Pavlovian Conditioning:  From Learning to Performance

 

 

     The question of 'How do we get from learned associations to performance?' is the next fundamentally important question that we need to address.  The issue has been approached in several different ways.  First, we consider some important factors that have been demonstrated to influence the form of the conditioned response (CR).  Next we will consider 3 different theoretical approaches to understanding why the CR takes the form that it does.  And, third, we will consider several different ways in  which the CR is functionally significant.

A.  Determinants of the Form of the CR

            1.  Nature of the US.  Perhaps the best evidence suggesting that the form of the CR is influenced by the nature of the US was provided by Jenkins and Moore (1971).  In this study, pigeons that were both hungry and thirsty participated in an autoshaping experiment in which two keylights were each associated with different appetitive USs (i.e., grain or water).  Jenkins and Moore measured detailed aspects of the keypeck CRs that occurred in response to the keylights paired with the different USs.  They observed that, irrespective of the specific color of the keylight, the CRs differed in the two cases.  In particular, the birds made keypeck responses to the CSs that resembled the unconditioned responses expected to occur to the USs themselves.  In other words, the birds pecks the keylight that was paired with grain in a manner similar to the way in which they would peck grain itself (i.e., by pecking forcefully, with wide open beaks).  Further, they would peck at the keylight paired with water in a manner more similar to how they would interact with water (i.e., by pecking much less forcefully, and with their beaks not wide open and with mumbling motions).  These data demonstrate that the form of the CR depends upon the nature of the US. 

 

            2.  Nature of the CS.  The Holland (1978) experiment makes the point simply that the form of the CR is affected by the nature of the CS.  In his experiment, Holland demonstrated with rats that when visual and auditory CSs are each paired (separately) with a food pellet US, the CRs that develop to these CSs differ in part.  A visual CS comes to evoke rearing behaviors, whereas the auditory CS comes to evoke startle and head jerk CRs.  In addition, both CSs evoke magazine approach CRs.  Since the same US is used in each case but topographically different CRs emerge, then  it can be concluded that the nature of the CS can also influence the type of conditioned responding one sees in Pavlovian learning.

 

            3.  CS-US Interval.  Timberlake, Wahl, & King (1982) demonstrated that the CS-US interval can also determine the form of the CR.  In their experiment, a ball bearing CS was paired with a food pellets with either a 2" or 6" CS-US interval in different groups of rats.  They observed their rats to quickly approach the food magazine when the ball bearing stimulus first appeared if a short CS-US interval was used.  However, if a long CS-US interval was used, then the rats approached the ball bearing, picked it up, and carried it towards the food magazine before eventually dropping it in order to get at the food once it arrived.  Clearly, the CR topographies differed in these two cases, suggesting that the CS-US interval can also influence the form of the CR.

 

B.  Theories of the CR

            1.  Stimulus Substitution Theory.  Pavlov observed that an unconstrained dog when presented with a visual CS would approach the light bulb and lick it.  Observations like this prompted him to suggest that the CS substituted for the US.  In other words, the CS acted as a surrogate US.  This idea suggests that the CR and the UR should be very similar, if not identical.  Indeed, the Jenkins and Moore (1971) experiment noted above produced results that are consistent with this view, because it demonstrated that the form of the CR closely matches the form of the unconditioned responses one observes to different USs.

            Timberlake and Grant (1975) put this idea to a test in a very unusual study.  In their experiment they used conditioned rats to anticipate a food pellet after a brief interval in two situations that differed in the type of CS that was used.  In one case, a wood block was secured to a platform and served as the CS for food.  In the other case, a live rats was secured to a the platform and served as  the CS for food.  They argued that Pavlov's stimulus substitution theory would lead to the prediction that the rats should approach and bite the CSs that were paired with food similarly in the two cases.  However, they observed that although conditioned biting CRs did occur to the wood block CS it did not occur to the live rats CS.  Instead the learning rats approached and engaged in social grooming type CRs to the live rat CS signaling food.  Thus, here is another example where the nature of the CS influences the CR form, and it more forcefully argues against stimulus substitution theory.

 

            2.  Konorski's Preparatory/Consummatory Approach.  Konorski suggested that animal behavior can often be characterized in terms of an on-going sequence of behaviors that ultimately lead to the animal obtaining or interacting with some goal object.  Behaviors that occur early in this sequence can be thought of as "preparatory" behaviors, whereas behaviors occurring late in the sequence can be thought of as "consummatory" behaviors.  Konorski further argued that the stimuli (internal and external) that control each of these two classes of behaviors also differ.  Particularly, it is suggested that general motivational states control preparatory behaviors that are themselves thought of as less focused, more general, and diffuse in character, whereas specific sensory stimuli or stimulus expectations are thought to control consummatory behaviors (i.e., those that occur when the goal object is either imminent or when the animal has contacted the goal object).

            With respect to Pavlovian conditioned responding, this distinction suggests that different CRs might be expected in a conditioning experiment depending upon which aspect of the US becomes associated with the CS.  If the CS associates with the motivational components of the US, then more diffuse preparatory CRs should be expected.  However, if the CS associates with a specific sensory representation of the US, then more specific consummatory CRs should be expected to occur.

 

            3.  Behavior Systems Theory.  Timberlake advanced a more sophisticated theory of behavior that shares some of the flavor of Konorski's approach.  Timberlake argued that the animal comes to any learning situation with an evolutionary history that has shaped and organized a pre-existing hierarchically organized network of behavior.  In fact, it is presumed that different networks have evolved for different motivational states.  One important component to this approach is that a given motivational system (like the feeding system) is broken down in terms of different sub-systems (e.g., predation or social foraging), modes (General Search, Focal Search, and Handling/Consuming), and more specific stimulus-response (S-R) modules.  The modes vary in their spatio-temporal relation to the goal object with the General Search mode being farthest from the goal object (food in this case), Focal Search being intermediate, and Handling/consuming being engaged by the immediate presence of the goal object.  The specific responses one sees depends upon, first of all, which specific mode is currently active, and, second of all, what specific stimulus the animal encounters.  Different modes are thought to be sensitive to different sets of "natural releasing stimuli," each of which are themselves assumed to evoke different specific responses.  For example, a moving stimulus is thought to be a natural releasing stimulus for chase and capture responses.  This S-R module is thought to be part of and activate the focal search mode.  Different modes, then, are partly defined by a different set of S-R modules.

            Within this framework, Pavlovian conditioning may operate in different ways.  However, one particular way in which the form of the CR may vary is in terms of which mode gets conditioned to the CS.  If the CS comes to associate, in some sense, with a general search mode, then behaviors that are distant to the goal object should be conditioned.  However, if the CS associates with a more specific focal search mode, then behaviors that are more proximal to the goal object should dominate.  Perhaps when the CS-US interval is varied what happens is that different search modes get conditioned to the CS.  In particular, Timberlake, Wahl, & King (1982) demonstrated that a ball bearing CS paired with food with a 6" (long) CS-US interval learned to approach, contact, and carry the ball bearing.  However, when a 2" CS-US interval (short) was used, the rats avoided contact with the ball bearing and instead directly approached the food magazine when the ball bearing CS appeared.  In other words, it looks as though a more general search mode was conditioned with the long CS-US interval and a focal search mode was conditioned with a short CS-US interval.

            A second important way in which the form of the Pavlovian CR may vary according to this framework is related to the similarity of the CS actually used in the experiment and the natural releasing stimulus that is unconditionally part of the behavior system.  For instance, a ball bearing CS may elicit approach and contact behaviors (with a relatively long CS-US interval) in part because this CS resembles a moving stimulus that is a natural releasing stimulus for such behaviors (e.g., a moving cricket in the ratÕs visual field).  A live rat CS, however, would come to evoke different CRs because this CS more closely resembles natural releasing stimulus that are part of a different aspect of the behavior system for feeding.  In particular, this releasing stimulus may be found within the social foraging subsystem of the feeding system and behaviors that were more appropriate for this sub-system would be expected.

 

C.  Functions of the CR.

            1.  Optimize interactions with the US.  One important function of Pavlovian conditioning is that it produces CRs that help the animal receive, process, and more generally interact with the US once it arrives.  Hollis demonstrated that the blue gourami fish was better prepared to defend its territory against intruders if it had the benefit of anticipating the intruder's arrival.  A visual CS was used to signal the sight of the intruder (10 s following the CS light turned on).  Subsequently, when the light signaled the presence of the intruder, the rats who learned about the meaning of the signal were more aggressive in defending their territory against the intruder than fish who did not learn about the meaning of the signal.  This is the "best defense is a good offense" approach.

 

            2.  Changing the affective evaluations of stimuli.  Another important function of Pavlovian conditioning is to influence the affective evaluation of stimuli.  We saw earlier how this was useful in feeding situations by allowing animals to distinguish among foods that are likely to cause them to be ill (conditioned food aversion learning) or feel pleasantly satiated (conditioned food preference learning).  Pavlovian learning ensures that animals can make more informative food choices based upon changes in the affective evaluations of food stimuli.  This is just an example.  You can imagine how affective evaluations may help the animal in other domains as well.

 

            3.  Changing the motivational status of stimuli.  We earlier considered the possibility that when a CS is paired with a US the motivational status of the CS has changed.  This should have an obvious adaptive function for the animal because it will be advantageous for animals to learn which stimuli do and do not have potent motivational meanings.  Earlier, the learning theorist Mowrer suggested a classification scheme whereby Pavlovian contingencies interacted with different motivational states to produce 4 basic motivational categories.  A positive Pavlovian contingency involving an appetitive US (e.g., food) would condition to the CS a motivational state called "Hope."  A negative Pavlovian contingency involving an appetitive US would condition to the CS a motivational state called "Disappointment."  Postive and negative Pavlovian contingencies involving an aversive US (e.g., shock) would condition to the CS, respectively, motivational states called "Fear" and "Relief."

            Konorski and Dickinson and Dearing (1979) went on to suggest that there are really two fundamentally distinct motivational states in this scheme.  They suggested that Hope and Relief are functionally equivalent and that Fear and Disappointment are functionally equivalent.  In essence, there are separate positive (hope and relief) and negative (fear and disappointment) motivational states.

            Several investigators went on to examine if these equivalences can be found experimentally.  The logic of the approach is to determine if a CS based on conditioning of one type of motivational state can block conditioning of a functionally equivalent motivational state to another CS (in a Kamin blocking design) even though the USs themselves may differ.  For instance, Kamin demonstrated that prior conditioning of a Light CS with a Shock US resulted in the Light being able to block aversive conditioning to a Tone CS when the Light+Tone compound later signaled shock.  Dickinson & Dearing (1979) reported that an appetitive inhibitor also acquired an ability to block aversive conditioning to a stimulus when a stimulus compound consisting of the appetitive inhibitor and a new stimulus signaled the shock US.  This result suggests that an aversive exciter (fear) and an appetitive inhibitor (disappointment) both evoke motivational states that have some commonality.  This commonality is presumed to be the basis for the blocking effect.

            Similarly, another pair of studies provided evidence to support the view that appetitive excitors and aversive inhibitors function similarly.  Rescorla (1971) demonstrated that when an aversive inhibitor was combined with a second CS with the compound now being paired with shock, there was stronger aversive conditioning to the new stimulus.  What Dickinson (1977) added to this was the finding that an appetitive excitor also had the same effect as the aversive inhibitor did in Rescorla's study.  In phase 1 of Dickinson's study, CS1 was paired with a food pellet.  During phase 2, CS1 was combined with a 2nd stimulus, CS2, and the stimulus compound (CS1+CS2) was paired with shock.  What Dickinson observed was that CS2 acquired more conditioned fear in this group compared to a control group that received random pairings of CS1 and food during phase 1.  In the two studies, it looks like an aversive inhibitor and an appetitive excitor both function to support greater aversive learning to CS2.  This presumably occurs because the aversive inhibitor and the appetitive excitor evoke functionally similar motivational states (relief and hope, respectively).