ANOTHER LOOK AT HARTFORD CIVIC CENTER COLISEUM COLLAPSE

By Rachel Martin1 and Norbert J. Delatte,2 Member, ASCE

ABSTRACT: Only a few hours after five thousand basketball fans had left, the roof of the Hartford Civic Center
Coliseum collapsed under a heavy snowfall. Fortunately, the arena was empty. The design of the space frame
roof had been based on an innovative and extensive computer analysis. However, when deflections twice as
great as those predicted by the computer analysis were observed during construction, the warning was ignored.
Overconfidence in computer analysis results played a large part in this failure. A useful lesson from this case is
that the computer is only an analytical tool and computed results must be checked by the designer with a careful
eye. The long, unbraced lengths of compression members made them highly susceptible to buckling. This case
serves as a lesson for engineering students and practicing engineers concerning the difficult technical, profes-
sional, procedural, and ethical issues that may arise during the design and construction of a complex, high-
Downloaded from ascelibrary.org by University of Leeds on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

occupancy structure.

INTRODUCTION concerns, another neglected topic in engineering education, in

real-life situations, as well as serving as a constant reminder
No one was killed or injured when the huge space truss roof of the repercussions of careless engineering (Delatte 1997).
of the empty Hartford Civil Center Coliseum collapsed under According to a 1987 survey conducted by the Education
a heavy snowfall at 4:19 a.m. on January 18, 1978 (Fig. 1). Committee of the Technical Council on Forensic Engineering
Had the failure occurred just a few hours before, however, the of the American Society of Civil Engineers, 63.2% of schools
death toll might have been hundreds, or even thousands. The indicated that they would consider teaching a course on failure
dramatic roof, designed with the aid of computers, had shown case studies if the appropriate materials were available. This
evidence of distress during construction, but the warnings had clearly demonstrates the need for case study material and
not been heeded. The building had been in service for five teaching aids to encourage the incorporation of failure case
years when it collapsed (Levy and Salvadori 1992). studies into the engineering curriculum (Rendon-Herrero
For the engineer and engineering student, knowledge of en- 1993). The objectives of this paper are to:
gineering’s failures is just as important as knowledge of its
successes. A success illustrates what engineering can make 1. Summarize what is known about the design, construc-
possible, while a failure demonstrates its limits. It takes nu- tion, and collapse of the Hartford Civil Center Coliseum.
merous successful structures to ensure the quality of a design 2. Examine the causes of the failure as well as the legal
or a construction method. One failure, however, can discredit ramifications.
an entire design or building technique. Because of this, the 3. Explore the technical, procedural, and ethical concerns
information that each failure has to offer should be carefully present, focusing on how the failure could have been
studied and applied to all future designs. As a result, similar avoided and how to prevent similar failures in the future.
failures, as well as their tragic consequences, can be avoided.
Because of their importance, failures should be incorporated This failure case study can be integrated into engineering
into engineering education. Unfortunately, undergraduate en- classes to introduce new topics or as the topic of a student
gineering students receive little exposure to engineering fail- research paper assignment.
ures in college. This approach to engineering education not
only leaves students less prepared for what they will face after
DESIGN AND CONSTRUCTION
college, but it also fails to show the importance of continuing
education (Delatte 1997). This may be one of the reasons that In 1970, Vincent Kling agreed to be the architect for the
a 1983 survey of ASCE section and branch presidents found Hartford Civic Center. Shortly thereafter he hired Fraoli,
that engineering failures are all too common (Bosela 1993). Blum, and Yesselman, Engineers (FB&Y), to design the arena.
Since undergraduate engineering students already face an In order to save money, FB&Y proposed an innovative design
overcrowded curriculum, rather than requiring a new class for the 91.4 ⫻ 110 m (300 ⫻ 360 ft) space frame roof 25.3
covering failure case studies, these case studies can be incor-
porated into existing classes throughout a student’s college ca-
reer. Not only will this approach capture the students’ interest
by showing how their classes relate to engineering, but it will
also inspire them to learn more about the history of the pro-
fession. In addition, it teaches them the importance of contin-
ued learning throughout their professional career. Finally, fail-
ure case studies provide a perfect opportunity to discuss ethical
1
Student, Washington Univ. in St. Louis, Civ. Engrg. Dept., Campus
Box 1130, One Brookings Dr., St. Louis, MO 63130.
2
Asst. Prof., Dept. of Civ. and Envir. Engrg., 1075 13th St. South, Ste.
120, The Univ. of Alabama at Birmingham, Birmingham, AL 35294-
4440.
Note. Discussion open until July 1, 2001. To extend the closing date
one month, a written request must be filed with the ASCE Manager of
Journals. The manuscript for this paper was submitted for review and
possible publication on January 31, 2000. This paper is part of the Jour-
nal of Performance of Constructed Facilities, Vol. 15, No. 1, February, FIG. 1. Hartford Civic Center Coliseum Roof Collapse, 1978
2001. 䉷ASCE, ISSN 0887-3828/01/0001-0031–0036/$8.00 ⫹ $.50 per (Construction Failure, Feld and Carper, 䊚 1997. Reprinted by
page. Paper No. 21952. Permission of John Wiley & Sons, Inc.)

JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / FEBRUARY 2001 / 31

J. Perform. Constr. Facil. 2001.15:31-36.

 m (83 ft) over the arena. The proposed roof consisted of two edges of the roof supported it instead of boundary col-
main layers arranged in 9.14 ⫻ 9.14 m (30 by 30 ft) grids umns or walls (Levy and Salvadori 1992).
composed of horizontal steel bars 6.4 m (21 ft) apart. Diagonal 5. The space frame was not cambered. Computer analysis
members 9.14 m (30 ft) long connected the nodes of the upper predicted a downward deflection of 330 mm (13 in.) at
and lower layers and, in turn, were braced by an intermediate the midpoint of the roof and an upward deflection of 150
layer of horizontal members. The 9.14 m (30 ft) members in mm (6 in.) at the corners (‘‘Space’’ 1978).
the top layer were also braced at their midpoint by interme-
diate diagonal members (Figs. 2 and 3). Because of these money-saving innovations, the engineers em-
This design departed from standard space frame roof design ployed state-of-the-art computer analysis to verify the safety
procedures in five ways: of the building.
A year later construction began. To save time and money,
1. The cross-section configuration of the four steel angles the roof frame was completely assembled on the ground.
making up each truss member did not provide good re- While it was still on the ground the inspection agency notified
sistance to buckling. The cross-shaped built-up section the engineers that it had found excessive deflections at some
had a much smaller radius of gyration than either an of the nodes. Nothing was done.
Downloaded from ascelibrary.org by University of Leeds on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

I-section or a tube section configuration of the same After the frame was completed, hydraulic jacks located on
structural members (Fig. 4). As a result, the buckling top of the four pylons slowly lifted it into position. Once the
load for the cross-shaped section was much lower than frame was in its final position but before the roof deck was
that of the other configurations. installed, its deflection was measured and found to be twice
2. The top horizontal members intersected at a different that predicted by computer analysis, and the engineers were
point than the diagonal members rather than at the same notified. They, however, expressed no concern and responded
point, making the roof especially susceptible to buckling that such discrepancies between the actual and the theoretical
because the diagonal members did not brace the top should be expected (Levy and Salvadori 1992).
members against buckling. When the subcontractor began fitting the steel frame sup-
3. The top layer of this roof did not support the roofing ports for fascia panels on the outside of the truss, he ran into
panels; short posts on the nodes of the top layer did. Not great difficulties due to the excessive deflections of the frame.
only were these posts meant to eliminate bending stresses Upon notification of this problem, the project manager ‘‘di-
on the top layer bars, but their varied heights also al- rected the subcontractor to deal with the problem or be re-
lowed water to be carried away to drains. sponsible for delays.’’ As a result, the subcontractor coped
4. Four pylon legs positioned 13.7 m (45 ft) inside the some of the supports and refabricated others in order to make
the panels fit, and construction continued (‘‘Design’’ 1978).
The roof was completed on January 16, 1973 (Feld and
Carper 1997). The next year, a citizen expressed concern to
the engineers regarding the large downward deflection he no-
ticed in the arena roof, which he believed to be unsafe. The
engineers and the contractor once again assured the city that
FIG. 2. Elevation of Space Frame Roof (Circled Section Is everything was fine (Levy and Salvadori 1992).
Shown Enlarged in Fig. 3)
COLLAPSE
On January 18, 1978, the Hartford Arena experienced the
largest snowstorm of its five-year life. At 4:19 a.m., the center
of the arena’s roof plummeted 25.3 m (83 ft) to the floor of
the arena, throwing the corners into the air. Just hours earlier
the arena had been packed. Luckily, it was empty by the time
of the collapse (Ross 1984).

CAUSES OF FAILURE
Hartford appointed a three-member panel to manage the in-
vestigation of the collaspe. This panel in turn hired Lev Zetlin
Associates, Inc. (LZA), to ascertain the cause of the collapse
and to propose a demolition procedure (Ross 1984). LZA is-
sued its report later that year (Lev Zetlin Associates 1978).
LZA discovered that the roof began failing as soon as it was
completed due to design deficiencies. A photograph taken dur-
ing construction showed obvious bowing in two of the mem-
bers in the top layer.
Three major design errors, coupled with underestimation of
FIG. 3. Section of Space Frame Roof (Figure Courtesy of LZA the dead load by 20% [estimated frame weight = 0.862 Pa (18
Technology, from Lev Zetlin Associates, 1978, Reprinted by Per- psf); actual frame weight = 1.10 Pa (23 psf)], allowed the
mission) weight of the accumulated snow to collapse the roof (‘‘De-
sign’’ 1978). The load on the day of collapse was 3.16–3.50
Pa (66–73 psf), while the arena should have had a design
capacity of at least 6.70 Pa (140 psf) (‘‘Collapsed’’ 1978b).
The three design errors responsible for the collapse are listed
below:

• The top layer’s exterior compression members on the east

FIG. 4. Compression Member Configurations and the west faces were overloaded by 852%.
32 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / FEBRUARY 2001

J. Perform. Constr. Facil. 2001.15:31-36.

 • The top layer’s exterior compression members on the
north and the south faces were overloaded by 213%.
• The top layer’s interior compression members in the east-
west direction were overloaded by 72%.

In addition to these errors in the original design, LZA dis-

covered that no midpoint braces were provided for the mem-
bers in the top layer. The exterior members were only braced
every 9.14 m (30 ft), rather than the 4.57 m (15 ft) intervals
specified, and the interior members were only partially and
insufficiently braced at their midpoints. The two members at-
tached to the midpoint of the top chord were both in the same
plane as the long axis of the chord, so they only provided
bracing in one direction. The perpendicular direction was ef-
fectively unbraced for the full 9.14 m (30 ft) length. This sig-
Downloaded from ascelibrary.org by University of Leeds on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

nificantly reduced the load that the roof could safely carry. In
addition, certain perimeter top chord members with a post
landing at midpoint were subjected to bending stress from the
roof load applied through the post. Since the members were
not designed for bending, this led to a considerable overstress
(Lev Zetlin Associates 1978).
Fig. 5 and Table 1 compare some of original details to actual
designs used in the building, demonstrating the reduction in
strength that these changes caused. Connection A was typically
used on the east-west edges of the roof, while connection B
was used on the north-south edges. Most of the interior mem-
bers used connection C, while a few used connection D. The
key difference between the original and the as-built details
may be seen in Fig. 5 and also by comparing the top and
FIG. 5. Comparison of Actual and Assumed Bracing (Figure
bottom rows of the table. The diagonal members were attached Courtesy of LZA Technology, from Lev Zetlin Associates, 1978,
some distance below the horizontal members. Thus, the flex- Reprinted by Permission): (a) Original Design Assumption; (b)
ibility of the connection reduced the effectiveness of the brac- Actual Design Condition
ing by introducing a ‘‘spring brace’’ instead of the hard brace
that had been assumed.
buckling is uncommon, it is often an overlooked mode of fail-
The most overstressed members in the top layer buckled
ure (‘‘New’’ 1979).
under the added weight of the snow, causing the other mem-
Hannskarl Bandel, a structural consultant, completed an in-
bers to buckle. This changed the forces acting on the lower
dependent investigation of the collapse for the architect’s in-
layer from tension to compression, causing them to buckle also
surance company. He blamed the collapse on a faulty weld
in a progressive failure. Two major folds formed initiating the
connecting the scoreboard to the roof. This opinion conflicts
collapse (‘‘Design’’ 1978). These were not the only errors that
with the opinions of all the other investigators (‘‘Hartford’’
LZA discovered. Listed below are the other factors that con-
1979). The LZA report’s findings were also disputed by FB&Y
tributed to, but probably were not solely responsible for, the
(‘‘Collapsed’’ 1978a).
collapse:
LEGAL REPURCUSSIONS
• The slenderness ratio of the built-up members violated the
American Institute of Steel Construction (AISC) code Six years after the collapse, all of the parties reached an
provisions. The spacer plates separating the individual an- out-of-court settlement. While this was beneficial to the parties
gles were placed too far apart in some of the four-angle involved, it did not provide the engineering profession with
members, allowing individual angles to buckle. the precedents that such a case could set (Feld and Carper
• The members with bolt holes exceeding 85% of the total 1997).
area violated the AISC code requirements for section re-
duction of tension members (‘‘Collapsed’’ 1978b).
TECHNICAL ASPECTS
• There were misplaced diagonal members (Feld and Carper
1997). The engineers for the Hartford Arena depended on computer
analysis to assess the safety of their design. Computer pro-
Loomis and Loomis, Inc., also investigated the Hartford col- grams, however, are only as good as their programmer and
lapse. They agreed with LZA that gross design errors were may tend to offer engineers a false sense of security (Shepherd
responsible for the progressive collapse of the roof, beginning and Frost 1995). The LZA report noted ‘‘the computer model
the day that it was completed. They, however, believed that used by the structural engineer only included the top and bot-
the torsional buckling of the compression members, rather than tom chords and the main diagonals. Roof loads were only
the lateral buckling of top chords, initiated the collapse. applied at top chord main panel points. If the computer model
Using computer analysis, Loomis and Loomis found that had represented the intermediate diagonals and horizontals and
the top truss members and the compression diagonal members had included the roof loads at the midpoint, subpanel points
near the four support pylons were approaching their torsional at the top chord, the instabilities and primary bending mo-
buckling capacity the day before the collapse. An estimated ments would have been detected by the designer’’ (Lev Zetlin
0.575–0.718 Pa (12–15 psf) of live load would cause the roof Associates 1978).
to fail. The snow from the night before the collapse comprised Instead of the cruciform shape of the rods, a tube or I-bar
a live load of 0.670–0.910 Pa (14–19 psf). Because torsional configuration would have been more stable and less suscepti-
JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / FEBRUARY 2001 / 33

J. Perform. Constr. Facil. 2001.15:31-36.

 TABLE 1. Comparison of Original Design and Actual As-Built Connections
Downloaded from ascelibrary.org by University of Leeds on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

ble to bending and twisting. The cruciform shape has the ad- only responsible for ensuring that the design was constructed
vantage of making the members easier to connect. Also, if the correctly and not the performance of the project (p. 202, Feld
horizontal and diagonal members intersected at the same place, and Carper 1997).
the bracing would have increased the buckling capacity in It is important for responsibility for the integrity of the en-
these members. The LZA report noted that ‘‘apparently, the tire project to rest with one person. Feld and Carper (pp. 202–
choice of the typical member as a cruciform, a section that is 204, 1997) offer an excellent discussion of the role that pro-
weak in bending and torsion, was based on the design as- cedural deficiencies played in this collapse.
sumption that such bending and torsion would be negligible As a result of the construction manager’s refusal to hire a
in the space truss’’ (Lev Zetlin Associates 1978). structural engineer for the purpose of inspection, no one re-
The LZA report further noted that ‘‘the investigation con- alized the structural implications of the bowing members. This
firms that space trusses and/or space frames are valid and safe collapse illustrates the importance of having a structural en-
structural systems. Two-way space trusses have been em- gineer, especially the designer, perform the field inspection.
ployed successfully on many projects. In the case of the Hart- The designer understands the structure that is being built and
ford Coliseum, unfortunately, certain aspects of the design and would best be able to recognize the warning signs of poor
construction were not implemented correctly’’ (Lev Zetlin As- structural performance and rectify them before they grow to
sociates 1978). catastrophic proportions. The LZA report noted ‘‘the inspec-
tion and/or quality control procedures utilized . . . were inad-
PROFESSIONAL AND PROCEDURAL ASPECTS equate and poorly handled. The absence of a full-time regis-
tered structural engineer experienced with the design and
The Hartford Arena contract was divided into five subcon- construction of long-span special structures was a serious mis-
tracts coordinated by a construction manager. Not only did this take. The visually apparent distortion or bowing of exterior
fragmentation allow mistakes to slip through the cracks, it also top chord members should have been a red flag to one of the
left confusion over who was responsible for the project as a inspecting parties that there was something seriously wrong
whole. Even though the architect recommended that a qualified with the Hartford Coliseum space truss structure’’ (Lev Zetlin
structural engineer be hired to oversee the construction, the Associates, 1978).
construction manager refused, saying that it was a waste of Finally, the Hartford department of licenses and inspection
money and that he would inspect the project himself. After the did not require the project peer review that it usually required
collapse he disclaimed all responsibility on the grounds that a for projects of this magnitude. If a second opinion had been
design error had caused the collapse. He asserted that he was obtained, the design deficiencies responsible for the arena’s
34 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / FEBRUARY 2001

J. Perform. Constr. Facil. 2001.15:31-36.

 collapse probably would have been discovered (Lev Zetlin As- how the theoretical ideas taught in their classes are actually
sociates 1978). Peer reviews are an essential safety measure applied by engineers. The only real obstacle that lies in the
for high-occupancy buildings and structures experimenting way of increased failure awareness at the undergraduate level
with new design techniques (Feld and Carper 1997). Today, is the absence of adequate resources, such as well-developed
Connecticut is one of the few states that requires peer review failure case studies and appropriate illustrations. This paper
of certain buildings. provides professors and students with a failure case study that
can be integrated into undergraduate classes.
ETHICAL ASPECTS How can educators use these aspects of this case? In struc-
tural analysis courses, they can be used to address technical
The excessive deflections apparent during construction were topics such as safety during construction, load paths, stability
brought to the design engineer’s attention several times. The of incomplete structures during construction, and stability of
engineer, confident in his design and the computer analysis structural members. Students may be assigned to research the
that confirmed it, ignored these warnings and did not take the literature in greater depth and support or criticize the available
time to recheck his work. The engineer should pay close at- theories. For engineering students, the legal ramifications of
tention to unexpected deformations and investigate their the case may be of even greater interest. Three additional im-
Downloaded from ascelibrary.org by University of Leeds on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

causes. They often indicate structural deficiencies and should portant points that may be made are the importance of fixing
be investigated and corrected immediately. Unexpected defor- overall responsibility on a project before difficulties are en-
mations provide a clear signal that the structural behavior is countered, the need for inspection during construction, and the
different from that anticipated by the designer. need to read the literature of the profession to keep up with
Kaminetzky (1991) quotes at length from a story in The technical and procedural advances.
Philadelphia Inquirer from May 28, 1978, about this incident, As a class example or homework problem, students may
headlined ‘‘Why The Roof Came Tumbling Down.’’ The story compare the moment of inertia for the cruciform, I, and the
suggests that the ironworkers knew from observing the defor- tube configurations of four angles, as shown in Fig. 3. Angle
mations during construction that the building was a death trap legs ranged from 89 to 203 mm (3 1/2 to 8 in.) long and were
and had vowed never to enter it once it was completed. It also 8 to 22 mm (5/16 to 7/8 in.) thick depending on loads, and
questions why the workers’ warnings were not listened to. the angles were separated by spacers 19 to 22 mm (3/4 to
Also, this collapse raises the important question of whether 7/8 inch) (Lev Zetlin Associates 1978). For numerical exam-
the factor of safety should be increased for buildings with a ples, 127 ⫻ 127 ⫻ 8 mm angles (L 5 ⫻ 5 ⫻ 5/16) may be
high occupancy. Should the impact of a possible failure be used. The torsional stiffness of these configurations may also
taken into account in determining the factor of safety (Kami- be calculated and compared.
netzky 1991)?
CONCLUSIONS
EDUCATIONAL ASPECTS
A useful lesson from this case is that computer software is
Petroski discusses this case in terms of the need for engi- only an analytical tool and that computed results must be
neers to be able to reason out whether or not computer results checked by the designer with a careful eye. Users must un-
make sense, through hand calculations and knowledge of derstand the theoretical foundations of the programs and the
structural behavior and performance. ‘‘Because the computer associated limitations. This case serves as a lesson for engi-
can make so many calculations so quickly, there is a tendency neering students and practicing engineers concerning the dif-
now to use it to design structures in which every part is of ficult technical, professional, procedural, and ethical issues that
minimum weight and strength, thereby producing the most ec- may arise during the design and construction of a complex,
onomical structure. This degree of optimization was not prac- high-occupancy structure. There is no substitute for a thorough
tical to expect when hand calculations were the norm, and knowledge of structural behavior, coupled with a healthy skep-
designers generally settled for an admittedly overdesigned and ticism toward the completeness and accuracy of computer soft-
thus a somewhat extravagant, if probably extra-safe, structure. ware solutions to unusual problems.
However, by making every part as light and as highly stressed
as possible, within applicable building code and factor of ACKNOWLEDGMENTS
safety requirements, there is little room for error—in the com- This research was supported by the National Science Foundation as
puter’s calculations, in the part manufacturers’ products, or in part of the University of Alabama at Birmingham Research Experiences
the construction workers’ execution of the design. Thus, com- for Undergraduates Site in Structural Engineering under grant EEC-
puter-optimized structures may be marginally or least-safe de- 9820484. Thanks are due to David Peraza of LZA Technology for pro-
signs, as the Hartford Civil Center roof proved to be’’ (p. 199, viding a copy of the firm’s 1978 investigation report to the City of Hart-
Petroski 1985). In the decade and a half since Petroski wrote ford. The paper reviewers made many excellent suggestions that were
incorporated into the final version of this manuscript.
these words, despite tremendous advances in computing power
and software, there is no sign that computer programs will
soon be able to envision failure modes that the designer has APPENDIX. REFERENCES
not foreseen, or check their own work. Bosela, P. (1993). ‘‘Failure of engineered facilities: academia responds to
Failure plays an important role in engineering practice. the challenge.’’ J. Perf. Constr. Fac., ASCE, 5.
Through failure analysis, engineers can learn to avoid similar ‘‘Collapsed roof design defended.’’ (1978a). ENR, June 29.
‘‘Collapsed space truss roof had a combination of flaws.’’ (1978b). ENR,
technical errors, allowing them to build stronger, safer struc- June 22.
tures. Since failure analysis plays such an integral role in a Delatte, N. J. (1997). ‘‘Integrating failure case studies and engineering
good engineer’s professional career, it only makes sense that, ethics in fundamental engineering mechanics courses.’’ J. Profl. Issues
in college, engineering students should be taught about fail- in Engrg. Educ. and Pract., ASCE, 123(3), 111–116.
ures, as well as their importance to any engineer’s professional ‘‘Design flaws collapsed steel space frame roof.’’ (1978). ENR, April 6.
life. In light of an already overcrowded undergraduate engi- Education Committee of the Technical Council on Forensic Engineering.
(1995). Failures in civil engineering: structural, foundation, and
neering curriculum, integrating failure case studies into already geoenvironmental case studies, R. Shepherd and D. Frost, eds., ASCE,
existing engineering classes is the most logical solution. New York.
This approach gives students a better idea of the obstacles Feld, J., and Carper, K. (1997). Construction failure, 2nd Ed., Wiley, New
that will face them after college, in addition to demonstrating York.

JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / FEBRUARY 2001 / 35

J. Perform. Constr. Facil. 2001.15:31-36.

 ‘‘Hartford collapse blamed on weld.’’ (1979). ENR, June 24. ‘‘New theory on why Hartford roof fell.’’ (1979). ENR, June 14.
Kaminetzky, D. (1991). Design and construction failures: Lessons from Petroski, H. (1985). To engineer is human, St. Martins Press, New York.
forensic investigations, McGraw-Hill, New York. Rendon-Herrero, O. (1993). ‘‘Too many failures: what can education do?’’
Lev Zetlin Associates. (1978). Rep. of Engrg. Investigation Concerning J. Perf. Constr. Fac., ASCE, 7(2), 133–139.
Causes of Collapse of Hartford Coliseum Space Truss Roof on January Ross, S., et al. (1984). ‘‘Hartford Civic Center, 1978.’’ Construction dis-
18, 1978, Hartford, Conn. asters, McGraw-Hill, New York.
Levy, M., and Salvadori, M. (1992). Why buildings fall down: how struc- ‘‘Space frame roofs collapse following heavy snowfalls.’’ (1978). ENR,
tures fail, W. W. Norton, New York. January 26.
Downloaded from ascelibrary.org by University of Leeds on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

36 / JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES / FEBRUARY 2001

J. Perform. Constr. Facil. 2001.15:31-36.