Working memory (WM) corresponds to a critical cognitive domain required for the representation of objects or places during goal directed behavior¹. Alternatively, reference memory (RM) is required for temporally stable representations of those objects or places. Working and reference memory have long been assessed in rodents using land based radial arm maze paradigms^2,3. However, these tasks frequently require pretraining, as rats are not predisposed to spontaneous maze running. This can significantly increase the time needed to complete an experiment and can interfere with time sensitive longitudinal designs. Other limitations of the land based eight-arm maze include requirements for food deprivation and difficulty controlling changes in scent cues left on the maze after each trial. Many of these limitations have been overcome by using the Morris water maze paradigm⁴, however historically these designs have been limited to testing reference memory and spatial learning (exceptions^5-9). The eight-arm radial water maze is a modified version of the eight-arm radial land maze, which has been used to assess both reference and working memory in rats and mice^5-7. In contrast to traditional land based mazes, the eight-arm radial water maze does not require food deprivation, minimizes potential confounding scent cues and utilizes the subjects' motivation for escape as an effective means to assess working and reference learning and memory simultaneously without the need for pretraining^5-7.

One application for this paradigm is to test working memory deficits following neonatal brain injury. Research in humans has shown that preterm infants at risk for brain injury exhibit working memory deficits later in life^10,11. Neonatal brain injury can be modeled in rodents by inducing hypoxia/ischemia (HI) early in postnatal development^12-15. More importantly, spatial learning and working memory deficits found in at risk infants are paralleled in rodents using the radial arm water maze, making it a robust model for the study of such impairments^8,9. The eight-arm radial water maze also allows for simultaneous quantification of reference and working memory making it ideal for the comparison of brain injured and noninjured subjects in time sensitive models (i.e. during discrete developmental widows) or for counter balanced longitudinal designs.

In this protocol, we describe testing procedures using an eight-arm radial water maze (see Figure 1) and example data for rats with and without neonatal hypoxic-ischemic injury. Hypoxia-ischemia (HI) was induced on postnatal day 7 following cauterization of the right common carotid artery and 120 min of 8% oxygen exposure. This surgical procedure has been extensively used to model pathology of prematurity and neonatal brain injury (for details see^12,13,16,17). We show that the radial arm water maze paradigm reveals deficits in reference memory and working memory capabilities following neonatal HI injury. In the current protocol, reference memory is assessed using four specific arms of the eight-arm radial water maze, which remain constant to extra maze cues throughout testing. These arms never contain an escape platform, thus, subject entries into these arms reflect poor reference memory performance and deficits in long-term learning. The remaining four arms all contain escape platforms at the beginning of the day, however each platform is removed once the animal successfully enters an arm and escapes the water^5-7. This requires the animal to remember which arms it entered over the four successive trials, increasing the working memory demand as each additional platform is removed. Working memory is assessed by examining re-entries into arms previously used for escape within the same testing day or re-entries into reference memory arms. Additionally, this design utilizes extra-maze cues to locate the escape platforms reflecting hippocampal dependent spatial learning. This method for assessing working and reference memory errors was originally used by Jarrard et al.¹⁸ and has since been used extensively to assess rodent models of aging related neurological disorders, learning disabilities and following hormonal manipulations^5,6,7,19,20.

The primary objective of this paradigm is to assess reference and working memory, thus, the eight-arm radial water maze design should not be limited for testing deficits due to developmental brain injury. Instead, previous studies using this paradigm show that a variety of rodent models can be evaluated to assess pathology of learning and memory across many experimental contexts^{5,6,7,18,19,20}.