As we navigate and engage with our world, we are constantly, automatically, and obligatorily encoding relations (spatial or otherwise), updating mental representations, and using that information in real-time to guide our behavior. The contribution of the hippocampus to spatial information and navigation has an extensive basis in the literature stemming from early evidence of location-modulated cells in the rodent hippocampus (for review, see Burgess et al., 2002). Evidence suggesting the hippocampus is important for spatial navigation also comes from patients with hippocampal amnesia (Maguire et al., 2006), as well as findings from functional neuroimaging studies (Ghaem et al., 1997; Maguire et al., 1997, 1998; Hartley et al., 2003; Kumaran and Maguire, 2005; Spiers and Maguire, 2006), especially when successful navigation requires access to detailed spatial representations of recently learned information. For example, Maguire et al. (2006) examined the involvement of the hippocampus in navigating an environment learned long ago in a taxi driver with hippocampal amnesia. While performance was relatively intact on general orientation in the city, knowledge of landmarks and their spatial relationships, and active navigation along some routes, hippocampal damage disrupted the ability to navigate complex environments that required the use of roads that were not major arties in the city, even though the information had been learned prior to the onset of amnesia. These findings are broadly consistent with a theory of hippocampal processing emphasizing the flexible and dynamic use of information, since representing spatial information requires constructing and maintaining relationships between different elements in the environment, establishing maps, layouts, and spatially arranged compositions of elements. Once such a configuration has been encoded (such as the relationships between the buildings and streets that make up the layout of a city, or the hallways and rooms that make up the layout of one’s own home), we must continually update our own position as we move through the map, and compare this location with the desired destination. These elements are intrinsic to spatial navigation and place a great demand on the flexible information supported by the hippocampus (Eichenbaum et al., 1999; Eichenbaum and Cohen, 2001).

Even outside the realm of navigation, the ability to tailor our behavior to meet current situational demands and incorporate immediate sensory input to guide upcoming actions and choices relies on contributions from both memory and executive control systems (Squire and Zola-Morgan, 1991; Smith and Jonides, 1999; Eichenbaum and Cohen, 2001; Tanji and Hoshi, 2008). Recent neuroimaging research suggests that the hippocampus and areas of the frontal cortex, including dorsolateral PFC, support active exploration of the environment and may lead to optimization of behavior for learning and memory of new information (Voss et al., 2011a,b). Consistent with these findings, the benefit of active control during learning is absent in patients with hippocampal damage, suggesting the hippocampus may actually be a critical component of the network that supports such behaviors (Voss et al., 2011a). In this task, patients with hippocampal amnesia studied an array of common objects arranged on a grid and viewed one object at a time through a small moving windows. When the patients with amnesia were tested for memory of the items and their spatial layout, their performance did not improve, and was actually worse, when they had active control of the moving window during the study portion of the task. Research in hippocampal amnesia also suggests the hippocampus has an active role in acquiring information about the environment and using that information during ongoing processing to guide what information should be obtained next based on previous experience (Voss et al., 2011a,b; Yee et al., 2014). Together, these findings suggest that actively learning about the environment optimizes interactions among specialized neural systems and relies critically on the involvement of the hippocampus. Furthermore, the contribution of the hippocampus is far more immediate than would be suggested by traditional descriptions of hippocampal function that are limited to long-term memory. Thus, the contribution of the hippocampus stems from the fundamental role of the hippocampus in the flexible use of relational representations.

The contribution of the hippocampus in (re)constructing, manipulating and updating relational information extends to imaginary and future events. Neuroimaging studies have consistently shown hippocampal activation during tasks that require participants to create fictional mental scenarios, especially when they draw upon or dynamically recombine previously encoded materials (e.g., Addis et al., 2007; Buckner and Carroll, 2007; Hassabis et al., 2007a; Schacter and Addis, 2007, 2009; Schacter et al., 2007; Szpunar and McDermott, 2008; Addis and Schacter, 2008). Consistent with these data, patients with hippocampal amnesia are impaired at generating descriptions of imaginary and future events, such that their descriptions are more fragmented, contain fewer episodic and semantic details, and are poorer in overall quality than matched comparison participants (Hassabis et al., 2007b; Kwan et al., 2010; Race et al., 2013). These findings suggest that the hippocampus is required to manipulate and flexibly express stored memories into novel combinations to create the elements of imaginary events.

The flexibility afforded by hippocampal representations that are important for imagination also plays a critical role in the ability to engage in creative thinking more generally. Creativity requires the ability to rapidly combine and recombine existing mental representations in order to create novel ideas and ways of thinking (Damasio, 2001; Bristol and Viskontas, 2006). While cognitive flexibility is considered to be an important component of creativity and is often attributed to frontal lobe function (Dietrich, 2004; Runco, 2004; Bogousslavsky, 2005; Kowatari et al., 2009; Dietrich and Kanso, 2010), we have shown that the hippocampus is also involved in representing ideas that are important for creativity and is part of a more broadly construed creative, constructive network (Duff et al., 2013). On a well-validated, standardized measure of creativity (Torrance Tests of Creative Thinking), we found patients with hippocampal amnesia are dramatically impaired, qualitatively and quantitatively, on measures of verbal and figural creativity, relative to matched comparison participants (Duff et al., 2013). For example, on the verbal portion, participants were asked to use written language to generate creative uses for cardboard boxes during a 10 min time period. Amnesic participant 2363 produced only two responses (e.g., recycling the boxes and making a fort), while the age, education, and IQ matched comparison participant produced 26 responses, 23 of which were determined to be unique, such as building a suit of armor. We observe the same pattern on the figural portion where, on one task, participants were presented with an oval shape and asked to think of a picture that includes the oval, adding new ideas to the make the picture tell as interesting and exciting a story as possible (see Figure 1). One healthy comparison participant made the oval into a giant tick or “tick-mobile” that, similar to a hot air balloon, takes people for rides above the city. Another comparison participant used the oval as part of a golf course complete with signs for parking and the clubhouse, the CBS sports truck, and Tiger Woods with this caddy. In striking contrast, amnesic participants 1846 and 1951, despite the same stimulus and amount of time (10 min), used the oval as an egg with a chicken above it and as a bug, respectively. This deficit in creativity in amnesia is consistent with the role of the hippocampus in representational flexibility and resonates with other similar findings that demonstrate the role of the hippocampus in imagination (e.g., Addis and Schacter, 2012), making comparisons (Olsen et al., 2012), and inferential reasoning (Zeithamova et al., 2012).

The role of the hippocampus in flexibly constructing and manipulating representations to imagine future possibilities and alternatives has implications for the contribution of the hippocampus in decision-making. On one such assessment, the Iowa Gambling Task (IGT, c.f. Damasio, 1994; Denburg et al., 2007), choices are associated with different amounts of rewards and punishments. In the task, participants select from four decks of cards that are overall advantageous or disadvantageous, such that some decks are associated with small rewards and also have small punishments, whereas other decks are associated with larger immediate rewards but also larger long-term punishments. The participant must learn to select from the decks that are overall rewarding. Thus, the IGT involves constructing an overall evaluation of the different decks based on integrating variable positive and negative outcomes, as well as selectively disregarding information that is inconsistent with the overall value of the deck.

These attributes make the IGT sufficiently more demanding on hippocampal representations as compared to other still complex tasks (e.g., Weather Prediction and the Wisconsin Card Sorting Task), on which patients with hippocampal amnesia perform successfully (Leng and Parkin, 1988; Janowsky et al., 1989; Shoqeirat et al., 1990; Knowlton et al., 1994). On the IGT (Gupta et al., 2009), patients with hippocampal damage differ from controls and from other patient populations. For example, patients with vmPFC and amygdala damage, known to have difficulties with real-world decision-making (Eslinger and Damasio, 1985; Stuss and Levine, 2002; Anderson et al., 2006), develop a preference for the disadvantageous decks on the IGT (Bechara et al., 1994, 1999, 2003; Fellows and Farah, 2005). Patients with hippocampal damage, however, do not develop a preference for either the advantageous or disadvantageous deck, even when there is no interposed delay after the card selection, suggesting these patients maintain only a momentary response to the outcome and employ a simplistic “lose-shift” strategy. Thus, the hippocampus is necessary to form, maintain, and update choice-outcome relations that unfolded over the course of the task, while the vmPFC and the amygdala are important to successfully integrate the information into a coherent, positive-payoff strategy. These findings are also consistent with other research in hippocampal amnesia (Gutbrod et al., 2006), patients with memory deficits resulting from Alzheimer’s type mild dementia (Sinz et al., 2008), and neuroimaging findings in healthy participants (Wimmer and Shohamy, 2012), which support the role of the hippocampus in effective decision-making.

Interestingly, one of the patients with hippocampal amnesia in Gupta et al. (2009) had more extensive bilateral MTL damage that also encompassed the amygdala. While patients with circumscribed amygdala develop a preference for the disadvantageous deck, this patient performed more like the patients with focal hippocampal damage than the patients with focal amygdala damage—failing to develop a preference for either the disadvantageous or advantageous decks. We have interpreted this additional finding to suggest that the contribution of the hippocampus may occur earlier and be more fundamental to advantageous decision-making, since patients with either amygdala or vmPFC damage are able to use hippocampal representations to develop a preference for one of the decks, albeit the disadvantageous one. Together, these findings may also explain real-world challenges that rely on similar complex decision-making abilities, such that patients with hippocampal amnesia are often unable to live independently, hold full-time employment, manage their finances, or navigate flexibility through the world.

The ability to flexibly represent information afforded by the hippocampus has an important role in a range of social behaviors. The ability to learn new information about a person, or ourselves, that is tied to a specific event or experience is a characteristic feature of hippocampal-dependent memory, and contributes to our ability to form relationships with others, influences our behaviors towards others, and affects our judgments and perceptions of others. For example, hippocampal representations enable us to access multiple lines of associated information, often remote in time and space, and flexibly integrate the information with new experiences, such that the way people have behaved towards us in the past will influence the way we expect them to act in the future (Cohen and Eichenbaum, 1993; Eichenbaum and Cohen, 2001; Croft et al., 2010); however, the role of the hippocampus in social behaviors has only recently been formally investigated (with limited exceptions, e.g., Johnson et al., 1985; Tranel and Damasio, 1993; Duff et al., 2007, 2008a,b, 2009; Todorov and Olson, 2008; Croft et al., 2010; Davidson et al., 2012; Beadle et al., 2013).

We investigated the contribution of the hippocampus in forming and updating character judgments by comparing the performance of patients with hippocampal amnesia to patients with damage to the vmPFC (a brain region that contributes to processing of emotional salience and moral information), as well as other brain damaged controls (Croft et al., 2010). In this study, patients made character judgments about unfamiliar persons before and after the presentation of scenarios in which a person was shown engaging in morally good, bad, or neutral behaviors. The ability to update the representation of the unfamiliar person, based on the behavior depicted in the scenario, was reflected by the change in the participant’s rating of the person (i.e., the change in the rating from before the presentation of the scenario to after the presentation of the scenario.) The patients with vmPFC damage exhibited the least change in character judgments, as expected due to their impairment in emotional processing; however, the patients with hippocampal amnesia demonstrated the greatest change and, after the presentation of the scenario, dramatically rated the persons as either very good or very bad (see Figure 2). These findings suggest that the hippocampus provides the specific contextual information from which to make appropriate character judgments, flexibly binding together information from multiple experiences, and without this signal (as well as on the IGT in Gupta et al., 2009) the patients with hippocampal amnesia overvalue the present event to make more polarized judgments.

The ability to flexibly represent everyday experiences and the relations among them also impacts the capacity to form relationships with other people and maintain them overtime. Indeed, research in patients with hippocampal amnesia suggests that the hippocampus contributes to establishing and maintaining social bonds (although see Duff et al., 2008c; Davidson et al., 2012; Warren et al., 2012). In the decades since the onset of their amnesia, patients report making only a few close new friends and are less involved with neighbors, as well as religious and community groups. As a result, their social networks are significantly smaller than matched comparison participants. The inability to form, update, and flexibly deploy hippocampal-dependent memory representations likely contributes to difficultly in maintaining and developing social relationships. That is, patients with hippocampal amnesia cannot consciously recollect shared experiences, learn the names of new people, or incorporate important new information about existing relationships into their mental representations. Consistent with this perspective, performance in healthy adults on hippocampal-dependent memory tasks has been shown to predict social network size (Stiller and Dunbar, 2007).

The ability to form and maintain social relationships may also involve contributions from hippocampal representations that support the ability to imagine and reflect upon experiences with other people. That is, to consider the social relationship from another person’s perspective and exhibit empathy. Empathy is an important ability that contributes to the quality of human relationships, life satisfaction, and well-being. The cognitive and neural substrates of empathy usually include brain regions involved in processing emotional experience and perspective taking, such as vmPFC, amygdala, anterior insula, and cingulate; however, we have shown that the hippocampus is also important (Beadle et al., 2013). In this study, we measured several aspects of empathy, including perspective-taking, emotion contagion, emotional responsiveness, and empathic concern, using a variety of standard questionnaires and in response to a series of empathy inductions. Relative to matched comparison participants, patients with hippocampal amnesia reported lower cognitive and emotional trait empathy on questionnaires, and reported no increase in empathy ratings or prosocial behavior in response to empathy inductions. For example, on one of the measures of cognitive trait empathy, perspective-taking, the ratings of the patients with hippocampal amnesia were three standard deviations, or more, below that of matched comparison participants. The perspective-taking subscale of the questionnaire was designed to assess the ability of the individual to adopt the mental perspective of another person (e.g., “When I’m upset at someone, I usually try to “put myself in his shoes” for awhile”). These findings suggest that empathy places a demand on the flexible use of information, especially in terms of the ability to engage in perspective-taking, and construct and update on-line representations that incorporate new information from recent interactions—all of which involves contributions from the hippocampus. Similarly, research in healthy adults suggests that the quality of hippocampal-dependent memory representations contributes to empathic responses, in terms of facilitating the desire to endorse prosocial intentions, such that participants report increased prosocial intentions when they vividly imagine an event, or remember a past event, helping another person (Gaesser and Schacter, 2014).

The contribution of the hippocampus also extends to what could be considered the most complex form of flexible cognition: discourse and language use in social interaction. We have proposed that the hippocampus is a key contributor to meeting many of the demands of social discourse and language use and processing (Duff and Brown-Schmidt, 2012). Spoken language unfolds over time requiring rapid and incremental processing as multiple sources of information are generated, gathered, integrated, and maintained in real-time to create meaning. Consistent with recent accounts of hippocampal involvement over very short delays, or no delays at all, (e.g., Hannula et al., 2006; Warren et al., 2011), we have found deficits in on-line referential processing in patients with hippocampal amnesia over very short discourse histories (e.g., within and across utterances) (Rubin et al., 2011; Kurczek et al., 2013). For example, in one study we had patients view a scene while listening to short dialogues introducing two characters; for example, Melissa is playing violin for Debbie/Danny as the sun is shining overhead. She is wearing a blue/purple dress (Kurczek et al., 2013). Healthy comparison participants and vmPFC patients rapidly identified the intended referent of the pronoun (she) when gender uniquely identified the referent, and when it did not, they showed a preference to interpret the pronoun as referring to the first-mentioned character. Patients with hippocampal amnesia, however, while exhibiting a similar gender effect, exhibited significant disruptions in their ability to use information about which character had been mentioned first to interpret the pronoun. Findings like this, and others (e.g., Rubin et al., 2011) suggest that patients with hippocampal amnesia not only have difficulty remembering a conversation they had earlier, they also have trouble maintaining and integrating linguistic representations as they unfold over time and using that information to guide language processing in the moment.

The hippocampus also contributes to social discourse, which often requires highly creative and flexible uses of language. Two examples of creative and flexible uses of language ubiquitous in social discourse include reported speech, in which speakers represent or reenact words or thoughts from other times and/or places (e.g., If I ever have kids I’m going to tell them, please don’t say mean things to me, Tannen, 1989) and verbal play, in which speakers play with the sounds and meanings of words through the use of puns, voices, teasing and telling funny stories (Crystal, 1998). Both of these discourse features require flexible access to previous knowledge as well as the ability to flexibly and creatively generate unique combinations of the reconstructed elements (what details to represent, what details to omit, to meet the specific interactional goals of this telling, on this occasion, with this communication partner). We have suggested that these processes are hippocampal dependent and we find deficits in the use of both reported speech and verbal play in individuals with hippocampal amnesia (Duff et al., 2007, 2009; Duff and Brown-Schmidt, 2012). That is, patients with hippocampal damage use significantly less reported speech and verbal play in their social interactions with either a clinician or familiar communication partner (see Figure 3) and when they do use reported speech or verbal play it is qualitatively different from healthy comparison participants (e.g., rotely produced). Furthermore there is some degree of hippocampal specificity; damage to vmPFC does not disrupt verbal play or reported speech (Gupta et al., 2012).

Hippocampal damage is associated with deficits across a range of linguistic and discursive abilities, although it does not, of course, affect all aspects of social discourse and language use (e.g., Gordon et al., 2014). Furthermore, hippocampal damage does not cause the devastating impairments in basic linguistic processing associated with aphasia. Yet, when the demands of social discourse and language use place sufficiently high demands on the processing features of the hippocampus, we observe deficits in the capacity to creatively and flexibility deploy the communicative and cognitive resources necessary to meet many of the moment-to-moment demands of everyday language use in social interactions.

The critical claim here is that representations supported by the hippocampus contribute not only to performance on memory tasks but also to a diverse set of cognitive abilities, which are engaged in accomplishing a variety of complex cognitive and social behaviors. The evidence reviewed above documents the striking deficits following disruption of the hippocampus or of its interconnections within a broad network of structures, impairing a broad array of abilities across multiple domains and paradigms. This leads to some hard questions: what is in common between the memory tasks typically associated with the hippocampus and the much broader range of abilities we now see as also dependent on the integrity of the hippocampus? And, what does this apparent expansion of the purview of the hippocampus mean with regard to accounts of the nature of the critical processing performed by the hippocampus?

On our view, the hippocampus implements a very basic functionality, but one whose reach has not always been fully appreciated: it binds together multiple items into compositional, relational representations that are stored, maintained, and updated, in the interconnections of the hippocampus with neocortical networks (more on this below), in such a way as to be available for retrieval by multiple brain processors, to be deployed in service of a wide variety of performances in a broad range of domains. While the functionality is basic—relational memory binding and reactivation—its reach is extensive, capable of being used in service of any performance that challenges or would benefit from the ability to construct, update, search, compare and contrast, and flexibly deploy relational representations across time.

Due to its capacity, in conjunction with the neocortical networks to which it is connected, to provide a rich relational database of information, the hippocampus plays an early and critical role in the formation, maintenance, and flexible deployment of representations that are then used by other neural systems in service of flexible cognition and complex social behavior. More specifically, we propose as an individual navigates through dynamic spatial and social environments in the world, the hippocampus is creating rich relational representations of the present while simultaneously and automatically recovering previous experiences that are similar in content and/or context and may generate novel scenarios of possible future events and outcomes (see Eichenbaum and Cohen, 2014; Wang et al., 2014). The flexibility afforded by these hippocampal representations allows them to be made immediately available to other neural systems promiscuously (Cohen, 1984). Thus, other structures in the network specialized for weighting social information and social decision-making (e.g., PFC and amygdala) can then use the information (i.e., representations of past, present, and possible future(s)) made available by the hippocampus to make decisions about the best outcome or course of action. To highlight just one example emphasizing the critical role of the hippocampus within a larger brain network, recall from the studies using the IGT that the hippocampus is necessary to form, maintain, and update choice-outcome relations that unfolded over the course of the task, while the vmPFC and the amygdala are critical to successfully integrate the information into a coherent, positive-payoff strategy.

The functionality described here is fully consistent with our previous accounts of the hippocampus (e.g., Cohen and Eichenbaum, 1993; Eichenbaum and Cohen, 2001, 2014; Eichenbaum, 2004; Wang et al., 2014). The emphasis here is on the idea that hippocampal function is not limited to particular domains of memory (e.g., just space or just explicit remembering), and not limited only to the domain of memory, but instead is available to and in service of any aspects of behavior or performance that place high demands on the formation or use of relational and flexible representations. This idea is also consistent with points of emphasis of other contemporary investigators, including ideas about the role of the hippocampus in scene (re)construction (e.g., Maguire and Mullally, 2013), and in the generation of simulated future events (Buckner and Carroll, 2007), as well as the functional relationship between the hippocampus and the PFC in service of episodic memory (Ranganath, 2010b). Moscovitch (2008) has also talked about the hippocampus in ways reminiscent of our views on relational memory binding, and about the critical interaction of hippocampus with the PFC. He emphasizes that the hippocampus is a “stupid” module, and attributes to the PFC subsequent and ostensibly more important processing of the information, with ultimate decision- and choice-making (Moscovitch, 2008). We would emphasize that the extent that the hippocampus successfully forms and recovers all the pertinent information of the past, present, and possible future options, in addition to maintaining that information online and making it available to other neural systems, has a profound influence on and plays a critical role in the outcome of what neural systems can accomplish. This is true not just for memory performance, narrowly, but as we have documented here, also for flexible cognition and social behavior more generally. So, while the hippocampus may not make decisions about what pieces of information to encode or make decisions about how to act on those representations, its binding, construction, updating, and reactivation of relational representations are critical to flexible cognition and social behavior.

A different thread in the current literature on memory and hippocampus is less focused on the interconnections of the hippocampus and larger brain networks, and more focused on the internal structure of, or anatomical subdivisions within, the hippocampus, emphasizing the memory-related sub-processes of pattern separation and pattern completion (e.g., Norman and O’Reilly, 2003; Bakker et al., 2008). But, here, too, such memory operations are thought to be applicable to all kinds of information, contributing to the creation and retrieval of the kinds of representations that we suggest here have such a broad impact on flexible cognition and social behavior.