While the brain works as a dynamic network, with no brain region solely responsible for any particular function, it is generally accepted that the hippocampus plays a major role in memory. Spatial memory operates through the hippocampus with communication with the prefrontal and parietal cortices. This chapter will focus on two separate reference frames involved in spatial memory, egocentric and allocentric, and outline the differences of these reference frames and associated search strategies with relevance to behavioural neuroscience. The importance of dissociating these search strategies is put forward, and steps researchers can take to do so are suggested. Neurophysiological and clinical differences between these spatial reference frames are outlined to further support the view that distinguishing them would be beneficial.
- spatial memory
Spatial memory is the cognitive process of noticing, encoding, and retrieving landmarks in the surrounding environment, to allow an organism to navigate and exist in the world. It is important for survival, by enabling searching and finding safety and food and being able to return to found places without issue. It is the domain of the hippocampus and medial temporal lobe, with links to the retrosplenial cortex and parietal cortex . Seminal studies in humans and animals have demonstrated the important role that the hippocampus plays in navigating the world around us [2, 3]. In humans, damage to the temporal lobe causes disturbances to spatial navigation , and similarly, humans employed in roles that require fantastic spatial navigation skills have enlargement of the hippocampus and its connections [5, 6]. In parallel, through multiple manipulations such as lesion, electrophysiological and optogenetic studies, the hippocampus has been shown to be equally important to animal spatial memory. Disruptions to hippocampal tissue or silencing of neurons in the hippocampus leads to spatial memory deficits [7, 8]. This parallel role of the hippocampus in both humans and animals allows research to be performed on these animals with the insights gained able to be extrapolated to humans.
2. Spatial memory in behavioural testing
Behaviourally characterising an animal model of disease often involves a battery of tests that investigate the animal’s motivation, locomotor activity, startle reflex, anxiety, fear response, social behaviour, learning, memory and other emotional and cognitive traits. Dysfunctions in these behaviours are used to infer structural and functional changes in the brain, and the recovery of performance on these tests is used to evaluate the effectiveness of potential therapeutics. These inferences are only accurate with the use of appropriate tests with high specificity both for the behaviour in question and in terms of the specific brain regions recruited during test performance. Therefore, behavioural tests that are specific to one domain or behavioural tests that can correctly dissociate multiple domains should be used. Rodent spatial memory tests, often mazes, are commonly used in preclinical drug development and fundamental science experiments. The use of these behavioural tests dates back over a century, and a plethora of maze designs have been developed since then to probe different aspects of learning and memory . Complex networks of brain regions and neuron populations are required to orientate and navigate using information such as environmental, vestibular and proprioceptive cues . The current general consensus is that spatial memory encompasses two distinct but related reference frames, egocentric and allocentric. Here, we outline the differences between these reference frames and their relevance in behavioural neuroscience and discuss the merits of placing a stronger emphasis on distinguishing egocentric and allocentric search strategies in spatial memory tests.
3. What are allocentric and egocentric search strategies?
The egocentric reference frame is also referred to as a fixed, self-centred or first-person perspective. Egocentric navigation is based on direction (left-right) responses and actions independent of environmental cues. Directional decisions are made at single or sequential choice points; however, these locations are not used as cues and are therefore still egocentric in nature . For example, memorising routes based on sequential turns would employ a mostly egocentric strategy (Figure 1A). Path integration, the summation of travelled vectors to deduce current position, is an example of an egocentric strategy that can navigate through novel paths. The allocentric reference frame, on the other hand, can be thought of as a third-person perspective. Allocentric navigation utilises external cues or landmarks in relation to each other to navigate and is independent of self (Figure 1B). Utilising compass directions (north, south, east, west) is an example of allocentric reference frame use as these directions are relative to the Earth and do not change depending on the orientation of the navigator . An advantage of allocentric navigation is the flexibility of being able to locate novel points from various start locations as long as the external cues remain the same. In situations where external cues are changing, minimal or absent, egocentric strategies become more salient .
Navigating environments outside of experimental settings requires the use of both allocentric and egocentric reference frames, with relative saliencies falling within a spectrum . Experiments in controlled settings with specifically designed spatial memory tasks aim to dissociate these reference frames; however, it is argued that complete dissociation is not achieved . Nevertheless, the employment of more precise tasks as well as the use of more rigorous analytical techniques allows greater dissociation and investigation into navigational strategy preference and specific dysfunctions in reference frames. Nonspatial strategies such as random or serial searches can often be successful in that they result in lower latencies to a goal. These, however, are not indicative of spatial memory, and measures should be put in place to detect such strategy use. The following section provides an overview of the various spatial memory tasks currently used in behavioural neuroscience and their ability to effectively probe egocentric and allocentric search strategies.
4. Spatial memory and navigation paradigms
There are a large variety of behavioural tests for both rodents and humans that provide a measure of spatial memory and navigation [9, 13, 14]. Generally, rodent spatial memory tests utilise maze apparatus that have a goal area that the animals must find, learn and remember. These goals can be positive reinforcements such as food rewards, escapes from negative stimuli such as water or bright light or a result of instinctive behaviour such as exploratory drive. Human spatial memory testing, on the other hand, is mostly conducted on virtual reality setups that create controlled three-dimensional environments with goals usually being explained to the subject by the researcher. More recently, steps have been taken to combine aspects from both animal and human tests to increase the similarity and therefore translatability of these tests. Virtual reality versions of rodent tests have been developed for humans , and virtual reality and touchscreen setups for rodents that were developed from human equivalents have also become popular [16, 17]. Distinguishing allocentric and egocentric reference frames and search strategies used in spatial memory tasks for rodents differs depending on the type of test. Some tasks are designed to encourage employment of a single strategy, and so performance on that task is reflective of the saliency of that particular reference frame. Other tasks can be completed with a combination of allocentric and egocentric strategies, and subsequent analysis or probe tests are needed to infer deficits or preferences in these reference frames. Consideration of what types of spatial navigation are being tested, and extra steps to dissociate these strategies are often overlooked, despite the relative ease of implementing such measures. Below we discuss popular maze apparatus used to investigate spatial memory and various tests, controls and analyses that can help distinguish egocentric and allocentric navigation.
Spatial memory can be investigated through a variety of tests on mazes such as the Y-maze, cheeseboard maze, Morris water maze, Star maze, Barnes maze, radial arm maze and T-maze. These mazes encompass investigation of a range of spatial memory, including long-term, short-term and working memory, as well as cognitive flexibility. Tests that probe allocentric reference frames include the use of static visual cues which the rodent can use to develop a cognitive map. Efforts are made to minimise proximal cues and create open, unobstructed spaces to avoid non-allocentric strategies. The opposite is true for egocentric tasks where visual cues are minimised or made irrelevant (incorrect or random). The most accurate way of testing for egocentric strategies is to perform a test in the dark, which ensures removal of visual distal cues that could be used for allocentric strategies . Many apparatus that are used to investigate egocentric navigation restrict movements to narrow channels or arms to create distinct choice points where egocentric strategies are encouraged .
Constructed in the shape of a capitalised ‘T’, the
5. Analysing search strategies to compare the use of egocentric or allocentric search strategies
Spatial memory proficiency is commonly measured through a range of parameters in the above-mentioned mazes including latency, distance and time spent in target quadrants. However, evidence suggests that these analyses are not providing sophisticated enough insights into cognition and behaviour . The Current trend is a deeper analysis of spatial navigation in order to produce more efficient research and more efficient use of research animals , moving beyond the well-known parameters of latency and distance. Research is now interested in the search strategy employed by research subjects and animals (Figure 7). Search strategy analysis can observe the complexity and dynamic nature of cognition employed in spatial memory mazes. For example, while different genotypes may have no significant differences in the typical parameters of latency, distance or target quadrant, a difference in approach to goal could exist and demonstrate changed cognition as a result of genotype. This may be more reflective of the innate differences that can exist in individual cognition despite similar anatomy. Of particular interest is the path trace analysis of allocentric tests in open field-type mazes, where movement is not restricted by walls (such as the MWM, CBM or Barnes maze). Although the absence of choice points aims to encourage allocentric strategies in these mazes, evidence suggests egocentric strategies can still be used; view-matching on distal cues can lead to egocentric cue guidance (e.g. face the star and then turn left) , which can successfully complete the task. Non-allocentric strategies such as serial strategies (visit all locations) and chaining (knowing distance from the edge of the maze) can also be successful strategies that also cannot be seen using traditional metrics (see Figure 7). These search strategies can be manually assigned through blinded categorisation or be analysed using automated algorithms. While historically latency and distance have been used as measures of cognitive disturbance in the MWM, time spent in the target quadrant on the probe day and search strategy are adjunct parameters that can provide a deeper analysis. Indeed, Rogers et al.  elegantly put forth how imperative investigating search strategy and setting up a high-powered experiment can be. Their study demonstrated not only the importance of high saliency cues but also the depth and breadth of information available through the analysis of search strategy.
The adoption of an allocentric search strategy is completely dependent on the quality of landmarks available . This adds another consideration to the design of experiments for researchers; the setup of the maze must be carefully considered. Additionally, Rogers et al.  demonstrated that the latency and path length parameters do not provide differentiation between the different search strategies and in fact do not provide a reliable analysis of spatial memory formation. From this arises the argument that not only does investigating search strategy allow for the elucidation of egocentric versus allocentric search strategies but that the saliency of distal cues allows the research animal to employ these strategies in the first place. It is important to note that more thorough methods for evaluating MWM performance have been suggested for a long time. The proximity measure, introduced in 1993, measures distance to the goal at a frequency of 10 Hz to get an average proximity throughout the trial. This measure was seen to be more sensitive than latency to the goal and was able to pick up subtle and otherwise masked effects . Unfortunately, this measure is still currently underreported and highlights the need to actively encourage extended analysis beyond latency, distance and time.
Building upon this, the study by Suzuki and Imayoshi  deftly investigated and presented a novel method of analysing navigation in the Barnes maze. The authors titled this ‘network analysis method’, which allowed for the visualisation of a rodent’s exploratory patterns. The method involves several algorithms which initially determine the search strategy employed by a rodent (spatial, serial or random). Following this analysis, Suzuki and Imayoshi  were interested in determining if particular networks were associated with particular search strategies. A local network is the exploratory behaviour pattern of one mouse of one experimental group. Once local networks are established for all mice of an experimental group, a global network can be created from this data and demonstrates the exploratory behaviour of the whole experimental group. For this study, Suzuki and Imayoshi  focused on eight different exploratory behaviours that formed dynamic nodes. Following algorithmic analysis, links between the different nodes (i.e. exploratory behaviours) were established. The authors observed that as spatial learning is established across the experimental days, the global network is simplified, and nodes surrounding the target area are stronger than indirect nodes with indirect links. Most importantly, as highlighted by Suzuki and Imayoshi , although significant differences in cognitions were subtle, these spatial navigation behaviours were able to be recognised and quantitatively analysed using the ‘network analysis method’. The capacity to apply quantitative statistics to patterns of behaviour provides a fantastic opportunity to apply strong, scientific investigation into higher cognitive processing. This is a strong example of utilising search strategy analysis in order to identify the more dynamic substrates of the cognitive underpinnings of navigation. The successful identification of strengthened spatial memory by Suzuki and Imayoshi  using the ‘network analysis method’ demonstrates the brevity of utilising similar approaches when investigating spatial memory.
6. Neurophysiology of allocentric and egocentric strategies
Studies investigating the neurological correlates of egocentric and allocentric navigation have utilised lesion, electrophysiological and optogenetic techniques to better understand the distinct mechanisms underlying them. In many experimental and clinical settings, specific deficits in one reference frame but not the other are observed, further indicating separate mechanisms.
6.1. Lesion studies for identification of allocentric and egocentric brain networks
A number of studies have investigated the cognitive consequences of lesioning the hippocampus using spatial memory tests such as the MWM. The overwhelming consensus is that allocentric learning is impaired after hippocampal lesioning. One of the first studies to demonstrate this was by Morris et al.  in rats. They demonstrated that lesioning the hippocampus of rats resulted in an inability to navigate the MWM. This is supported by numerous other studies [7, 39, 40], which all found significant deficits in traditional spatial memory measurements such as time to platform, distance to platform and time spent in target quadrant (probe trial). Other lesion studies indicate the perirhinal cortex, entorhinal cortex and parietal cortices to be involved in allocentric search navigation [41, 42, 43]. Maze apparatus that can be utilised to test egocentric search strategies include RAM , Cincinnati water maze and Star maze . While allocentric search strategies appear to be dependent majorly upon the temporal lobe components, egocentric navigation appears to have a broader network. A study using the RAM observed deficits in egocentric navigation after lesioning medial agranular cortices . Comparatively, a fascinating study by Wolff et al.  demonstrated that region-specific lesions of the thalamus impaired egocentric and allocentric navigation independently. They postulated that lateral thalamic lesions interrupt communication between the striatum and frontal cortex, by destruction of the intralaminar nuclei. This interrupted pathway manifested as deficits in egocentric navigation. Indeed, studies have indicated that the dorsal striatum and head direction cells are involved in egocentric navigation . The cerebellar-dentate nucleus has also been implicated in egocentric processes , demonstrating the complexity of the networks involved in these search strategies. While we have so far attempted to separate these two navigation strategies, they are not mutually exclusive. A fantastic review by Ekstrom, Arnold and Iaria  goes into detail on theories that describe transitions between allocentric and egocentric strategies, as well as the overlap between them.
6.2. Electrophysiological studies for identification of allocentric and egocentric brain networks
There has been extensive research into the neural correlates of spatial memory and navigation. In the seminal book,
Mechanistic differences between egocentric and allocentric reference frames are also observed in electrophysiological recordings. Theta oscillations, or the theta rhythm, are low-frequency (~7–9 Hz) local field potential oscillations that function as a temporal frame in which neurons fire action potentials . Both place and grid cells demonstrate theta phase precession effects to differing levels during navigation. That is, as an animal travels closer to the peak firing field of a certain place or grid cell, that cell will fire earlier in the theta phase . This adds an additional layer of encoded information that contributes to navigation. Furthermore, oscillatory activity has been shown to facilitate the coherency between brain regions involved in egocentric and allocentric navigation . Specifically, low-gamma oscillations (25–50 Hz) between the CA1 and CA3 and high-gamma oscillations (65–140 Hz) between the CA1 and entorhinal cortex. Indeed, these oscillatory frequency ranges in the CA1 are associated with changes in egocentric and allocentric behaviour .
6.3. Optogenetic studies for identification of allocentric and egocentric brain networks
Optogenetics is an outstanding technique to elucidate the functional relevance of particular neuron populations in specific brain regions and areas. A study by Andrews-Zwilling et al.  optogenetically inhibited hilar GABAergic neurons which led to a spatial memory retrieval impairment in the MWM. This study used the parameters escape latency and percentage time spent in target quadrant. However, there was no reported analysis of search strategy. As outlined by Rogers et al. , search strategy analysis is imperative to confirm spatial memory learning. For this study, it would be interesting to know the strategies employed by the mice and compare to controls, to see exactly how the optogenetic inhibition is affecting navigation. By knowing the effects upon search strategy, it provides further depth and breadth to understanding the cognitive processes occurring. Yamamoto et al.  further confirm a role for the hippocampus in spatial memory with their optogenetic inhibition of medial entorhinal cortex layer III (MEC) inputs to the CA1 of the hippocampus. This was demonstrated using the delayed nonmatch-to-place T-maze task, a working memory task that is based upon egocentric navigation, that is, it is based upon the successful alternation of turning left or right at a junction . Building upon this, the study by Perusini et al.  demonstrated that optogentically stimulating the dentate gyrus in aged mice improved memory retrieval in the contextual fear conditioning paradigm. This has great implications for the current problem of the world’s extended life span and associated neurodegenerative diseases such as dementias. The hippocampus is a hub for memory and is linked to multiple networks, as demonstrated especially by Ito et al. . Optogenetic inhibition of cells in the nucleus reuniens of the thalamus resulted in reduced trajectory-dependent firing of the CA1 region of the hippocampus. Projections from the medial prefrontal cortex to the nucleus reuniens which end in the CA1 hippocampus region are imperative to goal-directed map representation.
The studies examined above indicate that some regional differentiation exists between the individual networks involved in allocentric and egocentric navigation. Taken together, it would appear that the hippocampus and surrounding areas are strongly involved in spatial memory and in particular the allocentric search and egocentric navigation strategies. Understanding the effects upon spatial memory and navigation is enhanced by analysing the search strategies employed by research animals. Disruptions to normal functioning could result in compensatory mechanisms that disguise impairments to spatial memory, if the appropriate analyses are not performed. Future studies should use techniques such as optogenetics to specifically investigate cell populations in the hippocampus and associated areas and their role in spatial memory and allocentric and egocentric navigation strategies using specifically designed mazes such as the Star maze. It is widely accepted that the hippocampus has a role in spatial memory, but we are now starting to understand how disrupting spatial memory alters navigational pathways.
7. Search strategies and their relevance to ageing and disease
Further incentive to differentiate egocentric and allocentric navigation in spatial memory tests arises from evidence in studies of human ageing and disease showing that deficits are observed in specific search strategies. Studies in real-world environments such as supermarkets  and roads  confirm the anecdotally long-held belief that spatial memory performance worsens with normal ageing. Elderly humans also perform worse in virtual reality versions of mazes designed to investigate spatial memory  accompanied by changes in electrophysiological event-related potentials . Allocentric navigation seems to be affected more so than egocentric navigation [25, 67], and specific deficits arising only when switching to an allocentric from an egocentric strategy have also been observed . These behavioural changes may be a result of age-related changes in the hippocampus including decreased synapse function and long-term potentiation . Declines in other domains such as working memory and sensory perception most likely also contribute to the decreased spatial memory performance seen in ageing; however, the vulnerability of allocentric over egocentric strategies prompts the need for further investigation into the mechanism behind this deficit. Interestingly, allocentric-specific deficits also seem to manifest in the young (6–7 years old) as well as the elderly , suggesting the deficit may be related to cognitive load.
Alongside ageing is an increase in risk for neurodegenerative disorders such as Alzheimer’s disease (AD) and associated decline in memory. Topographical disorientation is an early symptom of AD that involves the inability to orientate in the environment and often leads to patients being prone to getting lost. A systematic review of egocentric and allocentric spatial ability in AD by Serino and colleagues  observed an allocentric deficit in both mild cognitive impairment and AD. Furthermore, a later study by Allison and colleagues showed allocentric-specific deficits can also be seen in asymptomatic preclinical AD, suggesting allocentric spatial memory tasks may be useful in the early diagnosis of AD . Similar allocentric-specific deficits are also observed in neurodevelopmental disorders such as attention deficit hyperactivity disorder . Although the ability to learn locations from allocentric representations has been shown to be decreased in patients with autism spectrum disorder (ASD) as well , there is sparse literature and agreement on this topic . Cognitive symptoms are an untreated aspect of schizophrenia, and allocentric-specific deficits have been observed .
Many spatial memory deficits in cognitive decline and disease seem to preferentially affect the allocentric reference frame and navigational strategy. Constructing an allocentric cognitive map of an environment would allow navigation from any start point to a goal location compared to an egocentric sequence, which would only be viable from a single start point to reach a goal. Intuitively, allocentric search strategies are more complex than egocentric strategies and therefore may experience loss of function before the onset of more severe deficits that then go on to affect the egocentric reference frame. In a similar vein, there is also evidence to suggest that perhaps the allocentric reference frame is a culmination of many egocentric frames, meaning egocentric frames are likely to exist without allocentric frames but not vice versa . This could explain the disproportionate dysfunction in allocentric abilities and the relative persistence of egocentric ones. Another possibility is that specific navigational deficits are a reflection of inaccurate (unconscious) selection of the search strategy most suited for the task at hand .
8. Why is the distinction important?
Animal models allow the investigation of specific forms of memory and dysfunctional neuro-components, as a way to parallel human illness. Since humans and animals have analogous brain regions with similar functions, it is helpful to the expansion of biological knowledge to investigate possible disruptions in order to understand the fundamental neuroscience.
Distinguishing egocentric and allocentric search strategies in spatial memory tests is important because:
Accuracy and integrity of experimental results would be stronger. Due to the fact that one strategy may be preferentially affected over the other, not considering the distinction has a similar effect to not measuring the effect of an unknown variable. Results may become skewed, diluted or even completely masked.
There is a potential to discover novel therapeutic targets. Coupling behavioural data with known physiological and molecular pathways underlying these search strategies could elucidate specific deficits in disease.
They can function as more precise outcome variables that can potentially be utilised in early diagnosis of cognitive impairments. Detection of subtle deficits may also be improved.
Understanding the inner workings of our brains will be advanced.
Reviewed here is evidence supporting the distinction of egocentric and allocentric reference frames in spatial memory. These reference frames and their respective search strategies are closely related and are often used in combination when navigating. We argue that because these reference frames involve different mechanisms and they are differentially affected by experimental manipulations and disease, they should be appropriately dissociated when investigated. Rodent mazes such as the Star maze have been developed to tackle this issue by directly probing egocentric and allocentric strategies. Other, more widely used mazes such as the Y-maze and RAM are able to probe these strategies with slightly modified protocols. Open arena apparatus such as the MWM, CBM and Barnes maze can provide different insights on spatial memory performance, but an often overlooked and informative parameter is the qualitative measurement of path traces and investigation of search strategies. Not only has the investigation of search strategy been shown to be required to confirm the creation of an allocentric map, it provides a depth and breadth to understanding the cognitive processes occurring post-experimental intervention or modification. We strongly encourage and recommend the adoption of search strategy analysis and comparison between experimental groups, in order to gain the most from your data.