December 9, 1993
Community Memory is a public message system created to facilitate community communication as an electronic "village square" in the San Francisco Bay Area. During the Fall 1993 semester at UC Berkeley, Dr. Howard Besser required use of the system by his class to facilitate focus group discussions and provide experience with a public communication system. The messages were posted by the four focus groups in non-public areas established by the students. Although Dr. Besser required two messages per week per student, it was understood that he desired utilization of the system as a means of communication that would extend class discussions and provide a virtual forum for the focus groups' discussions.
The purpose of this paper is to analyze the factors that influenced the use of Community Memory including hardware considerations, the software's architecture and user interface, group dynamics, and motivational factors. The information for this study is the messages themselves on Community Memory dated from 9/24 to 11/24, responses to a short written class questionnaire (Appendix 2), informal conversations with class members and my experiences participating as a student in the class.
The class accessed Community Memory through the UC Berkeley campus computer network, via Unix computer accounts. Community Memory's data structure and menu/option based user interface is designed to accommodate people with little or no previous experience with computers. Boxes frame menus and messages with natural language options or commands enclosed inside angle brackets (< >). Moving the cursor with keyboard arrow keys to the desired option on the current screen is the method of navigation. The F1, F2 and F3 keys can be used at any time to issue the commands OPTIONS, BACK-UP or HELP.
Each of the four class groups established a new and descriptively named discussion group. Selecting any one of these groups would display the three to ten index terms, assigned by group members, to provide a framework for the posting of messages related to the groups' designated topic. Messages fall into two categories, initial or original messages and responses. Selecting any index term accesses the initial messages chronologically, most recent initial message first. Response messages are attached to other messages and are viewed through selecting a <responses> choice under the current message. For example, the message shown below was entered under the Index term "capitalism" on 10/29 and has a response attached. To see this response, the user must first view the shown message, then choose <RESPONSES>.
Press F1 for OPTIONS; F2 to BACK UP; F3 for HELP
/================================<ABOUT MESG.>=<WRITE A MESSAGE >=\
| added by Xxxx Xxxx on 10/29/93
| I was reading U.S. News and World Report this week, and
| there is an editorial in the back entitled "communicopia"
| which has the usual blathering on about how great the
| communcation age will be, how it will further democratize
| society,etc. One interesting note in this cornucopia
| enthusiasm was a plug for the DE-regulation of the
| communication industry. According to the editorial,
|=> See <RESPONSES> to this message
Response messages may also have responses that must be viewed in the same manner. This structure creates discussion threads, in a binary tree or tiered structure, that may contain as many messages as the system can accommodate.
Figure B. Example of 3 tiered Discussion Thread
---A2 responds to A1 on 10/18
A1 posts message on 10/13
---A3 responds to A1 on 10/16 ---A2 responds to A3 on 10/18
Moving about through these message queues is done by going forward and backwards through the messages in order using arrow keys. Exiting a message queue, i.e. index term area, is possible at any time by using the F2 key. One other general feature of the system is the ability to assign a message more than one index term, after entering the text. For example, the message shown in Fig. B was entered on 10/13 a under "capitalism" and also assigned the index term "ideology" and thus the entire thread is accessible through either term's message queue.
The charts accompanying this paper show quantitative use of the system by group from the period 9/24 to 11/24. It must be noted that Group A's number of messages are much higher than shown because members of the group deleted a number of messages in late October. Estimates by members of the group suggest that 8-15 threads of 2 to 5 messages each were erased, i.e. anywhere from 16 to 75 unique messages. Although this set of data cannot be accurately reconstructed for quantitative analysis, information about the circumstances and motivations surrounding the deletion is available and will be considered by this paper.
As is expected the groups with more members, A and D, have a larger number of messages (Fig. 1). It is interesting that the smaller groups, B and C, have higher levels of participation by individuals outside their focus groups (Fig. 2). Scanning the content of the messages, confirms the fact that the two smaller groups, B and C, had less threads of discussions both in terms of the number of messages and in topics discussed. Group C had only one group member who posted with some consistency throughout the two month period, including a few messages in other group areas (Fig. 4C). Group B's messages were predominately posted by two individuals, although one of these was also the second largest contributor of non-group postings (Fig. 4B). Column two of Figure 3 shows that both these groups had higher levels of participation by non-group members. This combined with the fact that the 5 member groups had more threads of discussion (total of first column Fig. 4A and 4D) suggests that a critical mass for internal group discussion in this situation is five individuals.
Another idea generated by these tables is that those in smaller groups who desired more interaction in different topics of discussion would have to browse more outside, whereas those in larger groups would have more internal discussion threads and a lower motivation to browse outside the group. This idea may be supported or discounted by the fact that the three individuals who posted the highest number of messages (without exact figures for Group A's deletion) were D2, D1 and B2 respectively. These same people also had the highest levels of posting to other groups, D2, B2 and D1 respectively (Fig. 4). The individual from Group B, supports this hypothesis since that group consisted of three people and three index terms and this individual posted more messages to non-groups than to Group B itself. In addition, members of Group A (who did not erase postings to other groups) posted few messages, compared to Groups B and D, outside their group.
The two persons from Group D, however, may counter this position because Group D had the highest number of messages of any of the four groups, discounting the Group A deletion and the highest number of messages posted to other groups. However, it can be argued that Group D was more limited in threads of conversation since it had 5 index terms compared to Group A's 10 terms. Members of Group D commented on the need for more index terms, but were prevented by the technology. My experience was that the group's area became crowded because of inadequate index terms, since the group's discussion moved beyond the original terms and poor use of the cross-posting feature, which placed too many messages under inappropriate index terms.
Whatever the reasons, the data (Fig. 4) indicates that, regardless of group size, the individuals who entered a larger number of messages within their group were also those who posted the majority of the messages outside of their focus group.
The fact that more of the messages were responses than initial messages, and scan of message content, indicates that discussion along topics was occurring. The higher number of responses to initial messages as opposed to responses to responses may indicate that not all were willing to follow through the message queue to follow messages. For example, one thread began with a message on 10/25 contains five messages as illustrated in Figure C.
A6 message on 10/25 --Instr. responds to A6 on 10/29 --A2 responds to Instr.on 11/13
--D1 responds to A2 on 11/17 --A2 responds to D1 on 11/17
When a user first accesses the index term "capitalism", the messages displayed on the first screen are most recent initial messages, in this case dated 10/29 (not messages in Fig.5) and 10/25. To view the final 11/17 message of the 10/25 thread the user would have to view the 10/25 screen and choose the option "responses" four times to pass through the original and other messages entered on 10/29, 11/13 and 11/17 respectively to reach the desired message.
Another reason for the higher number of responses may be, as stated on a questionnaire ,"I mostly posted in response to others postings." This reflects a belief from my experience, that it is faster and easier to comment on a posting, than to compose an original idea or posting. Class members did report that they generally like the discussion thread structure, but were unhappy with the inability to jump between messages in this structure. However other factors such as unpredictable system down times or dissatisfaction with the software interface, which will discussed later, may also have contributed to this kind of attitude.
The use of index terms to "cross-post" messages was designed to allow the freedom to places messages under related topics. Use by the class led to a number of the same message threads being posted under multiple index terms, lengthening the time to browse through the message queue because the system does not permit display of a simple lists of messages. In my case this reduced my desire to browse through the different index terms, since I would see many of the same messages. Why waste time looking at the same messages?
Cross-posting also had the effect of giving the feeling that there is a good deal of information when there actually is not. For Group A 53% of the messages were unique with Group B 86%, Group C 74% and Group D 77% respectively (Col. 5, Fig. 1). This was especially true about group A, which had 10 subject terms, the most of any group. While taking notes on the messages in group, I found that four of their ten index terms comprised of entirely cross-posted messages, i.e. no unique messages to any of these terms. Since the group delete which occurred late October, no new messages have been added to these four index terms. Group dynamics had a large effect the on messages entered into Community Memory by the class. The Group A with the largest number of people began with 12 people, though dwindled to 5 by mid October. The smaller groups, B (with 3 members) and C (4-2 members) had the least number of messages. The median group for the majority of the semester, D (5 members) had the median number of messages. This is not surprising considering the two messages per week requirement, but there were other factors involved.
More individuals contributing substantive messages led to more topics and expression of opinions as well as diverse and convoluted discussions. In Group C, for example, primarily one person within the group was responsible for over 50% of the messages. Most people who responded to her messages were from other groups, primarily B and D. Group A, on the other hand reported (part of deletions) a high level of use by a core of 5 members which lead to more threads of discussion. Though not stated, I believe this reflects why this group had less messages outside their group. More internal content did not require exploring the other groups. Additionally it appears that groups B,C, and D did not contribute often to Group A, especially during the first month. This may reflect the time needed to navigate the myriad of Group A's messages which was disconcerting enough to the group itself and lead to the eventual group deletion.
Three questionnaire respondents stated that they preferred the weekly "in-person" meetings to discussion on Community Memory. These comments create the need to discuss the effect group meetings had on the use of Community Memory. It is not unexpected, due the different types of content areas assigned, that each group had a unique set of circumstances that resulted in diverse group experiences, each with its own modes of communication, socialization and scholastic goals..
Group A, as described earlier, experienced the most dramatic change in numbers, but also was also the most early organized with respect to assigning readings, and meeting other expectations. In addition, from discussion with group members, the established group of five members, three women and two men, a high level of interactivity and enjoyment through substantive discussions. In fact, this group is so motivated that it has plans to continue group meetings next semester, in private homes. The factors that led to mass deletion, inlcuding the server downtimes which will be discussed later, led the group to prefer its weekly meetings over Community Memory for discussion.
Groups B and D remained relatively stable throughout the semester, but experienced different types of interaction due to the difference in group size and personality differences. In both, groups there were only one male member who was also the oldest member, whose communication skills were considered weaker than the other female group members. In Group B, the smaller number of members magnified this effect. According to one of the women in the group, the fact that the man's native language was not English and other personality elements resulted in discussions with a great deal of speaking but very little listening. This apparent lack of communication may be reflected in Community Memory, where 12 of the 18 discussion threads and only 44% of the internal messages were posted by Group B members with the two women posting 80% of the groups messages.
In Group D, the four female members all had strong communication skills and generally reported satisfactory feelings towards group discussions, but stated that many of the comments interjected by the male were not on the current line of discussion. The male member volunteered during one meeting, when group dynamics as a topic were being discussed, that he knew his communications skills were not outstanding. As a member of this group, I felt that the four women were comfortable with each other speaking styles and could easily converse on any topic, but varied from accommodating to ignoring the ideas and statements of the man, depending on how they viewed the importance of his content. This attitude was reflected in Community Memory, since his messages were always strongly criticized or "flamed" by the other group members, which may be one cause for his diminishing contributions to Community Memory as the semester progressed.
The most prolific member of Group C reported a most confusing situation, where the group that began with four, but dropped to three and then inadvertently divided. Apparently one group member thought that he was the only person left in his group. The other two members thought the other member had dropped out and changed the weekly meeting time, to better fit their schedules, and continued discussions in person. Nevertheless, as described previously, only C1 entered messages regularly throughout the semester. I not have the opportunity to talk to the other two Group C members and investigate their motives (and I know they were not present the day I distributed the questionnaire), but obviously the class requirement was not sufficient motivation for participation. Comparing these four groups' experiences, it does seem likely that the degree of group fragmentation, negatively effected the internal group utilization of Community Memory.
It was stated by one member of Group D that individual messages are trivialized through the response structure, i.e. people think that what they have to say is important and want it read. In the response framework it is possible for a contribution to be missed except by diligent readers. The same individual said perhaps this is some of the appeal of email and listservs, since you know everyone (except those with filters) will at least receive your message. This idea is indirectly supported by statements from members of Group A who found Community Memory unreliable for broadcasting messages to the group. They said that Unix email was used when they wanted to ensure that everyone in the group received a message such as group assignments and meeting times. Feldenstein says that Community Memory was established to fill the community agora, i.e. to allow broadcast of information by community members to other community members. The dissatisfaction from these groups is caused by the fact that Community Memory is more like a bulletin board, where an ad for help is posted, and may or may not be read. For the purpose of this class, the groups often desired a broadcast tool that could ensure that everyone had received a certain message by a certain time.
One approach to analysis of motivational factors behind the adoption of the system, is found in Rogers' "Attributes of Innovation" from Communication of Innovations: A Cross-cultural Approach. He describes a generalized structure to determine the rate of adoption of innovations that includes five attributes of innovations which are relative advantage, compatibility and complexity, trialability and observability. These attributes provide a useful structure to examine the motivations and attitudes that influenced the class's use of Community Memory.
Relative advantage "is the degree to which an innovation is perceived as being better than the idea it supersedes." In the case of this use, the primary incentive to use the system was that it was required by the professor as a significant part of the grade. In other words, without using this system the individual student would receive a lower grade. All questionnaire responses to the question "what were your primary motivations for entering messages to CM" stated the class requirement as the primary motivating with a percentage scale ranging from 50%-100%. Two respondents connected the class requirement to the idea that they were fulfilling an obligation to their focus groups. Obviously the grade incentive was sufficient to ensure that the class made some use the system. At times students would incorporate items from class discussion, group meetings and public lectures, as was required, to address suggestions by the instructor for the content of group discussions. Within my focus group, it was actually discussed in meetings what we should discuss on Community Memory, in order to fulfull comments made by the instructor. I'm sure other similar scenarios occurred within the others groups.
Roger's discussion of incentives also states that once an incentive is removed, generally the adoption of an innovation ends unless the incentive is diffused into the other attributes. While it is possible that class members might continue to use the system for other reasons after the course has ended, comments both on the questionnaire and informally observed throughout the semester confirm that use of Community Memory will almost completely cease at the end of the term. This conclusion indicates that other attributes of the system do not provide enough incentive or advantages for adoption, i.e. continued use by the class. Compatibility is defined as "the degree to which an innovation is perceived as consistent with the existing values, past experiences and needs of the receivers." The two primary electronic communication systems with which the class was familiar prior to Community Memory were Unix email and the Library school Local Area Network (South Hall LAN). Unix accounts provided access to the Internet through a campus connection which gave access to a variety of Internet tools, information sources and discussion groups, such as listservs and newsgroup. Additionally the class was required to view some of the campus newsgroups and briefly join America Online, a commercial information system. The South Hall LAN had an IBM based mail system which is used infrequently since it is only accessible from terminals inside South Hall. All these experiences and the high level of computer literacy in a number of the class members, did not rate high standards of accessibility, interface and response time which Community Memory.
Throughout the process of collecting data for this study, I came to realize that efficient use of time was an underlying motivation for how and when Community Memory was used. The perceived level of acceptable accessibility to the class was the ability to schedule the time and location to log into the system. There were class members who had to use school facilities, others with computer access from home or work and those who used both. Students time limitations were imposed by campus computing facility hours and school and work schedules.
The most substantial factor that lowered the accessibility of Community Memory was that the server would be down for a few hours at unscheduled intervals during the day, throughout the month of October. The longest interval was from 10/30 to 11/5, which prevented entry of any messages for an entire week. While it is difficult, due to the time frame of data collection, to quantitatively measure if the number of messages entered per week dropped significantly after this week (Appendix 1, not included in Mosaic version), this down time did effect the class "morale" with respect to using Community Memory. According to the class survey every respondent except one would have posted more messages if the server had never gone down. Four of the nine respondents indicated indirectly that their content would have been different if they would have had the ability to spontaneously enter messages at will. "It was a hindrance to not be able to post when I had a clear message," was one comment. Once the server was down, the news spread word of mouth and soon as one student stated " [I] got frustrated and didn't even try to log on until [I] heard by word of mouth that it was up!".
Lack of convenience, meaning time and access, caused students to not use the system as much as required, despite the importance of its use a grade. It is my opinion, which is supported by other members of the class, that if the system downtimes had not occurred then its accessibility would have been viewed as reliable and use of Community Memory would have been more substantial. This scenario is unfortunate since it negatively effected class members overall attitudes towards the system and makes an unbiased assessment of Community Memory's structure and features difficult.
A personal complaint of limited accessibility is based upon three or four occasions when I was booted all the way off the system, while in the process of composing a message, because I had "exceeded the CPU time limit". Since the system does not retain these unfinished messages I was forced to login again and recompose the lost message which would not be the same as the original. In one case, I was so frustrated that instead of rewriting the message, I found another discussion thread and entered a completely different message in that area.
One other accessibility problem reported by various members of the class, predominantly Group A, is related to Community Memory's interface via a modem. At least half the class has the ability to login from home or work. Aside from the downtime problems, there was also a technical problem transmitting the correct F2 key characters from remote keyboard. Almost none of the class members solved this difficulty, making it impossible to back out of any area. Hence once an individual chose a group and subject area, the only way to gain access to a different area was to cut the connection and start over, a cumbersome and undesirable option. At least two people in informal conversation said that the ability to navigate without the F2 key via a modem would have increased their ability to view and write during a login session and would probably have contributed to more viewing of other groups messages.
Members of Group A when asked about reasons for the numerous cross-postings early in the use period, stated that the lack of an F2 key from home was the primary reason. Since it was impossible to back up, the same message would be posted to multiple index terms to it in more index groups without having to log back in and give it a higher level of accessibility by other group members, logging in from home with the same problem.
These problems illustrate that the accessibility from home is an important consideration for computer literate and modem-oriented people. One member from group A said that since she is used to doing email and similar type of work from home, she reserves time on campus to other tasks. Since she know Community Memory was accessible from home, but could not fully use its capabilities, this was a hindrance to being fully motivated to use the system. Again here is a example of how scheduling and efficient use of time are valued over the importance of an assigned task.
An acceptable level of interface for the class required word processing and data transfer tools that were compatible to previously used software. The word processing abilities of Community Memory were minimal. Text is typed into window that covers the previous message. The only methods for navigation within a message's text is through arrow key movement and utilization of the system functions of PRIOR or NEXT [screen] for longer messages (See Fig. A). Deletions and editing are done by moving the cursor with arrow keys to the appropriate spot and then typing more text or using the delete key. When compared with other word processors that the class was familiar with this interface was considered cumbersome and awkward.
Four questionnaire respondents commented that they would have preferred being able to view previous windows, i.e., messages, while entering text. To individuals who are used to writing in a sophisticated word processing environment, the ability to see previous text simultaneously, easily revise and move text appears to be essential to written thought. I believe this is because the process of writing in a highly sophisticated word processing environment is internalized in such a way that decreases linearity and increases spatial visualization of text, when compared to hand-written or typewriter technologies. Adjusting to a system that more closely resembles linear writing, is below expectations and is viewed as inconvenient and time consuming. While I believe this may be an intentional feature by the developer of Community Memory to accommodate non-technical users, it became another limitation of compatibility from the perspective of the class.
Comments about the inability to upload or download were made by three respondents, reflect the comparison of Community Memory to a Unix based email system. One said that "not being able to upload from another WP program into CM affected [the] length [of messages] particularly `with respect to the number of messages and their content." It is apparent that these individuals are used to the freedom to compose separately using a word processing application. The reasons for this were not directly identified by the respondents but the comments do imply circumvention of time restrictions and access issues and the desire to compose in a more friendly or familiar program. The desire to compose or incorporate text from a separate system indicates another example where there was not compatibility with previous experiences.
Other technological/interface barriers out of the users control that were described by different respondents included the inability to add more index terms to a discussion and the lack of flexibility to move between the messages outside of the chronological format which determined the sequence of messages within an index term. The later comment was clarified by comparison to a newsgroup, where it is possible to move directly to a desired message, without navigating through the entire message queue. All of these comments indicate that these members of the user group are familiar and used to technology that allows for greater flexibility.
Two response on the questionnaire highlighted the slowness of the system as another factor that contributed to frustration. Individuals who used CM from home experienced even slower response times over 2400 baud modems, requiring even longer periods of time to view queues of messages. The reflects a basic assumption by modern society that faster is better, since it is more efficient.
This perceived slowness whether at a school workstation or at home accumulates when an individual browses all the messages in a discussion thread. Responses are only added to the message queue, as the person looks at the responses to an original message. Once any message or response is read, it is added to the browse queue for the duration of the session in that index term. For example, in Fig. D the initial browse queue contains on two messages, the initial messages entered by A6 on 10/25 and A1 on 10/13. If the user chooses to look at all the responses to the first thread then the browse queue will contains 6 messages. To return to the original message requires backing up through 5 response messages screens. If all responses were read for both threads, then the message queue would include 10 messages, hence to return from the last response to the first message would backing up through 9 messages.
A6 message on 10/25 --Instr. responds to A6 on 10/29 --A2 responds to Instr. on 11/13
--D1 responds to A2 on 11/17 --A2 responds to D1 on 11/17
---A2 responds to A1 on 10/18
A1 message on 10/13
---A3 responds to A1 on 10/16 ---A2 responds to A3 on 10/18
It is only possible to remove response messages from the queue by backing up to the index term screen. Hence as illustrated a queue of 5 messages can easily expand to include over 20 message if there is significant discussion.
This was a major complaint by three respondents, who found it tedious to look for messages. The reason for this is not just the time involved queuing through messages, but that this process was often necessary to find the most recent comment. Why look at all the message you have already read, when all you want to see is the latest response in a long discussion? This was the primary reason that Group A decided to delete messages. They felt the need to eliminate fully discussed topics and limit the index areas to current discussion.
Complexity is defined as "the degree to which an innovations is perceived as relatively difficult to understand and use." With respect to Community Memory, except for a minority of the class, complexity did not present any difficulties. As a side example, I subscribed to a new listserv at its inception that was predominated by technical problems and questions for the first two weeks. Messages and conversations were dominated with questions of how to deal with technology, i.e. how to reply, how to subscribe, etc. As time passed and users become more familiar with the technology, messages become more topic oriented. With respect to the class use of Community Memory, most of the class was generally familiar with computer use, the transition to the new interface was relatively fast as indicated by questionnaire responses. All except two responded that they were facile with the user interface within one hour or less and the remaining two, were within two hours. Nevertheless four initial messages were posted commenting on technical problems during October and November. These dealt with more advanced functions such as creating a new index term, referencing another message and one in 11/12 was a direct complaint about the system's serve downtimes.
Trialability is "the degree to which an innovation may be experimented with on a limited basis". This is comparable to the idea of all or nothing. Slow commitment in stages increases the probability that an innovation will be adopted. In the case of Community Memory, it was an all or nothing scenario. Each student had to use system whether they like it or not as part of the class. Despite the requirement, there were many who only entered the required number of messages if that. This is also related to the down time of the server, but once some individuals decided that they did not like Community Memory and that it was unreliable, less attempts were made to use the system.
Observability is "the degree to which the results of an innovation are visible to others." With Community Memory observation is possible simply by browsing through the messages in Community Memory. The individual's opinion of the use of the system is determined by which groups and areas they directly observe or participate. Those who showed interest in the disucssions could see the benefits of particpation. Four respondents indicated that the their motivation changed, during the course of the semester, from primarily obligation to include more interest in the discussions. One commented that "my motivation did change slightly, in that I actually became interested in following the discussions". However, those who had little interest in the discussions on Community Memory viewed use as only an obligation. Another response to the change of motivation was " yes, changed [from 75% ] to 95% requirement/obligation to others because I had given up on Community Memory as a communication medium. My focus group discussions were much more interesting."
Throughout this discussion and analyis of student use and motivation, it is obvious that the level of adoption of Community Memory by the class is low. In general the reasons for this are based upon the fact the Community Memory did not meet the technological, scheduling and asthetic needs and expectations of the students. Another important reason, that has been suggested during the course of this paper, is that Community Memory is designed to be user friendly to non-technical people. Since most of the class is computer, if not network, literate, the features of Community Memory were viewed as less than adequate. In addition, I believe that at least half the class preferred the face-to-face discussion of the weekly meetings, becuase they provided more immediate feedback and social interaction than on Community Memory. If the groups had not met as often in person, and if the unfortuanted server downtimes had not given Community Memory the stigma of unreliability, I believe that the use of the system would have been more significant and the attitudes toward it more positive.
Besser, Howard. " Elements of Modern Consciousness ," Social Effects of Information Technology. Winter 1992, p. 1-28.
Brown, Hedy. People, Groups and Society. Philadelphia: Open University Press, 1985.
Carey, Jane M., ed. Human Factors in Information Systems: An Organizational Perspective. Norwood, NJ: Ablex Publishing Corporation, 1991.
Felsenstein, Lee. "The Commons of Information." Dr. Dobb's Journal. May 1993, p. 18-24.
Rogers, Everett M. Communication Technology : The New Media in Society. New York: The Free Press, 1986.
Rogers, Everett M. and F. Floyd Shoemaker. Communication of Innovations: A Cross-Cultural Approach. New York: The Free Press, 1971.
Felsenstein, p. 20-24.
Felsenstein, p. 22.
Everett M. Rogers and F. Floyd Shoemaker, Communications of Innovations: A Cross-Cultural Approach (New York: Free Press, 1971), p. 138.
Rogers and Shoemaker, p. 144-145.
Rogers and Shoemaker, p. 145.
Howard Besser, "Elements of Consciousness", p.10-13.
Rogers and Shoemaker, p. 154-155
Rogers and Shoemaker, p. 155.
Rogers and Shoemaker, pl 155-157.