Measures and Methods of Assessment - Method

Patterns of Attachment: A Psychological Study of the Strange Situation - Mary D. Salter Ainsworth 2015

Measures and Methods of Assessment


In the analysis of data, three kinds of measures were used: (1) incidence of specific behaviors (or combinations of behaviors) in specific episodes, indicated by the percentage of infants who manifested each; (2) frequency measures; and (3) special scores for dimensions of interactive behavior. These measures were useful in describing normative trends across episodes. They proved less useful in representing individual differences in strange-situation behavior than another method of assessment—(4) classification of infants according to the patterning of their behaviors. The classificatory system that reflects the organization of behavior in relationship to the mother may be seen in Chapters 6, 7, 8, 9, and 11 to yield robust, stable, and psychologically significant assessment of individual differences.

Percentage Measures of Behavior

An important feature of the data analysis is to ascertain the percentage of babies in the total sample who show each specific behavior in each episode of the strange situation, or specific combinations of behaviors. It would be redundant here to list all the behaviors and behavioral combinations that were examined, for these are implicit in the discussion of findings. In general, the definitions of behaviors are implicit in the definitions included in the Appendices, but where any doubt might arise, the definitions are given in the context of the report of findings.

Frequency Measures

Two types of frequency measure were used. One was preferred for behaviors that were discrete and of brief duration; the other for behaviors of extended duration or for those that consisted of a continuous sequence of coordinated separate behaviors. The first type was a simple frequency count of number of times the behavior was emitted in a given episode (or, in the case of Episode 3, in each of three subepisodes). This type of measure was used for smiles and vocalizations. The second type of frequency measure was based on the 15-second time intervals into which the narrative records were divided. For each behavior or class of behavior measured thus, the score was the sum of the time intervals in any given episode in which the behavior occurred, whether it occured continuously, intermittently, or momentarily. The standard length of an episode was 3 minutes, and therefore the highest obtainable frequency score was 12. If an episode were either longer or shorter than 3 minutes, the frequency scores were prorated to make them equivalent to those obtained from a 3-minute episode.

It is obvious that the measure of frequency per 15-second interval is a substitute for precise measure of duration, which was not possible to obtain with our method of recording. (See Omark & Marvin, 1978.) The degree of error introduced by the use of this substitute was deemed tolerable, however, for behaviors like crying and exploration. In the separation episodes once crying had begun it tended to continue, and exploratory behavior once it had stopped tended not to recur. In the reunion episodes once crying had stopped it tended not to recur. In all episodes exploratory behavior tended to operate over continuous blocks of time rather than to be emitted in sporadic “bursts” like smiles and vocalizations.

As a preliminary step to obtaining the frequency measures, the behavior shown by an infant in each 15-second time interval in each episode was coded according to a coding system described fully in Appendix II. The specific kinds of behavior coded were as follows: locomotion, body movement, body posture, hand movements, visual regard, adult contact behavior, infant contact behavior, crying, vocalization, oral behavior, and smiling. The coding yields a condensed sequential description of the behavior of an infant in each episode of the strange situation. It also lends itself to ready tabulation of the absolute frequency of behaviors. The original coding system was more complex than the one now in use; it attempted to deal in more detail with the behaviors shown by the infant in interaction with others. It emerged, however, that simple frequency measures were inadequate to deal with the complexities of interactive behavior. Nevertheless, some vestiges of the coding of such behavior remain for descriptive purposes and to aid in the use of the coding sheet as a convenient index to the narrative protocol.

Detailed instructions for tabulation of the frequency coding are also given in Appendix II. The tabulation yielded the following measures for each episode of the strange situation.

Exploratory Locomotion. This refers to locomotion that is clearly in service of getting to the toys or some other aspect of the physical environment in order to explore it, or that seems undertaken for the mere sake of the activity itself. It differs from the measures of locomotion and mobility used by others (e.g., Cox & Campbell, 1968; Maccoby & Feldman, 1972) in that it excludes locomotion with social, interactive, or distress implications, such as approaching a person, following, searching for the mother during the separation episodes, moving away from a person to avoid her, or moving about while highly distressed. For this analysis, as well as for the other two measures of exploration, we used the frequency per 15-second interval. In the case of episodes curtailed because of crying, it was assumed that no further exploratory behavior—whether locomotor, manipulative, or visual—would have occurred had the episode lasted the full 3 minutes.

Exploratory Manipulation. This refers, similarly, to hand movements that are clearly exploratory, and is intended to exclude hand movements that are used in social interaction or in physical contact, or those that are expressive gestures, whether with communicative intent or not. The coding system attempts to distinguish between fine and gross manipulation. With 1-year-olds, however, this distinction was found to be difficult and seemingly of little significance. Therefore, the measure used in our analysis is the total measure of exploratory manipulation.

Visual Exploration. Like the other exploratory measures, visual exploration refers to exploration of the toys or other aspects of the physical environment. It excludes looking at persons or looking at objects that in the separation environment may be considered to be of interest because they are associated with the mother rather than for their own intrinsic interest—for example, the door through which the mother departed, the chair in which she previously sat, or the handbag that she left behind her.

Visual Orientation. There are three scores representing visual orientation: looking at the mother, looking at the stranger, and looking at the toys or at some other aspect of the physical environment. The third score is identical with the visual-exploration score. Within any episode or when comparing one episode with another, it is possible to ascertain what proportion of the baby’s visual attention goes to the physical environment in contrast to persons, and to the mother in contrast with the stranger.

For this analysis we used the frequency-per-15-second-interval measure, because looking at aspects of the physical environment tends, like other forms of exploratory behavior, to be of extended duration, even though looks at persons—especially looks at the mother—tend, like smiles and vocalizations, to be momentary and sporadic. In the episode most important for comparison of orientation to mother versus stranger—Episode 3—Bretherton and Ainsworth (1974) used absolute frequency of looks. This latter measure also introduced some distortion, because looks at the stranger were often stares and of longer duration than the glances at the mother. Nevertheless both measures of frequency yielded essentially the same findings when comparison of the two target figures was the issue.

Crying. A distinction has been made between “real crying”—which includes screaming and fussing as well as crying, and includes intermittent and isolated cries as well as continuous crying—and “minimal crying,” in which there is a cryface without vocalization or unhappy vocalization, or protest without a cryface. This distinction has descriptive utility, but in our analyses we have used a total crying score, including both. Babies who cry minimally in regard to intensity also cry relatively infrequently in the strange situation, so that little or no information is lost by using the total crying score. Here our measure was frequency-per-15-second interval. In the case of episodes that were curtailed because of crying, it was assumed that the baby would have continued to cry had the episode lasted for the full 3 minutes.

Smiling. Although most interactive behaviors could not be dealt with meaningfully in terms of frequency measures, smiling has been included among them, as indeed are looking (visual orientation), vocalization, and crying. There are three measures of smiling: smiles clearly directed toward the mother, smiles clearly directed toward the stranger, and total number of smiles including those not clearly directed toward any person. These three measures were expressed as the absolute number of smiles.

Vocalization. It is more difficult to ascertain whether a vocalization is directed toward a person than to judge whether a smile is so directed. Nevertheless, the frequency of vocalization was handled in the same way as the frequency of smiling—the number of vocalizations clearly directed to the mother, the number clearly directed toward the stranger, and the total number of vocalizations including those not clearly directed toward any person. These were expressed as the absolute number of vocalizations.

Oral Behavior. It is assumed that, in a 1-year-old, oral behavior in a situation such as the strange situation is unlikely to be either nutritive or exploratory. It is assumed to be “autoerotic,” tension reducing, or, in ethological terminology, a “displacement” activity that occurs in a situation in which two or more incompatible behavioral systems are simultaneously activated (Bowlby, 1969). Therefore, it is of interest to compare one episode with another in regard to frequency of oral behavior. To be sure, sucking and chewing are not the only kinds of autoerotic or displacement activities that occurred, but they were the only ones that occurred commonly enough to be represented in our frequency measures. This class of behavior was scored in terms of frequency per 15-second interval.

Intercoder Agreement in Regard to Frequency Measures

Throughout this study our major efforts toward reliability have been: (1) to have all measures based on coding undertaken by at least two coders working independently, with any discrepancies resolved in conference; and (2) on training of personnel. Under these circumstances, conventional assessments of intercoder agreement provide a very conservative estimate of the reliability of the measures actually used in the statistical treatment of the data. The coding of frequency measures is very straightforward. When discrepancies between coders arise, they are almost invariably due to carelessness on the part of one coder or the other and are easily corrected by checking back to the protocols themselves. Nevertheless, because Bell in her 1970 study (which yielded Sample 2) had to work alone, a check was made of the agreement between her coding and that of a then-more-experienced coder (MDSA). Eight cases were selected at random for the reliability check. Product-moment reliability coefficients for four frequency measures obtained by these two independent coders were as follows: exploratory locomotion, .99; exploratory manipulation, .93; visual exploration, .98; and crying, .99. Connell (1974) and Rosenberg (1975) independently coded the frequency measures for 23 subjects, counting as a “match” scores that were not more than one whole number apart for an entire episode. A reliability of .950 was obtained (459 matches/483 scores).

Scoring of Interactive Behavior

Although the frequency measures and the percentage measures play a role in the analysis of data, they are not adequate to represent certain of the more significant infant behaviors, especially those involved in interpersonal interaction. To be sure, the frequency measures include measures of various behaviors commonly assumed to be attachment behaviors—smiling, vocalization, and looking. These measures, whether considered separately or together, do not provide an adequate assessment of infant attachment behavior directed toward the mother, to say nothing of sociable behavior directed toward the stranger. The “percentage measures” of behavior are of descriptive value when dealing with normative trends across the various episodes of the strange situation, and these can and do deal with combinations of behaviors. Percentage measures do not, however, lend themselves to the kind of quantification that is useful when one wishes to assess the interrelationships between behavioral measures (as, for example, in multivariate analyses) or the interrelationships between strange-situation behavior and either infant behavior or maternal behavior at home.

One of the major difficulties in relying upon frequency assessments of behaviors implicated in interpersonal interaction is that (perhaps especially in the last quarter of the first year, or after a baby’s behavior has become goal-corrected—to rely on the concept of Bowlby, 1969) behaviors that are superficially quite different may have a certain degree of interchangeability. If this is the case, then assessments based on specific behaviors separately considered may well fail to reflect the true state of affairs.

Furthermore, in any naturalistic situation and even in our laboratory situation, in which the behavior of the adults (mother and stranger) was only partially controlled through instructions, the behavior shown by an infant toward another person can scarcely be assessed without considering the context provided by that person’s behavior. Such considerations wreak havoc with comparisons across individuals or across situations, unless one can find some way of taking into account the contingencies of interchangeability of behavior and of reciprocal behavior (or lack of it) in the partners in the interpersonal transactions in question.

These difficulties were particularly apparent in the reunion episodes of the strange situation in which the meager instructions to the mother permitted much latitude in behavior and in which interchangeability of behaviors was particularly conspicuous in the case of the infant. Furthermore, the need to overcome the difficulties and to make accurate assessment of infant “interactive” behavior in the reunion episodes was increased by our emerging conviction that it was precisely the reunion episodes that afforded the most discriminating occasion for assessment.

Let us examine some considerations. It seems significant, for example, to note whether a baby approaches his mother when she returns, and whether he does so immediately or only after a delay—or whether he does so only after an invitation from her or fails to do so despite such an invitation. It is of interest whether he merely seeks increased proximity, or whether he seems to want close physical contact. If the latter, he may clamber up on his own initiative, or he may merely reach as a signal to elicit his mother’s help in gaining contact. On the other hand, he may not approach at all, but merely redouble his cries when his mother enters, obviously wanting her—and indeed, in most such cases, the mother goes quickly to the distressed baby and picks him up without delay.

Ainsworth and Wittig (1969), reporting on the first 13 cases of Sample 1, did not attempt to devise measures of interactive behavior that might take into account the various combinations of infant and maternal behaviors. Rather they chiefly relied upon percentage measures. So did Bretherton and Ainsworth (1974), who focused on infant behavior toward the stranger. Ainsworth and Bell (1970), however, based their report of interpersonal behavior on a then-new set of measures that we report here.

The first step toward analysing individual differences in strange-situation behavior was to devise a classification of infants into groups showing similarities in the ways in which their behavior was organized. In the earliest attempt at classification (Ainsworth & Wittig, 1969), first attention was given to the presence or absence of distress in the separation episodes; but subsequently the reunion episodes were perceived to be more significant. In the course of refining the classificatory system (first reported by Ainsworth, Bell and Stayton, 1971, and described later in this chapter) it was possible to identify four dimensions of behavior that seemed crucial in distinguishing the various classificatory groups and subgroups.

These four dimensions were: proximity- and contact-seeking behavior, contact-maintaining behavior, avoidance, and resistance. Although behavior to the mother was of especial interest, the same dimensions were implicit in behavior toward the stranger. The scoring system therefore comprehended both and was therefore applicable to all episodes save Episode 1, which was merely introductory, and Episode 6, when the baby was alone.

For each of these four behavioral variables, the protocols of 56 infants (Samples 1 and 2) were examined, and “behavioral items” were extracted from them. A behavioral item consisted of all the behavior in an episode that each child showed relevant to each of the behavioral variables, as that behavior was directed to the mother or, in a separate record, to the stranger. Each behavioral item was typed on a slip of paper, ready to be sorted. Each variable was then dealt with separately, without regard for the episode in which it was shown or for whether it was directed toward mother or stranger.

First, for proximity and contact seeking, for example, the behavioral items were sorted into seven piles, each representing a point on the proximity- and contact-seeking dimension. The dimension was defined in terms of the degree of initiative and active effort implicit in the specific proximity- or contact-promoting behavior displayed. Point 7 included those behavioral items in which the baby had shown most active initiative in seeking proximity to an adult—approaching without delay and without needing to be invited, and approaching fully to make contact, clambering up on the adult without needing to elicit her cooperation. Point 1 included all items in which there seemed to be no overt effort to gain proximity and no behavior that seemed to be a clear-cut signal inviting the adult’s approach. Point 2 included behavioral episodes in which a baby made an “intention movement” toward a person—a slight and incomplete approach. Points 6, 5, 4, and 3 represented various degrees of active initiative as contrasted with mere signalling.

The considerations entering into the disposition of a behavioral item to one point or another included: (1) the degree of activity and initiative of the behavior; (2) promptness of the behavior; (3) frequency of the behavior; and (4) duration of the behavior. Thus, for example, active approaching was considered a stronger proximity-promoting behavior than either reaching or a directed cry signal. An immediate approach was considered stronger than a delayed approach. Several approaches, other things being equal, were considered stronger than one approach. Or, to turn to contact-maintaining behavior, it was considered stronger if contact lasted 2 minutes than if it lasted for 1 minute, or for only 30 seconds.

Furthermore, the sorting had to consider the behavior of the adult. Thus, for example, some mothers picked the baby up, whereas others did not; and those that did might do so either immediately or only after a delay. A baby whose mother was promptly responsive to his desire for contact could scarcely be given the highest score, which was reserved for infants who achieved contact entirely on their own initiative; on the other hand, though, his active initiative in eliciting maternal behavior had to be given due weight. Similarly, in regard to contact-maintaining behavior, highest scores were given to infants who persistently resisted attempts of the mother to put them down. But if the mother held the baby for a long period and attempted to put him down only after he was fully soothed and reassured, then he had little opportunity to resist release. In such instances it was assumed that the baby somehow signaled to his mother his desire for continuing contact while she held him, perhaps by clinging or by “sinking in,” and thus if the contact was indeed maintained for a prolonged period, the baby was given a high score despite his lack of opportunity to resist release.

The seven-point scales resulting from this sorting procedure are shown in Appendix III. The behaviors that define each point of each scale are drawn from the actual behaviors shown by the infants of Samples 1 and 2. It was clearly impossible to comprehend in the scale every degree and combination of behavior that might be met empirically in subsequent data collection. Therefore, when scoring the behaviors of Samples 3 and 4, the scorer might not find a precise match for the behavior of a given child in the behavioral definitions of the scale points, and would have to find the best match that he could or assign an interpolated score.

The initial system for scoring interactive behavior included a fifth behavior variable that has not heretofore been mentioned. Although search for the mother in the separation episodes had not entered into the classificatory system as a differentiating variable, it lent itself to scoring by the method described above. A sixth variable—distance interaction—was added to the scoring system much later. This variable did in fact enter into the classificatory system as distinguishing between subgroups, but had not been dealt with initially.

Although the instructions for scoring these six behavioral variables are presented fully in Appendix III, here we give a brief indication of what is covered by each.

Proximity and Contact Seeking. As previously stated, this variable refers to the degree of active initiative a baby shows in seeking physical contact with or proximity to another person.

Contact Maintaining. This refers to the degree of active initiative a baby exerts in order to maintain physical contact with a person, once such contact is achieved. The highest scores are given to infants who repeatedly resist release and who, as a consequence, succeed in maintaining physical contact throughout most of the episode in question. Resisting release implies intensified clinging when the adult attempts to put the baby down (or merely to shift his position), or turning back immediately to clamber up again when put down. Mere protest, without active effort to maintain contact, is scored lower.

Resistance. The highest scores are given to babies who persistently manifest intense angry and/or resistant behavior to an adult. The resistance is shown by pushing away from, striking out at, or squirming to get down from an adult who has offered contact, or by pushing away, throwing away, or otherwise rejecting toys through which an adult attempts to mediate interaction. The highest scores imply an obviously angry emotional tone, although in lower scores the resistant behavior may be seemingly without negative affect. Resistant behavior is not incompatible with proximity seeking or contact maintaining. An angry, resistant infant may nevertheless strongly seek to gain and to maintain contact, although such a combination suggests ambivalence.

Avoidance. As reported by Ainsworth and Bell (1970), some babies actively avoid proximity and interaction with their mothers in the reunion episodes, in which a common response is to seek closer proximity or contact. Highest scores are given to infants who persistently ignore their mothers, continuing to play without acknowledging mother’s return despite her effort to invite the baby’s approach. Somewhat lower scores are given to infants who mingle greeting responses with moving away, turning away, or looking away. Appendix III presents parallel scales for mother and for stranger, although it is not meant to imply thereby that the behavior has the same dynamics for these two figures.

Search. When a baby is separated from his mother, two major classes of attachment behavior may be activated—crying and search. Crying constitutes a proximity-promoting signal—one that instructions to the mothers tended to make ineffective, except insofar as the episode would be curtailed if the infant seemed intensely distressed. Crying behavior was assessed in terms of the frequency measures described in an earlier section. Searching for the mother did not lend itself to a frequency measure, however. Search behavior is defined as behavior in which the baby, through means other than crying, attempts to regain proximity to his mother. The strange-situation procedure prevents a baby from succeeding in his search; so the most active behavior that an infant can manifest is going to the closed door (through which his mother left) without delay, attempting to open it, and furthermore remaining oriented toward the door throughout most of the rest of the separation episode. Some babies, who were scored lower, show a toned-down version of this behavior by merely looking at the door either persistently or frequently. Others manifest search behavior by approaching (or merely looking at) the place associated with the mother—her chair.

Distance Interaction. Some babies who are clearly attached to their mothers do not show heightened proximity or contact seeking in the reunion episodes, but rather show heightened interest in interacting with the mother across a distance—smiling or “talking” to her, pointing to things of interest, showing her toys, or offering them to her across a distance. In 1-year-olds this kind of reaction is substantially less frequent than either proximity seeking or avoidance or resistance, whereas in 2-, 3-, and 4-year-olds it occurs more often (Blehar, 1974; Maccoby & Feldman, 1972; Marvin, 1972). Nevertheless, even in this study of 1-year-olds it seemed desirable to include among our scores of interactive behaviors a measure of infant initiative in distance interaction.

Interscorer Agreement for Measures of Interactive Behavior

The scoring of interactive behaviors is obviously more complex than the routine tabulations involved in coding the frequency measures, and requires more judgment on the part of the scorer. This being so, it was necessary to demonstrate that a satisfactory degree of interscorer agreement could be achieved. Several formal assessments of interscorer agreement are reported here.

First, 14 protocols were selected from among those of Sample 1 to provide a full range of the behaviors in question. The degree of agreement between two independent scorers for behaviors directed to the mother, assessed by reliability (rho) coefficients, was as follows: proximity- and contact-seeking behavior, .93; contact maintaining, .97; resistance, .96; avoidance, .93; and search, .94.

Second, interscorer agreement for the two sessions of the 24 subjects of Sample 3 was found for six pairs of graduate-student judges in the course of their training in strange-situation procedure. Each pair dealt with six protocols—somewhat too few to represent an adequate range of scores in all variables. Agreement was assessed for behavior both to the mother and to the stranger in each relevant episode separately. The median coefficients of agreement for five variables were as follows: Proximity and contact seeking, .90; contact maintaining, .85; resistance, .88; avoidance, .75; and search, .87.

Third, Main (1973) reports the agreement for two independent scorers of two behaviors directed to the mother for Sample 4: resistance, .94 and avoidance, .93.

Finally, Connell (1974) and Rosenberg (1975) independently scored the interactive measures for 23 subjects, using a statistic consisting of the proportion of matches (i.e., less than one whole number) for four of the scales (excluding search) for behavior directed at mother and stranger in all relevant episodes considered separately. The overall index of interscorer agreement was .876.

The measure of distance interaction was devised later than the other scales and was assessed separately for interscorer agreement. Protocols for the total sample of 106 infants were scored by one research assistant, but in the course of training and checking the scoring of that assistant, one of us (MCB) independently scored 92 protocols. The correlation between the two sets of scores was .85.

In conclusion, a satisfactory degree of interscorer agreement can be achieved for these scales. It may be noted, however, that training and experience is required before a high level of interscorer agreement can be achieved. In our opinion such training is particularly important for the scoring of avoidant behavior. In the case of Sample 3, which was scored by students in training, the degree of reliability of the scores is better than the coefficients of interscorer agreement would indicate, because the final scores used were settled in conference, with any interscorer discrepancies resolved.

Classification of Infants in Terms of Their Strange-Situation Behavior

As implied earlier, classification of infants in terms of the patterning of their behavior in the strange situation preceded the identification of the dimensions in terms of which their interactions with the mother and with the stranger could be described. Indeed, classification was the first procedure that we used to help us make sense of the enormously complex variety of behaviors manifested by 1-year-olds in interaction with their mothers in the strange situation.

Contemporary psychologists tend to be biased against classificatory procedures, deeming them not only subjective and hence unreliable but also dedicated to a belief in a more or less rigid typological concept of the way in which human behavior is organized, with implications of discontinuity in the various quantitative dimensions that may be implicit in the description of types or categories. It is not in this spirit that we offer a classificatory system. On the contrary we view it as a first step toward grasping the organization of complex behavioral data. We agree with Hinde (1974) that one must first describe and classify when one sets out to study natural phenomena:

Description and classification may not seem very difficult tasks, but their neglect hampered many aspects of psychology for half a century. . . . This descriptive phase, essential in the development of every science, was bypassed by those experimental psychologists who attempted to model their strategies on classical physics. These workers overlooked the fact that classical physics was a special case in that its subject matter—falling apples, the apparent bending of sticks in water, floating logs—were everyday events, so that the descriptive phase was part of common experience, and not especially a job for the scientists. Of course, the way people behave is also part of everyday experience, but to describe behavior precisely is much more difficult than appears at first sight (p. 5).

Although Hinde was directing his comments toward the classification of behavior—for example, classifying behaviors into systems in terms of their common outcome—his comments also seem relevant to classification of individuals in accordance with the different patterns of behavior they manifest in comparable situations. He goes on to caution us, however, that there are limitations to classificatory systems. If pressed far enough, they do not work. The categories are tools, not “absolutes.”

Two ways in which a classification of patterns of behavior can be a useful tool are: (1) to identify the chief behavioral dimensions in terms of which the groups so classified differ; and (2) to raise the issue of how such disparate patterns of behavior happened to develop. The identification of dimensions of difference between classificatory groups may lead to a quantification of such dimensions and hence to a more precise description of individual differences in terms of new variables not previously considered relevant. This was indeed the case in the present study, for the primary step of classification led to the development of the measures described in the preceding section.

As for the second issue, one’s immediate response to a perception of patterning of behavior is to entertain the notion that there is some good reason that the component behaviors should be interrelated in this way—perhaps because they tend to serve the same function or have a common cause or interacting sets of causal influences. When faced with a species-characteristic patterning, the answer may be that this is the nature of the beast—which merely pushes the questions of why and how into a search to understand phylogenetic development, or the processes of evolution. When faced with a perception of patterns of individual variations, one can also attribute them to basic genetic differences between individuals. Because classificatory systems—usually termed typologies—have indeed appealed to ingrained genetic predispositions as explanation for patterning, it has become all too common to believe that this is characteristic of all. Another approach, which also should be explored, is to examine the developmental histories of the individuals in question for common antecedent experiences that may be hypothesized to have an influence in the development of similar patterns in one group of individuals that distinguish them from other groups of individuals who have other patterns. This is the way we have chosen to approach the issue of patterning. (The reader is referred to Chapters 7 and 8, in which differences in maternal and infant behavior at home during the first year are related to differences in strange-situation behavior patterns at the end of that year.)

Because the qualitative distinctions implicit in our classificatory system yielded significant quantitative dimensions, it might be asked why classification is not supplanted, instead of being merely supplemented, by quantitative measures. There are three reasons why we have chosen to retain classification. The first, and perhaps most important, is that the definitions of classificatory groups retain the picture of patterns of behavior, which tend to become lost in—or at least difficult to retrieve from—the quantification process. (This becomes evident in Chapter 6, which deals with the discriminant-function analysis of the three main classificatory groups.)

Second, we would be foolish to believe that the dimensions that we have so far subjected to quantification take into account all the behaviors that are important components of the patterning of individual differences in strange-situation behavior not only in this sample but also in other samples drawn from other populations. Although there may be themes common to a wide variety of samples, our sample of 106 1-year-olds cannot exhaust all possible variations on the common themes. A classificatory system can remain flexible, with the possibility of refining classificatory criteria in the light of further knowledge or, indeed, the possibility of elaboration in order to accommodate new patterns into new groups or subgroups. To abandon the classificatory system in favor of our present set of component behavioral scores (or in favor of the discriminant weights, which is discussed in Chapter 6) would freeze our knowledge in its present state.

Third, let us return to the issue of understanding why and how the patternings arose. We believe that the patternings described and differentiated within a classificatory system keep this issue to the forefront rather than burying it in a welter of refined statistics. It is perhaps a reflection of our own modest level of expertise that we so believe, but it seems that a preoccupation with measurement may lead one to forget that this is a tool and not an end in itself. Because the classificatory system was a useful tool in achieving a beginning of understanding, we wish to retain it in the belief that it will continue to be a useful tool in future research, supplemented but not supplanted by quantitative efforts. Ainsworth and Wittig (1969) made the first, relatively crude classification of 1-year-olds’ strange-situation behavior. When examining individual patterns of the first 13 infants of Sample 1, they found the presence or absence of separation distress to be the most conspicuous way in which differences might be ordered. They used it, therefore, in classifying infants into three loose groups. Not wishing to assign descriptive labels at this point, they called the groups A, B, and C. Group-A infants showed minimal disturbance in the separation episodes, whereas Group-B infants were distressed by separation. Even in this first attempt at classification, one dimension (i.e., crying in the separation episodes) proved insufficient, for Group-C babies were distinguished from Group B, even though they, too, were distressed by separation, in terms of the “maladaptive” nature of their behavior.

After 10 more infants had been added to Sample 1, it seemed wise to undertake a more careful and systematic classification. Working with the protocols of the total sample of 23 infants, we began by grouping infants whose behavior in all episodes was alike in as many respects as possible. This purely empirical exercise yielded seven clusters of infants. Similarities between the clusters were then sought, resulting in three main groups of infants, which we again designated as Groups A, B, and C. The seven clusters were retained as subgroups—two in each of Group A and Group C, and three in Group B.

The most conspicuous feature of the new Group A was avoidance of the mother in the reunion episodes, although many other behavioral features were common to the infants so classified, including minimal crying in the separation episodes. Only one of the infants classified by Ainsworth and Wittig (1969) in Group A was retained in the new Group A, together with five of the new subsample. The four infants originally classed in Group C were retained in the new Group C, and none of the new subsample were so classified. Instead of the loose designation of “maladaptive,” it was now perceived that Group-C infants shared, in addition to strong interest in proximity to and contact with the mother in the reunion episodes, a tendency to manifest angry resistance to the mother upon reunion. The new Group B consisted of 13 infants, including three previously classified by Ainsworth and Wittig in Group A, five previously classified in Group B, and five of the new subsample. Although not all of the new Group-B infants were distressed in the separation episodes, they shared a manifest interest in gaining proximity to and contact (or at least interaction) with their mothers in the reunion episodes, without evidence of antithetical behaviors, such as avoidance or resistance. The criteria for classification were then prepared in detail, on the basis of the behavioral similarities (and differences) that had led to the original sorting into subgroups and groups.

Before giving these criteria, however, two further steps must be reported. When Bell (1970) had observed her 33 subjects in the strange situation (Sample 2) she applied this new classificatory system to them. It proved applicable in all cases, except that she specified a further subgroup of Group B (Subgroup B4) in order to accommodate three of her sample. The same classificatory system was later used without change for the classification of infants in Samples 3 and 4. All but one of the 50 additional infants were accommodated in existing subgroups—and she was eventually classified in Group A, without being assigned to a subgroup. It may be noted that in the criteria for classification that follow, the emphasis has shifted from behavior in the separation episodes to behavior relevant to the mother in the reunion episodes—not because of preconceived theoretical convictions but because behavior in the reunion episodes contributed the most convincing evidence of clustering behaviors, in contrast to a continuous distribution along one or even two major dimensions.

Criteria for Classification

Group A:

  • —Conspicuous avoidance of proximity to or interaction with the mother in the reunion episodes. Either the baby ignores his mother on her return, greeting her casually if at all, or, if there is approach and/or a less casual greeting, the baby tends to mingle his welcome with avoidance responses—turning away, moving past, averting the gaze, and the like.
  • —Little or no tendency to seek proximity to or interaction or contact with the mother, even in the reunion episodes.
  • —If picked up, little or no tendency to cling or to resist being released.
  • —On the other hand, little or no tendency toward active resistance to contact or interaction with the mother, except for probable squirming to get down if indeed the baby is picked up.
  • —Tendency to treat the stranger much as the mother is treated, although perhaps with less avoidance.
  • —Either the baby is not distressed during separation, or the distress seems to be due to being left alone rather than to his mother’s absence. For most, distress does not occur when the stranger is present, and any distress upon being left alone tends to be alleviated when the stranger returns.

Subgroup A1

Conspicuous avoidance of the mother in the reunion episodes, which is likely to consist of ignoring her altogether, although there may be some pointed looking away, turning away, or moving away.

If there is a greeting when the mother enters, it tends to be a mere look or smile.

Either the baby does not approach his mother upon reunion, or the approach is “abortive” with the baby going past his mother, or it tends to occur only after much coaxing.

If picked up, the baby shows little or no contact-maintaining behavior. He tends not to cuddle in; he looks away; and he may squirm to get down.

Subgroup A2

The baby shows a mixed response to his mother on reunion, with some tendency to greet and to approach, intermingled with a marked tendency to turn or move away from her, move past her, avert the gaze from her, or ignore her. Thus there may be moderate proximity seeking, combined with strong proximity avoiding.

If he is picked up, the baby may cling momentarily; if he is put down, he may protest or resist momentarily; but there is also a tendency to squirm to be put down, to turn the face away when being held, and other signs of mixed feelings.

Group B:

  • —The baby wants either proximity and contact with his mother or interaction with her, and he actively seeks it, especially in the reunion episodes.
  • —If he achieves contact, he seeks to maintain it, and either resists release or at least protests if he is put down.
  • —The baby responds to his mother’s return in the reunion episodes with more than a casual greeting—either with a smile or a cry or a tendency to approach.
  • —Little or no tendency to resist contact or interaction with his mother.
  • —Little or no tendency to avoid his mother in the reunion episodes.
  • —He may or may not be friendly with the stranger, but he is clearly more interested in interaction and/or contact with his mother than with the stranger.
  • —He may or may not be distressed during the separation episodes, but if he is distressed this is clearly related to his mother’s absence and not merely to being alone. He may be somewhat comforted by the stranger, but it is clear that he wants his mother.

Subgroup B1

The baby greets his mother, smiling upon her return, and shows strong initiative in interaction with her across a distance, although he does not especially seek proximity to or physical contact with her.

If picked up, he does not especially seek to maintain contact.

He may mingle some avoiding behavior (turning away or looking away) with interactive behavior, but he shows little or no resistant behavior and, in general, seems not to have feelings as mixed as an A2 baby.

He is likely to show little or no distress in the separation episodes.

Subgroup B2

The baby greets his mother upon reunion, tends to approach her, and seems to want contact with her, but to a lesser extent than a B3 baby. Some B2 babies seek proximity in the preseparation episodes, but not again until Episode 8, and then perhaps only after some delay.

The B2 baby may show some proximity avoiding, especially in Episode 5, but this gives way to proximity seeking in Episode 8, thus distinguishing him from the A2 baby.

Although he accepts contact if he is picked up, he does not cling especially, and does not conspicuously resist release.

On the other hand, he shows little or no resistance to contact or interaction, and in general shows less sign of mixed feelings than A2 babies.

He tends to show little distress during the separation episodes.

He resembles a B1 infant, except that he is more likely to seek proximity to his mother.

Subgroup B3

The baby actively seeks physical contact with his mother, and when he gains it he is conspicuous for attempting to maintain it, actively resisting her attempts to release him. Most B3 babies show their strongest proximity-seeking and contact-maintaining behavior in Episode 8, but some do so in Episode 5 and are so distressed in the second separation episode that they cannot mobilize active proximity seeking and resort to signaling. Occasionally, a baby who seems especially secure in his relationship with his mother will be content with mere interaction with and proximity to her, without seeking to be held.

At the same time, the B3 baby may be distinguished from other groups and subgroups by the fact that he shows little or no sign of either avoiding or resisting proximity to or contact or interaction with his mother.

He may or may not be distressed in the separation episodes, but if he shows little distresss, he is clearly more active in seeking contact and in resisting release than B1 or B2 babies.

Although his attachment behavior is heightened in the reunion episodes, he does not seem wholly preoccupied with his mother in the preseparation episodes.

Subgroup B4

The baby wants contact, especially in the reunion episodes, and seeks it by approaching, clinging, and resisting release; he is, however, somewhat less active and competent in these behaviors than most B3 babies, especially in Episode 8.

He seems wholly preoccupied with his mother throughout the strange situation. He gives the impression of feeling anxious throughout, with much crying. In the second separation, particularly, he seems entirely distressed.

He may show other signs of disturbance, such as inappropriate, stereotyped, repetitive gestures or motions.

He may show some resistance to his mother, and indeed he may avoid her by drawing back from her or averting his face when held by her. Because he also shows strong contact-seeking behavior, the impression is of some ambivalence, although not as much as is shown by Group-C infants.

Group C

  • —The baby displays conspicuous contact- and interaction-resisting behavior, perhaps especially in Episode 8.
  • —He also shows moderate-to-strong seeking of proximity and contact and seeking to maintain contact once gained, so that he gives the impression of being ambivalent to his mother.
  • —He shows little or no tendency to ignore his mother in the reunion episodes, or to turn or move away from her, or to avert his gaze.
  • —He may display generally “maladaptive” behavior in the strange situation. Either he tends to be more angry than infants in other groups, or he may be conspicuously passive.

Subgroup C1

Proximity seeking and contact maintaining are strong in the reunion episodes, and are also more likely to occur in the preseparation episodes than in the case of Group-B infants.

Resistant behavior is particularly conspicuous. The mixture of seeking and yet resisting contact and interaction has an unmistakably angry quality and indeed an angry tone may characterize behavior even in the preseparation episodes.

Angry, resistant behavior is likely to be shown toward the stranger as well as toward the mother.

The baby is very likely to be extremely distressed during the separation episodes.

Subgroup C2

Perhaps the most conspicuous characteristic of C2 infants is their passivity. Their exploratory behavior is limited throughout the strange situation, and their interactive behaviors are relatively lacking in active initiative.

Nevertheless in the reunion episodes they obviously want proximity to and contact with their mothers, even though they tend to use signaling behavior rather than active approach, and protest against being put down rather than actively resist release.

Resistant behavior may occur, particularly in Episode 8, but in general the C2 baby is not as conspicuously angry as the C1 baby.

Interjudge Agreement in Classification

The classificatory system was established on the basis of the strange-situation behaviors exhibited by the 23 subjects of the main project. As a result of applying the classificatory system to her 33 subjects, Silvia Bell proposed the addition of another subgroup to Group B—the B4 subgroup. After the classificatory system had been expanded to accommodate this proposal, one of us (MDSA) classified all Bell’s subjects, without any knowledge of these subjects beyond the strange-situation protocols themselves. In regard to 31 infants there was virtually total interjudge agreement. One of the disagreements was indeed minor; one judge placed the baby in Subgroup B1 and the other in Subgroup B2. The other disagreement was more substantial, although obviously the protocol in question was difficult to classify. One judge placed the infant in Group A, whereas the other did not classify him at all.

A second test of interjudge agreement was undertaken for Sample 3 in conjunction with the test of interscorer agreement discussed earlier, in which students were trained in the use of the strange-situation procedure. For this assessment we consider only Session 1. The judges were divided into two teams, one of four and the other of three students. The members of each team were independently to classify 12 of the 24 subjects and then meet in conference to resolve any discrepancies and to decide on a final classification. If we take the conference classification as the criterion, there was 96% agreement with the criterion for the seven subjects finally placed in Group A, 92% agreement for the 15 finally placed in Group B, and 75% agreement for the two placed in Group C.

As might be expected, classification into subgroups was accomplished with somewhat less agreement. There was 100% agreement with the criterion for the two infants placed in Subgroup A1, and 76% agreement for the five placed in Subgroup A2. Most of the discrepancies between judges involved classifying as A1 infants finally placed in A2; the rest involved classifying them as B1. One infant was finally placed in Subgroup B1 with 100% agreement, four in B2 with 79% agreement, nine in B3 with 87% agreement, and one in B4 with only 50% agreement. Most of these discrepancies were between B2 and B3. One baby was finally placed in C1 with 100% agreement, and one in C2 with only 50% agreement. The discrepancies here were nearly all between C2 and B4.

These findings suggest that the classificatory system can be used with highly satisfactory reliability by experienced judges and with somewhat less agreement among less experienced judges. We suggest that training is necessary before a high degree of reliability of classification can be achieved. The final classification of Sample 3, achieved after the resolution of discrepancies in conference, is undoubtedly a closer match to the criteria for classification than the foregoing discussion of interjudge agreement would suggest.

The smallest group, Group C, presented the most difficulty to the judges, not only in Sample 3 but also in other samples. If a baby’s behavior precisely matches the specifications of one of the subgroups, there is likely to be a high degree of interjudge agreement, even though the judges have not previously been acquainted with that pattern of behavior. But the classificatory specifications in many cases can be only guidelines; and under these circumstances judges build on the experience they have gained with previous cases, as though they had to see several infants in each classificatory group (and subgroup) before they “get the feel” for the range of variation covered by each. However—as in the case of Group C and Subgroup B4 in Sample 3—if there are too few cases that match the specifications even approximately, then difficulties and disagreements are especially likely to arise.