ࡱ> =?:;<a #jbjb11 (~[[    {{{8 || }:}"}}}}~<~ -//////,lRz[-~}}~~[<}}<<<~}}-< <~-<<hx} {ކ^I4lY8<8 < 7DD7 D Collaborative Recall in Face-to-Face and Electronic Groups Justina Ohaeri Ekeocha William Paterson University and Susan E. Brennan State University of New York at Stony Brook Please address correspondence to: Justina O. Ekeocha Department of Psychology William Paterson University Wayne, NJ 07470 Email:  HYPERLINK "mailto:ekeochaj@wpunj.edu" ekeochaj@wpunj.edu Fax: (973) 720-3392 Telephone: 973-720-3398 Running Head: Effects of collaboration on recall Abstract When people collectively remember a shared experience, the amount they recall as a collaborating group is less than the amount obtained by pooling their individual memories. We tested the hypothesis that reduced group productivity can be attributed, at least in part, to content filtering, where information is left out of the group product either because individuals fail to retrieve it or choose to withhold it (self-filtering), or because the group rejects or fails to incorporate it (group-filtering). Three-person groups viewed a movie clip together and recalled its contents, first individually, then in groups, and finally, individually again. Although both kinds of groups recalled equal amounts, group-filtering occurred more often in face-to-face groups, while self-filtering occurred more often in electronic groups. This suggests that reduced group productivity is due not only to intrapersonal factors stemming from cognitive interference, but also to the interpersonal costs of coordinating the group product. ConverselyFinally, face-to-face group interaction facilitated subsequent individual recall. Acknowledgments This material is based on data collected as part of the first authors dissertation (Ohaeri, 1998) and supported by a W. Burghardt Turner Fellowship. We thank Richard Gerrig for his comments, Committee Members Suparna Rajaram and Marci Lobel for their expertise and guidance, and Ricardo Carrion, Tara Defantis, Christina Paul, and Amanda Pisino for their assistance in running the experiment, transcribing, and coding. Introduction Traditionally, research on human memory has treated recall as a purely intra-personal process, by focusing on cognitive processing by individuals recalling alone. In everyday life, however, recall is often an interpersonal process in which one person recounts an experienced event to another who was not present, or in which people reconstruct an event they have experienced together. What they recall does not take the form of an orderly list of items, but may consist instead of proposals, elaborations, comments, expressions of uncertainty, and invitations to agree or disagree. Consider this exchange among three volunteers in our collaborative memory experiment, recalling a movie they saw together: A: yeah, and that- the kid was like talking, something L: yeah he was talking thats why he got in trouble he was whispering in some kids ear about something? A: yeah L: um... then he gets punished or whatever? D: what was that, a wreath or L: yeah it was some kind of browny- A: yeah it was some kind of straw thing or something L: mhm D: around his neck L: so that everybody knew what he did or something? As A, L, and D reconstruct what they saw together, they take turns (in no predetermined order). A begins with a vague proposal, which L elaborates by providing details that she seems to invite A to ratify (and he does so). L continues with another installment, which D confirms by proposing a detail that is in turn confirmed with additional information provided by both L and A. The resulting product is richly detailed and reconstructed jointly, sometimes with one person completing anothers utterance. The speakers display their confidence (or lack of confidence) about what they present, using hedges and rising intonation. After each presentation, the others display acceptance or uptake either with direct evidence (e.g., an acknowledgment such as mhm) or indirectly (by offering the next installment). The point is that when people recall events together, the product is shaped by their interaction. Remembering in a social context is a function of both intrapersonal and interpersonal actions, including the cognitive and communicative behaviors by which members evaluate, pool, and assemble their resources (Steiner, 1972). It is evident from the example that people recalling together can cue each other, which supports the expectation that a person recalling in the context of a group might be more productive than one recalling alone. This reasoning was pursued by Edwards and Middleton (1986) who analyzed the collaborative recall of a popular film, finding examples where one persons recollections seemed to serve as a recall cue for another. Another way in which collaboration might boost collective memory performance, but at the encoding stage, has been labeled transactive memory by Wegner (1987); the idea is that pairs of collaborators who know each other well can judge how best to distribute the labor of encoding information, based on what each partner knows about the others expertise. Within the transactive memory framework, it has been shown that pairs of friends perform better together than do pairs of strangers (Hollingshead, 1998; Wegner, Erber, & Raymond, 1991). The transactive memory approach tends to focus on the process of communication during encoding; for instance, Hollingshead (1998) found that dating couples were better at pooling their knowledge when they interacted face-to-face than when they interacted electronically. Although two heads are better than one in that people working together recall more than one recalling alone, it is striking that collaborating groups do not recall as much as would be expected from combining individual products into nominal group products (formed by additively pooling non-redundant products of the same number of individuals who recalled alone). The finding that real groups recall less than nominal groups is highly robust (e.g., Basden, Basden, Bryner, & Thomas, 1997; N. Clark, Stephenson, & Kniveton, 1990; Finlay, Hitch, & Meudell, 2000; Hartwick, Sheppard, & Davis, 1982; Morrissett, Crannell, & Switzer, 1964; Stephenson, N. Clark, & Wade, 1986; Weldon & Bellinger, 1997; Weldon, Blair, & Huebsch, 2000; Wright & Klumpp, 2004). Additive models based on individual performance (e.g., Lorge & Solomon, 1955) simply do not scale up to what happens in real groups. What causes this loss of productivity? One possibility researchers have considered is social loafing. Social loafing is a reduction in motivation and effort that may occur when individuals work collectively (to produce one group product) compared with working alone. Variables associated with social loafing include diffusion of responsibility, evaluation potential, and evaluation apprehension (see Karau & Williams, 1993 for review). Weldon, Blair, & Huebsch (2000) investigated this explanation for under-performance in group recall by manipulating motivation, evaluation apprehension, personal accountability, and group cohesion. None of these manipulations eliminated the inequality between nominal and collaborative group products, leading Weldon et al. to conclude that lower-than-expected productivity in groups was not due to motivational factors. It is possible, however, that other manipulations may support a social loafing explanation; for instance, if t The submergence of individuals in a group provides a cloak of anonymity that allows them to reduce their effort, then electronic groups (where members interact by typing and have a harder time tracking who has typed what because they cannot see or hear one another) may recall less than face-to-face groups (see Williams, 1977; Kiesler, Siegel, & McGuire, 1984; and Diener, 1979 for some comparisons of face-to-face and computer-mediated behavior). As research on recall by collaborating groups has burgeoned in recent years, reduced group productivity has also been explained in cognitive terms, as collaborative inhibition (see Basden, Basden, Bryner, & Thomas, 1997; Weldon & Bellinger, 1997). The idea is that hearing other peoples contributions disrupts individual organizational and retrieval strategies, an effect similar to part-set cuing inhibition, whereby providing (at recall time) some items from a studied list inhibits the retrieval of the remaining list items (for a review of part-set cuing, see Anderson & Neely, 1996; see also Basden & Basden, 1995 on part-set cuing as strategy disruption and Diehl & Stroebe, 1991 on production blocking). In studies that involve recall of word lists, the words recalled tend to be ordered by category (both within an individuals contributions and across individuals contributions), suggesting that when one person switches from one to another category, this disrupts partners recall of remaining items from the first category (Basden, Basden, Bryner, & Thomas, 1997; Finlay, Hitch, & Meudell, 2000; Basden, Basden, & Stephens, 2002). Although disruption of the organization of memory is a plausible explanation for reduced group productivity for recall of arbitrary items such as lists of words, it is less satisfying for naturalistic material, such as the episode co-recalled in our opening example. Unlike a list of words, a story has meaning, coherence, and an inherent organization that typically depends on the logical flow of events. Indeed, there is some evidence that cueing inhibition is less of a problem for meaningful, well-integrated memories (Basden et al., 1997). Basden et al.s (1997) results show that when highly organized material (e.g., small categories of items) is used in recall, collaborating groups are sometimes capable of recalling as many category items as nominal groups. This result was attributed to well-organized materials leaving little room for idiosyncratic organization (Basden et al, 1997). However, Weldon and Bellinger (1997, Experiment 2) used an audio recording of the War of the Ghosts story (first used by Bartlett, 1932) to test recall during two sessions, performed by individuals either alone or in three-person groups. The puzzle is that even though groups had shared knowledge of the storys organization that they could use as a retrieval strategy, real groups still under-performed nominal groups. We propose that collaborative inhibition is not the only force behind reduced group productivity in real-world settings. Notably, most studies to date have considered the products of group recall, but not the interactive, interpersonal process by which these products emerge. In fact, some studies have eliminated interaction altogether by ruling out communication or limiting initiative by group members, forcing them to take ordered rather than spontaneous turns contributing items to the group product (e.g., Wright & Klumpp, 2004; Basden, Basden, & Henry, 2000). Some studies have emphasized stable input factors that are determined in advance, like group composition (e.g., Stephenson, Clark, and Wade, 1986; Stephenson, Brandsttter, & Wagner, 1983; Vollrath, Sheppard, Hinsz, & Davis, 1989; Weldon, Blair, & Huebsch, 2000) and encoding conditions (e.g., Finlay, Hitch, & Meudell, 2000; Weldon, Blair, & Huebsch, 2000), testing the impact of these input factors on group output. This highly controlled approach is useful for zeroing in on collaborative inhibition by itself, but rules out additional processes that, we propose, can fundamentally shape recall in small, interacting groups. To better understand collaborative recall, we look not only at what memories groups produce, compared to individuals, but also at what group members spontaneously say while coordinating recall. It stands to reason that collaborative processes can affect collective products. "Process," here, refers to what Steiner (1972) described as the individual or collective actions of people assigned to do a task, including actions both intrapersonal and interpersonal, as well as the intellectual and communicative behaviors by which they evaluate, pool, and assemble their resources. Here, we examine collaborative memory in the context of dynamically emerging behaviors through which group products are realized. Consider the reasons why elements that an individual member is capable of recalling alone might fail to make it into the groups collective recollection. Retrieval failure due to collaborative inhibition is only one of them. Individuals may also choose to withhold items that they nevertheless can recall correctlythat is, not to contribute them for the group productwhen theyre not confident enough about accuracy or relevance, when its too costly to express or explain something, or when they anticipate that it may be difficult to get the items accepted by the group. We label those cases in which individuals take the initiative to withhold material from the group product as self-filtered. According to Halbwachs (1980), recall in groups involves a process of social comparison. People compare their recollections to those of others with whom the experience was shared. When their memories coincide, their confidence in the memories is likely to increase. If memories diverge, people will probably be less confident. For example, if one person recalls an item that in some way contradicts what another has just presented, this might lead the first person to re-evaluate and maybe suppress (self-filter) the divergent idea. A second way in which items can fail to make it into the group product is through group-filtering; that is, an individual may recall and present an item, but the item fails to make it into the group product. This may happen in several ways; it may be unintentionally omitted if it doesnt get recorded and no one notices, or it may be intentionally omitted, because another group member either overtly or covertly rejects it. Furthermore, if a group sets a goal or criterion, formally or informally, members may feel pressured to go along with it. In the case of story recall, for example, findings by Stephenson, Clark, and Wade (1986) suggest that groups adopt a strategy of tracing the chronological development of events, with emphasis placed on the main ideas of the story. It is possible, however, that the "program" the group has set for itself might not accommodate all the memories of its members, leading to the suppression of certain items in deference to the group goal. Stephenson et al. (1986) subsequently concluded that collaboration led to the selection of ideas from the total pool of available individual ideas, based on the group's decision as to what was appropriate to talk about. This does not tell us, however, whether the initiative for selecting information occurred at the group or the individual level. Examining the contents of individual and group recall products and comparing them to transcripts of groups communication will, we propose, shed some light on the processes involved. In order to understand the impact of coordination upon collective recall, we will compare situations in which the coordination costs are different. When people interact, the medium in which they do so shapes how they coordinate their activities. A particular medium makes some things easier and others, harder (Brennan & Lockridge, 2006; Clark & Brennan, 1991). For instance, speech is fairly effortless for most people to produce (but is ephemeral), whereas typing is slower and more difficult (but creates a representation that can be reviewed and edited). These costs and affordances shape the ways in which people distribute their initiative within a group. Elsewhere, we have used this grounding framework (Clark & Wilkes-Gibbs, 1986) to predict and explain differences that emerge between electronic vs. face-to-face groups (Brennan & Ohaeri, 1999); we will appeal to that framework here. Grounding is the process of seeking and providing evidence to establish mutual understanding with a partner (Brennan, 2005), and holds that collaborators seek to minimize the effort they expend jointly (rather than just their individual effort; Clark & Wilkes-Gibbs, 1986). Collaborative recall involves not only retrieving information from memory (a step presumably sensitive to collaborative inhibition), but also evaluating that information as to its correctness and suitability for being included in a collective product. Initiative for the latter step can be taken by the individual (self-filtering) or ceded to the group (when the individual presents items, even low-confidence ones, for the group to accept or reject). The grounding framework predicts that when group members can see and hear each other and easily produce speaking turns and non-verbal cues, individuals should rely more on other group members to assess and filter items for the collective product. When the cost of producing a turn or getting a response is higher (as it is for typing and then waiting for a response to come back), then it is more efficient for individuals to self-filter their contributions to the collective product. When people interact, they are driven not only by task-related needs, but also by social needs or face management, which include preserving ones own self-esteem, not threatening that of others, providing options to others, not imposing, and showing solidarity (Brown & Levinson, 1978). If one group member presents an idea she is uncertain about (especially if the other members might recall it correctly), she risks losing face unless she indicates that uncertainty. And if another group member explicitly rejects her idea, he risks insulting her. In both cases, group members who have nonverbal options for expressing uncertainty about their own and each others ideas should find it easier to negotiate which propositions that come up during the interaction survive to make it into the collective recall product. Elsewhere we have found that face-management costs differ in face-to-face versus electronic interaction (Brennan & Ohaeri, 1999). Our main goals in this paper are to investigate the source of productivity loss in collaborative groups as compared to nominal groups, and to examine the impact of recalling in groups upon recall by individuals later on. We did this by having triads of participants (who were strangers) recall a movie that they had watched in groups of three. They recalled first alone, then in groups interacting spontaneously either face-to-face or electronically, and finally, alone again. This design provided an initial baseline for examining the impact of recall in groups, as well as an opportunity to test the impact of group recall on subsequent individual recall. We examined the content of collective recall products, as did previous studies, but we also compared the group product to what group members said as they discussed and produced it. Finally, we manipulated the instructions given to face-to-face groups: half of our face-to-face groups were explicitly instructed to come to a consensus on all submitted material and half were not. We added this manipulation to check for the possibility that group productivity might suffer due tofailure to achieve consensus. In many previous studies of collaborative recall (e.g., Hartwick et al., 1982, Warnick & Sanders, 1980, Vollrath, Sheppard, Hinsz, & Davis, 1989, Stephenson et al., 1983, Stephenson et al., 1986, Stephenson et al., 1991, Hollin & Clifford, 1983, and N. Clark et al., 1990), groups wereexplicitly instructed to reach consensus. However, we did not make any predictions with regard to this manipulation. We summarize our predictions as follows: Group recall products will be greater than initial individual recall products, not only in the quantity of propositions recalled, but also in their quality (with more correct and fewer incorrect items), due to groups having more resources for correcting errors than do individuals. As in previous studies, interacting groups will under-perform compared to nominal groups. This will be due to filtering mechanisms on the part of both group and individuals. If social loafing were a significant force, face-to-face groups should recall more proportions than electronic groups because members are more identifiable when they can see and hear one another. However, consistent with the grounding framework, we predict that both kinds of groups will adjust their joint effort to do the collective recall task equally well (producing products of equivalent quantity and quality). Group members will distribute their filtering efforts according to the costs of coordinating individual actions within the communication medium. There should be more group-filtering face-to-face, and more self-filtering with electronic text communication. If groups indeed have superior resources for filtering out incorrect items and reinforcing items that an individual member did not recall alone, then recalling in a group should improve the subsequent solitary recall (individuals should recall more after recalling in a group with other people than before). A post-group boost in recall by individuals should be especially likely when the communication medium affords easy-to-use mechanisms for proposing and ratifying contributions to the joint product (e.g., as non-verbal cues do when people interact face-to-face). Method Participants One hundred thirty Stony Brook University undergraduates volunteered to participate in exchange for two research credits that they could use to satisfy a requirement in a psychology class. Participants were required to be native speakers of English and to be able to type. They were also required not to have seen the movie The Secret of Roan Inish. The experiment took approximately two hours. Thirty nine participants were tested in three-person groups in the electronic condition (with consensus requirement), 39 were tested in three-person groups in the face-to-face condition (with consensus requirement), 39 were tested in three-person groups in the face-to-face condition (without consensus requirement), and 13 were tested alone in the control condition. Group members were grouped according to the appointment times they happened to sign up for, and groups were randomly assigned to conditions; no attempt was made to balance for gender composition. Approximately 70% of the participants were female. Design We tested the effects of repeated recall within-subjects; each participant in the experimental conditions was first tested alone (Session 1), then in collaboration with two other participants (Session 2), and finally alone again (Session 3). Session 1 provided a baseline for what each participant could recall alone, and Session 3 provided a measure of the strength of the groups influence on individual memories. The second variable, communication medium (face-to-face vs. electronic), was tested between-subjects; collaborating groups during Session 2 were tested in either the face-to-face (speech) or the electronic (text) condition. A third variable, instructions, was tested between subjects; the groups in Session 2 interacted with or without an explicit requirement to reach consensus. This variable was manipulated for face-to-face groups only, as a check on whether three people recalling together would attempt to reach consensus naturally, and so that the results would be comparable to previous face-to-face studies in which consensus was required. In sum, there were three group conditions for Session 2, each involving 13 three-person groups: Electronic (with consensus), Face-to-face (with consensus), and Face-to-face (without consensus). Thirteen additional participants were tested alone over three recall sessions to provide a control condition in order to provide a potential baseline for other known influences on repeated recall: (1) a possible fading of memory over time and (2) hypermnesia, or "net improvements in recall across tests" (Wheeler & Roediger, 1992, p. 241). Materials All participants viewed an 8-minute clip from a John Sayles film (The Secret of Roan Inish), of an old man telling a story to his granddaughter. The presentation was on a 19" color television screen. Electronic collaboration was carried out using networked Macintosh computers with 16" monitors and Aspects"!, a collaboration software program by Logic Technologies. For the electronic condition, a chat window and a blank document for text-editing were opened side-by-side. The chat window was used for idea presentation and discussion and indicated which individuals (identified only by the labels A, B, or C) produced which statements. The text document was used to paste and edit the group recall product. The two windows were displayed side by side on each group members monitor, so that information in the chat and text windows was shared by all three group members. At the bottom of the chat window was an edit window for typing contributions. This was the only space that was not shared; that is, the edit window was visible only to the person typing in it. Ideas were then sent to the chat window by hitting the 'Return' button. All editing was done in the edit window (and edits not sent to the chat window were not permanently recorded); once ideas were sent to the chat window, they could not be edited. The text document was fully editable, but only by the designated scribe. Procedure Study phase. Each three-person group watched the movie clip together in the same room, after being instructed to pay close attention to what happened in the movie. Then participants did a five-minute distractor task (listing as many countries as they could think of). In the control condition, each participant watched the clip alone and then did the same distractor task. Session 1 test: Pre-Group individual recall. Following the distractor task, each participant was seated in a separate room (with closed door) where they typed their recollections using Microsoft WORD. They were instructed to recall as much of the movie as they could and to be as complete and accurate as possible. There was no time limit imposed on this task. Group members who finished first stayed in their rooms and waited for the others. When all were finished, they rejoined the group. Session 2 test: Group recall. Following the pre-group session, participants were assigned to one of the three collaborative group conditions: face-to-face (with a consensus requirement), face-to-face (without a consensus requirement), or electronic (with a consensus requirement). In the face-to-face conditions, a tape-recorder and microphone were set up beside the computer to record all spoken discussion by the group, and the three group members sat around one computer, so that all could see the monitor. Before the session began, the experimenter asked for one member to volunteer as the session scribe, responsible for typing up the groups product. If no one volunteered, the experimenter randomly assigned this role to one of the members. Typing (rather than handwriting) was required to allow face-to-face groups the same editing capability as electronic groups. In the electronic condition, after brief instruction on how to use the collaboration software, participants were sent to their separate rooms. As in the face-to-face condition, one participant had the task of pulling the group product together onto the text document. While all participants could see and comment on what was written on the text document, only the assigned person could write in it. A transcript of the session (both chat and text windows) was saved. Group members were instructed to discuss the story with one another and to come up with a group account of the story. Groups with a consensus requirement were furtherinstructed to include only those items that they could all agree on; those without the requirement were not given explicit instructions about agreement. There was no time-limit set on this task. Session 3 test: Post-group individual recall. After Session 2, documents were again saved and closed, and blank ones were opened. Participants were then asked to recall the film clip alone in their separate rooms one last time. As in the other sessions, there was no time limit. After this session, participants were debriefed, thanked for their help, and dismissed. Coding and analysis Two undergraduate research assistants naive to the experimental hypotheses were trained by the first author to do the coding. First, the coders watched the movie clip a couple of times to become familiar with the story. They were also provided with a written transcript of the clip, to serve as a reference in coding what had been correctly recalled from the stimulus. Coding was done in several phases, as follows. Segmenting recall products. Products from Sessions 1, 2, and 3 were segmented into propositions that contributed new information. Noun phrases indicating place or sequence of events, descriptive adjectives, temporal and spatial locatives, and quantities were coded as independent propositions. Proper names or other references to individuals, as well as adjectives, were counted at first mention only (redundant expressions were not counted). For example, "an old wrinkly man" was counted as three propositions at first occurrence but was not subsequently segmented in reference to the same man. Phrases containing no information relevant to the content of the story (e.g., "she said something here but I forgot what it was") were not coded. Verb phrases were segmented into propositions as follows. For intransitive verbs, the propositional unit cutoff was after the verb. For sentences with transitive verbs the cutoff was after the object of the sentence. However, for sentences that also contained proper names, adjectives, etc, this transitive/intransitive rule was overridden in order to identify these items as independent propositions (see http://www.psychology.sunysb.edu/sbrennan-/memory.htm for a sample segmented file). Categorizing recall products. Propositions were coded into four mutually exclusive categories: Correct details of the story (whether central or peripheral); inferred details, consistent with the story, but not explicitly depicted; meta-statements, including, for instance, comments on story structure, expressions of the participant's attitude toward an event, or judgments about characters (N. Clark & Stephenson, 1990): and incorrect information. Note that mis-identifying a character at first mention was coded as an incorrect proposition. However, if the information subsequently provided in relation to this character was correct (even if the identity was still wrong), the information was coded as correct (to avoid cascading errors). Computing nominal products. For each group, the three pre-group individual products were pooled, including redundant ideas only once. This yielded a measure of quantity (total propositions) but not quality for nominal groups, as it is not clear for a given item how to meaningfully combine errors with correct propositions. Coding of group interaction transcripts. The audiotapes from the face-to-face group sessions were transcribed and double-checked for accuracy; the transcripts from electronic sessions were logged automatically. Word counts were computed for each group discussion (excluding unintelligible speech and nonverbal sounds such as laughter). Transcripts of the group sessions were organized into turns labeled with an identifying letter for each group member. Turns were segmented into propositions and classified using the same criteria and categories as for the recall products. Finally, coded transcripts from the interactions in Session 2 were compared with the coded recall products from Sessions 1, 2, and 3. Propositions presented by individuals during discussion but not included in the official group product were coded as having been group-filtered. Propositions appearing in the pre-group product but in neither the official product nor the interaction transcript from the group session were coded as having been self-filtered. Note that this category did not distinguish among items that failed to be retrieved, or that were recalled and forgotten before they could be presented, or that were recalled but intentionally withheld. Items recalled in Session 1 but not presented to the group in Session 2 were coded as restored if they reappeared in Session 3. Incorrect items were examined to determine how many errors from group products persisted in post-group products. Reliability. The coders received detailed written and verbal instructions from the lead author. They segmented a given recall product independently and then met to resolve any discrepancies, ending up with a single segmented file. Finally, they independently categorized the segmented propositions in this file. They did this for one entire individual and one entire group recall product, yielding 80% and 92% agreement, respectively (because individual protocols included more inferences, meta-statements, and peripheral information, additional instruction about those categories was provided). Then, both coders independently segmented and categorized an arbitrarily chosen set of 39 recall products and interaction transcripts (approx. 11% of all products and interaction transcripts from the experiment, which included 3 products from each of 13 controls plus 7 products each3 in Session 1, 1 in Session 2, and 3 in Session 3from the 39 triads, plus the 39 interaction transcripts from Session 2). To ensure that coding was consistent, coders then met to identify, discuss, and resolve any discrepancies in either segmentation or categorization, agreeing on a single version of the coding for each file. The coders were then each randomly assigned to code the rest of the folders; the same person coded the pre and post-group individual protocols, as well as the group product and session transcript, for a given group of three. Analyses. We analyzed recall products for both quantity and quality. Concerning quantity, mean numbers of propositions per product were compared for overall productivity. For quality, proportions of correct items, incorrect items, inferred items, and meta-statements were calculated for each product. Because some of the proportions were small for some of these categories (e.g., incorrect items), Arsin transformations were done on each proportion to normalize the distribution of values, as suggested by Lindman (1974). Unless otherwise specified, 3-person groups were treated as the random factor. Repeated measures ANOVAs with session as a within factor and communication medium as a between factor were used to compare products of the 39 groups to mean individual recall products (each groups mean from Session 2 compared to the mean of its 3 individual members in Sessions 1 or 3), as well as for comparing pre- and post-group individual products. ANOVAs (repeated measures wherever the same individuals were involved) were used for planned comparisons. Repeated measures ANOVAs were used to compare true group to nominal group products, as well as to look for hypermnesia by the 13 controls who recalled alone repeatedly. Consensus requirement. There was no difference in productivity between face-to-face groups with the consensus instruction (79.69 propositions) and those without (76.08), F(1, 24) = .19, MSE = 450.82, n.s. Neither were there any differences in the proportions of the different measurement categories. Consequently, to simplify the analyses, data from these two otherwise identical face-to-face conditions were collapsed for analyses (yielding 26 face-to-face groups and 13 electronic groups). Results Recall by Individuals, Groups, and Nominal Groups Total Productivity: As expected, three-person groups recalled more (with all measurement categories combined) from the film (Session 2) than did the average individual acting alone (Session 1), F(1, 37) = 14.68, MSE = 334.22, p < .001. Also as expected, collaborative group products were less than nominal group products (the pooled non-redundant items recalled by individual group members in Session 1), F(1, 37) = 160.37, MSE = 259.51, p < .001. Product Quality: Table 1 shows total numbers of items recalled, broken down into correct, incorrect, inferred, and meta-commentary. The proportion of correct items in group products was greater than that in pre-group individual products, F(1, 37) = 14.91, MSE = .0027, p < .001. A lower proportion of errors were made in groups than had been made by the same people recalling alone in Session 1, F(1, 37) = 10.83, MSE = .001, p = .002. This is evidence that group products were of higher quality than pre-group individual products, suggesting that groups have mechanisms for error-checking that individuals do not have. Individual products in Session 1 also tended to include more idiosyncratic commentary than did group products; the proportion of meta-statements was significantly greater in pre-group individual than group products, F(1, 37) = 8.27, MSE = .00049, p = .007. There was no difference in proportions of inferences made in Sessions 1 v. 2. _________________________________ Insert Table 1 about here __________________________________ Productivity of Face-to-Face versus Electronic Groups Total Productivity: Only half as many words were typed by electronic groups as were spoken by face-to-face groups (886 v 1646), t(24) = 3.79, p < .001. This is what we expected, as many of our participants were not fast typists (see Brennan & Ohaeri, 1999). However, electronic and face-to-face groups both did the task well, recalling equal quantities of propositions, F(1, 37) = .17, MSE = 448.15, n.s. This finding is consistent with studies of mediated communication; even when bandwidth is limited or production is more difficult, collaborators adjust their effort in order to meet the performance criteria of the task (Brennan & Lockridge, 2006; Brennan & Ohaeri, 1999; Clark & Brennan, 1991; Karsenty, 1999; Whittaker, 2002). That electronic group members, even with their increased opportunities for anonymity, recalled just as much propositions as face-to-face groups provides further evidence against a social loafing explanation. Product Quality: There were no significant differences between face-to-face and electronic groups (Session 2) in proportions of correct, incorrect, inferred, or meta-statement propositions (see Table 2). _________________________________ Insert Table 2 about here __________________________________ Effects of recalling in groups on subsequent recall by individuals Total productivity: Recalling the story as a group (Session 2) improved subsequent individual recall; individuals recalled an average of 13.36 more items in Session 3 than they had recalled in Session 1, F(1, 37) = 22.31, MSE = 164.06, p < .001 . Participants in the control condition who recalled the film clip alone for all three sessions showed no productivity differences between Sessions 1, 2, and 3 (68.61, 68.85, and 67.15 propositions respectively), F(2, 24) = .29, MSE = 37.86, n.s. So hypermnesia (see Wheeler & Roediger, 1992) is not a likely explanation for the boost in post-group recall. The communication medium used by a group did affect subsequent recall by its members; solitary recall after group recall interacted with the groups communication medium, F(1, 37) = 7.35, MSE = 92.09, p = .01 (see Figure 1). Individuals from face-to-face groups recalled 15.46 more items when working alone in Session 3 than did those from electronic groups, F(1, 37) = 6.03, MSE = 343.68, p < .05. This difference was apparently due to the experience of recalling in a face-to-face group, as there had been no initial productivity differences among these two randomly-assigned groups of individuals when they recalled alone in Session 1, F(1, 37) = .36, MSE = 222.59, n.s. In fact, after collaborating face-to-face, individuals actually recalled 6% more (4.63 more items) in Session 3 than their groups had recalled together in Session 2, F(1, 25) = 4.28, MSE = 64.98, p < .05. Collaborative recall conducted via text was not so helpful for electronic groups; in fact, they recalled 7.87 marginally fewer items (11%) when working alone again in Session 3 than they had recalled together in Session 2, F(1, 12) = 2.71, MSE = 148.55, n.s. _________________________________ Insert Figure 1 about here __________________________________ Product Quality: Overall, groups were still more accurate than individuals were when recalling post-group, F(1, 37) = 6.37, MSE = .0019, p < .02. Recalling in a group improved individuals recall performance from Session 1 to Session 3, with an increase in the proportion of correct items in post-group individual products compared to pre-group individual products, F(1, 37) = 4.62, MSE = .0017, p < .05 (see Table 1). Converging evidence comes from a small but significant reduction in the proportion of incorrect items in individual recall in Session 3 compared to Session 1, F(1, 37) = 7.45, MSE = .0009, p = .01. Collaborative recall in groups had not only benefits for subsequent individual recall, but also costs. We examined the contents of all errors made in any of the three sessions by particular triads and discovered that 44% of the errors in group products persisted; that is, group members carried over the errors from their group into their post-group individual products. This means that although collaborative recall in Session 2 led to improved individual recall in Session 3, it also led to some distortions. Another way to look at this finding is that 54% of the errors made by an individual in Session 3 involved information that had been left out of that individuals product in Session 1, but was introduced in incorrect form by the group. The rest of the errors in Session 3 either originated in that individuals own Session 1 (38%), or had actually been correct in that individuals pre-group recall but were distorted by the group (8%). Content filtering We returned to the transcribed interactions from the electronic chat windows and the recordings of face-to-face sessions, in order to compare these to the lists of recalled items that the groups produced (official group products). Recall that based on comparison with these transcripts and Session 1 products, propositions from the Session 2 products were coded as to whether they were unaltered, group filtered or self-filtered. Group-Filtering versus Self-Filtering. In the group sessions, there was evidence of both group-filtering and self-filtering (see Table 3). Keep in mind that an item was coded as group-filtered if it was presented by an individual during the text or spoken discussion from group session but failed to be included in the official group product. On the other hand, an item was coded as self-filtered if it was included in an individuals pre-group recall but was not at all presented during the group session (so as a coding category, self-filtering does not distinguish forgotten from recalled but withheld propositions). Overall, group filtering occurred marginally more often than self-filtering, F(1, 37) = 3.30, MSE = 73.01, p < .08. That there was so much group-filtering suggests that reduced productivity in collaborative recall (compared to nominal recall) is not due to cognitive interference alone, but is also affected by group interaction. Only 17% of all group-filtered propositions were explicitly filtered (as evident from verbal evidence in the transcripts in which one member explicitly questioned or debated what another member presented); the rest appear to have been simply ignored. Note that explicit filtering occurred equally often whether face-to-face groups were told to reach consensus or not, F(1,24) = .50, MSE = 4.92, n.s.. _________________________________ Insert Table 3 about here __________________________________ Filtering in Session 2 was not based on accuracy, nor on whether information was central to the story line. Of the group-filtered items, 77% were actually correct (and of these, 64% were central); the rest were incorrect items, inferred items, and meta-statements. Of the self-filtered items. 70% were correct (and 56% of these were central). Restoration. Propositions were coded as restored if they had been filtered out in the group session (by self or group) but reappeared in individuals products in Session 3. Of items filtered out in Session 2, 27% were restored by individuals in Session 3; these included 31% of all self-filtered items and 24% of all group-filtered items. With regard to accuracy, 77% of all restored items were correct, 14% were incorrect, and 9% were inferred. This pattern suggests that individuals were attuned to accuracy even when their groups made errors. Filtering and restoration in face-to-face versus electronic groups. Overall, there was no difference in total filtering between face-to-face and electronic groups (with group-filtering and self-filtering during Session 2 considered together). However, consistent with the grounding framework, group-filtering was higher in face-to-face than in electronic groups and self-filtering was higher electronically than face-to-face (see Figure 2). This interaction of medium and filtering was marginally significant, F(1, 37) = 3.53, MSE = 73.01, p < .07, driven by a much lower amount self-filtering face-to-face than in the electronic medium, F(1, 37) = 8.34, MSE = 31.29, p = .006. There were similar numbers of items group-filtered in face-to-face and electronic groups, F(1, 37) = .57, MSE = 75.24, p = .45, n.s.. There was no difference in items restored by face-to-face and electronic groups. _________________________________ Insert Figure 2 about here __________________________________ Discussion The main goals of this study were (1) to investigate the source of the productivity loss in collaborating groups as compared to nominal groups, as a function of group communication medium and (2) to discover the effects of recalling in groups upon subsequent solitary recall. We began by establishing that in our naturalistic recall task, groups recalled a higher proportion of correct, central ideas and a lower proportion of incorrect propositions and meta-statements than their individual members had in the pre-group recall session. As expected, spontaneously interacting groups underperformed nominal groups, both in the quantity and quality of recall products. These findings confirm our first two predictions that aimed to establish consistency with previous studies, most of which used list learning tasks and allowed little or no interaction among group members. Concerning the reason for recall underperformance in groups, our results fail to support a social loafing explanation (prediction #3), as electronic groups recalled just as many propositions as did face-to face groups. Moreover, the results demonstrate that in spontaneously interacting groups, there is a reduction in productivity in addition to the reduction that can be attributed to one members recall interfering with anothers (when a group member recalls a proposition but fails to report it because she is derailed by having to wait while hearing her partners discuss another proposition). The advantage of examining not only the official recall product, but also the transcript of the groups interaction while they did the task together (and by comparing the two records), is that we can tease apart the existence of distinct mechanisms for how content comes to be filtered out of a group product: there is significant filtering at both the individual and the group levels. There is no reason to suppose that there should be more cognitive interference in electronic groups than in face-to-face groups; if anything, there should be less, since group members are not distracted by seeing and hearing one another. So even though our self-filtering category cannot distinguish cognitive interference from the intentional withholding of propositions, the fact that there is so much more self-filtering in electronic than face-to-face groups supports the strategic allocation of collaborative effort predicted by the grounding framework (prediction #4). In particular, it was confirmed that group-filtering was more likely in the face-to-face medium, whereas self-filtering was more likely in the electronic medium. This finding shows that the distribution of initiative during interpersonal coordination is a relevant factor. When the currency of interaction is text, producing an utterance and completing a conversational exchange takes more time and effort than does speaking. And when partners are not co-present to one anothers intonation or facial expressions, their non-verbal cues are limited. Speakers use such cues to display their commitment to what they are saying, and addressees pick up on those cues (Brennan & Williams, 1995; Smith & Clark, 1993; Swerts & Krahmer, 2005). It should be less effortful for someone in the face-to-face medium to present everything she can recall, including propositions she is not confident about (accompanied by the intonation or facial expression to mark this lack of confidence) and let the group filter it out (either explicitly or via non-verbal cues) than it would be for someone in the electronic medium to offer a proposition and more painstakingly explain that it may be dubious. In the latter situation, she might well choose to withhold the proposition, as it takes so much effort to negotiate its acceptance into the group product. Similarly, groups interacting face-to-face have more mechanisms with which to indirectly reject each others proposals without risking insult, as we have shown elsewhere (Brennan & Ohaeri, 1999). When non-verbal cues are limited, our transcripts show that electronic groups still try to use such cues occasionally, even when they have to type them out: P3: and the grandmother told him to quit rambling superstitions.. P2: his wife scolded him for scaring the child P3: nod P1: does that sound ok? [Electronic Group 7] P3: or maybe it was just a nice way to close the story.. you know.. a bed time story.. the girl goes to sleep.. the grandmother lights the fire.. and then the fade out P2: whateva P3: shrug P3: the end [Electronic Group 7] The finding that collaborators distribute the responsibility for rejecting low-confidence or incorrect items among individuals themselves (predominating in the electronic medium) versus the other group members (predominating in the face-to-face medium) is consistent with the grounding framework. Grounding predicts that individuals flexibly shift their resources and distribute their effort in order to reach the performance criterion for a collective task, and that who does what depends on the affordances of the communication medium (Brennan & Lockridge, 2006; Clark & Brennan, 1991). With regard to group filtering, groups appeared to include only the propositions they believed were necessary for the task. Group products contained significantly higher proportions of correct central ideas from the story, whereas individual products contained significantly higher proportions of meta commentary. This is consistent with Stephenson, Kniveton, and Wagner (1991)s suggestion that while individuals give accounts of events in "characteristically different ways reflecting their idiosyncratic notions of what is appropriate to talk about, groups produce accounts which focus strictly on the action and events, but ignore situational comment on the setting and the motives and character of those they have observed" (p. 465). In other words, collaboration leads to the selection of propositions from the total pool of available individual propositions, based on the group's decision as to what is appropriate to talk about (Stephenson et al., 1991). The transcripts of the interactions showed that groups sometimes do set a task criterion. Criterion setting was explicitly negotiated in 14 of the 39 groups, and occurred about equally in face-to-face and electronic groups. In some cases the criterion was set explicitly by a partner saying, for instance, "let's get this done quickly and get out of here." In other cases criterion negotiation was done less directly, for example, by a partner asking "do we really need to go to that much detail?" or "do you think that x is important?" or "do you want to include x?" The answer was sometimes "yes" and sometimes "no." But even in cases where a criterion was not explicitly set, it is likely that group members judged what was appropriate to talk about based on the kind of information their partners were presenting. Recall that relatively few group-filtered propositions were filtered explicitly (17%), usually by partners saying that the proposed item was incorrect or that they did not remember it that way. Most other filtered items were simply ignored (or at least generated no verbal evidence of explicit filtering in the transcript). Concerning our second main goal (and final prediction): It was confirmed that recall by an individual was improved by the experience of having recalled in a collaborating group. Although groups (Session 2) recalled a higher quantity of propositions than did individuals in Session 1, individuals in Session 3 recalled just as many propositions as had their groups (when face-to-face and electronic groups were combined). Moreover, the experience of recalling in a group seemed to have, overall, a positive effect on the quality of recall products, with individuals in Session 1 recalling a higher proportion of incorrect items than in Session 2 groups, but the same proportion in those groups as subsequently in Session 3. Even so, groups did sometimes introduce distortions into post-group individual products; this was most likely to happen when a proposition had been entirely absent from an individuals pre-group recall product. The good news is that individuals who recalled a proposition correctly the first time rarely let the groups incorrect propositions distort their own post-group recall. The resemblance of post-group products to the group product is consistent with previous studies showing that groups have greater confidence in their collective memory, correct as well as incorrect, than individuals do in their own memories (Hinsz, 1990; Stephenson, N. Clark & Wade, 1986; Warnick & Sanders, 1980; Alper, Buckout, Chern, Harwood, & Slimovits, 1976). It has also been suggested that the act of communication seems to transform or reconstruct the cognitive representation of relevant information such that memory thereafter remains consistent with the transformed version (Loftus, 1975; Higgins, 1992; Zimmerman & Bauer, 1956). This effect of communication on memory is consistent with a cognitive dissonance interpretation (Festinger, 1957; Festinger & Carlsmith, 1959: Brehm & Cohen, 1962); in this case, participants who have publicly committed to the group product, by accepting it as an accurate representation of the story, are likely to adhere to it (see Kiesler, 1971). Concerning communication medium, recalling in face-to-face groups boosted post-group solitary recall in Session 3, whereas recalling electronically did not; this was the case even though the groups in both media conditions recalled just as many propositions during Session 2. Why should face-to-face spoken interaction be so much more beneficial to individual memory than electronic text? We suggest the following account: Before group members get together to recall collaboratively, their memories of the co-experienced event are relatively idiosyncratic; this is the reason nominal group products, by definition, turn out to include more propositions than do individual products (that is, if the individual members all independently recalled the same items, then their nominal product would be no higher than their pre-group individual products). What happens during collaboration by interacting groups is that this original diversity is reduced; while creating a collective product, individual partners can rehearse, elaborate, and improve their own memories for the event, and these memories become more similar. This should happen to a greater degree when the medium affords mechanisms that make it clear whether individual presentations have been taken up or ratified by the group. We propose that the non-verbal cues that are available face-to-face for grounding contributions to the collective product are particularly likely to support a boost in post-group recall by individuals. One limitation of the current study is that the measures made on the interaction transcripts fail to establish precisely and directly which amounts of self-filtering are due to being unable to retrieve a proposition, or to retrieving it and meaning to present it but failing to do so, or to retrieving it but intentionally withholding it (although it seems clear from the discussion above that all three of these are going on). Likewise, group-filtering may be due to intentional and explicit rejection of one persons recollection by the others, or to more tactful ignoring or implicitly rejecting a proposed memory, or simply to the failure of the scribe to record all the propositions proposed as part of the collective product (although the fact that all 3 partners monitor the emerging product may minimize this somewhat). All of these are ways in which coordination costs could tax an interacting group, resulting in under-performance compared to a nominal group. Conclusion Memory is not solely an intrapersonal cognitive process, but often, an interpersonal one. When a group of individuals witnesses an event, each encodes it somewhat differently; this may be a function not only of probabilistic influences on what any one person is likely to recall, but also of individual differences in knowledge, focus, biases, and perspective. When individuals recollect by grounding their idiosyncratic versions within a group, the memory is co-created and transformed to be part of their common ground. The resulting collective version is distinct from the original versions to the extent that it has been effectively filtered, grounded, and enhanced by the group, it represents a more certain product, and under some circumstances may improve subsequent recall. The act of coordinating interaction has the potential to influence both the quantity and the quality of collective memory. References Alper, A., Buckout, R., Chern, S., Harwood, R., & Slimovits, M. (1976). Eyewitness identification: Accuracy of individual vs. composite recollections of a crime. Bulletin of the Psychonomic Society, 8, 147-149. Anderson, M. C. Neely, J. H. (1996). Interference and inhibition in memory retrieval. InE. Bjork, B. Ligon, & A. Robert (Eds.), Memory (pp. 237-313). San Diego, CA: Academic Press. Bartlett, F. C. (1932). Remembering: An experimental and social study. Cambridge, UK: Cambridge University Press. Basden, D. R. and Basden, B. H. (1995). Some Tests of the Strategy Disruption Interpretation of Part-List Cuing Inhibition. Journal of Experimental Psychology: Learning, Memory, and Cognition, Vol. 21, No. 6, 1656-1669. Basden, B. H. , Basden, D. R., & Henry, S. (2000). Costs and benefits of collaborative remembering. Applied Cognitive Psychology, 14, pp. 497-507 Basden, B. H., Basden, D. R., Stephens, J. P. (2002). Part-set cuing of order information in recall tests. Journal of Memory and Language, Vol 47(4), pp. 517-529 Basden, B. H., Basden, D. R., Bryner, S., Thomas, R. L. (1997). A comparison of group and individual remembering: Does collaboration disrupt retrieval strategies? Journal of Experimental Psychology: Learning, Memory, and Cognition, Vol 23(5), pp. 1176-1189. Brehm, J. & Cohen, A. (1962). Exploration of cognitive dissonance. Wiley: New York. Brennan, S. E., & Lockridge, C. B. (2006). Computer-mediated communication: A cognitive science approach. In K. Brown (Ed.), ELL2, Encyclopedia of Language and Linguistics, 2nd Edition (pp. 775-780). Oxford, UK: Elsevier Ltd. Brennan, S. E., & Ohaeri, J. O. (1999). Why do electronic conversations seem less polite? The costs and benefits of hedging. Proceedings, International Joint Conference on Work Activities, Coordination, and Collaboration (WACC '99) (pp. 227-235). San Francisco, CA: ACM. Brennan, S. E., & Williams, M. (1995). The feeling of another's knowing: Prosody and filled pauses as cues to listeners about the metacognitive states of speakers. Journal of Memory and Language, 34, 383-398. Brown, P. & Levinson, S. C. (1978). Politeness: Some universals in language usage. Cambridge, UK: Cambridge University Press Clark, H. H. & Brennan, S. E. (1991). Grounding in communication. In L. B. Resnick, J. Levine, & S. D. Teasley (Eds.), Perspectives on socially shared cognition (pp. 127-149). Washington, DC: APA. Clark, H. H., & Wilkes-Gibbs, D. (1986). Referring as a collaborative process. Cognition, 22, 1-39. Clark, N. K., Stephenson, G. M., & Kniveton, B. H. (1990). Social remembering: Quantitative aspects of individual and collaborative remembering by police officers and students. British Journal of Psychology, 81, 73-94. Diehl, M., & Stroebe, W. (1991). Productivity loss in brainstorming groups: Tracking down the blocking effect. Journal of Personality and Social Psychology, 61, 392-403. Diener, E. (1979). Deindividuation, self-awareness, and disinhibition. Journal of Personality and Social Psychology, 37, 1160-1171 Edwards, D. & Middleton, D. (1986). Joint remembering: Constructing an account of shared experience through conversational discourse.  HYPERLINK "javascript:__doLinkPostBack('detail','ss%257E%257EJN%2520%252522Discourse%2520Processes%252522%257C%257Csl%257E%257Erl','');" \o "Discourse Processes" Discourse Processes, Vol 9(4), Oct-Dec 1986. pp. 423-459 Festinger, L. (1957). A theory of cognitive dissonance. California: Stanford University Press. Festinger, L., & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology, 58, 203-210 Finlay, F., Hitch, G. J., & Meudell, P. R. (2000). Mutual inhibition in collaborative recall: Evidence for a retrieval-based account. Journal of Experimental Psychology: Learning, Memory, & Cognition, 26, 1556-1567 Halbwachs, M. (1980). The collective memory. Trans. F. J. Ditter, Jr., & V. Y. Ditter. New York: Harper & Row. Hartwick, J., Sheppard, B. H., & Davis, J. H. (1982). Group remembering: Research and implications. In R. A. Guzzo, (Ed.), Improving group decision making in organizations (pp. 41-72). London:Academic Press. Higgins, E. T. (1992). Achieving 'shared reality' in the communication game: A social action that creates meaning. Journal of Language and Social Psychology, 11, 107-131. Hinsz, V. B. (1990). Cognitive and consensus processes in group recognition memory performance. Journal of Personality and Social Psychology, 59, 705-718. Hollin, C. R., & Clifford, B. R. (1983). Eyewitness testimony: The effects of discussion on recall accuracy and agreement. Journal of Applied Social Psychology, 13, 234-244. Hollingshead, A. B., (1998). Retrieval processes in transactive memory systems. Journal of Personality and Social Psychology 74 (1998b), 659671. Karau, S. J., & Williams, K. D. (1993). Social Loafing: A Meta-Analytic Review and Theoretical Integration. Journal of Personality and Social Psychology, 65, 681-703. Karsenty L (1999). Cooperative work and shared context: An empirical study of comprehension problems in side-by-side and remote help dialogues. Human-Computer Interaction, 14, 283315. Kiesler, C. (1971). The psychology of commitment. London and New York: Academic Press. Kiesler S., Siegel, J., & McGuire, T. W. (1984). Social psychological aspects of computer-mediated communication. American Psychologist, 39, 1123-1134 Lindman, H. R. (1974).  HYPERLINK "http://web.ebscohost.com/ehost/viewarticle?data=dGJyMPPp44rp2%2fdV0%2bnjisfk5Ie46bZRs6m2S7Wk63nn5KyI8e%2fhgL6prUuupbBIrq2eTLirtVKypp5oy5zyit%2fk8Xnh6ueH7N%2fiVauntki0prRPtqi3PurX7H%2b72%2bw%2b4ti7iPHv5j7y1%2bVVv8Skeeyzr1G1q6tItKq3Taumrkik3O2K69fyVeTr6oTS2%2faM&hid=13" \o "Analysis of variance in complex experimental designs." Analysis of variance in complex experimental designs. Oxford, England: W. H. Freeman & Co. Loftus, E. F. (1975). Leading questions and the eyewitness report. Cognitive Psychology, 7, 56-572 Lorge, I. & Solomon, H. (1955). Two models of group behavior in the solution of eureka-type problems. Psychometrika, 20, 139-148. Morrissett, J. O., Crannell, C. W., & Switzer, S. A. (1964). Group performance under various conditions of work load and information redundancy. Journal of General Psychology, 71, 337-347. Ohaeri, J. O. (1998). Group processes and the collaborative remembering of stories. Unpublished doctoral dissertation, Stony Brook University. Smith, V. L., & Clark, H. H. (1993). On the course of answering questions. Journal of Memory and Language, 32, 25-38. Steiner, I. D. (1972). Group process and productivity. New York: Academic Press. Stephenson, G. M., Brandstatter, H., & Wagner, W. (1983). An experimental study of social performance and delay on the validity of story recall. European Journal of Social Psychology, 13, 175-191. Stephenson, G. M., Clark, N. K., & Wade, G. S. (1986). Meetings make evidence? An experimental study of collaborative and individual recall of simulated police interrogation. Journal of Personality and Social Psychology, 50, 1113-1122. Stephenson, G. M., Kniveton, B. H., & Wagner, W. (1991). Social influences on remembering: Intellectual, interpersonal and intergroup components. European Journal of Social Psychology, 21, 463-475. Swerts, M. & Krahmer, E. (2005). Audiovisual prosody and feeling of knowing. Journal of Memory and Language, Vol 53(1), pp. 81-94. Vollrath, D. A, Sheppard, B. H, Hinsz, VB,& Davis, J. H. (1989). Memory performance by decision-making groups and individuals. Organizational Behavior and Human Decision Processes, 43, 289-300. Warnick, D. H., & Sanders, G. S. (1980). The effects of discussion on eyewitness accuracy. Journal of Applied Social Psychology, 10, 249-259. Wegner, D. M. (1987). Transactive memory: A contemporary analysis of the group mind. In B. Mullen & G. R. Goethals (Eds.), Theories of group behavior (pp. 185-208). New York: Springer-Verlag. Wegner, D. M., Erber, R., & Raymond, P. (1991). Transactive memory in close relationships. Journal of Personality and Social Psychology, 61, 923-929. Weldon, M. S., & Bellinger, K. D. (1997). Collective memory: Collaborative and individual processes in remembering. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 1160-1175. Weldon, M. S., Blair, C., & Huebsch, P. D. (2000). Group Remembering: Does Social Loafing Underlie Collaborative Inhibition? Journal of Experimental Psychology: Learning, Memory, and Cognition, 26(6), 1568-1577 Williams, E. (1977). Experimental comparisons of face-to-face and mediated communication: A review. Psychological Bulletin, 8, 963-976 Wheeler, M. A., & Roediger, H. L. (1992). Disparate effects of repeated testing: Reconciling Ballard's (1913) and Bartlett's (1932) results. Psychological Science, 3, 240-245. Whittaker S. (2002). Theories and methods in mediated communication. In Graesser, A., Gernsbacher, M., & Goldman, S. (Eds.), The Handbook of Discourse Processes (pp. 243-286). Hillsdale, NJ: Erlbaum. Wright, D. B. & Klumpp, A. (2004). Collaborative inhibition is due to the product, not the process, of recalling in groups. Psychonomic Bulletin & Review, 11(6), 1080-1083 Zimmerman, C. & Bauer, R. A. (1956). The effect of an audience upon what is remembered. Public Opinion Quarterly, 20, 238-248. Table 1 Individual, Collaborative Group, Nominal Group, and Post-Group Recall. _________________________________________________________________________ Recalled Propositions Included in Product _________________________________________________________________________ Condition Total Correct Incorrect Inferred Metaa _________________________________________________________________________ Pre-group Individual Products M 64.00 55.73 (.87)b 4.58 (.07) 2.47(.04) 1.17(.02) SD 14.79 13.48 1.84 1.15 1.43 Group Products M 76.90 69.26 (.90) 4.23 (.06) 2.87(.04) .79(.01) SD 20.94 19.53 2.51 1.72 1.06 Nominal Group Products M 124.05 106.08 (.86) 8.95 (.07) 5.15(.04) 2.87(.02) SD 26.51 24.90 4.38 2.84 4.11 Post-Group Individual Products M 77.36 68.49(.88) 4.54(.06) 3.24(.04) 1.03(.01) SD 19.73 18.74 2.07 1.27 1.04 _______________________________________________________________________ a Meta-statements b Values in parentheses represent proportions of total product. Table 2 Individual and Collaborative Recall for Face-to-Face and Electronic Groups _________________________________________________________________________ Recalled Propositions Included in Product ____________________________________________________________ Session aTotal Correct Incorrect Inferred Metab _________________________________________________________________________ FTF Groups N = 26 groups Pre-group Individual Products M 62.99 55.72(.88)c 3.91(.06) 2.40(.04) .94(.02) SD 16.65 15.00 1.57 1.11 .86 Group Products M 77.88 70.62(.90) 3.81(.05) 2.88(.04) .96(.01) SD 20.89 20.29 2.14 1.77 1.22 Post-group Individual Products M 82.51 73.60(.89) 4.46(.06) 3.35(.04) 1.10(.01) SD 20.84 19.79 1.79 1.33 1.01 Electronic Groups N = 13 groups Pre-group Individual Products M 66.03 55.74(.84) 5.92(.09) 2.62(.04) 1.64(.03) SD 10.41 10.31 1.61 1.27 2.15 Group Products M 74.92 66.54(.89) 5.08(.06) 2.85(.04) .46(.01) SD 21.75 18.40 3.04 1.68 .52 Post-group Individual Products M 67.05 58.26(.87) 4.69(.07) 3.03(.05) .87(.01) SD 12.46 11.20 2.62 1.14 1.13 __________________________________________________________ aTotal propositions in recall product (correct + incorrect + inferred + meta-statements) b Meta-statements cValues in parentheses represent proportions of total product. Table 3 Content-filtering in Face-to-Face and Electronic Groups ________________________________________________________________________ Recalled Propositions Included in Product ___________________________________________________________ Filtering aTotal Correct Incorrect Inferred Metab ________________________________________________________________________ Group-filtered Items (Session 2) FTF Groups M 16.77 13.04(.76)c 2.42(.15) 1.31(.09) 0.00 SD 7.62 6.56 1.90 1.23 0.00 Electronic Groups M 14.54 10.38(.77) 2.38(.13) 1.77(.10) 0.00 SD 10.54 7.32 2.63 1.88 0.00 Weighted Mean of Both Groups M 15.65 11.71(.77) 2.41(.14) 1.54(.09) 0.00 SD 9.08 6.94 2.26 1.55 0.00 Self-filtered Items (Session 2) FTF Groups M 9.18 6.72(.71) 1.49(.20) .68(.07) .31(.04) SD 5.18 3.85 1.05 .71 .52 Electronic Groups M 14.67 9.97(.67) 3.00(.22) .77(.06) .85(.05) SD 6.36 5.25 1.36 .50 1.87 Both Groups M 11.92 8.34(.69) 2.24(.21) .69(.06) .58(.04) SD 5.77 4.55 1.20 .60 1.19 Restored Items (Session 3) FTF Groups M 3.62 2.83(.75) .38(.15) .37(.09) 0.00 SD 2.82 2.36 .41 .58 0.00 Electronic Groups M 3.59 2.90(.81) .28(.12) .41(.07) 0.00 SD 2.90 2.35 .30 .67 0.00 __________________________________________________________________________ aTotal propositions in recall product (correct + incorrect + inferences + meta-statements). b Meta-statements cValues in parenthesis represent the proportion of the total product. Figure Captions: Figure 1. Mean number of propositions in pre-group, group, and post-group products of participants in face-to-face and electronic media. Figure 2. Mean group-filtered and self-filtered propositions in face-to-face and electronic groups.    Note that repeated measures comparisons of pre- and post-group individual products yielded the same results whether individual recall was compared for the 117 individual participants or collapsed into 39 3-person means, so for consistency we report the latter.     PAGE  PAGE 1 Ekeocha, J. O., & Brennan, S. E. (in press). Memory. Copyright 2007 Justina O. Ekeocha - ALL RIGHTS RESERVED DO NOT CITE WITHOUT PERMISSION 9:abcuv| $ .    $ F S W [ d f }  3 > K L ^ o  0 Z r .h xZhP6hahP6hPhEjk_HaJhP_HaJh xZhP_HaJ hP0J<hPh xZhP]h xZhP0Jjh xZhPUjh xZhPU h xZhP8=>?Vrv!12wgdP0d]0 $da$gdP0@ d]0^@ `gdP$0d@&]0a$gdP!"#M$0Eƀ0]0a$gdP0]0gdP 0d]0gdP 0d@&]0gdP0d]0gdPF G H I J K M ] ^ wii d1$7$8$H$gdP 0d]0gdP$0d]0a$gdPIdEƀi&gdP# 5p@  P !$`'0*dgdP -N3:M 0d]0gdPdhx1$7$8$H$^gdPdh1$7$8$H$^gdPdh1$7$8$H$`gdP0d]0`gdP $da$gdPJ Q ###5$C$$o%%%%&'\)f)***v++,V.[.|.}...(/0/i/p/000088;);mCCFFsIJJJ-L.LMMO PhPB*ph333h xZhP_HaJh xZhPh]]zhAVhPi&*6 hahPh xZhPaJh xZhP6hP h{hP hP0J<h xZhPha h xZhPh xZhP_HaJ7t$S)|..35:@HLSV0dEƀG]0`gdPoG0d]0`gdP P$PDPEPFPPQDQFQNQWQURVRRRRSS[[[[\]y`{`LdfdhOR[\Z[ 4Uxyŝ$8+,<h xZhP5h xZhPh5h xZhPh]]zh xZhP6 h xZhPhPR͙Ι$4Wèz:0d]0`gdP d`gdP dgdP$a$gdP d`gdPdgdP<?IJ]^"ʤˤۤޤӥԥWh¦æӦ֦Ƨǧקڧ¨è@C89:a!+679Fai8I\h xZhPhah xZhPB*phh xZhPhJj_h xZhP>*hP h xZhPh xZhP6N9\ۺ;<Gv$da$gdPodgdPI$Eƀ`a$gdP$a$gdP$da$gdPoed&d d`gdP ۹ܹ^_nq{|im *.!% +,jh xZhP0JUh xZhP5h xZhP5>*hAVhP6h xZhPhAVhAVhPj&*6h xZhPhgHhAVhPpj&*6h xZhP6 h xZhPhP:r# ]| dgdP^gdP ^gdP d`gdPYdC$Eƀmj&`gdPoUd&]#qr}0]0gdP$0d]0a$gdP$da$gdPoed&dd`gdPoj&d&x d`gdPdgdP,qr|}qy >@ip>µµwb)hahPOJQJP*OJQJ5hahPCJ OJQJaJ P*CJOJQJ hahPh xZhP0J6CJaJh xZhP0JCJ]aJh xZhP0JCJaJh xZhP0J5]h xZhP6h xZhPhAVh xZhPhgHh xZhPB*phhP h xZhP&Ww')/8>Q`i+Kn<S#EQFoufh xZhPB*CJaJph333h[hPB*ph333hP0JCJaJh xZhP0J5]h xZhP0J6CJaJh xZhP6 h xZhPh xZhP0JCJaJhahP0JCJaJ hahP)hahPOJQJP*OJQJ,hahP6OJQJP*OJQJ'(),-gh|}PQ 0@&]0gdP0]0gdP 7$8$H$gdPGw,12=23FGQ%R](45K` P9aln̼h xZhPB*aJ phh xZhP6aJ h xZhPaJ hlShP6h3hP6h xZhP0J6>*B*phjh xZhP6Uh xZhP0J6\h>hP6h xZhP6 h xZhPhP2mn\]45uv!"lmgdP0]0gdP 0@&]0gdPdgdPo"h&d&d9:XYZdy/HSU\cdlmc{j  ٹ٘{٬shEjkhP6h xZhP0JB*phh xZhP0J6>*B*phjh xZhPUh xZhP0Jh xZhP0JCJaJh xZhP0J5]h>hP6hPh xZhP6 h xZhPh xZhP6_HaJh xZhP_Hh xZhP_HaJ-deTU XY$%  X0]0gdPgdP0]0gdP 0@&]0gdP  $ %   v w 7 8     kldgdPo!h&d&d 7$8$H$gdP0]0gdPgdP 0@&]0gdP       #     C k w       C f g  _j"Elsčulh xZhPCJh xZhP6B*ph333h xZhPB*ph333h>hP6 hP6h xZhP_HaJh xZhP6_HaJh xZhP_HaJh xZhP6hPh xZhP6]h xZhP0J6CJaJh xZhP0JCJaJh xZhP0J5] h xZhP*mn: 0d]0gdP dgdP 0@&]0gdP0]0gdPJ0Eƀ)]0gdPQRfg23jmLMNO_`a.FGefyz?@qt̶쪜h xZhP5CJEHaJh xZhP5CJaJhPh xZhPEH h xZhPEH h xZhPh xZhPCJEHaJh xZhP>*CJaJh xZhPCJaJh xZhP>*=:i3Q2kLMN` 0]0^gdP0]0gdPBq.Ge ?r*+ 0]0^gdP $0]0a$gdP0]0gdP :;kn?AB3f.1op"UVǿǿh xZhPCJEHaJh xZhP5CJaJh xZhPCJEHaJ hPCJh xZhP>* h xZhPh xZhPEHh xZhPEHh xZhPCJh xZhP>*CJaJh xZhPCJaJ:+:l?@A3|f 0]0^gdP0]0gdP/QRo ?@U#1 0]0^gdP0]0gdP#12\_% & o }          y!!!!!!!!!!!!!""""""""#ӽӽjhPUjhP0JUj h xZhPU hvhPjh xZhPUhPh xZhP>* h xZhP hPEHh xZhPEHh xZhPEHh xZhP5CJaJh xZhPCJaJh xZhP>*CJaJ21]|}%       x!y!!!!!!!!!!!gdPdgdP0]0gdP 0]0^gdP!!"""""""""# # # ###M######$a$$a$gdP$h*@&]ha$h]h $&`#$a$0]0gdP#### # # #######D#E#L#M#W#X#^######ȷ h xZhP hPhPp*6CJhPhPCJhPhPp*CJhP0JmHnHuhP hP0JjhP0JU###0]0gdP&1h:pP/ =!"#$%DyK ekeochaj@wpunj.eduyK Lmailto:ekeochaj@wpunj.eduyX;H,]ą'c6 Dd :0  # A" W$Y~ +@= W$Y~h$'Td8v*xZ]h\E>wMMMbbnQRjQ&j5ؘXbX/)I!(>XZ$AЀ!Z|R93ɝn ̙{=@PW{`;tTLq*nK%l|3,ŷp<=/G8VPJPv,fʸ.:fffy oĐEwz%:1#@ކ.vxheU?gP =nl~axC<}ǔXf?i4 k~~Jk9ׇ'u/`GWҕ!Vzt<^ZQ6Hc|j@~[Ъ* Kc(=>:k^šBƪg3(b39ߑx!32&''WM!e=&\]okxNmFAنM3lH*ڧC[{Ru{R],PM>I p+7ҟu\;=+8>% O|Lc'Nw <)k^Fl7f?qcg>!N+~H]AO`GkajEǧ0kZ9z+XST˵n쥳Jʿ܇{1GI@UQf 3$xr OfmF=]Ԇ2dՃ9CJe>Ԏ2o`μ@G{i=t!>3œsJe>?9%IZcL_:Ÿ^4or s~>% O|Lc'Nw <)kBĭ5S&Ѿ~Qz(己Fesw >fs}$X^h.D9 NrpBD5%}eޤwy?umf3!OXx{^={0w}:Ł[żu!UOk=7Dl$[&3Ì n#[ݻw?#s,k,l<8 G=<Rn'>N7_轲o-|C> |Kцwؖ6-L˲x&rl]t1Ǵ>y`xŖ+1κp|NxKaw|c%8+Z[ r?! sOاFsi _|O泷ͧh"˹x|g3pF_L;/NΗ9|, ax)w.@u =gW}sF*+?u=ּ؛7YۧoI(׶Q/\K"ߵYXȺvsbп6Ki3]Čk)?+fpCo #fHBF'Td8v*)xY=LA~Fm 9 s3F,A$hbD1FhPL;4ƟB+ i(ւXRmqvqyޛ[[.W=- TY;zGy742<@tjh Gmǭ:n7PߑK2-bw /| M璕{UljT/R3ݣQI?c%(Rj/Q[m-ݨ66<ݨ66,ݨ66ݨ6G?D7jM ?D7jMe/эaj;)эZWS[E4>{R ե?Q)'MeɍI5&<O4{N5؎fҰ-N/qc羋YfF֘vV6",Nhu]eQmnSc_W&3>uey} YmnS 1 ˫~^U~mnSVcץOe߾q]>/{:uYWX^uQ͌#@mY u).ū+,/Y t5ʃhZS@7Z .ЍV&7J*'Uv)F7cPvp]jPgov&|mN~o:qwMjL~M;@==-EGouՈ.I[T:N"#GFt&rE~Ñ#HaD}ѭ>1o5mV炱`xϫSzumTWW{oBݵzo[nY[ɴ~{@úߑl/>bD&~zZ@eL+~k5klo%H`&f3t}2Mo[{&Zo>>O`_h"[x?&l=.U=!'U'r Dd 0  # A" _k"m ISʷ a @= _k"m ISʷ!\G(xZ]h\E>sSğvIXl[B@0B# ɃHĦ5CԫB+ Hŀ%A9wNN޽l6{pg͙9sܙ{cK܏'^m8v> \MH!.?]c(PaE5,LPQnKʅ{(>Ǎtg\?4yMA>} ÜfpQ9`P7k+K106 Ӹh36Gep_ƈ߬8mQ!}U_`L+Æl>2 0 /~⤯>8Fr7M@y?>=\r ?=t0 k2z~V_WS~gy}.D7oI~Nn||&C~wS[*bﶂ>VyӵYnUʨ1f:8^qS^A&Җxni ڍvu[|x<1%2,T$Lg|T< OOZcw0Y#oT5X>'iO <->,~ <)OTf1,&A Ykw 'cG{o/u.RuѶI Z_e]dUXSgfTغq eUznF%}O48k{5U |Q>-/OX֝#^L|=d|W"p&+Sm5r8Y6~wDccyes9) y#ne,8y9?p^mpDy9sp&JqcL5l{˼Hgmemdf oc$? h x^tۻqSJC9C%'mMs\G(xXKTQ>W41E"u ,SwBD֦VE-DHFvq[Bp ZBhw3* }p}&=U*DL\{q:4z0щN3vۼ?`>.2-,5Eܕb Zs{cjTC5f 1(CAܶ|Ѥաih(Vn:u5\Zj=kWKԤ2%qsjmE95V@^7&KԤ#%qsj-D=/яSn7)я[Y5v3tXvHͼ|^Bx0os|)}cxgͩI(^ssUbO3T_W4q?Opqq9u߳6EjNjZc_qh(~1vF՚گx\==<^7 Jƕ֯5cF\\3.(K_1ki-%%d.>6g^d$~P:4(Џzk|k% m> 㵜Z6ZLxͧ@~ < (ЏW>5m> 5 ?~„Wm} $ʧv)Lm#..ᖵq*Kžvխ^F\\ J6),z]ZԕרvvsJ/:6Lg-:#5nW_S.1a?X(iZhZv}<-OnNOfwju\2 `N(NsMaeUR،Ru,{'F<-~wi䆹O"i@V۬ydj3ϙ>*{D3=WM_!ZĻQH0Kۜ%ɤzAUͩ'@> knComment ReferenceCJ44 kn Comment Text:j: knComment SubjectD@"D kn Balloon TextCJOJQJaJ@@2@ {AHeader  ! OJQJaJ.)@A. {A Page Number@ @R@ {AFooter  ! OJQJaJlC@bl Jj_Body Text Indentd7$8$^`CJOJQJ_HaJ.X@q. ~fEmphasis6]4U@4 ~f Hyperlink >*ph8O8 ~f medium-font1CJaJ*W@* ~fStrong5\6@6 5 Footnote Text@&@@ 5Footnote ReferenceH*߈ )      !"#$%&'()   !"#$%&'( ~) z z z z z z z z z z z z z z z z z z z z z z z z z z z z z z z z !z "z #z $z %z &z 'z (z )zL !&/8AJRR[clu~Jj49   I  C V xLx   !"#$%&'(=>?Vrv!12wFGHIJKM]^   - N 3 : M tS#|(.-/4:BFMNNoPoQSSSWWW\]] ^xcyccdf(lmooopuuxy{܁%ۑ5ՖU(Jekիj .Ilmx 9T "=>T8DEFG#$YZ]^*+89efRS23PQ78UV UVhi$%K L          ! k    d      ( K L c      5 } ~       ()s_x-PQp8[\k45pqr !*,dN2`!Qpq34Tb V  $%'(*+-.01:;<GH~000000>0>0>0>0>0>0>0>0>0>0>0>0>0>0>0>0>0>0>00000000000000000000000ؤ000000000000000000000000000000 0 0 0 0 00000S00W0W0W00^0^00zc0zc0zc0zc0zc0000o00u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u000000000000000:0:0:0:0:0:000g000000T0T0000300000000000000000000000W00000000000000j00000000000000M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0M 0ɀ00P0ʀ0R0ʀ0R0ʀ0R0ʀ0R0P0P0P0@@0@0ɀ0ɀ0ɀ0P0ɀ0P0P=>?Vrv!12wFGHIJKM]^   - N 3 : M tS#|(.-/4:BFMNNoPoQSSSWWW\]] ^xcyccdf(lmooopuuxy{܁%ۑ5ՖU(Jekիj .Ilmx 9T "=>T8DEFG#$YZ]^*+89efRS23PQ78UV UVhi$%K L          ! k    d      ( K L c      5 } ~       ()s_x-PQp8[\k45pqr !*,dN2`!Qpq34Tb V  $%'(*+-.0:;<GH~00000000000000000000000000000000000000000000000000000000000000000000000000000 0 0 0 0 000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000