Although response-dependent shock often suppresses responding, response facilitation can occur. In two experiments, we examined the suppressive and facilitative effects of shock by manipulating shock intensity and the interresponse times that produced shock. Rats' lever presses were reinforced on a variable-interval 40-s schedule of food presentation. Shock followed either long or short interresponse times. Shock intensity was raised from 0.05 mA to 0.4 mA or 0.8 mA. Overall, shock contingent on long interresponse times punished long interresponse times and increased response rates. Shock contingent on short interresponse times punished short interresponse times and decreased response rates. In Experiment 1, raising the range of interresponse times that produced shock enhanced these effects. In Experiment 2, the effects of shock intensity depended on the interresponse times that produced shock. When long interresponse times produced shock, low intensities increased response rates. High intensities decreased response rates. When short interresponse times produced shock, high shock intensities punished short interresponse times and decreased response rates more than low intensities. The results may explain why punishment procedures occasionally facilitate responding and establish parameters for future studies of punishment.

Azrin and Holz (1966) defined punishment as the process by which a response-dependent stimulus reduces the future probability of the response. In the typical procedure for studying punishment, food occasionally is delivered contingent on a rat's lever presses. Once responding stabilizes, a punishment contingency is added by delivering response-dependent electric shock through a grid floor. Punishment is assessed by comparing responding before and after the addition of the shock contingency.

Although adding a shock contingency often decreases response rates, it can also increase rates (Appel, 1968; Arbuckle & Lattal, 1992; Filby & Appel, 1966; Galbicka & Branch, 1981; Lande, 1981; Sizemore & Maxwell, 1985). This paradoxical effect has been observed with a variety of subjects (e.g., monkeys, rats, and pigeons), operants (e.g., lever presses and key pecks), and reinforcers (e.g., monkey chow, pellets, and access to grain). Whether shock increases or decreases response rates may depend on several factors including the schedule of shock, the schedule of reinforcement, the manner of introducing shock, shock duration, shock frequency, shock intensity, and the interresponse times (IRTs) followed by shock (Arbuckle & Lattal, 1992; Azrin & Holz, 1966; Baron, 1991; Boe, 1966; Church, 1963; Sizemore & Maxwell, 1985). Of present interest are two of these factors: shock intensity and the IRTs followed by shock.

Typically, response facilitation is more likely when the shock intensity is low (Appel, 1968; Arbuckle & Lattal, 1992; Filby & Appel, 1966; Lande, 1981; Sizemore & Maxwell, 1985). For example, Filby and Appel reported that 0.2-mA and 0.4-mA shocks increased response rates, but shock intensities of 0.6 mA and higher decreased response rates. Similar results were reported by Appel (1968), Arbuckle and Lattal (1992), Lande (1981), and Sizemore and Maxwell (1985).

Whether shock increases or decreases response rates may also depend on the IRTs followed by shock (e.g., Arbuckle & Lattal, 1987; 1992; Galbicka & Branch, 1981; Sizemore & Maxwell, 1985). Galbicka and Branch used shock schedules with and without an IRT contingency to assess the role of IRTs in determining responses rates. They reinforced squirrel monkeys' lever presses according to a variable interval (VI) 60-s schedule and arranged 1.0-mA shock according to a differential IRT schedule or a random-ratio (RR) schedule. Under both shock schedules, approximately one out of every thirty responses was followed by shock, but only one schedule included an IRT contingency. In the differential IRT condition, shock depended on long IRTs, whereas in the RR-30 condition, shock was independent of IRTs. When long IRTs produced shock, response rates increased. Long IRTs were punished, and shorter IRTs became more frequent. Expanding the range of IRTs followed by shock to include shorter IRTs caused response rates to increase further. By contrast, when shock delivery was independent of IRTs, response rates decreased. Short IRTs were punished, and long IRTs became more frequent. Under the RR schedule, expanding the portion of responses followed by shock decreased response rates further. Arbuckle and Lattal (1992, Experiment 2) extended these findings by delivering shock contingent on either long or short IRTs. Although the effects on overall response rates were inconsistent across subjects, shock affected the IRT distributions by punishing the IRTs that produced shock.

The results of Galbicka and Branch (1981) are interesting not only because they demonstrate that response-dependent shock can increase response rates but also because they establish that IRTs can serve as functional units under punishment contingencies. Their results suggest that cases of response facilitation on response-dependent shock schedules may be attributable to the punishment of long IRTs rather than responses. Given that the facilitative effects of shock on response rates seem to be limited to low shock intensities, one may wonder whether the punishment of long IRTs is specific to low intensities as well. In Galbicka and Branch's study, shock intensity was held constant at 1.0 mA. This prevented any interactive effects of shock intensity and the IRTs that produce shock. To assess any interactive effects of these variables, shock intensity must be manipulated when IRTs are the functional units.

Sizemore and Maxwell (1985) investigated the combined effects of shock intensity and the IRTs that produce shock. They reinforced rats' lever presses with food on a VI 40-s schedule. Across conditions, shock intensity was raised from 0.1 mA to 0.4 mA in 0.1-mA steps. Two experiments differed in terms of the IRTs specified by the shock contingency. In the first experiment, all “long” IRTs (IRTs greater than 8 s) produced shock. In the second experiment, all “intermediate” IRTs (IRTs greater than 8 s and less than 12 s) produced shock. The lowest intensity of 0.1 mA increased response rates regardless of the IRTs that produced shock. By contrast, the effects of higher shock intensities depended on the IRTs that produced shock. When long IRTs produced 0.2-mA shock, response rates increased slightly. Higher shock intensities decreased and occasionally eliminated responding. When intermediate IRTs produced 0.2-mA, 0.3-mA, or 0.4-mA shock, shock decreased response rates but did not eliminate responding. Analyses of the IRT distributions showed that shock punished the targeted IRTs in both experiments, however, punishment of these became less differential as the intensity was raised.

Sizemore and Maxwell's (1985) results provide additional support that IRTs can serve as functional units under punishment contingencies. They also show that shock intensity and the IRTs that produce shock interact to determine whether shock suppresses or facilitates responding. When high-intensity shock was contingent on long IRTs (Experiment 1), response rates decreased. One reason for the decrease in response rates may be that shock elicited other responses, such as freezing, which interfered with the ability to complete the operant response. Another possibility is that the temporal reach of shock depends on the shock intensity. If so, raising the shock intensity may also raise the temporal reach of shock and increase the number of IRTs captured by the punishment contingency. At low intensities, the effect of shock may be localized to the IRTs closest in time to the shock. If long IRTs produce shock, long IRTs decrease in frequency, and response rates increase (e.g., Arbuckle & Lattal, 1992; Sizemore & Maxwell, 1985). If short IRTs produce shock, short IRTs decrease in frequency and response rates decrease (e.g., Arbuckle & Lattal, 1992). At high shock intensities, the temporal reach of shock may extend to a longer sequence of IRTs, which most likely includes a variety of IRT durations. Under these circumstances, various IRT classes would decrease in frequency, and response rates would decrease.

There is no reason, however, to assume that changes in the temporal reach of a behavioral consequence are limited to cases where the consequence is a punisher. Raising the intensity, or magnitude, of any behavioral consequence may increase its temporal reach regardless of whether the consequence is a punisher or a reinforcer.

Doughty and Richards (2002) demonstrated how the magnitude of a consequence affects the temporal reach of its behavioral effect. They examined the effects of reinforcer magnitude on responding under differential-reinforcement-of-low-rate (DRL) schedules. On a DRL schedule, an IRT longer than a specified criterion produces a reinforcer. In the first experiment, rats' lever presses produced either 30 μl or 300 μl of water on a DRL schedule. The experiment was divided into three phases. In the first phase, IRTs longer than 72 s produced 30 μl or 300 μl of water, depending on the rat. After responding stabilized, the amount of water was reversed such that rats that were previously given 30 μl were given 300 μl and vice versa. In the second phase, the reinforcement schedule was changed to a DRL18-s schedule. In the third phase, the DRL remained unchanged but the water amounts alternated each session. Across phases, 300 μl of water produced higher response rates than 30 μl, indicating that 30 μl was more effective in the differential reinforcement of long IRTs. Analyses of the IRT distributions revealed that as the size of the reinforcer increased, the distributions became more positively skewed, and intermediate and short IRTs increased in frequency. In other words, when the reinforcer was large, the differential reinforcement of long IRTs became less differential and generalized to other classes of IRTs. Similar results also were obtained by Kirshenbaum, Brown, Hughes, and Doughty (2008).

The experiments of Doughty and Richards (2002) and Kirshenbaum et al. (2008) support the idea that the temporal reach of a behavioral consequence increases as the magnitude of the consequence is raised. Arbuckle and Lattal (1992) suggested that a similar outcome may occur with punishment. Results from Sizemore and Maxwell's (1985) first experiment, in which high shock intensities punished all classes of IRTs, are consistent with but do not directly establish the idea that more IRTs are affected by shock as shock intensity is raised. More direct evidence of the changes in the temporal reach of shock as it applies to punishment is needed.

To fully understand the interactive effects of shock intensity and the IRTs followed by shock, and consequently, to understand how behavioral units might vary as a function of shock intensity, several issues must be clarified. One is the role of shock intensity, specifically high intensities. In Sizemore and Maxwell's (1985) first experiment, the subjects' contact with 0.4-mA shock was limited. This made it difficult to assess the effects of the highest intensities. Furthermore, high shock intensities were confounded with increases in the cumulative exposure to shock. This was due to manipulating shock intensity in ascending order. Gradually raising, as opposed to lowering or randomizing, the shock intensity is a common experimental strategy in studies of punishment because it minimizes the risk of sensitization (e.g., Azrin & Holz, 1966). Nevertheless, manipulating shock intensity in this way makes it difficult to disentangle the effects of shock intensity from the effects of cumulative exposure to shock. Additionally, in Sizemore and Maxwell's study, the range of shock intensities was small, spanning only 0.3 mA. Using a broader range of intensities might reveal effects that were absent with narrow ranges. Sizemore and Maxwell's study also leaves unanswered questions about the IRTs that produce shock. For example, in their experiments, shock targeted long and intermediate IRTs. Fewer studies have included shock contingencies targeting short IRTs (cf. Arbuckle & Lattal, 1992). Lastly, Sizemore and Maxwell defined the IRTs that produced shock in absolute terms. This may have produced inconsistent contact with the shock contingencies across subjects, as the IRT distributions presumably varied across individuals.

The purpose of the present study was to clarify the conditions under which shock suppresses or facilitates responding by examining the effects of shock intensity and the IRTs that produce shock. To address previous concerns regarding the limited range of shock intensities and contact with shock contingency, the present study included five shock intensities varied across a 0.75-mA range. This is more than twice the range used by Sizemore and Maxwell (1985). Additionally, we used two different arrangements of the experimental manipulations to separate the effects of shock intensity from those of the cumulative exposure to shock. To fully understand the role of the IRTs that produce shock, we delivered shock contingent on short IRTs as well as long IRTs and defined the IRTs that produced shock in relative terms to maintain more consistent contact with the shock contingencies across conditions, subjects, and shock intensities. Lastly, we manipulated a third variable, the range of IRTs susceptible to shock, not studied by Sizemore and Maxwell. Previous research by Galbicka and Branch (1981) indicates that the range of IRTs susceptible to shock may influence whether shock suppresses or facilitates responding, but few studies have examined how this variable interacts with shock intensity.

In two experiments, rats' lever presses were reinforced with food on a VI 40-s schedule. Once responding stabilized, response-dependent shock was added. We manipulated the range of IRTs susceptible to shock and raised the shock intensity from 0.05 mA to 0.4 mA or 0.8 mA. Shock delivery was contingent on either long IRTs or short IRTs. Because previous research with shock has shown differences in suppression levels according to how shock intensity is raised (e.g., Azrin & Holz, 1966), the experiments differed in their arrangements of the experimental manipulations. In the first experiment, the portion of the IRT distribution susceptible to shock was raised prior to raising the shock intensity. In the second experiment, shock intensity was raised before the portion of the IRT distribution susceptible to shock was raised.