Bowlby's theory

Most ground-living primates seek an “attachment figure” when frightened, rather than simply fleeing, attacking, or freezing as do the majority of mammals. This given, Bowlby emphasized the likely existence of a universal biologically based behavioral program that would lead human infants raised in all but extremely anomalous circumstances to form an attachment to a parental figure during the first 3 years and usually by the end of the first year of life (Bowlby, 1969/1982). Given the strong relation between maintenance of proximity to an older conspecific and survival for infant primates, including humans, infants and young children were expected not only to maintain reasonable proximity to the “attachment figure” (usually but not always the biological parent) in unthreatening circumstances, but also greatly increase efforts to gain proximity or contact to that figure in times of alarm. Thus, the normal functioning of the attachment system should include its marked activation when cues to threat or danger are present. A corollary of Bowlby's proposal is that frightening behavior on the part of the parenting figure should therefore lead to disruptions in the functioning of this system under stress (Main & Hesse, 1990).

To clarify this line of reasoning, it is necessary to consider our phylogenetic heritage. Bowlby pointed out that most primates differ from other animals (including most mammals) in that gaining proximity to a protective conspecific, as opposed to a place (e.g., a den or burrow) provides our primary solution to situations of fear (Bowlby, 1958, 1969/1982). Thus, for the human infant and many other ground-living nomadic primates it is not a place, but rather a selected individual who provides the primary “solution” to conditions that elicit fright. Because this propensity is biologically based, when frightened, the attached young human will seek proximity to the attachment figure even if, paradoxically, that figure is the source of its fear.

At first glance an inherent propensity to approach the location of alarm of course appears irrational. However, the powerful influence of phylogeny and the sometimes unusual outcomes it can produce is well illustrated by Darwin's observations regarding the Galapagos sea lizard. Darwin noted that this animal, able to move about with ease on both land and sea, exhibited the peculiar behavioral feature that, when frightened, it would not enter the water. To further his understanding of this phenomenon, Darwin repeatedly threw a sea lizard into the ocean. Strikingly, although possessed “of perfect powers of diving and swimming,” the sea lizard invariably returned in a direct line to the promontory of land on which its “attacker” (Darwin) stood. Darwin (1839/1972, pp. 334–335) speculatively solved this conundrum by consideration of the animal's evolutionary history or “phylogeny”:

Perhaps this singular piece of apparent stupidity may be accounted for by the circumstance, that this reptile has no [natural] enemy whatever on shore, whereas at sea it must often fall a prey to the numerous sharks. Hence, probably, urged by a fixed and hereditary instinct that the shore is its place of safety, whatever the emergency may be, it there takes refuge.

Returning to human evolution, parallel to Darwin's sea lizard, the human infant alarmed by its attachment figure, and constrained by its phylogeny, has no inherent (i.e., instinctively organized) means for separating the location of its attacker from the location of its haven of safety. Confronted with circumstances unanticipated within its evolutionary history, then, it should experience strongly contradictory propensities regarding solutions to threat. A natural correlate of this condition is, then, that anomalous forms of alarming parental behavior, including as the example par excellence direct maltreatment in the form of physical abuse, should lead to a malfunctioning of the attachment system, created by simultaneous and contradictory infant propensities to approach, and to flee from, the (same) attachment figure. This should create a “looping” of behavior and attention, because the source of alarm and haven of safety are one and the same. As noted, this appears to be a situation involving “fright without solution” (Main & Hesse, 1990, 1992) or more specifically, fright taking the alarming behavior of the phylogentically established “haven of safety” as its source. Repeated experiences of fright without solution should be inherently disorganizing, and, effecting emotion and attention as well, cannot be “resolved” at the behavioral level. As such, it is not surprising if experiences of this kind, when repeated, increase the individual's vulnerability to psychopathology (Main & Hesse, 1990, 1992; see also Liotti, 1992, and Sroufe, Egeland, Carlson, & Collins, 2005).

Thus, we proposed some years ago that, observed with the parent in stressful settings such as the Ainsworth Strange Situation procedure, maltreated infants would become disorganized and disoriented (see Main, 1981; Main & Weston, 1982). A first test of our hypothesis regarding the probable role of fear of the parent in creating D infant behavior under stress was conducted by Cicchetti and his colleagues, and independently by Lyons-Ruth and her colleagues. Here, as predicted, a very strong majority of maltreated infants were indeed found disorganized with the parent in the Strange Situation (Carlson, Cicchetti, Barnett, & Braunwald, 1989; Lyons-Ruth, Connell, Zoll, & Stahl, 1987). D behavior in low-risk samples where the majority of disorganized infants have presumably not ordinarily been maltreated had, however, yet to be explained.

D infant attachment: A collapse of attentional and behavioral strategies under stress

By 1990, infant response to the Ainsworth Strange Situation had been videotaped in many laboratories investigating low-risk and well as high-risk dyads. This well-known procedure introduces infants to several of Bowlby's “natural clues to danger,” namely an unfamiliar environment, a stranger, and two separations from the parent (Ainsworth et al., 1978). One secure and two insecure (insecure–avoidant and insecure–ambivalent) “organized” responses to the parent during the Strange Situation had been identified by the late 1960s and related to maternal behavior toward the infant in the home (e.g., Ainsworth, Bell, & Stayton, 1971). Later, Main (1990) emphasized the role of organized (flexible/secure vs. inflexible/avoidant or inflexible/ambivalent) behavioral and attentional strategies within these patternings. In conjunction with her 1973 doctoral thesis, Main had already noted the occurrence of “conflict” behaviors in some insecure infants, a finding twice replicated by 1981 (Main & Stadtman, 1981; these behaviors had been termed “disorganized/disordered” by 1979). Main and Solomon later undertook an exacting description of these behaviors, creating a fourth, “disorganized/disoriented” (D) infant Strange Situation category based upon what often appeared to be a collapse in behavioral and attentional strategy (Main & Solomon, 1986, 1990).

Infants are placed in Main and Solomon's (1986, 1990) D category if, in the parent's presence during the Strange Situation, they display any of an array of D behaviors, including substantially misdirected movements, contradictory movements, stilling or even freezing of all movement for 20 s or longer, stereotypies and anomalous gestures and postures, and direct indices of confusion or apprehension (for a summary of exemplar behaviors, see Table 1; for a more complete overview of the D category and its correlates see Hesse & Main, 2000). The D classification is always assigned together with a secondary, best-fitting “organized” classification, for example, D/secure, or D/insecure–avoidant.

Disorganized/disoriented behavior observed during the Strange Situation

Some of the sequelae to early D attachment status with the mother include unfavorable outcomes that do not reach the clinical level, for example, parent–child role inversion, “catastrophic” stories told in response to pictured parent–child separations, and bizarre/frightening elements seen in family drawings by age 6 (Main, Kaplan, & Cassidy, 1985; each of these findings have been replicated in other laboratories). Infant D also predicts later affect dysregulation (see Bakermans-Kranenburg, van IJzendoorn, & Juffer, 2005, for review), and in an interesting series of case studies, difficulties in peer relations (Jacobvitz & Hazen, 1999). Finally, young adults D with mother in infancy are found to be having the most difficulties overall in couple relationships, and to be the most hostile to their partners (Sroufe et al., 2005).

However, infant D also predicts varying forms of psychopathology (van IJzendoorn et al., 1999; see also Bakermans-Kranenburg et al., 2005 for references to more recent “clinical” sequelae). For example, following a suggestion made by Liotti (1992), Carlson (1998) found infant D attachment status with mother predictive of dissociative behavior in elementary through high school settings, and in the DSM-like K-SADS interview at age 19, a finding that the author termed “unpredecented.” Infant D also predicts externalizing and internalizing disorders in both high-risk and low-risk samples. (Interestingly, however, and at least in preschoolers, these disorders may include a substantial dissociative component, as recent studies by Macfie, Cicchetti, & Toth [2001a, 2001b] have demonstrated.) Most recently, infant D attachment has also been found predictive of self-mutilating (self-injurious) behaviors in young adulthood (Sroufe et al., 2005). Strikingly, despite several rigorous investigations (e.g., Lakatos et al., 2000), genetic contributors to D have yet to be successfully established (see Bakermans-Kranenburg & van IJzendoorn, 2003; Bokhorst, Bakermans-Kranenburg, van IJzendoorn, Fonagy, & Schuengel, 2003).

Earlier it was noted that, given their repeated experiences of “fright without solution,” a high percentage of maltreated infants had been found disorganized within the Strange Situation. However, a strikingly substantial minority of infants in low-risk samples (averaging 15%, but ranging to 33%; van IJzendoorn et al., 1999) are also judged disorganized. Although some infants in low-risk samples are undoubtedly maltreated, proportions ranging to 33% are unlikely, and an explanation for the appearance of D Strange Situation behavior in such samples initially seemed problematic. Surprisingly, the first clues to identifying forms of FR parental behaviors other than physical abuse emerged in the finding that disorganized Strange Situation behavior in low-risk samples was linked to unusual linguistic responses made by parents of D infants to queries regarding loss (or abuse) experiences during a semistructured interview, the AAI (see George et al., 1984, 1985, 1996; Hesse, 1999).

U/d attachment status as identified on the AAI: Lapses in reasoning or discourse during the discussion of loss or abuse experiences

The AAI focuses upon a participant's description and evaluation of early childhood experiences (particularly, relationships with each parent or primary caregiver) and their effects upon current personality and functioning. Like the Strange Situation, most AAI transcripts differ systematically in ways which permit classification into one of three “organized” categories. As is the case with the Strange Situation, AAI transcripts can be viewed as reflecting differences in the flexibility (secure–autonomous) or inflexibility (insecure–dismissing or insecure–preoccupied) of the speaker's attention (Hesse, 1996; Main, 1993, 2000). Each adult attachment category is considered equivalent to, and, in parents, has repeatedly been found predictive of, the corresponding “organized” category of infant Strange Situation response (see van IJzendoorn, 1995, for a meta-analysis regarding the “organized” adult to infant links). However, as is the case for infant Strange Situation behavior, disorganizations in language usage during this interview also sometimes appear.

Global disorganization of interview texts does occur, especially within high-risk samples, in which case the texts are termed “cannot classify” (CC; see Hesse, 1996). In low-risk samples disorganization most frequently takes a more restricted form termed a “lapse in the monitoring of reasoning or discourse” appearing during the discussion of loss (including miscarriage; see Bakermans-Kranenburg, Schuengel, & van IJzendoorn, 1999; as well as stillbirth; see Hughes, Turton, Hopper, McGauley, & Fonagy, in press) or abuse. These lapses can and usually do occur in otherwise well-organized texts, and, frequently appearing in only a few sentences, are used to identify “U” states of mind (see Table 2). Such transcripts are termed unresolved/disorganized or “U/d.” As with the infant D system, a second best-fitting alternate “organized” category is always assigned, for example, U/secure or U/preoccupied. For many years, we have attributed these brief failures in reasoning or distortions in discourse to brief collapses or failures in working memory, which can, importantly, be negatively effected by fright (Arnsten, 1998).

Identifying unresolved/disorganized attachment status in the Adult Attachment Interview

Examples of lapses in the monitoring of reasoning include statements implying that the lost person is simultaneously dead and alive in the physical sense. This in turn suggests an intrusion from a “segregated system” wherein a dead person is also held to have remained alive. The concept of “segregated systems” was utilized by Bowlby (1980), following upon Hilgard's (1977/1986) work concerning the existence of “multiple controls on action” found in some hypnotized subjects. A real-life exemplar was described in the case of “Geraldine,” a young woman who occasionally entered wandering “fugue” states that Bowlby attributed to the reemergence of a “segregated system” that believed her deceased mother was still alive, and, searching for her, had taken executive control. The subtle reasoning slips occasionally found within the AAI indicating that a dead person is believed alive are usually brief, are often discovered in individuals who are high functioning, and normally appear to go unnoticed by the speaker. The momentary intrusion of this secondary “system” of belief is most likely elicited by the persistent questioning regarding loss found within the AAI and is currently believed unlikely to result in actions approaching extremes such as that described in the case study above.

Examples of lapses in the monitoring of discourse often consist in indices of high-level absorption that appear to have drawn the speaker out of a speech register appropriate to the interview context, including extreme and inappropriate attention to detail, sometimes accompanied by funereal speech (for a summary of exemplar statements, see Table 2). Absorption is considered to be a relatively normative aspect of dissociation, indicative of diminished awareness of the surround. A direct link between this index of an “U/d” state as assessed in the AAI and Tellegen's Absorption Inventory (Tellegen & Atkinson, 1974; Tellegen, 1992) was first uncovered in two independent samples by Hesse and van IJzendoorn (1999; see Gribneau, 2005, for replication).

Linking parental U to infant D attachment status: Cross-generational (parental U/d to infant D) links are continuing to be identified, and infant D is beginning to predict U/CC status on the AAI

By the late 1980s AAI texts classified as U/d for loss and/or abuse were found predictive of infant D attachment to the speaker in two independent samples of mother–infant dyads living in poverty (van IJzendoorn et al., 1999; see also Carlson, 1990). At the same time, U/d attachment status in parents was also found associated with infant D attachment within a low-risk sample (Main & Hesse, 1990). These findings were replicated in a second low-risk sample 1 year later (Ainsworth & Eichberg, 1991). By 1995, nine studies of the U to D link were included in a meta-analysis, yielding an overall moderate effect size (d = .64, r = .31; van IJzendoorn, 1995), while an extremely strong relation was found between the effect size and the amount of training the coders had had in identifying infant D attachment status (as well as its mid-childhood “D-controlling” equivalent; see Main & Cassidy, 1988). At least 10 further studies of the relations between parental U attachment status and infant D attachment status have been conducted, and in several of the most recent studies the effect size has been remarkably higher (e.g., Abrams, Rifkin, & Hesse, 2006; Behrens, 2005; φ = .56 and .59, respectively). Among the recent studies with which we are familiar, an association has failed to appear in only one, when the infants were 18 months of age (Lyons-Ruth, Yellin, Melnick, & Atwood, 2005). The expected U to D link was however significant for this same sample at 12 months of age (Lyons-Ruth, Yellin, Melnick, & Atwood, 2003).

It should be noted that, whereas this presentation focuses upon the prediction of D attachment status with a given parent from circumscribed U lapses observed in the same parent's AAI, a second path from the AAI to infant D (and/or infant CC, a fifth infant Strange Situation category being developed in our laboratory) is also viable, and is currently being explored. Hesse's (1996) identification of a fifth, CC adult attachment category is based upon “global disorganization” found throughout the interview, usually, but not always, readily identified via evidence for conflicting (e.g., both dismissing and preoccupied) discourse strategies found throughout the text as a whole. Like U/d adult attachment status, albeit in only a few cases and studies, CC to date appears to predict infant D/CC behavior toward the CC speaker within the Strange Situation (Ammaniti & Speranza, 1994; Hughes & McGauley, 1997; Jacobvitz, personal communication, 2005). Interestingly, in two case studies (Hughes & McGauley, 1997; Minde & Hesse, 1996) CC status on the AAI has predicted sharply contradictory, hence most likely unpredictable and FR, behaviors toward the infant in the home.

The above set of findings refer to the relation between the parent's U attachment status and D attachment status in their offspring. However, two laboratories are now able to examine the association in a different way entirely, considering the follow-up AAI status of subjects who had been seen in the Strange Situation with the mother 18 to 25 years previously. The two samples are different in important ways, one consisting of 40 dyads observed in our own low-risk Bay Area study (e.g., Main et al., 1985; Main, Hesse, & Kaplan, 2005), and the other consisting of well over 100 dyads observed in the Minnesota poverty sample (e.g., Sroufe et al., 2005; Weinfield, Whaley, & Egeland, 2004). In both samples, infant D attachment status with the mother was recently found significantly predictive of U and/or CC AAI status (the latter, again, signals a more global disorganization within the transcript; see Hesse, 1996).

Systems for identifying FR parental behavior

Following our 1990 study (Main & Hesse, 1990), we began to develop our system for identifying FR parental behavior. It presently consists of six subcategories, including dissociative behavior as well as frightened and threatening behavior as the three subcategories expected to lead directly to infant disorganization. Independently and simultaneously, two friends and colleagues began drafting similar systems. All three systems successfully predict infant D behavior.

One, developed by Jacobvitz, based directly upon the anecdotes and theory provided within our 1990 chapter, was applied to a large sample of infant–mother dyads (p < .001, two-tailed test; Hesse & Jacobvitz, 2003) differing from the sample included in this special section. The second (described initially by Lyons-Ruth, Bronfman, & Atwood, 1999) was intended to capture a wider group of “atypical” parental behaviors, and included the entire early set of items within our system (Main & Hesse, 1992), which Lyons-Ruth, Bronfman, and Atwood (1999, p. 52) termed FR+. This system included five subcategories, one exclusive to Lyons-Ruth et al.'s thinking (“affective errors”), two initially composed entirely of FR items (“disorientation” and “negative-intrusive behavior”), and the final two including a mix of FR items and other forms of atypical maternal behavior (“role confusion” and “withdrawal”).

Although Jacobvitz left her early data unpublished, moving to the use of the 1995 version of the FR system within a second sample (Jacobvitz, Leon, & Hazen, 2006), Lyons-Ruth (1999) have continued to develop their system of “atypical/disruptive” maternal behaviors into a system termed AMBIANCE, inclusive of (but still separately crediting) items from the FR system. Despite the obvious overlap, the three systems are by no means identical, and their independent success in predicting infant disorganization underscores the absence of any “method bias” interfering with tests of the essential hypothesis.

FR-like behaviors not expected to produce alarm in the infant: Normative expressions of fear, anger, and mild absorption

There are several common frightened (fearful), threatening (angry or disciplinary) and pseudodissociative (mildly absorbed) parental behaviors that would not be expected to produce disorganization. For example, fearful parental expressions can indicate impending danger (e.g., the appearance of a large aggressive looking dog; the toddler beginning to move toward oncoming traffic). In the former case, it is the mother's perceptions of potential “outside” threat that lead to expressions of fear, while in the latter it is the infant's actions themselves that she understands to have immediately dangerous consequences. In both of these examples, however, what is alarming, that is, the source of the alarm, is external to the parent. The alarming stimulus will therefore ordinarily be both discernible and comprehensible. This will be made obvious in the parent's orientation and action and, in addition, the infant will usually be free to approach the parent. Further, when most parents accidentally do something that frightens the infant, they are likely to take immediate actions that provide comfort and contact, or (in clinical terms) “repair” (see especially Lyons-Ruth & Jacobvitz, 1999; Lyons-Ruth, Bronfman, & Parsons 1999). Another well-known example of parental expressions of “fright” that alter infant behavior without leading to disorganization are found in Campos' studies of infant “visual cliff” performance, where infants are observed to monitor and respond to parental expressions of apprehension or alarm as the infant approaches what appears to be a dangerous “edge” by inhibiting further movement (see Kermoian & Campos, 1988; Klinnert, Campos, Sorce, Emde, & Svedja, 1983).

Threatening parental behavior often arises in ordinary contexts. Thus, it is not unusual for a parent to become angry and/or threatening in disciplinary interactions, for example, when the child does something impulsive and dangerous (as above), or fails to respond to parental requests for compliance. At such times, the parent may harshly raise their voice, or even spank the child. Here again, however, the origin or stimulus for the parent's behavior is both external to the parent and is comprehensible. In addition, via compliance the infant or child should to able to terminate the “frightening” parental behavior. Moreover, as just noted, the child will ordinarily be able to (or even encouraged to) approach the parent. This, too, provides an opportunity for “repair.”

Finally, all people engage in expressions of behavior that could, but do not, in fact, represent dissociation, or anomalous forms of absorption. The completely unmoving and “glazed” eyes of someone in a trance has an entirely different quality than the slight “blankness” (accompanied by slight eye movements) that may appear on a parent's face as they enter a brief daydream or try to remember everything they need to write down on a shopping list. Given that the former suggests an abnormal alteration in consciousness, it is likely to be alarming, whereas the latter are simply everyday examples of “drifting” or mildly absorbed minds. Thus, in the situations just described, conditions leading to parental expressions of fright, threat, or pseudodissociation are discernible and comprehensible, and the expressions themselves are not anomalous.

The six subcategories of parental FR behavior: Primary (threatening, frightening, and dissociative) and secondary (timid/deferential, sexualized, and disorganized)

Our system for identifying parental FR behaviors is divided into six subcategories, which appear in two sets. The first constitutes three “primary” behavior subcategories that are expected to be alarming in themselves. As noted at the outset of this presentation, we believe that the parent exhibiting FR behavior in the infant's presence is often likely to be experiencing fear themselves. This given, it is unlikely to be coincidental that this primary set of behaviors parallel the three “classic” mammalian forms of response to fright which are (e.g., Nijenhuis, Vanderlinden, & Spinhoven, 1998), respectively, attack (parental threatening behavior), flight (parental frightened behavior), and freezing (dissociative behavior, which often takes the form of trancelike stilling).

The remaining three (“secondary”) FR behavior subcategories are not necessarily alarming in themselves, but have been observed in a substantial proportion of the parents of D infants. They have clear connections to the primary forms of FR behavior named above in that, for example, “timid/deferential” behavior may appear most frequently in parents who are overtly frightened (particularly, of the infant) at other times, whereas both “sexualized” and “disorganized/disoriented” adult behavior would seem most likely to appear in parents prone to entering dissociative states. The full set of FR behaviors are provided with examples in Table 3.

Precis of the six central categories of the System for Coding FR Behaviors

Parental behavior likely to be directly frightening and/or disorganizing: Dissociative behavior, anomalous threat, and anomalous expressions of fright

Unlike the “normal” frightened, threatening, and pseudodissociative parental behaviors noted briefly above, parental behaviors that originate in internal factors related to unresolved experiences of trauma and fear should normally disrupt the regulated functioning of the offspring's attachment behavioral system. The central reason for this outcome is that the parent's psychological state will most likely be “altered” when the behaviors occur. As will be discussed below, this in itself is cause for infant alarm.

“Secondary” behaviors likely to be associated with U parental states: Timid/deferential, sexualized, and D behavior

The final three subcategories within in the FR coding system most likely would not in themselves lead directly to an approach–flight paradox for the infant. Nonetheless, they each imply alterations in normal consciousness. This should, in principle, increase the likelihood that the anomalous behaviors capable of directly producing disorganization will appear at other times.

Studies utilizing the FR system

To date, four published studies have utilized Main and Hesse's (1992–2006) FR coding system. Three were conducted in home settings, with the fourth consisting in a laboratory observation of videotaped free play separated by 1 week from the Strange Situation procedure. All home observations were conducted from 1 week (generally, several weeks) to over 1 year prior to the Strange Situation.

The pioneering study investigating our hypothesis regarding probable relations between U, D, and FR parental behavior was undertaken by Schuengel and colleagues (Schuengel, 1997; Schuengel, van IJzendoorn, & Bakermans-Kranenburg, 1999). Schuengel made two visits to the homes of 85 low-risk mothers who were known to have experienced the death of someone important to them, videotaping behavior with their 10- to 11-month-old infant for 2 hr per visit. The mothers were asked to go about their day as usual., and the videotapes of the resulting 4 hr of maternal and infant behavior were studied for FR behavior. The AAI was conducted at 12 months of infant age, and the Strange Situation at 14–15 months.

In this study, intercoder reliability was established over 16 cases for the FR scale as a whole (r = .83). Reliability for dissociative behavior was r = .73, as was reliability for threatening behavior (r = .73). Frightened behavior had to be combined with a similar scale presently termed timid/deferential behavior to obtain satisfactory reliability (r = .68). One-tailed t-tests showed that while overall FR behavior in the home did predict infant disorganization (p < .05), the subscale for dissociative behavior predicted infant D the most strongly (p < .01). Frightened maternal behavior predicted infant disorganization at a trend level when combined with the “minor” but conceptually similar subscale for timid–deferential behavior.

Maternal U attachment status predicted FR behavior only when the mother's alternative best-fitting attachment category was insecure (dismissing or preoccupied), and for these dyads, maternal FR behavior, as expected, mediated the relation between maternal U and infant D. However, U mothers whose alternative attachment category was secure did not show substantial FR behavior, and FR did not mediate the relation to infant D. In Jacobvitz et al.'s (2006) study, U/F mothers also showed significantly less FR behavior than U/non-F mothers. However, in contrast to Schuengel, van IJzendoorn, and Bakermans-Kranenburg's (1999) study, both U/F and U/non-F mothers in Jacobvitz et al.'s study showed significantly more FR behavior than parents falling in the “organized” AAI categories.⁴

A second study using the FR system was conducted with 44 dyads in the Dogan ethnic group in Mali, West Africa (True, Pisani, & Oumar, 2001), with the sample consisting of 27 dyads from a rural town setting and 15 from two different agrarian villages; infants ranged from 10 to 12.5 months at the first of the five assessments. Mother–infant interactions were videotaped during two brief well-baby exams, and for two more “natural” periods in which the infant, always in near proximity to the mother, or else carried, was bathed, or nearby as the mother cooked. Subscales were not utilized, and interrater reliabilities for the home and field observations for FR behavior across 10 observations was r = .79, whereas for the very brief well-baby exams FR reliabilities for 13–24 observations ranged from .71 to .84. Strange Situations were conducted at an average of 12.5 months in semiopen westernized buildings, separated from the previous assessment by 1 to 4 weeks. A majority of the infants were secure, but one quarter were disorganized. Although the level of maternal FR behavior observed was low, disorganization was nonetheless predicted by maternal FR behavior (p < .01). No babies were classified as insecure–avoidant. Mean FR scores for the three C (insecure–resistant/ambivalent) babies were calculated separately, and did not differ from those for B (secure) babies.

In a third study, Jacobvitz et al. (2006) administered the AAI to 116 mothers in a low-risk sample during the mother's third trimester of pregnancy. Infants were videotaped interacting with their mothers at 8 months, and mothers were “directed” to complete a set of normal interactive tasks with their infants, including a diaper change, feeding, and play for about 45 min in total. Half the videotapes were co-coded by two raters, who conferenced with a third coder in the case of substantial disagreement. The intraclass correlation between the two first raters was .68, and subscales of FR were not utilized. Mothers' U AAI status strongly predicted FR behavior during interactions with the infant (p < .000). Using one-tailed tests, U/insecure mothers were observed to be more frightening than U/secure mothers, but both were significantly more frightening than non-U mothers. For loss of alternative attachment figures, U mediated the relation between loss and FR behavior, but for loss of parents it did not.

In the fourth study, Abrams et al. (2006) conducted Strange Situations for infants at 12 or 18 months of age. Parent–infant interaction was observed during a 30-min free play session for 75 dyads (50 mother–infant dyads, 25 unrelated father–infant dyads, i.e., not married to the mothers) 1 week later. These research participants were part of the Bay Area Social Development Project, and observations were made in a 30-min laboratory play session separated from the Strange Situation by 1 week. Two coders established agreement across 30 of the 75 videotapes. Intraclass reliability for the FR scale overall was .80. Reliability for the threatening subcategory was .70, for the dissociative subcategory it was .75, and for the frightened subscale it was .70. Among the “secondary” FR subcategories, frequencies were too low for computation, with the exception of the disorganized subscale, which was r = .85 (for all four subscales, p < .001). Thirty-two of the parents in this study had participated in the AAI when their offspring had reached 6 years of age, and had experienced loss or other trauma prior to the play session.

Both paternal and maternal FR status was found strongly related to infant D attachment status (p < .001, d = 1.26). As in the case of Schuengel's (1997) study, the dissociative subscale was the best predictor of infant D (p < .001, d = 1.12), whereas the threatening subscale was also significantly predictive (p < .01, d = 0.80). In contrast to Schuengel, van IJzendoorn, and Bakermans-Kranenburg's (1999) study, the frightened subscale did not predict infant disorganization. For the subsample of 32 parents where AAIs were available (n = 32), FR behavior was found significantly associated with U attachment (φ = .59, p < .001). This was the first study to use FR classifications (ratings of 5.5 or above, with ratings of 5 left to the discretion of the coder) as well as scores. None of the parents of the 28 B babies were classified as FR. Only 1 parent of the 28 A babies and no parents of the 3 C babies were classified as FR. Thirteen of the 23 parents of D infants were classified as FR.

Studies utilizing AMBIANCE/FR+

Main and Hesse (1990) founded their speculations entirely upon anecdotal parental behaviors observed inside the Strange Situation itself. Perhaps for this reason, Lyons-Ruth's first test of AMBIANCE/FR+ was conducted within the Strange Situation, and to our knowledge, studies published to date utilizing this system have analyzed maternal behavior either within the Strange Situation, or in the first few minutes following.

AMBIANCE presently includes five subcategories of “atypical/disrupted” maternal behavior: affective communication errors, role-boundary confusion, withdrawal, intrusiveness/negativity, and fearful/disoriented/disorganized/dissociative behavior. The AMBIANCE system carefully marks items taken from the FR system, and all subcategories excepting “affective communication errors” contain elements of the early (1992) FR system (Lyons-Ruth, Bronfman, & Parsons, 1999). Among its five subscales, the third, titled fearful/disoriented (or more fully, “fearful, disoriented, disorganized, and dissociative parental behavior” is derived exclusively from the FR system, containing not only FR-fearful (frightened) items but also FR-dissociative and FR-disorganized items. Intrusiveness/negativity is based largely on early FR descriptions of threatening behavior, while withdrawal includes timid/differential behavior as well as some elements of dissociated behavior (e.g., Goldberg, Benoit, Blokland, & Madigan, 2003).

In constructing AMBIANCE/FR+, Lyons-Ruth, Bronfman, and Atwood (1999) and Lyons-Ruth, Bronfman, and Parsons (1999) reasoned that in addition to direct FR behaviors, simple extreme insensitivity, as, for example, seen in complete lack of responsiveness to the infant, could be frightening as well. In addition, they suggested that when extreme insensitivity is not followed by “repair,” the insecure infant will be left without access to an “organized” strategy for influencing the attachment figure in times of stress or alarm. Further, Grienberger, Kelley, and Slade (2005) describe AMBIANCE as stressing the “competing strategies,” which may arise when caregivers feel simultaneously compelled to heighten, and reject, infant attachment. Notably, however, competing strategies have been observed as well in case studies of CC parents and their infants in the home, and the infants of CC parents are to date being found either D or CC within the Strange Situation, see below).

Goldberg et al. (2003) offer a particularly useful comparison of the FR and AMBIANCE/FR+ systems, comparing Lyons-Ruth's thinking to Solomon and George's (1999) suggestion that the “key element in D attachment is the caregiver's failure to terminate attachment needs when they have been aroused.” Thus, they conclude that an important theme in both Lyons-Ruth's and Solomon and George's model is that “lack of response (withdrawal) can be as fear-provoking for the child as behavior that is frightening in and of itself” (pp. 242–243). Except for cases in which extreme “lack of response” actually stems from a parental dissociative state, this conceptualization is very different from the bouts of FR behavior and the resultant paradox with which the FR system is concerned. However, when a parent is extremely unresponsive without accompanying indices of a dissociative state, the authors would more likely expect unorganized than disorganized infant Strange Situation behavior. Exploring this hypothesis may necessitate use of a fifth, CC category of infant Strange Situation behavior presently being developed in our laboratory.

In their first test of AMBIANCE, Lyons-Ruth, Bronfman, and Parsons (1999) observed 65 high-risk infant–mother dyads in the Strange Situation, finding that AMBIANCE as a whole predicted D/insecure but not D/secure Strange Situation behavior, with the subscales for affective communication errors, fearful/D behavior, and withdrawal behavior serving as the strongest predictors. Results for two substudies of the AMBIANCE items including (a) only those separate from FR, and (b) those marked as exclusively FR, were essentially the same, with both identifying D/insecure, but not D/secure infants. However, in a subject-level analysis, the mothers of D/secure infants were found to be less hostile, negative, or intrusive than the mothers of D/insecure infants, and more likely to exhibit withdrawal.

Goldberg et al. (2003) attempted to replicate Lyons-Ruth's “high-risk” Strange Situation findings by analyzing 197 Strange Situation procedures within two low-risk samples. Strange Situations were conducted at 12 months, and AAIs had been administered prenatally to all subjects. Prenatal AAIs predicted AMBIANCE in the Strange Situation, and U/d mothers were higher on AMBIANCE (p < .05). In contrast to Schuengel, van IJzendoorn, and Bakermans-Kranenburg's (1999) findings and Jacobvitz et al.'s (2006) partial replication, no significant differences were found between the U/autonomous and the U/nonautonomous groups in relation to AMBIANCE/FR+. Within the Strange Situation itself, the scale constructed for overall AMBIANCE/FR+ was strongly related to infant disorganization (p < .0001); however, although the authors did not consider their interjudge agreement strong enough for reliable assessment of subcategories, two scales closely related to FR (fearful/disoriented/disorganized/dissociated; p < .01, and withdrawal; p < .02) made the strongest showing in discriminating not-U from U mothers.

However, in a later analysis of the same sample, Benoit, Madigan, and Goldberg (2003) reported that AMBIANCE/FR+ in fact had not discriminated D from C resistant/ambivalent infants. They concluded that AMBIANCE/FR+ might be “more sensitive than specific,” pointing to insecure rather than U infants.

Grienberger et al. (2005) examined relations between AMBIANCE/FR+ and infant Strange Situation classifications for 45 infant–mother dyads. Perhaps following upon Benoit et al.'s (2003) reanalysis of the Goldberg et al. (2003) report, Grienberger and colleagues predicted that insecure infants generally, rather than D infants specifically, would have higher scores. Interrater reliability was calculated only for the scale as a whole, and reliabilities for subscales were not tested. The intraclass correlation was .85 for the AMBIANCE total score. Here, as in the Benoit reanalysis, D infants could not be discriminated from resistant (C) infants, and had identical mean scores on AMBIANCE/FR+. However, avoidant infants had scores virtually identical to secure infants.

The latest study of AMBIANCE/FR+ (82 adolescent mothers and their infants) was conducted by Madigan, Moran, and Pederson (2006). Many of the mothers had been sexually abused, and 59% of the sample was D, 9% avoidant (A), and 33% secure (B). As in Lyons-Ruth, Bronfman, and Parsons's (1999) study, all infants classified as C were also classified as disorganized (D/C), composing 31% of the group of D infants.

Although AMBIANCE/FR+ checklists the occurrence of each set of behaviors, Madigan et al. (2006) reworked the five checklists, emerging with five 7-point Likert scales. Interrater reliabilities were established across 18 cases and were high both for the “general” scales for AMBIANCE/FR+ and for all five subscales. In contrast to previous studies, this AMBIANCE/FR+ assessment was conducted in a 6-min play session designed by Pederson, which followed upon, rather than occurring within, the Strange Situation procedure. In the first 3 min, mothers were asked to play with their infants in the absence of all toys: in the second 3 min, toys were brought in. In the subsession of the play procedure without toys, which can readily be regarded as the more stressful, all findings were stronger, and form the basis for the findings reported here. The overall AMBIANCE/FR+ scale was related to U maternal AAI status (p < .05), while the subscales were not. AMBIANCE/FR+ was strongly related overall to infant D attachment status (p < .0001). This may have been accounted for in part by the third subcategory of AMBIANCE/FR+, fearful/D/dissociative, which was found very strongly related to infant D attachment (p < .0001 in the session without toys) with intrusive negativity appearing as the second strongest contributor (p < .001). Other relations to subscales were modest, and withdrawal was not significant.

Because two previous studies found that AMBIANCE/FR+ predicted resistant (C) equally as well as D infant attachment status, (underlying) resistant attachment was tested against the overall scale (“disrupted communications”), and was again found significantly related to AMBIANCE/FR+ scale. The mean level of disrupted communications was, in fact, significantly higher for underlying or alternate C infants than for non-C infants (p < .001). Importantly, planned t tests also revealed significant differences between C and non-C infants specifically on affective communication errors (p < .01), the only subscale not inclusive of FR items, but the other AMBIANCE/FR+ subscales failed to point to group C infants. Fearful/D/dissociated behavior, intrusive negativity, and role-boundary confusion were, in contrast, unique to identifying D attachment.

Other sources of D behavior

From the above review, it is apparent that our hypothesis regarding the ability of repeated experiences of FR parental behavior to specifically yield D infant Strange Situation behavior (Main & Hesse, 1990) has obtained considerable support, both from studies using the FR system alone and from studies utilizing FR+. At this point, however, it is important to note that some D behaviors occur in response to neurological difficulties (see Barnett et al., 1999; Main & Solomon, 1990), although others can originate in pharmacological intervention (see Hesse, 1999, for an overview) and thus may not involve concommitant experiences of fear.

In addition, if sufficiently frightened by the situation in which they are placed, infants who are normally “organized” in their attachment to the primary caregiver(s) can become disorganized. For example, when Ainsworth (attempting an early stability study) subjected infants who were insecure–avoidant in the Strange Situation at 50 weeks to a repeat of the procedure at 52 weeks, almost all these formerly avoident infants were considered “unclassifiable” by her student coders (the “D” classification, of course, did not yet exist). At the outset of the second Strange Situation fright was highly apparent in these now overly stressed infants. This was no doubt due to the very short interval between administrations of the proceedure. Some of the infants began crying simply on seeing the door to the Strange Situation room while still in the hallway. This distress led to distorted proximity-seeking making the best-fitting “forced” classification “secure” in all seven cases (Ainsworth et al., 1978).

As we have already pointed out (Main & Solomon, 1990), it seems highly unlikely that the mothers of these avoidant Baltimore infants, already observed to be consistently aversive to tactual contact with their infants in the home (Main & Stadtman, 1981), had suddenly become frightening over the 2-week period. Rather, the “disorganization” seen in the infant's approach was probably more the result of meeting with a “problem without solution” (True, 2003): how to approach a mother who is not frightening but has stably shown a dislike of physical contact, when you are frightened. This is quite a different “tangle” than the experience of fright without solution. Similarly, True et al. (2001) found FR behavior predicted infant D in the Strange Situation among the Dogon ethnic group of Mali. Dogon infants, however, are continually carried by their mothers, and the Strange Situation conducted in a semiwesternized building was no doubt especially frightening. True later discovered that scores for mother's observed aversion to physical contact in the home substantially increased her power to predict D behavior within the Strange Situation (True, 2003).

The same increased likelihood of the appearance of D behavior holds, of course, for infants observed in stressful long-term separation settings (as see case histories presented by Heinicke & Westheimer, 1966; note, however, that although disorganization appeared in the separation setting, organized avoidance of the parent appeared upon reunion). However, disorganization in Strange Situation behavior itself does appear following repeated or long separations (Main & Solomon, 1990; see also Solomon & George, 1999). In addition, the majority of toddlers raised in orphanages absent the assignment of special caregivers are coded as disorganized within the Strange Situation (Vorra et al., 2003). Thus, despite the appearance of D behavior in the Strange Situation, none of these conditions, that is, excessively long separations from the mother, placement in a context sufficiently frightening to necessitate approaching a usually unapproachable mother, or orphanage rearing, involve fear of the parent, or are self-perpetuating “feedforward” experiences.

In that the above findings underscore a large set of conditions associated with D infant behaviors beyond the repeated experiences of “fright without solution” associated with FR parenting emphasized in this presentation, they exemplify the principle of “equifinality” (as see Cicchetti & Rogosch, 1996). At the same time, and at our present state of knowledge, only parental FR behavior appears to involve the infant in a looping of attention. If so, as opposed to other situations associated with disorganized Strange Situation behavior, repeated interactions involving FR parenting may be more likely to result in increased dissociative propensities. Thus, we propose that persons who had repeated FR interactions with the primary caregiver during infancy may be more vulnerable than others to dissociation. This is suggested by the longitudinal associations between infant D with mother and U/CC adult attachment status in the same individual found in two independent samples (Main et al., 2005; Sroufe et al., 2005).⁵

Dissociative aspects of U/d attachment during the AAI

As noted earlier, self-reported tendencies to enter states of absorption (Tellegen, 1992; Tellegen & Atkinson, 1974), a normative index of dissociation, have been linked to U status on the AAI in three independent, low-risk samples (Gribneau, 2005, unpublished data; Hesse & van IJzendoorn, 1999), while scores on Carlson and Putnam's (1993) Dissociative Experiences Scale scale have functioned similarly in a low-risk sample (Schuengel, van IJzendoorn, & Bakermans-Kranenburg, 1999). This suggests a direct link between dissociative or dissociative-like experiences and lapses in the monitoring of reasoning or discourse during the discussion of potentially traumatic events, which are the two chief indices used to identify U/d (U) status in the AAI.⁷

Lapses in the monitoring of reasoning or discourse during discussions of potentially traumatic experiences: Indications of lapses into mildly to more fully dissociative states

In our view, all indices identifying AAI transcripts as U/d fit to a dissociative model, as indicated directly below. However, there is much variability in the intensity and extremity of these lapses, and some approach the normative. This point is critical to our aim of creating a more incisive means of identifying early vulnerabilities to dissociation.

1. Dissociation is associated with traumatic/frightening experiences. Similarly, the anomalous patterns of speaking and reasoning discovered in interviews conducted with the parents of D infants in low-risk samples normally appear specifically during their discussions of potentially traumatic/frightening events.

2. Dissociation is identified through “alterations” in consciousness and behavior: correspondingly, our directions for identifying U/d responses to trauma refer to “alterations”/lapses in the monitoring of reasoning or discourse (Main et al., 2002).

3. Our directions for identifying “lapses in monitoring of reasoning or discourse” appear compatible with those which might also be used to identify (a) efforts to dissociate memories from awareness, as when as speaker remarks that memories of a dead person have been put away in a special place in their mind, or that when confronted with traumatic experience, their mind has just “gone off somewhere”; (b) current interference from partially dissociated memories; and (c) evidence of coexisting but incompatible and dissociated “memories” for past events (Main et al., 2002). We consider each of these in turn.

As is clear from the above, not all U/d lapses are indicative of extreme dissociation. For example, the use of funereal speech, or unusual attention to detail, merely suggest elevated levels of the most “normative” component of dissociation (Spiegel, 1990), absorption. In contrast, most of those cited under section (c), above, suggest the presence of real dissociative phenomena such as “segregated systems,” although in most cases we assume these are unlikely to involve multiple executors capable of guiding action. Future investigations should attempt to discriminate whether individuals exhibiting more extreme dissociative components in their discourse lapses are more frightening during interactions with their offspring, and have offspring who are more prone to dissociation.

FR behavior: Will differences in its expression be found to be associated with differences in vulnerabilities to dissociation?

The theme surrounding the primary subcategories of parental “FR” behavior is that they are directly frightening to the infant, thereby invoking the approach–flight paradox which, if experienced frequently enough at sufficient intensities, is expected not only to produce infant disorganization, but also to leave the infant vulnerable to developing dissociative tendencies if faced with intervening trauma at later ages (Liotti, 1992; Ogawa, Sroufe, Weinfield, Carlson, & Egeland, 1997). Above we reported that the dissociative subcategory of FR parental behavior was the strongest predictor of infant D for both Schuengel (1997) and Abrams et al. (2006), whereas for AMBIANCE/FR+, the “D” subcategory (fearful/disoriented/D/dissociative) was also the strongest predictor of infant D (Madigan et al., 2006).

However, aspects of the two other primary subcategories of FR also may fit to a dissociative model, and should not be overlooked in this regard. For example, a dissociative element may be present when the parent treats the infant as prey (seen in some “threatening” behaviors) or as a powerful and potentially harmful stimulus (some “frightened” behaviors). Because dissociative FR behavior appears to be the strongest predictor of infant D, and infant D is presumed to result from a disorganizing fear of the parent, it may be that dissociative behavior is particularly frightening and constitutes a theme connecting the subcategories of FR. Because, however, fear may not be readily observed in videotapes of young infants, physiological indices (see Montoya, Campos, & Schandry, 2005) might be utilized. Thereafter, following the model provided by Macfie et al. (2001a, 2001b), short-term longitudinal studies may assist in determining whether the dissociative component of FR behavior is, in fact, the most frightening to infants, and if so, the most likely to lead to vulnerabilities to dissociative disorders. At present, however, we need to leave open the possibility that it is simply the frequency and intensity of varied types of FR behavior that lead to increased likelihoods of D attachment, and hence to dissociative tendencies.

D infant Strange Situation behaviors particularly fitting to a dissociative model

Although, as noted earlier, many D behaviors identified as disorganized are unlikely dissociative, as hiding under the chair at the entrance of a clearly frightening mother), some D behaviors (chiefly, trancelike behaviors and seemingly “dissociated” actions) do seem to fit to a dissociative model. Indeed, at the phenotypic/descriptive level, a review of a subset of Main and Solomon's (1990) directions for identifying D behaviors reveals some remarkable resemblances to behavior indicative of dissociative states:

Summary

In sum, our directions for identifying U adult attachment status are, insofar as we are aware, well fitting to a dissociative model, albeit at varying intensities. FR behavior shown by U parents (e.g., Jacobvitz et al., 2006) is probably frequently dissociative, but may effect the infant only sufficiently to produce conflict behavior, or, as in the cases just cited, become associated with trancelike states and/or dissociated actions. A worthwhile endeavor for developmental psychopathology will be to examine the exact form taken by U/d lapses shown in parents during the AAI and the intensity and type of FR behaviors displayed and to compare these to the forms of D behavior exhibited by their infants. If possible, this should be undertaken utilizing several home observations, and more than one Strange Situation. Such studies must be followed up by “outcome” data assessing childhood dissociation but, as recently demonstrated, this can be assessed as early as preschool (Macfie et al., 2001a, 2001b).