Beruflich Dokumente
Kultur Dokumente
0272-4944/86/030205 + 27 $03.00/0
206
G.T. Moore
2O7
question is, how does the spatial character and configuration of activity areas or
behavior settings influence these activities, if at all?
Only a few studies have looked at activity settings in child care environments.
Rosenthal (1974) studied the behaviors of a heterogeneous population of preschoolers balanced in terms of gender, race, and age during 37 child care sessions.
Settings differed significantly in their attractive power as measured by the percentage
of children involved in them, as well as their holding power as measured by the
length of involvement. Settings for art, block play and novel ventures were the most
attractive settings, while role playing settings had the greatest holding power.
Somewhat similar findings were reported by Shure (1963), who found that the most
popular areas were block play and art, with the block play area having the greatest
holding power, although there were significant gender differences. No theoretical
discussion was offered in either study to account for the differences in attractiveness
or holding power of different settings. We don't know, for instance, whether the
behaviors have anything to do with the characteristics of the physical environment,
the materials provided, staff characteristics, or some combination.
On the other hand, it is known that children playing outdoors tend to congregate
in groups of less than five children with a mean of just under two children in a
setting (Aiello et al., 1974). Similarly, experts recommend that the best size for an
indoor preschool play group is two to four children (Millar, 1968). A large sample
national study in the U.S.A. indicated that the quality of child care programs as
measured by the Preschool Inventory and the Peabody Picture Vocabulary Test is
related to the size of the group, with small groups working best (Abt Associates,
1979).
At a more specific level, the now widely cited work of Smith and Connolly (e.g.
Smith, 1978; Smith and Connolly, 1980) on the ecology of preschool environments
in England has begun to shed considerable empirical light on child-environment
linkages in early childhood indoor environments. Using experimental designs and
naturalistic observation techniques, they have found, among other things, that more
fantasy play is observed in smaller play groups though aggressive behavior does not
vary with group size. Running, chasing and vigorous activities occur more frequently
in larger spaces (those, for instance, around 6-75 m 2, or 75 ft 2 per child) while
more physical contacts happen between children in smaller preschool spaces (2.2 m 2
or about 25 ft 2 per child), though again aggressive behavior was not found to vary
with absolute space size. When more rather than less play equipment is available,
children tend to play in smaller subgroups including playing alone some of the time;
also with more equipment available, there is less sharing of equipment and less
aggressive behavior and overall a less stressful situation for the children. Other
findings are reported in Smith and Connolly's book, comparing structured activities
to free-play conditions, and variations in the ratio of children to staff members, but
for the most part these latter studies were conducted without explicit attention to
the possible role of physical environmental variables.
208
G . T . Moore
behavior settings (i.e. those with clear boundaries from circulation space and from
other behavior settings, and with at least partial acoustic and visual separation)
should decrease classroom interruptions and contribute to longer attention spans
and greater involvement with developmentally appropriate activities (Moore et al.,
1979, Pattern 908).
Well-defined behavior settings are areas limited to one activity, but not completely
cordoned off from other activities. They are sized to accommodate two to five
children plus one teacher (typically 5 to 10 m 2 or 50 to 100 ft2), and typically
include storage, surface area, equipment plug-ins and display for the activity. In
the better child care centers, one behavior setting is provided for each major
developmental activity.
The notion of spatially well defined settings implies that the setting has a high
degree of spatial differentiation from other settings and therefore from other activities. Thus, for example, spatially well-defined settings may be characterized by any
or all of the following:
(1) partially surrounding walls or partitions;
(2) bookcases, storage cabinets, or shelves used as partial dividers and able to be
moved and changed as staff and children wish in order to accommodate varied
group sizes;
(3) changes in levels, either the floor or the ceiling;
(4) changes in floor coverings or textures;
(5) hangings and placement of overhead lighting to define spaces; and
(6) implied boundaries suggested by the placement of columns, posts, or other
strong visual elements, or the visual completion of space by the implied visual
connection between wall stub ends [i.e. the Gestalt principle of pragnanz or, more
specifically, the law of perceived closure of almost closed figures (Koffka, 1935)].
Spatially well-defined settings are contrasted with poorly defined behavior settings, i.e. areas where the actual or implied spatial definition is low, where the area
is too large or too small for the group size, and/or where the resources and work
surfaces are not readily available for the particular activity. Diagrams showing the
evolution of spatially well-defined behavior settings are shown in Figs 1-4.
The purpose of this study was to investigate the effects of the degree of spatial
definition of behavior settings on a number of cognitive and social behaviors in
child care centers.
Hypotheses: the two sets of hypotheses behind this study were as follows:
H 1 There will be more indicators of cognitive and social development (e.g.
engagement, child-imitation of behavior, exploration, social interaction and cooperation) in child care centers with architecturally well defined behavior settings
than in those with less well-defined behavior settings.
H2 As neither the social nor the physical environment acts independently of the
other, there will be complex group-place interactions, specifically interactions
between the backgrounds of the children, the teaching philosophy and style of the
teachers, and the character of the physical environment will affect various cognitive
and social behaviors.
209
(b)
I
~"~t
~:::~
..
I n , definition
~ . ~ = ~ / "
t "~.~[~,., '. ~
without full
room walls
~ Partitions
Storage
Enclosed spaces
Hangings
Ceiling height
Changes
Circulation
FIGURE I. Diagrams showing the evolution of spatially well-defined behavior settings for child care
centers: (a) The U.S. National Day Care Study finding that small groups work best, translated into (b)
the beginning definition of activity pockets or behavior settings inside the total space available. (Illustrations in Figs I-4 by Tim McGinty from Recommendations for Child Care Centers by G. T. Moore, C. G.
Lane, A. H. Hill, U. Cohen and T. McGinty, 1979, Milwaukee: University of Wisconsin-Milwaukee,
Center for Architecture and Urban Planning Research. Illustrations copyright 1979 by Tim McGinty,
reprinted by permission.)
(b
"~3-4 small
activity
spaces
Storage of
manipulative
toys
Soft
places
m a t e r i ~
Cross visibility
Easilly
/~/"NJ
visibile . ~ E ~ ,.1
by an
~'~.1_a
Recorders
i.J
etc.
adult
X-Sense of
enclosure
2-5
children
-J
I ~ , , ~t
"Y'~
Relation to
other
activities,
views, etc.
FIGURE 2. Qualities of well-defined behavior settings. (a) The provision of a sense of enclosure and of
resources in the behavior setting, and (b) the clustering of three or four behavior settings in proximity to
each other.
Method
210
G. T. Moore
(a)
(b)
-)
*Light
:
,~
* Lofts and
level changes
{
Level
changes
(c)
(d)
* Defined but
non-interfering
circulation
Defined
space
View
....
*Implied boundries
made by changes in
materials
Low
movable
ca binets
5'or 4'
FIGURE 3. Ways to spatially define behavior settings (law of pragnanz applied to building design),
including level changes, skylights and defining areas by pools of light, lofts and child-only crawl space
underneath, separation of circulation from activity settings by storage units or other vertical elements,
and implied boundaries suggested by changes in materials or textures.
cultural aspects of environments. While much of the work in the environmentbehavior field has been exploratory~lescriptive in nature or has been survey research
with little intervention and little control (see, for example, the studies reviewed in
Zeisel, 1981), only a few studies have been able to develop experimental procedures
while maintaining the relative integrity of settings and events (e.g. Carp, 1966; Trites
et al., 1970; Smith and Connolly, 1980). Though the field should move toward
multilateral and multivariate causal explanations, adoption of a strict causal model
does not seem appropriate at this time. An ecological model where the total ensemble
of social and physical variables are related holistically seems more appropriate-careful observation and systematic description must preceed explanation (Proshansky, 1972; Wapner et al., 1973; Bronfenbrenner, 1977; Wohlwill, 1980).
The problem of making causal inferences in field settings has been dealt with at
length in a series of handbook chapters and books by Campbell and his colleagues
211
:~
FIGURE4. An example of a spatially well-defined behavior setting, this one to double-function as a quiet
nook for reading/listening, a breakaway and retreat area, and occasionally an informal napping area
(Pattern 1016 from Recommendationsfor Child Care Centers, 1979).
OA
O8
OA
Xz
O8
OA
X~
O8
FIGURE5. Diagram of the quasi-experimentalpost-test-only control group design with multiple levels of
treatment and proxy pretest measures, using the nomenclature of Cook and Campbell (1979).
(Campbell, 1957; Campbell and Stanley, 1966; C o o k and" Campbell, 1976, 1979).
They have systematically explicated the threats to valid causal inference when the
major features of laboratory-centered experimental research are not present (e.g.
when randomization, physical isolation of respondents, short duration of the
experimental treatment and strict control over all extraneous independent variables
cannot be achieved). In research concerning everyday field settings, it is the
researcher's task to systematically rule out as m a n y alternative explanations limiting
valid causal inference as possible. The Campbell approach and quasi-experimental
research designs form the methodological basis for the current study, and the threats
to internal and external validity form the basis for caution in interpreting the
findings.
Research design and rationale
It would have been ideal to redesign existing child care centers in the directions
suggested by the hypotheses, to use other centers as controls, and to randomly
assign children to conditions, thus permitting a true experimental design. Practical
limitations working in naturalistic field settings did not make this feasible. Thus, a
number of existing centers were located that were characterized by the physical
212
G.T. Moore
O
.
,,,._-~
~,
.
,_,_.--,
%.~.,,~,."~:,
,.
.,
rr
~o ~
l~IE____I~I~
""
~-;-"(I
~.~...
~.~
,/ / ' ~ \ ~
~.~
: ~
If-H
FIGURE 6. Example of a child care center with spatial well-defined behavior settings. Note the two
overhead lofts, the well-defined kitchen accessible to children underneath the loft, the stairways from the
lofts providing partial separation between behavior settings, and the use o f bookcases and different
carpet textures to shield and differentiate 'activity pockets' or behavior settings.
i :.--'."i.
s t ~
'
Up
N
oowo /-H ]
,r [[
L
I
I0
ft
FIGURE 7. Example o f a child care center with partially defined (transitional) behavior settings. Note the
lack of overhead differentiation or any strong vertical elements, but also note the clustering o f similar
types of furniture together and the use of carpets and some low furniture to partially differentiate behavior
settings.
patterns hypothesized to have impacts on children's behavior, and other centers that
did not have the patterns.
Inasmuch as it was impossible to assign subjects randomly to conditions, subject
groups were matched and proxy pretest data were collected to aid in matching
groups, to check on the equivalency o f groups, and to use as covariates to remove
remaining initial group differences in subsequent data analyses. To determine the
\Ill
office
-~--'-~
~
13
IlLunge~
213
~
o
tcL_/I
1o
~.......
'~
]-(-:::ill: i!J
z--'I .........
L-I
s' Ir
I
>N
I
0
I
I0
fi
FIGURE 8. Example of a child center with poorly defined behavior settings. Looking at the three child
care rooms on opposite ends of the building, note the wide open spaces and the lack of any dividers,
overhead elements, book cases, or other furniture used to differentiate behavior settings.
relative effects of different treatment levels, three levels of the independent environmental construct (definition of behavior setting) were selected. Thus, following the
nomenclature of Cook and Campbell, the research design would be referred to as a
combination of two quasi-experimental nonequivalent control group designs, the
post-test-only design with multiple levels of treatment and the untreated control
group design with proxy pretest measures (Cook and Campbell, 1979, pp. 98-99,
112-115) and would be depicted diagrammatically as shown in Fig. 5, where X
stands for the treatments in three ordinal levels from poorly defined to well defined
spatial settings, O stands for proxy pretests and post-treatment observations, where
the subscripts A and B refer to different measures, and where the dotted lines refer
to non-equivalent groups, intact rather than randomly assigned.
Settings
The settings were 14 child care centers in Milwaukee County selected to represent a
minimum of two sets of centers for each of the three levels of the spatial definition
of behavior settings: well-defined, moderately defined (transitional) and poorly
defined (see Figs 6-8). The degree of spatial definition was measured in terms of ten
variables, including the degree to enclosure, degree of visual separation from other
settings, appropriate size, degree of separation from circulation, and appropriateness
of the amount of storage, work surfaces, and display space to the activity in the
setting. Independent verification was achieved using judges' ratings on the ten-item
Early Childhood Physical Environment Scale.
The centers were selected to be similar and were matched as well as possible in
terms of size, socio-economic status of the children, general educational philosophy,
214
G . T . Moore
and four specific dimensions of teacher style. Thus for example, three centers similar
in size, stated educational philosophy, and observed teacher styles were identified
within a five-block radius in a lower-middle income mixed ethnic area that had,
respectively, well-defined, partially defined and poorly defined behavior settings. A
large middle-income center was also chosen that had a varied clientel
but had well defined, transitional, and poorly defined settings in different rooms.
The effectiveness of this matching strategy was assessed by a Center Profile, a
Teacher Style Rating Scale, and a Dimensions of Education Rating Scale (described
below). No significant differences were found between centers, children or teachers
on 23 of the 28 measures (82.14%). Remaining differences (center size, open versus
traditional philosophy, and socio-economic status) were used as proxy pretest or
control variables in subsequent analyses of covariance thus removing most of the
initial group differences and increasing the interpretability of the quasi-experiment
(Cook and Campbell, 1974, p. 114).
Subjects
The subjects were selected on the basis of random space and time sampling. They
ranged from 2.5 to 6 years of age. Each center was mapped and divided into a grid
of cells. The days when each center were observed was determined randomly, the
time of day and order of observing different rooms and spaces were determined
randomly, and the time and order of observing particular cells were determined
randomly. Every child starting in or entering a cell during an observational period
was observed. All cells were observed equally over equal periods of time. Cases were
the observational cells. A total number (N) of 1,061 cases were observed.
Physical environmental variables
To ensure construct validity for the physical environmental variables, a detailed
operational definition and rating scale were prepared for each level of spatial definition. The requirement that these scales focus on specific physical environmental
variables meant that the Early Childhood Environment Rating Scale of Harms and
Clifford (1980) and our own Facility Inventories used in an earlier study of child
care centers (Cohen et aL, 1978) were not appropriate. Inspection of the Harms and
Clifford scale indicated that it deals more with the social than the physical environment, and in fact makes no distinctions between different types of physical settings.
Our own earlier inventories were too general for the present study. The new pair of
scales developed for this study are called the Early Childhood Physical Environment
Scales. Using these scales, the validity of the selection of settings was tested by
having a panel of three judges not familiar with the hypotheses independently rate
each of the settings on a series of ten five-point items. The results of this pilot test
indicated that the behavior settings were significantly different from each other in
terms of spatial definition and always in line with the characterization given them
by the principal investigator (paired sample t-tests ranging from 1-97 to 4.15, df =
59 to 119, P-values ranging from <0.05 to <0.001).
Subject group variables
As the dependent behavioral consequences measured in this study were based on an
ongoing 'treatment' (the ongoing program and design of each center), proxy 'pre'test
215
* This strategy is recommended by Cook and Campbell (1979, pp. 112-115) for the 'untreated control
group design with proxy pretest measures'. As this particular quasi-experimentalresearch design has not,
to the best of my knowledge, been used previously in the environment-behavior literature, a brief
description may help. From Cook and Campbell: 'Different pretest measures are also needed when
evaluating the consequences of an ongoing practice, since in this case it may not be possible to collect
any measures other than those found in the archives or those where changes are not likely to be affected
by the treatment (e.g., stable characteristicslike age, sex, or socioeconomicstatus). Such variables function
in the design and analysis as proxies for the pretest' (p. 113).
216
G . T . Moore
217
field study such as the present one. The closest observational schedules to that
needed for the present study have been reported in the environment-behavior
literature by Prescott et al. (1972), Harms and Clifford (1980), Perkins (1980), and
Smith and Connolly (1980).
Based on review of the above, an Environment/Behavior Observation Schedule
for Early Childhood Environments appropriate for the present study was constructed. The main data recording sheet is comprised of three types of observations:
(1) setting, (2) individuals and (3) observed behaviors.
The major portion of the observation schedule is given over to space for
recording observed behaviors in seven categories: (a) general type of behavior
(engagement, transitional, functional, random and withdrawn); (b) initiation of
behavior (spontaneous free, individual directed and group-directed); (c) quality of
exploratory behavior (immersed, somewhat involved, not involved); (d) type of social
interaction (cooperation, competition, aggression, affection); (e) degree of social
interaction (reciprocated, acknowledged, not acknowledged, no social interaction); (f) type of teacher involvement (co-action, encouragement, control, information, observation, presence but no involvement, no teacher present); and (g)
type of teacher-teacher interaction (group, colleague, observation, more than one
teacher but no interaction, one or no teacher present). Ordinal and interval scales
were developed within categories (e.g. distracted, attending and immersed under
the category of engagement; parallel, associative and cooperative under cooperation).
Each of the categories and scales was based on the existing research literature for
that behavior type. For example, Parten's (Parten, 1932; Parten and Newhall, 1943)
conceptualization of the stages of social participation of preschool children has
become one of the classics of child psychology. She introduced six categories of
social participation in play behavior: unoccupied behavior, solitary play, onlooker
behavior, parallel play, associative play, and cooperative play. This category system
has been used in many studies and has been simplified and re-examined (e.g. Smith,
1978). In the present case, Parten's six categories have been divided into two sections
of the Observation Schedule. The two behaviors that are not strictly play behaviors
(unoccupied and onlooker) were recorded under 'general type of behavior', while
the four types of social play (solitary, parallel, associative and cooperative) were
recorded under 'type of social interaction--degree of cooperation'.
While space does not permit giving all the operational definitions here, a sample
of the most important observational categories used for later analyses can be summarized:
Engagement
Child is visually and/or physically involved with a point of focus. Point of focus
may be another person, object or activity. Behaviors include looking at, listening to,
participating in a prescribed or spontaneously initiated activity with, and/or touching
or manipulating the point of focus.
Random behavior
Behavior that is nondirected and shifts rapidly from one setting or object to another.
Includes behavior that is hyperactive, impulsive, fast moving, and ineffective, and
actions that are incomplete.
218
G . T . Moore
Withdrawn
Behavior that is not focused on any activity. Characteristic behaviors include
vacant staring, staying close to adults without visually or physically exploring the
environment, and indications of fearfulness such as crying, hiding, thumb-sucking,
auto-manipulation, and trembling.
Exploration
Behavior that is directed toward investigating, examining, studying, or searching for
an object, activity, person, setting or other points of focus, including inspection,
asking questions, manipulation and producing effects, etc.
Based on these and similar considerations, a detailed set of operational definitions
and coding b o o k were prepared for training observers, for checking reliablility, and
for use in the observation sessions.* For ease of recording, the schedule was reduced
to one page, as shown in Fig. 9.
After three training sessions, interobserver agreement was significantly high (percentage of exact agreement = 85.71, P <0.001; kappa k = 0-86, P <0.001).
Procedure
The study was conducted in three phases. First, centers were located and contacted,
permission forms for the teachers and parents of children were completed, and
demographic profiles were filled out by the center director and parents. Second,
each teacher in the 14 selected centers was asked to fill out the Teacher Style Rating
Scale and the Dimensions of Education Rating Scale. Third, the children's and
staff's behaviors were observed in each of the settings on a r a n d o m space and time
sampling basis.
The observer stationed herself in a position in each r o o m that would provide
views of all portions of the room, but would not interfere with the children's
behavior (e.g. the corner of a large loft, a chair behind a bookcase, in a corner of
the room). This was done with sufficient time before the beginning of observation to
allow the children to adapt to the new person in the room. A mini-tape-recorder
with unobtrusive ear-plug, clipboard with observation forms, and pencils were all
the equipment needed by each observer. There were no obvious signs of disruption
or complaints caused by our observations.
Observations were conducted by two trained observers over a period of a month.
They were done for 20 min sessions, then a 5 min break, and then additional sessions
of 20 min each for a 2.5 hour observational period. Each observation lasted for 10 s
(timed by beeps on the tape-recorder) with 1 min and 50 s for recording and a deep
breath. At each time beep of the recorder, the observer moved her attention to the
next randomly selected observation cell.
* The behavior mapping instrument, called the Environment/Behavior Observation Schedule for Early
Childhood Environments, together with operational definitions and coding book and the other scales
used in this study are available from the author.
ENVIRONMENT/BEHAVIOROBSERVATIONSCHEDULEFOREARLYCHILDHOODENVIRONMENTS
Observer_ _
Date_ _
ENVIRONMENTALSETTING
The Locationof the ObservedBehavior
Center
[]
[]
Observational Cell
Genders
Ages
Ethnicity
SECTION2: CHILD-INITIATEDVSSTAFF-DIRECTEDBEHAVIOR
Initiated
[]
[]
Adults
[]
Girls
[]
Boys
[]
2 to 3
[]
3to4
[]
4 to5
[]
5to6
[]
6 and over
[]
White
[]
Black
[]
Hispanic
[-7
Other
[]
Exploration
[]
Attending
Transitional
[]
[]
[]
Interaction
[]
EmptyCell
[]
[]
[]
[]
Not Acknowledged
[]
[]
SECTIONS: COOPERATIONC
, OMPETETIONA
, GGRESSION,AFFECTION
Cooperation
[]
CooperativeActivity [ ]
Associative Activity [ ]
Parallel Activity
[]
AbsoluteGains
[]
[]
Relative Gains
[]
[]
[]
ThreatenedAttack
[]
[]
Verbal Abuse
[]
Intimate Physical
[]
[]
[]
Primarily Engaged
[]
Only Functional
[]
Affection
[]
[]
[]
Not Applicable[ ]
Involvement
[]
[]
Vacant Staring
[]
Intermlttant Focusing [ ]
Passive Observation
[]
Unclear
[]
Friendly Physical
[]
Verbal
[]
Unclear
[]
SECTION6: TYPEOFTEACHERINVOLVEMENT
No SustainedActivity [ ]
Not Applicable[ ]
Co-Action
[]
Encouragement
[]
Control
[]
Information
[]
Observation
[]
No Involvement
[]
Unclear
[]
SECTION7: TYPEOFSTAFF-STAFFINTERACTION
Interaction
[]
Archilccture .U0iversityof
& Ur4b~n Wisconsin.
Planning Milwaukee
Reciprocated
Acknowledged
Spontaneous Interest [ ]
Withdrawn
[]
[]
[]
Distracted
Directed Interest
[]
PhysicalAttack
Partially Functional [ ]
Random
Immersed
SECTION4: SOCIALINTERACTION
Aggression
Only Transitional
Primarily Engaged
[]
Partially Transitional[ ]
Functional
[]
Rivalry
SECTIONi: GENERALTYPEOFBEHAVIOR
I .... sed
Group Directed
Unclear
SomewhatInvolved
OBSERVEDBEHAVIORS
ObservableBehaviorsCharact
as a Wholeor for Most of the ObservationSegment
[]
[]
SECTION3: EXPLORATION
Children
Competition
Engag.... t
SpontaneousFree
Individual Directed [ ]
[]
INDIVIDUALSINVOLVED
The Numberand Characteristics of Children and
Adults Involved
[]
Seq # - -
OBSERVEDBEHAVIORS{continued)
Room/Area
Group Size
Time_ _
Not Applicable[ ]
Group
[]
Colleague
[]
Peer Observation
[]
NO Interaction
[]
Unclear
[]
FIOURE 9. The Environment-Behavior Observation Schedule (behavior map) for Early Childhood
Environments. Complete instructions for its use are available from the author.
220
G . T . Moore
Data analysis
The results of interactions between physical environmental variables (the spatial
definition of behaviour settings), subject group variables (the various measures of
teacher style and philosophy o f teaching), and resultant dependent behavioral
variables (social and cognitive b e h a v i o r s ) w e r e analysed using multivariate
parametric and non-parametric statistics available in the SPSS package (Nie et al.,
1981). Included were nominal and ordinal data analysed by three-way Z2 tests
controlling for subject and environmental variables and Kruskal-Wallis one-way
analysis of variance with ranks.
The majority of the data was interval in level and was analysed using three-way
factorial analyses of covariance (ANCOVA) on the adjusted means on the dependent
variables with two covariates, for example, settings x teacher styles x dimensions
of education, with socio-economic status and center size as covariates *. This analysis
was deemed most appropriate for the data collected for a number of reasons, among
them that the physical environmental variables (the spatial definition of behavior
settings) and the subject group variables (the scales of teacher style and o f open
versus traditional philosophy of education) were independent of each other, that the
factorial analysis allows the important and too-often neglected analysis of interaction
effects, that the covariates though gathered concomitantly with the dependent
measures served as appropriate proxies for pretest variables (Cook and Campbell,
1979), that the proxy pretest covariates are stable characteristics and therefore not
likely to be affected by the treatment, and that the analysis of covariance allows the
analysis of F-scores based on adjusted means, SS, MS and df, or, said differently,
the analysis of covariance is appropriate for eliminating any initial group differences
remaining after the matching of centers and subject groups. Power was increased
by using multiple covariates which were also un-correlated with each other. The
covariates and dependent variables were assumed to be colinear, though this was not
tested directly.
The A N C O V A procedures are not without difficulty and potential sources of
bias, however, when used with intact groups and the non-equivalent control group
design. This has been discussed in detail by Reichardt (1979). Despite the matching,
multiple treatment levels, and analysis of covariance of some of the expectedly
most important remaining initial group differences, the presence of other selection
differences between the groups is inevitable. Unfortunately, various forms of
experimental control, randomization and design elaboration are not possible in
many field settings, leading to a balancing act between internal and external validity.
A degree of uncertainty always remains in quasi-experimental non-equivalent control
group studies, and despite relatively powerful quantitative analysis of the data, the
results and conclusions below must be rendered as tentative and suggestive of causal
connections, not as conclusively demonstrative.
Results
The findings are presented in terms of four subsets o f hypotheses, those pertaining
* All reported Fs and df in the analyses of covariance pertain to adjusted means on the dependent
variables due to the use of the covariates within the statistical analyses. Though collection occurred
concomitantlywith that of the dependent variables, the proxy pretest variables used as covariates meet
the assumption that scores on the covariate not be influenced by the treatment variables as they are
stable characteristicsof the subjects and centers.
221
to (a) style of behavior and degree of engagement, (b) child-initiated and exploratory
behavior, (c) group sizes and cooperative behavior, and (d) teacher involvement.
Each will be examined in turn.
But first, a binomial test of proportions indicated that child care centers with well
defined behavior settings have significantly more such settings (sometimes referred
to as 'activity pockets') than centers with moderately and poorly defined areas (Zvalues ranging from 1.73 to 2.04, P-values ranging from < 0-05 to < 0-02). That is,
not only do some centers tend to have better defined behavior settings, but they also
tend to have more of them. In many cases this is because the same overall square
footage is subdivided into smaller, and therefore better defined 'pockets'. It remains
to be seen whether this has any impact on behavior.
Engaged behavior
Well-defined
Partially
defined
Poorly
defined
38-9
3-3
25.0
3.1
26-2
3.5
performed to see if, in the absence of main effects, there might be any interaction
effects. The results indicated a small but statistically significant relationship. For
centers with teachers scoring high on the scale of open educational philosophy,
there was a significant interaction between the definition of the behaviour setting
and the philosophy of education jointly affecting engaged styles of behavior (~(z =
4-68, df = 1, P < 0.05). The physical environment and teaclfing philosophy appear
to reinforce each other in affecting engaged versus non-engaged behavior only at
the 'highest' levels of both spatial definition and open educational philosophy.
Inspection of the three-way analysis of covariance (ANCOVA) in Table 2 indicates
that, where it does occur, the degree of engagement is more intense depending on
the spatial definition of the behavior setting (F = 6-51, df = 2, 654, P < 0.01) and the
overall size o f the center (F = 20.40, df = 1,654, P < 0-001). A follow-up multiple
classification analysis (Nie et al., 1981) and an ordinal Kruskal-Wallis one-way
analysis of variance with ranks confirmed that it was indeed in the spatially welldefined behavior settings where the most immersed forms of engaged behavior
occurred (Kruskal-Wallis; Z2 = 30-59, df = 2, P <0.001).
222
G . T . Moore
TABLE 2
Adjusted df
Adjusted MS
1
1
1.89
10.71
3.60
20.40"**
2
2
1
7
654
3"41
0.42
0-54
2.93
0.53
6.51 **
0.81
1-02
5.58 ***
Covariates
SES
Center size
Main effects
Behavior setting
Teacher styles
Open-traditional
Explained
Residual
Note. Due to empty cells or a singular matrix, interactions could not be calculated.
** P < 0-01, *** P < 0.001.
TABLE 3
Initiation of behavior in spatially well-defined, partially defined and poorly defined behavior
set tings
Variation
Well-defined settings
Transitional settings
Poorly defined settings
Source of Variation
Covariates
SES
Center size
Main effects
Behavior setting
Teacher styles
Open-traditional
Explained
Residual
s.o.
419
296
308
1.93
1.91
1-98
0.81
0.80
0.81
Adjusted df
Adjusted MS
1
1
1.45
3.70
2.31
5.88
2
2
1
7
1009
0.04
1.54
11.09
2-57
0"63
0.06
2.44
17.61 ***
4.08 ***
Note. Due to empty cells or a singular matrix, interactions could not be calculated.
* P < 0.05, *** P < 0.001.
223
TABLE 4
Exploratory behavior in spatially well-defined, partially defined and poorly defined behavior
settings
Variation
S.D.
434
304
317
0"78
0.59
0"65
0.91
0.79
0'79
Adjusted df
Adjusted MS
1
1
1-20
7-54
1.61
10-84"**
2
2
1
7
1009
4-10
1.60
2.23
2-85
0"70
5.90 **
2.30
3-21
4.09 ***
Well-defined settings
Partially defined settings
Poorly defined settings
Source of Variation
Covariates
SES
Center size
Main effects
Behavior setting
Teacher styles
Open-traditional
Explained
Residual
Note. Due to empty cells or a singular matrix, interactions could not be calculated.
** P < 0.01, *** P < 0.001.
TABLE 5
Relative group size, number of children, and number of adults in spatially well-defined, partially
defined and poorly defined behavior settings
Welldefined
Partially
defined
Poorly
defined
Group size
M
S.D.
Residual
5.54
5.02
5-83
6"90
5.16
4.39
No. of children
M
S.D.
Residual
4-93
4"58
No. of adults
M
S.D.
Residual
0.64
0'75
4.86
6-21
0.97
1.21
df
MS
34.82
1.17
1058
29-86
35-35
1058
23.56
9.87
1058
1-01
1.50
4-34
3"57
9.74*** -
0"83
1"10
224
G . T . Moore
TABLE 6
Degree of social interaction in spatially well-defined, partially defined and poorly defined
behavior settings
Variation
S.D.
436
303
315
1.56
0.95
1.10
1.38
1-23
1.28
Adjusted df
Adjusted MS
1
1
7-20
0.17
2
2
1
7
621
12.81
10-76
5.01
7-77
1.66
Well-defined settings
Partially defined settings
Poorly defined settings
Source of Variation
Covariates
SES
Center size
Main effects
Behavior setting
Teacher styles
Open-traditional
Explained
Residual
4-34"
0.10
7"72"**
6.49 **
3.02
4.68 ***
Proportion of cooperative versus competitive behavior in spatially, welldefined, partially defined and poorly defined behavior settings
Competition
Cooperation
Well-defined
Partially defined
Poorly defined
0-2
41 "8
2-1
25.6
1"4
29.0
225
TABLE 8
Level of teacher involvement with children in spatially well-defined, partially
defined and poorly defined behavior settings
N of instances
M ranks
Well-defined
Partially defined
Poorly defined
205
252.31
166
253.12
159
295.42
226
G . T . Moore
227
styles and philosophy of education) that the physical environment does not--and
maybe cannot--have an independent impact. This interpretation gains some support
from the pair of findings that both of the negative results--initiation of behavior and
type of behavior--were significantly related in the first case to an open philosophy
of education and in the second case to a mild interaction between open philosophy
and type of setting. Nothing behind this study presupposes an environmental determinist, or empiricist orientation, and these findings are in fact not supportive of a
strictly empiricist theory (this general point is discussed in more detail in Moore, in
press).
Another possibility is that only certain types of behaviors, perhaps those that are
inherently more active adaptations to the environment (like immersion in activities
and exploratory behavior) and those that are more small-group social (like reciprocated social interaction, mutual cooperation, and co-action between staff and
children), are subject to the influences of well articulated spatial settings that are
acoustically and visually separated, provide a stage-set for less interruptions and
longer attention spans, and thus to greater involvement in cognitive and social
developmental activities. The social behavior side of this interpretation gains some
tangential support from the findings of Smith and ConnoUy (1980) that children
play in smaller social groups or switch from larger parallel subgroups to smaller
parallel subgroups when resources are increased; it may be that they are affected in
this way both by the definition of the space and by the amount of resources, but this
remains to be tested. Similarily, the possibility that the spatial definition of behavior
settings may effect social and exploratory behaviors through the mediating variables
of less interruptions and longer attention spans was not tested in the present study
and awaits empirical investigation.
For many of these issues, however, significant interaction effects were found for
group x setting interactions. For example, we saw that there was a significant
interaction between the definition of behavior settings and the philosophy of
education jointly affecting engagement in developmentally supportive behaviors.
Similarly, the degree of spatial definition of behavior settings and teacher styles
(open styles of preschool education) combine to affect the amount of behavioral
episodes initiated by children versus by teachers. And third, the type of social
interaction among children---cooperation in particular--is related to the definition
of the setting in interaction with the teaching style of the staff members present.
A caveat must, however, be offered. The work of Smith and Connolly (1980)
indicated that not only is the amount of space important in predicting the type of play
observed in preschools, but also one must take into account the scarcity or abundance
of resources within that space. They found, for example, that when more equipment
was available, children play in smaller subgroups or switch from large parallel subgroups to smaller parallel subgroups or solitary play. The present study included a
rough measure of the amount of resources available within the behavior setting as one
of the ten defining characteristics of well defined settings. Due to the quasi-experimental
nature of the present study, where the independent environmental treatment variables
could not be directly manipulated or changed, but where existing centers were used as
the basis for study, it was not possible to systematically vary the amount of resources
independently of the degree of implied spatial enclosure. These two important facets of
what we have previously termed 'Resource-Rich Activity Pockets' (Pattern 908 of
Moore et al., 1979) remain to be disentangled in future studies.
228
G.T. Moore
This study was supported by a grant from the National Endowment for the Arts
No. AZ15533-80. An earlier version of the paper was presented at the Division of
Population and Environmental Psychology, American Psychological Association
annual convention, Anaheim, California, August 1983.
My thanks to Seymour Wapner and Ina Uzgiris for their advice on the Study, to
Naomi Leiseroff and Marleen Sobczak for assistance in data collection, to Thomas
Laurent for assistance in data analysis, and to Sherry Ahrentzen, David Canter,
Roger Hart, Daniel Stokols, David Stea and two anonymous reviewers for
comments on an earlier draft.
References
Abt Associates. (1979). Children at the Center: Final Report of the National Day Care Study.
Cambridge, MA: Abt Associates.
Aiello, J. F., Gordon, B. and Farrell, T. J. (1974). Description of children's outdoor activities
229
230
G . T . Moore
231
for understanding the transactions of men and environments. Environment and Behavior,
5, 255-289.
Wicker, A. W. and Kirmeyer, S. L. (1976). From church to laboratory to national park. In
S. Wapner, S. B. Cohen and B. Kaplan (eds), Experiencing the Environment. New York:
Plenum.
Wohlwill, J. F. (1980). The confluence of environmental and developmental psychology:
signpost for an ecology of development? Human Development, 23, 354-358.
Wright, R. J. (1975). The affective and cognitive consequences of an open education
elementary school. American Educational Research Journal, 12, 449-468.
Zeisel, J. (1981). Inquiry by Design. Monterey, CA: Brooks/Cole.