David A. Kenny
April 18, 1998

    The Social Relations Model is a model for dyadic data. Data from two-person interactions and rating or sociometric studies can be used. The level of measurement should be interval (e.g., seven-point scales) and not categorical (e.g., yes or no).  Generally the data are collected from people but the dyadic units can be animals, groups, organizations, cities, or countries.

    There are three major types of effects in the Social Relations Model: actor, partner, and relationship effects. The actor effect represents a person's average level of a given behavior in the presence of a variety of partners. For example, Sandy's actor effect on the variable of trust measures the extent to which she tends to trust others in general. The partner effect represents the average level of a response which a person elicits from a variety of partners. Sandy's partner effect measures the extent to which other people tend to trust her. The relationship effect represents a person's behavior toward another individual in particular, above and beyond their actor and partner effects. For example, Sandy's relationship effect toward Mark on the variable of trust measures the extent to which she trusts him controlling for her general tendency toward trusting others and his general tendency to be trusted by others.  Relationship effects are directional or asymmetric, such that Sandy may trust Mark more, less, or the same as he trusts her. To differentiate relationship from error variance, multiple indicators of the construct, either across time or with different measures, are necessary.

    Actor and partner are generic terms. In social perception data, actor might be more appropriately called perceiver and partner might be called target.  In nonverbal communication, the actor might be called a receiver and the partner a sender. In studies of liking, the actor effect measures how much the person likes others, and the partner effect measures how much a person is liked, i.e., popularity.

    The focus in Social Relations modeling is not on estimating the effects for specific persons and relationships but in estimating the variance due to effects. So there is a study of how intelligent people see each other, the interest is in whether there is actor, partner, and relationship variance.  Actor variance would assess if people saw others as similar in terms of intelligence, partner variance would assess whether people agree with each other in their ratings of intelligence, and relationship variance would assess the degree to which perceptions of intelligence are unique. (To learn more about the mathematics of the Social Relations Model.)

    There has been extensive research using the model. Consult the "Interpersonal Perception" page to see the type of questions that can be answered using the model. The bibliography also lists published papers that have used the model.  Any variable that is dyadic can be studied using the model.

    The most common Social Relations design is the round-robin research design. In this design, each person interacts with or rates every other person in the group, and data are collected from both members of each dyad. The scores of the two people are usually different. Usually there are multiple round robins.  SOREMO is a computer program that is used to analyze data from round-robin designs. To be able to estimate the parameters of the Social Relations Model, all groups must have at least four people. The documentation for computer program SOREMO is contained in the file SOREMO.DOC.

    Consider the Warner, Kenny, and Stoto study of speech (this data set can be downloaded). The data file is called WARN.DAT. The user might create a file called SWARN.SET that would contain the instruction to SOREMO on how to analyze the data.  The output might be written to a file called WARN.OUT. To run SOREMO, the user would then create the following three-line file, called perhaps SOREMO.SET:

SWARN.SET
WARN.DAT
WARN.OUT

The files WARN.SET, WARN.DAT, and SOREMO.SET can be downloaded from this site. When prompted for the file with "I/O instructions," the user types SOREMO.SET.

    SETSRM, written with Patrick Sullivan and Katherine LaFontana, is a computer program that is used to assist in the running of the computer program SOREMO. It prompts the user with basic job information to create a file such as SWARN.SET. One can use the program SETSRM to change files for SOREMO. One can also run SOREMO from SETSRM. The one task that SETSRM cannot do is change which variables are analyzed. To run SETSRM, the user need only type SETSRM. Unlike SOREMO, it is interactive.

    The program BLOCKO presumes that the design is a block design. In this design a group is divided into two groups and members from each group interact or rate the members of the other group. Such a design is referred to as a block design. There must be at least two members of each subgroup. The Warner et al. study can be considered a block design in that four men interact with four women. So the opposite-gender interactions can be considered a block design.

    The documentation for BLOCKO is contained in the file BLOCKO.DOC which can be downloaded. The test data are contained in the file WARN.DAT and the file to run the program is called BWARN.SET. Like SOREMO, BLOCKO requires a file that contains the names of the file with instructions, the data file, and the output file. For the example, that file might be called BLOCKO.SET and the instructions file is BWARN.SET. The file WARN.DAT can be used as the data.

    The program AID_SRM is used to determine the power of Social Relations designs and is useful the planning of studies: what design to use (round robin versus block), how many groups, and how many persons per group. It was written with Brian Lashley and is considered to be in development. Advice about improving this program or any others would be appreciated.