Abstract
Research examining the impact of video games on neural systems has largely focused on visual attention and motor control. Recent evidence now shows that video games can also impact the hippocampal memory system. Further, action and 3D-platform video-game genres are thought to have differential impacts on this system. In this review, we examine the specific design elements unique to either action or 3D-platform video games and break down how they could either favor or discourage use of the hippocampal memory system during gameplay. Analysis is based on well-established principles of hippocampus-dependent and non-hippocampus-dependent forms of learning from the human and rodent literature.
During the past decade, the impact of video-game playing on the brain has received a large amount of scientific attention. It is now well established that certain genres of video games (e.g., action video games such as first- and third-person shooting games) can have a positive impact on visual attention. For example, habitual action-video-game players display higher levels of accuracy when presented with a task that asks them to detect the location of a target stimulus (e.g., a circle) among an array of non-target distractor stimuli (e.g., squares; Bavelier, Achtman, Mani, & Focker, 2012; Belchior et al., 2013; Dye, Green, & Bavelier, 2009; Feng, Spence, & Pratt, 2007; Gozli, Bavelier, & Pratt, 2014; Green & Bavelier, 2003; Mishra, Zinni, Bavelier, & Hillyard, 2011; West, Al-Aidroos, & Pratt, 2013; West, Stevens, Pun, & Pratt, 2008). These findings have been used to support the idea that training with action video games can remediate visual-attention deficits in children (Dye et al., 2009), adults (Bavelier & Davidson, 2013), and older adults (Belchior et al., 2013).
A large amount of attention has been given specifically to the effect of action video games on cognition; however, the importance of video-game genre when considering the specific impact that video games can have on the brain has recently been recognized. For example, one recent cross-sectional analysis revealed that playing certain action video games (action role-playing games; e.g., Borderlands 2) was correlated with lower volume in the entorhinal cortex, a neural structure that, with the hippocampus, is important for spatial memory (e.g., learning a new route to take when going to work). Conversely, it was found that people who played logic/puzzle games (e.g., Tetris) and 3D-platform games (e.g., Super Mario 64) had increased volume in the entorhinal cortex (Kuhn & Gallinat, 2013). Together, these studies have begun to suggest that different game designs could have opposite impacts on the hippocampal memory system and its related structures. The impact that our experiences, such as video-game playing, have on the hippocampus is important to consider because lower gray matter in the hippocampus is a significant biomarker for numerous neurological and psychiatric disorders across the life span, including disorders that specifically impact older adults, such as mild cognitive impairment and Alzheimer’s disease (Albert et al., 2011; Apostolova et al., 2006; Swan & Lessov-Schlaggar, 2007).
The hippocampus shares an inverse relationship with the caudate nucleus of the striatum (part of the basal ganglia; Bohbot, Lerch, Thorndycraft, Iaria, & Zijdenbos, 2007; Lerch et al., 2011). The hippocampus is centrally involved in allocentric memory (i.e., the coding of information about the location of one object based on the relative position of other objects) and supports the formation of a cognitive map (i.e., memory for relationships between environmental landmarks) irrespective of the position of the individual (Bohbot et al., 1998; Eichenbaum, Stewart, & Morris, 1990; O’Keefe & Nadel, 1978; Scoville & Milner, 1957). In contrast, the striatum is involved in egocentric memory (i.e., the encoding of the locations of objects in space relative to a person’s specific position) and habit formation through making rigid stimulus-response associations (Alvarez, Zola-Morgan, & Squire, 1995; McDonald & White, 1993; Packard, Hirsh, & White, 1989; Packard & McGaugh, 1996; Wolbers, Weiller, & Buchel, 2004). Further, the caudate is part of the brain’s reward pathway and has been implicated in the formation of procedural memories (e.g., how to ride a bicycle; Squire & Zola, 1996). The caudate is also involved in stimulus-response learning by facilitating action in response to an environmental trigger (Lerch et al., 2011; McDonald & White, 1993). Importantly, increased volume and activity in the caudate is associated with decreased volume and activity in the hippocampus (Bohbot, Iaria, & Petrides, 2004; Iaria, Petrides, Dagher, Pike, & Bohbot, 2003). In other words, a deficit in one of these parallel memory systems allows the other one to grow and control behavior. A behavioral consequence of this dichotomy is that people who use the relationship between landmarks to navigate in their environment have more gray matter and functional activity in their hippocampus, whereas people who use a rigid path to navigate have more gray matter and activity in their caudate nucleus (Bohbot et al., 2007; Iaria et al., 2003).
The designs of 3D-platform and action video games differ in several distinct ways that could have different impacts on the brain’s parallel memory systems. In 3D-platform games such as Super Mario 64, Super Mario Galaxy, and Banjo-Kazooie, one of the main goals is to explore the virtual environment to find in-game tokens and solve environmental puzzles. When a player locates and collects enough tokens, new environments are unlocked and can then be explored. The environments presented to the player are rich with proximal and distal landmarks, multiple paths to the same location, and interactive platforms and buttons that can open new areas or change the layout of already visited locations. Players therefore need to have a rich internal representation of the environment to successfully complete in-game goals and progress to new areas. The overlaid head-up display (HUD) is constantly visible to the player and provides information about the player’s progress (e.g., number of tokens collected, number of lives remaining). Importantly, the HUDs in 3D-platform games do not have mini-maps or an in-game GPS to guide the player to the next game event or goal. Instead, the player must use the environmental landmarks and scenery to learn where in-game goals (e.g., tokens, entrances to new areas) are present. This type of learning relies on the hippocampus (O’Keefe & Nadel, 1978). For example, players need to use the environmental information (e.g., mountains, trees, rivers) to remember where objects are located in space. They are often required to return to previous locations by navigating using only the game’s environment.
In contrast, in action video games such as Borderlands 2, Dead Island, and Counter-Strike, the main goal is to make speeded reactions to environmental enemies. This can involve aiming and shooting or taking cover and evading hostiles in the game environment. Action-video-game environments are rich with landmarks, but the overlaid HUD displays in-game GPS to direct players to their next location or toward in-game goals and events (e.g., items such as weapons or health packs and the locations of enemies). Because of this, players navigate without relying on the relationships between landmarks, a process that depends not on the hippocampus but on the caudate nucleus of the striatum (Mishkin & Petri, 1984; Squire & Zola, 1996). For example, players must quickly locate enemies and react to them, leaving little time to direct attention to the game’s environment. The inclusion of the in-game GPS compensates for this reality and allows players to quickly move to the next game event without encoding environmental information. Further, many action video games played online with other players, such as Counter-Strike, repeat the same environments numerous times, requiring players to overlearn the environments to adequately perform. It is thought that when a new environment is repeated many times, the caudate of the striatum is recruited to facilitate the efficient, however rigid, memorization of a series of movements and turns (Mishkin & Petri, 1984; Squire & Zola, 1996).
Evidence for the dichotomous effect of video-game genres on the hippocampal memory system has been highlighted in two recent studies. The first study examined habitual action-video-game players’ performance in a virtual maze compared to that of non-video-game players. It found that young adults who were action-video-game players were significantly more likely to use navigation strategies that were not dependent on the hippocampus (West et al., 2015). In other words, action-video-game players were less likely to use their hippocampus to remember the locations of items in a new virtual environment. In the second study, young adults were trained on a 3D-platform game (Super Mario 3D World), and this training led to increased hippocampus-dependent and episodic memory performance (Clemenson & Stark, 2015). This finding supports previous evidence that training on a 3D-platform game (Super Mario 64) results in increased volume in the hippocampus (Kuhn, Gleich, Lorenz, Lindenberger, & Gallinat, 2014). Together, these findings suggest that different video-game designs can either discourage or favor the use of the hippocampus during learning.
Hippocampus-Dependent Spatial Learning and Non-Hippocampal Response Learning
Previous evidence suggests that different video-game genres, which have key differences in design features, can have differential impacts on the hippocampal system, including the functionally and structurally connected entorhinal cortex. This differential impact could be explained by examining which design aspects of certain video-game genres can bias people to learn a new environment using either hippocampus-dependent or non-hippocampus-dependent learning. Non-hippocampus-dependent learning typically involves response learning, which is a type of learning that relies on the caudate nucleus of the striatum and involves the rigid memorization of a series of movements (e.g., left and right turns) from given positions that act as stimuli (e.g., a post office). Response learning is also a process mediated by the brain’s reward pathway, as the striatum is centrally involved in the formation of habits that emerge over repeated occurrences of a given behavior (Mishkin & Petri, 1984; Squire & Zola, 1996). For example, when one first learns to ride a bicycle, the process is not automatic but requires a high level of attention to coordinate the various muscle movements needed to master the task. Over time, this process becomes automatic, and the conscious effort to achieve this coordination is no longer required. The caudate nucleus of the striatum takes part in establishing this automated type of behavior. This same system is used in navigation when participants are not using their hippocampus. When people navigate using response learning, it resembles a “GPS-like” strategy. For example, when one learns a new location using a response-learning strategy, a rigid series of directions are used to follow a sequence of left and right turns. This process will become automatic over time (i.e., habit formation) and is dependent on the caudate nucleus of the striatum. Because of this, response learning in an environment can be faster and more efficient (Iaria et al., 2003). However, response learning is also inflexible. For example, if the route learned through response learning were blocked—for example, because of construction—and a new route were now needed, the individual would not have the necessary information to accurately reach his or her destination. In other words, the relationships between landmarks in the environment were not learned; for this reason, it is not possible for people using stimulus-response learning to derive a novel route to a target destination.
Conversely, spatial learning relies centrally on the hippocampus and involves building relationships between landmarks to create a detailed internal representation of the environment, or a cognitive map (O’Keefe & Nadel, 1978). For example, learning the relationships between landmarks (buildings, trees, boulders, rivers, etc.) allows for a flexible use of this information to navigate to a destination point that is independent of the position of the observer. In other words, a rigid route is not required, as it is with response learning, because the internal representation of the environment is detailed enough to allow one to accurately locate a destination from any starting point. In contrast to caudate-mediated habit learning, acquisition of cognitive maps is thought to be driven by curiosity and novelty detection, not reward and punishment.
There is now a large amount of evidence that shows a close relationship between response learning and gray matter within the hippocampus. Humans who have less hippocampal gray matter preferentially use response strategies in situations in which both response and spatial strategies could be used (Bohbot et al., 2007; Iaria et al., 2003; Konishi & Bohbot, 2013) and also show lower functional activity in the hippocampus (Etchamendy, Konishi, Pike, Marighetto, & Bohbot, 2012; Iaria et al., 2003; Konishi et al., 2013). This relationship was shown to be causal in rodents. Lerch et al. (2011) used a water maze to train mice to find a target platform by using either distinct visual landmarks (spatial learning) or a single “beacon” stimulus that directly indicated the location of the platform (response learning). In other words, the mice were trained on either a spatial-learning task or a response-learning task. Mice in the spatial-learning group showed post-training increases in hippocampal gray matter relative to mice in the response-learning group. This study demonstrated that the environment can have a significant impact on gray-matter levels in the hippocampus. Further, levels of gray matter in the hippocampus and in the caudate nucleus of the striatum have been shown to have an inverse relationship, such that increased gray matter in the caudate nucleus is associated with lower gray matter in the hippocampus in both mice (Bohbot, Del Balso, Conrad, Konishi, & Leyton, 2013) and humans (Bohbot et al., 2007). Thus, promoting the caudate through repeated response-learning experiences could contribute to lower gray matter in the hippocampus.
Why is hippocampal gray matter important? Higher levels of gray matter in the hippocampus are associated with reduced risk of neuropsychiatric illness in younger and older adults alike, including disorders such as depression (Amico et al., 2011), posttraumatic stress disorder (Gilbertson et al., 2002), and Alzheimer’s disease (Apostolova et al., 2006; Swan & Lessov-Schlaggar, 2007). How does this relate to action-video-game playing? We know that action-video-game players are more likely to use response strategies (West et al., 2015). It could therefore prove beneficial to modify action-video-game designs to encourage spatial learning instead of response learning.
Video-Game Design and Spatial Learning
It is now clear that different video-game genres (puzzle/logic, action, 3D-platform, etc.) have differential impacts on the brain (Clemenson & Stark, 2015; Kuhn & Gallinat, 2013; Kuhn et al., 2014; West et al., 2015). The key to understanding why this is true is to know how the general designs of different game genres encourage different types of learning. This can be accomplished by breaking down the tested principles first introduced by O’Keefe and Nadel (1978) that outline how environments can encourage either response or spatial learning. These environmental conditions can be directly linked to design choices made in video games. We identify three general principles below (see Table 1).
Video-Game Design Principles Hypothesized to Discourage or Favor Spatial Learning
Cognitive-map formation
3D-platform games encourage exploration of the virtual environment. Players can use the relationships among environmental cues, such as landmarks, to memorize where in-game goals are present. Over time, when the spatial relationships among these cues are encoded, an internal representation of the environment is formed and used to direct navigation when locations need to be revisited. This internal representation is called a cognitive map and is dependent on the hippocampus (O’Keefe & Nadel, 1978). Further, response learning is discouraged through the lack of in-game wayfinding or GPS cues that encourage route following, which relies on the caudate nucleus of the striatum. The absence of in-game GPS within 3D-platform games is hypothesized to promote the development and use of a cognitive map through spatial learning (Bohbot et al., 2007; O’Keefe & Nadel, 1978) and increased gray matter in the hippocampus (Kuhn et al., 2014). In contrast to 3D-platform games, the modern action video games in which in-game GPS and other wayfinding cues are ever present are hypothesized to promote response learning through biasing navigation along rigid routes and thus do not recruit the hippocampus (Bohbot et al., 2007; Lerch et al., 2011). In other words, when people rely on a GPS system to locate in-game goals, they are using their striatum (response learning) and not using their internal cognitive map (spatial learning) to progress through the game.
Task speed
Tasks that are required to be completed at faster rates promote response learning (O’Keefe & Nadel, 1978; Schwabe et al., 2007). In action-video-game design, it is often the case that players are asked to make many stimulus-response-goal (S-R-G) actions per minute, as exemplified by a player’s need to quickly locate in-game enemies, aim at and eliminate said enemies, and proceed to the next goal, sometimes all within a matter of seconds. Because these types of actions are repeated frequently in action video games, this type of behavior becomes more automatic with time. This type of interaction with the environment actively discourages hippocampus-based learning (Schwabe et al., 2007), because spatial learning requires more time than many action video games afford to complete the task at hand (Iaria et al., 2003). Further, an increased number of S-R-G actions per minute increases autonomic stress (Hasan, Begue, & Bushman, 2013). This type of induced stress also encourages response learning, because it is more efficient in terms of shorter latencies, and has been shown to reduce long-term potentiation in the hippocampus of rodents (Kim, Lee, Han, & Packard, 2001). Specifically, Kim et al. (2001) found that stressed rodents displayed reduced long-term potentiation in hippocampal neurons and lower spatial-memory performance compared to non-stressed rodents. Further, it was found that this effect is mediated by the amygdala. Stressed rodents with amygdala lesions did not display reduced hippocampal long-term potentiation or impaired spatial-memory performance compared to stressed non-lesioned rodents. This inhibition of the hippocampus creates a condition in which the other, caudate memory system is favored to direct navigation.
This is again in contrast to 3D-platform games, which more often than not do provide ample time to explore a new game environment without the use of GPS and allow for the building of an internal cognitive map, thereby stimulating the hippocampus (Bohbot et al., 1998; Eichenbaum et al., 1990; O’Keefe & Nadel, 1978; Scoville & Milner, 1957). In other words, when stress is not induced through speeded responses, the hippocampal memory system has the opportunity to direct navigational behavior.
Environmental diversity
Learning a new environment promotes exploration and curiosity, which supports spatial learning (Lerch et al., 2011; O’Keefe & Nadel, 1978). Exposure to an increased number of varied game environments (i.e., environmental diversity) without the use of rigid S-R-G mediated routes (i.e., GPS), which are typically not present in 3D-platform games, is hypothesized to support the creation of a cognitive map (Lerch et al., 2011; O’Keefe & Nadel, 1978). Conversely, when an environment is learned and repeated, inflexible routes are efficiently memorized through response learning. Importantly, this reflects a common design aspect of commercially available action video games (e.g., Counter-Strike) in which gameplay is focused on only a few environments that are repeated often during online gameplay and speeded gameplay (e.g., shooting skirmishes) is emphasized. In other words, these gameplay demands do not allow for exploration or for the use of spatial strategies, and the repeated environments promote route learning that is central to the response strategy (Bohbot et al., 2007; Lerch et al., 2011; O’Keefe & Nadel, 1978).
Because the way we interact with different environments can change gray matter in the hippocampus (Lerch et al., 2011) and action-video-game players are known to predominantly be response learners (West et al., 2015), the design of action video games could be modulated to favor hippocampus-dependent learning. Such changes in design could bias players to favor spatial learning during gameplay, which would in turn encourage hippocampal stimulation (Bohbot et al., 2007; Kuhn et al., 2014; Lerch et al., 2011):
Action video games could be designed without in-game GPS or wayfinding cues to encourage the construction of internal cognitive maps of environments that rely on the hippocampus (Bohbot et al., 2007; Kuhn et al., 2014; O’Keefe & Nadel, 1978).
The reduction of the speed that dictates the number of S-R-G actions per minute in action video games is unlikely. However, if these events are spread out among in-game tasks that encourage spatial learning (e.g., in-game logic puzzles that require spatial reasoning, exploration without GPS), both types of learning can be encouraged.
The environmental diversity of action video games, especially in online multiplayer games that repeat the same game environments often, can be increased with an increased number of varied environments to encourage spatial learning.
We should note that it is possible that employing these game-design changes could make action video games less attractive to people who were drawn to these types of games in the first place as a result of their “non-hippocampal” style of gameplay. The rewarding dopaminergic effects promoted by the caudate response or habit-learning system may be one of the main reasons why people are attracted to action video games. As a result, game designs that favor hippocampal-based learning might not be appealing to these individuals.
It should also be noted that the caudate memory system is also important for optimal cognitive function and that degradation of this system occurs in certain neuropsychiatric illnesses such as Parkinson’s disease. Therefore, an ideal balance would be for both the hippocampal and caudate systems to be used during video-game playing. There are some aspects of caudate-guided behavior that can arguably never be removed from the action-video-game genre (e.g., aiming and quick stimulus-response reactions toward in-game enemies). Our considerations above are aimed at bringing about more balance when it comes to encouraging the use of both hippocampal and caudate memory systems during action-video-game play.
Future Research Directions
The current challenge is to directly test these hypotheses. We know that not all action video games are created equal. For example, a recent title released in 2016, Rise of the Tomb Raider, is considered an action video game and features many S-R-G style shooting skirmishes, but it also includes a high number of logic and spatial-reasoning puzzles. Similarly, a game released in 2004 called Half-Life 2 was a first-person shooting game; however, it also encouraged players to navigate through the environment without the aid of in-game GPS and contained environmental puzzles. This is in contrast to more modern video-game designs that mostly rely on in-game GPS to guide players toward goals that allow them to progress through the game. These examples highlight the need to experimentally isolate the above-identified aspects of game design thought to encourage or discourage spatial learning. With this said, experimentally isolating these variables could prove to be a challenge without the help of the gaming industry. An ideal experiment would, in collaboration with game designers, modulate existing games thought to have an impact on the hippocampal system to test these outlined hypotheses. This option would arguably be more ecologically valid than creating a degraded version of a game for experimental use only. This, coupled with the fact that training studies need a lot of content to fill 60 hours of training, makes modulating existing games a more attractive option. Further, many PC versions of video games have options to customize game levels and maps that do not require heavy coding abilities.
We would also like to note certain limitations of our proposal based on the limited research available. First, as previously noted, the evidence that video game genres have differential impacts on the hippocampal memory system consists of data from cross-sectional or ex post facto designs. This means that these findings are correlational in nature. Further, it is possible that people are differentially attracted to certain video-game genres depending on whether they are response learners with higher levels of gray matter in the caudate or spatial learners with more gray matter in the hippocampus. It is for this reason that we underscore the importance of conducting longitudinal research in the future to establish a causal relationship between components of video games thought to recruit the hippocampal versus the caudate memory system and growth within these neural systems. This will then allow for the direct testing of a causal relationship between the presence of specific design elements and changes in hippocampal gray matter through longitudinal training studies. Results from such studies have the possibility to greatly increase our understanding of how different environmental elements can impact hippocampal integrity.
