Group Membership and Diffusion in Virtual WorldsDavid A. Huffaker, Chun-Yuen Teng, Matthew P. Simmons, Liuling Gong, Lada A. Adamic

School of Information, University of Michigan, Ann Arbor, MIEmail: huffaker@google.com, (chunyuen,mpsimmon,llgong,ladamic)@umich.edu

Abstract—Virtual goods continue to emerge in online commu-

nities, offering scholars an opportunity to understand how social

networks can facilitate the diffusion of innovations. We examine

the social ties for over one million user-to-user virtual goods

transfers in Second Life, a popular 3D virtual world, and the

unique role that groups play in the diffusion of virtual goods.

The results show that individuals – especially early adopters –

are more likely to adopt a virtual good when they belong to

the same groups as previous adopters. We also find that groups

exhibit bursty adoption, in which many individuals adopt in short

succession. In addition, we show that adoption activity within a

group depends on the group’s size and interactivity. Our work

provides insights into theories of social influence and homophily.

I. INTRODUCTION

Virtual goods continue to become incorporated into socialnetwork services, gaming platforms and virtual worlds, allow-ing users to enhance their online experiences or customizetheir online identities. These virtual items can be purchasedfor a small payment or for free, and often manifest ascommodities (e.g., fashion, brand name or celebrity gear) orgame accessories (e.g., a World of Warcraft shield or a wagonin Zynga’s Farmville) [1], [2]. Virtual goods represent a wayto measure information being transmitted over social networksand provide broader insights into our understanding of socialinfluence and diffusion.

While previous research in communication studies hasdemonstrated the important role that social ties play in thespread of information [3], little empirical work has been ableto test these propositions on a large-scale. More recent researchon diffusion has highlighted the intricate and intertwined rolesof social influence (i.e., a friend or opinion leader impactsone’s propensity to adopt an idea or innovation) and homophily(i.e., sharing similar interests with others increase one’s expo-sure to new innovations and propensity to adopt) [4], but itis unclear which variables best measure these complex socialbehaviors.

This research examines the role of social influence and ho-mophily in large-scale virtual goods adoption with a particularfocus on the role that groups play in the process of diffusion. Itis motivated by the following research questions: What socialfactors increase the likelihood that a user will adopt a virtualgood? What role does group membership play in the adoptionof virtual goods, and to what extent do group characteristicsimpact adoption?

To answer these questions, we examine how virtual goodsspread along social networks using a unique data set of onlineinteractions in a popular 3D immersive virtual world called

Second Life (SL). SL allows users to create and distributevirtual goods, and uses an in-game currency that can bepurchased with real money. Because 95% of the content inSL is actually created by the users, and because an astoundingnumber of paid and zero-cost transactions occur in SL [5], itremains an ideal venue for studying why some virtual goodsare adopted by users while others fail to diffuse. Second Lifealso allows us to examine the unique role of groups in theprocess of diffusion. Users can join up to 25 formal groupsbased on common interests or identities. SL users can sendmessages to other group members, hold events and meetings,and share information about products and locations inside thevirtual world.

II. THEORETICAL FRAMEWORK

Two popular theoretical frameworks have been used toexplain diffusion: (a) theories of social influence ; and (b)theories of homophily. Theories of social influence suggestthat exposure to influential individuals increases the likelihoodof adopting similar beliefs or attitudes [6]. Friends, peers,family, or more distant leaders, cultural figures or celebritiesall have the potential to impact another person’s attitudesor behaviors. Examples in recent research include evidencethat obese friends or partners can increase one’s own weightgain [7] or that local social networks can influence othersto quit smoking [8]. Research in online settings also showsupport for theories of social influence: digital informationtends to spread within the same social circles [9]; adoptionrates increase as friends begin adopting an item [10], [11];neighbors provide social reinforcement of a particular healthbehavior [12].

Theories of homophily posit that individuals seek out otherswho follow the same interests and self-categorization (e.g.,auto restoration hobbyists, punk rockers), or belong to thesame formal or informal groups (e.g., ethnicity, IT work-ers) [6]. Homophily not only increases exposure to ideasor innovations as one coalesces with the group, but groupmembers are more likely to adopt because they already havea propensity towards the attitude, behavior or idea. In otherwords, it is not that hanging with the wrong crowd willincrease bad behavior, but that individuals with a propensityfor bad behavior might gravitate toward the wrong crowd [13].Recent online studies lend support for theories of homophilyand diffusion: a study of mobile app adoption finds thathomophily explains more than half of diffusion [14]; a studyof email product recommendations finds that the most popularitems represent niche interests, suggesting that there is a low

GROUP CGROUP A

GROUP B

Fig. 1. Conceptual model of how diffusion spreads via social influence andhomophily. The origin of the virtual good (in black) influences the adoptionof several friends, one of whom is able to spread the asset further because ofshared interest among fellow group members.

probability of crossing over to individuals who do not sharethat particular interest [15].

These theories are not independent. Scholars argue that ho-mophily selection results in correlated attitudes and behaviors,making it difficult to determine a causal relationship betweensocial influence and diffusion [4]. Some even argue that socialinfluence and homophily are so highly intercorrelated thatthey can not be distinguished from one another [16]. Thisis not unique to offline settings; a study of social influencein Wikipedia shows that users become rapidly similar aftertheir first online communications, and this keeps increasingover time, making it difficult to tease social influence andhomophily apart [17].

This research intends to provide further insights into theways in which we can disentangle social influence and ho-mophily in online interactions. In Second Life, users candistribute virtual goods to one another, communicate througha chat system, add each other to a friend list, or join up to25 formal groups based on interest or self-categorization. Thisallows us to study the various types of interaction that usersengage in, and measure the extent to which diffusion is relatedto friendship or shared group membership. We argue thatwhile social ties impact adoption, being a member of a groupincreases the diffusion of virtual goods. This is illustrated inFigure 1: the node in black shares a virtual good to severalfriends, but the asset continues to diffuse because nodes ingray are in groups based on shared interests. This even allowsthe asset to spread across multiple groups with a potentialboundary. Since Group C is focused on a completely differenttopic, the asset does not penetrate.

We also argue that not all groups share the same commu-nication patterns or structural signatures, and it is likely thatsome attributes such as the size and communication patternsshould impact diffusion as well. For example, participationequality and dynamism, group size and diversity, maturityand cohesiveness all contribute to seeking and sharing in-formation [18]. The structural signatures necessary for thehealth and stability of large and small groups also differ, in

which small groups can rely on strong connections of a fewindividuals while large groups require continuous changes inmembership [19]. Our secondary contribution is to providemore insight into the relationship between group-level socialstructure and diffusion.

III. METHODOLOGY

A. DataOur data includes all virtual goods adoptions that took place

in SL between November and December 2008 for two typesof assets, provided the user still held the asset at the timeof data collection. The first, landmarks, are bookmarks thatavatars can use to visit a specific location within the virtualworld. There may be multiple landmarks with different anno-tations that correspond to identical coordinates, facilitating thetracking of diffusion of information about a location throughdifferent initial landmark instances. The second kind of assetis a gesture, which allows the avatar to execute a sequenceof motions, or to emit a particular sound. Landmarks andgestures were selected because they are two types of non-divisible virtual goods that can easily be traced.

In addition, we have weekly snapshots of the social network(buddy) graph for SL users, as well as weekly updates on theirgroup affiliations for the same Nov.-Dec. 2008 time period. Wealso have data on fee transactions that have been described inprevious studies [5], [10], [20], but which here are used onlyto determine whether a user is active in SL during the periodin question. Data on individual users includes informationon their experience and activity level: when they joined SL,how many days they were active and how much revenue theygenerated.

B. SampleOur goal is to be able to predict which users will adopt

which asset, given certain types of information such as thenumber of their friends who had previously adopted a particu-lar asset, or the number of groups the user shares with previousadopters of the asset. At first it may seem difficult to determinewhich of hundreds of thousands of users will adopt one ofthe millions of assets. However, the problem becomes muchmore scoped once we know that a user’s friend has adoptedan asset. In previous work, we showed that a large portion ofasset transfers occurs between users who are friends in SL,and users are more likely to adopt after their friends do, evenif they do not obtain the asset from the friend directly [10].

We chose our first sample accordingly by examining onlyadopters and non-adopters who both have at least one adoptingfriend. This allows us control for a user’s knowing someonewho adopted while examining our variables of interest de-scribed below. In other words, for each asset we select oneadopter who had a friend subsequently adopted the same asset,and then record that adopting friend and as well as one friendwho did not adopt. We omit assets where no pair of adoptersare friends. In terms of asset adoptions, we see 1,092,094adoptions, which represent 546,047 unique assets that wereexchanged among 235,467 unique users.

1 5 10 50 100 500 1000

0.05

0.10

0.20

0.50

1.00

# of adopters per asset

P(X>=x)

gesturelandmark

Fig. 2. The distribution of number of adopters for two types of virtual goods.

As a control, we also construct a sample where neitheradopters nor non-adopters share an adopting friend. For this,we select one adopter per asset, and to balance the sample,we select one non-adopter who actively transacted during thetime that the asset was being adopted. This excludes userswho may have been non-adopters due to inactivity. However,because there is no record of adopting users subsequentlydiscarding the item. predictive accuracy is compromised bylack of complete information about all adoptions.

For our examination of group characteristics conducive todiffusion, we analyze the network characteristics of 61,722unique groups. We limit our analysis to groups with ten (10)or more members, and for virtual goods with at least five (5)adopters.

From Figure 2, we can see that while a few assets enjoywidespread popularity, most were adopted by just a few indi-viduals. The distribution of gestures displays a step functionbecause avatars by default have a certain number of gesturesthat tend to remain unless discarded from inventory. It is alsointeresting to note that adopters tend to occupy a distance oftwo degrees in the buddy graph from the first adopter.

C. Variables of InterestOur dependent variables focus on the likelihood of adoption,

as well as total adoptions within a group. Our independentvariable selection captures many unique social behaviors in-cluding relationships with friends, groups and other adopters.We also control for how long users have been in SL, their pre-vious adoption behavior, and the adoption rates of the groupsthey inhabit. Because SL users can join up to 25 differentgroups, we develop an aggregated measure of group similaritybetween adopters and non-adopters. We also present a varietyof social network variables used to measure characteristics ofthese groups.

Asset Adoption. This dichotomous outcome variable repre-sents whether a unique asset has been adopted by a particular

user.Number of Days in SL: Total number of days a user has

spent logged into Second Life.Number of Previous Adoptions: Total number of other

virtual assets that the user has adopted in the past.Adopt Time: Ratio of the adopter’s position in the adoption

chain (i.e., first adopter, second adopter, nth adopter) dividedby the total number of adopters of the asset. For example, t =0.1 means that the user was among 1/10th of users to adopt.

Number of Previously Adopting Friends: Total number offriends that previously adopted the same virtual asset. Friendsare users the user has added to their buddy list.

Distance to First Adopter: Length of the shortest pathfrom the first adopter of an asset in SL to any other adoptionsthat occur afterwards. The first adopter is also likely to be thecreator.

Number of Friends Shared with Previous Adopter: In thesample where the paired adopter and non-adopter both sharethe same adopting friend, this is a proxy for the strength oftie with the adopting friend.

Number of Groups: Total number of SL groups a userbelongs to during the sampling frame.

Group Similarity: Cosine similarity between the groupmembership of an adopter and the groups of previous adopters.

Si,j =|Gi ·Gj |

� Gi �� Gj � (1)

We assign the following weight Wkto each group k:Wk = log[(total # users)/(# users in group k)]. When com-paring one user against all previous adopters, we take theaverage pairwise similarity between that user and all previousadopters.

Crowding Factor: Percentage of adopters in each of theuser’s groups. It is calculated as:

Crowdingfactor =�

k

Wk# adopters in Group k

# users in Group k(2)

Burstiness: Binary variable specifying whether a burst eventis detected in the asset adoption time series.

Group Adoptions: Total number of assets adopted by allusers in the group. The asset can be exchanged with any otheruser in SL.

Group Transfers: Total number of assets transferred be-tween group members.

Number of Group Members, N: Total number of usersthat have joined or were already members of a group duringthe one-month period.

Number of Edges E: Total Number of connections betweengroup members in the social graph or through transactions.

Average Degree: E/N.Clustering Coefficient: The number of closed triads as a

proportion of all connected triples. It represents how cliquishsocial ties are within a group.

Largest Connected Component (LCC): The LCC is thelargest subset of the network by which one can reach any nodefrom any other node within the subset by traversing edges.

Change in Size: Positive or negative change in group mem-bership between November and December 2008, normalizedby the size by the group membership in November.

Number of Active Users: Users in the group who activelycommunicated or traded during the one-month period.

The groups in this sample demonstrate a large amount ofvariability in the above characteristics. While 36% of groupshave at least 10 members, the top 1% of groups have 500or more members. There is a high rate of overall adoptionsby group members (M = 1531.89), as well as within-grouptransfers (M = 856). There is a high degree of clustering(M = .29), a medium average degree (M = 5.6), and a largeconnected component (M = .51). These groups also show highvariability in their membership turnover, ranging from losingtwo members in a month to gaining 4,000 members.

IV. RESULTS

A. Predicting Virtual Goods AdoptionIn order to predict adoption, we include attributes of the

individual users, their interactions with other adopters, andadoption activity within the groups to which they belong. Werely on two data sets to build our models. In the first test dataset, we select two users for each asset who have a friend thathas previously adopted it. One user adopts the asset after thefriend, and the other user does not. This allows us to utilize alogistic model where asset adoption serves as a binary outcomevariable.

As shown in Table I, all the variables in the logisticregression are significant. We note that because our sampleis very large, some of the variables — while significant —do not explain much of the variance. In order to examinethe individual impact of each variable on the model, werely on cross-validation measures [21]. Cross-validation (CV)randomly selects ten “folds” in the data, removes each fold andre-fits the regression model repeatedly, in order to estimatehow well the predictive model performs. Values above 0.5suggest that the variable(s) is predicting better than a randomguess, i.e., that everyone in our balanced data set would have50% chance of adopting.

The model shows that group similarity (CV = 0.63) andcrowding factor (0.61) are the most predictive individualvariables in determining if an adopter’s friends will adopt.The group similarity measure shows that the more groups thata potential adopter shares with those who previously adopted,the higher the likelihood of adoption (Odds Ratio = 1.26). Thecrowding factor also reveals that the more adoption activityfound in a users’ groups, the more prone its members are toadopt (Odds Ratio = 1.11).

The next three most predictive variables represent one’spropensity to adopt virtual goods in general, how embeddedan individual is in a particular social circle with the previousadopter, and the social distance from the first adopter. Thenumber of friends that are shared with the previous adopter,a proxy for tie strength, shows a positive impact on adoption(Odds Ratio = 1.40). The more items a user has adopted inthe past, the more likely she is to adopt in the future (Odds

Ratio = 1.62). The further removed a user is from the firstadopter, who is also potentially the creator, the less likelythat the former will adopt an asset (Odds Ratio = .93). Thelength of individual users have been in SL does not affect theirlikelihood of adoption (Odds Ratio = .99).

The number of adopting friends tells us whether the userhad more than just one friend who adopted before him. Inthe full model, having additional friends who have adoptedcorresponds to a lower likelihood of adopting (Odds Ratio= .64). However, in a simple logistic regression the numberof adopting friends is a positive predictor. One interpretationis that both the adopter and non-adopter in the test set havepotentially been exposed to the asset through the shared friendwho adopted. Once we account for the user’s potential interestin the asset through the group similarity variable, the numberof adopting friends actually signals that this user has resistedadoption through multiple potential exposures. Hence he isless likely to eventually adopt.

We then apply the same analysis to the second data set,where the sampled adopter and non-adopter do not sharea common adopting friend. We find qualitatively consistentresults with one important exception. As shown in Table II,the number of adopting friends variable has a much higherCV accuracy in the second sample (0.73 vs. the earlier 0.53),which is expected since the previous sample had guaranteed atleast one adopting friend, but this one does not. The variablescapturing the connection between groups and adoption alsohave a higher individual predictive ability, and the combinedmodel achieves a CV accuracy of 0.86.

We also measured specifically how much prediction im-proved when group variables are included. In our first sampleddataset, we find that the full model including group similarityand crowding factor shows roughly 5.5% greater accuracy(CV = 0.68) than the full model without group similarity (CV= 0.63). In our second sampled dataset, the accuracy of thefull model with group similarity and crowding factor is stillhigher (CV = .86) than the model without group similarity andcrowding factor (CV = 0.83).

TABLE IVARIABLES PREDICTING WHETHER FRIENDS OF AN ADOPTER WILL ADOPT

Estimate Std. Error z value CV acc.# Days in SL -0.01 0.00 -10.13 0.50# Adopting Friends -0.44 0.00 -140.46 0.53# of Groups -0.52 0.00 -239.48 0.53Distance from 1st Adopter -0.07 0.00 -17.72 0.56# other Adopted Assets 0.48 0.00 438.47 0.58# Friends with Prev Adopter 0.34 0.00 316.52 0.58Crowding Factor 0.10 0.00 225.86 0.61Group Similarity 0.23 0.00 341.87 0.63Combined model 0.68

In sum, the results suggest that group similarity and crowd-ing factor have a strong impact on the likelihood of adoption.To better illustrate this, we plotted the fitted regressions forboth group similarity and crowding factor, as well as thenumber of adopting friends. As Figure 3 shows, the impactof the number of adopting friends has a flatter slope than

TABLE IIVARIABLES PREDICTING ADOPTION OF A VIRTUAL GOOD WHERE

ADOPTERS AND NON-ADOPTERS DO NOT NECESSARILY SHARE ANADOPTING FRIEND

Estimate Std. Error z value CV acc.# Days in SL 0.01 0.00 6.04 0.50# of Groups -0.40 0.00 -113.94 0.66# Adopting Friends 3.11 0.02 169.93 0.73# Other Adopted Assets 1.04 0.00 424.24 0.74Distance from 1st Adopter -8.39 0.04 -238.68 0.75Crowding Factor 0.21 0.01 27.64 0.77Group Similarity 0.94 0.00 341.61 0.78Combined model 0.86

Fig. 3. Probability of adoption based on group similarity, crowding factorand number of adopting friends among adopters and non-adopters whohave a friend who previously adopted the virtual good. All predictors werestandardized to the same scale.

the impact of these two group-related variables. In the nextsection, we examine how these patterns differ among earlyand late adopters.

B. Differentiating Early vs. Late Adoption and Item Popularity

Groups may differ in their ability to explain early asopposed to late adoption. If groups serve as active distributorsof information about an item, they may be instrumental indispersing early word of a new available item, or they maygive late-comers an opportunity to catch up with what theyhad missed. On the other hand, even if groups are simply aproxy for user interests, it may be that early adopters tendto share more characteristics with one another relative tolate adopters [14]. To understand this further, we separate alladoptions into early, middle, and late. Given differing timespans over which different assets are adopted, we normalizethe time span from the first to last adoption event. Earlyadoptions occur in the first 1/5 of the time interval, lateadoptions in the last 1/4, and the rest are labeled as being

in the “middle”.Predicting whether someone adopts an asset might further

depend on the overall popularity of that asset. Therefore, weclassify the assets by the number of adopters into popular(i.e. more than 100 adopters), less popular (between 6 and12 adopters), and least popular assets (fewer than 6 adopters).As can be seen in Table III, when we try to predict whichof two friends of an adopter will themselves adopt, it isthe early adopters who are easiest to predict. However, it isthe late adopters who can most easily be differentiated froma non-adopter who does not necessarily have an adoptingfriend. We believe there are two contributing factors. The firstis that groups may be more instrumental in early adoption.However, having an adopting friend is also an important factorin adoption. Therefore, in the case where we control for havingan adopting friend, it is the early adopters who are morepredictable using information on group affiliation. By contrast,late adopters are more likely to have had a previously adoptingfriend, and are therefore more predictable when compared tonon-adopters who do not have adopting friends.

We also observe for both sampling methods that adoptionof the least popular assets is most predictable. It is the nicheitems that are most easily predicted by information on whetherone’s friends and groups are adopting.

TABLE IIICROSS-VALIDATION ACCURACY FOR PREDICTING ADOPTION BROKEN

DOWN BY ADOPTION TIME AND ASSET POPULARITY. F DESIGNATES THATBOTH THE SAMPLED ADOPTER AND NON-ADOPTER SHARE A FRIEND WHO

PREVIOUSLY ADOPTED; NF DESIGNATES THAT THE SAMPLED ADOPTERAND NON-ADOPTER DO NOT NECESSARILY SHARE AN ADOPTING FRIEND.

Adoption time Popular Less popular Least popularEarlyF 0.758 0.736 0.787MiddleF 0.682 0.714 0.726LateF 0.667 0.671 0.682EarlyNF 0.820 0.828 0.841MiddleNF 0.852 0.862 0.873LateNF 0.892 0.880 0.900

C. Exploring Asset Adoption Through Group Coverage

The regressions in the previous sections show a correspon-dence between sharing groups with adopters and being likelyto adopt. This effect did not merely reflect overall engagementwith the virtual world and its groups, since the number ofgroups to which a user belongs is not a significantly predictivevariable. Rather, adopters of the same assets are likely toshare the same specific groups. To illustrate this further, wecomputed a minimum number of groups required to “cover”all users who adopt an asset and who are also members of atleast one group.

Omitting users who are not members of any group reducesthe per-asset adoption count by less than 1% for 91% of assets.We then find a minimal subset of groups where the membersof those groups cover all adopters of the asset, according to thegreedy algorithm developed in [5]. We then create a baselineby calculating the number of groups it would take to cover anyrandom set of users who adopted any asset of a given size.

Number of users adopting asset (grouped only)

% o

f gro

up c

ontri

bute

d to

cov

erag

e fo

r lar

gest

gro

up0.

020.

040.

060.

080.

100.

12

! ! !

!

!

!

! ! ! ! ! ! ! ! ! !!

!

! !

!

!

!

!

!

!!

!

! !

!

!

!

!

!

!

10 100 1000 10000

!

!

Data Source! random set of adopters! adopters of same asset

Fig. 4. For each asset, we select the group that contains the most adopters ofthat asset. We then show what fraction of the group is comprised by the asset’sadopters. We compare this against equally sized random sets of adopters.

As can be seen in Figure 5, it takes on average a smallernumber of groups to cover users who are adopting the sameasset than to cover users who are adopting assets in general.This pattern, consistent with our regression results, suggeststhat the individuals who choose to adopt an asset are morelikely to share the same groups. Furthermore, as Figure 4shows, adopters can comprise a small but significant portionof an individual group’s membership. As expected, randomlychosen users do not comprise a significant portion of anygroup.

D. Group Membership and Adoption BurstsAn asset adoption time series gives the adoption frequency

of an asset over a large time scale. A burst in the timeseries suggests an abnormal behavior in the series where anunexpectedly large number of events, e.g. adoptions, occurwithin some time window. One might expect friends andmembers of the same group to adopt during a narrower timewindow than all adopters of an asset. We therefore ask whetherthose individuals involved in the bursts are more likely to belinked through friendship or group ties.

Here we select two methods of identifying such bursts. Thefirst, a fixed window burst detection, looks simply at whethera high proportion of all adoptions occurs within a short, 7-day time window. We identify those assets where 90% ofadoptions occur within a week, those where a maximum ofbetween 20% and 90% of adoptions occurred within any single7-day period, and assets with no such period of concentratedadoption. This definition can be applied regardless of the totalnumber of adoptions, although assets with fewer adoptions aremore likely to have a majority of those adoptions occur withina short time window.

The second method, elastic window burst detection usinga wavelet technique [22], must have a sufficient number ofobservations for a burst to be detected. We randomly sampledone fifth, or 11,398 such assets adopted at least 250 times,among which 6809 assets (or about 60%) contained a burstevent. The assets are then classified into three categoriesaccording their total number of adopters, i.e., 3,562 large assets

with ≥ 1000 adopters, 2,754 medium assets with between500 and 1000 adopters, and 5,082 small assets with < 500adopters. Figure 5 shows that the minimum required numberof groups to cover adopters is smaller for those assets whichare adopted in bursts. This implies that bursts of adoptiontend to occur within groups. This further suggests that groupsare either a rapid medium for the diffusion of information, orthat they in other ways correspond to pockets of susceptibleindividuals in the social graph through which information canspread rapidly.

Number of users who adopted asset (group members only)N

umbe

r of g

roup

s re

quire

d to

cov

er1

1010

010

0010

000

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●

●●

●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●

●●●●●●●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●●●

●●

●●●●

●●●●

●●

●●●●

●●●●

●●

●●●●●●

●●●●

●●

●●●●●●●●●●

●●●●●●●●●●●●

●●●●●●

●●●●●●●●●●●●

●●●●●●●●

●●

●●●●

●●

●●

●●●●●●

●●

●●

●●●●●●●●●●

●●●●

●●

●●●●

●●

●●●●●●

●●●●●●●●

●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●●●

●●●●

●●●●●●

●●●●

●●

●●

●●●●●●●●●●●●●●●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●

●●●●

●●●●

●●

●●●●●●●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●●●

●●●●

●●●●●●●●●●

●●

●●

●●●●

●●●●●●

●●●●

●●

●●

●●

●●●●●●●●●●●●

●●

●●●●

●●●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●

●●●●●●●●

●●

●●●●

●●

●●

●●●●●●●●●●●●●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●●●

●●●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●

●●●●●●

●●

●●●●●●

●●●●●●

●●

●●

●●

●●●●

●●●●●●●●●●

●●●●●●

●●

●●●●

●●●●

●●●●

●●●●

●●

●●

●●

●●●●

●●

●●●●●●●●●●

●●

●●●●●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●●●●●

●●

●●●●

●●●●●●

●●●●●●

●●●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●●●

●●

●●●●●●●●

●●●●●●

●●

●●●●

●●●●●●

●●

●●

●●

●●●●

●●●●

●●

●●●●●●●●

●●●●●●●●

●●●●

●●●●●●●●

●●●●

●●

●●

●●●●

●●●●

●●

●●●●●●●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●●●

●●●●

●●●●

●●

●●●●

●●●●

●●

●●

●●●●

●●●●●●

●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●●●●●

●●

●●

●●

●●●●

●●●●●●●●

●●

●●●●●●

●●●●●●

●●

●●●●

●●

●●●●●●

●●●●

●●

●●●●●●●●●●

●●●●●●

●●

●●●●

●●●●●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●●●

●●

●●

●●●●

●●

●●

●●●●●●●●●●

●●

●●

●●

●●●●●●●●●●●●

●●●●

●●

●●●●●●●●●●●●●●

●●

●●●●●●●●●●

●●

●●●●●●●●●●●●●●

●●

●●●●

●●

●●●●●●

●●

●●●●

●●●●●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●●●●●●●●●●●

●●●●

●●●●●●●●●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●●●●●

●●

●●●●

●●

●●●●

●●●●

●●●●

●●●●●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●●●●●

●●●●●●

●●

●●●●●●

●●●●

●●●●

●●●●

●●●●●●

●●

●●●●

●●

●●

●●●●

●●

●●●●●●

●●●●●●●●●●●●

●●●●●●●●

●●

●●

●●●●●●●●●●●●

●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●

●●●●●●

●●

●●●●

●●●●

●●

●●

●●

●●●●

●●●●●●●●

●●

●●●●

●●●●●●

●●●●

●●●●●●●●●●

●●●●

●●●●

●●

●●●●

●●

●●●●●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●●●

●●●●

●●●●

●●●●

●●●●●●●●●●

●●●●

●●

●●●●●●●●●●●●

●●●●●●●●

●●●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●●●●●

●●●●

●●

●●●●

●●

●●●●

●●●●●●●●●●●●

●●

●●●●●●●●

●●

●●●●●●

●●

●●

●●

●●

●●●●

●●●●●●●●

●●

●●

●●●●

●●

●●●●

●●●●●●●●

●●

●●●●●●

●●●●●●●●

●●●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●●●●●

●●●●

●●

●●

●●●●●●●●●●●●●●●●

●●

●●

●●●●●●●●●●

●●●●●●

●●●●●●●●

●●

●●

●●●●

●●

●●●●●●

●●

●●●●●●●●●●●●

●●

●●●●●●●●

●●●●●●

●●

●●●●●●

●●

●●●●●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●●●●●

●●

●●●●

●●

●●

●●●●

●●

●●●●

●●●●

●●

●●

●●

●●●●

●●●●●●

●●●●●●●●●●

●●●●●●●●

●●●●●●

●●

●●

●●●●●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●●●

●●●●

●●

●●●●

●●●●●●

●●

●●

●●●●●●

●●●●

●●

●●●●●●

●●●●

●●●●

●●

●●

●●

●●

●●●●

●●●●●●

●●●●●●●●

●●

●●

●●

●●●●

●●

●●●●●●●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●●●

●●●●●●●●●●

●●

●●

●●

●●●●

●●●●●●

●●

●●●●●●

●●●●●●

●●

●●

●●

●●

●●

●●●●

●●●●●●●●

●●

●●

●●

●●

●●●●●●

●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●●●●●●●●●

●●●●●●●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●●●●●●●

●●●●

●●●●

●●

●●

●●

●●●●●●●●

●●●●●●

●●

●●●●●●

●●

●●

●●●●

●●●●●●

●●

●●●●●●●●

●●●●●●●●

●●

●●●●

●●

●●●●●●●●●●●●●●●●

●●

●●●●

●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●●●

●●●●●●●●

●●

●●●●

●●

●●

●●●●

●●

●●●●●●●●●●

●●

●●●●●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●●●●●●●

●●●●

●●

●●●●

●●●●●●

●●●●

●●●●

●●

●●

●●●●●●●●

●●●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●●●

●●●●●●●●

●●

●●

●●●●

●●●●

●●●●

●●

●●

●●●●

●●●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●●●●●

●●

●●●●●●●●

●●

●●

●●●●●●●●●●

●●●●●●●●

●●

●●●●

●●●●●●●●●●●●●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●●●●●●●

●●●●●●●●●●

●●

●●

●●

●●●●●●

●●●●

●●

●●

●●

●●●●

●●

●●●●●●●●

●●●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●●●

●●●●

●●

●●●●●●

●●

●●

●●

●●●●●●●●●●

●●

●●●●●●●●

●●●●●●

●●

●●

●●

●●

●●●●●●●●

●●●●

●●

●●

●●

●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●●●●●●●

●●

●●

●●●●

●●

●●●●●●

●●●●●●●●

●●

●●●●●●

●●●●

●●

●●●●●●●●●●●●

●●●●

●●●●●●●●●●●●

●●

●●●●●●

●●

●●●●●●

●●

●●●●●●●●●●●●●●●●

●●

●●●●●●

●●●●

●●●●

●●●●

●●●●●●●●●●

●●●●●●

●●

●●

●●●●●●●●●●

●●

●●●●●●

●●

●●●●●●●●

●●

●●●●●●

●●●●

●●●●●●

●●●●

●●●●●●

●●●●

●●●●●●●●

●●

●●●●

●●●●●●

●●●●●●●●●●

●●●●

●●●●

●●

●●●●●●

●●

●●●●●●●●●●●●●●

●●

●●●●

●●

●●●●

●●●●●●

●●●●

●●

●●

●●

●●●●●●●●●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●●●●●

●●

●●●●

●●

●●●●

●●

●●●●

●●

●●●●

●●

●●●●●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●●●

●●●●●●●●

●●

●●●●●●

●●

●●●●●●

●●

●●●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●●●

●●●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●●●●●●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●●●

●●●●●●●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●●●●●●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●●●

●●●●

●●●●

●●

●●

●●

●●

●●●●●●

●●●●●●

●●

●●●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●●●

●●●●●●

●●

●●●●●●

●●

●●●●

●●

●●

●●●●●●●●●●

●●●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●●●●●

●●

●●

●●●●

●●●●

●●●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●●●●●●●

●●●●

●●

●●●●●●●●●●

●●

●●

●●

●●●●

●●●●●●●●

●●

●●

●●

●●

●●●●

●●

●●●●●●●●

●●

●●

●●●●●●●●●●

●●●●●●

●●

●●

●●●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●●●●●●●

●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●●●

●●●●

●●●●

●●

●●

●●●●

●●●●

●●●●

●●●●●●

●●

●●●●

●●

●●

●●●●●●

●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●●●●●

●●

●●●●●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●

●●

●●●●●●

●●

●●●●

●●●●●●

●●

●●●●●●

●●●●

●●

●●●●

●●●●

●●

●●●●

●●●●

●●●●

●●●●●●●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●●●●●●●

●●●●

●●

●●

●●

●●

●●●●●●

●●●●●●●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●●●

●●

●●●●●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●●●●●

●●

●●●●●●

●●●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●

●●●●

●●

●●●●●●

●●●●●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●●●

●●

●●●●●●

●●

●●

●●●●

●●●●

●●●●●●●●

●●

●●●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●●●

●●●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●

●●

●●

●●●●●●

●●●●

●●

●●●●

●●●●●●●●●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●●●●●

●●

●●

●●

●●●●

●●

●●

●●●●●●

●●●●

●●●●●●●●

●●

●●●●●●

●●●●●●

●●

●●●●●●

●●●●

●●

●●●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●●●●●

●●

●●●●

●●●●

●●

●●

●●●●●●

●●●●

●●●●

●●●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●●●●●●●

●●●●

●●●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●●●●●

●●

●●

●●●●

●●●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●●●●●

●●●●

●●

●●●●

●●●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●●●●●

●●●●

●●

●●

●●●●●●●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●●●

●●●●

●●

●●●●●●

●●●●●●●●

●●

●●●●

●●●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●●●●●●●●●●●●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●●●

●●●●●●●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●

●●

●●●●●●●●

●●●●

●●

●●

●●●●

●●●●

●●●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●●●●●●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●●●

●●

●●

●●

●●●●●●

●●●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●●●●●

●●

●●●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●●●●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●●●

●●●●●●●●●●

●●●●

●●●●●●

●●●●●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●●●●●

●●

●●

●●●●

●●●●

●●

●●●●

●●●●●●

●●●●

●●●●●●●●

●●●●

●●

●●●●

●●

●●

●●●●●●●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●●●

●●

●●

●●

●●●●

●●●●

●●●●

●●●●●●●●●●

●●

●●●●

●●

●●●●

●●●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●

●●●●●●●●●●

●●

●●●●●●

●●●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●●●●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●●●●●

●●

●●●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●●●●●●●●●

●●

●●●●

●●

●●●●

●●

●●●●●●

●●

●●●●●●

●●●●

●●●●

●●

●●●●●●

●●

●●●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●●●

●●

●●●●●●●●

●●●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●●●●●●●●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●●●●●

●●●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●●●

●●●●●●●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●●●

●●●●

●●●●●●●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●●●

●●

●●

●●●●

●●

●●●●●●●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●●●

●●

●●●●●●●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●●●●●

●●●●●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●●●

●●

●●

●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●●●

●●●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●●●●●

●●

●●

●●●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●●●●●

●●

●●

●●

●●●●

●●●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●●●

●●

●●●●

●●

●●

●●●●●●●●

●●

●●

●●●●●●

●●

●●

●●●●

●●

●●

●●●●

●●●●●●●●

●●●●●●

●●

●●

●●

●●●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●●●

●●

●●●●

●●●●

●●●●

●●

●●●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●●●●●

●●●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●●●●●

●●

●●

●●

●●●●●●●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●●●

●●

●●●●●●●●●●●●●●

●●

●●●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●●●

●●●●

●●●●●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●

●●

●●●●●●●●●●●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●●●

●●●●

●●●●

●●●●

●●●●

●●●●

●●●●●●

●●●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●●●

●●●●

●●

●●●●

●●

●●●●●●

●●●●●●

●●●●●●●●

●●

●●●●●●

●●

●●

●●●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●●●●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●●●

●●

●●●●●●●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●●●●●

●●●●●●

●●

●●

●●

●●●●

●●

●●●●●●●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●

●●

●●

●●●●

●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●

●●●●

●●●●●●●●●●●●

●●

●●●●●●●●

●●

●●

●●

●●●●●●●●●●●●●●

●●

●●●●●●

●●

●●●●●●●●

●●●●●●

●●●●

●●

●●●●

●●

●●●●

●●●●●●

●●●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●●●●●●●●●

●●

●●●●

●●●●

●●●●

●●●●

●●●●●●●●

●●

●●●●●●

●●

●●●●●●

●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●●●●●

●●

●●●●

●●

●●●●●●●●●●

●●●●

●●

●●●●●●●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●●● ●●

●●

●●

●●●●●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●

●●

●●●●

●●

●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●

●●●●●●

●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●

●●●●●●●●●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●●●●●●●●●●●●●

●●●●●●●●

●●●●

●●●●●●●●●●●●

●●

●●

●●

●●●●●●

●●●●

●●●●

●●

●●●●●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●●●

●●

●●

●●●●

●●

●●

●●●●●●

●●

●●●●

●●●●

●●●●

●●●●●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●

●●

●●●●●●

●●●●

●●

●●●●●●●●●●●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●●●●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●●●●●

●●●●

●●●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●●●●●

●●

●●

●●

●●

●●●●●●

●●

●●●●

●●●●

●●●●●●●●●●●●●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●●●●●●●●●

●●

●●

●●

●●

●●●●●●●●●●●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●

●●

●●●●●●●●●●

●●

●●●●

●●●●

●●●●

●●

●●●●●●

●●

●●●●●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●●●

●●

●●●●●●●●●●

●●●●

●●

●●

●●●●●●●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●

●●

●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●●●

●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●

●●

●●●●●●

●●

●●●●●●●●●●●●●●●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●●●●●●●

●●

●●

●●

●●●●●●●●●●

●●

●●

●●

●●

●●●●●●

●●

●●●●●●●●●●

●●

●●

●●●●

●●

●●

●●●●

●●●●

●●●●●●

●●

●●●●●●

●●●●●●

●●

●●

●●

●●

●●●●

●●

●●●●●●

●●●●

●●●●

●●

●●

●●●●

●●

●●●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●●●

●●

●●●●●●

●●

●●

●●

●●●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●●●●●●●

●●●●●●●●●●

●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●

●●●●

●●●●

●●●●

●●

●●

●●●●

●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●●●●●●●●●●●

●●

●●●●

●●●●

●●

●●●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●●●●●●●

●●

●●●●

●●

●●

●●●●

●●●●●●

●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●●●●●●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●

●●

●●●●

●●

●●●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●

●●●●

●●●●

●●●●

●●●●

●●●●●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●●●

●●●●●●

●●

●●

●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●●●

●●●●

●●●●●●●●●●●●●●

●●

●●●●

●●

●●

●●

●●●●

●●●●●●●●●●

●●

●●

●●

●●

●●

●●●●

●●●●

●●●●

●●●●●●●●

●●●●●●●●

●●

●●●●●●●●●●

●●●●

●●

●●

●●

●●●●●●●●●●

●●

●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●

●●

●●●●

●●

●●●●●●

●●

●●

●●●●●●●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●●●

●●

●●

●●●●●●●●

●●●●

●●●●

●●●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●

●●●●●●

●●

●●

●●

●●

●●●●●●

●●

●●●●

●●

●●

●●●●●●●●●●●●●●●●●●

●●●●●●

●●●●

●●

●●●●●●●●

●●

●●

●●

●●●●●●

●●

●●●●●●

●●

●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●

●●●●●●●●●●

●●

●●●●●●

●●●●●●●●●●●●●●●●

●●

●●●●●●

●●

●●

●●●●●●●●

●●●●●●

●●

●●●●●●

●●

●●●●●●

●●

●●●●●●

●●

●●

●●

●●●●●●

●●

●●●●●●

●●●●●●

●●

●●

●●●●

●●

●●

●●

●●●●

●●●●

●●

●●

●●

●●

●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●

●●●●

●●

●●

●●

●●

●●

●●

●●

●●●●●●●●●●

●●

●●●●

●●

●●●●●●●●●●●●●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●

●●

●●

●●

●●●●●●●●●●●●●●●●

●●

●●

●●●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●

●●

●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●●●●●●●●●●

●●●●●●

●●●●

●●●●●●●●

●●

●●

●●

●●

●●●●●●●●

●●

●●

●●

●●●●

●●●●

●●●●

●●

●●

●●●●●●●●●●●●

●●●●

●●

●●●●●●

●●

●●●●●●●●

●●●●

●●

●●

●●●●●●●●●●●●

●●

●●●●●●●●

●●●●●●

●●●●●●●●●●●●●●●●

●●●●●●●●●●●●●●●●●●

●●●●

●●●●

●●●●●●

●●●●

●●

●●●●●●●●

●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●

●●

●●●●●●●●●●●●●●●●●●

3−7 8−20 20−50 50−150 150−400 400−1.1k 1.1k−3k 3k−8k 8k+

dataset20% burst90% burstNo burstRandom set of adopters

Fig. 5. The number of groups required to cover all adopters of a particularasset, provided they are members of least one group. The number of groupsrequired to cover the adopters of an asset is smaller than the number thatwould be required to cover a randomly assembled group of the same size.

In addition to being more likely to share groups, adoptersof bursty assets are more likely to be friends. We quantifythis by constructing equivalent networks using a configurationmodel proposed by Chung and Lu [23].The model keeps thedegree of each node fixed, but randomly matches the endpointsof the edges. One can then calculate, given the number offriends each adopter has, the number of connections onewould expect between adopters. Table IV shows that friendshipconnections are more common relative to what is expectedfor assets being adopted in bursts than for those who don’texhibit bursty adoption. Furthermore, users who adopt in closesuccession, i.e., those who are part of the same burst event,are more likely to be friends (see Table V). Moreover, it isnotable that those who end up holding smaller (i.e. less-widelyadopted assets) are more likely to be friends. They may havebeen brought together by a unique interest, or influenced oneanother directly. This illustrates that social variables are moreimportant for diffusion in the niche.

E. Group Characteristics that Increase AdoptionSo far we have provided several pieces of evidence that

groups play an important role in the diffusion of virtualgoods. Yet groups tend to vary in their size, attrition ratesand interconnectedness. A second goal in our analysis isto determine what characteristics of groups, if any, impactdiffusion. In order to examine this, we performed a linearregression of the total virtual goods transfers among memberswithin a group, as well as a linear regression of the total

TABLE IVMULTIPLE COMPARISON OF THE ACTUAL-TO-EXPECTED RATIOS OF

CONNECTIONS BETWEEN ALL-ADOPTERS FOR BURST-ASSETS AND FORNON-BURST-ASSETS

Mean 99% Conf. intmean diff.

Large Assets Non-burst 7.0040 [-1.6326, 0.4308]Burst 6.4031

Medium Assets Non-burst 14.2801 [1.7275, 8.4618]Burst 19.3747

Small Assets Non-burst 23.0261 [1.8695, 16.9693]Burst-assets 32.4454

TABLE VMULTIPLE COMPARISON OF THE ACTUAL-TO-EXPECTED RATIOS OF

CONNECTIONS BETWEEN ALL-ADOPTERS AND BETWEENBURST-ADOPTERS FOR BURST-ASSETS

Mean 99% Conf. intmean diff.

Large Assets All-adopters 6.4031 [3.5348, 10.1310]Burst-adopters 13.2359

Medium Assets All-adopters 19.3747 [15.4171, 53.9804]Burst-adopters 54.0735

Small Assets All-adopters 32.4454 [35.5050, 146.2946]Burst-adopters 123.3453

adoptions by group members from anywhere in Second Life.As shown in Table VI, we find several group-level attributes tobe correlated with the transfer of virtual goods within groups.

In particular, the number of group members (β = −.06)is slightly negatively correlated with total transfers. A simpleexplanation could be that large groups are more easily joinedand therefore might be composed of users who are not asactive in SL. However, we fail to find a correlation betweenthe size of a group and the percentage of its members whoare active, as judged by last login. We also do not find thatchanges in the size of groups impact the volume of transfers.The number of active users (β = .11) does show a strongpositive relationship with transfers. This shows that it is theactive group size that matters.

Although the total number of social ties between membersfollows the trend of total number of members in beingnegatively correlated with transfers (β = −.04), the averagein-group degree per member, measuring how connected eachmember is within the group, is positively related with transfers(β = .17). Further measures of social connectedness, such asthe clustering coefficient (β = .21) and the strongly connectedcomponents (β = .24) are also positively correlated withtransfers. This suggests that groups that show signatures ofsocial interactions are more likely responsible for the spreadof assets.This is illustrated in Figure 6, in which low adoptinggroups (top row) show very few links between members,especially with regard to social ties (in gray).

When we examine all adoptions that are made by groupmembers, some of the group characteristics show a differentrelationship. First, as shown in Table VII, change in group sizehas a small but significant impact on total adoptions (β = .05).This suggests that as groups grow and membership changes,there are new opportunities for adoption. For example, a new

Fig. 6. Network visualizations of transaction ties (in black) and social ties(in gray) for lowest (top) and highest (bottom) frequency of adoption for arandom sample of groups with 20 members.

TABLE VIVARIABLES PREDICTING TOTAL WITHIN-GROUP TRANSFERS

Estimate Std. Error t value Pr(>|t|)(Intercept) -1.902 0.122 -15.601 0.000Num of Members -0.002 0.000 -10.106 0.000Num of Edges -0.000 0.000 -8.477 0.000Clustering Coefficient 12.275 0.234 52.549 0.000Average Degree 0.326 0.011 30.835 0.000LCC 8.924 0.189 47.224 0.000Change in Size -0.002 0.008 -0.252 0.801Num of Active Users 0.008 0.001 16.432 0.000

member of the group may come with new ideas or innovations,which are attractive to members who have been exposed tothe same items for a period of time. However, there maybe some internal boundaries: the results show that averagedegree has a negative impact on adoption (β = −.22), thatas the interconnectedness of group members increases, theless likely new innovations are to be adopted. For example,members may be accustomed to certain types of virtual goodsand uninterested in investing time or money in new ones, orcertain cultural practices or norms may be in place that lendthemselves to some types of innovation but not to others.

TABLE VIIVARIABLES PREDICTING TOTAL ADOPTIONS FROM GROUP MEMBERS

Estimate Std. Error t value Pr(>|t|)(Intercept) 5.048 0.162 31.206 0.000Num of Members -0.006 0.000 -27.383 0.000Num of Edges -0.000 0.000 -2.430 0.015Clustering Coefficient 9.860 0.310 31.809 0.000Average Degree -0.547 0.014 -38.994 0.000LCC 26.434 0.251 105.413 0.000Change in Size 0.133 0.011 12.063 0.000Num of Active Users 0.016 0.001 23.807 0.000

V. CONCLUSION

We explored the role of group membership and structurein the adoption of virtual goods in a large-scale online com-munity. Prior work had focused primarily on the influence ofsingle social ties, or at most the entirety of an individual’s ego

network. By contrast, we examine the role of larger socialcollectives, namely groups, and demonstrate that they areuseful in discerning whether a given individual will eventuallyadopt an asset. We show that when an individual belongs tomany of the same groups as other adopters, and conversely,when that individual’s groups are populated by adopters, thenthe individual is more likely to adopt. These findings areparticularly germane for assets that are adopted by a smallnumber of users. Individual groups are also more likely toplay a role in bursty adoption, when many individuals adoptin rapid succession, and for adoptions that occur shortly afterthe asset is first created.

From these findings we can infer that groups represent nicheinterests, and that assets may be more likely to diffuse initiallythrough a small group. Groups might also facilitate rapiddiffusion of an asset during a short time period, through one-to-many announcements, events, or other ways in which groupmembers interact. We gain further insight that group structureis conducive to diffusion. Smaller groups tend to show higheroverall adoption activity, as well as within-group transfers.Denser, more interconnected groups, are related to exchangeswithin the group, but not in terms of virtual goods adoptionamong group members and the rest of the community.

There are clear limitations to our approach. Group activity isreflective of both social ties and shared interests, two variablesthat contribute to adoption, highlighting the complex interplayand potentially confounding nature of social influence andhomophily [4], [16]. While we have demonstrated the utility offactoring in information about group membership to predictingadoption events, controlling for variables such as the numberof friends in one’s social circle who adopted, the present dataanalysis does not allow us to distinguish how much of theexplanatory power of group membership is due to groupsreflecting interests that would make an individual susceptibleto adoption, and how much is due to activities within a grouppromoting adoption. We think this is an interesting avenue forfuture work.

In conclusion, we have demonstrated the important role thatgroups and group membership play in the diffusion process,and provided evidence that the adoption of virtual goods inSecond Life is more likely to occur when users share similargroups. We also show that the communication behaviors andnetwork structure of these groups are tied to the number ofadoptions that occurs among group members. These findingsprovide further support that individual influence is not thesole factor in the diffusion process and that considering largersocial collectives such as formal groups can better explain thelikelihood of adopting an innovation.

ACKNOWLEDGMENT

We thank Linden Lab for sharing Second Life data. Thiswork was supported by the Intelligence Community (IC) Post-doctoral Research Fellowship Program, NSF IIS-0746646, andMURI award FA9550-08-1-0265 from the Air Force Office ofScientific Research. We also thank Robert Horowitz.

REFERENCES

[1] E. Castronova, D. Williams, C. Shen, R. Ratan, L. Xiong, Y. Huang, andB. Keegan, “As real as real? Macroeconomic behavior in a large-scalevirtual world,” New Media & Society, vol. 11, no. 5, pp. 685–707, 2009.

[2] T. Malaby, “Parlaying value: Capital in and beyond virtual worlds,”Games and Culture, vol. 1, no. 2, p. 141, 2006.

[3] E. Rogers, Diffusion of innovations. Free Pr, 1995.[4] M. Jackson, “An Overview of Social Networks and Economic Applica-

tions,” The Handbook of Social Economics., ed. Jess Benhabib, AlbertoBisin and Matthew O. Jackson. Elsevier Press. forthcoming, 2009.

[5] E. Bakshy, M. Simmons, D. Huffaker, C. Teng, and L. Adamic, “TheSocial Dynamics of Economic Activity in a Virtual World,” in FourthInternational AAAI Conference on Weblogs and Social Media (ICWSM-10), vol. 1001, 2010, p. 48103.

[6] P. Monge and N. Contractor, Theories of communication networks.Oxford University Press, USA, 2003.

[7] N. Christakis and J. Fowler, “The spread of obesity in a large socialnetwork over 32 years,” New England Journal of Medicine, vol. 357,no. 4, p. 370, 2007.

[8] ——, “The collective dynamics of smoking in a large social network,”New England journal of medicine, vol. 358, no. 21, pp. 2249–2258,2008.

[9] F. Wu, B. Huberman, L. Adamic, and J. Tyler, “Information flow insocial groups,” Physica A: Statistical and Theoretical Physics, vol. 337,no. 1-2, pp. 327–335, 2004.

[10] E. Bakshy, B. Karrer, and L. Adamic, “Social influence and the diffusionof user-created content,” in Proceedings of the Tenth ACM Conferenceon Electronic Commerce. ACM, 2009, pp. 325–334.

[11] L. Backstrom, D. Huttenlocher, J. Kleinberg, and X. Lan, “Groupformation in large social networks: membership, growth, and evolution,”in KDD’06, 2006, p. 54.

[12] D. Centola, “The Spread of Behavior in an Online Social NetworkExperiment,” Science, vol. 329, no. 5996, p. 1194, 2010.

[13] M. McPherson, L. Smith-Lovin, and J. Cook, “Birds of a feather:Homophily in social networks,” Annual Review of Sociology, vol. 27,pp. 415–444, 2001.

[14] S. Aral, L. Muchnik, and A. Sundararajan, “Distinguishing influence-based contagion from homophily-driven diffusion in dynamic networks,”Proceedings of the National Academy of Sciences, vol. 106, no. 51, p.21544, 2009.

[15] J. Leskovec, L. Adamic, and B. Huberman, “The dynamics of viralmarketing,” ACM Transactions on the Web (TWEB), vol. 1, no. 1, p. 5,2007.

[16] C. Shalizi and A. Thomas, “Homophily and contagion are genericallyconfounded in observational social network studies,” Sociological Meth-ods & Research, vol. 40, no. 2, p. 211, 2011.

[17] D. Crandall, D. Cosley, D. Huttenlocher, J. Kleinberg, and S. Suri,“Feedback effects between similarity and social influence in onlinecommunities,” in Proceeding of the 14th ACM SIGKDD internationalconference on Knowledge discovery and data mining. ACM, 2008, pp.160–168.

[18] J. Gastril, The group in society. Sage Publications, 2009.[19] G. Palla, A. Barabasi, and T. Vicsek, “Quantifying social group evolu-

tion,” Nature, vol. 446, no. 7136, pp. 664–667, 2007.[20] D. Huffaker, M. Simmons, E. Bakshy, and L. Adamic, “Seller activity

in a virtual marketplace,” First Monday, vol. 15, no. 7-5, 2010.[21] R. Kohavi, “A study of cross-validation and bootstrap for accuracy

estimation and model selection,” in International joint Conference onartificial intelligence, vol. 14. Citeseer, 1995, pp. 1137–1145.

[22] Y. Zhu and D. Shasha, “Efficient elastic burst detection in data streams,”in Proceedings of the ninth ACM SIGKDD international conference onKnowledge discovery and data mining. ACM, 2003, pp. 336–345.

[23] F. Chung and L. Lu, “Connected components in random graphs withgiven expected degree sequences,” Annals of combinatorics, vol. 6, no. 2,pp. 125–145, 2002.