***** To join INSNA, visit http://www.insna.org ***** Scott: Thanks for these very interesting-looking cites. (Everyone else too. And just when my readme stack was getting lower than the doorknob.) A correlation of r=.55 is indeed considered moderate by common rules of thumb. That is enough "signal" for some purposes and I don't aim to indict the study you described. Still, that is in the neighborhood of a quarter of variance being accounted for, far below what I would personally expect to find if the reports were some kind of simple reflection of the behaviors. The omission errors you describe are interesting. In the Corman & Bradford (1993) study I cited earlier in the thread, we found that omission errors were related (logarithmically) to communication load. I don't think we checked status but that's probably correlated with load anyway. We also found commission errors (i.e. relations reported that couldn't be observed). These tended to reproduce the group structure of the organization. A study by Freeman, Romney, & Freeman (1987, American Anthropologist, vol. 89) reported similar results. So we can agree that there is a host of structural influences on individual level reports, and that is one reason why they diverge from behavior. The influences are collective-level and probably stem from the group "processing" its own behavior over time. We also know that occasions for behavior are structured, and not necessarily in a way that reproduces an existing network. So in Organization X some exigency requires representatives from departments A, B and C to meet. These people might not know one another or they might even be sworn enemies, yet their co-behavior is activated by circumstances. Similarly, two people might be selected for a project from a pool of people with more or less equivalent (say, peer) relationships, such that the variance in networks links can't possibly explain the resulting behavior. This is another reason why behaviors might diverge from perceived relationships. In my humble opinion, all this says that a complete explanation of how networks are produced and reproduced over time we needs to consider: (1) A network of perceived relationships that is abstract and relatively stable, (2) concrete but ephemeral behavior that manifests (elements of) the network, (3) a process of social cognition that "updates" the perceived network based on behavior manifested, and (4) a social activation process that structures occasions where the network (elements) might become manifest. I suspect that Dave's cognitive nets are good measures of #1, and that what you get with traditional network questionnaires are individual (and biased) views of #1. Best, Steve -----Original Message----- From: Social Networks Discussion Forum [mailto:[log in to unmask]] On Behalf Of Scott Gest Sent: Monday, January 23, 2006 1:12 PM To: [log in to unmask] Subject: Re: [SOCNET] CSS & "A Million Little Pieces" ***** To join INSNA, visit http://www.insna.org ***** Michael, Steve, Don (and others): Regarding the difficulty of directly observing networks, you may be interested in related research on classroom-based adolescent peer groups. The late Robert Cairns, a developmental psychologist, conducted observational research (including numerous animal behavior studies) and was well aware of the the value, difficulties and limitations of such data. In the early 1980's he developed a "social cognitive map" method (similar to Krackhardt's CSS method) in which each child in the network (typically a classroom) is asked: "Are there any kids in your (grade/class) who hang around together a lot?" Each child identifies as many groups as he/she can, including groups in which he/she participates. (Unlike the CSS method, children are not required to classify all peers into a group.) Data are summarized in a (valued) multi-informant matrix in which each cell represents the number of times two individuals are named to the same group. Cairns used a fairly simple algorithm to extract groups from this matrix: he did not use formal SNA tools, but his approach is conceptually very similar to structural equivalence methods. Cairns conducted two studies in which he validated groups derived from the multi-informant matrix against direct observations of classroom interactions. In a pilot study (Cairns, Perrin & Cairns, 1985), he found that 7th graders interacted with members of their SCM- identified groups at rates 3 to 4 times higher than with other same- sex peers. In a much more extensive study of 78 adolescents spread across dozens of classrooms (each of whom was observed for several hours across multiple days), virtually the same relative interaction rates were obtained (i.e., 3-4x higher). More generally, for each adolescent we constructed an "interaction profile" (a vector of observed interaction rates with classmates) and a "co-nomination profile" (a vector summarizing number of times nominated with each classmate): the median correlation between these profiles was r = . 55. I had the good fortune to work in Cairns' lab and recently published the latter results with several of his long-term collaborators (Gest, Farmer, Cairns & Xie, 2003). In other words, we find that the aggregated reports of members of a network can provide a reasonably valid estimate of "observable" behavior in the network. Given that observations were only conducted during classroom instructional time and therefore failed to capture behavior in other potentially relevant settings (e.g., bathrooms, hallways, buses, playgrounds), I think one could make the argument that the aggregated peer reports are more likely to reflect actual behavior than any feasible observation system. At any rate, we find some comfort in the fact that the peer report procedure has some demonstrable link to directly observed behavior patterns. Regarding biases in ego reports, a few researchers using Cairns' method have examined the correspondence of ego-reports to the groups derived from the multi-informant matrix (Leung, 1996). The general finding is that ego reports (at least among children in the 4th to 7th grade range, ages 10 to 13) are reasonably accurate but are subject to self-enhancement biases. These biases mainly take the form of "errors of omission" in which ego omits relatively low-status network members from self-reports; falsely including higher-status peers is less common. These biases are more evident when ego himself/ herself is relatively low-status in the network. Like Krackhardt, we find large individual differences in the completeness and accuracy of children's perceptions of the classroom social network. A graduate student in my lab (Alice Davidson) is beginning to use SNA methods to explore whether the correlates of these individual differences may be similar to those proposed by Krackhardt (e.g., do individuals with greater centrality provide more complete/accurate reports?). We are newcomers to SNA but are optimistic that theories and concepts may be quite useful in thinking about children's peer group dynamics and the correlates of varyng perspectives on the network. best regards, Scott References: Cairns, R.B. & Cairns, B.D. (1994). Lifelines and Risks: Pathways of Youth in Our Time, New York, NY, US: Cambridge University Press. Cairns, R. B., B. D. Cairns, H. J. Neckerman, S. D. Gest, and J. L. Gariepy. 1988. Social Networks and Aggressive-Behavior: Peer Support or Peer Rejection? Developmental Psychology 24:815-823. Cairns, R. B., J.E. Perrin, and B.D. Cairns. 1985. Social structure and social cognition in early adolescence: Affiliative patterns. Journal of Early Adolescence 5:339-355. Gest, S. D., Farmer, T. W., Cairns, B. D., & Xie, H. (2003). Identifying children's peer social networks in school classrooms. Links between peer reports and observed interactions. Social Development, 12, 513-529. Leung, M-C. (1996). Social networks and self-enhancement in Chinese children: A comparison of self reports and peer reports of group membership. Social Development, 5, 146-157. On Jan 23, 2006, at 12:05 AM, Don Steiny wrote: > ***** To join INSNA, visit http://www.insna.org ***** > > Steven, > > Your comment on the difficulty of getting observable measures > resonates. I just had the pleasure of attending a conference in > Linz with Ivan Chase and spending several days traveling and > general hanging out. He presented work he is currently doing on > pecking orders in chickens and he studies fish and other > vertebrates. He describes his work as incredibly tedious > analyzing the interactions of chickens or fish over 12 hours and > recording every contact. This requires multiple viewings of video > tapes and takes a very long time. His most recent (unpublished) > paper is on dominance hierarchies in chickens and he shows, in a > new way, something he has shown before: that the intuitive notion > of dominance, that the biggest and strongest it the top, the second > strongest second and so on is not the way it works. It is also > true that intransitive relationships are rare, in that the top > chicken pecks those below, but they do not peck up (or do so very > rarely). One experiment he did he separated the animals for a > few weeks so they forgot the dominance order, put them back > together and in more that 2/3 the cases a new hierarchy formed. > In other words, the linear hierarchies that form do not form for > the obvious reasons. > The point here is that even with relatively simple creatures > like fish and chickens in small groups (4), it is very time > consuming and difficult to really observe a single aspect of their > behavior. > If you are familiar with Paul Eckman's work on emotions and how > we influence each other at a subconscious level (discussed in the > book recommended at last year's Plenary Session by Ron Breiger at > SunBelt -- Looking for Spinoza by Antonio Damasio), it starts to > become clear how daunting a "behaviorial" analysis of networks > would be. We do rapid (1/25th of a second) stereotypical > behaviors that seem to be communicating which neither we or those > we are with are aware of at a conscious level. It may be best to > look at this at an other level of analysis than as networks, such > at Harrison White's disciplines, but however you look at it it > reinforces you (Steven's) point about the difficultly of getting > observable measures. > > Gibson's work on turn taking in group discussions is pretty cool > and a step in that direction (Harrison White discusses this in the > new edition of Identity and Control). > I think with tools like digital video that can be slowed down > becoming cheap we can possibly look at behavior in small groups, > but we would need to know what we were looking for. It is much > easier said that done. > > -Don >> ***** To join INSNA, visit http://www.insna.org ***** >> >> Michael, >> >> "I do believe that something like actual, 'objective' human behaviors >> happen in the world." Wow, how un-PC. Can't you get your cultural >> anthropologist license revoked for saying something like that? :-) >> >> My dissertation (way back when) was on this subject of perceived >> versus >> observable behavior in network research, and it's been the main >> thrust >> of my research program since. One important point is that it is >> extremely hard to even *get* observable behavior measures, >> especially in >> a network of any appreciable size. So the studies that have looked at >> this have used very small groups, strange contexts (e.g. HAM radio >> operators who keep logs), questionable observation schemes (walk >> through >> an organization every half hour and write down everyone you see >> talking), or incomplete records of behavior (i.e. e-mail flows). >> >> In cases when people have compared such observations to standard >> network >> questionnaire responses, they have found correlations that range from >> bad to terrible. Some people think these data are noisy (more or >> less) >> and by mathematical manipulation we can recover valid info about the >> actual behavior. I disagree with this and believe (as you seem >> to) that >> they are different phenomena. I have done one study showing that >> self-report data is systematically biased by factors such as one's >> communication load and position in an organization structure, and >> another showing that people vary widely in terms of the "evidence" >> they >> use for formulating their answers to network questionnaires. >> Based on >> this research I've theorized (borrowing Giddens's notion of duality) >> that networks exist in the domain of social structure, behavior >> exists >> in the domain of social interaction, and processes of social >> activation >> and social cognition continuously transform one into the other. >> (Cites >> to these studies are below and I will send on request.) >> >> If you believe this model then you cannot really say that one of the >> phenomena is more valid or important than the other. People act >> on the >> basis of the network they perceive, so you can't hope to explain >> their >> behavior by just looking at what they do. Yet the things they >> actually >> do have consequences (like transferring information to someone else) >> whether they accurately perceive this or not, so you can't assess the >> impact of a network by just looking at people's perceptions, either. >> Unfortunately, this creates quite a research challenge because you >> need >> good data on both phenomena and as I said above the observable is >> very >> hard to get. >> >> Regarding Krackhardt's cognitive nets, my guess is that they yield a >> high-quality representation of the collective perceived network >> rather >> than a valid measure of behavior. But I know of no study that has >> done >> a three-way comparison between perceives, cognitive, and observable >> networks. If you find one please let me know. >> >> Good luck! >> >> Steve >> >> Corman, S. R., & Scott, C. R. (1994). Perceived communication >> relationships, activity foci, and observable communication in >> collectives. Communication Theory, 4, 171-190. >> >> Corman, S. R., & Bradford, L. B. (1993) Situational effects on the >> accuracy of self-reported organizational communication behavior. >> Communication Research, 20, 822-840. >> >> Corman, S. R., & Krizek, R. L. (1993) Accounting resources for >> organizational communication and individual differences in their use. >> Management Communication Quarterly, 7, 5-35. >> >> ________________________________________________ >> Steven R. (Steve) Corman >> Professor, Hugh Downs School of Human Communication >> Arizona State University >> http://www.public.asu.edu/~corman >> Chair, Organizational Communication Division >> International Communication Association >> http://www.icahdq.org >> >> >> -----Original Message----- >> From: Social Networks Discussion Forum >> [mailto:[log in to unmask]] On >> Behalf Of Michael Reed >> Sent: Sunday, January 22, 2006 5:55 AM >> To: [log in to unmask] >> Subject: [SOCNET] CSS & "A Million Little Pieces" >> >> ***** To join INSNA, visit http://www.insna.org ***** >> >> I'm a cultural anthropologist who is new to new SNA; hence, I've >> turned >> to >> Krackhardt's High-Tech Managers data (including the 1987 article, >> "Cognitive >> Social Structures")--recommended by Wasserman and Faust (1994)--as >> a way >> to >> learn about a "simple," one-mode, 3-relation set of SNA data. (I'm >> about >> to >> conduct similar data collection with a group of 40 women >> entrepreneurs.) >> >> As I was reading the 1987 article (expecting, modestly, to learn some >> basic >> SNA skills), I was struck by the actual focus of Krackhardt's >> research, >> i.e., the contention that "...what people say...bears no useful >> resemblance >> to their behavior" (Bernard et al., 1982). The article bears down on >> BEHAVIOR ("what actually happened") vs. COGNITION/PERCEPTION >> ("people's >> perceptions, often in retrospect, of what actually happened"). >> >> I am intrigued by Figure 3 in the article, which shows Person 15's >> "slice" >> (how he/she sees relations between pairs of the 21 managers), vs. >> Figures 1 >> & 2 (the "locally aggregated" and "consensus" structures). It is >> clear >> that >> Person 15's perceptions are wildly different from the more >> "objective" >> measures. I personally would be concerned if my own perceptions of >> "reality" >> varied that much from "actual reality"! I would be tempted to say >> cynically >> that Person 15 is "living in a bubble"! (but that's no doubt unfair). >> For >> example, in my daily life, I frequently try to do "reality checks" to >> make >> certain that my thoughts and perceptions jibe, more or less, with >> those >> of >> other people. >> >> Finally, I just happened to read the Krackhardt article right after >> reading >> Mary Karr's op-ed piece, "His So-Called Life," in the Jan. 15 >> NYTimes. >> Here >> she weighs in on the recent uproar about James Frey ("A Million >> Little >> Pieces") and the question: Should a memoir be held to higher >> "factual" >> standards than a piece of fiction? As someone who wrote a daily >> research >> journal in Africa and who is now in the midst of trying to finish a >> novel, I >> am very interested in whether or not it is even possible for a >> memoirist >> to >> accurately document on paper "the way life and behavior ACTUALLY >> occurred at >> some past time." (As a novelist, I am most concerned with what I >> would >> call >> "emotional truth," although getting the "facts" straight is >> important, >> too.) >> >> I know my own memory to often be extremely "inaccurate"; I don't >> know if >> this inaccuracy is a function of my advancing age (54) or simply >> of the >> fact >> that I didn't pay as much attention to memory when I was younger and >> thus >> didn't see how problematic it is. Sometimes I'm nearly resigned to >> believing >> that all human memory is basically a "creative >> reconstruction" (done in >> the >> present according to present needs and wants) of the past. That's why >> historians turn to written, archival sources for help (not that >> they are >> without bias or error--we can never escape the fact that fallible >> humans >> are >> involved). >> >> Still, I do believe that something like actual, "objective" human >> behaviors >> happen in the world. The question is, How accurately can we humans >> measure >> or remember or understand those behaviors, i.e., "what really >> happened"? >> Krackhardt ends his article by stating, "But the task of future >> research >> should not be to show that behaviors are more important than >> cognitions, >> nor >> that cognitions are more important than behaviors. Rather, our >> task will >> be >> to show the consequence of each--behavior and cognitions." >> >> As someone who believes that there IS an important difference >> between a >> memoir and a piece of fiction, I would have to say that, in some >> sense, >> the >> behaviors must take precedence (although I admit that "behavior" is >> itself a >> cognitive creation; we never escape from our mental jail): we need to >> make >> certain that our cognition about the past doesn't willfully (or even >> unintentionally) distort past behaviors. >> >> Michael C. Reed, Ph.D. >> Independent Consultant & Cultural Anthropologist >> Kalamazoo, Mich., USA >> [log in to unmask] Tel. 269-342-4025 Cell phone 269-808-8983 >> >> _____________________________________________________________________ >> SOCNET is a service of INSNA, the professional association for social >> network researchers (http://www.insna.org). To unsubscribe, send >> an email message to [log in to unmask] containing the line >> UNSUBSCRIBE SOCNET in the body of the message. >> >> _____________________________________________________________________ >> SOCNET is a service of INSNA, the professional association for social >> network researchers (http://www.insna.org). To unsubscribe, send >> an email message to [log in to unmask] containing the line >> UNSUBSCRIBE SOCNET in the body of the message. >> >> >> > > _____________________________________________________________________ > SOCNET is a service of INSNA, the professional association for social > network researchers (http://www.insna.org). To unsubscribe, send > an email message to [log in to unmask] containing the line > UNSUBSCRIBE SOCNET in the body of the message. > --- Scott D. Gest, Ph.D. Assistant Professor, Human Development & Family Studies Penn State University 814-865-3464 office _____________________________________________________________________ SOCNET is a service of INSNA, the professional association for social network researchers (http://www.insna.org). To unsubscribe, send an email message to [log in to unmask] containing the line UNSUBSCRIBE SOCNET in the body of the message. _____________________________________________________________________ SOCNET is a service of INSNA, the professional association for social network researchers (http://www.insna.org). To unsubscribe, send an email message to [log in to unmask] containing the line UNSUBSCRIBE SOCNET in the body of the message.