to the US (without the CS) prior to the conditioning phase, relative to a control group not pre-exposed to the US.
c. Learned irrelevance Ð This phenomenon demonstrates that learning is severely impaired in a group pre-exposed to a random contingency between CS & US prior to a conditioning phase. The impairment is more severe than that produced by CS or US preexposures by themselves. This gives rise to the question of whether learned irrelevance reflects the sum of the CS & US preexposure effects or something different.
2. Stimulus Intensity - The rate at which learning occurs, as well as its asymptote, has been shown to be influenced by stimulus intensity.
a. CS intensity primarily affects the rate of learning. More intense CSs are learned about more rapidly.
b. US intensity affects both the rate and asymptote of learning.
3. Temporal Contiguity - The time between CS & US (measured by CS-US interval or Inter-Stimulus-Interval) has been long thought to be critical for Pavlovian learning to take place.
a. The graph showing conditioned responding as a function of the CS-US interval illustrates that temporal contiguity is important in Pavlovian learning.
b. Relative temporal contiguity. Kaplan's study (1985) illustrates that CS-US contiguity relative to the intertrial interval influences learning. He demonstrated that the same trace conditioning procedure could result in excitatory (conditioned approach to the keylight) or inhibitory (conditioned withdrawal from the keylight) conditioning, depending upon the time between trials. If absolute temporal contiguity were fundamental, then the two groups should have displayed similar learning. Instead, relative temporal contiguity seems important here.
4. Spatial Contiguity - Closeness of the CS & US in space is also important. Rescorla & Cunningham (1979) demonstrated that for 2nd order conditioning, at least, learning was faster when the 2nd and 1st order keylights were presented to the same key compared to when they were each presented to spatially distinct keys.
5. CS-US Contingency - This idea gave us a common framework within which to understand excitatory and inhibitory conditioning. According to this view, excitatory and inhibitory learning occurred whenever the conditions of the experiment mapped onto different regions of the contingency space (above or below, respectively, the zero contingency line). The framework also gave us some ideas about what the best control group might be for assessing excitatory or inhibitory learning Ð the zero contingency control.
a. Rescorla's (1966; 1968) studies clearly demonstrated that CS-US pairings were not the only thing that mattered for learning to occur. The number of USs occurring in the inter-trial interval is also important. He showed that a group exposed to CS-US pairings with no intertrial USs learned more than a group exposed to the same number of CS-US pairings but with additional intertrial USs. The likelihoods of the US given a CS and given no CS were equal in the zero contingency group, but the likelihood of the US given a CS was greater than the likelihood of the US given no CS in the positive contingency group. Because each group was exposed to the same number of CS-US pairings but learned to different degrees, it can be said that temporal contiguity is not sufficient for learning to occur. Maybe contingency is?
6. US Surprise - Kamin (1969) argued that Pavlovian learning will occur to the extent that the US is surprising.
a. This argument was made on the basis of his discovery that learning will not occur to a stimulus that is conditioned along with another stimulus that, itself, previously received conditioning. In his study, one group of rats was trained with Tone -> Shock pairings in phase 1. A second group did not receive any tones or shocks during this phase. During phase 2, both groups were trained with Tone+Light -> Shock pairings. In the test phase, the second group displayed more fear conditioning to the Light CS than did the first group. It is said that prior conditioning to the Tone blocked learning to the Light in the pretrained group. This may occur, as Kamin suggested, because the US was no longer surprising during phase 2 in the pretrained group.
b. Another interpretation is based on Rescorla's contingency analysis. Since the pretrained group in Kamin's experiment was exposed to the shock occasionally during phase 1 in the absence of the Light, it could be argued that this would degrade the contingency between Light & Shock. Thus, the pretrained group should show less learning to Light, not because the US is surprising, but because the Light-Shock contingency is less in the pretrained group. Rescorla (1971) examined this by comparing three groups of rats, each of which was exposed equally to shock (in the absence of Light) during phase 1. The first group received Tone -> Shock pairings (as in Kamin's study). The second group received Shock alone presentations, and the third group received Tone and Shock explicitly unpaired. During phase 2, all three groups received Tone + Light -> Shock pairings. Of most interest were the results in the test phase. Group 1 showed the least amount of fear conditioning to Light, Group 3 the most, and Group 2 somewhere in the middle. This experiment strongly suggests that the blocking effect is not caused by a decreased Light -> Shock contingency in the pretrained group. Instead, it looks like US surprise is more fundamental in Pavlovian learning. In this experiment, Group 1 should be the least surprised when they receive the shock during phase 2. Group 3 should be the most surprised (because the explicitly unpaired procedure should ensure that the tone was a signal for no shock), and Group 2 should be somewhere in the middle.
c. Another important aspect to this experiment is that (just like Rescorla's contingency study) the results imply that CS-US contiguity is not sufficient for Pavlovian learning to occur. This is so because the Light was equally contiguous with Shock in each of the groups studied in the Kamin and Rescorla (1971) experiments. If temporal contiguity was sufficient for learning to occur, then the groups should not have differed in their learning to Light. They clearly did, and this establishes that something else is more fundamental for learning - US surprise!
d. Wagner, Logan, Haberlandt, & Price (1968) - Relative cue validity. This experiment fits with the emerging theme that temporal contiguity is not sufficient for learning, and it supports the US surprise idea. In their experiment, two groups of rats were trained with two stimulus compounds. In the first group, AX+ and BX- trials alternated. In the second group, AX+/- and BX+/- trials alternated. When the X stimulus was tested subsequently, they discovered that the second group learned more about X than the first group. This occurred because the X stimulus was relatively a less valid predictor of the US in the 1st group compared to the A stimulus. In the second group, X was at least as valid a predictor of the US as was A. If the ÒvalidityÓ of the CS relative to other available cues is low, then less learning to that CS should occur. Notice how this result is consistent with the US surprise notion. In group 1, since A is a reliable predictor of the US, the occurrence of the US is not so surprising, ultimately, on AX+ trials. The US will be surprising on AX trials more often in group 2. Thus, the 2nd group should display more learning to X.
7. CS-US Relevance - The search for the necessary and sufficient condition for learning continues because Garcia and Koelling's (1966) experiment demonstrated that temporal contiguity is not necessary and that US surprise is not sufficient for learning. In particular, they demonstrated that certain CS-US combinations are better learned about than others. Their rats were trained with a bright, noisy, tasty solution as a signal for illness (in one group of rats) or as a signal for shock (in a second group of rats). When the rats were subsequently tested for which component of the 3-stimulus compound gained control over their behavior, they discovered that the nature of the US was important. In particular, rats conditioned with illness as the US learned to avoid the 3-stimulus compound on the basis of the taste of the stimulus. In contrast, rats conditioned with the shock US learned to avoid the 3-stimulus compound on the basis of auditory and visual components of the stimulus. Since it can be argued that the shock and illness USs were equally surprising from the outset of conditioning, there is no basis for expecting that different components of the 3-stimulus compounds should have been learned about. Instead, it is as though evolution has created a species that is more ready to learn about certain CS-US combinations than others. In other words, maybe a gut learning system has evolved in the rat that finds taste stimuli to be more relevant than auditory or visual stimuli, to gastrointestinal stimuli. Alternatively, maybe a predatory defense learning system has evolved to consider auditory and visual signals as more relevant than taste to shock USs.