STATE OF PRACTICE OF RESIDENTIAL FLOOR SLAB FLATNESS

By Kenneth D. Walsh,1 Howard H. Bashford,2 and Brad C. A. Mason3

ABSTRACT: Forensic engineering investigations of residential structures often include measurements of ele-
vation differences across the slab-on-grade floor as an indicator of movement (settlement or heave) of the
underlying soil. Such assessments are predicated on an assumption that the slab was flat or very nearly so at
the time of initial placement. Despite the relatively frequent use of such measurements in practice, confirmation
of the underlying assumption is scarce in the archival literature. This paper describes a testing program to
measure the deviation from level of a large number of residential concrete slab-on-grade floors in Phoenix,
Arizona, within a few days of concrete placement. An average maximum elevation change of 13.5 mm was
observed. Average and maximum slopes are also presented. The construction of the residential structure atop
the slab was found to have no statistically significant effect on the levelness of the slab, nor did the season in
which the concrete was placed. Residential slab-on-grade construction methods are reviewed, concluding that
this degree of deviation from level is to be expected. Comparison of the data with recommendations in the
literature suggests that caution must be taken in evaluation of cosmetic or structural impairment of residential
structures based in large part on measurements of interior floor slab levelness.

INTRODUCTION sparse. There are a number of standards for such measure-

This paper presents the results of an investigation of the ments, but very few actual measurements. Significantly more
degree to which residential floor slabs deviate from level when attention has been paid, especially in the recent literature, to
they are placed. The study is motivated by the general obser- the measurement of the degree to which slabs in homes that
vation that measurements of deviation from level of floor slabs have experienced some damage are out of level. In this sec-
are commonly used as an indication of movement of the floor tion, the technical literature on measurements of both newly
slab. This observation has also been made in the relevant lit- constructed and serving residential floor slabs will be re-
erature [for example, Koenig (1991); Marsh and Thoeny viewed, along with a brief discussion of applicable standards.
(1999)]. Measurements of the deviation from level of the floor The most straightforwardly applicable literature is that
slab are a common feature of forensic investigations of homes which contains measurements of newly constructed floor slabs
that have experienced some level of cosmetic or functional of the sort obtained in this study. Remarkably few such results
distress, and conclusions are often based on these results. could be found in the literature. Koenig (1991) is the most
These conclusions are almost always based on the assumption extensive such study we identified with description of new
that the floor slab was flat and level at the time of construction, slabs as a primary goal. Koenig presents data from 54 newly
and therefore any deviation from level must result from placed residential slabs in Texas. The elevations of nine points
postconstruction movements. on each slab (one in the center and eight around the perimeter)
However, the engineering community has also recognized were measured with a water level manometer. Koenig relates
that this assumption is likely incorrect. Specialized screeding that his sample exhibited a range of 3.2 to 25.4 mm (1/8 to 1
and finishing equipment and processes must be adopted to in.) of maximum elevation difference across the slab. The av-
achieve high degrees of flatness and levelness in industrial erage maximum deviation from level was 13.7 mm (0.54 in.),
floors, requiring a much higher standard of care and different with a standard deviation of 5.33 mm (0.21 in.). Marsh and
processes than are employed in residential construction. While Thoeny (1999) report a range of 15.2 to 25.4 mm (0.6 to 1
qualitative understanding of this issue exists, quantitative in- in.) and an average of 19.1 mm (0.75 in.) for a small study
formation about the deviations from level of newly placed res- conducted on only six newly constructed homes in southern
idential slabs has not been systematically obtained and made California. Bondy (2000) reports a range of 12.7 to 22.9 mm
publicly available. To date, the technical literature on this topic (0.5 to 0.9 in.) and an average of 17.8 mm (0.7 in.) for a
has included only a relatively small number of actual slab sample of four one-day-old slabs in California.
measurements. In this paper, the technical literature will be By contrast, the literature contains a considerable amount
reviewed in some detail. This will be followed by a description of work on the relationship between different types of slab
of the measurement program for this study, with analysis and distortion and the type and severity of structural damage that
statistical evaluation of the results. Finally, conclusions about may be related. The classic work in this area is Skempton and
residential floor slab flatness will be presented. MacDonald (1956), who used extensive studies of many dif-
ferent types of buildings over an extended period of time to
LITERATURE REVIEW determine the maximum settlement before damage is likely to
The literature on the degree to which newly placed residen- occur. The specific focus of their research was on the total and
tial concrete slabs-on-grade may be out of level is relatively differential movement of foundations, as opposed to slabs, but
to the extent that the performance of interior walls that bear
1
Assoc. Prof., Del E. Webb School of Constr., Arizona State Univ., on the slab would be affected by similar distortions, the
Mail Code 0204, Tempe, AZ 85287-0204. E-mail: ken.walsh@asu.edu Skempton and MacDonald (1956) results would be applicable.
2
Assoc. Prof., Del. E. Webb School of Constr., Arizona State Univ.,
Mail Code 0204, Tempe, AZ 85287-0204. Furthermore, their work is frequently cited as threshold criteria
3
Grad. Res. Asst., Del E. Webb School of Constr., Arizona State Univ., for the appearance of cosmetic damage in a structure.
Mail Code 0204, Tempe, AZ 85287-0204. Skempton and MacDonald (1956) point out that the dam-
Note. Discussion open until April 1, 2002. To extend the closing date aging movements for the panels and finishes are actually the
one month, a written request must be filed with the ASCE Manager of distortions of those panels, rather than simple translations.
Journals. The manuscript for this paper was submitted for review and
Based on their evaluation of the damage observed in a number
possible publication on September 19, 2000; revised May 5, 2001. This
paper is part of the Journal of Performance of Constructed Facilities, of buildings, they reported a range of permissible slopes de-
Vol. 15, No. 4, November, 2001. ASCE, ISSN 0887-3828/01/0004- pending on the type of building or foundation. The most com-
01270134/$8.00 $.50 per page. Paper No. 21977. monly selected criterion from this work is 1/300 as a threshold
JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / NOVEMBER 2001 / 127
rotation before cracking of panels and finishes can be expected types and conditions, along with the number, location, and
in framed structures. Strictly speaking, this recommendation severity of different types of cracks in the slab and structure.
was not developed or intended for light residential structures, With these observations and other results from the literature,
but it has nonetheless been repeated many times in the tech- Marsh and Thoeny (1999) related a number of measures of
nical literature. slab levelness and crack size to the damage level. These mea-
Day (1990) studied the angular distortion and slab differ- sures are shown in Table 1 along with the Boone (1996) re-
ential elevation of 34 residences in the San Diego metropolitan sults. Marsh and Thoeny (1999), while clearly derivative,
area. These slab characteristics were linked to the observed show general agreement with Boone.
level and severity of cosmetic and structural damage to the Boone (1996) emphasizes the importance of descriptions
building. A range of observed damage from negligible to very and measurements of cracks in determining building damage
severe was reported for the homes in the study. It was con- level. Marsh and Thoeny (1999) agree with crack width as an
cluded through this analysis that cracking of panels in wood- important measurement, but add a number of other measures.
frame structures on lightly reinforced concrete foundations Additional measures of severity presented by Marsh and
may occur when the angular distortion of the foundation Thoeny include the total slab differential elevation and the
reaches or exceeds 1/300, and structural damage may result angular distortion of the slab. They provide a range of values
when the rotation exceeds 1/100. for these variables in different damage severity classifications,
Boone (1996) was generally critical of the practice of em- depending on the specific type of soil problem at the site. They
phasizing individual measurements in the evaluation of struc- divide potential soil problems into three categories: expansive,
tural damage. Boone proposes instead a systemic evaluation, settlement, or slope stability.
including a number of additional parameters such as the size, The measurements presented later in this paper were made
location, and severity of cracks that occur (Table 1). in new residential units in the Phoenix, Arizona, metropolitan
Marsh and Thoeny (1999) used the Boone approach in eval- area. In Arizona, as in many jurisdictions, residential slab lev-
uating patterns of movement for different soil conditions and elness is not strictly controlled at the time of construction.
residential slab-on-grade systems. As opposed to previous However, the Arizona Registrar of Contractors (AROC) may
studies, this research considered not only the slab and struc- be asked to evaluate construction quality in the event of a
tural damage and the amount of slab movement, but also the dispute between a licensed contractor and a customer of the
types of soil conditions and foundations. These characteristics contractor. In these cases, the AROC will use its minimum
were then related to the specific type and severity of associated workmanship standards (AROC 1997). These standards pro-
damage to the structure and slab. The research design involved vide, among other things, a value for the maximum angular
measurements of the levelness of lower-level floor slabs of distortion of a new concrete slab. While not precisely a stan-
one- and two-story detached residential homes and one- and dard, the AROC values are intended to describe the minimum
two-story attached and detached condominium/townhouse tolerance the contractor should be able to achieve.
units using a water level manometer. Additionally, visual ob- A number of damage threshold criteria are presented in Ta-
servations of the slabs were used to determine the amount and ble 2. The values shown in the table are the thresholds reported
types of distresses that these slabs had experienced since being by each study before damage of some type, even cosmetic, is
constructed. likely. Also shown for comparison is the AROC workmanship
A total of 401 homes were reviewed over a 12 year period, standard, which is considerably flatter than any of the thresh-
122 of which were available to assess the condition of the old values reported. It must be kept in mind that all of the
slab. Information recorded included visual observations of soil results in Table 2, with the exception of the AROC standard,

TABLE 1. Severity Characterizations from Literature

Boone (1996) Marsh and Thoeny (1999)
Damage level Description Damage level Description
Negligible Hairline cracks Negligible None
Very slight Fine cracks appear that can be easily resolved with Slight or low Aesthetic damage; some doors and windows stick;
simple redecoration; slight exterior cracks visible cosmetic repairs most likely necessary
upon close inspection and/or isolated slight frac-
ture in building
Slight Cracks appear that can be easily filled; redecoration Low to moderate More pronounced aesthetic damage; some doors
probably required. Several slight fractures may and windows stick; accelerated weathering to
appear showing inside structure of building; ex- external features; repairs to slab and foundation
ternal cracks are visible and may need physical often necessary
repair
Moderate Cracks appear that may need opening up and can Moderate Distinct patterns of interior and exterior distress
be patched by a mason; external cracks present; appearing; pervasive cracking may be apparent;
doors and windows stick; service pipes may frac- tile in floor usually perceivable; repairs to home,
ture; weather tightness may be impaired slab, and foundation usually necessary
Severe Damage includes large cracks requiring extensive Moderate to se- Distinct patterns of distress exist; window and
repair work, including breaking out and replacing vere door frames distorted; perceivable tilt in floor;
sections of walls (especially over doors and win- serviceability of building affected; repairs to
dows); distorted door and window frames; lean- home, slab, and foundation necessary
ing of building walls; noticeably sloping floors;
disrupted service pipes
Very severe Damage often requires major repair job involving Severe Large pervasive cracks throughout; door and win-
partial or complete rebuilding; beams lose bear- dow frames severely distorted; wall frames lean-
ing; walls lean and require shoring; windows are ing noticeably; repairs to home, slab, and foun-
broken with distortion; danger of structural insta- dation are necessary and often extensive
bility Very severe Increasing risk of structure becoming dangerous;
requires major repair and often removal of struc-
ture

128 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / NOVEMBER 2001

 TABLE 2. Damage Threshold Criteria where z = average elevation difference (positive uphill) be-
Maximum Maximum tween points 3.05 m apart; and sz = standard deviation of sam-
differential angular ple of z values.
Source elevation distortion The ACI standard further defines a local FL in which only
Skempton and McDonald Not given 1/300 a single pair of elevation measurements 3.05 m apart is used.
(1956) For this case, (1) can be rewritten as follows:
Day (1990) Cosmetic damage: 31.8 mm 1/300
(1.25 in.) 317.5 mm
Structural damage: 88.9 mm 1/100 FL = (2)
(3.5 in.) z
Marsh and Thoeny (1999) 25.427.9 mma (11.10 1/4801/330a
in.)a If the pair of points with the largest elevation difference (z)
AROC (1997) workmanship Not given 1/580
standards
were chosen, (FL)min would therefore result from (2). ACI 117-
ACI 117-90 (1990) toler- FL > 13 1/96 90 provides tolerances for FL and (FL)min of 13 and 10, re-
ances for concrete con- spectively, for the type of finishing common in residential con-
struction and materials
struction. Using the local value, the standard thereby sets a
a
Depending on soil conditions. tolerance on zmax between two points 3.05 m apart of 31.75
mm (1.25 in.), corresponding to a slope of 1/96. The index
procedure has the advantage of being standardized, but in the
arise largely from measurements made on structures that had writers experience it is more common in residential practice
experienced some distress. to focus on the maximum total differential elevation across the
One inherent aspect of the studies reported in Table 2 is that slab.
any type of observation that measures the slope or elevation
change of the slab and reports it as postconstruction movement SUMMARY OF FIELD STUDY
is based on the assumption that the floor was level when
placed. In the study by Marsh and Thoeny (1999), levels of A water level manometer device (Sarek Company model
angular distortion and differential movement of the slab were 465C) was used to measure elevation differences in 98 con-
measured using a combination of manometer measurements crete slabs at many locations across the Phoenix metropolitan
and information derived from the damage observations (ob- area. The measurements cover a number of builders and con-
servations of the types, extent, and severity of cracking) in the crete subcontractors and were made at several different times
slabs. Although an educated guess can be made as to the char- of year. The concrete subcontractors did not know the slab
acteristics and dynamics of these slabs given the loading, mea- measurements were going to be made, although in some cases
sures of levelness, and damage profile, much of the informa- the builders superintendent was aware of the upcoming mea-
tion given in Table 2 is indirect and not truly derived from surements.
pure observation and analysis. The water level manometer consists of a large reservoir of
Marsh and Thoeny (1999) acknowledged that information water on an adjustable stand, connected by a plastic tube to a
about newly placed concrete slabs must be obtained before graduated rod. The water level remains at the elevation in the
truly accurate studies of this type can evolve: During inves- reservoir, which is transmitted by hydraulic continuity through
tigations of distressed homes it is often necessary to assess the tube and read on the graduated rod. The water level at the
how much differential movement has occurred as a result of rod can be read through the tube against the graduated scale
on the rod. The elevation difference between two points can
soil influence. To do this, one must know some details of how
then be found by comparing the readings on the graduated rod
the slab was first constructed, such as how level it was ini-
when it is held vertically at each point. The water level ma-
tially (Marsh and Thoeny 1999). The Koenig (1991) data
nometer is shown schematically in Fig. 1.
represent one source for new slab levelness and point to the
The manometer was used to obtain readings on concrete
possibility that new slabs may be out of level on the order of
slab-on-grade floors. Readings were taken on a 3.05 m (10 ft)
12.7 mm (0.5 in.) maximum differential, on average.
spacing. The slab was marked using a snap line around the
Maximum elevation difference across the slab (that is, the
grid, beginning at a point 0.3 m (1 ft) inside an outer corner.
elevation difference between the highest point and the lowest
An example layout is shown in Fig. 2. No measurements were
point) is a fairly commonly used indicator of distress in resi-
made in the outer 0.3 m of the slab to avoid irregularities
dential slabs. This is the measure used by Koenig (1991) and
arising from the screeding process itself. In addition to the
frequently found in the relevant literature. The angular distor- points of intersection on the grid, additional points were mea-
tions reported in Table 2 are slightly different, representing the sured near utility penetrations through the slab to identify po-
slope between adjacent measurement points on the slab. In tential problems in these areas that are difficult to screed and
other words, they represent the elevation difference between finish.
neighboring measurements, divided by the horizontal distance The 3.05 m grid was selected based on timely completion,
between the measurements. convenience, and compatibility with the typical number of
A different approach to slab geometric quality is presented points for forensic investigations in the writers experience. In
in the ASTM (1996) E1155-96 standards and the American practice, measurement patterns on serving homes are usually
Concrete Institute (ACI 1990) 117-90 specifications. Here, in- heavily controlled by the arrangements of interior partitions
dex values are used to assess the flatness and levelness of a and furniture, and great variability exists in the number of
newly placed slab. The indices are known as FF and FL, re- points commonly measured by different investigators. The
spectively, and are calculated as dimensionless ratios involving writers are currently studying the impact of measurement spac-
the mean and standard deviation of elevation measurements ing on the accuracy and reproducibility of the results.
on the slab surface made along 3.05 m (10 ft) lines following A typical measurement would proceed as follows. The ma-
the layout guidelines in ASTM E1155-96. The most appropri- nometer would be set up near one corner of the slab surface.
ate index to the current discussion is FL, defined as A 3.05 3.05 m grid was laid out across the slab, starting
from this corner. Each grid intersection point within the slab
317.5 mm 12.5 in. was taken as a point of measure. A rough sketch of each slab
FL = = (1)
3sz z 3sz z and measurement grid was drawn in the field. Measurements
JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / NOVEMBER 2001 / 129
 FIG. 1. Water Level Manometer Measurement System

FIG. 3. Typical Condition of Interior Floor Space Prior to Final Grad-

ing

Access to organize and group the information gathered in these

analyses. Subsequent data reduction included identification of
the lowest point on the slab and calculation of the elevation
difference between this point and the other measured points
on the slab.
The sites included 89 slabs measured within four days of
concrete placement, and nine more measured after structural
loading using roughly the same technique. A loaded slab in
this context refers to a slab over which a house has been
framed, enclosed, and loaded with roofing tiles for at least one
week before water level measurement commenced, but prior
FIG. 2. Example Showing Measurement Grid System to installation of cosmetic flooring materials. These basic el-
ements of single-family home structures represent the majority
of dead loads that will be transferred to that slab. Because of
of elevation difference for each point were recorded on the constraints associated with built-in units, interior walls, and
sketch. The subdivision name and homebuilder were noted for other fixtures, using a regular grid was not practical for the
each slab measured. The address of each was noted when slabs with structure in place. Instead, an average representative
available. The concrete subcontractor was also documented number of measurements were taken at points within the slab.
when that information was available. Concrete slab measurable These points were chosen so that their spacing from one an-
square footage was noted and ranged from 110 to 465 m2 other, position relative to exterior and interior load-bearing
(1200 to 5000 ft2) in this sample. Weather conditions at the walls, and position relative to the central and outer areas of
time of measurement, including temperature and wind condi- the slab itself were representative of the number and spacing
tions, were noted. A database was developed using Microsoft of the unloaded slab measurements using the 3.05 m grid.
130 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / NOVEMBER 2001
 The slabs in this study were all generally representative of ually screeded using a lumber straightedge, to the level of the
the typical concrete slab-on-grade (not posttensioned) methods top of the stem wall. The screeding process is conducted using
in use in the Phoenix metropolitan area, which are similar to the top of the stem wall as a guide around the edges of the
methods in use in at least the desert southwestern United floor slab, and then using the screeded zones near the edge in
States, as well as many other portions of the country. The the middle portions (Fig. 4). Obviously, using this method re-
footings and stem walls are formed and placed first, with the quires considerable skill to obtain a plane surface, and very
top of the stem wall being set to the grade of the finished floor flat conditions are not extremely likely. Interior joints are usu-
elevation. The footing trench spoil is usually placed on the ally installed after concrete placement by saw cutting, but pre-
floor side of the stem wall before placement of the footing, fabricated joints may also be used.
and to this loose soil is added the spoil from any subfloor
utility trenches that must be installed (Fig. 3). After inspection ANALYSIS AND RESULTS
of the subfloor utilities, the concrete subcontractor is charged
with shading and backfilling the interior trenches, removal of Each point of measurement on slabs tested for this study
any excess spoil, and placement of an interior base course yielded a difference (either positive or negative) between the
layer that is commonly about 102 mm (4 in.) thick. In many elevation of the reference point and the point of measurement.
cases, the trench backfilling and placement of the base course To compute the maximum elevation difference noted for any
are completed without systematic attempts to compact or mois- given slab, the maximum (most positive) and minimum (most
ture-control the materials. negative) differences between measurement points and the ref-
The concrete contractor arranges for application of termiti- erence point were obtained for each slab. The maximum ele-
cide the evening before the slab is to be placed. The fresh vation difference for the slab is then obtained using (3).
concrete is then placed inside the stem walls (often integrally
max = max(ref) min(ref) (3)
with any interior bearing footings). The concrete is then man-
where max = maximum elevation difference across slab; and
ref = difference between measurement point and reference
point.
This process was conducted for every slab, and the maxi-
mum elevation differences are presented as a histogram in Fig.
5. The vertical axis represents the relative frequency of oc-
currence of measured values of maximum slab elevation dif-
ference in a bin centered on the value shown on the x-axis.
Relative frequency is the number of samples in each bin di-
vided by the total number of samples (89), expressed as a
percentage. Only a small fraction of the slabs in the sample
were measured to have a maximum deviation from level of
over 25 mm (1 in.). Also shown on the histogram, for com-
parison, is a normal distribution with the same average and
standard deviation, which can be seen to be a reasonable fit
to the measured results.
FIG. 4. Typical Residential Slab-on-Grade Screeding Operation The maximum angular distortion for each slab was calcu-

FIG. 5. Newly Placed Residential Slab Elevation Difference (n = 89)

JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / NOVEMBER 2001 / 131

lated using a similar procedure, but in this case comparing the TABLE 4. Confidence Intervals for Mean of Population
elevation difference between adjacent points instead of the ex- Confidence Confidence
trema. The elevation change is then divided by the distance level Critical z interval for
between the two points to arrive at the slope of the slab be- (percent) [z(/2)] mean
tween the two points. This calculation is completed using (4).
90 1.645 [12.80, 14.22] mm
For small angles ([0.504, 0.560] in.)
95 1.96 [12.67, 14.35] mm
rise
tan = (4a) ([0.499, 0.565] in.)
run 99 2.58 [12.40, 14.63] mm
([0.488, 0.576] in.)
refi refj
= (4b)
Lij
draws from a population, it is statistically possible that the
where refi = elevation difference to reference point for ith nearness of the means is only an accidentthe luck of the
measurement; and Lij = plan distance between ith and jth draw, as it were. To evaluate this possibility, the t-testing pro-
points. cedure was used. This procedure allows one to evaluate the
The calculation of the maximum angular distortion required likelihood that two samples are the same. In this case, we will
that all pairs of adjacent points be considered. A routine was be checking the hypothesis that the mean maximum deviation
developed in the Access database to step through the data and from level of the population of slabs immediately after they
calculate the maximum angular distortion for each pair, and are placed is the same as the mean of the population of slabs
then to identify the maximum value for each slab. after the structures are completed; in other words, the hypoth-
Summary statistics for the maximum slope and maximum esis is that the construction of the home does not affect the
elevation difference are presented in Table 3. Note that the deviation from level of the floor slab.
average values for the loaded and unloaded slabs are remark- There are two possible errors that we could make in eval-
ably similar. Further, note that the average maximum slope is uating this hypothesis. The first, known as the error, is that
nearly as high as Day (1990) reported for cosmetic damage we reject the hypothesis when it is true. In this case, that
and that the range includes values nearly as steep as Day de- would mean that we conclude that the sample shows that the
scribes for a threshold before structural damage will begin construction of the home does affect the deviation from level
(Table 2). Only the maximum elevation difference is presented of the floor slab, when in fact it does not. The second or
for the slabs measured after completion of the residential struc- error is to accept the hypothesis when it is false; this error
ture (loaded slabs). The distances between the points where would mean that we conclude that construction of the home
elevations were calculated for these slabs were not accurately does not affect the slab when in fact it does. The sample sta-
measured. tistics in Table 3 allow us to evaluate the confidence we have
The results summarized in Table 3 indicate that, on average, about the conclusions that might result from comparing the
the slabs in the sample showed about 13.5 mm (0.53 in.) max- means. In essence, the goal is to establish a probability that
imum deviation from level. These data can be used to estimate we are right about the means being identical, or a probability
the mean of the actual population. If the mean of the popu- that in fact the home construction does not affect the levelness
lation is given by , then the measured results for the sample of the slab.
of newly placed slabs allow us to estimate the maximum de- The hypothesis is written in statistical terms as
viation from level of the population, along with the confidence
H0:L U = 0 (6)
interval for the estimate (Table 4). This estimate can be made
using the statistical function for estimating confidence inter- where L = mean of population of loaded slabs; and U =
vals for a large sample (n > 30), using (5): mean of population of unloaded slabs.
The hypothesis H0 will be rejected if the calculated t-value
s
= x z(/2) (5) for the samples is less than the critical t-value for the number
n of degrees of freedom and the desired confidence level. The
critical t-value can be obtained from a t-distribution, which is
where x = mean value of sample; z(/2) = critical z value for tabulated in most elementary statistics books (we used Devore
confidence level desired (Devore 1991); s = sample standard 1991). The calculated t-value assuming the hypothesis H0
deviation; and n = number of observations in sample. From comes from (7).
Table 4, it can be seen that, to very high degrees of confidence,
the average deviation from level of concrete slabs is about 13.5 xU xL

mm (0.53 in.) when the slabs are newly placed. t= (7)
1 1
The smaller sample of the slabs measured after the struc- sp
tures were completed above them shows an average value very nU nL
similar to that of the sample of newly placed slabs. One is
tempted to conclude that the results show that the averages are where xU = mean of sample of unloaded slabs; xL = mean of
the same. However, because the samples represent random sample of loaded slabs; n = number of observations in each
sample; sp = (nU 1)s U2 (nL 1)s L2 /(nU nL 2); sU =
TABLE 3. Summary Statistics for Slab Level Measurements in Study sample standard deviation for unloaded slabs; and sL = sample
Maximum Elevation Difference across Slab
standard deviation for loaded slabs.
For the 95% confidence limit, and for the appropriate num-
Sample Standard
ber of degrees of freedom (89 9 2 = 96), the critical t-
Type size Range Mean deviation
value is 1.989. The calculated t-value using the values in Table
New placed slabs
max 89 6.3530.0 mm 13.5 mm 4.06 mm
3 and the equations above is 0.0318. As the calculated t-value
(0.251.18 in.) (0.532 in.) (0.160 in.) is well below the critical t-value, there is no statistical reason
max 89 1/8571/101 1/334 1/796 to reject H0 to the 95% confidence level. In other words, the
Loaded slabs max 9 11.717.0 mm 13.5 mm 1.70 mm likelihood that we ought to conclude that the means are alike
(0.460.67 in.) (0.530 in.) (0.067 in.)
is very high.
132 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / NOVEMBER 2001
 The error is more difficult to evaluate. It requires an as- yields = 0.04. The hypothesis H0 would therefore be rejected
sumption about the standard deviation of the population, incorrectly only 4% of the time. Taken together, these statistics
which is in truth unknown to us. However, it is likely that it indicate that there is a high probability that the means of the
is close to the value for the unloaded slab sample, since that two populationsthat is, loaded and unloaded slabsare the
sample is so large. If we assume that the standard deviation same. In other words, it appears likely that loading the slab
of the population () is about 3.8 mm (0.15 in.), we can find by construction of the house, including the interior partitions
out the probability that the means of the unloaded and loaded and the roofing, does not affect the maximum deviation from
slabs are more than a set amount different, which amount will level.
be called . This probability is known as and is calculated Another way to look at this sample is to calculate a confi-
using standard charts and the d-value, which is calculated us- dence interval for the difference between the means of the two
ing (8). samples. This is done using (9).

d=


nUnL
(nU nL)(nU nL 1)
(8) I = xU xL (t/2,nUnL2)sp
1
nU

1
nL
(9)

Using this equation and the 95% confidence level, we arrive

Obviously, the smaller the potential difference in the means at the conclusion that, to the 95% confidence level, the means
, the larger the probability that the means are different. are no more than 2.8 mm (0.11 in.) apart. Again, this leads us
Therefore, one would ordinarily choose a such that a dif- to the conclusion that the construction of the homes has no
ference of that much would have engineering significance. In statistically significant effect on the degree to which the slabs
this case, we chose to evaluate the likelihood that the true are level.
means are within = 5 mm (0.2 in.) of each other. This leads Similar analyses can be used to check other kinds of ques-
to a d-value of about 0.4 and at the 95% confidence level tions one might ask about the results of this study. Two im-

TABLE 5. Summary Statistics for Maximum Elevation Difference in Comparison Studies

Sample Standard
Type size Range Mean deviation
Newly placed slabs
All data 89 6.430.0 mm 13.5 mm 4.06 mm
(0.251.18 in.) (0.532 in.) (0.160 in.)
Hot months 49 6.430.0 mm 13.4 mm 4.50 mm
(0.251.18 in.) (0.527 in.) (0.177 in.)
Cool months 40 8.926.9 mm 13.7 mm 3.61 mm
(0.351.06 in.) (0.538 in.) (0.142 in.)
Loaded slabs (max) 9 11.717.0 mm 13.5 mm 1.70 mm
(0.460.67 in.) (0.530 in.) (0.067 in.)
Koenig (1991), Texas 54 3.225.4 mm 13.7 mm 5.33 mm
(0.1251.00 in.) (0.54 in.) (0.21 in.)
Marsh and Thoeny (1999), California 6 15.225.4 mm 19.1 mm
(0.61.0 in.) (0.75 in.)
Bondy (2000), California 4 12.722.9 mm 1.78 mm
(0.50.9 in.) (0.7 in.)

FIG. 6. Normal Probability Distribution for Maximum Elevation Difference on Newly Placed Residential Concrete Slab-on-Grade Floors

JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / NOVEMBER 2001 / 133

portant questions include the degree to which these data match slabs on which no structure had yet been built, these values
results presented in the literature, and the degree to which the do not indicate any cosmetic, functional, or structural danger
results vary by season. The maximum elevation difference val- to the structure. Only those elevation changes that occur after
ues are summarized in Table 5. The hot month data are a completion of the structure are in fact likely to produce dam-
summary of the results obtained in June, July, or August, and age to the supported residence.
are separated out to determine if curing in the heat of summer These results point out the potential inaccuracies that may
may introduce a larger maximum deviation from level, perhaps arise in forensic engineering studies that rely heavily on the
by virtue of slab curling. maximum elevation difference or the maximum angular rota-
Using the same analysis procedure outlined above, it can be tion of the residential floor slab. If the forensic engineer uses
shown that the data from the hot and cool months do not have such measurements to make conclusions about the likelihood
statistically significant differences from each other, the overall of any damage to the structure, they must consider the possi-
result, or the loaded slabs. Furthermore, these results are not bility (indeed the likelihood) that the floor slab was not level
statistically different from the Koenig (1991) data obtained for at the time of placement. Furthermore, the results of this study
newly constructed slabs in Texas. The Marsh and Thoeny can be used to address the question of postconstruction move-
(1999) and Boone (2000) results did not include the standard ments probabilistically. Fig. 6 shows the probability, based on
deviation and constitute only a very small sample. As a result, a normal distribution having the same mean and standard de-
it is impossible to make quantitative conclusions about the viation observed in this study, that any given value of maxi-
statistical significance of the difference between these data and mum elevation difference might have existed at the time of
the other rows. However, the range of results obtained by placement of the floor slab. Clearly, these results can be taken
Marsh and Thoeny are generally similar to those obtained by as supportive of the general approach recommended by Boone
Koenig and in the present study. (1996), in which the elevation difference and slope measure-
ments from a residential floor slab of unknown initial flatness
CONCLUSIONS are taken in the context of observed distresses to the structure.
This paper has presented a summary of an investigation of
residential floor flatness for newly placed floors in metropol- ACKNOWLEDGMENT
itan Phoenix, Arizona. The study included a review of the This work was supported in part by a grant from the Arizona Con-
relevant literature and a field testing program to measure the sulting Engineers Association.
flatness of 89 newly placed floors and 9 recently loaded floors.
Two studies reported average maximum deviations from REFERENCES
level of 12.7 to 19 mm (0.5 to 0.75 in.) More literature was American Concrete Institute (ACI) Committee 317. (1990). Standard
found with measurements of homes that have experienced dis- specifications for tolerances for concrete construction and materials
tress, without information about the levelness of these floors (ACI 117-90) and Commentary (ACI 117R-90), Detroit.
at the time of concrete placement. The literature indicates that Arizona Registrar of Contractors (AROC). (1997). A guide for contrac-
some level of distress may be expected when the maximum tors: Minimum workmanship for licensed contractors, Phoenix, Ariz.
deviation from level across a serving residential slab-on-grade ASTM. (1996). Annual book of standards, American Society for Testing
and Materials, Philadelphia, Pa.
floor exceeds about 28 to 38 mm (1.1 to 1.5 in.). Furthermore, Bondy, K. B. (2000). Performance evaluation of residential concrete
the literature indicates that damage may begin to become ap- foundations. Tech. Notes, 9, 16.
parent when floor slab slopes exceed 1/300 and may become Boone, S. J. (1996). Ground-movement-related building damage. J.
severe when slopes exceed 1/100 (Day 1990; Marsh and Geotech. Engrg., ASCE, 122(11), 886896.
Thoeny 1999). Day, R. W. (1990). Differential movement of slab-on-grade structures.
The data collected from concrete slab-on-grade floors im- J. Perf. Constr. Fac., ASCE, 4(4), 236241.
Devore, J. L. (1991). Probability and statistics for engineering and the
mediately after placement in this study resulted in an average sciences, 3rd Ed., Brooks/Cole Publishing Company, Pacific Grove,
maximum deviation from level at the time of placement of Calif.
13.5 mm (0.53 in.), with a maximum observed value of 30.0 Koenig, A. S. (1991). Relative elevation data analysis on newly con-
mm (1.18 in.). The average maximum slope observed on floors structed residential foundations. Paper Presented to the ASCE Texas
in this sample was 1/334, with a maximum observed value of Section Spring Meeting, San Antonio, Texas, Available through the
1/101. The extrema from both the maximum elevation differ- Expansive Clay Research Institute, San Antonio, Tex.
Marsh, E. T., and Thoeny, S. A. (1999). Damage and distortion criteria
ence and the maximum angular rotation exceed the threshold for residential slab-on-grade structures. J. Perf. Constr. Fac., ASCE,
values found in the literature indicated in the previous para- 13(3), 121127.
graph, but compare favorably with the tolerances set forth in Skempton, A. W., and MacDonald, D. H. (1956). The allowable settle-
ACI 117-90. Obviously, since these extreme values came from ments of buildings. Proc., Instn. Civ. Engrs., 5(3), 727768.

134 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / NOVEMBER 2001