Chapter Outline

Chapter Outline

1. THE GENERALIZATION PROCESS

   Generalization is the process that allows responding to stimuli that are similar to the cues used in original training. Discrimination is the process of responding to some stimuli but not others. Generalization and discrimination are studied by plotting a generalization gradient which shows the relationship between the degree of change in the original conditioning stimulus and response strength.

   1. Generalization Gradients: Generalization gradients have been plotted for stimuli based upon excitatory conditioning. There is also a limited amount of research studying inhibitory generalization gradients.

      1. Excitatory-Conditioning Generalization Gradients

         Much of the research on excitatory generalization gradients has used pigeons as subjects. Guttman and Kalish (1956) performed the classic excitatory generalization experiment. They trained different groups of pigeons to peck a stimulus key that varied in its wavelength. After training, each group was given test stimuli, including wavelengths that were higher and lower than the original stimulus. Response levels were measured for each test stimulus. Guttman and Kalish consistently observed symmetrical generalization gradients with the highest level of responding occurring at the original training stimulus.

         The same type of excitatory generalization gradients found by Guttman and Kalish are common in other species, including humans.

         Flat generalization gradients can be observed under certain circumstances.

      2. Inhibitory-Conditioning Generalization Gradients

         Weisman and Palmer (1969) studied inhibitory generalization gradients. Pigeons were first trained to peck a green disk (S+) to receive food reinforcement. When a white vertical line (S-) was presented, the pigeons were not reinforced for key pecking. Inhibitory generalization tests were conducted with other lines that differed from the original vertical line by various degrees of arc. The results showed the greater the difference between the original vertical line and the test line, the less was key pecking inhibited.

   2. The Nature of the Generalization Process: The Lashley-Wade theory of stimulus generalization proposes that generalization occurs when the subject cannot distinguish between the test stimulus and the conditioning stimulus. Several lines of research support this explanation.

2. DISCRIMINATION LEARNING

   Discrimination learning involves the use of environmental cues to signal the availability or unavailability of positive and negative reinforcement as well as punishment. In discrimination training, the symbol S^D is used to indicate the availability of reinforcement contingent upon the occurrence of an appropriate response. The symbol S ^Δ is used to indicate that reinforcement is not available and a response will not be effective in producing reinforcement. When operant behavior is under the control of a discriminative stimulus (either S^D or S^Δ ), Skinner called the behavior a discriminative operant.

   1. Discrimination Paradigms

      1. Two-Choice Discrimination Tasks

         In the two-choice discrimination learning situation, the S^D and the S^Δ are within the same stimulus dimension. In this task, subjects usually begin by responding equally to the S^D and the S^Δ . However, with continued training, the response rate to the S^D increases and the response rate to the S^Δ decreases.

      2. Conditional Discrimination Task

         In a conditional discrimination task, the availability of reinforcement to a particular stimulus depends upon the presence of a second stimulus.

   2. Behavioral Contrast: In discrimination tasks, the increased responding to the S^D and decreased responding to the S^Δ is called behavioral contrast. Flaherty suggested that two types of contrast occur. Local contrast is activated by the sudden change in stimulation (S^D to S^Δ or vice versa) and is due to a temporary emotional change (elation or frustration). Sustained contrast (sometimes called anticipatory contrast) occurs when animals anticipate an impending change to more or less preferred reinforcement contingencies.

   3. Occasion Setting

      1. The Occasion-Setting Function of a Conditioned Stimulus

         The ability of a CS to enhance responding to a second stimulus is called occasion setting because it can facilitate the effectiveness of the second stimulus's ability to elicit a conditioned response. This facilitation effect (occasion setting) apparently is not due to the excitatory properties of the CS. That is, it is not due simply to the ability of the occasion setter to elicit a conditioned response. Instead, the facilitating or occasion-setting property of the CS is separate from conditioned excitation. To have a stimulus function as an occasion setter, it must precede a second CS that is then paired with an UCS.

      2. The Occasion-Setting Function of a Discriminative Stimulus

         Rescorla and colleagues (1988) have shown that an occasion setting CS can raise the level of operant responding. Moreover, Pavlovian occasion setters and operant discriminative stimuli are interchangeable, suggesting a common mechanism of action.

      3. Conditioned Stimuli and Operant/Instrumental Behavior

         CSs appear to exert mediational control over instrumental performance. Rescorla and Solomon (1967) proposed that the mediational effects of CSs are tied to their ability to arouse central motivational states. Research has indicated that a CS associated with food arouses the appetitive motivational state but inhibits the aversive motivational state. Also, a CS associated with a painful event arouses the aversive motivational state while inhibiting the appetitive motivational state. Additionally, a CS predicting the absence of a reward inhibits the appetitive motivational state but activates the aversive motivational state. Finally, a CS predicting the absence of a painful event inhibits the aversive motivational state but excites the appetitive motivational state.

   4. The Nature of Discrimination Learning: How do we discriminate the S^D from the S^Δ ?

      1. The Hull-Spence Theory

         The Hull-Spence theory of discrimination learning provides an associative description of some essential features of discrimination learning.

      2. Development of Conditioned Excitation and Inhibition

         According to the Hull-Spence view, discrimination learning develops in three stages. First, conditioned excitation develops to the S^D as the result of reinforcement. Second, conditioned inhibition develops to the S^Δ as the result of non-reinforcement. Finally, conditioned excitation and inhibition have generalization gradients. The algebraic summation of excitatory and inhibitory strength determines the response rate to test stimuli.

      3. The Peak Shift Phenomenon

         The Hull-Spence theory predicts the peak shift phenomenon which refers to the development of maximum responding to a stimulus other than the S^D that is in a direction opposite to the S^Δ . Experimental support exists for the peak shift effect.

      4. The Aversive Character of S^Δ

         Terrace (1964) claims that exposure to the S^Δ produces an unpleasant frustration emotion that has the effect of increasing the response rate to other stimuli.

      5. The Transposition Effect

         Kohler developed an alternative view to the Hull-Spence theory. According to Kohler, stimuli are evaluated in relative, not absolute, terms. Kohler's view predicts the transposition effect which proposes that subjects learn the relationships between stimuli in discrimination training. Kohler's relational view is supported in discrimination tasks in which the subject must choose between two presented stimuli. However, the Hull-Spence theory is an adequate explanation when subjects are tested with only one stimulus at a time.

      6. Errorless Discrimination Training

         Under certain experimental arrangements, a discrimination can be learned with few or no errors. a result known as errorless discrimination learning.

      7. Training Procedure

         Errorless discrimination training was demonstrated by Terrace (1963) whose pigeons pecked at a red illuminated key (S^D) to receive food reinforcement. When the pigeons received early and progressive introduction to the S^Δ (a dark key lasting a short period of time), they learned to ignore the S^Δ even when it reached its full effect.

      8. Application: Errorless Discrimination Training in Humans

         Errorless discrimination training has also been observed with humans using the fading technique.

      9. Nonaversive S^Δ

         The behavioral characteristics found with standard discrimination training are not observed with errorless discrimination training. The peak shift phenomenon is not found in errorless discrimination training suggesting that the S^Δ has not acquired aversive properties.

      10. Sutherland and Mackintosh's Attentional View

          Sutherland and Mackintosh developed an attentional view of discrimination training that operates in two stages. In the first stage, the subject's attention to the relevant stimulus dimension is strengthened. In the second stage, the relevant stimulus is associated with a response.

      11. The Recognition of the Relevant Dimension

          The Sutherland-Mackintosh attentional theory claims that each stimulus can activate an analyzer that actively scans the environment searching for the relevant aspects of a stimulus configuration. Once a particular analyzer has been aroused, the output from the analyzer is associated to a particular response.

      12. Predictive Value of Discriminative Stimuli

          Results by Wagner and associates (1968) suggest that the relative predictiveness of the S^D, not the percentage of trials on which the S^D is associated with reinforcement, determines its ability to control behavior.

      13. Continuity versus Discontinuity

          The Hull-Spence view claims that excitation and inhibition gradually increase during the acquisition of a discrimination. This position is called the continuity theory of discrimination learning. Krechevsky (1932) and Lashley (1929) presented an alternative view that denies that discrimination learning is gradually acquired. Instead, subjects are presumed to form hypotheses about the relevant dimension that predicts reinforcement. The Krechevsky and Lashley view predicts that discrimination performance is exceptional once the relevant stimulus dimension is discovered. Thus, their view has been referred to the noncontinuity theory of discrimination learning.