Discriminative Stimuli, Stimulus Control, and the Three-Term Contingency

Introduction to Discriminated Operants and the Three-Term Contingency

• Operant contingencies have historically been discussed in terms of a two-term contingency involving a response and its consequence (Response-Consequence).

• However, most operant contingencies are actually three-term contingencies.

• While differential reinforcement usually focuses on the dimensions of the response, responses can also be differentially reinforced with respect to the dimensions of the stimuli in the presence of which they occur.

• A discriminated operant refers to a behavior that is under the control of a specific stimulus.

• The introductory concept is illustrated via a rat saying: "Oh, not bad. The light comes on, I press the bar, they write me a check. How about you?" (Cheney).

The Three-Term Contingency Framework

• The three-term contingency consists of a Discriminative Stimulus ($S+$ or $S^D$), a Response, and a Consequence.

• Rat Example:

  • Condition 1: Light ($S^D$) $\rightarrow$ Lever Press (Response) $\rightarrow$ Food (Consequence).

  • Condition 2: No Light ($S^\Delta$) $\rightarrow$ Lever Press (Response) $\rightarrow$ No Food (Consequence).

• Human Example:

  • Scenario 1: A friend stops speaking ($S^D$) $\rightarrow$ You reply (Response) $\rightarrow$ Continue Social Interaction (Consequence).

  • Scenario 2: While a friend is speaking ($S^\Delta$) $\rightarrow$ You reply (Response) $\rightarrow$ "Don't Interrupt!" (Consequence).

Characteristics of Discriminative Stimuli

• Discriminative stimuli do not automatically elicit responses in the same way that unconditioned or conditioned stimuli do in classical conditioning.

• Instead, they signal the availability of reinforcement; they do not "trigger" the response.

• Discriminative stimuli are said to "set the occasion" for the response.

Measuring Discrimination and Stimulus Control

• In a laboratory setting, a light may function as a discriminative stimulus ($S+$ or $S^D$) correlated with a Variable Interval (VI) reinforcement schedule.

• The absence of the light functions as a discriminative stimulus ($S_{delta}$ or $S^\Delta$) correlated with extinction.

• Over successive sessions, lever pressing typically increases during the presence of current light and decreases during its absence.

• Discrimination Index: This measures the degree to which responses become sensitive to the presence of the light. It is calculated as the percentage of total responses occurring in the presence of the light.

  • $\text{Discrimination Index} = \frac{100 \times S^D \text{ rate}}{\text{Total response rate}}$

• Descriptive terminology for this phenomenon:

  • Lever pressing in the presence of light is a "discriminated operant."

  • The light functions as a "discriminative stimulus" for lever pressing.

  • Lever pressing is under the "stimulus control" of the light.

Stimulus Control Gradients and Peak Shift

• Generalization Gradients: If a response is reinforced in the presence of a single stimulus and a specific property of that stimulus (such as color wavelength) is varied, the response rate will depend on the physical difference between the original stimulus and the current stimulus.

• Post-Discrimination Gradients: These are stimulus control gradients observed after an organism has been trained to discriminate between two or more stimuli along a single dimension.

• Peak Shift (Hanson, 1959):

  • In a study with pigeons, the $S^D$ was light at a wavelength of $550\,nm$.

  • The generalization gradient (red) shown in the graph peaks precisely at the intended $S^D$ ($550\,nm$).

  • The post-discrimination gradient (green) appears after reinforcement at $550\,nm$ ($S^D$) and extinction at $570\,nm$ ($S^\Delta$).

  • The extinction at the second stimulus ($S^\Delta$) "pushes" the generalization away from the $570\,nm$ end of the light spectrum.

  • Maximum responding (the peak) shifts off the precise $S^D$ wavelength and away from the $S^\Delta$.

  • Peak shift suggests that discrimination involves sensitivity to the physical relationship between the two stimuli rather than just responding to two independent stimuli.

  • Note: Because of this shift, the pigeon is actually not earning the maximum amount of reinforcement available at the original $S^D$.

Fading: Stimulus Control by Successive Approximations

• Just as shaping changes response properties, fading gradually changes the stimulus properties that define a discriminated operant class.

• Training begins with easy-to-discriminate stimuli and then gradually moves to more difficult stimuli rather than exposing the organism to difficult discriminations immediately.

• Fading Procedures with Pigeons:

  • Step 1: Easy-to-discriminate Stimuli: $S+$ (Red) and $S-$ (Green).

  • Step 2: Superimpose Difficult-to-discriminate Stimuli: $S+$ (Red with a Vertical Line) and $S-$ (Green with a Horizontal Line).

  • Step 3: Gradually fade the colors (Red and Green) until the animal responds solely to the lines.

• Teaching Children to Read:

  • Letters are initially written in easy-to-discriminate colors.

  • These colors are gradually faded until the child is reading the black letters alone.

Discriminable Stimuli in the Natural Environment

• Discriminated operants are behavior classes defined by the stimuli that occasion responding in daily life.

• A stimulus is regarded as a class of functionally related events rather than one particular instance.

• Traffic Light Example:

  • $S^D$: Red Light (generalizes across high and low brightness levels).

  • Response: Stop.

  • Consequence: Avoid Crash.

• Telephone Example:

  • $S^D$: Phone Ring (generalizes across high/low loudness and varying tones).

  • Response: Answer.

  • Consequence: Speak to Friend.

Complex Non-Arbitrary Discrimination and Concept Formation

• Concepts are behavioral phenomena understood in terms of generalization within a class of stimuli and discrimination between different classes of stimuli.

• The Tree Concept (Herrnstein, 1979):

  • Pigeons were trained using slides. Key pecks were reinforced in the presence of pictures with trees ($S^D$) but not those without trees ($S^\Delta$).

  • Training Stage: Photos with trees $\rightarrow$ Peck Key $\rightarrow$ Food; Photos without trees $\rightarrow$ Peck Key $\rightarrow$ No Food.

  • Testing Stage (Generalization): Pigeons were shown NOVEL photos. They pecked for new tree images and did not peck for new non-tree images.

• Theoretical Explanation (Herrnstein, Loveland & Cable, 1976):

  • Recognizing a tree does not require a single list of common elements (e.g., green, leafy, vertical, woody).

  • No single feature is necessary or sufficient.

  • Instead, pigeons respond to clusters of features comprising a complex list of probabilistic conjunctions and disjunctions.

• Discrimination of Paintings (Watanabe, Sakamoto, & Wakita, 1995):

  • Pigeons successfully learned to discriminate between color slides of paintings by Monet and Picasso.

  • Pigeons generalized to novel paintings by the same artists not used in training.

  • Generalization occurred from Monet to similar artists (Cezanne, Renoir) and from Picasso to similar artists (Braque, Matisse).

  • Upside-down images disrupted discrimination of Monet's paintings (Impressionism) but did not disrupt Picasso's (Cubism), suggesting control by depicted objects in Impressionist works but not in Cubist works.

Advanced Concepts: Conditional Discrimination and Matching-to-Sample

• Matching-to-Sample (MTS) Procedure:

  • An inter-trial interval (ITI) of $t$ seconds begins the trial.

  • A sample stimulus (e.g., Green) appears on a center key.

  • A peck on the center key activates two side keys (comparison stimuli).

  • One side key matches the sample (e.g., Green); the other does not (e.g., Red).

  • If the pigeon pecks the matching comparison ($S+$), food is delivered.

  • If it pecks the non-matching comparison ($S-$), no food is delivered and a new trial begins.

  • The location of the match (left or right) varies across trials.

• Generalized Matching:

  • If a subject reliably matches blue-to-blue and yellow-to-yellow without specific training on those colors after learning the general rule, it has learned generalized matching.

  • Established easily in many mammals (monkeys, dolphins, humans) but is difficult for pigeons.

Arbitrary Matching to Sample

• Arbitrary Relations: Matching relations can be arbitrary rather than based on formal physical properties (like color or shape).

• Example: A pigeon is trained to select a Green comparison when the sample is a Circle.

• Because the experimenter selects these stimuli arbitrarily, the subject cannot show formal generalization.

• Significance: The study of Arbitrary Matching to Sample is foundational to understanding complex human cognition and Derived Relational Responding.