Abstract
Goldfish (Carassius auratus) represent an alternative to using rodents for a psychology learning course. One difficulty, however, in using goldfish as subjects is that of suitable equipment. The present paper investigated the extent to which operant manipulanda designed for rodents could be used with goldfish. Attempts were made to shape the responses of goldfish to an omnidirectional ceiling rod as a manipulandum using a reinforcer pellet intended for mice, which was not successful. A previously described goldfish reinforcer was used successfully to shape a response by two goldfish, maintained by continuous reinforcement to obtain a low but consistent response rate over several sessions. Difficulties associated with the reinforcer as well as the rodent manipulandum appear to limit their utility in a novel application such as the one presented here.
Alternatives to using laboratory rodents in a psychology learning laboratory or course are being sought and proposed as the traditional “rat lab” is increasingly expensive to the point of being impractical or even prohibitive (Miskovsky, Beckner, Hillker, & Abramson, 2010). With the use of laboratory rodents, compliance with numerous animal housing, care, and welfare issues and regulations is required (Miskovsky, et al., 2010). The use of invertebrates does allow some freedom for instructors, but many students and others hold negative attitudes towards invertebrates (Looy & Wood, 2006) as well as a “vertebrate-centered view” of animals (Mather, 2011) and may be inclined to see limited generalization of principles (Bekoff, 1994; Mather, 2011).
Alternatively, an instructor could resort to using a computer simulation of a behaving organism that simulates or approximates behavior change with varying success; two widely used programs are CyberRat (Ray, 2003) and Sniffy the Virtual Rat ® (Alloway Wilson & Graham, 2005). While the computer programs do offer convenience, there are liabilities associated with the convenience. When asked to compare learning from the Sniffy program relative to the experience of learning about behavior from a living rat, students reported that they preferred learning from observing an actual organism more and the living animal's behavior presented a superior learning experience (Trench, 2011). Students asked to evaluate an actual classical conditioning experience with living organisms in comparison to a computer simulation reported similar observations (Abramson, Onstott, Edwards, & Bowe, 1996). One college instructor and evaluator of CyberRat as a teaching program reported dissatisfaction with the very limited range of behaviors upon which contingencies were applied (Iversen, 2011/2012). Based on the student and instructor-evaluator reports, there are reasons to consider the use of live animals for teaching students about learning and behavior change.
The common goldfish has been used as a vertebrate alternative to rodents as animal subjects and using goldfish as subjects can necessitate different IACUC regulations, but this varies considerably from one institution to another. The use of fish is governed by the Public Health Service Policy, which covers all live vertebrate animals, through the NIH's Office of Laboratory Animal Welfare (National Institutes of Health, 2002). Goldfish have been used successfully in a number of research publications (Hogan, 1961; Rozin & Mayer, 1961; Hogan & Rozin, 1962; Rozin, 1968; Woodward & Bitterman, 1974; Wolach & Roccaforte, 1976; Gee, Stephenson, & Wright, 1994). After an extensive search of various databases, to the author's knowledge, none of the apparatuses used in these studies is currently available if a researcher were to consider trying to equip a goldfish learning laboratory.
By way of possible alternatives, Miskovsky, et al. (2010) proposed a relatively easy-to-construct and inexpensive apparatus to use for operant contingencies with goldfish, even going so far as to arrive at a recipe/formula for a reinforcer that can be dispensed with their fish-training apparatus. Miskovsky, et al. (2010) pointed out that their fish-training apparatus was intended primarily for instruction, not for research. Another somewhat similar fish-training apparatus is the commercially available “R2 Fish School” system, which does allow for shaping of various operant behaviors, according to the company's on-line video demonstrations. However, this device has to be withdrawn from the water and repeatedly reloaded with the reinforcer, which amounts to a discrete-trial method instead of a free-operant method (Skinner, 1938; Perone, 1991); the former necessarily has an interruption in on-going behavior from trial to trial (Skinner, 1969).
In the past few decades, more and more colleges are discontinuing their animal research and teaching facilities due to either costs, faculty turnover, or ethical issues (Akins & Panicker, 1995; Plous, 1996; Abramson, et al., 1999), so operant-conditioning equipment formerly used with rodents is now available. The practical question of trying to use operant equipment designed for rodents with goldfish may be difficult to justify given the nearly prohibitive costs of much operant equipment (Miskovsky, et al., 2010). The attempt to use rodent equipment described here resulted from situational factors that unfortunately are becoming more common: an instructor may not have a facility approved or suitable for rodents although rodent-operant equipment is readily available to use. The current paper proposes an attempt to use operant manipulanda designed for use with rodents, but in water, so that operant contingencies can be applied to the behavior of the common goldfish.
Method
An operant learning situation requires, at a minimum, a two-term relationship between a measurable response and a consequence (Skinner, 1938). The operant response typically measured with rodents as subjects has typically been a lever press, with less commonly measured operants being wheel running, a nose-poke into a photobeam detector, or the push of a omnidirectional ceiling rod. Of these possibilities, a lever partially submerged at the surface of the water seemed to represent a possible response manipulandum to employ with goldfish but the force required to press a lever would make this operant difficult to acquire. Some researchers, however, have successfully shaped fish to lever-press on a customized lever for a thermal consequence as well as a liquid reinforcer (Hogan, & Rozin, 1961; Rozin & Mayer, 1961; Rozin, 1968).
As a possible alternative, the experimenter subjectively assessed the force required to operate an available rat lever compared to the force requisite to operate an omnidirectional ceiling rod, and the latter clearly required less force. A typical omnidirectional ceiling rod can be extended below the surface of the water with the actual microswitch out of the water, and can be operated with minimal force in any direction. The essential component of this manipulandum is a thin metal rod, approximately 4 mm in width and with a length of approximately 12.7 cm.
As for a reinforcer to be delivered contingent upon the operant, the author speculated that with sufficiently large goldfish or Koi as subjects, (approximately 7.6 cm) food-pellets designed for use with mice, a 20-mg grain-based food pellet, could be delivered with a conventional food pellet dispenser. The author's thinking was based on the fact that goldfish and closely-related species such as Koi will readily consume a variety of foods (Baensch & Riehl, 2007) including commercially available foods and pellets given on the surface of the water at ponds and aquariums (Kleinholz, 2000). To assess the BioServ® grain-based pellets for mice as potential reinforcers, the BioServ Corporation® provided a free sample bottle of the 20-mg pellets, which were used in a home aquarium to verify that goldfish and other fish would eat the pellets and they would putatively function as a reinforcer. Following some unforeseen difficulties with these reinforcers discussed below, a reinforcer proposed by Miskovsky, et al. (2010) was adopted for use.
Miskovsky, et al. (2010) proposed a formula for a semi-solid paste reinforcer and provided the details for raw materials and preparation. This reinforcer could be dispensed manually from a 10 ml syringe with the hypodermic needle removed and replaced with a length of standard aquarium airline 0.63 cm tubing; the tubing ended with a 90° airline tubing elbow as a nipple. Figure 1 illustrates the omnidirectional ceiling rod with tubing attached as configured for this application. Initially, the Miskovsky et al. apparatus and reinforcer were not the method of choice in the present investigation, but methodological difficulties soon pointed to the utility presented in the 2010 paper.
The omnidirectional ceiling rod with syringe and tubing attached used in the present study. The apparatus is shown in a metal stand and clamps for holding chemistry beakers and test tubes.
A 37.85 L aquarium served as the experimental tank. This aquarium was continually filtered but lacked any gravel. This aquarium was modified with semi-transparent plastic sheets applied to the inside surfaces to eliminate distracting reflections on the glass and the outside of the aquarium was wrapped in cardboard to further eliminate external distracting events. The experimental tank was further modified by adding 0.31 cm thick plastic barriers to confine the subjects to an area of approximately 161 square cm closest to the operant conditioning apparatus; tank water circulated in and out of this confinement area. In addition, two 8.25 square cm abutments were added to the front corners of the tank so that the fish could not access the omnidirectional rod by swimming into it from the outer perimeter of the tank. Confining the fish to an area close to the apparatus and requiring a direct frontal approach to the apparatus minimized the incompatible behavior of swimming far away from the apparatus and increased the probability of bumping the rod where the reinforcer would be delivered.
Participants
Four goldfish (Carassius auratus) were obtained from a local retailer and kept in a 75.70 L community tank in the author's laboratory. The tank was maintained at room temperature, had approximately 2.5 cm of gravel for a bottom, was continually filtered and was lit on approximate 12 hr. light: 12 hr. dark cycle. The fish lived in the community tank for approximately 3 wk. prior to any use as subjects. Two of the fish that most readily took flake food from the surface of the community tank with the experimenter's hand near the surface of the water served as experimental subjects.
Procedures
Following 24 hr. without food, the fish were individually transferred to the experimental tank and allowed a 15-min. adaption period. After the adaptation period, the initial attempts to establish the pellets as a reinforcer as a basis for shaping the desired behavior began. Initially, the pellets were delivered individually at the surface of the water from the pellet dispenser. The proximity of the pellet dispenser's delivery tube and the effect of a pellet on the surface of the water elicited flight (swimming away) by the fish, but this behavior appeared to undergo habituation in a few minutes. The sinking fish pellets elicited approach-pursuit behaviors which replaced the swimming away and it was observed that the fish would consume a small number of the pellets, stop feeding and in a few minutes, regurgitate the pellets. Thus, their function as reinforcers was poor.
Next, attempts were made to deliver goldfish food flakes in the food-pellet dispenser, but the flakes either jammed in the pellet dispenser and or an inconsistent amount of food was delivered. Neither the attempts at using conventional food-pellets for rodents nor delivery of flake food in the conventional apparatus were successful, which compromised the stated purpose of adapting rodent-apparatus for use with goldfish. Clearly, a consequence under the control of a researcher is needed for operant conditioning, which necessitated another approach in this study. The delivery syringe containing the Miskovsky, et al. (2010) reinforcer and tubing from the delivery syringe was taped to the side of the omnidirectional rod, as shown in Fig. 1. This allowed the reinforcer to be delivered as a small bead of food (approximately 1–2 ml) on the manipulandum, to allow Pavlovian contingencies to elicit responses directed upon the manipulandum (Brandon & Bitterman, 1979). The reinforcer was manually dispensed when the fish was within 5–6 cm of the manipulandum and the fish soon readily approached the apparatus when a bead of the reinforcer was visible. All of this worked adequately to allow the shaping of the operant response of a snout bump by the goldfish upon the manipulandum in a matter of a few training sessions of approximately 15 min. each. A picture of a fish bumping the apparatus is shown in Fig. 2. With the use of the Miskovsky, et al. paste reinforcer manually dispensed, no wiring nor circuitry was needed, a response was defined by observing the subjects bump the omnidirectional ceiling rod.
One of the two goldfish is shown bumping the apparatus to obtain the reinforcer.
Results
Two subjects were shaped to bump the manipulandum, and using a continuous reinforcement schedule, a low but fairly consistent rate of responding was observed for several sessions; both subjects responded approximately 1.5 times per min for repeated short sessions of approximately 10 min. The rate of responding is shown in Fig. 3. As shown in this figure, both subjects had experimental sessions in which no measurable responses were emitted and both subjects had their highest rates of responding at approximately 2.5 responses per minute in only one session. The experimental sessions were kept short because of difficulties with the Miskovsky, et al. reinforcer. Although the reinforcer was a semi-solid paste, as the fish consumed the reinforcer the tank water became cloudy with particulate matter. An on-going filter was not able to remove the matter adequately and water quality quickly deteriorated over the course of repeated reinforcer deliveries. Miskovsky, et al. (2010) did not report this outcome and it possibly resulted from the small volume of the 37.85 L experimental tank used in this investigation. When the author attempted to increase the response requirements to an intermittent schedule of reinforcement, responding by both subjects exhibited apparent ratio strain, responding ceased, and both subjects had to have the operant response shaped again. The failure to acquire responding on an intermittent schedule of reinforcement may have been due to the difficulties incurred from the reinforcer or the manipulandum used in this investigation.
The response rate in terms of responses per minute of two subjects on a continuous schedule of reinforcement in daily 10-minute sessions. Data for Subject 1 (black symbols) and Subject 2 (Pink symbols).
Discussion
This stated purpose in this investigation was to demonstrate that operant equipment designed for use with rodents could be adapted for use with goldfish. This demonstration was successful, within limits; some of the manipulanda available for use with rodents can be used in a fish tank. An omnidirectional ceiling rod is suitable for this application. The behavior of goldfish can be shaped to perform on operant response on this or another apparatus constructed by an experimenter (Miskovsky, et al., 2010). In order to conduct operant conditioning, an experimenter-controlled reinforcer must be delivered contingent upon a response (Skinner, 1938), and with respect to this independent variable, it is likely that the liquid-paste reinforcer of Miskovsky, et al. could be employed with the use of a reinforcer apparatus specifically manufactured to control and deliver liquid reinforcers. The present investigation did not attempt this, not having access to any manufactured, liquid-reinforcer delivery apparatus. Attempts to use operant contingencies to shape the behavior of goldfish will be complicated by the problem of delivering the consequence, such as the Miskovsky, et al. (2010) reinforcer-paste or another alternative (Hogan & Rozin, 1961). Rozin (1968) trained fish to lever press for thermal stimulation as a reinforcer but this again raises the question of what to use as a simple, readily available apparatus.
One way to address the difficulty associated with reinforcer delivery would be to use a reinforcer, i.e., not consumed, as in Rozin (1968) but with simpler technical demands. Higa and Simm (2004) demonstrated that the behavior of Siamese fighting fish (Beta splendens) could be shaped with a reinforcer consisting of the brief presentation of a mirror contingent upon swimming through a hoop. The mirror-image of the fish was made contingent upon the first response performed after a specified delay and served to maintain responding on several different fixed interval schedules of reinforcement. If the experimental arrangement of events and contingencies discussed by Higa and Simm (2004) can be easily replicated, then the use of this fish and the use of a mirror-image as a reinforcer represents an alternative to more conventional operant conditioning arrangements for teaching demonstrations.
As another alternative answer to the question posed above regarding appropriate apparatus, there is low-cost technology available with which to connect an omnidirectional ceiling rod with an interface-controller for purposes of having an automated goldfish operant chamber. The rodent omnidirectional ceiling rod described here was obtained from the MED-Associates equipment company; the microswitch and wiring of this omnidirectional ceiling rod is easily accessed with the removal of four screws on the lower surface of the cube-shaped enclosure at the top of the rod. With access to the wires and microswitch, this manipulandum could be interfaced with a computer or microcontroller to accept inputs from such a lever and the microcontroller could operate an output such as a pump to deliver the Miskovsky, et al. (2010) reinforcer. A low-cost and relatively simple to use microcontroller for this application was described in a recent paper (Varnon & Abramson, 2013). At the time of writing, an application of the Varnon and Abramson (2013) microcontroller to the manipulandum and to a liquid-reinforcer pump has not been attempted but it would appear that this experimental set-up is a likely prototype for a simple, readily-available apparatus, i.e., needed to facilitate research and teaching demonstrations of this sort.