In this paper, we describe a study in which we attempted to autoshape three horses to make a leverpress response. This study was motivated by observations made during earlier studies (Dougherty & Lewis, 1991b, 1992) in which we observed horses directing consummatory behaviors toward stimuli signaling the availability of reinforcers. The procedure was similar to that of other autoshaping procedures. Horses received forward conditioning trials in which stimuli (an illuminated panel and a retractable response lever) were periodically presented together for 10 sec and followed by a grain reinforcer. Each horse's stall served as an experimental chamber. A response panel was positioned in the stall's doorway, and a computer located nearby controlled and monitored experimental events. Each day, for 14 days, 100 stimulus-reinforcer pairings were presented. Despite 1,400 stimuli-reinforcer presentations, none of the horses were autoshaped. A few possible reasons for these results are discussed.

In previous signal-control experiments, several types of stimuli elicited pecking when paired with peck-contingent grain. Here, we compared the effectiveness of an auditory stimulus and five visual stimuli. For 12 pigeons, the first keypeck to follow the offset of a 4-sec. signal was reinforced with grain. We examined the following signals: a tone, a white keylight, a dark keylight, a keylight that changed from white to red, houselight onset, and houselight offset. All signals acquired strong control over responding. According to one measure, percent of signals with a peck, houselight offset showed less control than the others; according to another measure, pecking rate, the white keylight showed greater control than the others. In this experiment, we found that a wide variety of stimuli can elicit strong pecking in the signal-control procedure. The present findings increase the chances that in past conditioning experiments, some keypecks thought to be due to contingencies of reinforcement were in fact elicited.

Six pigeons were exposed to a procedure in which a tone preceded response-dependent grain. Pecks during the tone omitted grain for that trial. The magnitude of reinforcement, as manipulated by feeder duration, had no effect on the frequency of omission responding. The results are interpreted in the context of Scalar Expectancy Theory.

In two experiments, rats were shocked every 30 sec. Before a response, a bar press, shocks were long (2 sec); for 3 min after a response, shocks were short (0.1, 0.5, or 1 sec). When responding reduced shocks from 2 to 0.1 sec, bar pressing was acquired, and the shorter the shocks the more time spent with the short-shock condition in effect. In another procedure, the duration of individual shocks following a response was controlled so that the first shock was as long as those before the response (2 sec), but the remaining shocks in the 3-min period were short (0.1 sec). Bar pressing was maintained in some rats and acquired in others showing that, even when delayed, a reduction in shock duration is reinforcing. The latter findings question the generality of a two-factor, safety-signal interpretation of negative reinforcement. These results plus others imply that to predict responding in aversive situations it is necessary to integrate, for at least several minutes, the parameters of aversive events that follow a response.

Furedy and Biederman cited unreported silent parameters as the reason for their failure to replicate one of Badia and Culbertson’s (1972) experiments. Badia and Culbertson reported that rats prefer signaled electric shock to unsignaled. The purpose of the present experiment was to determine if the preference for signaled shock would emerge, using the Badia and Culbertson procedure, in a different laboratory and, hence, to evaluate the Furedy and Biederman speculation about silent parameters. Rats were shocked, in one of two conditions, at unpredictable times throughout each session. In the signaled conditions, each shock was preceded by a tone; in the unsignaled condition, no shock was preceded by a tone. Later, when given the opportunity, each rat changed the unsignaled condition to the signaled condition. The results are very similar to those of Badia and Culbertson. It was concluded that the changeover procedure is valid, reliable, and interpretable.

Following a shaping procedure, pigeons’ pecks were maintained on concurrent, independent variable-interval schedules in which pecks on either of two keys produced different durations of shock-free time. Shock-free times scheduled as a consequence of pecking the two keys were at different times, 3 min, 1 min, or 10 sec. The relative rates of pecking a key increased as the relative duration of shock-free time for that key increased. Results are consistent with the view that the effects of negative reinforcement are parallel, but opposite in sign, to positive reinforcement and that shock-free time is analogous to food duration in positive reinforcement experiments.

By responding on a changeover key, subjects could change the prevailing unsignalled VI 65-sec. schedule to a signalled, but otherwise identical, schedule. On the signalled schedule a brief tone preceded reinforcement. Regardless of the duration of the signal (5.0, 3.0, 1.0, 0.5, or 0.1 sec.), subjects strongly preferred the signalled schedule. Withdrawal of the signal produced a rapid decline in the changeover key response rate, demonstrating that production of the signalled condition maintained the responding. Results support the information hypothesis.

Four rats were conditioned to avoid shock on a fixed-cycle (15-sec) avoidance schedule. A 6-sec tone was introduced preceding shock. The tone could be avoided by a single response within the cycle preceding the tone’s onset. None of the four rats consistently avoided the tone, and this effect did not depend upon whether the tone could be terminated by the S. These results suggest that a warning signal does not function as an aversive stimulus, if the aversiveness of a stimulus is defined in terms of maintaining avoidance behavior. Finally, possible reasons for the inconsistency between the present experiment and that of Logan & Boice (1968) are noted.

Two experiments are reported in which distress vocalizations to shock (UR) were studied while varying signaled (CS) and unsignaled shock (US), strains, and seven different CS-US intervals (ISI) from 100 msec to 6 sec. Measures analyzed were total vocalizations, vocalizations on Trial 1, and distribution of vocalizations over the 2 sec shock. ISIs of 300 msec or less did not suppress vocalizations while ISIs from.5 to 6 sec did (p <.001), but not differentially. CS suppression of the UR occurred on Trial 1 (p <.001) and continued for all trials. Albinos vocalized more than hoods (p <.01). A non-associative interpretation was supported.

Two experiments were performed to explore the effect of illumination on rat vocalizations to shock and to repeat a previous experiment in which the presence of a CS significantly reduced squealing to shock. The first experiment replicated the suppressive effect of a CS on vocalizations and indicated the effect may be obtained using onset or offset of illumination as the CS. Reduced squealing due to illumination alone was found in Experiment I using an independent group design but not found in Experiment II using a repeated measures design. A marked increase in vocalizations over trials was observed in both experiments.

Two studies are reported using rat vocalization to electric shock as the behavioral measure. Study 1 used a classical conditioning procedure with an auditory CS for one group while presenting only the UCS to a second group. The second study investigated the same phenomenon but used a within- instead of a between-groups design. In both studies, Ss given CS-UCS pairings vocalized significantly less to UCS than when the UCS was given alone. Both studies also produced gradual UCR curves but no detectable vocalizations to the CS occurred.

Rat vocalization to electric shock was studied using a classical conditioning procedure for one group and presenting only the UCS to a second group. Ss given CS-UCS pairings vocalized significantly less to the UCS than Ss given only the UCS. Results were discussed and predicted from an information-preparation framework. Unexpected gradual OCR curves were obtained but no vocalization to the CS occurred.