Int J Legal Med (2011) 125:437–444

DOI 10.1007/s00414-011-0551-z

ORIGINAL ARTICLE

Body mass and corrective factor: impact on temperature-based

death time estimation
Michael Hubig & Holger Muggenthaler & Inga Sinicina &
Gita Mall

Received: 5 October 2010 / Accepted: 10 January 2011 / Published online: 1 February 2011
# Springer-Verlag 2011

Abstract Model-based methods play an important role in the body mass is equal to the quotient of the estimated
temperature-based death time determination. The most death time tD and the body mass m (D(tD)/D(m) ≈ tD/m).
prominent method uses Marshall and Hoare's double Additionally, formulae and their approximations are derived
exponential model with Henssge's parameter determination. to quantify the influence of Henssge's body mass corrective
The formulae contain body mass as the only non- factor c on death time estimation. In a range of body masses
temperature parameter. Henssge's method is well estab- between 50 and 150 kg, the relative variation of the body
lished since it can be adapted to non-standard cooling mass corrective factor is approximately equal to the relative
situations varying the parameter body mass by multiplying variation of the death time (ΔtD = (tD/c)Δc). This formula
it with the corrective factor. The present study investigates is applied and compared to computations and to experi-
the influence of measurement errors of body mass m as well mental cooling data with good results.
as of variations of the corrective factor c on the error of the
Marshall and Hoare–Henssge death time estimator tD. A Keywords Time since death . Henssge model . Body mass .
formula for the relative error of tD as a function of the Measurement errors . Estimation errors . Corrective factor
relative error of m is derived. Simple approximations of
order 1 and 0 nevertheless yield acceptable results validated
by Monte Carlo simulations. They also provide the rule of Introduction
thumb according to which the quotient of the standard
deviations D(tD) of the estimated death time and D(m) of Temperature-based death time estimators are most com-
monly used for determining death times in the early post-
mortem phase. The majority are model-based estimators:
They use any form of temperature–time curve T(t)
Electronic supplementary material The online version of this article describing the temperature of the cooling body as a
(doi:10.1007/s00414-011-0551-z) contains supplementary material,
which is available to authorized users. function of the time t since death, e.g. [1–7]. The deep
rectal temperature has been established since it can easily
M. Hubig : H. Muggenthaler : G. Mall (*)
be measured and only slowly converges to the ambient
Institute of Legal Medicine, University Hospital Jena,
Fürstengraben 23, temperature [8]. The most prominent model is the double
07740 Jena, Germany exponential model of Marshall and Hoare [3, 9] with the
e-mail: [email protected] parameter functions by Henssge [4]. To cope with several
types of non-standard cooling conditions such as clothing,
I. Sinicina
Institute of Legal Medicine, University of Munich, covering, ground, and convection differing from the
Munich, Germany experimental standard cooling scenario [4], Henssge devel-
438 Int J Legal Med (2011) 125:437–444

oped the body mass corrective factor [10]. Multiplying the t D . The estimated time since death t D satisfies the
real body mass m by this factor c leads to a different following condition:
“virtual” corrected body mass m′ = cm. The idea is to
T ðTE ; T0 ; m; tD Þ ¼ TM ð2:1Þ
compensate the difference between the non-standard cool-
ing situation N and the standard cooling situation S by The Henssge formula cannot be solved analytically but only
using the model curve T(m′,t) of the standard situation S numerically or by the well-known nomogram [10] as a
with the modified body mass m′ = cNm to describe the non- combination of numerical and graphical methods. Every
standard situation N. Henssge presents non-standard cool- numerical method is among other error types subject to
ing cases Ni in tables with the corresponding corrective errors in the input data, mainly systematic and stochastic
factors cNi [10–12]. errors occurring during measuring temperatures and the body
mass or presupposing the rectal initial temperature.
Death time estimation by the double exponential model There are several approaches to study the influence of
of Marshall/Hoare and Henssge errors in estimating the time since death. Kanawaku et al.
[16] studied the effects of rounding errors on post-mortem
The double exponential model can be expressed by: temperature measurements in the external auditory canal
caused by thermometer resolution. Other groups compared
real cases with exactly known death times to determine the
ðT ðtÞ  TE Þ=ðT0  TE Þ ¼ ½p=ðp  Z Þ expðZtÞ precision of numerical death time estimation models [1, 4,
10–14, 17–19]. While theoretical analyses run into some
 ½Z=ðp  Z Þ expðptÞ ð1:1Þ mathematical and statistical difficulties, experimental
approaches cannot differentiate between the different
error types—measurement errors, errors in the model
with the rectal temperature T, the environmental temper- parameters, systematic errors in the model approach—
ature TE, the rectal temperature at death T0, and the post- and cannot exclude contingent influences (e.g. errors
mortem time t. From their experiments and physical which emerge from specific circumstances at the exper-
considerations, Marshall and Hoare [3] decided to use imental site). The present study chooses a theoretical
only two parameters in their double exponential model: approach combining methods of mathematical analysis
Based on the equation (Eq. 1.1), they introduced two and Monte Carlo simulation like in [15]. It deals with the
parameters, p and Z(S), where S represents the so-called influence of measurement errors Δm of body mass m on
size factor of the human body S = 0.8A/m with the the death time estimator's error ΔtD as well as with the
effective body surface A (unit: square centimeter) influence of the corrective factor c on the death time
involved in heat loss and body mass m (unit: kilogram). estimator tD.
Henssge's cooling experiments and physiological as well An electronic version of this article containing the
as physical considerations [4] made him assume propor- formulae derivations in greater mathematical detail is
tionality of body surface and body mass. He published a provided as electronic supplementary material (ESM).
different parameter definition which is adapted here to a
more convenient form: Z = r + sm−5/8 with r = −0.0284,
s=1.2815 kg5/8 and p = w(TE) Z with w(TE)=5 if TE ≤23. Method
3°C, w(TE)=10 if TE >23.3°C.
The temperature model T(t) of Marshall and Hoare with Analysing the influence of measurement errors is based
the parameter definition by Henssge depends on the on the error propagation law. The algorithm calculating
parameters TE, T0, and m. the output data from the input data is a mathematical
function with the input data in its range and the output
Algorithm for back-calculating time of death and its error data in its domain. The standard deviation of the output
types data with respect to the true values can be approximated
by a Taylor series expansion of this function. There is no
Reconstructing the death time tD, which is more exactly explicit mathematical estimation formula tD = F(TE,T0,m,
defined as the time between death and rectal temperature TM), so that the Taylor series expansion cannot be
measurement, simply means looking for a time since computed directly. To overcome this problem, the implicit
death value tD which makes the temperature model T(t) function theorem and a Monte Carlo simulation are
assume the measured rectal temperature value TM if t = applied [15].
Int J Legal Med (2011) 125:437–444 439

Law of error propagation distribution with expectation value E(Δm)=0 and variance
V(Δm). Concerning possible error sources the stochastic
Equation 2.1 cannot be solved analytically for tD necessary parameters tD » þ ΔtD and m» þ Δm, the expected values
for a direct approach to error propagation. The temperature EðtD » þ ΔtD Þ ¼ tD »; Eðm» þ ΔmÞ ¼ m», and the varian-
curve T(t) is assumed to be a first order continuously ces V ðtD » þ ΔtD Þ ¼ V ðΔtD Þ; V ðm» þ ΔmÞ ¼ V ðΔmÞ are
differentiable function defined on the set of real numbers. of major interest. They can now be approximately calcu-
The mathematical estimation formula F or system function lated using the system function's Taylor series expansion.
representing the death time calculation algorithm hypothet- Since the measurement error in body mass m is stochasti-
ically computes tD = F(TE,T0,m,TM). Applying a Taylor cally independent from measurement errors in the temper-
series expansion and the implicit function theorem on F atures TM, TE, T0, and time tM, covariance matrices have
yields an analytical expression as a linear approximation for not to be considered.
small deviations ΔtD of the estimator tD = F(TE,T0,m,TM)
from the true value tD* of death time as a function of small Monte Carlo simulation
deviations Δm of its input parameter body mass m from its
true value m*. Applying the chain rule and some easy A Monte Carlo simulation is used as independent approach
transformations provides: estimating the variance V(ΔtD) and verifying equation
h
i (Eq. 3.1). A temperature time curve T(t) is computed for
ΔtD =tD ¼ ð5=8Þ= ðr=sÞm5=8 þ 1 Δm=m all values t = ti = i × h (where h=1min) according to the
Henssge model (Eq. 1.1) using fixed values of TE, T0, and
¼: mðmÞΔm=m
m. A random sample {m1, ..., mK} of size K from the
ð3:1Þ domain governed by the stochastic Gaussian distribution
The estimator error value ΔtD which can be computed from PΔm of body mass values with moments E(Δm)=0 and V
(Eq. 3.1) adds to the error value from measurement errors of (Δm) = V is produced using a PC-based random generator.
the input parameters TE, T0, TM, and tM [15]. The factor μ(m) For every simulated body mass mk (where k=1,...,K) a
in (Eq. 3.1) can be linearized in m by a first order Taylor death time estimation is performed assuming T(t,m) as
approximation in the neighbourhood of any body mass value correct curve and interpreting T(t,mk) as temperature
m0. Since our approach concentrates on body masses in the measurement resulting in the corresponding k
k  deathk time tD
interval (50, 150 kg) the Taylor approximation is computed defined by the usual condition TtD ; m ¼ T t; m . On the
in the neighbourhood of the mass m0 =100 kg. This leads to basis of the simulated samples tD1 ; . . . ; tDK of size K, the
the following numerical formula: standard deviation D(tD) = D(ΔtD) of the difference ΔtD
can be estimated using the usual momentum method
mðmÞ  1:03151 þ 0:00419kg1 ðm  m0 Þ ð3:2Þ formula in statistics.
For m=50 kg, the value of μ is approximately 0.82201; for a
Influence of the corrective factor
body mass of 150 kg, the value of μ increases to only
1.24101. Therefore, it is justified to set μ(m) approximately
The corrective factor c was introduced by Henssge [10] as a
to the constant value μ(m)≈1. Inserting this into Eq. 3.1
means to cope with so-called non-standard cooling conditions.
leads to the rule of thumb formula:
For model parameter calibration, Henssge [4] performed
ΔtD =tD  Δm=m ð3:3Þ cooling experiments under standardized environmental con-
ditions (body lying on its back on a blanket on top of a metal
Stochastic interpretation of the error propagation law trolley, no moving air, no irradiation sources). To extend his
standard model parameters to conditions that differ from the
The error propagation law interprets the variables tD and m experimental conditions, he introduced the hypothesis [10]
as random variables assuming an additive stochastic model that the temperature–time curve T(t,m,N) of a body of mass m
[15]: tD* and m* are the constant (correct) values of the cooling under non-standard conditions N is identical to the
variables and added to the stochastic error variables ΔtD temperature–time curve T(t, m′, S) of a body of a different
and Δm. The input body mass therefore is m* + Δm and mass m′—called the corrected mass—cooling under standard
the output death time is tD* + ΔtD. The probability conditions S. This assumption makes the corrected mass m′ a
distributions of the error variables are PΔtD and PΔm function of the original mass m and the non-standard
respectively and—consistent with the central limit theorem conditions m′ = m′(N,m). Henssge partitioned the function
in probability theory—PΔm is assumed to be a Gaussian m′ into two factors: the real body mass m and the corrective
440 Int J Legal Med (2011) 125:437–444

factor c = m′/m which depends on the non-standard condition for those assumptions by cooling experiments using
N. But, it turned out that the correction factor c, at least in dummies [11] and by real world case studies [11, 14] which
certain cases, shows an additional dependence [12] on m: m′ provided lists of the form:
(m,N) = c(N m)m. Henssge and others presented evidence

Case 1 Body mass m1 Non-standard cooling condition N1 Corrective factor c1

* * * *
* * * *
* * * *
Case K Body mass mK Non-standard cooling condition NK Corrective factor cK

by which he was able to compute the correct death time tD the diagram of Fig. 1. The graph of the exact formula
for each case k by using a specific correction factor ck given (Eq. 3.1) is displayed as a drawn line, whereas the
the specific non-standard conditions Nk and the body mass approximation formula (Eq. 3.2) is plotted as a dashed
mk. He proposed to use those lists in future cases with non- line. The linearized curve matches the exact curve very well
standard conditions N and body mass m by choosing a case with a small maximum deviation of 0.03. This proves the
k with most similar non-standard conditions Nk ≈ N and validity of the formula (Eq. 3.2) for computing μ(m).
similar body mass mk ≈ m from the list and using its Henssge's model (Eq. 1.1) is strongly nonlinear concerning
corrective factor ck for death time back-calculation. the mass m at first sight. It is essential to control the
By a short calculation, we can see how powerful this influence of this nonlinearity on the error propagation of m
method is: Since the corrective factor c transforms the expressed in (Eq. 3.1). The diagram in Fig. 1 presents the
real mass m by multiplication of a factor c into a virtual factor mðmÞ ¼ ðΔtD =tD Þ=ðΔm=mÞ on the abscissa and the
one, it provides a mass deviation Δm of the sort we body mass m on the ordinate. In the body mass interval (50,
considered in Eq. 3.1. Now, it is possible to derive an 150 kg), the factor increases almost linearly from μ=0.84
expression for the deviation ΔtD of the death time (m=50 kg) to μ=1.27 (m=150 kg). The narrow range of
estimator tD which is caused by a deviation Δc of the μ(m) justifies the previously formulated rule of thumb
corrective factor c. Let m# be the true body mass and m be (Eq. 4.2) approximating μ≈1.
the true corrected body mass which yields the true
correction factor c = m/m# and let m′ be the erroneous Stochastic results (Fig. 2)
“corrected” body mass according to a false corrective
factor c′ = m′/m# . With Δm = m′ − m, one obtains after The curves presented in Fig. 2 are graphs of the function D
some transformations from Eq. 3.1: (tD)—the standard deviation of the death time estimator tD
  as function of the true time of death. The true time between
ΔtD =tD ¼ m cm# Δc=c ð4:1Þ death and temperature measurement is indicated by the
This formula provides control of the death time recon- ordinate, the standard deviation by the abscissa. The
struction error caused by a numerically fixed uncertainty estimation was performed twice, once by the error
of the corrective factor. With the approximation μ=1, we propagation law (DEP) and also by Monte Carlo simulation
come to the following rule of thumb: (DMC) with a sample size of K=1,000. Cooling scenarios
with TE =18°C and T0 =37.2°C under reference standard
conditions were computed: (a) with body mass m=50 kg,
ΔtD =tD  Δc=c ð4:2Þ
(b) with body mass m=75 kg, (c) with body mass m=
100 kg, and (d) with body mass m=125 kg. The different
body masses m were inserted in the Henssge formula for
Results computing DMC whereas DEP was computed using formula
(Eq. 3.1). Though the parameters T0 and TE had to be fixed
Analytical results (Fig. 1) for the Monte Carlo computations, the results of the Monte
Carlo simulations and of the error propagation law were
The formula (Eq. 3.1) is solved for the proportionality independent of T0 and TE. This becomes evident in the
factor μ to display its dependence on the body mass m in formula (Eq. 3.1) which does not contain the variables T0
Int J Legal Med (2011) 125:437–444 441

Fig. 1 Factor μ(m) = (ΔtD/tD)/

(Δm/m) as a function of m for
computation of the influence
of errors in body mass and of
errors in corrective factor c. Full
line, μ(m) exact curve. Dashed
line, μ(m) ≈ a + b(m−100 kg)
Taylor approximation of order 1
in the neighbourhood of m=
100 kg with a=1.03151 and
b=0.00419 kg−1

and TE. The influence of “random noise” in the input The proportionality constant of this linear relation has the
variable body mass m on output death time estimate tD is term (r/s)m13/8 +m in the denominator, decreasing body
presented in Fig. 2 for a realistic standard deviation of the masses thus leads to steeper slopes. Accordingly, the line
body mass measurement of only D(m) = m/100. The with the steepest slope belongs to cooling scenario (a) with
estimator values DEP and DMC are indicated on the abscissa a body mass m=50 kg and the line with most moderate
in relation to the timespan tD between death and measure- slope to cooling scenario (d) with a body mass m=125 kg.
ment on the ordinate and the body mass m in the different The graphs DMC(tD) of the Monte Carlo simulations are in
cooling scenarios (a), (b), (c), and d. The graphs of the good accordance with the graphs DEP(tD) in all scenarios
death time estimator's standard deviations DEP(tD) comput- (a)–(d). This justifies the first order Taylor series approach
ed by the error propagation law (Eq. 3.1) as functions of deriving formula (Eq. 3.1). The standard deviation D(tD)
death time tD are straight lines. Equation 3.1 shows that the of the death time estimator tD following errors in body
proportionality factor μ establishing the relation between mass measurement m with a standard deviation D(m)=
ΔtD/tD and Δm/m is constant for fixed m. Solving for ΔtD 0.01 m increases from D(tD)=0.01 h at tD =1 h post-
and inserting Δm = D(m) produce a linear equation in tD. mortem up to D(tD)=0.46 h in case (a), D(tD)=0.415 h in

Fig. 2 Standard deviation D

EP(tD) (solid lines) and DMC(tD)
(dotted lines) of the death time (a)
estimator tD as functions of time
(b)
between death and measurement
for standard deviation D(m)=m/ (c)
100. Computations for fixed
body masses m: (a) m=50 kg, (d)
(b) m=75 kg, (c) m=100 kg,
and (d) m=125 kg
442 Int J Legal Med (2011) 125:437–444

case (b), D(tD)=0.375 h in case (c), and D(tD)=0.335 h in 1.3 (H3) and the environmental temperature TE(t) in the
case (d) at tD =40 h. The standard deviation D(tD) of the climatic chamber. The beginnings of the Henssge cooling
death time estimator tD is in the order of magnitude 1% of curves THi(t) are adjusted to the true death time. The curve
the death time tD estimated given the standard deviation D TH0(t) fits the data curve T(t) best. With increasing
(m)=m/100 of the body mass m measurement. From the corrective factor ci, the temperature curves THi(t) stay
diagram in Fig. 2, the following rule of thumb in above the data curve. The deviations can easily be
accordance with Eq. 3.3 can be deduced: The quotient of converted graphically into the corresponding errors ΔtD of
the standard deviations D(tD) of the estimated death time death time estimation given a fixed point in time tD0 and the
and D(m) of the body mass value is approximately equal computed temperature TH0(tD0) respectively by shifting the
to the quotient of the estimated death time tD and the body curve THi(t) to the left until the time–temperature point (tD0,
mass: D(tD)/D(m) ≈ tD/m. TH0(tD0)) meets the curve THi(t); the shifted time distance is
ΔtDi. Taking, e.g. TH0(tD0)=25°C and tD0 =27.1 h, the
Case results (Fig. 3) procedure provides the following tDi values: tD0 =27.10 h,
tD1 =30.00 h, tD2 =33.33 h, and tD3 =36.52 h. Therefore, we
We present the cooling curve of a middle-aged male of 177- have the following death time estimation errors ΔtD0 =0 h,
cm height and 103-kg weight wearing T-shirt, cotton shirt ΔtD1 =tD1 –tD0 =2.9 h, ΔtD2 =tD2 –tD0 =6.43 h, ΔtD3 =tD3 –
with long sleeves, underpants, leggings, jeans, socks, and tD0 =9.51 h in reasonable accordance with the rule of thumb
slippers. He collapsed in the cabin of his truck and was error predictions (Eq. 4.2): ΔtD0 ≈0 h, ΔtD1 ≈2.71 h, ΔtD2 ≈
unsuccessfully resuscitated in an ambulance, where he died. 5.42 h, ΔtD3 ≈8.3 h.
The environmental temperature at the death scene in the
ambulance was TE =17°C. The deceased was transferred to
a climatic chamber 1.48 h after death and the rectal cooling Discussion
curve T(t) was recorded under strictly controlled environ-
mental parameters (TE =17°C, no air movement, no radia- The body mass m plays a crucial role in Henssge's model
tion, back position on a metal trolley). Several cooling for death time determination [4]. It represents the only non-
curves THi(t) were computed using Henssge's model temperature dependent model parameter, located in the
(Eq. 1.1). Since the body was covered by two to three argument of the exponential function and bears the
layers of clothing, corrective factors c ranging from c0 =1.0 exponent −5/8 (Eq. 1.2) causing a distinct nonlinearity in
to c3 =1.3 were applied [20]. Figure 3 shows the measured the cooling model. In real case work, body masses should
cooling curve T(t), the cooling curves TH0(t), TH1(t) TH2(t), be measured by gauged scales to avoid systematic errors.
and TH3(t) according to the Henssge model with correction Nevertheless, body mass measurement is inevitably subject
factors of c0 =1.0 (H0), c1 =1.1 (H1), c2 =1.2 (H2), and c3 = to measurement noise. Moreover, changes of the body mass

40

Henssge TH0 c = 1.0

Henssge TH1 c = 1.1
Henssge TH2 c = 1.2
35
Henssge TH3 c = 1.3
Measured T
Environmental TE
Temperature [°C]

30

25

20

15
0 1 3 4 5 7 8 9 11 12 13 15 16 17 19 20 21 23 24 25 27 28 29 31 32 33 35 36 37 39 40 41 43 44
Time since death [h]

Fig. 3 Climatic chamber cooling experiment. Temperatures: measured T(t); environmental TE(t); and Henssge with correction factor c0 =1.0:
TH0(t); c1 =1.1: TH1(t); c2 =1.2: TH2(t); c3 =1.3: TH3(t)
Int J Legal Med (2011) 125:437–444 443

from the death scene to the autopsy when the body mass is easy estimation of death times with various corrective factors
measured can occur due to the loss of blood from wounds at the death scene without the necessity of using a computer or
during transportation of the corpse. The present study the nomogram. The rule also demonstrates the immense
investigates first the influence of measurement errors of the power of the corrective factor. An estimated time since death
body mass m (quantified by the standard deviation D(m)) on will be doubled by assuming a corrective factor of 2. This
the standard deviation D(tD) of the death time estimator tD strong influence of the corrective factor on the outcome of the
and second, the influence of the corrective factor c multiplied death time estimation is contrasted by the uncertainty of
to body mass m. Two independent approaches are used for choosing the appropriate corrective factor since only relatively
this investigation: first, analytical computation of the error few tables with cooling experiments are available to cope with
propagation law and second, numerical estimation via Monte an infinite variety of real death scene boundary conditions.
Carlo simulation. Identical results from both approaches It is not possible to perform the usual stochastic error
prove the validity of our results. According to our results, the analyses for input errors of the corrective factor c according to
relative error in body mass (e.g. 1%) corresponds to the the law of error propagation since the input random variable
relative error in the estimated death time (e.g. 1% as well). c is not defined on a proper probability space with a well-
The influence of the body mass errors on death time known error probability distribution. Actually, the variable c
estimation can be compared to the influence of other input is defined on the class N × IR of ordered pairs (N,m) with N
variable errors [15], first, to errors in the presumed initial being any possible non-standard cooling condition and m
rectal temperature at death T0 and second, to errors in the being any possible body mass, where N is an infinite space
environmental temperature TE. The error assumptions were without a clear set definition and with no quantifiable
adapted to realistic numbers with a standard deviation of mathematical structure or even a topology. Therefore, we
the measured body mass D(m)=0.01 m, of the initial rectal only present the above formulae (Eqs. 4.1 and 4.2) to provide
temperature D(T0)=0.5°C and of the environmental tem- a tool for error estimation if the amount of possible error in
perature D(TE)=1°C. Graphically, the T0-induced standard the variable c is known. Additionally, we present a real
deviation curve of the death time estimate decreases with cooling case and show the validity of the formula (Eq. 4.2).
time instantly from high values to a low constant socket Data curves of a cooling experiment in a climatic chamber
value. The m-induced standard deviation curve of the death are presented and compared to Henssge's model cooling
time estimate on the contrary starts from small negligible curves for different corrective factor values c=1.0, c=1.1,
values and continuously increases with time. Initially c=1.2, and c=1.3. The appropriate corrective factor taking
during the first hours of cooling, the m-caused standard into account the clothing of the body would have been at
deviation is small compared to the T0-caused standard least c=1.1 or c=1.2, while the model curve computed with
deviation. After 5 h (for smaller body masses) up to 10 h a corrective factor of c=1.0 fitted the data curve best. Faint
(for high body masses), the m-caused standard deviation air movement in the climatic chamber necessary to keep the
reaches the socket value (0.4–0.6 h) of the T0-caused air temperature in the chamber constant is no argument,
standard deviation. For death times >10 h, the body mass since the fan is positioned in a considerable height above the
m-caused standard deviation outweighs the T0-caused trolley with the cooling body and the air movement—
standard deviation since it linearly increases with time. measured with a thermoanemometer—produced mean air
Although there have been several approaches [21, 22] to velocities of only 0.018 m/s, which according to thermody-
establish models T(t) which do not necessarily require the namics literature [23] can be neglected compared to (the
environmental temperature TE, accurate knowledge of this always present) natural convection. In usual routine case
temperature is essential for death time estimation [15]. The work, the real time of death would not have been known and
TE-caused standard deviation of the death time estimate application of higher correction factors would have resulted
exponentially increases with time. Initially, from 0 to 10 h in an overestimation of the time since death. The results
post-mortem, it amounts to less than half of the m-caused obtained from the graphical comparison of the data curve to
standard deviation. From 10 to 25 h post-mortem, the TE- the four model curves for the four different corrective factors
caused standard deviation reaches to the level of the m- support the analytical results and the results from the rule of
induced standard deviation. In later cooling phases, the m- thumb approximation.
induced standard deviation becomes negligible compared to
the TE-caused standard deviation.
Apart from measurement errors of body mass, the References
influence of the corrective factor was investigated. Accord-
ing to our derived rule of thumb, the relative deviation of 1. De Saram GSW, Webster G, Kathirgamatamby N (1955) Post-
the corrective factor is approximately equal to the relative mortem temperature and the time of death. J Crim Law Criminol
deviation of the death time estimate. This rule allows an Police Sci 46:562–577
444 Int J Legal Med (2011) 125:437–444

2. Lundquist F (1956) Physical and chemical methods for the estimation mit Berücksichtigung von Einflußfaktoren. Beitr Ger Med 42:323–
of the time of death. Acta Med Leg Soc (Liege) 9:205–213 329
3. Marshall TK, Hoare FE (1962) Estimating the time of death. The 14. Albrecht A, Gerling I, Henßge C et al (1990) Zur Anwendung des
rectal cooling after death and its mathematical expression. J Rektaltemperatur-Todeszeit-Nomogramms am Leichenfundort. Z
Forensic Sci 7(1):56–81 Rechtsmed 103:257–278
4. Henßge C (1979) Die Präzision von Todeszeitschätzungen durch 15. Hubig M, Muggenthaler H, Mall G (2010) Influence of measurement
mathematische Beschreibung der rektalen Leichenabkühlung. Z errors on temperature-based death time determination. Int J Leg Med.
Rechtsmed 83:49–67 doi:10.1007/s00414-010-0453-5
5. Henßge C (1988) Death time estimation in case work. I. The rectal 16. Kanawaku Y, Kanetake J, Komiya A, Maruyama S, Funayama M
temperature time of death nomogram. Forensic Sci Int 38:209– (2007) Effects of rounding errors on postmortem temperature
236 measurements caused by thermometer resolution. Int J Leg Med
6. Hiraiwa K, Ohno Y, Kuroda F, Sebetan IM, Oshida S (1980) 121:267–273
Estimation of postmortem interval from rectal temperature by use 17. Henßge C, Althaus L, Bolt J et al (2000) Experiences with a
of computer. Med Sci Law 20(2):115–125 compound method for estimating the time since death. I. Rectal
7. Morgan C, Nokes LD, Williams JH, Knight BH (1988) Estimation temperature nomogram for time since death. Int J Leg Med 113
of the post mortem period by multiple-site temperature measure- (6):303–319
ments and the use of a new algorithm. Forensic Sci Int 39(1):89– 18. Al-Alousi LM, Anderson RA, Worster DM, Land DV (2002)
95 Factors influencing the precision of estimating the postmortem
8. Lim CL, Byrne C, Lee JK (2008) Human thermoregulation and interval using the triple-exponential formulae (TEF). Part II. A
measurement of body temperature in exercise and clinical settings. study of the effect of body temperature at the moment of death on
Ann Acad Med Singapore 37(4):347–353 the postmortem brain, liver and rectal cooling in 117 forensic
9. Marshall TK (1969) The use of body temperature in estimating cases. Forensic Sci Int 125(2–3):231–236
the time of death and its limitations. Med Sci Law 9:178–182 19. Verica P, Janeska B, Gutevska A, Duma A (2007) Post mortem
10. Henßge C (1981) Todeszeitschätzungen durch die mathematische cooling of the body and estimation of time since death. Soud Lék
Beschreibung der rektalen Leichenabkühlung unter verschiedenen 52(4):50–56
Abkühlungsbedingungen. Z Rechtsmed 87:147–178 20. Madea B, Brinkmann B (2004) Handbuch gerichtliche Medizin
11. Henßge C, Madea B (1988) Methoden zur Bestimmung der Band 1, 1st edn. Springer, Berlin
Todeszeit an Leichen. In: Weinig E, Berg S (eds) Arbeitsmethoden 21. Green MA, Wright JC (1985) Postmortem interval estimation
der medizinischen und naturwissenschaftlichen Kriminalistik, Vol- from body temperature data only. Forensic Sci Int 28(1):35–46
ume 18. Max Schmidt–Römhild, Lübeck. ISBN 3-7950-0624-4 22. Mall G, Eckl M, Sinicina I, Peschel O, Hubig M (2005)
12. Henßge C (1992) Rectal temperature time of death nomogram: Temperature-based death time estimation with only partially
dependence of corrective factors on the body weight under stronger known environmental conditions. Int J Leg Med 119(4):185–
thermic insulation conditions. Forensic Sci Int 54:51–56 194
13. Stipanits E, Henßge C (1981) Präzisionsvergleich von 23. Cengel YA (1998) Heat transfer a practical approach. WCG/
Todeszeitrückrechnungen aus der Rektaltemperatur ohne und MCGraw-Hill, New York
Copyright of International Journal of Legal Medicine is the property of Springer Science & Business Media
B.V. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright
holder's express written permission. However, users may print, download, or email articles for individual use.