Family Process - March 1964 - HALEY - Research On Family Patterns An Instrument Measurement

_____________________________________________________________________________________________________________
Fam Proc 3:41-65, 1964
Research on Family Patterns An Instrument Measurement

JAY HALEYa
aResearch Associate, Mental Research Institute of the Palo Alto Medical Research Foundation. This research was partially supported by
the U.S. Army Medical Research and Development Command (grant no. DA-49-193-MD-2396). Recording equipment was provided
by the Ampex Foundation. Computer programming was conducted by Janet Beavin, Research Assistant. Gratitude is expressed to the
National Science Foundation under Grant Number NSF-Gp948 for support of the computers at Stanford University.
This is a report of an investigation to determine whether, and on what dimensions, the patterns of interaction between
members in one family can be differentiated from those in another family. Can it be shown, for example, that the
interchange in a family where one or more members suffers from some form of psychopathology is different from that in a
family without evident psychopathology? In this presentation there will first be a clarification of this particular area of
family research and a discussion of the kinds of problems which arise. Then there will be a report of a study of 80 families
with a measurement which appears to differentiate families from one another, which differentiates most "disturbed" from
"normal" families, and which indicates the possibility of measuring change before and after Family Therapy in the most
simple way possible with the use of an instrument rather than human judgment.
TYPES OF FAMILY RESEARCH

Although the family has been a subject of investigation in the past, only recently has there been a focus upon the family
system: the habitual transactions between family members as they interact with one another. This focus on family patterns
can be differentiated from three other major kinds of family research to avoid confusion about the purpose of the research
reported here.
Statistical Study. One area of family research has been the compilation of information, or facts, about families, including
such information as how many divorces there are per year in a particular society, the average age at time of marriage, the
number of children per family, and so on. The data are largely the records of governmental bureaus and institutions, and the
questions posed are in terms of correlations between different items, such as a correlation between age at marriage and
frequency of divorce or the presence of a delinquent and broken homes. This approach does not include observation of
actual family members.
Anthropological Study. Another area of family study has been the observation of families in different cultures as part of a
general attempt to portray a culture and examine its influence upon the individual. The data has been largely the
observations of the investigator and the reports of his native informants. The questions posed center in the relationship
between the family and cultural patterns, with particular emphasis upon types of kinship system. Attempts also were made
to correlate the development of individual personality with child-rearing customs, and the family unit, or lack of one, was
seen as relevant to that development.
Individual Study. A third major approach in family research has been, paradoxically enough, the investigation of
individuals. Once it was assumed, without evidence, that the individual was a reliable informant about his family, it became
common to base descriptions of the family upon the reports of individual family members. The emphasis in such studies is
upon the person's attitude about, or perception of, other family members, family relationships, family values, and so on. The
procedures in this type of research have included giving survey questionnaires to individuals, interviewing family members,
and giving members projective, personality, or attitudinal tests. Similarly, each parent is exposed to hypothetical child
rearing situations and asked how he would deal with them, or the family life cycle is studied by asking individuals what life
is like in their families now or to recall what it was like when they were young. The results of such studies are, by
definition, a comparison of the responses of individualshow this person answers in contrast to this person in this other
familyrather than a comparison of family systems. At times comparisons are made in this type of study which can only be
called illegitimate: to map how each family member perceives the relationships in his family according to his self report,
and then to combine his answers with those of other family members and call this an interpersonal portrait of relationships
in the family is not proper. Such a combination is a totaling of the perception of individuals and may or may not be related
in any way to the actual transactions which take place in that family.
Interactional Study. Recently a number of investigators have attempted to describe the interaction pattern in families by
bringing the members together as a group and studying the ways they actually deal with one another. The families are
observed in conjoint interviews where they are asked standard questions, in family therapy sessions, or they are exposed
jointly to a task, such as playing a game together, filling out a questionnaire together and reconciling differences, or giving
their impressions of TAT or Rorschach cards. The data is collected by an observer or on tape or film. Two sorts of
measurement are attempted in these studies: the counting of the responses to the task by frequency or time, and the
1
15455300, 1964, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1545-5300.1964.00041.x by National Institutes Of Health Malaysia, Wiley Online Library on [26/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
_____________________________________________________________________________________________________________
categorization of the verbal interchange of the family conversation. It is this type of research, where the family members are
brought together to interact, which is the focus of the study being reported here.
PROBLEMS OF CLASSIFYING INTERACTION

Although it seems reasonable to study a family system by recording it in operation, the problem of categorizing the
interchange in such a way that families can be rigorously contrasted is exasperatingly difficult. The goal of such an
endeavor is to make reliable measurements of typical family interchanges while introducing as little bias as possible into
the system being studied. Yet one is faced with extremes. Merely observing the family in its natural habitat introduces the
least amount of bias, but the measurements can only be the subjective reports of the observer. Placing families in structured
situations so that their communication with one another is limited, for example to pushing buttons, one can achieve accurate
measurements of the interchange but an extreme bias may have been introduced since the family members do not habitually
communicate in that limited way. Since family members typically communicate by conversation, one would like to make
measurements of that conversation, and yet how can such complex communications be categorized and quantified with any
reliability?
The ideal data in family research should be a recording of observable events which are accurately measurable in some
way so that comparisons and contrasts between families can be made. There should be no guesswork or inference in the
data itself, although what the data "means" must involve inference. That is, we should not have to infer whether something
happened or not in the interchange, we should be able to state that as a fact, although we might then make inferences about
the nature of a family from that fact.
In psychiatry and clinical psychology in general, and in most family research to date, the problem has resided in
collecting and categorizing data which cannot ultimately be rigorous because it is inferential by definintion. A category is
"inferential" whenever one must infer whether something happened or whether or not a certain state exists. All
investigations of thought processes, perception and emotions are inferential; we cannot know what someone perceives, or
even if he does, we can only infer it. Similarly all descriptions of what someone "means" are inferential, as is any
description of motivation. We cannot verify why a person does what he does, we can only infer the reason and support our
argument with the help of other inferences.
In family research, investigators have assumed that they were dealing with problems similar to those of individual
research and therefore past methods should continue to be used. When observing a family conversation and attempting to
categorize it, the observer's goal has been to avoid subjective bias. Since the investigator might be looking at the
conversation in a certain way, his inferences might be unreliable. Therefore it has become common to have several people
look at the conversation to see if they all classify the interaction in the same way. This is thought to be more scientific, and
yet even if several raters agree, which is unusual unless they have been trained to look at the data in the same way, there is
still doubt whether the family is actually doing that something or whether the raters are merely making the same inference
because they have a common point of view. One has the nagging thought that not many years ago any number of raters
would have agreed the world was flat, and the question of the shape of the world was not answered by taking a vote. Having
several people listen to a tape of a family conversation, or watch a film, and attempt to compress the interaction into a set of
simple categories is further complicated by the extraordinary speed and complexity of such an interchange. One has only to
look at a film of a family in slow motion to realize how much is overlooked at normal speed. To say that two or more
people are "doing" any particular thing, such as "agreeing" or "disagreeing," when one takes into account all the body
movement, the vocal intonations, and the complex verbal statements, is to do considerable violence to the complex multiple
level patterns of communication which occur in a conversation. Finally, the results obtained from such an approach can
only be based upon inference; the raters are guessing. This is certainly true of such categories as mother "dominating" or
"directing" father or whether two people are being "hostile" but it is also true of more specific acts. If one tabulates the most
elementary sort of behavioral category, such as "who is speaking to whom" when observing a family interchange, the results
are inferential. The investigator must "guess" whether father spoke to mother at that point. It may be a good guess based
upon the fact that father looked at mother while he spoke, yet perhaps father was only apparently speaking to mother but
"really" speaking to the child. If that doubt exists, we cannot state it as a fact that father spoke to mother, even if any number
of raters should agree.
The importance of getting inference out of the collection of data is not mere quibbling but the most basic problem in
research. If our measurements are a guess, and we then build further inferences upon them, the entire structure is built upon
quicksand. This does not mean we should either collect "facts" or abandon research, because at a particular stage we might
not be able to collect ideal data. However, often the use of crude guessing seems to be done as an end in itself rather than as
a first step toward building more rigorous measurement. The opportunity posed by family research is not merely that we
might increase the reliability with which we study the human being, but that we can develop a different concept of
reliability. When the temperature of water was categorized as hot, cold, or warm with the use of the elbow, reliability could
be increased by having several elbows immersed. The introduction of the thermometer was not merely an increase in
reliability but a different way of going about the measurement. Study of the individual in isolation in the past might have
2
_____________________________________________________________________________________________________________
required inference because the individual remains a "black box." In family research, we appear to have the opportunity of
shifting the nature of our measurements because we are dealing with the interchange of observable acts.
THE BASIC QUESTIONS

Recently the author decided to make a fresh start on the problem of categorizing family conversations by formulating
what seemed to be the most basic questions faced in family research and then attempting to answer those questions with the
most reliable and non-inferential measurements possible. The questions chosen to be investigated were these:
1) Can we demonstrate that a family is an organization by showing that family members follow repetitive patterns in
their interaction? If it is to be argued, for example, that a family with a schizophrenic member is organized
differently from, or follows different patterns from, a family with only normal members, we should first present
reliable evidence that any family follows patterns which repeat. To phrase this question another way: Can we
demonstrate that family members are not behaving randomly in relation to each other?
2) Can we demonstrate that any one family is different from or similar to any other family on some base line which is
common to them both? Often an incident in one family is described and contrasted with an incident in another
family, but there is no common scale on which to place the different families in relation to a common point.
3) Can we, on some scale, differentiate a "disturbed" family from a more "normal" family? This is a different order of
question since it presumes two types of families exist. A "disturbed" family can be defined as one where one or more
members is in treatment or in conflict with community authorities, and a "normal" family is one where the members
have neither had treatment nor been arrested, i.e., they have confined what difficulties they might have within the
family.
4) Can we, on some scale, measure changes in the family system if they occur as a result of family therapy? This
question emphasizes change in family patterns rather than changes in the individual members. Presumably if family
therapy has been effective, the families must differ on a significant measurement taken before and after therapy.
With these questions formulated, it was decided to begin with the most simple measurement, and if the questions were
not satisfactorily answered to proceed to the next more complicated measurement. This report deals with the answers
provided by the most simple measurement possible.
THE SAMPLE
A group of eighty families was selected and interviewed. This group included forty "normal" families, primarily selected
randomly with the use of a high school directory, The parents were asked to come in with one child of their own chosing for
an interview. The children ranged in age from 10 to 20 and were living at home with their natural parents. Those families
which contained a member who had contacted, or been advised to contact, a psychotherapist or who had been arrested were
excluded from the normal category. The "disturbed" group consisted of forty families and included families in which some
member (a) was suffering from schizophrenia, (b) had committed a delinquent act, (c) had a school problem which brought
him to the attention of the authorities, as well as (d) families in which some member sought treatment for some neurotic
problem and (e) families where parents had sought marriage or family therapy. By definition a "disturbed" family was one
which had not been able to contain its difficulties and had become involved in community attention. The children ranged in
age from 10 to 20, but four children older than 20 were included. In the "disturbed" group, if the child was the identified
patient the parents were asked to bring in that child. (A list of the families, with the diagnoses, is given in Table I.) Some of
these families were tested both before and after family therapy. Additionally, there was a group of six families with a gifted,
or outstanding, child according to the school authorities. When these families had not come in contact with authorities
because of a problem, they were classed in the normal group. The families were largely middle class and were at least
second generation American.
Table 1
Deviation from Equal Use of the Six Categories
40 Normal Families 40 Abnormal Families
R Dev Age Sex R Dev Diagnosis Age Sex
1.63 11 M 2.84 Marital problem 13 M
5.53 15 M 6.93 Psychotic mother 16 M
6.01 11 M 10.69 Neurotic child 18 F
6.10 19 M 12.02 Psychotic child 10 M
9.39 17 F 14.04 Neurotic child 18 M
3
_____________________________________________________________________________________________________________

11.92 16 M 16.24 Delinquent 14 M
12.30 14 M 20.13 Neurotic child 15 M
12.74 18 M 21.72 Delinquent 14 M
13.59 18 M 24.61 Withdrawn child 19 M
13.98 18 F 24.72 Psychotic child 11 M
14.11 16 M 24.84 Acting out child 10 M
14.88 14 M 25.61 Delinquent 20 M
15.66 Gifted 16 F 25.73 Gifted Underachiever 16 M
15.87 16 F 27.30 Psychotic child 14 M
15.93 16 M 27.41 Acting out child 16 M
16.45 12 M 27.82 Delinquent 16 M
18.25 16 M 28.56 School problem 16 M
20.96 12 M 28.93 Marital problem 19 M
21.01 Gifted 16 F 29.05 Neurotic child 15 M
21.49 Gifted 16 M 29.52 Psychotic child 10 F
21.90 16 F 30.28 Neurotic father 17 F
22.06 14 M 31.79 Gifted underachiever 16 F
23.52 18 F 31.92 Sibling of psychotic 18 M
23.93 16 F 32.59 Marital problem 17 M
24.06 16 M 34.26 School problem 12 M
24.40 18 M 35.28 Psychotic child 34 F
24.49 Gifted 15 M 35.44 Delinquent 13 M
25.35 10 F 35.99 Delinquent 14 M
25.36 19 F 37.07 Neurotic child 10 M
25.84 Gifted 17 M 37.28 Neur. child (schizoid) 24 F
26.65 15 M 38.61 Neurotic child 13 M
29.67 17 F 39.07 School problem 13 M
31.63 Gifted 16 M 41.28 Acting out child 14 M
32.01 17 F 47.64 School problem 12 M
33.08 17 M 61.71 Psychotic father 14 M
 
766.21 1178.34
Mean = 19.16 Mean = 29.46
The families are ranked in order of increasing deviation from a random distribution. The computation of the R Deviation index
number is explained in footnote 3.
THE PROCEDURE
For purposes of this research, it was essential that the families have a conversation in which all three members have a
right to speak about the subject of conversation posed for them. A pilot sample of families was asked to "plan something to
do together" on the assumption that if they all were to do it together they would all have a right to speak about the subject.
However, long runs of conversation were needed and a more structured situation was sought. At this point, the writer
discovered that a research project being conducted by Dr. Antonio J. Ferreira and Dr. William D. Winter provided the sort
of situations that seemed to be well suited for the kind of measurement under consideration. In their project, family triads
were being studied while interacting towards a family-decision over relatively neutral items on a questionnaire and during
4
_____________________________________________________________________________________________________________
the telling of a story based upon TAT cards. The procedure promoted two different sorts of conversations; the questionnaire
discussion and the TAT discussion. In the questionnaire discussion the families were speaking about precise items on a list,
while in the TAT story making the conversation ran more freely. This seemed to meet the specific purpose of the
measurement reported here since it provided long runs of conversation among family members where each family member
had an equal right and opportunity to enter the discussion. Accordingly, the sharing of families was instituted between the
two projects.1
THE THEORETICAL APPROACH

When we examine a series of family conversations and wish to make a measurement which is as noninferential as
possible, there is only one set of observations of which we can be certain. We cannot be sure what the family members are
saying, what they mean by what they say, or even to whom each is speaking, but we can be reasonably certain that each
family member speaks. In fact, one can be sure of which one is speaking by placing on each person a microphone which
only fires if that person makes an audible sound.
Taking a conversation of mother, father and one child and recording the order in which the members speak, one obtains a
stream which looks like this: MCFMCFCFMFMFCFM ... and so on. (At times the family members speak simultaneously,
and a set of categories could be added for that, but in this research it was decided to break down simultaneous speech into
who spoke first and who last and so put the speeches in sequence. It was also decided that no person would follow himself,
and so if a person spoke and was silent for a period and spoke again this was counted as one speech.) Examining this
stream of conversation, one finds a variety of frequency counts are possible. One could total how often mother speaks, or
father speaks, or child speaks. Such a count would not help us compare families; we would only be comparing how an
individual in one family contrasts with an individual in another family. The most elementary count which measured an
interchange would be of sequences of two persons speaking. Examining this stream and making this most simple family
count, one can note the possibilities. Father can be followed by child or mother (FC or FM), mother can be followed by
father or child (MF or MC), and child can be followed by either father or mother (CF or CM). That is, this most elementary
frequency count of who follows whom provides six possible categories.
To compute the frequency of use of these six categories a "Family Interaction Analyzer" was devised by the Alto
Scientific Company of Palo Alto. Each family member at the table has a lavalier microphone which is connected to the
Analyzer, a small box with six windows in it. In each window is a counter which runs up a score as the people talk. When
father speaks, nothing happens until mother follows him, and then a count is made on the FM counter. Then when child
follows mother, a count appears on the MC counter, and so on. Therefore, as the family has a conversation, the sequence in
which they speak is immediately ordered by the machine so that totals appear automatically in each of the six categories. In
addition, each microphone fires a pen on an event recorder so that a visual record of the conversation is obtained.2
The Hypotheses
The theoretical approach of this study is based upon a change in the approach to classification of family patterns. In the
past an attempt was made only to contrast one type of family system with another, while in this approach types of family
systems are contrasted with non-organized behavior. It appears possible to use random, or non-organized, behavior as a
point of reference from which to compare families.
The study is based upon a simple definition of "organization": Organization requires limitation. If any group of people is
following certain patterns repetitively, and is therefore manifesting organization, they must be utilizing some patterns more
frequently than others. That is, given n possible things three people in a room might do, the frequent use of some
possibilities more than others indicates a limited use of the range of possibilities, and so organization. Therefore the first
hypothesis to be tested was the following:
1) When one examines the order in which family members speak, this order should differ from random if the family
is following repetitive patterns.
This hypothesis can be tested in several ways with this type of data. If the family is not following repetitive patterns, then
on an infinite run when any member spoke there would be a fifty percent probability that he would be followed by either of
the other two members. When one examines the behavior of an actual family and finds, for example, that father is followed
by mother more often than by child, then if that difference is greater than the count which would occur if their behavior was
random, it has been demonstrated that the family is following patterns which repeat. To put the matter another way, if
families were not following patterns, then a frequency count of these six possible categories would show equal use of them
because it would be a toss of the coin who followed whom.
The second question being asked in this research can also be answered with this type of data. Families can be
differentiated from one another on a common base line by creating a scale of deviation from randomness. Phrased in terms
5
_____________________________________________________________________________________________________________
of a hypothesis, we have the following:
2) If one creates a scale with the zero point being that distribution of the six sequences which would occur if the
behavior was random, then every family will deviate from that zero point and each family will differ from another
on the basis of how far they extend away from the zero point.
If father, mother, and child followed each other randomly, the distribution of the six sequences on a frequency count
would be equal: Each of the six sequences would occur one-sixth of the time on an infinite run. When a frequency count is
made of the use of each of the six sequences in the conversation of a particular family, the count will differ from the equal
distribution. Some families should tend toward a more equal distribution, and some families toward a more unequal
distribution, and so families should differ from each other.
The third basic question was also phrased in terms of a hypothesis:
3) Organization means limitation, and the more pathological the more limited. Therefore on this frequency count the
more normal families will tend to use more of the possible sequences more often, and the disturbed families will use
fewer of the possibilities and use some of them more often than others. Therefore on a scale of deviation from
random behavior, the normals will tend toward randomness and the disturbed will tend away from randomness.
This hypothesis was logically derived as well as produced by observation of disturbed families. A family which is
disturbed appears to be more "rigid." Certain sequences in the interaction occur again and again, often despite the valiant
efforts of a family therapist to encourage a greater range of sequences. Mother and child, for example, will develop an
entente and shut father out, with his cooperation, so that each interchange between father and child is broken up by a
mother-child interchange. This hypothesis is essentially similar to a possible way of looking at neurosis in the individual:
the more emotionally handicapped the person the more narrow the range of activity he exhibits in comparison with normal
individuals.
One could expect that in the middle class American culture where children are encouraged to express themselves it
would be "normal" for all members to participate about equally. The child would be as likely to speak after father as mother
would. (In contrast, there is a myth that in European families the child speaks when spoken to, which would induce an
unequal distribution for "normals.") If the members participated equally, they would tend toward random on this frequency
count. Of course the sequence would be affected by the subject matter of the conversation. If they were discussing
something only the parents knew about, the child would naturally participate less. Therefore it was essential for this
measurement that the subject of conversation be one on which all family members had a right to speak.
The final question was also phrased in terms of a hypothesis:
4) When a family falls in the disturbed range on the scale of deviation from random behavior and is then treated
successfully by family therapy, the family will move toward a normal range.
This is a count of verbal participation, and the distribution of speeches becomes skewed when one family member
follows the other two members unequally. It is in the nature of Family Therapy to encourage more equal verbal
participation by family members. For example, if father is shut out, the therapist tends to work toward bringing him more
into the conversation with wife and with child. Presumably if therapy is successful, all dyads in the family are occurring
with more equal frequency and therefore the family will move toward randomness on this measurement.
RESULTS
The data from the interviews with eighty families was processed and a variety of frequency counts were made: the
frequency each individual spoke, the frequency with which one person followed another, and the frequency of longer
sequences of speech, i.e. sequences of three, four, five, six and seven speeches. The results will be presented in relation to
the hypotheses being tested.
1) To answer the question whether we can demonstrate that famines families follow repetitive patterns, it was
hypothesized that if the order in which family members speak differs from that order which would occur with random
behavior then the family is following patterns which repeat.
When we compute the frequency with which father speaks after mother and child, mother speaks after child and father,
and child speaks after mother and father, we can arrive at a percentage of the use by each family of the six possible
categories: MF, CF, CM, FM, MC, FC. If the family members were behaving randomlyfor example if three robots were
tossing coins to see who spoke nextall of these six categories would occur equally on an infinite run. That is, each
category would occur 16.66 percent of the time on an infinite run because it would be equally likely that any one person
6
_____________________________________________________________________________________________________________
would follow any other. Shorter runs of random responses would not all show a distribution of one-sixth in each category
but would fluctuate. If we take the equal use of the six categories, or a distribution of one-sixth each, as a zero point, then
randomness would be indicated by a fluctuation away from that zero point. If the actual count of a sample of families
distributed near that zero point, or the equal use of the six categories, it could be said that family members were speaking
randomly and not following patterns. When the actual distribution of the eighty families in this study is plotted, as shown in
Figure 1, it can be seen that they compose a fairly normal curve around a mean deviation from equal use of the six
categories of 24.31.3 Therefore the distribution differs from that which would occur if the family members were behaving
randomly and the first hypothesis is supported.4
FIGURE 1
2) The second hypothesis suggested that if one created a scale with the zero point being a random distribution, then every
family will deviate from that zero point and any one family can be contrasted with any other on this common base line.
When the 80 families are ranked in order of deviation from an equal use of the six categories, it can be seen that any
family can be contrasted with any other family in reference to a point common to them both. This is evident in Figure 3
where the families are placed in rank order with the height of each histogram indicating that family's deviation from the
common zero point.
3) The third hypothesis proposed that normal families would use more of the possibilities more often and therefore
would tend toward randomness on this scale, and disturbed families would use some possibilities more often than others
and so tend away from randomness.
The results with 80 families show that the normal group tends toward the random end of the scale, using each of the
categories more equally, and the abnormal group tends away from the zero point, showing a more skewed distribution of
the six categories. The two groups differ significantly at the .00003 level5 and the hypothesis is supported. Figure 2
diagrams the differences between the two groups. In Figure 3 the distribution is shown with the deviation from equal use of
each category by individual families represented by the height of each column. In Table I the 80 families are grouped into
normal and disturbed and ranked in order of increasing deviation from randomness. The table also gives the presenting
symptoms of each of the abnormal families.
7
_____________________________________________________________________________________________________________
FIGURE 2
Examination of Table I and Figure 3 shows that 10 abnormal families fall below the median family (the dotted line in
Table I), or in the "normal" range, and 10 normal families fall above that point in the "abnormal" range. One would expect
to find normal families falling in an abnormal range on almost any measurement because of the way these families are
selected. The criteria does not exclude abnormality, it merely excludes abnormality which has come to the attention of the
community. Many of the normal families might contain an undiagnosed disturbed member, and many might ultimately
develop a disturbed member but have not as yet. It is of interest that of the 10 abnormal families falling in the "normal"
range, four of them contain a psychotic member while the remaining five families with a psychotic member fall in the
abnormal range.
FIGURE 3
When this approach was conceived, it was assumed that this simple count of six categories would show differences
between normal and abnormal families, but that counts of larger patterns would show greater differences. Therefore the
8
_____________________________________________________________________________________________________________
family data was processed for computer analysis. Interestingly, this elementary count of sequence of two speeches
differentiates abnormal from normal families more sharply than more complex counts. A count of three speeches, such as
FMC, FMF, and so on, where there are twelve possible sequences reveals less difference, but a significant difference,
between the two types of families on this scale. Similarly, patterns of 4, 5, 6, and 7 speeches provide significant but less
differentiation. Table II shows the mean differences of these various sequences.
Table 2
R Dev Means6
Normal Abnormal Difference C. R.
Individual speeches 9.27 14.47 5.20 3.82 (P = < .00009)
Sequence of two 23.24 38.39 15.15 4.42 (P = < .00003)
Sequence of three 64.18 83.15 18.97 3.64 (P = .00023)
Sequence of four 26.97 171.42 44.45 4.40 (P = < .00003)
Sequence of five 283.09 364.00 80.91 4.39 (P = < .00003)
Sequence of six 677.05 837.36 160.31 3.87 (P = .00009)
Sequence of seven 1803.63 2103.43 299.80 3.20 (P = .00058)
It would appear that the two-count provided by the Family Interaction Analyser offers sufficient answers to the
questions posed, without the elaborate work necessary for computer preparation.
4) The final hypothesis suggested that if a family falls in the abnormal range on this scale and then is successfully treated
with family therapy, the family will move toward the normal range on this scale.
Sufficient data is not yet available to test this hypothesis. Figure 4 represents six families who were tested before
treatment and then were given exactly the same test with the same interviewer six months later. The first four families were
treated by training therapists and their supervisor reported independently that he felt none of these families had undergone
any basic change. The results here support that conclusion; all the families moved toward the normal range of the scale, but
only slightly. Family A, a family with a formerly psychotic mother, originally fell in the normal range and changed only
from an R Deviation number of 6.32 to one of 5.06. Families B and C also appear to have changed only slightly and remain
in the abnormal range. Family D moved thirty points toward the normal range but remained abnormal (the change
apparently represented father participating more in arguments with mother). The fifth family, Family E, was treated by a
more experienced therapist and therapy was being recessed, but not completed at the time of the second test. (The parents
were going to Europe and leaving their 23 year old son behind, a change which would seem represented in the higher
mother-father pairing on the second test.) Family F fell in the "normal" range originally and was also, treated by a more
experienced therapist. The "normal" range originally and was also treated by a more experienced therapist. The child has
improved sufficiently to leave the hospital, but little change is shown on this measurement; the family moved only three
points on the R Deviation scale. Until families with more clearly successful treatment are measured, this hypothesis cannot
be adequately tested.
9
_____________________________________________________________________________________________________________
FIGURE 4
To determine whether families change with therapy, it is necessary to consider how normal families change on test and
re-test without the intervention of a therapist. At this point six normal families have been re-tested and the results are
shown in Figure 5. The families change only slightly less than the group of treated families. Taking the mean percentage
change in each category, the normals change 2.36% and the treated families 3.15%. change 2.36 and the treated families
3.15. The change in R Deviation score shows the normal mean change to be 5.64 and the treated families 9.64. Whether
this difference is significant is difficult to determine on so small a sample. It is also possible that normal families represent a
different species from abnormal families and therefore their fluctuation on test and re-test might not be comparable to the
results with abnormal families who might be more "rigid." Ideally one should compare abnormal treated families with
abnormal families who have not been treated. This raises the sampling problem of finding abnormal families who are not
undergoing treatment. However, the test and re-test results on these few families indicate the possibility of measuring
therapeutic change in families with an instrument and therefore with high reliability.
10
_____________________________________________________________________________________________________________
FIGURE 5
We do not know as yet whether families which have undergone "successful" therapy will change on this measurement,
but it does appear that the measurement picks up a family pattern which is surprisingly consistent over time. Three of the
normal families and two of the abnormal families do not change on re-test more than two points on this scale, and the
percentage change in each category is slight for several of the families. As a group the 12 families change only 2.75 in each
category and the mean change in deviation from random is 5.73. One would expect greater fluctuation on the basis of error,
chance, or the disposition of family members on those particular days.
Other Computation
When this research was begun it was predicted that abnormal families would show an unequal distribution of these
categories, but it was not expected that they would show any particular pattern for a type of family. For example, it was not
assumed that there would be a greater frequency of mother-child interchange7 in abnormal families because the mothers are
said to be overprotective and the fathers passive. The differences in interchange between the two groups is shown in
percentages in Table III.
Table 3
FC CF FM MF MC CM
Normal 14.95 15.34 16.91 15.58 18.15 17.94
Abnormal 14.99 14.74 19.34 19.59 15.55 15.78
Normal Abnormal
Father-child 9 9
Mother-child 17 12
Father-mother 14 19
In the normal group the mother-child interchange is highest and father-child is least, while in the abnormal group the
mother-father interchange is highest and the father-child is least.
A similar distribution occurs if one merely counts the dyads which have the greatest frequency of speeches, as shown in
Table III.
The father-mother interchange in the abnormal group is greatest twice as often as the father-child interchange. Although
this means that in the abnormal group father and mother talk after each other more often than father and child, just what this
difference means is difficult to determine. It might mean that father and mother are "closer" than father and child, or it might
11
_____________________________________________________________________________________________________________
mean quite the opposite. Insofar as the families are assigned a task in the interview and talk until that task is accomplished,
it could be argued that frequency of pairs of speeches indicates disagreement, because with agreement the task and the
conversation end. Therefore a more frequent interchange between two people could mean greater disagreement between
them. Lacking a non-inferential measure of "disagreement," we can only note the difference at this point.
Individual Participation
It might be argued that the significant differences found between the two groups on the interactional measurement are
merely a result of an abnormal member participating less in the conversation. Even if the disturbed member did speak less
than someone in the normal group, and granting that individual participation influences group participation, the individual
emphasis does not seem to be the most useful way to examine this kind of data. The argument that individual behavior
should be given the important emphasis is usually either a product of past tradition or a belief that individual behavior is the
result of some innate, perhaps organic, characteristic. Yet it is equally possible to argue that what the individual does is a
response to what the other individuals are doing. The muteness of a disturbed child can be seen as a product of the ways the
parents deal with him and he with them, and so a measure which includes his muteness is measuring the habitual
operational patterns in that family. What the individual does is not separable from what the other two individuals are doing
and so a measurement which implies independent individual behavior is doubtfully legitimate. As an exaggerated example,
one might note that mother speaks less often in a particular family and hypothesize that this is a result of her character or
personality. Yet when one examines the conversation in which she speaks, it could be noted that whenever father turns to
speak to her the child cuts in and speaks to father and when child turns to speak to her father cuts in and speaks to child.
When the frequency of mother's speech is totaled, it could appear that she is more unresponsive than father or child, but to
measure her "unresponsiveness" as if it is independent of the behavior of the others is a distortion of the situation.
However, in response to the argument that the behavior of the identified patient "causes" the difference between the two
groups of families, two sorts of computation can be presented. Table IV shows the percent distribution of individual
speeches.
Table 4
Father Mother Child Total Speeches
Normal 31.87 34.85 33.28 37,502
Abnormal 34.32 35.14 30.52 35,932
In the normal group the mother speaks most often and the father least often, while in the abnormal group the mother
speaks slightly more often and the child least often. The two groups do not differ significantly on this measurement since the
child in the abnormal group does not speak significantly less than the father in the normal group?8
A further breakdown shows that of the 36 families in which the presenting problem was one member who was
considered the identified patient, that person spoke an average of 29.80% of the time, speaking least in 20 out of the 36
cases (52.94%). In the 33 families where the child was the identified patient, he spoke an average of 30.05% of the time,
speaking least in 18 out of the 33 cases (54.55%). In contrast, the person speaking least often in the normal group was the
father, and he spoke least in 18 out of the 40 cases (45.00%). It would appear that the identified patient tended to be the
person who speaks least in the abnormal group, but in about half the cases the person who spoke least was not the
identified patient.
In summary, if one raises the question whether there is a more unequal distribution of frequency of speech by individuals
in the abnormal group than there is in the normal group, this is so. Taking as a base line a random distribution of speech, or
each of the three family members speaking 33.33% of the time, the normal and abnormal group of families differ
significantly (as shown by the "individual speeches" in Table II). However, this difference does not seem attributable to the
behavior of the identified patient. It would appear that whatever the measurement, individual or interactional, if one
approaches the data from the point of view of a deviation from the equal use of all possibilities, then normal and abnormal
families will differ. As individuals the abnormals participate more unequally, as dyads they show an unequal distribution,
and larger frequency measurements show a similar skewness.
Questionnaire vs. TAT

The results reported to this point have included the total conversation which was composed of both a conversation about
a questionnaire and a conversation about three sets of TAT pictures. When the two parts of the total conversation are
examined separately, somewhat different results are obtained. The questionnaire conversation provides essentially the same
results as the total conversation; normal and disturbed groups differ with a P of .00003 and there are 11 abnormal families
below the median in the "normal" range and 11 normal families above the median. Examining only the TAT conversation,
12
_____________________________________________________________________________________________________________
one finds that normal and disturbed groups differ with only a P of .005 and there are 15 abnormal families below the
median and 15 normal families above. It is possible that this different result with the two conversations is a difference in the
nature of the conversation since the questionnaire involves achieving agreement on a specific list of items while the TAT
conversation is more free flowing. However, it is also possible that an extraneous influence intruded; the TAT conversation
was always presented to the family last and so an individual fatigue factor may have entered. One could only find this out by
alternating the conversations.
When individual families are examined, it is found that some of them follow exactly the same pattern on both
conversations as well as the total of the two. In fact some families follow the same pattern from the first four minutes
throughout the conversation. Other families show different patterns on the questionnaire and TAT conversations. This
variation is indicated by the fact that of the 11 normal families who fall above the median on the questionnaire, only four of
those families also fall above the median on the TAT conversation. Of the 11 abnormal families falling below the median
on the questionnaire, eight of them also fall there on the TAT conversation. There is one normal and one abnormal family
which falls out of place on the total of the two conversations but not on either one separately. That is, the distribution is not
unequal on each conversation but shows that pattern when the two conversations are totaled. The reverse case, where the
family was skewed in each part of the conversation but fell in the "normal" range when the two parts were combined,
occurred only once and that was with a normal family.
Since it is possible that the more a family deviates from a random distribution the more "rigid" their family patterns, it
should follow that those families which deviate furthest from random would have similar patterns on both questionnaire and
the TAT conversations. Conversely, those families which approach most closely to a random distribution should differ
more often in the two conversations, showing greater "flexibility." This hypothesis can be tested by taking as the same
pattern the occurrence of the same dyad with the greatest frequency of interchange in the questionnaire and TAT
conversation. That is, if the mother-child interchange is greatest in both conversations the family is following the same
pattern. If the dyad with the most frequent interchange is different in the two conversations, the pattern can be said to be
different. Examining the distribution in this way, we find that of the families below the median, those tending toward
random, 30 families follow different patterns and 10 follow the same pattern, while of those above the median, tending
away from random, 12 families are different and 28 follow the same pattern in the two conversations. The difference is
significant at greater than the .001 level (although a question can be raised whether these are independent measures).
When the two types of conversation used here are combined, a conversation of some length is achieved, with the families
averaging 918 exchanges. This would seem a sufficiently long conversation to obtain typical family patterns. The reliability
of the measurement is indicated not only by test and re-test similarities but by the small amount of change which occurs
when additional families are added to the sample. The computations were first made for 30 normal and 30 abnormal
families. When an additional 10 normal and 10 abnormal families were added to make a sample of 80 families, it was found
that no measurement changed as much as one percent.9
CONCLUSION
This report has presented the results of a study of 80 families as part of an ongoing investigation to develop
non-inferential measurements of patterns of family interaction. The general theoretical approach is based upon the premise
that organization means limitation; if responses are organized, then of all possible responses some will be used more than
others. Therefore on any measurement it should be possible to classify ongoing organizations against the results which
would occur if the members of the organization were behaving randomly. This approach provides a range, or common base
line, from which families can be classified into types.
The particular measurement chosen as a first step in this approach was the most simple interactional measurement
conceivable: the order in which three family members speak with a frequency count of which person follows another. It is
possible to make this frequency count with an instrument and thereby provide high reliability. The results support the
hypotheses proposed and the sample is large enough to indicate confidence in the method. Apparently this measurement
demonstrates that families follow patterns, differentiates one family from another on a common base line, differentiates
most "normal" from most "disturbed" families, and suggests that a measure of family change before and after therapy is
possible. A larger sample is being developed and the next more complex measurement which is non-inferential is being
attempted. Although the general theoretical approach is different from, and difficult to correlate with, the usual
psychodynamic descriptions of families, it appears to solve many of the problems which have handicapped family research
to date.
1Thanks are due to the NIMH Grant MH-06560-01, under whose auspices Ferreira and Winter initiated the project that led to
this fortunate collaboration.
2For the research reported here, the results were not entirely obtained from this analyzer and so human judgment was involved
and there is a possibility of error. The investigator listened to tape recordings of the interview and laid out the order in which the
13
_____________________________________________________________________________________________________________
members spoke. In most cases he combined the listening with observation of the visual record of the conversation from the event
recorder. This procedure was used partly because a difficulty developed when it was found that if family members sufficiently
increased their volume they fired each other's microphones, thereby creating a spurious count. It was also necessary in order to
make use of some interviews obtained in a different laboratory, and to prepare the data for computer processing for more elaborate
counts of the interchange. Currently the analyzer is connected to throat microphones which work on vibration and therefore family
members cannot fire each other's microphones and the count is automatic. The analyzer was also revised to include four persons so
that parents and two children could be tested as well as a therapist with a family of three.
3This deviation from an equal use of the six categories can be computed in several ways. A particular way was used in this study,
and throughout this report an index number of deviation from random, or equal use, of the six categories will be used. This will be
called the R Deviation number of a particular family and the means and medians discussed will be based upon these R Deviation
numbers. This index number is computed by as "zero" the equal use of the six categories, or 16.66% frequency of use of each
category. When a count is made of a particular family, the deviation in frequency of use of each category from 16.66% when
totaled (ignoring the sign) provides the R Deviation number for that family. In this way a scale is established with a common zero
point for all families. As a simplified example, if a family exhibits a 1% difference from 16.66% in each category, their R Deviation
number would be 6.00 computed as follows:
FC 17.66 - 16.66 = 1.00
FM 15.66 - 16.66 = 1.00
MF 17.66 - 16.66 = 1.00
MC 15.66 - 16.66 = 1.00
CM 17.66 - 16.66 = 1.00
CF 17.66 - 16.66 = 1.00
TOTAL  6.00
4The total group has a standard deviation of 12.50. The Critical Ratio for the difference between the obtained mean and zero, the
estimated mean of a random distribution, is 1736. The P value for the C.R. is infinitesimally small. This may be because the
hypothetical random distribution would be skewed rather than normal because it could not go below zero. However, the difference
is so great that it appears to support the hypothesis, since the effect of violation of the normal curve assumption here is to use a
standard deviation which is too large.
5Mean of the normal group = 19.15, with a standard deviation of 8.26. Mean of the abnormal group = 29.45, with a standard
deviation of 13.82. A one-tailed test was appropriate here.
6This computation was made for 60 rather than 80 families and the means were not computed in terms of a deviation from
16.66 in each category but according to a matrix in which each person would be followed by one of the other two 50% of the time
if the behavior was random. This figure represents a percentage deviation from that 50%. The two methods of computation provide
essentially similar results.
7This measurement tends to pick up an interchange between two people since it is based upon who speaks after whom and so
includes one person speaking to another person. However, it is also possible for this count to appear to be an interchange between
two people when it is not. For example, if mother speaks to child and he does not answer, and father speaks to child and he does not
answer, and mother speaks to child again, and so on, it would appear to be a conversation between mother and father on this count
when actually it was not. However, in these results any pair will be termed "an interchange" to avoid the more cumbersome
phrasing of one speaking after another.
8P = .11 (two-tailed test). Normal fathers spoke a mean of 31.87% of the time (standard deviation = 3.89). The 33 abnormal
children spoke a mean of 30.05% of the time (standard deviation = 5.58).
9Since these computations were made an additional 20 abnormal families have been tested and the mean R. Deviation score for
the group changed only from 29.46 to 29.79. More families with a psychotic child have been added and the group of 12 such
families differs from the normal group at a .005 level of significance with three families falling in the "normal range" and 9 in the
"abnormal range."
14

Family Process - March 1964 - HALEY - Research On Family Patterns An Instrument Measurement

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Family Process - March 1964 - HALEY - Research On Family Patterns An Instrument Measurement

Uploaded by

Copyright:

Available Formats

_____________________________________________________________________________________________________________

Fam Proc 3:41-65, 1964

Research on Family Patterns An Instrument Measurement

TYPES OF FAMILY RESEARCH

PROBLEMS OF CLASSIFYING INTERACTION

THE BASIC QUESTIONS

10.58 10 M 15.13 Psychotic child 25 M

THE THEORETICAL APPROACH

of a hypothesis, we have the following:

Questionnaire vs. TAT

You might also like