By Ellen Isaacs and John C. Tang

Published in 1994 in the Springer-Verlag journal Multimedia Systems, Volume 2, pp. 63-73. A PDF version is available of a slighty earlier version presented at the Multimedia '93 conference in Los Angeles, CA.

Abstract

Keywords: Remote collaboration, desktop video conferencing, computer-supported cooperative work, user interfaces, conversation.

Introduction

There has been particular interest in the use of video to enhance remote collaboration, which has traditionally been supported by voice-only (phone) or text-only (e-mail) interactions. Although video is often intuitively presumed to greatly improve the quality of interactions among remote participants, many studies have found no evidence that groups are more effective or efficient at solving problems or making decisions when they are connected through a video and audio link than when they use only an audio link [Chapanis et al., 1972; Gale, 1990; Ochsman & Chapanis, 1974; Short et al., 1976; Williams, 1977].

Did these previous studies somehow miss finding the effect of video, or are our intuitions about the value of video misleading? By re-examining some of the assumptions and conditions of the previous studies, we identified three reasons why they might not have detected any effect of video in support of interaction.

The previous studies measured the product (e.g. decisions, quality of solutions, completion times) of short, problem-solving interactions. The effects of video are more likely to be visible when studying the process of interactions. For example, video is likely to be useful for managing the mechanics of conversations, e.g. turn taking, monitoring understanding, noting and adjusting to reactions [Clark & Schaefer, 1987; Clark & Wilkes-Gibbs, 1986; Isaacs & Clark, 1987; Sacks et al., 1974; Williams, 1977]. If video is effective at enhancing the process of interaction, people may perceive their interactions to be more satisfying, and it may encourage coworkers to collaborate more frequently. If the process is important to collaboration, then the mechanics of interaction must be facilitated in the user interface so that users may take advantage of their rich set of existing skills in a natural and intuitive way.

Secondly, the effects of video in supporting interaction may be most visible over the long term, and may be too subtle or difficult to capture in short, laboratory experiments. As Gale [1990] notes:

The structure of groups is continually changing. The effects of technology on a group may take weeks, months, or even years before becoming apparent. These sort of effects cannot be fully explored in a one hour experiment (p. 187).

We would expect that richer interactions would lead to more productive and/or higher quality results in the long term, although more research would be needed to demonstrate this hypothesis.

Thirdly, most of the previous studies were among strangers who were asked to accomplish an artificial task for the purposes of the study. That is, the participants did not have working relationships with each other, and were not dealing with an issue that highly motivated them. Yet the interactional cues that video transmits are likely to play a more important role among people who know each other and are accomplishing real work that requires complex social negotiation. Again, Gale [1990] notes:

The results from this study suggest that by adding audio and video to the communications medium we allow groups to perform more `social' activities. ... A possible reason for the lack of difference in the quality of the output of the groups is that the tasks used in this study were not sensitive to social factors (p. 186).

Thus, previous studies may have missed the effects of video because they are too subtle to see among strangers carrying out impersonal tasks.

In this context, we believe there is still good reason to pursue video as an integral part of collaborative technology. To study the user interface implications of using video for remote collaboration, we observed a team of engineers who were using a desktop video conferencing (DVC) prototype. The prototype enabled digital audio-video connections between workstation desktops. Rather than conducting a broad survey of users' reports of their perceptions in using this technology, we focused on studying the details of one group's behavior when using video and audio as compared with audio-only and face-to-face interactions. Our intention is to describe the evidence we found for the potential benefit of video in remote conversations compared with audio alone, and to point out how video interactions can fall short of, and in some ways offer advantages over, face-to-face interactions. We then discuss how our results may be applied to the design of effective video conferencing systems.

Method

The DVC prototype provided a simple interface for requesting a desktop conference that followed a telephone model for making connections. The interface enabled a user to request a conference with up to two other people, and those people had to accept that conference request before audio-video connections were made. The interface allowed users to specify what kind of connections they wanted for a conference: audio, video, and/or a shared drawing tool called Show Me. Show Me allowed users to share an image of anything they could display on their screens. They could draw on top of shared images or construct a joint drawing from scratch. Within Show Me, users could type or draw at the same time, they could erase anyone else's work, and they could always see where everyone else was pointing with their cursors.

The default setting was to request audio, video, and Show Me connections with one other participant. Once a conference request was accepted, several windows popped up on each participant's workstation screen. Each person saw a video window showing the image of the other participant, plus a smaller preview window of the video image being sent to the other person. A full duplex audio connection was made with the other person. A Show Me window also popped up on each screen. (In a three-way connetion, each person would see an additional video window of the second remote participant.)

The DVC prototype ran on Sun SPARCstation computer workstations with a prototype add-on board that enabled real-time video capture, compression, and display. For audio, the prototype used the 8 KHz mlaw encoding that is built into the workstation. At the time of this study, the prototype video board used the Intel RTV 1.5 video compression algorithm. The video windows had a video resolution of 120 x 128 pixels, although that resolution could be scaled to any arbitrary size window. The default video frame rate was five frames per second, due to some long-distance network bandwidth limitations, but the users could request a different video frame rate before starting a conference (although they did so only once). The quality of the video image was less than that of broadcast television and certain analogue desktop video conferencing systems such as Cruiser [Root, 1988] (due to the lower refresh rate and lower effective resolution). But it was of higher quality than video conference systems than run over ISDN or conventional phone lines. More details on the technical description of the prototype may be found in [Pearl, 1992].

The data from this study was drawn from a larger study that observed the team's work activity under three conditions: before installing the DVC prototype, to understand their baseline collaborative activity; with the DVC prototype fully installed, including audio, video, and the shared drawing tool; and with the video channel subtracted from the DVC prototype, leaving audio and the shared drawing capabilities still in place. (See Tang and Isaacs, 1993 for a further description of the results from the larger study.) Since the team that was observed was not involved in the development of the DVC prototype, they had never experienced using the prototype before it was installed during the second condition of the study. Each team member was given a short (less than 15 minutes) demonstration of the prototype to familiarize them with the prototype's capabilities and user interface.

Although we took many measures of their work activity, the data for this paper are based on videotapes of six interactions in three different modalities: two desktop video conferences, two face-to-face interactions and two telephone conferences. Comparing interactions among the same people using different tools enabled us to better isolate the effects of the tools on their interactions. One of the desktop video conference meetings included all five group members (call them Kate, Jeff, Jack, Dave and Craig) and one was between just Kate and Jeff. Likewise, the two face-to-face meetings included the same sets of participants. We could not obtain phone conference data among the same sets of people, so instead we studied a four-way call between Kate, Jeff, Jack and Dave, and a three-way call between Jeff, Craig and Dave. Table 1 shows the people in each interaction we observed.

DVC	Meeting	Phone
Kate, Jeff	Kate, Jeff	Jeff, Craig, Dave
Everyone	Everyone	Kate, Jeff, Jack, Dave
Table 1. People involved in each observed interaction.

The five-person DVC was actually a three-way connection where two people crowded around one camera and workstation at each of two sites. The four-person phone conference actually connected three sites; two people were in the same office sharing a speaker phone.

The videotapes of the six interactions were analyzed in the tradition of interaction analysis [Tatar, 1989; Tang, 1991a] to look for any changes in pattern among the three modalities. The qualitative analysis involved creating a detailed account of all the interactional behaviors among the group members, with a particular emphasis on behaviors that took advantage of audio and visual cues. The quantitative analysis involved comparing the mechanics of conversational turn-taking (i.e., duration of turn, frequency of turn changes) between the face-to-face and DVC modalities.

Benefits of Video over Audio

Expressing Understanding

For instance, during the two-way DVC, Kate explains a technical issue. At first, Jeff tilts his head and looks puzzled, but eventually he gives a slight head nod as he grasps the concept. Then he sighs and shakes his head, acknowledging the issue as difficult. All these subtle reactions give Kate a running commentary on the state of Jeff's understanding. Later, Kate asks him to confirm his understanding of an idea and he says "Uh huh," but then he looks down and purses his lips as he considers the issue. Kate proceeds to elaborate, apparently responding to the visual rather than the auditory feedback.

In contrast, during the phone conferences, speakers often explicitly asked for confirmation. In one instance, Dave says, "...we should probably take, like, the first part of the meeting and just go through and see what questions you guys have." After a three second pause, he says, "Okay? Then you can at least get your questions answered. [1 sec. pause]. And then we can hit you up for stuff that we want to know. [1 sec.]. Okay? [1 sec.] All right?" Finally, Jeff says "Yep" and continues. With no visual feedback, Dave had to explicitly request a response four times before getting one.

In DVCs, the video provided an effortless and ongoing feedback channel that gave the participants a fluid sense of each other's understanding throughout the conversation. Addressees could visually give feedback on the level of their understanding without interrupting the speaker's talk. Without the video, the participants had to work harder to get much less information about each other's understanding.

Forecasting responses

Later, Kate starts to nod in response to Jeff's comment but then stops abruptly, indicating she thought she agreed but now is not sure. When she gives no indication of agreement at the end of his utterance, he prompts her with "Right?" He seems to ask for explicit feedback because she stopped nodding in the middle of his utterance. Forecasting negative responses was just one way that participants seemed to use the visual channel to express and handle disagreement. Others will be discussed in the examples below.

Obviously it is impossible to use head gestures and expressions to forecast responses on the telephone. As a result, participants are unable to read each others' gestures and adjust their utterances in mid-course. Of course, addressees may recognize that their reactions are not being forecast and therefore explicitly express their reactions verbally. But doing so requires more effort, and thus people may be more inclined to let subtle problems pass. In particular, participants may prefer not to verbally express disagreement that might have been reflected on their faces. The speaker may therefore be unaware of a potential problem and cannot take steps to work out the disagreement.

Enhancing Verbal Descriptions with Gestures

In many cases, the gestures appeared to be made unconsciously, sometimes outside the view of the camera or when the other person was not looking. Many people gesture while talking on the phone, apparently because it helps them express themselves verbally. As a result, when people cannot see each other (as when on the phone), they may not express verbally the subtleties conveyed through their inadvertent gestures.

Conveying Purely Non-verbal Information

In another example of using visual information, Jeff notices that another person, Ted, is walking behind Craig and Dave as they are discussing a technical matter. Ted happens to be knowledgeable about the matter, so Jeff suggests asking him to join the conversation, which he does. Clearly, it would be impossible for a phone conference participant to draw someone at a remote site into the conversation; only the person on that end could do so.

Participants could not convey information purely non-verbally over the phone. One interesting incident occurred in the four-way phone conference. During this call, two of the participants, Jack and Kate, were in the same office sharing a speaker phone, and so they could see each other. At one point, Jeff asks Jack, "I forget, how big of a pain it is to add new built-ins, Jack?" After a three second pause Kate observes, "He doesn't look too happy," and Dave bursts out with a laugh. Had Kate not been able to see Jack, the pause would have indicated only that he was considering the answer; Jack's spontaneous unhappy expression would not have been communicated.

Expressing Attitudes In Posture and Facial Expression

Later in this conference, Jeff tells the group that he has written a software utility they can use. They express interest, but then Craig teases Jeff, "As usual, no documentation." Jeff smiles and says, "It's not even done yet!" Craig throws his head back while smiling broadly, as shown in Figure 2. Jeff's words could be construed as defensive, but the smiles and Craig's response make it clear to everyone that the conversation is in fun. In contrast, in the three-way phone conversation, Jeff teases Craig about how his wife will react when she learns he is planning to spend three straight weeks on business trips. He gets a minimal response that doesn't acknowledge the tease, so he exaggerates it. Again, he gets a non-committal response, so he explicitly asks Craig to address the tease. Craig does, but without revealing whether he finds the whole topic amusing. Had Jim been able to see Craig, he would have been better able to interpret the response and adjust if in fact he had hit a sore spot.

It was particularly interesting to see how participants used visual cues to convey disagreement. In many cases, participants looked away from a speaker when they disagreed with what she was saying, sometimes returning their gaze as soon as she said something they agreed with or when the topic changed. In other cases, they responded in understated terms but looked down and sat back in their chairs while doing so. Previous research shows that people prefer to use unspoken cues to handle topics that raise politeness issues because it enables them to handle potentially threatening situations more gracefully and effectively [Brown and Levinson, 1978, Isaacs and Clark, 1990]. Over the phone, either the disagreement or misunderstanding is never communicated or the addressee must raise it explicitly, which can make it clumsy to resolve.

Of course, losing the shades of meaning conveyed in expressions and body position does not often cause dramatic effects, although there are cases when it would be critical to realize someone is not agreeing or that a comment is intended to be humorous. Our basic argument is that over the course of a conversation or series of conversations, seeing each other's facial expressions, gestures and posture generally increases participants' level of mutual understanding without requiring extra effort. Although any single instance of conveying these cues is expendable, the aggregation of them over time makes a substantial difference in the rapport and kind of interactions that occur between colleagues.

Managing Pauses

The video also made it easier to manage extended pauses, which generally must be explained in phone conversations. In one dramatic example during the five-way DVC, the two Billerica participants spend over two minutes looking for an electronic mail message while the others wait. There are extremely long pauses, punctuated by the other three teasing the two in Billerica and having a casual conversation among themselves. The Mountain View participants are able to monitor the other two members' progress and adjust their expectations accordingly.

There were certainly instances of non-problematic pauses during the phone conference as well. In fact, one lasted as long as 28 seconds. However, on the whole they were more likely to be explained explicitly. At one point in a phone conversation, Dave says "I'm trying to look down things that are open bugs," meaning that he is consulting a list. For the next seven minutes, his participation in the conversation is minimal, until he says "I can't find anything else in here."

Design Implications of Adding Video

One implication of this finding is that, relative to the phone, video should be most helpful in those situations when people's rich set of interaction skills are most in demand. Our data suggest that one such case is the resolution of conflicts. Cultural norms tend to discourage people from handling disagreements directly, requiring people to rely more heavily on subtle unspoken cues to interpret another person's attitude. Through video, a speaker may notice an addressee's unconscious expression or shift in posture, and adjust her utterance in mid-stream to head off a misinterpretation. This finding suggests that, relative to audio only, video would also be of use for handling other highly interactive situations when nonverbal cues are most helpful, such as negotiating or creating rapport. Finally, video should be more effective than the phone for people who are working together from different locations over a long period of time. If remote collaborators can communicate richer information more easily, they are likely to have fewer misunderstandings and more effective interactions. Of course, it would be better still to carry out such activities face-to-face, but these are at least a few areas where video and audio offer an advantage over audio alone.

It is important to note that although these subtle cues arrive through the visual channel, participants often use the audio channel to respond to the information. For example, after seeing someone show doubt, the participants in our study often verbally explained more fully, asked about the other person's concern, etc. Notice, also, that much of the speaker's adaptation depends on tightly integrated verbal exchanges. Previous studies have shown that small delays in the audio can seriously disrupt participants' ability to reach mutual understanding and reduce their satisfaction with the conversation [Krauss & Bricker, 1967; Tang & Isaacs, 1993]. This presents a design trade-off, because synchronizing video with audio is typically accomplished by delaying the audio until the more computationally-intensive video is processed. However, delaying the audio reduces the participants' ability to make use of the information in the video. In effect, delaying audio to provide synchronized video and audio generates a rich set of visual information, but people cannot effectively respond to it because of the introduced delay. We have found that users of such a system feel far more frustration about this delay than they do over of a lack of synchronization [Tang & Isaacs, 1993].

In our DVC prototype, we transmitted the audio as fast as possible, without attempting to preserve synchrony with the video. One-way audio delays ranged from 0.32 to 0.44 seconds, while video arrived noticeably later. We found that, although the participants found it slightly disturbing when the video did not match the audio, they still had well-timed interactions that were far richer than those we have observed among people using a commercial video conferencing system, which delayed the audio by about 0.57 seconds (one-way) to synchronize with video [Tang & Isaacs, 1993]. In fact, one group who was using this audio-delayed commercial system decided to turn off the audio and use a half-duplex speaker phone connection instead, demonstrating their strong preference for instantaneous audio over synchronized audio and video.

It was somewhat surprising that the participants accomplished rich interactions using the DVC prototype with audio delays as long as 0.44 seconds. Still, our experience is consistent with a previous study that showed minimal detrimental effects of 0.3 second audio delays (one way) compared to 0.9 second delays [Krauss & Bricker, 1967]. On the other hand, Wolf [1982] found that participants who interacted with a 0.420 second one-way audio delay rated the audio and interaction quality significantly lower than those who experienced 0.167 second delays. However, that study reported only participants' ratings of audio quality and simultaneous speech rather than measuring actual audio problems and overlapping speech. Our experience concurs with Wolf's findings because our participants did notice and complain about the 0.32-0.44 second audio delays. Nonetheless, we found that they were able to effectively compensate for audio delays within that range.

Limitations of Video

Managing Turn Taking

Exchanging shorter turns more frequently indicates that in the face-to-face encounters, the participants were able to more tightly coordinate their utterances, which research has shown enhances their ability to reach mutual understanding [Clark & Schaefer, 1987; Clark & Wilkes-Gibbs, 1986; Isaacs & Clark, 1987]. It is unclear why the participants in the two-way DVC and face-to-face meetings did not differ in their turn duration even though they exchanged turns more rapidly. Apparently, there must have been more silence in between turns during the DVCs. Nonetheless, in both cases, the turn rate indicates that the participants more tightly coordinated their turn-taking. This finding indicates that while video improves the ability to handle conflict and confidential issues compared with the phone, face-to-face interactions are even better than video conferences for handling those types of sensitive issues.

It should be noted that this turn-taking finding is inconsistent with some other similar research. Sellen [1992] did not find a significant difference in number of turns when comparing two video conditions to face-to-face interactions. It seems plausible that the difference stems from the fact that her video setup used analog audio and video over short distances, which resulted in nearly no transmission delay. This would suggest that difficulties in managing turn taking are primarily an artifact of the audio delay and not an inherent limitation of video. However, Krauss and Bricker [1967] varied the audio transmission delay for an audio-only task, and they showed a difference in turn length only when the delay lasted 0.9 seconds, but not when there was no delay or a 0.3 second delay. They also found no difference in turn frequency in any condition. The difference in our findings may also be caused by our different measures of "turns." Sellen [1992] did not count as turns short "backchannels" that lasted less than 1.5 seconds, whereas we used a cruder definition of turns that included any utterance that lasted at least .5 seconds. Perhaps the participants in our study used more backchannels when they were face-to-face than when they were talking over the DVC, which could account for the discrpancy in our findings.

Controlling the Floor

In contrast, in our desktop video conferencing prototype, it was impossible to direct attention toward a specific person in a multi-way conference. Everyone sees you through the same camera, so if you are looking at one person's video image, it appears to everyone as if you are looking at all of them. Not surprisingly, participants in DVCs did not seem to use body or eye position to control the floor. (But see [Sellen, 1992] for one way to overcome this obstacle.)

Instead, people tended to use each other's names to address each other. For instance, at one point, Jack and Craig start talking at the same time and Jack gets the turn. As Jack starts speaking, Jeff overlaps with, "I didn't hear you, Craig" in an attempt to direct the next turn toward Craig. However, Jack holds on to the turn, and after he is done Jeff again explicitly asks Craig to take the next turn. Had they been face-to-face, Jeff might have used gestures to help Craig win the previous turn from Jack.

Using Peripheral Cues

For example, during a 30 second sequence of Jeff and Kate's face-to-face interaction, Jeff is talking and Kate is looking down as she takes notes. Three times, Jeff looks up at Kate for confirmation, and each time, she nods or replies "Yeah," without looking up or interrupting her writing. She is obviously able to sense his head position and eye gaze and recognize that he is seeking a response.

We did not see this kind of subtle coordination based on peripheral cues in DVCs. If anything, we saw many instances when the participants just missed each other's glances. (See [Heath & Luff, 1991] for a discussion of similar problems.) In one typical example, Jeff glances at Kate as he finishes speaking, but looks away too soon to catch Kate's nod in response. At another point, Jeff misses Kate's smile, so he responds to her comment seriously.

Having Side Conversations

In the five-way face-to-face meeting, the conversation occasionally broke into two parallel conversations and then seamlessly transitioned back to a single conversation. For example, at one point Jack makes a joke and everyone but Kate laughs. While the others continue with the conversation, Kate looks at Jack and asks him to repeat what he said, which he does. She comments on his joke and then they both refocus on the group's conversation. This side conversation can be accomplished because participants can "open" a second audio channel and because the visual cues enable everyone to understand who is participating in which conversation when.

Pointing

We saw few instances of pointing in either the two-way or five-way DVC, even to indicate items in their own space. We saw one instance when Jeff pointed to his image of the two people in Billerica, but from the other participants' perspective, he simply appears to be pointing to his screen. It is difficult for them to determine exactly which image he was indicating.

In contrast, we saw many instances of pointing during both face-to-face meetings. During the five-way meeting, the participants repeatedly pointed to places in their own documents and at times reached over to each other's documents to point out a particular line or diagram. In the two-way meeting, Kate pushes part of the document between her and Jeff, and they repeatedly point to different parts of it as they talk about it. We did see instances of this kind of pointing when the pariticipants used the Show Me shared drawing tool, but it wasof course accomplished through a cursor on the shared window rather than over the video channel.

Manipulating Real-World Objects

We saw instances of both of these during face-to-face interactions, whereas no equivalent behavior was possible using our DVC, again unless they used the shared drawing tool. For example, when Kate pushes the document to the middle of the table, she and Jeff write and draw on it, at times building on each other's sketches or comments. They also continue to write on their own pads, transitioning easily between their own space and the shared space. In another simple example from the five-way meeting, Kate leans over to look at Jack's copy of the document to see where he is looking.

Advantages of Video over Face to Face

We see an interesting parallel with electronic mail, which people use, among other reasons, when they want to handle certain factual or practical matters, perhaps without "bothering" with accompanying social interaction. Using e-mail does not mean people do not also use other communication techniques to handle more social or interactional matters. It merely provides another option when textual content is most important.

Secondly, participants in DVCs are normally in their own offices, with all those resources at their disposal. All participants can spontaneously bring into the discussion both on-line and off-line materials if they become relevant. In addition, if one person is looking for something or handles an interruption (a phone call, a person dropping by, or even an incoming e-mail message), the other members can draw on their own private space to use the time productively. As a result, meetings can and were used at times more like loose connections akin to sharing an office. In some cases, individual meetings smoothly shifted between focused conversations and loose, intermittent interactions. Users of other desktop video conferencing systems have also been reported to open up video connections between offices to create virtual shared offices, while at other times they used the connection for focused interactions [Bly et al., 1992; Fish et al., 1992].

This kind of interaction may be inappropriate at times, and in fact members of the team we observed said they were sometimes annoyed when one member stopped participating as he read or answered an incoming e-mail message. But this type of "shared space" can be a useful environment for certain types of activities.

Design Implications from the Limitations of Video

One potential danger of such an approach is that it may force people to take explicit actions to carry out behaviors that are normally negotiated unconsciously. For example, requiring users to indicate explicitly when they want the next turn eliminates their ability to manage the politeness issues around floor control. Doing so may also eliminate cues about the degree of spontaneity and enthusiasm in a participants' desire to contribute. In addition, artificial behaviors may be interpreted differently by other participants. For instance, a person who would have been seen as enthusiastic might be perceived as dominating if she uses an explicit mechanism rather than a socially negotiated one to manage floor control.

In general, we recommend thinking in terms of enabling a wider range of collaborative tasks by broadening the shared space among participants. This can be done by integrating other collaborative tools with a video-based system. Such a system may entail providing one or more mechanisms to enable particular collaborative activities (e.g. pointing, noticing motion), but it should also expand the participants' ability to handle collaboration issues through the standard social negotiation process.

The integration of the Show Me shared drawing program into the DVC prototype is an example of such an approach. Previous studies had shown that the ability to draw shared diagrams and pictures is an important aspect of many interactions [Olson & Bly, 1991; Tang & Minneman, 1991; Tang, 1991b]. We have mentioned some ways in which Show Me enabled a wider range of collaborative activities not available through video by increasing the nature of the shared space among the participants. Not only could participants bring into discussion any document or image from their workstations, but they could also use the cursor to point to parts of the image, and they could track each other's attention through their cursors. We did not build in protocols to prevent people from erasing each other's work, relying instead on the audio connection and social negotiation for people to manage its usage. Our intention was to enable a new type of activity (shared drawing), which involved building technology to support certain behaviors (showing certain objects, pointing, tracking attention) as well as relying on existing collaborative behaviors to handle many of the social interaction issues.

On the other hand, the tool was not as successful as we would have liked because it allowed for sharing of only one bitmap image at a time. If two people wanted to jointly edit a document, they could not work on the actual document. One person would have to make changes and then transmit a bitmap of the updates. To move on to another page, one person had to page the actual document and then transmit the image of the next page. The essential problem was that the shared space was not as broad as we would have liked, and that limitation did appear to reduce the usefulness of the tool.

Our observations lead us to conclude that tools designed to supplement a video conferencing system should:

• broaden users' shared environment;
• enable behaviors associated with particular collaborative tasks;
• take advantage of users' existing collaboration skills;
• not require conscious actions for behaviors that are normally done unconsciously.

We hope that we have drawn attention to some of those limitations so that we may have more realistic expectations of video systems that do not specifically address them, and so that we may focus our development efforts on tools that help compensate for these drawbacks. In addition, our study identified some specific ways in which video affects the processes rather than the products of interaction compared to audio only and face-to-face. It remains to be shown if and how these process effects accumulate over time into productivity or product quality effects. More research is needed to explore these longitudinal effects of video support for remote collaboration.

Acknowledgements

References

Brown, P. and Levinson, S. Universals in language usage: politeness phenomena. In E. Goody (Ed.), Questions and Politeness (pp. 56-311). Cambridge: Cambridge University Press, 1978.

Chapanis, A., Ochsman, R.B., Parrish, R.N. and Weeks. G.D. Studies in interactive communication: I. The effects of four communication modes on the behavior of teams during cooperative problem-solving, Human Factors, 14 (6), 1972, pp. 487-509.

Clark, H.H. and Schaefer, E.F. Collaborating on contributions to conversations, Language and Cognitive Processes, 2(1), 1987, pp. 19-41.

Clark, H.H. and Wilkes-Gibbs, D. Referring as a collaborative process, Cognition, 22, 1986, pp. 1-39.

Duncan, S. Some signals and rules for taking speaking turns in conversation, Journal of Personality and Social Psychology, 23(2), 1972, pp. 283-292.

Fish, R.S., Kraut, R.E. and Chalfonte, B.L. The VideoWindows system in informal communications. In Proceedings of the Conference on Computer-Supported Cooperative Work (Los Angeles, CA, 1990), pp. 1-11.

Fish, R.S., Kraut, R.E. and Root, R.W. Evaluating Video as a technology for informal communication, Proceedings of CHI `92 Human Factors in Computing Systems, (Monterey, CA, 1992), pp. 37-48.

Francik, E., Ehrlich Rudman, S., Cooper, D. and Levine, S. Putting innovation to work: Adoption strategies for multimedia communication systems, Communications of the ACM, 34(12), 1991, pp. 53-63.

Gale, S. Human aspects of interactive multimedia communication, Interacting with Computers, 2, 1990, pp. 175-189.

Grudin, J., Why CSCW applications fail: Problems in the design and evaluation of organizational interfaces, Proceedings of the Conference on Computer-Supported Cooperative Work, (Portland, OR, 1988) pp. 85-93.

Heath, C. and Luff, P. Disembodied conduct: Communication through video in a multimedia environment. In Proceedings of the CHI `91 Conference on Human Factors in Computing Systems (New Orleans, LA, 1991), pp. 99-103.

Isaacs, E. A. and Clark, H. H., Ostensible invitations, Language in Society, 19, 1990, 493-509.

Isaacs, E.A. and Clark, H.H., References in conversation between experts and novices, Journal of Experimental Psychology: General, 116, 1987, pp. 26-37.

Krauss, R.M. and Bricker, P.D., Effects of transmission delay and access delay on the efficiency of verbal communication, Journal of the Acoustic Society of America, 41, 1967, pp. 286-292.

Kraut, R.E., R.S. Fish, R.W. Root and B.L. Chalfonte, Informal communication in organizations: Form, function and technology, People's Reactions to Technology, S. Oshkamp and S. Spacapan, Eds., Newbury Park: Sage Publications, 1990, pp. 145-199.

Mantei, M.M., Baecker, R.M., Sellen, A.J., Buxton, W.A.S. and Milligan, T., Experiences in the use of a media space, In Proceedings of the CHI `91 Conference on Human Factors in Computing Systems (New Orleans, LA, 1991), pp. 203-208.

Ochsman, R.B and Chapanis, A., The effects of 10 communication modes on the behavior of teams during co-operative problem-solving, International Journal of Man[sic]-Machine Studies, 6, 1974, pp. 579-619.

Olson, M.H. and Bly, S.A., The Portland experience: A report on a distributed research group, International Journal of Man[sic]-Machine Systems, 34(2), 1991, pp. 211-228. Reprinted: Computer-supported Cooperative Work and Groupware, Saul Greenberg (Ed.), London: Academic Press, pp. 81-98.

Pearl, A., System support for integrated desktop video conferencing, Sun Microsystems Laboratories, Inc. Technical Report, TR-92-4, 1992.

Root, R. W., Design of a multi-media vehicle for social browsing, Proceedings of the Conference on Computer-Supported Cooperative Work (Portland, OR, 1988), pp. 25-38.

Sacks, H., Schegloff, E. and Jefferson, G., A Simplest systematics for the organization of turn-taking for conversation, Language, 50, 1974, pp. 696-735.

Sellen, A.J., Speech patterns in video-mediated conversations, Proceedings of CHI `92 Human Factors in Computing Systems, (Monterey, CA, 1992), pp. 49-59.

Short, J., Williams, E. and Christie, B., The social psychology of telecommunications, London: John Wiley & Sons, 1976.

Tang, J.C., Involving social scientists in the design of new technology, Taking Software Design Seriously: Practical Techniques for Human-Computer Interaction, Karat, J. (Ed.), Boston: Academic Press, 1991a, pp. 115-126.

Tang, J.C., Findings from observational studies of collaborative work, International Journal of Man [sic]-Machine Studies, 34(2), 1991b, pp. 143-160. Reprinted: Computer-supported Cooperative Work and Groupware, Saul Greenberg (Ed.), London: Academic Press, 1991, pp. 11-28.

Tang, J.C. and Isaacs, E.A., Why do users like video? Studies of multimedia-supported collaboration, CSCW: An International Journal, Volume 1, 1993, 163-196. Also a Sun Microsystems Laboratories, Inc. Technical Report, TR-92-5, 1992.

Tang, J.C. and Minneman, S.L., VideoDraw: A video interface for collaborative drawing, ACM Transactions on Information Systems, 9(2), 1991, pp. 170-184.

Tatar, Deborah, Using Video-Based Observation to Shape the Design of a New Technology, SIGCHI Bulletin, Vol. 21, No. 2, October 1989, pp. 108-111.

Williams, E., Experimental comparisons of face-to-face and mediated communication: A Review, Psychological Bulletin, 84(5), 1977, pp. 963-976.

Whittaker, S., Frohlich, D., & Daly-Jones, O. (1994). Informal workplace communication: What is it like and how might we support it?, Proceedings of the Conference on Computer Human Interaction, (131-137). Boston: ACM Press.

Wolf, C.G., Video conferencing: Delay and transmission considerations, in Teleconferencing and Electronic Communications: Applications, Technologies and Human Factors, L.A. Parker

Wulfman, C.E., Isaacs, E.A., Webber, B.L., and Fagan, L.M. Integration discontinuity: Interfacing users and systems, Proceedings of Architectures for Intelligent Interfaces: Elements and Prototypes (Monterey, CA, 1988) pp. 57-68.