Social Networks Analysis' Methods and Applications Chapter 2' Social Network Data: Collection and Ap

1
Social Networks Analysis. Methods and
ApplicationsChapter 2. Social Network Data
Collection and Applications

• Wasserman and Faust

2

• What are network data?
• Type of social network variables
• Modes
• Affiliation Variables
• How to sample actors and relations?
• What is your population? Sampling and Types of
  networks
• One-mode networks data Definition, Actors (or
  social entities), Relations, Kinds of relations,
  and Actor Attitudes
• Two-mode networks Definition
• Two set of actors Definition, Actors (or social
  entities), and Relations.
• One set of actors and one set of events
  (Affiliation network) Definition, Actors (or
  social entities), Events, and Attributes
• Special dyadic networks Definition
• Ego-centered networks Definition
• In sum
• The methods of social networks assume that there
  are no theoretical limitations on interactions
  among actors. A social network arises when all
  actors can, theoretically, have ties to all
  relevant actors.
• The primary object of study for social networks
  methods is to define the boundary of a complete
  collection of actors (one or more sets) and the
  ties among them.

3

• Network data, Measurement and Collection
• Measurement
• Data Collection
• Questionnaire
• Interview
• Observation
• Archival records
• Others techniques ego-centered, small world, and
  diaries.
• Cognitive social structure design
• Experimental data
• Ego-centered or local networks
• Small world
• Diary
• Longitudinal Data Collection
• Measurement validity, reliability, accuracy, error

4

• Accuracy
• Remember that social network data is measured and
  collected by having people report on their own
  interactions. The informant accuracy of these
  data is a concern, people are not very good at
  reporting on their interaction in particular
  situations (Bernard, Killworth, and Sailer, 1980,
  1982).
• However, dont forget that social networks
  studies are primarily concern on the true
  structure or stable patterns of interaction
  (long-term patterns of interactions not
  particular relations).
• Other accuracy problem is if the data collected
  from individuals represents organizational or
  corporative relations.
• Validity
• If the measurements measure what they intended to
  measure.
• Construct validity hypothesis testing about
  theoretical propositions, deductions,
  expectations or concepts.
• Concurrent validity how different or opposite
  cases correspondence with the concept measured.
• Reliability
• Consistency of measures stability in time,
  internal reliability, inter-oberver consistency.
• Stability (test-retest reliability) test-retest
  comparison, comparison of alternative question
  formats, and the reciprocity of sociometric
  choices. For social networks, it is problematic
  to conduct retest.

5

• Reliability
• Internal reliability or multiple-indicators
  measures (index) split-halves or alternative
  forms, Cronbachs Alpha that calculates the
  average of all possible split-half reliability
  coefficients (Empirical rule more than 0.80).
• Sociometric questions using ratings or full rank
  orders are more reliable (higher test-retest
  reliability) than fixed choice designs in which
  just a few responses are allowed.
• Response to sociometric questions about intense
  or intimate relations have higher rates of
  reciprocation than sociometric about less intense
  or intimate relations.
• Reliability of aggregate measures is higher than
  reliability of choices made by individual
  actors.
• Measurement error
• Discrepancy between the true score or value of
  a concept and the observed (measured) value.
• Understanding the implications of measurement
  error for different levels of study social
  network actor, pairs of actors, subset of
  actors, and network as a whole.
• Fixed choice data collection designs present some
  particular errors or bias in the way we restrict
  choice or nomination.

6

• Issues
• For what analysis do you need the data? For
  exploratory analysis or determining causalities.
• Evaluate cost and non-response of each method.
• Evaluate accuracy of closed, pre-coded, or fixed
  choice questions. There advantages in time and
  cost but there are disadvantages in variability.
• Evaluate easiness of data processing.
• Reducing interviewer and participant observation
  variability structured methods reduce
  variability.
• Question order and clear instructions are always
  important.
• Reducing coding variability define clearly
  coding frame.
• Pay attention to probing and prompting issues for
  each technique.
• Likert scales usually present some response set
  problems acquiescence (tendency to agree or
  disagree with a set of questions without
  evaluation) and social desirability (tendency to
  answer according to a social perception, not
  individual opinion).
• Data sets
• Krackhardt.s high-tech managers
• Padgetts Florentine Families
• Freemans EIES network
• Countries trade data
• Galsskiewiczs CEOs and clubs network
• Other data

7
Network Data and Measurement

• Peter V. Marsden (1990)

8

• Conceptual questions
• Measuring existing relations (objectivist or
  behavioral networks) or perceived social
  relations by actors (cognitive networks). It
  depends on what we are studying external
  constraints vs internal constraints.
• Temporal elements in the definition of social
  ties. Episodic, transient, momentary relations or
  transactions vs routinized and recurrent patterns
  of relations.
• Relationship between concepts and measures
  (descriptions-indicators). Instead of evaluate
  the traditional validity-reliability, we have to
  be concerned more with the robustness of analytic
  methods to errors.
• Network study designs
• Levels of analysis
• Boundary specification
• Network sampling

9

• Sources of network data
• Surveys
• Questionnaires soliciting self-reports
• Archival methods
• Diaries
• Electronic traces
• Observation
• Informants
• Experiments

10

• Surveys and questionnaires
• Delimit population.
• Different formats dichotomous indicators, fixed
  number of links, scales or ranks, or paired
  comparison.
• Techniques for collecting egocentric networks by
  name generators/interpreter sequence. There are
  three name interpreters
• Reports on attributes of person or alters
  enumerated (age, gender, race, etc).
• Reports on properties of the tie between
  respondent and alter (frequency of contact,
  duration of acquaintance, intensity of relation,
  etc.)
• Reports on the intensity of ties between pairs of
  alters, which can be measure the structure of the
  egocentric network (density).
• Affective and role relations were used as name
  generators
• Multiple name generators approaches.
• Wellman (1981) noted that most of the instruments
  for network measurement seek to elicit supportive
  ties and ignore difficult, disruptive, or
  conflictual connections.
• Using different name generators with different
  subsamples of name interpreters affective, role
  of position, specific exchange.
• Other instruments for measuring properties
  different than name generator/interpreter
  sequence net range (basis of contacts with
  categories of people), distance scale (desired
  relations to social categoriesoccupations or
  ethnoreligious groups). Interorganizational
  studies use intensity of relations, inflows and
  outflows of information, resources, support,
  joint activities, or indicators of dyadic
  interorganizational relationships such as domain
  similarity, resource dependence, communication,
  and formalization. The main problem in these
  types of measurements is the respondent selection
  in terms of individuals reporting on
  organizations.

11

• Archives
• This type o source is inexpensive and useful for
  longitudinal, past or interlocking directorate
  studies.
• Records describing roasters, list of membership,
  lists of participants on committees, events or
  meetings, etc.
• Co-citation is also useful.
• Archives are useful too for study international
  and interurban networks (world systems, trading,
  treaties, diplomatic exchanges, etc).
• Little has been done in archival network data
  collection methods but there are some pieces for
  triangulation studies showing how co-citation or
  shared affiliations are good interaction
  indicators.
• Other data sources
• Observational data increases naturalness and
  descriptive accuracy but it is costly,
  time-consuming and restrictive to small
  populations.
• Few studies have used experimental data
  collection using small-world technique.

12

• Discussion
• Enumerating networks and the informant accuracy
  issue
• Comparing survey responses to known standard
• Reciprocation of survey responses
• Test-retest studies
• Quality of name interpreter items
• Reports on attributes of alters cited
• Reports on properties of relationships
• Recent developments for basic indices and
  measures
• Network size
• Network density
• Centrality and centralization
• Tie strength
• Network range

13

• Enumerating networks and the informant accuracy
  issue
• There is an objective existence beyond respondent
  cognition.
• The accuracy or reliability of self-reported
  information about a respondents network ties can
  be assessed by
• Comparing responses to an observed or known
  standard.
• Interviews with alters cited.
• Stability of responses over time studies.
• Use of multiple indicators checking name of
  alters mentioned in different name generators.
  But its difficult because there may be different
  kinds of relationships for a common link.

14

• Comparing survey responses to known standard
• Bernard, Killworth, and Sailer (1976, 1977, 1981)
  BKS strategy comparing different network data
  sources.
• There is a low correspondence of descriptive
  accuracy among different data sources
  questionnaires, diaries, monitoring, radio
  communication, observers, electronic monitoring.
  The reports are significantly different among
  sources.
• Recalling may be low because respondents usually
  participate in various networks and not only that
  one under study. In this matter, reliability is
  concern about the correlation between responses
  and observations (more than 0.8).
• Observational data vs self-reports data discuss
  issues about the appropriateness of data
  collection in terms of interpretative or
  subjective viewpoints.
• Shifting from identifying the gap between
  recalling and observing data to understanding
  these two measurements. The difference is the
  purpose of the study measure routinized ties or
  time-bound transactions.
• Systematic difference in accuracy among
  informants. Freeman and Rommey (1987) found that
  informant errors will be biased toward routine,
  typical structure. People are incapable of
  reporting accurately on very specific time frames
  than recall typical and recurrent social
  relations.
• Some studies found that recognition methods yield
  substantially larger estimates of size than do
  recall methods. Others found that they have
  similar estimates on average. However, question
  order becomes an issue to improve accuracy when
  respondents were asked first about their close
  ties instead of about their acquaintance.

15

• Reciprocation of survey responses
• Multi-sources is difficult to conduct in large
  populations.
• One alternative is to assess a rate of
  reciprocation between respondents and the group
  of alters first-cited. There is a correspondence
  between close or intense ties and higher
  reciprocity ties.
• However, reciprocation differs in forms and
  rates. Failures to assess reasons of
  reciprocation would bring inaccuracy or
  unreliability in the data or genuine asymmetry in
  the relationships under study.
• What is the cause of failure of reciprocation or
  asymmetry? Definitional or design considerations,
  the name generator/interpreter sequence, the
  structural characteristics of the network like
  density, or the nature of the study.
• This problems suggest a consideration of
  measurement errors.
• Test-retest studies
• Interpreting the results of over-time studies of
  network measures.
• Stability tests are more suitable for routinized
  and intense rather than time-bound transactions.
• There is more turnover or low stability rate in
  larger networks.
• There is greater stability in relations among
  respondents and the first-cited zone of alters in
  their listing than in the second.

16

• Quality of name interpreter items
• The quality of respondents reports on attributes
  of alters or properties of their relationships to
  alters, given that they are correctly enumerated.
• Reports on attributes of alters cited
• Demographic characteristics can be reported with
  substantially greater accuracy than attitudes.
• However, there are some reports not too accurate
  such as family income, use of contraception,
  desired family size, sexual attitudes, political
  party preference.
• Reports on properties of relationships
• Correspondence between respondents and alters on
  descriptions of the relationship between them
  (agreement rather than accuracy is the standard).
• Agreement was less common for less close ties.
• Hammer (1984) found very high concordance on
  frequency of contact, duration, kinship, and
  intensity of relationship.
• Happier couples had higher concordance than
  unhappy ones, except for topics of relatives.
• Recent developments for basic indices and
  measures
• Recent network studies are focused more on
  structural measures.

17

• Network size
• Definition the number of direct ties involving
  individual units.
• Self-reports tend to understate network size.
• Network density
• Definition the mean strength of connections
  among units in a network or (for dichotomous
  measurements) the proportion of links present
  relative to those possible.
• Using density is a problematic index of
  structural cohesion if the network has subgroups
• Making comparisons of density across networks
  that differ in size is misleading.

18

• Centrality and centralization
• Several definitions or modalities of centrality
• Freeman (1979) centrality measures for
  dichotomous network data
• Degree-based measures (network size) focus on
  levels of communication activity.
• Betweenness measures stress control or the
  capacity to interrupt communication.
• Closeness measures reflect freedom from the
  control of others.
• Stephenson and Zelen (1989) centrality measures
  based on information. Using direct and indirect
  ties between pairs of units.
• Bonacich (1972) centrality measures using valued
  measurement.
• Knoke and Burt (1983) prominence measures for
  centrality or prominence in affiliated units.
• Mizruchi (1986) difference of positional
  centrality between hub locations and bridge
  locations, and centrality over time.
• Mizruchi and Bunting (1981) sensitivity to
  directionality and differential tie strength are
  closer to historical accounts.
• Mariolis and Jones (1982) directionality
  presents higher reliability and stability (in
  time) coefficients for centrality measures.
• Freeman (1979) shows that there is a
  corresponding network-level of centralization for
  each centrality measure.

19

• Tie strength
• Multiple-item indices or measures of properties
  of individual dyads
• Scales for love, commitment, and investment in
  close relationships.
• 69-item concerning a single tie (factors of
  empathy, congruence, etc.).
• Its difficult for respondents to complete long
  batteries of items.
• Examining correlation of name interpreters
  closeness (or intensity), frequency, and
  duration.
• Closeness is one of the best indicators for
  strength of ties. Duration tends to overstate
  strength of relations. Frequency was weakly
  associated with closeness and duration.
• For job environment searches, we have intimacy,
  formality, and leisure as measures of strength of
  ties. Closeness, duration and frequency are
  highly correlated with intimacy.
• Network range
• Definition the extent to which a units network
  links it to diverse other links.
• Range can be measured by network size and network
  density (less dense networks having higher
  range).
• Some other indicators
• Indices of diversity in the characteristics of
  alters units.
• Positional similarity sensitizes the strength of
  links between focal units and alters.

20
Recent Developments in Network Measurements

• Peter V. Marsden (2001)

21

• Network Study Designs
• Whole-network
• Egocentric (ego-alters)
• Set-theoretic, graph-theoretic, and in matrix
  terms.
• Minimal network
• Sociomatrix
• One-mode and two-mode
• Longitudinal networks
• Cognitive social structure (CSS)
• Modalities and multi-methods
• Setting Network Boundaries
• Doreian and Woodard (1992) expanding selection
  by adding subjects or units to a list.
• Survey and Questionnaire Methods
• Batchelder (1989) comparing network data of
  different scales types (dichotomous, ordinal,
  interval, ratio, absolute) and the inferences
  about network-level properties (reciprocation,
  presence of cliques, etc.). This may affect
  dichotomous measurements.
• Feld and Carter (2002) referred to this as
  expansiveness bias

22

• Name generator instruments for egocentric
  networks
• McCarty et al. (1997) sought to measure features
  of total personal networks.
• Name generators instruments differ from
  conventional survey because interviewers can
  assist respondents (probing and prompting).
• Little research has examined differences in data
  quality collection mode.
• Examining the in-practice performance of
  instruments.
• Comparing name generators
• Varying name generator influences egocentric
  network size.
• Multiple-generator exchange-based instruments
  produce appreciably larger networks
• Predictability of network size and composition by
  using battery of name-generators.
• Recall, recognition, an forgetting
• Brewer (2000) compared free recall list of
  persons and then to supplement their list after
  consulting an inventory listing to include
  forgotten subjects. Likelihood of forgetting.
• Brewer makes clear it clear that name generators
  elicit only a fraction of those persons having a
  relation to a respondent. Intimate name
  generators have larger fraction of elegible
  alters than do weaker ones.
• Brewer and Webster (1999) reported high
  correlation between measures of centrality,
  egocentric network size, local density and
  recalled alters.
• Brewer (2000) suggested several steps to reduce
  the level of forgetting.

23

• Test-retest studies
• Brewer (2000) reviewed test-retest studies.
• Analyze stability and time effect on network
  size.
• Patterns in the free recall of persons
• Fiske (1995) and Brewer (1985) analyze patterns
  of free recall in terms of similarity of
  individual features, social proximity and
  attributes.
• Order nominations of alters by ties strength.
• The meaning and interpretation of name generators
• Bailey and Marsden (1999), Straits (2000)
  investigated how respondents interpret name
  generators.
• McCarty (1995) investigated respondent judgments
  of how well they know others by using
  indicators of the strength closeness, duration,
  friendship, kinshipwere associated with knowing
  alters well.
• Interview context effects
• Bailey and Marsden (1999) assessed the influences
  of context of respondents on the interpretation
  of a name generator.
• Interviewer effects
• Van Tilburg (1998), Marsden (2003) and Straits
  (2000) probed on the effects of interviewer on
  the egocentric network size and the use of name
  generator methods.
• Name interpreters
• Comparing self-reports and proxy reports.

24

• Additional instruments for egocentric networks
• Instruments for measuring extensive network size
• Summation method (McCarty, 2001) for global
  network by asking respondents sixteen
  relationships (ex family, friendship,
  neighboring, etc.) to estimate the total size of
  the network.
• Killworth (1998) developed scale-up methods to
  integrate subpopulations sizes to estimate
  population size.
• Reverse small world (RSW) is another approach to
  measuring extensive networks
• Position generators
• Measures linkages to specific locations directly.
  It asks respondents whether they have
  relationships with persons in each of a set of
  social positions (usually measure range and
  composition). Some examples
• Lin, Fu, and Hsung (2001) asked among relatives
  who hold fifteen different occupations.
• Erickson (1996) use position generators to
  identify weak and strong contacts.
• Smith (2002) found that position generators
  instruments have larger impact on strong
  relations like friendship than name generators
  and global generators.

25

• The resource generator
• Var der Gaar and Snijders (2004) proposed
  resource generators as instruments for measuring
  individual-level social capital (resources owned
  by a member of a individuals personal social
  network). Ex repair vehicles, knowledge of
  literature, high income, etc.
• Cognitive Social Structure Data (CSS)
• Definition It consists of judgments by each of
  several perceivers about each dyadic relationship
  in a whole network.
• Three designs
• A single observers slice of judgment,
• A locally aggregated structure of judgments by
  the two actors directly involved in each dyad.
• A consensus structure based on all judgments
  about a given dyad.
• Informant biases in network perception
• Informant reporting is not in vacuum. There are
  patterns or schemes on which informants draw when
  describing their social environments.
• Johnson and Orbach (2002) found some ego bias
  when we locate ourselves in central location of
  the network.
• Freeman (1992), and Kumbasar et al. (1994)
  pointed out that peripheral informants, adjacent
  alters, or some informants simplify observations
  of interactions, imposing a group or balance
  schema by selectively creating or neglecting
  relations among alters.

26

• Informant accuracy and competence
• Bernard, Killworth and Sailer (BKS 1981)
  questioned the utility of self-reported network
  data.
• BKS questioned the validity of respondent reports
  on social ties and lack of correspondence between
  survey reports of interaction frequencies and
  contemporaneous observations.
• Correspondence between reports and observations
• Freeman and Webster (1994) compared cognitive
  data vs observations of interactions. They found
  correspondence with more clustering effect in
  observational data than in cognitive data.
• Corman and Bradford (1993) compared real
  interaction and recalling correspondence. They
  found that participants tend to omit units in
  self-reports.
• Kashy and Kenny (1990) found a limited
  correspondence between cognitive data between
  observations and self-reports.
• Studies of informant competence
• Informant accuracy is the investigation of
  variation of informants in reporting on a whole
  network.
• Bondonio (1998) pointed to proximity as a source
  of competence informants were more competent in
  reporting on the networks of close alters than
  distal alters.

27

• Prospective uses of informants
• Key informants as third-parties that are not part
  of the network (neither alters) provide valid
  information about the relations of actors, dyads,
  subgroups or whole network.
• Archival network data
• This research modality has been used more
  frequently.
• The validity of archival data rests on the
  correspondence between measured connections and
  the conceptual ties of research interest.
• Observation
• Difficulty of coding (Who was addressing whom?)
• Large populations participating in one event
  require multiple observers.
• What can be called as a relation and what not.

28
The Social Structural Basis of the Organization
of Persons in Memory

• Devon D. Brewer (1995)

29

• Purpose
• Studying the recall of persons in socially
  bounded communities to explore the social
  structural influences on memory organization and
  whether individuals in such communities share
  common cognitive structure of community members.
• Nature of the study Empirical
• Integration
• Main issue The relationship between cognition
  and social structure by focusing on the social
  structural influences on how people remember and
  think about persons (the organization of persons
  in memory).
• Conceptual framework
• Evidence indicates that when subjects free-list
  acquaintances they tend to cluster, or mention
  successively, persons from the same social
  context (family, work, school, church, etc.).
  This demonstrates quite clearly that, at a very
  general level, persons are organized in memory
  according to social structural principles.
• Types of response patterning association,
  frequency, and serial order.
• Four possible general structures social
  relational structures (kinship or social
  interactions), persons individual
  characteristics (gender, ethnicity/race, or
  personality), persons spatial/geographic
  locations,
• Alphabetic/acoustic similarities of persons
  names.

30

• Research method of three study cases
• Monte Carlo tests with low p mean that subjects
  tend to recall persons who were in the same
  cohort adjancently.
• Inter-response times (IRTs) short IRTs signify
  strong connections between persons in memory
  where as longer IRTs signify weaker connections.
• Graphical analysis of two-dimensional
  correspondence analysis Map of subjects
  associative patterns.
• Graphical analysis of maximum link hierarchical
  clustering (0minimum, 1maximum).
• Graphical analysis using scatterplot between IRTs
  and associative factor for pairs of persons.

31

• CASE 1
• Graduate Academic Program
• 15 graduate students recalled the names of their
  fellow students.
• They clustered by cohort (year in the program).
• Monte Carlo tests p.0008 (range from lt.0001 to
  .2072 p for 12 subjects lt.05).
• In this case IRTs were shorter for persons who
  were in the same cohort than those in different
  cohort.
• Graphic map of subjects associative patterns
  exhibits a cohort structure.
• Results Frequency and serial order patterns were
  related to the cohort structure.

32

• CASE 2
• Taiwanese and Taiwanese-American Religious
  Fellowship
• 25 members of a church in California recalled the
  names of religious fellowship members.
• They clustered by perceived social proximity as
  the general associative factor in subjects
  recalls.
• Monte Carlo tests p.0001 (range from lt.0001 to
  .3380 p for 22 subjects lt.05).
• In this case IRTs were shorter for persons who
  were socially closer or social proximity (in
  terms of subjects perception of social
  interaction) than those socially distant.
• Graphic of maximum link hierarchical clustering
  exhibits a social proximity clustered structure.
• Scatterplot of a log transformation for IRTs and
  social proximity of pairs of persons shows that
  IRTs are very short for persons with closer
  social proximity with each other than those
  distant.
• Results Frequency and serial order patterns were
  related to the social proximity structure but two
  cases showed alphabetical patterns related to the
  organization of memory of persons. However,
  serial order alone is not significant to explain
  clustering by social proximity. More social
  network centrality for socially closer subjects.

33

• CASE 3
• Department in a Formal Organization
• 13 members of a department of public affairs
  division of a university in the southwestern of
  U.S. recalled the names of co-wokers.
• They clustered by perceived social proximity as
  the general associative factor in subjects
  recalls.
• Monte Carlo tests p.0009 (range from lt.0001 to
  .0148 p for 22 subjects lt.05).
• In this case IRTs were shorter for persons who
  work closely or work proximity (in terms of
  subjects perception of departmental employees at
  work but not status interactions that could not
  account for this work proximity clustering, such
  as hierarchy, work sections within department,
  spatial locations of persons offices) than those
  organizationally distant.
• Graphic of two dimensional depicts a common
  structure between the work proximity and
  adjacency in recall. The lines connect each
  persons position for work proximity to the
  subjects position. The lines tend to be short.
• Graphic of maximum link hierarchical clustering
  exhibits a work proximity clustered structure.
• Scatterplot of a log transformation for IRTs and
  work proximity of pairs of persons shows that
  IRTs are very short for persons with closer work
  proximity than distant positions.
• Results Frequency patterns were related to the
  work proximity structure in terms of more
  physical presence. Serial order patterns were
  related with work proximity and higher status
  than those with less status and worked less
  closely. The status serial order pattern did not
  account for highly significant clustering by work
  proximity observed in subjects recalls.

34

• CASE 3
• Department in a Formal Organization
• The author test-retest reliability (stability)
  with a reinterviews 2-3 weeks after the first
  recall. The authors also test reliability using
  two groups of gathering data free recall and
  alphabetically directed subjects, both with very
  similar results in terms of associative,
  frequency, serial order patterns to those
  observed in the first interviews.
• Summary of results
• The underlying cognitive structure of persons in
  a community is based on the communitys social
  structure (as revealed by subjects associative
  patterns).
• The salience (in memory) of persons in a
  community is related to a small number of
  variables, including proximity to a subject in
  the communitys social structure, visibility, and
  status (as indicated by subjects frequency and
  serial order patterns).
• Patterns in the recall of persons are uniform
  across subjects and subjects exhibit reliable
  recall patterns over time (as revealed by log
  transformations and graphical analysis of IRTs
  and reliability test-retests).
• The associative patterns in all studies provide
  systematic confirmation of the existence of
  clustering persons relationships.

35

• Summary of results
• Why interaction patterns serve as the basis for
  the cognitive structure of persons in a socially
  bounded community?
• Persons inter-act each other every day
  (spatiotemporal contiguity).
• Ability to perceive information about social
  networks/affiliations among persons offers a
  competitive advantage (networks knowledge or
  evolutionary perspective).
• Ability to perceive information about social
  networks/affiliations among persons can be
  measured by recalling persons relations, ties,
  names, characteristics, localtions,
  alphabetic/acoustic similarities, etc.
• But persons organize this information within
  clusters and subclusters of persons information
  from the same social context and temporal
  patterns.
• Visibility and status (as measures of centrality
  in the social network) and their relationship
  with power. Centrality is positively related to
  perceived power in a wide variety of settings.
• The person dominance (status and centrality in
  the hierarchy) is just essential for affiliation
  patterns. The evolutionary importance of
  hierarchy and status as a cognitive manifestation
  that also offers competitive skills.

36

• Other research
• People retrieve names quicker by referring social
  context than by personality.
• When there is no social context reference, people
  retrieve names by personal attributes (age,
  gender, occupation, etc).
• Persons status and centrality (dominance) and
  social closeness to an individual subject
  (affiliation) determine the nature of the degree
  of cognitive abilities.