evaluating quality in mother-infant interaction: situational effects

14
NFANT BEHAVIOR AND DEVELOPMENT 12, 451-464 (1989) Evaluating Quality in Mother-Infant Interaction: Situational Effects MARION O’BRIEN University of Kansas JEANNE M. JOHNSON Washington State University DIANA ANDERSON-G• ETZ University of Kansas Although the quality of maternal interaction is widely assumed to influence children’s development, attempts to relate measures of interaction quality to later developmental status have yielded inconclusive results. In the studies reported here, situational factors that may confound the measurement of mather- infant interaction quality were examined systematically. Results indicated that situational factors significantly influenced the quality of interaction as assessed by a rating scale incorporating an evaluation of emotional involvement as well as the nature and amount of stimulation provided the infant. Despite observed differ- ences, however, some dimensions of interaction quality, particularly emotional involvement and verbal stimulation, were highly correlated across situations. mother-infant interaction environment infancy measurement The quality of mother-infant interaction is widely assumed to have a major in- fluence on child development. Sensitive, responsive, and stimulating maternal care is considered to be the primary component of an environment that is most This project was supported in part by the University of Kansas Biomedical Sciences Support grant #4543 awarded to the first author. The second author was supported by a postdoctoral train- eeship from the Department of Education (grant #GOO8300899) to the Child Language Program, University of Kansas, and the third author by a predoctoral tralneeship from NICHD (grant #HDO7173) to the Department of Human Development, University of Kansas. The authors thank Michelle Knoll, Carrie Winterowd, Nancy Glinski, and Frances Johnson for assistance with scor- ing videotapes. Portions of this article were presented at the International Conference on Infant Studies in Washington, DC, April, 1988, and at the annual convention of the American Speech- Language-Hearing Association, Boston, November, 1988. Correspondence and requests for reprints should be sent to Marion O’Brien, Human Develop- ment and Family Life, University of Kansas, Lawrence, KS 66045. 451

Upload: marion-obrien

Post on 28-Aug-2016

214 views

Category:

Documents


1 download

TRANSCRIPT

NFANT BEHAVIOR AND DEVELOPMENT 12, 451-464 (1989)

Evaluating Quality in Mother-Infant Interaction:

Situational Effects

MARION O’BRIEN University of Kansas

JEANNE M. JOHNSON Washington State University

DIANA ANDERSON-G• ETZ University of Kansas

Although the quality of maternal interaction is widely assumed to influence children’s development, attempts to relate measures of interaction quality to later developmental status have yielded inconclusive results. In the studies reported here, situational factors that may confound the measurement of mather- infant interaction quality were examined systematically. Results indicated that situational factors significantly influenced the quality of interaction as assessed by a rating scale incorporating an evaluation of emotional involvement as well as the nature and amount of stimulation provided the infant. Despite observed differ- ences, however, some dimensions of interaction quality, particularly emotional involvement and verbal stimulation, were highly correlated across situations.

mother-infant interaction environment

infancy measurement

The quality of mother-infant interaction is widely assumed to have a major in- fluence on child development. Sensitive, responsive, and stimulating maternal care is considered to be the primary component of an environment that is most

This project was supported in part by the University of Kansas Biomedical Sciences Support grant #4543 awarded to the first author. The second author was supported by a postdoctoral train- eeship from the Department of Education (grant #GOO8300899) to the Child Language Program, University of Kansas, and the third author by a predoctoral tralneeship from NICHD (grant #HDO7173) to the Department of Human Development, University of Kansas. The authors thank Michelle Knoll, Carrie Winterowd, Nancy Glinski, and Frances Johnson for assistance with scor- ing videotapes. Portions of this article were presented at the International Conference on Infant Studies in Washington, DC, April, 1988, and at the annual convention of the American Speech- Language-Hearing Association, Boston, November, 1988.

Correspondence and requests for reprints should be sent to Marion O’Brien, Human Develop- ment and Family Life, University of Kansas, Lawrence, KS 66045.

451

452 O’BRIEN, JOHNSON, AND ANDERSON-GOETZ

likely to promote positive social, emotional, and cognitive development. To date, attempts to measure quality in mother-infant interaction and relate these measurements to later developmental status have yielded mixed results. One source of difficulty in measuring interactions stems from the potential effects of such situational variables as familiarity of the location in which interaction is observed and the context of the interaction session as determined by the in- structions given by the experimenter and the presence or absence of toys.

Weisz (1978) contended that if the purpose of the research is to identify robust aspects of human development, then the focus should be to identify behaviors that are transcontextually valid. When researchers are seeking to ob- tain a picture of “typical” interactions during their observations, they gener- ally attempt to use measurement processes that are minimally intrusive. It can be argued, however, that any form of assessment is likely to influence the phe- nomenon being observed. It is therefore important both to determine which observation procedures produce the least distortion of typical interactions and the most consistency across situations (Weisz, 1978) and to look for ways in which behavior is altered depending on context alone (Bronfenbrenner, 1979).

Most studies of mother-infant dyads have been conducted in the participants’ homes or in a laboratory. Usually, no explicit rationale for one setting versus another is given. The implicit advantage of using the home setting is that it is more familiar; thus, the probability of witnessing a more natural interaction may be maximized in this condition. Bronfenbrenner (1977), among others, suggested that the home setting is likely to produce more ecologically valid samples of behavior. The laboratory, on the other hand, has the advantage of standardization across observations, elimination of distractions for the dyad, and convenience for the investigator. In a review of methodology in interac- tional studies, Lytton (1977) found that researchers chose the laboratory setting and task cognizant, in some manner, of the compromise between control over extraneous variables and typicality of the interaction. To date, few data are available to help an investigator decide which environment yields the most useful data for a particular purpose.

A few studies have addressed the issue of situational effects on mother and infant behavior. Brookhart and Hock (1976) found that the frequency of in- fant contact with the mother was higher in the lab and that infant proximity with a stranger was higher at home. These differential effects were attributed to location. Similarly, in a study of separation protest, Ross, Kagan, Zelazo and Kotelchuck (1975) found less crying at home than in the lab when infants were left alone. Klein and Durfee (1979) found significantly less proximity to mother, looking, and vocalizing in l-year-old infants tested in the laboratory versus those tested at home. When maternal behavior was introduced as a covariate, however, these effects were found to be associated entirely with variation in mothers across the two locations. Because different instructions were given to mothers in the two situations, the interpretation of the results re- mains unclear.

EVALUATION OF QUALITY 453

For those studies that found an effect related to location, several possible scenarios can be postulated. In the laboratory, all distractions are removed and mothers are typically asked to focus in some way on their infants. This may create a novel situation in that mothers rarely spend uninterrupted minutes with their infants. The amount of responsiveness and stimulation provided is accordingly far higher in the laboratory. Additionally, the laboratory may create tension for mothers who feel that they expected to get their infants to “perform.” Observations carried out at home do not avoid all these problems, however. The presence of an observer or of equipment such as a video camera may create an intrusion that may be equally as disruptive as in the laboratory. Similarity of behavior on the mother’s part in both locations would indicate that maternal behavior is controlled primarily by the presence of the infant.

In 1980, Belsky carried out a study to compare two interactional situations: the “free-play” task typically used in many laboratory studies and “normal” interactional patterns observed in the home. He found that the behavior of the infants did not change across the two situations; however, the mothers were more active and responsive in the free-play laboratory setting than at home. In- herent in the method was a difference in the instructions given in the laboratory versus at home. In the lab, Belsky asked the mothers to pretend they were at home with one half hour of free time to spend with their infants. In the home, mothers were asked to go about their daily routines as if the observer were not present. Thus, the lab task obligated the mother to play with her infant; the home task did not. Because the validity of the free-play type of task was the focus, any conclusions about situational effects must consider this inherent difference in task structure as well as the different locations of observation.

Lytton (1977) addressed the elusive nature of task structure and how it may predispose a parent to behave in a particular way. The amount of structure in- herent in seemingly open-ended instructions is revealed by predictable patterns of interaction. For instance, Field (1977) demonstrated that a small change in instructions to mothers led to quite different behavior in high-risk infants. After obtaining a baseline of the infant’s gaze maintenance and gaze aversion, she asked mothers to try to get their infants to talk. The frequency of gaze aversion rose in relation to the increased activity on the mother’s part. Field then asked mothers to imitate their infants’ sounds and movements. The fre- quency of gaze aversion dropped below baseline as did the mother’s level of activity.

Instead of controlling the mother’s behavior through instructions, other re- searchers have used standard sets of toys during observations, assuming that this lends a degree of control (e.g., Mahoney, Powell, & Finger, 1986; Russo & Owens, 1982). To date, systematic manipulation of situational variables such as toys has not received much attention. O’Brien and Nagle (1987) reported that parents’ speech to toddlers differed more as a function of the toy placed with than by gender of parent or child, suggesting that play context exerts a powerful influence on context of interaction.

4% O’BRIEN, JOHNSON, AND ANDERSON-GOETZ

The purpose of the two studies reported here was to evaluate systematically the effects of situational variables on ratings of quality of maternal interaction with their infants. At question was (a) the degree of consistency in the quality of maternal interaction across situations and (b) differences in maternal be- havior associated with particular interaction situations. In Study 1, both the location and instructions given mothers were manipulated to evaluate both familiarity of location and context-of face-to-face play and “normal routine” -within the familiar home setting. In Study 2, the location was varied but the context-toy play-remained constant.

DESIGN

Two different groups of 3- and 6-month-old infants and their mothers were observed. In Study 1, 31 mothers and infants were observed three times: in a face-to-face session in the laboratory with no toys available; in a face-to-face session at home; and at home with instructions to “do whatever you normally do this time of day.” In Study 2,22 mothers and infants were observed in two different locations, home and laboratory playroom, but the interaction con- text was held constant by providing the same toys in each location.

STUDY 1

Subjects Thirty-one mother-infant dyads (11 3-month-olds and 20 6-month-olds; 15 males, 16 females) were observed. All infants were full-term with no postnatal complications. The mean age of mothers was 28.4 years (range = 18-38 years), and the mean number of years of education for mothers was 15 with a range of 11 to 21 years.

Procedures Mothers and infants were randomly assigned to be videotaped first either at home or in the laboratory; the order of location was counterbalanced within each age group by sex of the infant. The face-to-face interaction in the labora- tory took place in a small playroom with a one-way mirror. The infant was placed in an infant seat on a low table and the mother was asked to sit in a small chair facing her infant. No toys were provided nor were any personal effects brought into the room. The mother was asked simply to “play with the baby for 10 minutes” and could hold the infant or interact with the infant in the seat.

The home interaction was videotaped under two conditions. During the first 10 min, the parent was asked to play with the infant without using toys. Mothers selected a place in the home that was comfortable for them and where the observer with the video camera could record the interaction; most mothers chose the living room couch or floor. The face-to-face interaction was always

EVALUATION OF QUALITY 455

TABLE 1 Dimensions of Moternal Interaction

Dimension

Average

Interobserver Agreement

1. Overall rating: the rater’s impression of the quality of interaction. 2. Interest or involvement: intensity of mother’s attention to the infant;

distractibility. 3. Amount and quality of verbal stimulation: amount of conversation,

variability of voice tone, oppropriateness to infant’s state. 4. Tone of voice: degree to which mother’s pleasure in the infant is made

obvious through her voice. 5. Content of verbalizations: variety of topics, use of games. 6. Quality of tactile stimulation: amount of touching, especially affectionate

touching. 7. Maintenance of joint attention: mother’s efforts to attract and hold in-

fant’s visual attention and willingness to follow infant’s focus of attention. 8. Responsiveness to infant vocalization: frequency of mother’s response to

noncrying vocalizations, variability and appropriateness of response. 9. Positive emotion: mother’s expressions of pleasure.

10. Negative emotion: mother’s expressions of displeasure, fatigue, irritation, or frustration.

.93

.76

.79

.90

.BO

.BO

.B5

.66

.B3

.91

presented first, to make it as similar as possible to the laboratory session. Following this, the mothers were asked to do whatever they normally did dur- ing that time of day, and the infant was videotaped for the next 30 min. The in- vestigator did not interact with either the mother or the infant during the videotaping. The lab and the home observations occurred within 1 week of each other.

Rating of Interaction Quality. Videotaped sessions were later scored using a lo-item mother-infant interaction rating scale. This 7-point scale evaluates the responsiveness of a mother to her infant and the quality of stimulation pro- vided by the mother during interaction. A brief description of the categories rated appears in Table 1. Similar types of scales have been used by others to capture aspects of interaction, both in home and laboratory contexts, that relate to later cognitive, linguistic, and social outcomes (Bakeman & Brown, 1980; Mahoney et al., 1986). According to Bakeman and Gottman (1986), ratings of specific behavior are more useful than frequency counts in capturing individual differences in behavioral style, which was the question of interest in this study.

Because of the subjectivity inherent in evaluating quality of interactions, each tape was scored by three observers who did not participate in the data col- lection. Raters worked independently and rated a S-min segment from each of the IO-min home and laboratory tapes (Minutes 2-6). The scores from the three raters were then averaged. (Mean percentage agreements across observers

456 O’BRIEN, JOHNSON, AND ANDERSON-GOETZ

in both studies on each item in the rating scale are presented in Table 1.) Tapes were scored in random order without subject identifiers to eliminate carry-over effects.

The 30-min “normal-routine” sessions were observed in random order at a separate time. Because the rating scale was intended to evaluate actual inter- action quality and not overall amount of interaction, only portions of the tape in which both mother and infant were visible were rated. Many mothers took a brief break from face-to-face interaction after the lo-min session; therefore, the first 5 min of the normal-routine sessions were omitted. Raters then scored the first 5 min of interaction. If the mother moved in and out of the infant’s range, episodes of interaction were accumulated and timed with a stopwatch until the rater had observed a total of 5 min of interaction. Raters then scored the interaction and also recorded both the time and the presence or absence of distractions or potentially interfering activities (siblings, TV, dog, etc.) in the immediate environment at any time during the interaction observation.

HOME Scale. The Home Observation for Measurement of the Environ- ment (HOME; Caldwell & Bradley, 1978) has been used widely as an evalua- tion of the quality of a child’s home environment and has been correlated to outcome variables at later ages in a child’s development. The measure uses both observation and parental interview to obtain information about the emo- tional and verbal responsiveness of the mother, the amount of restriction and punishment in the home, the organization of the home environment for the child, the kinds of play materials provided, the degree of maternal involve- ment with the child, and the opportunities for variety in daily stimulation. The infant version of the scale is intended for use in the homes of children from birth to age 3.

The investigator who visited the home to videotape completed a HOME scale during and after the visit. This provided a measure of external validity for the mother-infant rating scale and an independent measure of the child’s environment.

Results Initial analyses were run to evaluate possible effects of child age and gender on the ratings of maternal interaction quality. Separate Age (2) x Sex (2) MANOVAs were run for each of the three locations. Although there were no main effects or interactions involving child gender, significant main effects were found for age, with the mothers of 6-month-old infants receiving higher ratings in both face-to-face interaction situations: lab, F(1,20) = 12.60, p< .OOl; home, F(1,20) = 3.40, p< .05. Age was therefore used as a covariate in further analyses of situational differences.

The average ratings across the three situations are shown in Table 2. Situa- tional effects on the ratings of maternal interaction quality were evaluated by separate repeated-measures MANOVAs on each item with age as a covariate.

EVALUATION OF QUALITY 457

TABLE 2 Mean Maternal Interaction Ratings in Three Situations, F Values,

and Differences Between Means as Shown by Post-Hoc Tests

Lab Home

Face-to-Face Face-to-Face

Home

Normal Routine I=

Overall rating Interest/involvement

5.2’=b 5.3b

4.Bb B.Ol** 4.9b 19.55***

Verbal stimulation 5.B0 5.4O 4.5b 20.07’*’ Tone of voice 6.1’ 5.6b 5.6b 6.86” Content of verbalization 5.1 4.8 4.6 3.18*

Tactile stimulation 5.4 5.3 4.9 4.10’ Joint attention 5.9O 5.5b 4.9b 9.76***

Response to vocalization 5.0 4.7 4.1 4.26* Positive emotion 5.9P 5.8’ 5.3b 12.17*** Negative emotion 6.7 6.7 6.5 3.BB* Average rating 5.B’= 5.4ab 5.lb 17.82’**

Note, For each rating category, meons with dissimilar superscripts differ significantly, p<.Ol.

l p<.o5. ** p<.Ol. *** p<.ool.

Significant differences were found across settings for all components of the rating scale (Table 2). On all dimensions of quality, the lab interactions were rated highest and the normal routine interactions lowest. Post-hoc tests indi- cated situational differences (at the .Ol level) between lab and home face-to- face interactions in interest/involvement, tone of voice, and joint attention and between home face-to-face and home normal routine in verbal stimulation and positive emotion.

The consistency of mothers’ behavior across situations was evaluated by calculating Pearson product-moment correlation coefficients across the entire sample. The correlations are shown in Table 3. Six of the individual items on the rating scale, as well as the average rating, were significantly correlated (at the .Ol level) across all three situations. The lowest correlations were found for tactile stimulation and joint attention. To evaluate the effects of distractions in the normal-routine home situation compared with the face-to-face situation in the laboratory, partial correlations between lab and normal routine with the presence of distractions partialled out were also calculated. These correlations were essentially the same as the zero-order correlations, with the exception of the interest/involvement factor which was slightly higher (.46 compared to 40) when controlling for distractions.

Correlations between the maternal interaction rating scale items and the HOME scale scores were also carried out to determine whether the two scales were measuring independent or overlapping dimensions. In these middle-class households, some dimensions of the HOME scale showed little variability. Therefore, only the total HOME score, which averaged 39.9 across the 31 pm- ticipating families (range = 30.0-45.0), was used. The correlations between the

458 O’BRIEN, JOHNSON, AND ANDERSON-GOETZ

TABLE 3 Zero-Order Correlations Among Ratings Across Three Settings

Overall roting

Interest/involvement Verbal stimulation

Tone of voice Content of verbalization Tactile stimulation

Joint attention Response to vocalization Positive emotion

Negative emotion

Average roting

l p<.Ol.

Lob/Home

Face-to-Face

56’ .39

.62’

.57’

.43’

.32

.53*

.60*

.4B*

.M’

.52’

Lob/Home Normal Routine

.42*

.40

.29

.72*

.53’

-.04 .21 .49’ .44’

.59’

.5B’

Home

Face-to-Face/

Normol Routine

.42’

.33

.44’

53’ .57f

- .09

.02

.4B*

.53*

.43’ .52*

TABLE 4 Correlations Between HOME Scale Totol Score and Moternal Interaction Ratings

in Three Settings

Lab Home

Face-to-Face Face-to-Face

Home Normal

Routine

Overoll roting

Interest/involvement Verbal stimulotion

Tone of voice Content of verbalization

Tactile stimulation Joint ottention Response to vocolizotion

Positive emotion Negative emotion Average rating

.27

.25

.22

.lO

.30

.lB

.31* -36

.21 .47** .36*

.oo - .oa

.I5

- .02 .14 .20

.02 -.12

.31 l

.lB .lO

.46**

.49**

.45**

.45”

.41*

.03 .28 .29

.47”

.54”

.45**

l pc.05. l * pc.01.

HOME scores and the rating scale items in each situation are shown in Table 4. The HOME scores were highly related to the maternal interaction ratings only during the normal routine segment of interaction.

Discussion The results from Study 1 indicated that although maternal interaction quality differs across situations, some dimensions of quality remain consistent. Ratings were consistently highest in the laboratory, where distractions were eliminated and mothers may have felt the most obligated to “put on a show.”

EVALUATION OF QUALITY 459

Given the same instructions for face-to-face interaction at home as in the lab, mothers’ levels of involvement with their infants were lower. Some mothers brought out toys to entertain the child rather than interact face-to-face, as instructed.

Maternal interaction quality was rated lowest when mothers were asked simply to “do what you always do at this time of day.” Some of the mothers left their infants alone for considerable periods or remained nearby but inter- acted only as the child became fussy or needed care. One mother did not inter- act with her infant at all during the 30 min of videotaping and so could not be scored. This finding may be related to the fact that the normal-routine segment was always taped last to maximize the similarity between the lab and home face-to-face sessions. Because mothers had just spent 10 min interacting, they may have felt free to attend to other household chores. To allow mothers a break, we did not begin coding until 5 min after the normal-routine session was begun and then coded the first 5 min of interaction occurring anytime dur- ing the rest of the 2%min session. The average time required to obtain 5 min of interaction was 11 min (range = 5-24).

The implicit assumption behind many observations of parent-infant inter- action in the laboratory is that the more intense lab situation simply “com- presses” normal interaction bouts into a single session. The procedures followed in this study essentially tested that assumption; periods of actual interaction were electronically compressed to produce a 5-min period of interaction. By contrast with typical face-to-face interaction, however, the normal-routine in- teraction sessions taped in this study were of significantly lower quality. Fre- quently, the periods of interaction were very brief and functional, involving repositioning the infant, directing attention to a different toy, wiping the in- fant’s nose or mouth. The rating scale was devised to score face-to-face inter- action and so may not have been entirely suited to such short bouts of contact and caregiving. The general dimensions scored by the scale are widely consid- ered to be important components of quality in maternal caregiving, however, and presumably should be observable during daily activities. Further and more extended comparisons of daily routine situations with face-to-face interactions should be carried out to determine the frequency with which face-to-face types of interactive behavior occur in normally occurring parent-infant interchange.

Although the ratings varied according to the situation, there was also con- sistency across most dimensions of the scale. The components reflecting emo- tional involvement of the mother with the infant (interest/involvement, tone of voice, positive and negative emotion) were particularly highly correlated across the three situations. In fact, when three of these four items (omitting negative emotion, which showed little variability in ratings) were combined into an emotionality score, the correlations across settings were quite high: lab with home face-to-face, 58; lab with home normal routine, .6I; home face-to-face with home normal routine, .55. Similarly, a verbal stimulation score, derived by combining verbal stimulation, vocal responsiveness, and content of verbahza-

460 O’BRIEN, JOHNSON, AND ANDERSON-GOETZ

tions, was consistently highly correlated across locations: lab with home face- to-face, .63; lab with home normal routine, 53; home face-to-face with home normal routine, .63. By contrast, the tactile stimulation and joint attention dimensions showed little similarity across situations, particularly when com- paring face-to-face interactions with the normal-routine segment. Most mothers did not hold their infants during this portion of the home observation but set them on the floor or in a walker. As the mother came and went, neither touching nor joint attention occurred very often.

Interaction quality did not appear to be highly influenced by distractions at home, such as older siblings watching TV in the same room as mother and in- fant. Mothers’ relative rankings in the lab situation, with no distractions, and the normal-routine at home situation were not affected by whether there were distractions in the environment.

The HOME scale, intended to evaluate the overall quality of a child’s home environment, was most closely related to the ratings of interaction quality dur- ing the normal-routine segment. The observer who videotaped the home inter- actions also completed the HOME scale; thus, these evaluations were based in part on what occurred during the face-to-face interaction at home and the half-hour of normal-routine interaction. The HOME scale scores were corre- lated only with the normal-routine segment of interaction, suggesting that the face-to-face interaction context may tap some dimensions of the child’s social environment that are different from those evaluated by the more global ratings of the HOME scale. A combination of the two measures, maternal interaction ratings during a face-to-face session and a HOME scale rating, may give a more complete and balanced picture of the quality of the child’s environment than either measure alone.

STUDY 2

Subjects Twenty-two mother-infant dyads (10 3-month-olds and 12 6-month-olds; 11 males, 11 females) participated in Study 2. All were healthy full-term infants. Maternal age averaged 30.5 years (range = 19-42), and years of education averaged 15 (range = 12-20).

Procedures Mothers and infants were videotaped for 10 min of play with a standard set of toys, both at home and in the same laboratory playroom used in Study 1. The order of location was again counterbalanced. Toys included several rattles, a roly-poly clown with a weighted base, a squeeze toy, and several small plastic age-appropriate toys. Parents with older children who had been or were to be involved in the home videotaping session were asked to bring them to the session in the laboratory as well. This was done in an effort to make the two

EVALUATION OF QUALITY 461

TABLE 5

Mean Ratings for Toy Ploy Tosk in Two Settings ond Correlations Across Settings

Lob Home r

Over011 roting 5.4 5.4

Interest/involvement 5.1 5.1

Verbal stimulation 5.3 5.3 Tone of voice 5.8 5.5

Content of verbolizotion 5.3 5.1 Tactile stimulation 4.9 4.9 Joint attention 5.9 5.7

Response to vocalization 4.5 4.2 Positive emotion 5.5 5.3

Negotive emotion 6.0 6.7 Averoae rotinq 5.5 5.4

.M’

.73’

.71*

58’ .71 l

.31

.29

.B2* .74’

.Bl l

.66*

* p<.ool.

settings as similar as possible. The mothers were instructed to play with their infant using the toys provided.

Rating of Interaction Quality. The home and laboratory tapes from Study 2 were observed in random order, with identifying information concealed, by one of the observers in Study 1 and two new observers. The same rating scale used in Study 1 was used in this study. Data represent averages across the three raters.

Results A Sex (2) x Age (2) x Situation (2) multivariate analysis of variance was run to evaluate differences in ratings related to age and sex of the child. Significant main effects for both Sex and Age were found for the ratings made at home: Sex, F(1,15)=12.7O,p<.Ol; Age, F(1,15)=10.9O,p<.Ol. UnivariateFtests showed no single rating scale item to be significantly different across groups, however.

The overall ratings across the two situations were highly similar, as shown in Table 5. The zero-order correlations across the two settings (shown in Table 5) were also higher than those found during Study 1 when no toys were pro- vided for play. The two composite dimensions found to be stable in Study 1, emotionality and verbal stimulation, were also highly consistent across these lab and home situations: emotionality, r= .73; verbal stimulation, r= .90.

Discussion The ratings of maternal interaction quality were highly similar in the lab and at home when the same set of age-appropriate toys was provided in both settings. The context of toy play served to reduce the variability associated with dif- ferences in the location of observation. In addition, mothers seemed more

462 O’BRIEN, JOHNSON, AND ANDERSON-GOETZ

comfortable with the toy play situation than when asked to interact face-to- face. Middle-class mothers commonly use toys as mediators of interaction, even with very young infants. Thus, the toy play interactions may represent more “natural” interaction situations for parents of 3- and 6-month-old in- fants than the face-to-face sessions commonly used in research. The results of the present study strongly suggest that such toy play interactions conducted in the more controlled and efficient laboratory environment are as useful in eval- uating the quality of the mother’s interaction with her infant as toy play in- teractions carried out at home.

GENERAL DISCUSSION

In many studies, parent-child interaction observations are considered to be representative of the “typical” behavior of the dyad. Clearly, however, behav- ior differs across situations. The researcher is therefore confronted with the problem of defining what “representative” means and how to conduct obser- vations that capture meaningful aspects of the parent-child relationship and yet are reasonably efficient in terms of time and money.

A further question concerns the extent to which situations place behavioral demands on individuals. Research to describe the factors within home or laboratory environments that serve to elicit different patterns of interaction would be useful in isolating characteristics of the dyad from characteristics of the situation.

From the results reported here, it is evident that observations made at home, when mothers are asked to do what they normally do, produce a differ- ent picture of mother-infant interaction quality than do either face-to-face or toy play interaction observations carried out either in the laboratory or at home. Much of the interaction between mothers and infants during the normal- routine home situation in the present study consisted of brief and functional interchanges, caregiving episodes, or verbal interaction from a distance. Dur- ing face-to-face and toy play observations, the middle-class mothers in these studies engaged their infants in longer bouts of interaction, focusing their at- tention on their infants even when there were distractions.

Researchers interested in carrying out detailed analyses of parent-child in- teractive behavior should consider the context in which they are observing in- teractions. If joint attention between parent and infant is the focus of the study, for example, an instruction to “do what you normally do” is not likely to elicit the kind of interaction in which joint focus on objects occurs. A toy play situation, on the other hand, is likely to be a more useful context in which to elicit joint attention.

Despite the observed differences across situations, however, some aspects of interaction quality appeared to be stable, most notably the emotional tone of the mother and the amount and nature of verbal stimulation provided. Other factors that are commonly included in measures of interaction quality, such as

EVALUATION OF QUALITY 463

tactile stimulation, were quite unstable across situations. If an investigator wishes to capture aspects of quality occurring across all the daily activities of an infant without devoting many hours to observation, then those dimensions that remain stable should be the focus.

It would also be helpful for researchers studying parent-infant interaction to define general conditions under which parents’ behavior actually influences children’s development and design their observations to capture such critical aspects of parent-child interactions. Tomasello and Farrar (1986), for example, found that episodes of joint attentional focus, although brief, could be used effectively in language teaching. It seems probable that a mother’s tendency to use such episodes as teaching opportunities can be observed more efficiently within a structured observation setting rather than across a full day of “natural” home observations. The definition of other such critical aspects of interactions would allow investigators to test assumptions about the typicality of their laboratory sessions without embarking on extensive observation and analysis projects.

For overall evaluation of the quality of an infant’s environment, including both social interaction and physical environmental variables, a combination of measures is probably needed. For example, combining a structured interaction observation, such as that provided by the kind of standard toy play task used in Study 2, and a home observation and interview, such as the HOME scale, would be both efficient and thorough. The play task allows direct observa- tion of maternal and child behavior in a typical context, and the HOME scale provides a broader evaluation of the infant’s environment. In combination, the two kinds of observations provide a considerably more comprehensive pit- ture of the child’s environment than either measure alone.

REFERENCES

Bakeman, R., & Brown, J.V. (1980). Early interaction: Consequences for social and mental devel- opment at three years. Child Development, 51, 437-441.

Bakeman, R., & Gottman, J. (1986). Observing interaction: An introduction to sequential analy- sis. Cambridge, MA: Cambridge University Press.

Belsky, J. (1980). Mother-infant interaction at home and in the laboratory: A comparative study. Journal of Genetic Psychology, 137, 37-41.

Bronfenbrenner, U. (1977). Toward an experimental ecology of human development. American Psychologist, 32, 5 13-53 1.

Bronfenbrenner, U. (1979). The ecology of human development. Cambridge, MA: Harvard Uni- versity Press.

Brookhart, J., & Hock, E. (1976). The effects of experimental context and experiential background on infants’ behavior toward mothers and a stranger. Child Development, 47, 333-340.

Caldwell, B.M., & Bradley, R.H. (1978). Manual for the Home Observation for Mea.wrement of the Environment. Little Rock: University of Arkansas.

Field, T. (1977). The effects of early separation, interactive deficits, and experimental manipma- tions on infant-mother face-to-face interactions. Child Development, 48, 763-771.

Klein, R., & Durfee, J. (1979). Comparison of attachment behaviors in home and laboratory. Psychological Reports, 44, 1059-1064.

464 O’BRIEN, JOHNSON, AND ANDERSON-GOETZ

Lytton, H. (1977). Observation studies of parent-child interaction: A methodological review. Child Development, 42, 651-684.

Mahoney, Ci., Powell, A., & Finger, I. (1986). The Maternal Behavior Rating Scale. Topics in Early Childhood Special Education, 6, 44-56.

O’Brien, M., & Nagle, K. (1987). Parents’ speech to toddlers: The effect of play context. Journal of Child Language, 14, 269-279.

Ross, G., Kagan, J., Zelazo, P., & Kotelchuck, M. (1975). Separation protest in infants in home and laboratory. Developmental Psychology, II, 256-257.

Russo, J., & Owens, R. (1982). The development of an objective observation tool for parent- child interaction. Journal of Speech and Hearing Disorders, 47, 165-173.

Tomasello, M., & Farrar, M. (1986). Joint attention and early language. Child Development, 57, 1454-1463.

Weisz, J. (1978). Transcontextual validity in developmental research. Child Development, 49, 1-12.

22 November 1988; Revised 9 June 1989 W