General Biases That Operate in Animal Models of Cognition

Measures of cognition in animal models may be influenced by biasing factors that are unrelated to cognition. Such factors include motivation, motility, and left-right preferences.

An animal must be adequately motivated to attempt a cognitive task and to do well on it. In other words, performances that are deemed to represent learning efforts must be driven by motivation to learn rather than by random, exploratory behavior. Motivation is generally ensured through reward (e.g., a food pellet) or punishment (e.g., a footpad electric shock). Food-motivated tasks may require the rat to be on I hour/day restricted feeds for 2-3 days prior to the learning experiments; feeding on the days on which learning is assessed is permitted after the learning tasks for the day have been completed.

(If the learning tasks span several days, it may be advisable to permit feeding only when several hours have elapsed after exposure to the task; otherwise, the rat may show poor motivation to attempt the task, having already learned that food will be provided after exposure to the task.)

A problem with such food deprivation is that the rat may become sluggish; as a result, the rat may not attempt the task at all. For example, we have observed that rats on restricted feeds sometimes do not move out of the stem of the T maze. The learning of these rats cannot be assessed, and the results of the experiment may consequently suffer bias. Rats that are sluggish but that do attempt the task may show biased performances on time-based variables. Another problem with food deprivation is that hypoglycemia can itself compromise learning. The interaction between this variable and the experimental variable cannot be estimated. These limitations of food-motivated tasks with rats on restricted diets must be kept in mind when conducting and interpreting research based on such paradigms.

Learning tasks that are time dependent are biased by variations in basal animal motility. For example, in the Hebb-Williams maze, the dependent variable that estimates learning performance is the speed by which the rat reaches the reward chamber. This variable is influenced by the rat's basal motility: a sluggish rat will "learn" more slowly while a restless rat will "learn" more quickly. Hence, a treatment that alters basal motility will produce spurious changes in teaming performance. The direction of the error will depend on the nature of the task and the effect that the treatment has on motility. A treatment that decreases motility will falsely enhance learning performances and will fail to adequately identify amnestic effects in learning tasks, such as passive avoidance paradigms, in which an absence of response from the animal indicates learning. Such a treatment will fail to adequately identify learning, and will falsely inflate amnestic effects in learning tasks such as the Hebb-Williams maze and active avoidance paradigms, in which an active response from the animal indicates learning. Treatments that increase motility will produce errors in the opposite direction. For example, Posluns and Vanderwolf (1970) found that retrograde amnesia in passive avoidance tests after ECS may be partly due to a deficit in the ability to suppress motor activity.

Possible solutions to motility-related problems in time-based learning tasks are either to previously study and rule out an effect of the treatment upon basal motility before proceeding with the task or to use a task such as the T maze, the results of which are not influenced by motility factors. Another possibility is to deliberately use a learning task that predisposes to a false negative error when studying a putatively procognitive drug that affects basal motility. The logic here is that it may be better to err on the side of caution during screening. While including a sham treatment control group by itself will not help, the use of a factorial experimental design can reduce (but not necessarily eliminate) motility related errors. In a study intended to test the antiamnestic effects of a drug in ECT-treated animals, the experiment would include drug/ECT, sham drug/ECT, drug/ sham ECT, and sham drug/sham ECT groups. The last two groups serve as internal controls to the main experimental and main control groups. In the analysis of results, the interaction effect between the drug and ECT would indicate the antiamnestic action of the drug.

Many of the ingredients of the herbal formulations that we studied are claimed to have tranquilizing properties and would consequently be expected to reduce motility. We therefore focused on T maze paradigms in our recent research and used factorial designs in all our studies to minimize motility-related errors.

For over two decades it has been recognized that rats show clear left-right preferences, and it has recently been recognized that these preferences influence rats' choices on spatial tasks such as the T maze. Very recently (Andrade et al., unpublished observations), we showed that the bias in T maze arm preference was substantial: 22.2% of the rats that we studied showed a left preference, and 52.8% showed a right preference. This bias was spontaneous and was consistent over two testing sessions 30 days apart. Left- and right-biased rats learned rapidly when trained to enter the arm ipsilateral to the bias; learning was significantly poorer or did not occur contralaterally. This contralateral learning difficulty was particularly evident when transfer of learning was assessed, especially with right-biased rats. Interestingly, unbiased rats (25%) also showed some difficulties in attaining the criterion for learning in one or the other arm of the T maze. This finding is probably a result of the broad definition that we used for absence of bias in contrast with the strict definition used for the presence of bias. Actually, some of the unbiased rats also showed bias albeit to a lesser extent, and this bias may have been responsible for the learning confusion observed. Our findings suggest that unless spontaneous laterality preferences are taken into consideration, spurious results may be obtained in spatial learning tasks.

In our own research, described in an earlier section, we attempted to ensure validity of results by screening all rats for the ability to learn in both arms of the T maze, and by randomizing rats into groups based on their learning performances. We consider that there are three prerequisites for valid use of the T maze in cognitive research: Animals should be preselected for capacity to learn in both arms, randomization into experimental and control groups should be stratified for spontaneous arm bias, and original learning should be directed towards the arm contralateral to the bias while transfer of learning, if required, can be directed into the ipsilateral arm. These prerequisites are unfortunately likely to make T maze research time-consuming and unattractive.