Memory and Common Ground Processes in Language Use

Abstract

During communication, we form assumptions about what our communication partners know and believe. Information that is mutually known between the discourse partners—their common ground—serves as a backdrop for successful communication. Here we present an introduction to the focus of this topic, which is the role of memory in common ground and language use. Two types of questions emerge as central to understanding the relationship between memory and common ground, specifically questions having to do with the representation of common ground in memory, and the use of common ground during language processing.

1. Introduction

Any two people bring to a conversation a different perspective on the world, including different beliefs, knowledge, and life experience. Through the process of communicating, we form assumptions about the perspective of each other. Knowledge that is shared with a communication partner, and that the communication partners know each other know, forms their common ground. These representations of common ground serve as a backdrop for successful communication.

According to the classic view (Clark & Marshall, 1978, 1981), conversational partners assume common ground based on several sources of evidence, including information that is culturally co-present (e.g., most Americans know who the president is), information that is physically co-present (e.g., in face-to-face conversation, the immediate surroundings), and information that is linguistically co-present (e.g., information that has been exchanged in the dialog). Common ground is a subset of one’s own knowledge in that it is the information that is known to both partners in a conversation, and that is known (or assumed) to be mutually known. Which aspects of common ground are and are not relevant to a given conversation are likely to depend on the nature of the conversation. For a conversation about a political election, the major candidates for office are likely to be considered common ground. When one of the conversational partners shares her political views, the fact that she holds these views is also now part of their common ground, regardless of whether the other partner agrees with that view. Once some common ground is established, conversational partners have a solid footing on which to proceed with the conversation—they know what language to speak, what information can already be assumed, and what information cannot.

A critical area of inquiry, then, concerns the memory processes that support the representation and use of common ground during communicative exchanges. How do we encode and subsequently retrieve information about what specific other individuals do and do not know? An emerging consensus in the literature is that the way in which representations of common ground are stored in memory are likely to be relevant to the way common ground does or does not guide language processing.

The focus of this topic is how common ground is represented in memory and how it is used. It is trivial that common ground must be represented in memory somehow, but while this much is taken for granted in the literature, what is lacking are strong empirical and theoretical arguments for how it is represented, as well as the implications for language processing. The goal of this topic, then, is to examine this issue from multiple distinct perspectives in order to lay the groundwork for a broader understanding of how common ground is represented and used in communication.

2. Common ground plays a role in many domains

Representations of common ground play a critical role in one of the most basic aspects of the human experience—the ability to appreciate the perspective of another being. Perspective-taking is fundamental to communication as it facilitates the efficient exchange of information and ideas. For example, imagine Otto asks Eleanor, Where is the book? It is by appreciating the fact that one’s communication partner may have a different perspective that Eleanor can successfully understand Otto’s question to be asking about information that he does not know but that she does know—that is, Otto is asking about Eleanor’s private knowledge—her privileged ground (Brown-Schmidt, Gunlogson, & Tanenhaus, 2008). Likewise, when Eleanor designs her response to this question, she must formulate it to be interpretable from Otto’s perspective, a process known as audience design (Clark & Murphy, 1982). Audience design critically involves distinguishing what is common ground with one’s conversational partner from what is privileged ground.

Theoretical treatments of common ground grew out of research traditions in philosophy and linguistics (see Stalnaker, 2002; 1978), and play a central role in theories of language use (Clark, 1996). More generally, the fact that we understand that others can have beliefs that are distinct from our own—that we have a theory of mind—features prominently in theories of child conceptual development (Baillargeon, Scott, & He, 2010) and language acquisition (Bloom, 2001). Indeed, perspective-taking processes are relevant to a broad range of issues, including imagining the future, global conflict resolution, mathematical reasoning, navigation, space exploration, and nuclear waste disposal. Success in each of these domains involves forming distinct representations of an alternative perspective on the world. For that reason, it has been proposed that at least some of these domains, such as prospection and theory of mind, may involve similar neural mechanisms (Buckner & Carroll, 2007; Ryskin, Brown-Schmidt, Tullis, & Benjamin, 2015).

Consider the issue of nuclear waste disposal. How can today’s society communicate to future societies that the waste is toxic, given that the structure of society and, in particular, our writing systems are bound to change over time? A “Keep out” sign might not be translatable in the distant future, and a skull and crossbones may be misinterpreted as pirate booty (Lapidos, 2009). A U.S. Department of Energy commissioned report titled “Expert Judgment on Markers to Deter Inadvertent Human Intrusion into the Waste Isolation Pilot Plant” (Trauth, Hora, & Guzowski, 1993) was one government attempt to design a labeling system. Among other possibilities, the group proposed a sign that featured a cartoon version of Munch’s “The Scream.” The sign reads “DANGER / POISONOUS RADIOACTIVE WASTE BURIED HERE / DO NOT DIG OR DRILL HERE BEFORE A.D. 12,000” (p. F-115). Moving forward, as our societies generate long-lasting poisons, figuring out how to appreciate the perspective of future society in order to generate appropriate warning signs will be critical. To do so will require making assumptions (hopefully accurate ones) about the possible perspectives of future society and what signs and signals will successfully communicate the relevant danger. Similar assumptions and design issues are in play in the design of artifacts intended to communicate with possible entities in outer space: The Pioneer plaques were placed on board spacecraft in 1972 and 1973 and were intended to communicate with extraterrestrial life in the event they were intercepted. The plaques feature pictures of nude humans as well as information about Earth’s location (NASA, 2007). The Voyager spacecraft from 1977 similarly contained golden records which, if played, contain information about life on Earth, including music, natural sounds, and greetings in 55 languages. In each of these cases, successful design of a communicative instrument requires estimating the perspective of an intelligent recipient in the distant future.

Communicating with future civilizations and with extraterrestrial life are extraordinary examples of perspective-taking. A more commonplace example is navigation. Imagining alternative perspectives is a critical component of navigating throughout the world, as well as when giving directions to other individuals. For example, when receiving directions from another person, understanding their direction terms, for example, “go left,” requires making assumptions about which perspective the speaker has adopted. Whether you should go left or right depends on which spatial perspective your partner is using, that is, your perspective or theirs.

3. Common ground in language use

Common ground is thought to provide the basic context for language (Clark, 1996). Experimental research on common ground in language processing typically focuses on how physical and linguistic cues to common ground are used to guide language production and comprehension.

In any face-to-face conversation, some of the objects in the physical environment are visible to both parties, whereas other objects are visible only to one of the conversational partners. Imagine a situation in which two individuals, Charles and Forrest, are playing at the beach and there are two buckets, a small bucket which is visible to both boys, and larger bucket which is concealed in Forrest’s backpack. If Forrest said, “Hand me the bucket,” Charles would likely understand him to mean the small bucket in view. Forrest would not need to specify which bucket because the larger one is not in common ground and there would be no reason to think that Charles would be aware of it. By contrast, if both the large and the small buckets were mutually visible, Forrest would need to say “Hand me the small bucket.” Studies of language production in similar situations show that speakers consistently specify size in situations in which both buckets are in view, whereas they are significantly less likely to mention size when the second object is hidden (Nadig & Sedivy, 2002; Yoon, Koh, & Brown-Schmidt, 2012). These findings show that speakers are sensitive to the perspective of the listener, and design expressions with respect to a context that includes objects in common ground; that is, speakers readily engage in audience design.

Of course, conversational partners do not limit their discussions to just those entities in common ground. During conversation, common ground grows as partners discuss previously privileged information, establishing it as common ground (Clark & Schaefer, 1989). For example, in the beach scenario, if Forrest asserted, “I have a larger bucket in my backpack,” and Charles replied, “Oh great,” they would have successfully established the large bucket as common ground. Evidence from studies that monitor language processing by measuring eye-gaze in conversation show that this type of linguistic mention is an effective way of establishing information as common ground. Once common ground is established, addressees typically interpret requests such as “Pick up the block . . .” as referencing objects in common ground (due to either physical or linguistic co-presence; Hanna, Tanenhaus, & Trueswell, 2003; Brown-Schmidt, 2012), whereas questions such as “What’s below the cow . . .” are typically interpreted as asking about items in the addressee’s privileged ground (Brown-Schmidt et al., 2008).

Not only do conversational partners represent information as common and privileged, but they additionally bind these representations to the individuals with whom common ground is held. For example, if Forrest and Charles have a private discussion in which they invent a name for an abstract image such as “the dancer,” they only attribute this knowledge to each other. When talking with a new person who was not privy to the original conversation, Forrest would likely go into more detail if he wished to reference the same image, thereby accommodating the new partner’s lack of common ground for the name (Brennan & Clark, 1996; Wilkes-Gibbs & Clark, 1992), and facilitating their understanding (see Schober & Clark, 1989). Further, when addressing a mixed audience in which some addressees do have common ground for the name, and some do not, speakers are sensitive to the presence of a naïve individual and provide more detail to establish reference (Yoon & Brown-Schmidt, 2014). These findings show that common ground is formed with specific partners and is distinct from one’s egocentric knowledge. Addressees similarly tailor their interpretation processes to the knowledge that is shared with the current speaker. If one speaker knows the name of a novel toy, but a second speaker does not, addressees expect knowledgeable, but not unknowledgeable speakers to use the toy’s name (Metzing & Brennan, 2003; Brown-Schmidt, 2009; cf. Barr & Keysar, 2002). Successfully taking the perspective of one’s discourse partner therefore requires memory for whether discourse-relevant information, such as the name of a toy, is or is not common ground, and critically, with whom.

4. The role of memory in common ground

The central issue, and focus of this topic, is how common ground is represented in memory and how these representations guide language processing.

Evidence that perspective-taking is sensitive to the distinctiveness of previous experiences with specific partners (Horton & Slaten, 2012; Horton & Gerrig, 2005b; cf. Heller, Gorman, & Tanenhaus, 2012; Gorman, Gegg-Harrison, Marsh, & Tanenhaus, 2013) suggests that understanding when perspective-taking will and will not be successful must consider the ease with which the relevant, person-specific common ground can be retrieved from memory. Distinctiveness of sources is known to improve source memory (e.g., a male and a female speaker vs. two female speakers; Ferguson, Hashtroudi, & Johnson, 1992; Fischer, Schult, & Steffens, 2015); thus, in conversation speakers may have more success at remembering what is common ground with a specific person if that individual is distinct. Studies in the memory literature show that when source memory errors do occur, participants are likely to misattribute a statement to a person from the same social category as the true source (Klauer & Wegener, 1998); extended to conversation, this suggests that speakers may misattribute common ground to people who are confusable or belong to the same social category. Indeed, Horton and Gerrig (2005a) suggest that memory errors may cause perspective-taking failures, for example, when information that is associated with another person is erroneously assumed to be common ground when it is not. Other work points to age-related declines in the use of common ground (Horton & Spieler, 2007). Perspective-taking failures can also occur when information in privileged ground is highly salient (Wardlow Lane & Ferreira, 2008), emphasizing the importance of understanding domain-general cognitive constraints, including individual differences (Lin, Keysar, & Epley, 2010; Wardlow, 2013) in these processes.

Despite being central to language, it is not clear how common ground is stored in memory. For example, it is not known when common ground is encoded—during the initial encoding of an event, or whether common ground is inferred or assumed later. It is also not known what biological memory systems support common ground, with candidate systems, including episodic memory and the declarative memory system, as well as non-declarative memory mechanisms.

The classic view posits a central role for memory of past events, including the happenings and event participants. The idea is that memories for events provide the evidence that one needs to determine whether common ground can be assumed. If another person experienced an event with you, he or she can be presumed to have common ground for the event (Clark & Marshall, 1978, 1981). Clark and Marshall use the metaphor of a diary and a reference text such as an encyclopedia to describe how one’s memory for past events (like a diary) and one’s knowledge of the world (like an encyclopedia) can allow a person to determine whether something is or is not common ground. For example, Clark and Marshall (1978) write that common ground requires diary-like memory for an event:

What we need, in summary, is a diary of the significant events in our own personal experience, supplemented by cultural histories and atlases for cultural co-presence and by various reference texts for indirect co-presence. (p. 63)

Clark and Marshall (1981) elaborate on this idea, proposing that when interpreting definite reference, one searches through memory for an event that provides the necessary basis (“G”) to assume common ground for the thing that is being referred to. They give the example of Bob interpreting Ann’s definite expression “the man in the red shirt”:

When Ann uses the reference the man in the red shirt, Bob must find in memory an individual who fits that description . . .he must seek out an event that he can use along with certain auxiliary assumptions as the basis G for inductively inferring mutual knowledge of the identity of that man. (p. 53)

On this event-based view, access to memory for an event allows a person, working with a few assumptions, to determine whether something is or is not common ground. While this view continues to be influential, it is unclear, however, how to view Clark and Marshall’s (1978, 1981) metaphor of a diary and encyclopedia through a modern lens. As we shall see, this metaphor has been interpreted in different ways, and has inspired multiple lines of inquiry that follow up on different aspects of this proposal.

Horton and Gerrig (2005a,b; also see Pickering & Garrod, 2004) react to Clark and Marshall’s proposal that use of common ground relies on reference diaries. Specifically, they challenge the notion that special-purpose, person-specific memory structures or mechanisms are necessary to account for common ground. These approaches also raise the question of whether explicit, consciously accessible representations can be accessed rapidly enough to guide language processing. Horton (2007) suggests that sensitivity to common ground may be possible in the absence of explicit recall of what a conversational partner does and does not know, an idea that is quite different than the deliberative and explicit consideration of event memory described by Clark and Marshall’s early view.

Horton and colleagues (Horton, 2007; Horton & Gerrig, 2005a,b) alternatively propose a resonance-based theory of common ground where information in working memory (such as the identity of the current discourse partner) resonates with information in long term memory in a fast and automatic fashion. Inspired by the psychological construct of resonance (Ratcliff, 1978), Horton and Gerrig suggest that information becomes accessible to language users when stored memories, and the overlapping features among cues and memory traces, reach some activation threshold. For example, Horton (2007) suggests that some common ground effects are mediated via ordinary mechanisms of implicit memory:

the observed priming effects were mediated primarily via implicit memory mechanisms [. . .] these findings show how ordinary memory mechanisms can influence the accessibility of information associated with particular interpersonal contexts (p. 1131)

On Horton and Gerrig’s view, a combination of automatic and strategic assessment of commonality—processes which are based on ordinary memory search—support a variety of common ground based processes (also see Shintel & Keysar, 2009). One unintended consequence of this challenge to the notion of explicit calculation of common ground was that some viewed Horton and Gerrig’s proposal as arguing for common ground representations that did not depend on episodic/declarative memory (i.e., the conscious recollection of previous events). In the present issue of topiCS, Horton and Gerrig elaborate and clarify their view on these issues. More generally, Horton and Gerrig’s findings have been provocative, inspiring a range of new explorations in the literature examining the interactions of memory and common ground in language processing (e.g., Bard et al., 2007; Barr, Jackson, & Phillips, 2014; Knutsen & Le Bigot, 2012).

Our own work was inspired by Clark and Marshall’s hypothesis that memory for events is critical for common ground. As a result of this emphasis on event memory, we have interpreted Clark and Marshall’s metaphor as a proposal for the role of episodic memory (a form of declarative memory that includes memory for events) in common ground (Duff & Brown-Schmidt, 2012). However, such claims are not explicitly made in their paper. Further, beyond one publication (Rubin, Brown-Schmidt, Duff, Tranel, & Cohen, 2011), no study has investigated whether or not episodic memory is, in fact, necessary for common ground. In our view, multiple forms of memory support common ground (Duff & Brown-Schmidt, 2012; Duff, Gupta, Hengst, Tranel, & Cohen, 2011; Duff, Hengst, Tranel, & Cohen, 2006; Rubin et al., 2011). Indeed, modern theories characterize memory as being manifested by multiple functionally distinct memory systems supported by anatomically distinct brain systems (Eichenbaum & Cohen, 2001; Henke, 2010). Furthermore, memory research suggests that complex human behavior is not the product of a single memory system but rather is achieved through orchestration, competition, and intermixing of multiple forms of memory (Poldrack & Packard, 2003; Voss, Lucas, & Paller, 2012).

Linking common ground to episodic memory, a form of declarative memory, makes good sense. This memory system has the processing capacity to support rich, complex representations about the diverse range of information that could be encoded in common ground (e.g., people and objects present, what was said and by whom, and contextual details), to update and integrate new information across time, speakers, and situations, and for speakers to consciously introspect and access the content of representations (i.e., I know that you know that I know; Rubin et al., 2011). Just as Clark and Marshall (1978, 1981) point to memory for events as an essential building block in common ground-based inferences, we see episodic memory for past events as a key component to many (but not all) common ground-based processes. In our own work addressing the question of the memory representations that support common ground, we take a neuropsychological approach in which we examine the ability of patients with hippocampal amnesia, who have a severe and selective impairment in episodic/declarative memory, but intact non-declarative memory, to develop and use common ground. These patients exhibit severe deficits in the acquisition of new declarative memories, including impairments in memory for new facts and events in daily life. In contrast, non-declarative learning, including skill acquisition and priming, are relatively unimpaired.

If, as Clark and Marshall (1978, 1981) proposed, explicit recollection of events is critical to inferences about common ground, then patients who have episodic/declarative memory impairment should be impaired in the use of common ground. Studies of healthy participants cannot identify the unique contribution of one biological memory system over another because in healthy participants multiple memory systems can engage, even in implicit tasks. Studying neurological patients allows for a direct test of the contribution of a given memory system to representations of common ground and offers biologically and psychologically plausible accounts of the role of memory in common ground. The first study in this line of work investigated the ability of individuals with hippocampal amnesia to acquire and use referential labels for a set of abstract tangram images while interacting with a familiar communication partner (Duff et al., 2006). Using a variant of Clark and Wilkes-Gibbs’s (1986) referential communication task, Duff et al. (2006) found that patients with hippocampal amnesia demonstrated robust learning, arriving at increasingly concise and stable labels (e.g., siesta man; angel) for the tangram images despite profound impairments in explicit/declarative recall for completing the task or with whom they interacted. This result suggests that at least some forms of common ground can be supported outside of explicit recollection and declarative memory, opening the possibility that non-declarative mechanisms (which are intact in this participant population) play some role in acquiring and representing information in these communicative interactions.

An unanswered question is whether or not the amnesic patients represented these labels as common ground with their discourse partner. One reason to think that it is possible to form common ground in amnesia is that these patients successfully use visual cues to another person’s perspective (i.e., whether or not my partner can see this thing) when interpreting a discourse partner’s expressions (Rubin et al., 2011). That finding shows that, in principle, it is possible to form representations of what another person knows in the face of severe declarative memory impairment.

On the other hand, an analysis of the form of the expressions that amnesic patients produced suggests that when producing these newly learned labels, amnesic patients did not fully appreciate that the tangram had already been discussed in the discourse. When healthy adults discuss information that is identifiable within their common ground, they typically use definite referring expressions as in the movie; when common ground cannot be assumed, speakers are more likely to use an indefinite expression as in a movie (Isaacs & Clark, 1987; also see Clark & Marshall, 1981). Unlike healthy adults, amnesia patients described shared tangrams with a definite expression only 56% of the time, whereas healthy comparison pairs described shared tangrams with definite reference to 90% of the time (Duff et al., 2011). The failure to consistently use definite reference could be taken as evidence that common ground is encoded in declarative memory, a form of memory not available to these patients. It is also possible that some, possibly nascent, partner-specific common ground representations are formed via the non-declarative memory system. Indeed, patients with hippocampal amnesia have been shown to acquire a range of person-specific information (e.g., accent, political preference, personality attributes; Coronel et al., 2012; Johnson, Kim, & Risse, 1985; Tranel & Damasio, 1993; Trude, Duff, & Brown-Schmidt, 2014). One possibility is that when evidence for what is and is not common ground is immediately available in the environment, it can be formed and used through non-declarative memory mechanisms (e.g., Rubin et al., 2011), but when faced with generating labels, the non-declarative memory system does not provide sufficient cues to implore amnesic speakers to mark mutually known information as common ground (e.g., Duff et al., 2011). Taken all together, these data suggest that different components or aspects of common ground may be encoded by distinct memory systems. Future work is necessary to fully delineate the nature, time course, and orchestration of declarative and non-declarative memory contributions to common ground.

5. Further open questions

In addition to questions about how common ground is stored in memory are a variety of related and unanswered questions. Unknowns include what information is stored: Everything having to do with joint experience? Only contextually relevant things? According to one view (Pickering & Garrod, 2004), the mechanisms of language use in conversation serve to align speakers’ and listeners’ representations of the discourse. Indeed, as common ground grows, speakers and listeners tend to have more overlap in their visual attention and better understand what is said in conversation (Richardson & Dale, 2005; Richardson, Dale, & Kirkham, 2007). However, generation (e.g., speaking) promotes memory (Marsh, Edelman, & Bower, 2001); thus, speakers are likely to better remember what was said in conversation than listeners (McKinley, Brown-Schmidt, & Benjamin, 2015; Yoon, Benjamin, & Brown-Schmidt, 2016). Biases in recall may be relevant as well, and they may even work against this memorial benefit for speakers. When participants are asked to free-recall what was said in a conversation, there can be a bias to report more of what was heard than what was said (Stafford & Daly, 1984; cf. Ross & Sicoly, 1979), possibly because what one’s partner said seems more newsworthy than what one said themselves. As a result, the true discourse record may offer a false basis for understanding how conversation is shaped by the linguistic context. Instead, each partner’s distinct memory for the discourse history, along with their beliefs and biases for what aspects of the discourse history are relevant, may provide a better starting point for understanding how the discourse history guides language use (e.g., see Heller & Chambers, 2014; Knutsen & Le Bigot, 2012, 2014).

Other open questions concern the nature of what is represented. While a number of proposals have been put forth, there has been little work directly comparing and contrasting these proposals, and few details have been offered. For example, Brown-Schmidt (2012) suggested that representations of common ground are gradient, but that proposal is vague about which aspects of the representation are gradient. By contrast Brennan and colleagues (Brennan & Hanna, 2009; Galati & Brennan, 2010) propose that the evidence for common ground can be one-bit (e.g., Charles can see this bucket, or not), but it is not clear if the resulting representations are one-bit as well, or whether they may vary in a more gradient fashion as envisioned by Brown-Schmidt. Both Horton and Gerrig (2005a) and Arnold (2016) propose that common ground is encoded not as a distinct representational tag, but rather as an emergent property of our memories. Yet another proposal is that perspective-taking reflects probabilistic weighting of information in common and privileged ground (Heller, Parisien, & Stevenson, 2016). Whether these are incompatible theories, or instead whether they are orthogonal and speak to different aspects of common ground or memory processing, has not been addressed in the literature. Related to this issue is how and when common ground is determined. While Clark and Marshall (1978, 1981) describe a very explicit process for determining mutual knowledge on the basis of event memory, it is not clear if they envisioned mutual knowledge being encoded as a part of the event or not. Horton and Gerrig (2005a,b) alternatively postulate distinct processes of commonality assessment (during which common ground is assessed) and message formulation (during which common ground is used to guide construction of utterances).

Finally, many current theoretical debates focus on the time-course of when common ground is used by the language processing system. According to the constraint-based perspective, common ground guides the on-line processing of language (Hanna et al., 2003; Heller, Grodner, & Tanenhaus, 2008). Such constraint-based views posit no architectural limitations on when, in principle, common ground might be brought to bear on the process of language production or comprehension. However, it is likely that different sources of information influence processing at different time-scales depending on the accessibility of that information in memory. Alternatively, others have argued that common ground does not routinely guide on-line processing (Keysar, Lin, & Barr, 2003), though it may have a role in anticipatory processes (Barr, 2008). On this view, using an egocentric heuristic in conversation would save resources and lead to successful communication much of the time (Keysar, Barr, Balin, & Brauner, 2000; Keysar, Barr, Balin, & Paek, 1998; Pickering & Garrod, 2004). Evidence that perspective-taking errors occur when the system is taxed have been used to motivate egocentric-first theories (Horton & Keysar, 1996; Kronmüller & Barr, 2007). Understanding what memory systems are involved and the time-course of accessing information within these systems will likely offer insights that will be relevant to resolving these outstanding theoretical debates.

6. The current topic: Memory and common ground processes in language use

Thus far, we have outlined the topic of common ground and memory, emphasizing its breadth and how this issue has been examined in the domain of language processing. Many outstanding issues are being explored now in the literature, in many different ways by different labs. In addition to our introduction, this issue of topiCS contains five articles which further describe different empirical and theoretical approaches to the relationship between memory and common ground. In addition, a future issue of topiCS will contain commentaries on these articles from those in the research community as well as author responses to those commentaries.

Arnold explores the idea that representations of common ground, and information status more generally, are emergent rather than explicit. She contrasts the notion of how this information is represented with how it is used by the language processing system–processes that include both explicit and emergent mechanisms. Duran, Dale, and Galati describe a synthesizing dynamical systems approach to perspective-taking in contexts where conversational partners hold different spatial perspectives on a scene. In their view, progress in understanding complex interactional processes will be made by understanding the dynamic interactions between multiple informational constraints, including common ground. Horton and Gerrig use this target article as an opportunity to expand on their ordinary memory theory of common ground, including clarifying the claims of the theory and addressing mischaracterizations of the theory in the literature. These authors join Arnold in eschewing specialized mechanisms for representing common ground, and instead favor a view in which the memorial record binds together different people and experiences. Knutsen, Ros, and Le Bigot examine speakers’ repetition of content in both face-to-face dialog and phone conversations. Their empirical findings illustrate how language use is guided by contextually constrained memorial representations of the discourse history, and more generally, align with Duran et al.’s proposal that conversation is guided by multiple sources of probabilistic information. Finally, Masson-Carro, Goudbeek, and Krahmer explore the process of grounding, in which conversational partners interactively establish information as common ground. They focus on the roles of gesture and an individual’s cognitive constraints in this process and, in doing so, illustrate the influence of common ground across multiple communicative modalities. Together, the set of papers in this special issue illustrates a broad range of exciting new research directions for the study of common ground in language use. A unifying thread throughout this work is that communicative processes are multiply determined by representations that emerge from our past experiences. These experiences involve the people in the world around us, and it is the memorial record of our joint experiences that forms the basis of processes such as audience design.