The Structure of Positive Interpersonal Relations in Small Groups 1

1、Abstract

The authors sought to test Homans’s proposition that small groups inevitably generate a social structure which combines subgroups ( cliques ) and a ranking system . We present a graph theoretical model of such a structure and prove that a necessary and sufficient condition for its existence is the absence of seven particular triad types . Expected frequencies of the seven triad types in random graphs are deduced from elementary probability theory , and we suggest that a reasonable operational statement of Homans’s theory is that in most groups , the seven key triads are less frequent than the random model would predict . A data pool of sociograms and sociomatrices from 427 groups was collected from diverse published and unpublished studies . Random samples of 30 school and 30 adult groups were drawn from the pool and analysed . Significant majorities of both samples showed deviations from chance in the directions predicted . As a check , 60 simulated groups with truly random relationships were analysed and found to be close to the chance expectations and quite different from the real data samples .Overall , we claim support for Homans’s theory .
作者试图验证霍曼斯命题：即小群体不可避免地会产生一种由小群体（团）和一个排名系统组成的社会结构。我们提出了这种结构的一个图理论模型，并证明了其存在的一个必要和充分条件是没有七种特定的三元组类型。随机图中七种三元组类型的期望频率由初等概率论推导，我们认为，霍曼斯理论的一个合理的操作性陈述是，在大多数群体中，这七个关键的三元组比随机模型预测的频率要低。从不同的已发表和未发表的研究中收集了427组的社会图和社会心矩阵的数据。从数据库中提取并分析了30个学校和30个成人组的随机样本。这两个样本中的绝大多数都显示出与预测的方向有偏差。作为检验，对60个具有真正随机关系的模拟组进行了分析，发现它们与机会期望值非常接近，与真实数据样本非常不同。总体而言，我们声称支持Homans的理论。
In The Human Group ,George Homans presents a set of closely linked propositions about subgroup formation and ranking.In paraphrase , his argument is this:

• 1 ) In any group the external system (loosely, the group’s environment) makes it inevitable that frequencies of interaction will be unevenly distributed among the member pairs.
• 2 ) Because differential frequencies of interaction, interpersonal liking, and similarity in other sentiments and activities go together, pairs and larger subsets with initially higher rates of ineraction come to be increasingly differentiated from the rest of the group, forming subgroups (cliques) characterized by high rates of voluntary interaction，positive interpersonal sentiments, and normative consensus
• 3 . Nevertheless, the members are more nearly alike in tho norms they hold-than in their conformity to these norms (p. 126) and since the closer a person’s activities come to-the norm, the higher his rank will be (p. 141), all groups develop systems of ranking.
  在 Human Group 中，George Homans 提出了一组关于子群形成和排名的紧密联系的命题，换句话说，他的论点是：
  1）在任何群体中，外部系统（松散地，群体的环境）使得互动频率不可避免地在成员对之间分布不均
  2）由于互动频率的差异、人际喜好以及其他情绪和活动的相似性交织出现，最初互动率较高的对和更大的子集与群体其他部分越来越不同，形成了以高速率为特征的子群（团）自愿互动、积极的人际情感和规范共识
  3）然而，这些成员在他们所持有的规范上更接近于相似，而不是在遵守这些规范方面，并且由于一个人的活动越接近规范，他的等级就越高，所有群体都制定了排名系统。

It is hard to avoid the inference that if we examine voluntary interaction and sentiments in small groups, Homans expects us to find two sorts of structures, differentiation into cliques and elaboration into ranks, and he expectsus to find them in group after group after group.Furthermore, he expects us to find both structures with the same variables.Not only do subgroup members have higher rates of interaction, but so do higher ranking members.Not only do subgroup members have higher frequencies of liking, but higher ranking persons are better liked.
很难避免这样的推论：如果我们考察小群体中的自愿互动和情感，霍曼斯希望我们找到两种结构，分化成团（团）和细化成等级，他希望一组一组地找到他们。此外，他希望我们找到具有相同变量的两个结构。不仅子组成员的互动率更高，排名高的成员也是如此。不仅子组成员有更高的喜欢频率，而且排名高的人更喜欢。

These propositions are as well known as any in sociology , yet we have little systematic evidence for them .Homans himself says：
…let us be clear that it (the association between interaction and liking) is only a hypothesis, not a theorem.We have offered no proof, except what is provided by the behavior of the Bank Wiremen, and a statistician would say that a single instance is not nearly enough.Plenty of confirmatory evidence could be found in anthropological and sociological studies of small groups
这些命题和任何社会学一样广为人知，但我们几乎没有系统的证据来证明它们。霍曼斯自己说：
…让我们明确一点，它（互动和喜欢之间的关联）只是一个假设，而不是一个定理。除了银行电报员的行为提供的证据外，我们没有提供任何证据，而统计学家会说单个实例是远远不够的。在小群体的人类学和社会学研究中可以找到大量确证证据
This paper aims to test Honans’s structural Propositions using simple statistical models developed from graph theory and applying them to a data pool of interpersonal relations measures ( sociograms and sociomatrices ) for 427 groups . To the extent that our model is plausible , our probabilistic reasoning is valid, and our 427 groups are representative, the results provide favorable evidence for the propositions
本文旨在使用从图论发展而来的简单统计模型来检验霍曼斯的结构命题，并将其应用于人际关系测量数据库（社会图和社会矩阵）对于427组。在我们的模型可信的程度上，我们的推理是有效的，我们的427组具有代表性，结果为命题提供了有利的证据

2、A Graph Theoretical Modelwhile

Honans’s definitions are notoriously crisp , he here defines the total structure which is implied by his twin principles , We take the liberty of sketching such a model, hoping that it does justice to the original
We begin with the notion of a “positive relation” and say that person i has a positive relation to person j if he :

• 1 ) frequently interacts with j on a voluntary basis ( formal authority is excluded from the hypotheses)
• 2 ) expresses a positive sentiment about j
• 3 ) would prefer to interact with j on a voluntary basis
• 4 ) claims that j is his friend

一个图论模型
而Honans’s的定义是出了名的清晰，他在这里定义了双核原则所隐含的总体结构，我们冒昧地画出这样一个模型，希望它对原作公正
我们从“积极关系”的概念开始，假设i与j之间存在积极关系：
1）经常在自愿的基础上与j互动（假设中排除了形式权限）
2）表达对j的积极情绪
3）在自愿的基础上愿意与j互动
4）声称j是他的朋友
Note that the opposite of a positive relation , nonpositive maybe neutral ( indifferent ) or negative ( dislikes ,avoids , etc . ) Note further that positive relationships are not defined as symmetrical . If i has a positive relation to j , j may or may not reciprocate at . There are three logical possibilities which we will call M for mutual positive relations ,A for asymmetric relations in which there is a positive relation i to j or j to i but not both , and N for mutual non-positive relations . In graphs , we nay draw then as follows
$M:i\leftrightarrow jA:i\to j|i\leftarrow jN:i\leftrightsquigarrow j$
注意，积极关系的反面，或许中性（冷漠）或消极（不喜欢、避免等）。进一步注意，积极的关系被定义是不对称的。如果i与j有积极的关系，j可能会或可能不会回报他。存在三种合乎逻辑的可能性，我们将其称为M型，表示相互正关系，A型表示不对称关系，其中i与j或j与i之间存在正关系，但不同时存在，N型表示相互非正关系。如图.
Having granted that pair relations may be symmetric or asymmetric , we should now turn to the triads produced by all possible combinations of members taken three at a time . The logical heart of our model will consist of a set of propositions about these triads.However, our discussion will be clearer, though less rigorous, if we skip ahead to the sort of group structure which is implied by our yet unstated triad propositions.
假设对关系可以是对称的或不对称的，我们现在应该转向由所有可能的成员一次取三个组合产生的三元组。我们模型的逻辑核心将由一组关于这些三元组的命题组成。然而，如果我们跳到我们尚未说明的三元命题所暗示的那种群体结构，我们的讨论会更清楚，尽管不那么严格。
We begin by treating a group’s ranking structure as a series of ordered levels, which is another way of saying that there may be more people than status distinctions in the group.It is useful to think of the levels as stories in a building, in the sense that people on a given floor do not differ in level, any two persons on different floors are unambiguously ordered by level , and the stories form a complete order
首先将一个群体的排名结构视为一系列有序的级别，这是另一种说法，即群体中的人数可能多于地位差异。将级别视为建筑物中的楼层是有用的，因为在给定楼层上的人级别没有不同，不同楼层上的任何两个人都有明确的层级排序，楼层形成完整的顺序。
The building analogy is useful , but misleading in one important sense . while floors and ceilings mark the levels in a building , in structural theory we seek to generate features from the pattern of pair relations themselves .indeed one may think of social structure as those characteristics of a group which may be deduced from the characteristics of pair relations within the group . This leads us to one of the main ideas of the model :
Relations of the sort we have called A are assumed to connect persons in different levels.while M and N relations are assumed to connect persons in the same level.Further we assume that in pairs connected by A relations the recipient of the positive relationship is in the higher level
用建筑类比是有用的，但在一项重要意义上是有误导性的。虽然地板和天花板标志着建筑的层级，在结构理论中，我们试图从成对关系本身的模式中产生特征。实际上有人可能会将社会结构视为一个群体的特征，这些特征可以从群体内成对关系的特征中推断出来。这导致我们得出该模型的一个主要思想：我们称之为 A型 的那种关系被假定为将不同层级的人联系起来。而假设 M型 和 N型关系连接同一层级的人员。进一步，我们假设A型对关系中，积极关系的接受者处于更高层次
we are claiming that if you and I like each other or if neither of us likes the other we are probably in the same status level in our group , but if I like you and you do not like me , you are probably in a different and higher level.We may think of such A relations as “admiration” and summarize the whole business with the slogan,“admiration flows up levels.”
我们声称，如果你和我喜欢对方，或者如果我们都不喜欢对方，我们可能处于同一层级状态，但如果我喜欢你而你不喜欢我，你可能处于与我不同且更高的层级。我们可以把A型关系看作是“倾慕”，用“钦佩是向上流动的”的口号来概括整个业务。
The second major idea of the model is that within a level there may be disjoint subsets of people ( cliques or subgroups). analogous to people in different rooms on a floor of a buliding : Again , these must be defined relationally , All itra-level relationships , of course , must be type M or type N if all A relations lie between levels , which leads us to thesececond main idea of the model : M relations are assumed to connect persons in the same clique within a level . N relations are assumed to connect persons in different cliques with a level . Using our building analogy we advance the slogan " N relations make partitions.“
该模型的第二个主要思想是，在某个层级中，可能存在不相交的人群（团或子群），如同一楼层不同房间的人：同样，这些关系必须以关系的方式定义，当然，如果所有A型关系都位于不同层级之间，则层内关系必须是类型N或类型N，这导致我们得出模型的第二个主要思想：假设M型关系在一个层级内连接同一个团的人。假设N型关系连接同一个层级种不同团的人。使用我们的建筑类比，我们提出了口号“N 型关系进行分区”。

Putting both ideas together , the heart of the model is the notion that in small groups the members tend to be divided into levels by the pattern of their A relations and within levels they tend to be divided into cliques by the pattern of their M and N relations , Figure 1 puts the idea in rough schematic form :
将这两种观点放在一起，该模型的核心是这样一个概念，即在小群体中，成员往往按照其A型关系的模式划分为不同的层级，在层级内，他们往往按照M型和N型关系的模式分为不同的团，如图1所示

Remembering that we have not yet stated any principles which guarantee that such a strueture must emerge and be consistent ,let us examine Figure 1 . We see that it has three levels and five cliques , though the number of cliques and levels is not fixed by the model . within each clique all relationships are of the M type , where cliques differ in level there are always A relationships with the arrows pointing up , and between cliques at the same level there are N relationships . We further note that the top level has only one clique , which is to say that all cliques are assumed to lie within a level but it is not assumed that every level has a clique , None of the levels in figure 1 has more than two cliques , but this is simply because the diagram would be too cluttered .
记住，我们还没有阐明任何保证这种结构必须出现并保持一致的原则，让我们检查图1。我们看到，它有三个层级和五个团，尽管团和层级的数量并不是由模型确定的。在每个团内部，所有关系都是M型的，不同层级的团之间，总是有箭头指向上的A型关系，而在同一层级的团之间，是N型关系。我们进一步注意到，最高层级只有一个团，也就是说假设所有团都位于一个层级内，但并不意味着每个层级都有一个团，图1的层级中没有一个团超过两个，但这只是因为图表太杂乱。

Let us now ask whether Figure 1 is a plausible translation of Honans’s ideas .Considering cliques ( subgroups ) first , Homans defines them as follows :
If we say that individuals A,B,C,D,E… form a group, this will mean that at least in the following circumstances hold.Within a given period of time, A interacts more often with B,C,D,E, than he does with M,N,L,O, P… whom we choose to consider outsiders or members of other groups.B also interacts more often with A,C,D,E, than he does with outsiders, and so on for the other members of the group.
现在让我们问一下图1是否是霍曼斯思想的合理翻译！首先考虑团（子群），霍曼斯将其定义如下：
如果我们说个人 A,B,C,D,E… 组成一个群体，这将意味着至少在以下情况下成立。在给定的时间段内，A 与B,C,D,E 的互动比他与 M,N,L,O, P…的互动频率更高，我们选择将 M,N,L,O, P视为局外人或其他群体的成员。 B 与 A、C、D、E 的互动也比他与外人的互动频率更高，以此类推，适用于该组的其他成员。
In other words cliques are subsets of individuals with higher rates of positive relationships among themselves than with outsiders.This, of course, is the definition of cliques in the theory of structural balance and clusterability. our cliques have this property since each pair wthin a clique has two positive relationships , while each inter-clique has either one positive relationship ( if i and j are in different levels ) or none ( if i and j are in the same level ) .
换言之，团是个体的子集，他们内部比与局外人之间的积极关系比率更高。这当然是结构平衡和聚类理论中对团的定义。团具有这种性质，因为一个团中的每一对都有两种积极关系，虽然团之间要么有一个正关系（如果i和j处于不同的层级），要么没有（如果i和j处于同一的层级）。
Because Homans nowhere gives a formal definition of ranking， it is harder to say that our version fits his second principle.However, the model has two properties which seem natural for a ranking
因为霍曼斯没有对排名给出正式的定义，所以很难说我们的版本符合他的第二个原则。然而，该模型有两个对于排名来说似乎很自然的属性。
First , people in higher levels receive more positive relations , Consider person i in level i and person j in some lower level j . Mr . j receives positive relations from everyone in level l,l-1 plus anyone in his own clique in level j.Lofty Mr. i , however , receives all of these plus relations from everyone in lavel j whether or not they are in j’s own clique , plus some from everyone in levels which might occur above j and below i , Plus those from anyone in i’s own clique .
In the limiting case where i and j are the only members of adjacent levels , i still receives one more positive relation than j , the one from j to i , which is not , by definition ,reciprocated . Thus , in general , if two persons differ in level ,the one in the higher level will receive more positive relations.This property is not only consistent with Homans’s statement that leaders are more popular , but when interpreted in terms of interaction , it squares with his proposition , " the higher a man’s social rank, the larger will be the number of persons that originate interaction for him.Men that are not highly valued must seek others rather than be sought by them“
首先，层级较高的人获得更多的积极关系，考虑i层级的人i和较低层级的人j。Mr.j从层级l中获得积极关系，l-1加上j层级中自己团中的任何人的关系。然而，高高在上的i从级别j中的每个人获得所有这些积极关系，无论他们是否在j自己的团中，再加上一些来自在j层级以上且i层级以下的关系，再加上i与自己团中其他人的关系。
在有限的情况下，i和j是相邻级别的唯一成员，i仍然比j多收到一个更积极的关系：从j到i的关系，根据定义，这不是相互的。因此，一般来说，如果两个人的层级不同，高层级的人将获得更多积极的关系。这一属性不仅符合霍曼斯的领导人更受欢迎的说法，而且从相互作用的角度来看，它与他的命题一致：“一个人的社会等级越高，向他发起互动的人数就越多。不被重视的男人必须寻求别人而不是被别人寻求”
second , positive relations are transitive . If i has a positive relation to j and j has a psoitive relation to k , then i will always have a positive relationship to k.The proof is simple,but tedious , and will not be presented in detail .One takes all the possible triads permitted under the triad propositions to be stated later , examines the six possible three-step paths in each ( e . g , i to j to k , i to k to j,j to i to k…) . and sees that there are none in which the first two are positive and the third non-positive .
第二，正关系是可传递的。如果i与j有正关系，而j与k有正关系，那么i将始终与k有正关系。证明很简单，但很乏味，并且不会详细介绍。我们取三元命题下允许的所有可能的三元组，检查每个路径中的六个可能的三步路径（例如，i到j到k，i到k到j，j到i到k…），并发现没有前两个是正的，第三个是非正的。

since differential popularity and transitivity are the most common definitions of ranking systems , we feel that the model has some plausibility .
We have not , however , successfully reflected every structural proposition in The Human Group . There is at least one near miss and one clear difference of opinion.
The near miss is the claim that all cliques are ranked. In our model , the cliques are partially ordered(e . g . in Figure 1,the partial ordering is $1>(2,3)>(4,5)$> ),but cliques within a level are not ordered .
由于差异受欢迎度和传递性是排名系统最常见的定义，我们认为该模型具有一定的合理性。
然而，我们还没有成功地反映人类群体中的所有结构命题。其中至少有一个近似差错和一个明显的意见分歧。
近似差错是宣布所有团都被排序（第13页）。在我们的模型中，团是部分有序的，如图1中不同层级的团有序，但同一个层级内的团是不有序的。
The difference of opinion is worth some discussion .Homans states:
the more nearly equal in social rank a number of men are, the more frequently they will interact with one another…if a person does originate interaction for a person of higher rank, a tendency will exist for him to do so with the member of his own subgroup who is nearest him in rank

意见的分歧值得讨论。霍曼斯表示:
人们在社会地位上越接近平等，他们之间互动的频率就越高……如果一个人确实向层级更高的人发起了互动，那么他将有一种倾向，即与自己的子群中地位最接近他的人进行互动
In a group with three or more levels,this proposition implies N relationships between persons whose levels are not adjacent (Mr. Low seeks out Mr. Middle who seeks out Mr. High, but Messrs. High and Low do not seek out each other).Our model does not allow this.It does imply that the very highest rates of positive relations will be in the same level(intra- clique relations), but so will the lowest(inter-clique relations)。
在具有三个或更多层次的群体中，这个命题暗示了层次不相邻的人之间的 N 种关系（Low 先生寻找 Middle 先生，Middle 先生寻找 High 先生，但 High 和 Low 不相互寻找）。我们的模型不允许这样做。这确实意味着最高的积极关系等级将处于同一水平（团内关系），但最低的（团内关系）也是如此。

similarly , Homans implies that if i directs a positive relation to someone above him in rank , it will not go to everyone in that rank , while our model implies that everyone directs positive relations to everyone in every rank above him . The issue is of some interest because Homans uses it to argue , in effect , that interaction ( and thus presumably sentiments ) tend toward the structure graph theorists call a " tree from a point" ,most common structural model for formal organizations .
类似地，霍曼斯暗示，如果i与高于他的层级的人建立正关系，那么这不会影响到该级别的所有人，而我们的模型则暗示每个人都与高于他的每个层级的每个人建立积极关系。这个问题有一定的意义，因为霍曼斯用它来论证，实际上，这种互动（以及由此产生的情感）倾向于图理论家称之为“点树”的结构，这是正式组织最常见的结构模型。
Homans does go on to soften his proposition by saying that it is less true of smaller groups And those in less severe environments. The question is an empirical one and the reader who concerned about it should watch carefully in the data analysis for the results on what will be called " 0-2-1-b " triads . For now , we merely note that our model assumes groups so small or in such benign environments that positive relations do not “go through channels" .
霍曼斯确实软化了他的观点，他说，对于较小的群体和环境不那么恶劣的群体来说，这是不太正确的。这个问题是一个实证问题，关注它的读者应该在数据分析中仔细观察所谓的“0-2-1-b”三元组。目前，我们只注意到，我们的模型假设群体非常小，或者处于良性环境中，正关系不会“通过渠道”。
Me end our preliminary discussion of the model by noting its logical ties to some other models . It can be shown that by varying our assumptions about the presence of M , A , and N relations,the model can be changed into other well known structures .
1 ) It all pair relations are symmetrical , M or N , the structure consists of a single level and is equivalent to clusterability or structural balance .
2 ) If all pair relations are antisymmetrical , A , the structure consists of as many levels as people , all cliques are size one , and it is a transitive tournament
3 ) If all pair relations are M or A,there is only one clique at each level and the structure is what Hempel calls a “quasi-series.”
在结束我们对该模型的初步讨论时，我注意到它与其他一些模型的逻辑联系。可以表明，通过改变我们对M、A和N关系存在的假设，该模型可以转变为其他众所周知的结构。
1）如果所有对关系都是对称的，M型或N型，则该结构由单个层级组成，并等价于聚类性或结构平衡。
2）如果所有的对关系都是反对称的，A型，该结构人与层级一样多，所有的团都大小为1，并且它是一个传递竞赛图（transitive tournament，每对顶点之间都有一条边相连的有向图称为竞赛图）。
3）如果所有的对关系都是M型或A型，那么在每个层级上只有一个团，这种结构就是Hempel所说的“准级数”