Physiology & Behavior 165 (2016) 136–145

Contents lists available at ScienceDirect

Physiology & Behavior

journal homepage: www.elsevier.com/locate/phb

Adaptation of mastication mechanics and eating behaviour to small

differences in food texture
Benjamin Le Révérend ⁎, Françoise Saucy, Mireille Moser, Chrystel Loret
Nestlé Research Center, Vers Chez Les Blancs, CH-1000 Lausanne 26, Switzerland

H I G H L I G H T S

• Seven cereals foods have been tested for masticatory behaviour

• Coupled EMG and Jaw kinematics were analyzed in a time resolved manner
• Fracture force is not a good predictor of oral processing of cereal foods
• Results suggests that food oral processing can guide texture development

a r t i c l e i n f o a b s t r a c t

Article history: Eating behaviour is significantly modified with the consumption of soft or hard textures. However, it is of interest
Received 1 November 2015 to describe how adaptive is mastication to a narrow range of texture. ElectroMyoGraphy (EMG) and Kinematics
Received in revised form 12 July 2016 of Jaw Movements (KJM) techniques were used simultaneously to follow mastication muscle activity and jaw
Accepted 15 July 2016
motion during mastication of seven cereal products. We show that parameters such as the time of chewing ac-
Available online 18 July 2016
tivity, the number of chewing cycles, the chewing muscle EMG activity and the volume occupied for each
Keywords:
chewing cycle are amongst others significantly different depending on products tested, even though the textural
Eating behavior product space investigated is quite narrow (cereal finger foods).
Muscle activity In addition, through a time/chewing cycle dependent analysis of the chewing patterns, we demonstrate that dif-
Food structure ferent foods follow different breakdown pathways during oral processing, depending on their initial structural
Cereals properties, as dictated by their formulation and manufacturing process. In particular, we show that mastication
Food oral processing behaviour of cereal foods can be partly classified based on the process that is used to generate product internal
structure (e.g. baking vs extrusion). To the best of our knowledge, such time dependent analyses have not yet
been reported.
Those results suggest that it is possible to influence the chewing behaviour by modifying food textures within the
same “food family”. This opens new possibilities to design foods for specific populations that cannot accomplish
specific oral processing tasks.
© 2016 Published by Elsevier Inc.

1. Introduction
In recent years, health benefits related to the increase of the resi-
dence time of the food in mouth have been studied, and have been
linked with for example food intake control [1–3]. In addition to those
benefits in adults, several benefits of texture appropriateness during
the development of chewing in children have been identified, from tex-
ture acceptance in adulthood [4] to potentially dentition [5]. In order to
define food properties tuned to a certain oral processing time it is nec-
essary to understand how foods are broken down in mouth and this
topic has attracted interest from the food science community in the
⁎ Corresponding author.
E-mail address: benjamin.lereverend@rdls.nestle.com (B. Le Révérend).
last two to three decades [6–8]. At the heart of this interdisciplinary sci-
ence lies the anatomy and physiology of eating [9], and the food physical
properties (e.g. brittleness and fracture strength) [6,10]. This under-
standing requires in vivo methods as ElectroMyoGraphy (EMG) or Kine-
matics Jaw Movements (KJM). EMG techniques have been used since
late 80′s mainly to link mastication mechanics to sensory properties
and in-vitro characterization [11–13] on all kinds of products going
from jelly to chocolate, cheese, meat, vegetables and biscuits. Since
2000, the interest to investigate the process of food breakdown in the
mouth has raised and the dynamic dimension was introduced [14].
Since mastication is not a static process and that there is continuous
feedback between the central nervous system and the mouth to control
mastication up to the point of swallowing [7], describing the dynamics
of mastication mechanics appears of relevance. Several research groups

http://dx.doi.org/10.1016/j.physbeh.2016.07.010
0031-9384/© 2016 Published by Elsevier Inc.
 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145 137

Table 1 2.1. Study setup and product description

Weight and number of pieces given to study participants to maintain a constant volume of
4.2 cm3 across samples.
Weight Number of pieces
E1 0.68 5 formed about the objectives of the study. In compliance with the
E2 0.94 10
E3 0.73 1 untary informed consent was obtained from participants prior to partic-
C1 3.6 1
B1 2.45 1
B2 3.14 1 Each subject tested each of the seven food products 4 times
B3 2.66 1 monadically and sequentially, in randomized order, across two sessions.
have successfully combined both techniques to improve understanding
in this area [15,16]. Both techniques [16–19], which we argue are very
complementary.
Although several papers have been published in recent years on the
development of mastication in early childhood [20] using EMG [21–23]
and KJM [24–26], the community lacks the understanding of the mature
mastication behaviour of adult subjects., especially when considering
commercial cereal based foods, ranging from breakfast cereals to grano-
la bars. Such data would constitute the limit towards which one would
expect data to tend to after development of such skills, and also to give
an idea of the ability of the scientific protocols to differentiate between
food textures when investigating mature mastication behaviour. In ad-
dition most of the data presented in children mastication development
studies are collected with the aim to compare foods that are widely dif-
ferent in texture [25] (e.g. banana, gelatin, breakfast cereals). Although
those differences are relevant in the context of mastication develop-
ment from an academic perspective, they do not constitute a reference
frame for specific food product development. One recent study, carried
out by Hedjazi et al. [17], showed differences between different types of
cereal foods, but this study only reports data from a single adult subject,
which prevents any meaningful statistical treatment of the data, and
thus scientific interpretation beyond the understanding of this single
subject. In this study we aimed to fill both gaps identified in the litera-
ture; (i) define using both KJM and EMG adult mastication on a restrict-
ed range of texture (controlled, commercially available cereal food
product typically used by children) and (ii) use KJM and EMG to
probe mastication mechanics of such foods in a time resolved manner.
2. Material and methods
In this work we used simultaneously the EMG and KJM techniques to
follow mastication muscle activity and the jaw motion during mastica-
tion of seven cereal based food products.
10 healthy volunteers (5 men, 5 women, aged 26–50, not consulting
for dental treatment at the time of the study) were recruited and in-
World Medical Association Declaration of Helsinki (2008), written vol-
ipation and data were anonymized.
The first repetition of the four carried out was discarded in order to pre-
vent any transient effect coming from the sensory discovery of the
product.
The seven food products were commercial cereal based products
that were either extruded, such as breakfast cereals (E1, E2, E3),
baked biscuits (B1, B2, B3), or a chewy/granola type cereal bar (C1). A
piece of the C1 cereal bar was taken as a reference volume (4.2 cm3).
From this volume, weight and number of pieces was determined for
all the products using the matrix density (see Table 1). The objective
was to obtain sample sizes yielding the same bolus volume by the end
of the chewing process.
2.2. Data acquisition
Data was acquired simultaneously by both techniques. All subjects
involved in this study were equipped with the reflective markers
(KJM) and surface electrodes (EMG) as shown in Fig. 1.
2.2.1. EMG
Data was collected at 1000 Hz using a Noraxon Myosystem 1400
fitted with Noraxon bipolar electrodes (Noraxon, Cologne, Germany)
which were placed on top of the four different closing masticatory mus-
cles (Right Temporalis (RT), Left Temporalis (LT), Right Masseter (RM),
Left Masseter (LM)) and the group of opening muscles (Anterior Belly of
the Digastric (ABD)) relevant to characterizing chewing behaviour (see
Fig. 1). In Fig. 2, one can see a typical EMG recording (amplitude (μV) as
function of time (s)) of all the muscles monitored during the chewing
process.
The signal of the ABD is very weak compared to the other signals.
It was found difficult to accurately position markers to segment the
data in this channel. Thus it was decided to remove this channel
from our analyses and to focus on the jaw closing muscles (RT, LT,
RM, LM).

Fig. 1. Recording setup.

138 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145
Fig. 2. Typical EMG data collected during mastication of a cereal product (a). A close-up view of the rectangle in (a) shows the synchronism of the closing muscles (RT, LT, RM and LM) and
their dephasing with the opening muscle (ABD) (b).
2.2.2. KJM
Data was collected using a NDI Polaris (Northern Digital Imaging,
Waterloo, Canada) motion capture system at 60 Hz. A reflective marker
was placed on the chin to allow recording the motion of the lower com-
pared to the upper jaw. Since the upper jaw constitutes a rigid body
with the entire skull, the reference rigid body was attached to the top
of the skull of the subject using an adjustable head band (see Fig. 1).
Fig. 3 shows a typical KJM recording (amplitude (mm) as function of
time (s)) containing all chewing cycles (a) and a close-up view of a sin-
gle chewing cycle (b).
Successive chewing cycles can be observed in 3D in Fig. 4(a) where
one can see that the successive chewing cycles do not superimpose.
This is even more apparent when data is projected in the frontal plane
as shown in Fig. 4(b). One can then see the rotary motion mentioned
in the literature [25–28] as well as the overpopulation of points close
to the (0,0,0) coordinate. This (0,0,0) point is the occlusal point, which
is reached when the mouth is fully closed. From such 3D data, the work-
ing space of the mastication can be calculated by computing the convex
hull of the 3D trace and then calculate its volume [25].
One limitation of this technique is that KJM markers are stuck on the
skin, and that the jaw and the skin do not form a rigid body. There is
thus a possibility that movements of the skin do not represent accurate-
ly the movement of the jaw. However, KJM is still the method of choice
in most of the recent literature with good reproducibility across sessions
[29]. Other methods have been described in the literature [30] that are
Fig. 3. Typical KJM data collected during mastication of a cereal product (a). A close-up view of the rectangle in (a) shows the synchronism of the Y and Z signals and the de-phasing with
the X signal (b).
less commonly used and may offer the advantage of direct measure-
ment of the jaw motion. However their lesser use makes it difficult to
compare results with other studies, hence our choice of a marker
based methodology.
2.2.3. Synchronization of EMG and KJM
To synchronize both signals, a signal was sent to both equipment
using a custom built electrical device containing a light bulb and an elec-
trical output. When this device was turned on, a pulse of light was sent
via the light bulb, which was recorded by the KJM recorder, as well as a
5 mV square pulse to the EMG recorder. The signal was sent both at the
start and at the end of the experiment, allowing synchronizing the data
on the same time scales over both recorders. Data was then processed so
that only the data recorded between both synchronized signals was
kept. Accuracy of the synchronization method was verified and the pro-
cedure was successful since residual differences in timing were below 1/
60th of a second and limited only by the frequency of the least resolved
signals (KJM, 60 Hz).
2.3. Post processing
For each EMG burst in each of the six channels (5 muscles + sync
trace), markers were placed manually using the dedicated function in
the MyoResearch XP software (Noraxon, Cologne, Germany). Similarly
to the EMG data, KJM data were segmented into individual chewing
 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145 139

Fig. 4. Typical KJM data collected during mastication of a cereal product in 3D (a) and 2D (YZ plane) (b). One can notice the consistency of the occlusal point.

cycles using the vertical axis, since each vertical maximum marks the
full closure of the mouth, and its timeline adjusted using the synchroni-
zation device that is detailed in the previous section.
Using those markers we defined and computed a range of parame-
ters that quantitatively describe mastication, which were either aver-
aged over the whole mastication process (without the first chewing
cycle, since we found that it was very different from the rest of the mas-
tication cycles), or time resolved by isolating the first cycle and first, sec-
ond and third thirds of the mastication process (based on the number of
cycles) [31,32].
be the most variable of all the measurement and a post hoc HSD test
(α = 0.05) was used to show that except for product B3, all product
means were not significantly different from each other. Dry matter
was contained in a very narrow range between 92 and 98%, so that all
products can be considered “dry” when put in the global range of com-
mon foods. Density (bulk and crushed) was the parameter showing the
most variation in this set, mostly driven by the manufacturing process of
the samples (extruded had lower density than baked as expected). Both
densities were highly correlated (r = 0.92) and only bulk/whole density
will be discussed throughout the article.

2.4. Products physical characterization

3.2. Verification of consistency in biomechanical data
All seven products were characterized in texture using a compres-
First, we verified that the data we collected in this study was consis-
sion test between a flat surface and a tooth like geometry (the moving
tent with the existing literature on chewing biomechanics and physiol-
part of the three point bend rig HDP/3PB, Stable Microsystems, UK). In
ogy. In Fig. 2(a), all chewing cycles are represented, showing the
addition to this test, dry matter and density of the products were also
relatively regular patterns of chewing. In Fig. 2(b), a close up view of
tested. Dry matter was obtained by measuring water loss after drying
the successive chewing cycles shows that the closing muscles are in
under vacuum at 60 °C for 72 h. The whole density of the products
phase with one another, whilst the opening muscle ABD is out of
was measured by sand displacement, whilst the matrix density was
phase (activated when the closing muscles are at rest and at rest
measured gravimetrically after the products had been crushed using a
when the closing muscles are activated). This is characteristic of the
mortar and pestle.
mastication pattern as to minimize simultaneous activity of antagonist
muscles [21]. Similarly, for the KJM data in Fig. 3, since the mastication
2.5. Statistical analyses
involves a rotational motion in the sagittal plane around the temporo-
mandibular joint [27,28], it is not surprising that we find that the Y
Statistical analysis were performed using the R statistics software
and Z motions are in phase whilst the X and Y motions are off-phase
[33]. For each parameter, the median over the repetitions (same prod-
due to the rotary motion that exists in the frontal plane.
uct, same subject) were performed and on those medians, the means
of the different products were computed. A mixed model was per-
formed with product as fixed effect and subjects as random effect. 3.3. EMG
A Fisher's Least Significant Difference (LSD) was computed based on
this model to assess which products means are different one from the The total duration of consumption of the seven different products
other. As those methods do not take into account the multiplicity of measured by EMG is presented in Fig. 5. One can see that the E1 is the
tests, in average 5% of significant results are due to chance. This is fastest eaten food and significantly different from E2. B2 and B3 cereal
taken into account in the interpretation of the results by considering products belong to the same group. Finally, C1 takes significantly
the ratio range over LSD within the products/variable sets; it exceeds more time to be eaten than any other products.
1 when products are significantly different. If the value range/LSD was
higher than 1 but below 1.3, the results were still not considered signif- Table 2
icant. On the different graphs, products with means differing by less In vitro product characteristics across samples. For the fracture force, products with differ-
than LSD (not significantly different at 5%) are covered by the same ent letters are statistically different from one another.
grey bar. This formalism is kept throughout this article.
Fracture force (N) Dry matter (%) Density (g/cm3)

E1 34.7a 96.7 ± 0.1 0.161 ± 0.001

3. Results
E2 23.9a 98.2 ± 0.1 0.224 ± 0.002
E3 37.7a 97.8 ± 0.1 0.174 ± 0.004
3.1. Products physical characterization B1 33.7a 96.7 ± 0.04 0.584 ± 0.006
B2 24.2a 97.1 ± 0.1 0.636 ± 0.01
Results from the physical measurements of the different food prod- B3 56.2b 95.4 ± 0.1 0.749 ± 0.033
C1 30.9a 92.7 ± 0.2 0.86 ± 0.02
ucts used in this study are shown in Table 2. Fracture force was found to
140 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145
Fig. 5. Total duration of the mastication process as measured using EMG.
Fig. 6. Total number of chews of the mastication process as measured using EMG.
In addition to the time spent to process the foods in mouth prior to
swallowing, the number of chewing cycles was also measured using
the EMG data. Again the E1 and the C1 stand out in the chewing behav-
iour they elicit. The E1 needs the lowest number of chews (15) whilst
the C1 requires the largest number of chews (29). However, less signif-
icant difference is observed between the different products.
Considering the order of ranking of the different products in Figs. 5
and 6, one can see that there is a good correlation between the number
of cycles and the duration (r = 0.97). This suggests a weak effect of the
food on the chewing period that ranged between 0.88 and 1 s (and its
inverse; the chewing frequency which ranged between 1 and
1.13 Hz). We found no significant difference in chewing frequency for
Fig. 7. EMG activity rate spent over each chew of the mastication process as measured using EMG.
these 7 products. We also found a high correlation coefficient between
the total number of chewing cycles and the total EMG activity that is
spent to breakdown the foods during the mastication process (r =
0.97). Again, the E1 is the product that required the least amount of
EMG activity to be orally processed whilst the C1 required the most.
This suggests that the average EMG activity rate per chewing cycle
was weakly linked to the food products and quantitatively ranged
only between 122 and 148 μV⋅s.
In Fig. 7, we present the data of the EMG activity rate per chew spent
over the whole mastication process, split between first bite (FB), and the
three following thirds of the mastication process. With this transient
view, we are able to emphasize different paths of destructuring. C1
stays always one of the most demanding products to chew throughout
the mastication process, explaining why (combined with its longer du-
ration/number of chewing cycles) it is classified so much higher than all
the other foods in total EMG activity.
Fig. 8 shows that the duration of each chew tends to increase to-
wards the end of the mastication process (similar data was collected
using the EMG), except for the first bite, which is much longer. That
view confirms that we were right to exclude this first bite value from
the averages that were computed to represent the overall mastication
(see Materials and methods). Towards the end of the mastication pro-
cess, the period of chewing cycles increases significantly. This final
phase corresponds to the food bolus preparation and no longer to the
reduction of particle size [34]. It is interesting to notice this increase is
true for all products and is not product dependent. This can thus be
used to characterize the change in the mastication process from one re-
gime to another, to quantify the duration of the different phases of food
oral processing.
3.4. KJM
From the KJM analysis, very similar data can be reported on the over-
all chewing duration, number of chews and thus average chewing cycle
period to that collected with the EMG analysis (r N 0.9 for all three pa-
rameters): E1 and E2 are significantly different in terms of total dura-
tion. The C1 is also significantly different compared to the rest of the
product space, since it requires a longer chewing duration and larger
number of chews during oral processing. Similarly with the KJM, we
found little dependence of the chewing cycles period on the product
being orally processed, and the changes in chewing cycle durations dur-
ing oral processing were the same as those presented in Fig. 8.
If the two methods are very similar when looking for integrated re-
sults such as duration and frequency, when analyzing time dependent
variables, the KJM techniques offers new variables that are simply not
 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145
Fig. 8. Time dependent chewing period of oral processing as measured by EMG.
captured by EMG. In particular, when it comes to analyzing what hap-
pens during each chewing cycle, such as the opening and closing phases,
the working space and the vertical/horizontal amplitudes during
chewing. We want to emphasize particularly those aspects, since
whether one wants to understand mastication mechanics, build an arti-
ficial masticator or design more human like fracture tests, mastication is
not simply a matter of mechanical force but also direction and time be-
tween successive stresses.
The chewing period can be segmented between the opening and the
closing parts of the process. We analyzed both phases and found that
the jaw closing duration is not differentiating between products
(range/LSD = 1.47, barely significant), whilst the jaw opening duration
is (range/LSD = 2.26). In addition, along the product consumption, the
jaw closing duration does not change and stays around 0.4 s. By con-
trast, the jaw opening duration increases significantly from 0.3 s to
0.5 s towards the end of the chewing cycle.
Fig. 9 shows the overall working space that is used during the oral
processing of the different products. One can see that the E1 and E2
do not trigger a large working space whilst the C1 is again highlighted
as the most stimulating/challenging product.
As the different products do not need the same number of chews to
be orally processed, we also looked at the average working space per
chewing cycle in Fig. 10. Less difference between the products is
highlighted by this analysis, meaning that single chew spaces are not
very differentiated by products but since they are not superimposable
onto one another, we see an accumulation effect due to more chewing
cycles during oral processing of C1 in Fig. 9.
In Fig. 10(a), we present the decrease in working space used during
the chewing process from first bite to swallowing. One can see that the
C1 is not the one that requires the most working space in each of the
Fig. 9. Overall working space used during oral processing estimated by computing the
convex hull of the KJM data.
141
section and the B2 or B3 cereal based foods need more working space
than the C1 during most of the chewing process. We assume that B2
and B3 yield more superimposable paths that C1 which thus do not con-
tribute much to the overall working space analysis in Fig. 9, again this
highlights the additional understanding that can be gained by treating
such data as time series rather than as averages. One can see that the
E1, E2 and B1 products break down faster as they are those which
have the largest volume at the start and the least at the end of the pro-
cess, whilst being those that triggered the least number of chews. This is
confirmed by the lower slope coefficient when the working space is
fitted against time after log-transform (Fig. 10(b)). Also interesting is
that E1 yields a significantly lower working space than E2 by the end
of the mastication process (3/3).
In Fig. 11, we decoupled the overall working space by projecting oral
processing paths on the XY plane and we investigated the changes in
vertical and horizontal amplitude. In both directions, a reduction occurs.
Fig. 11(a) and (b) shows that the working space reduction with time of
consumption is due to both reductions in vertical and horizontal ampli-
tude of chewing movements, as previously suggested for the vertical
amplitude but not for the horizontal amplitude [35]. Although it appears
that much more decrease occurs in horizontal rather than vertical am-
plitude, this is mainly an effect of scale due to the large vertical ampli-
tude of the first bite (to fit the product into the mouth). When looking
at the actual variations, reduction of both vertical and horizontal ampli-
tudes goes from 8 to 9 mm in the first thirds to 5–6 mm in the third
third.
We also considered the jaw movements in terms of average horizon-
tal and vertical amplitude during the mastication process (except the
first bite), as shown on Fig. 12. If those two variables were dependent,
the correlation between both variables would yield a coefficient
r N 0.8. It is not the case, we obtained a correlation coefficient of r =
0.53. This indicates that both variables are not linked. This means that
some products yield more vertical chewing motion whilst others yield
more horizontal chewing motion.
Since the vertical and horizontal amplitudes can vary with products,
the velocity at which the jaw moves during oral processing could be
product dependent. Fig. 13 shows that the vertical velocity (range/
LSD = 2.61) is less differentiating than the horizontal velocity (range/
LSD = 3.54). The C1 product is chewed with a highest horizontal and
second highest vertical velocity. This is coherent with the fact that this
product has the largest working space (see Fig. 9) since more distance
is covered in about the same amount of time (constant chewing fre-
quency between products).
Finally, the distance between successive occlusal points was not
found to be product dependent (data not shown, d ≈ 2.5 mm). It was
already suggested in the literature that the consistency of the occlusal
142 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145
Fig. 10. Reduction of the working space used during oral processing estimated by computing the convex hull of the KJM data; raw data (a) and fit of the working space against time after log
transform (b) (the lower the value the faster the decrease).
point is expected in adult mastication behaviour [24]. It is interesting to
notice that even though the path of the jaw during mastication can be
influenced (see Fig. 12), the jaw always closes in the same position.
4. Discussion
In this work, we show significant differences in the chewing behav-
iour of commercial products even if they have similar formulation. The
total duration and the number of chewing cycles are the most obvious
drivers of those differences. This confirms the initial sensory data (not
shown) that was collected where the C1 was the most ‘chewy’ (requir-
ing more chewing). These data are interesting since it is not in agree-
ment to what could be expected from the study of model system in
the literature [32] since we observe large differences in oral processing
duration in spite of little or no difference in Fracture Force (see Table 2),
a typical analysis that is carried out when analyzing material properties
of solid food products. It is supposed to characterize the first and succes-
sive bites of the chewing process. But it is likely that the sensory ratings
of texture of a food as well as the important physiological parameters
related to oral residence time correlate with the total EMG activity rath-
er than simply with the hardness (amplitude/EMG activity at the first
bite). In the context of this study, a not so good correlation (r = 0.59)
was found between the total EMG activity and EMG activity spent dur-
ing the first bite for the seven products. This shows that differences in
oral behaviour cannot be characterized from only the first bite but
that a more global view of the mastication process is necessary, which
also questions the relevance of texture tests as they have been carried
out until this day.
Since we neither see a difference based on fracture force, nor the
EMG activity of the first bite, what is driving differences in food oral pro-
cessing between those samples? We suppose that differences in saliva
softening of the products depending on product structure (reflected
by dry matter and density and fracture force) are the source of these ef-
fects. By analyzing individual chewing cycles, we show that one of the
probable drivers of this rapid in mouth processing is the ability for
some extruded products to break down very fast in mouth, as shown
in Figs. 10, and 11. E1, E2 (and B1 to a lesser extent) trigger a drop of
the chewing space after the first bite. We hypothesize that E1, E2, E3
and B1 are probably sensitive to salivary action and/or mechanical
breakdown due to their structure. This rapid saliva uptake could come
either from the high porosity of low density products or the low water
content (high dry matter value) of the product matrix, values for both
parameters are reported in Table 2. Using correlations we hypothesize
that this is due to the low density, most likely driven by a high porosity
(low density) rather than due to the water content of the matrix (mea-
sure reported as dry matter). An attempt to quantify this relationship
can be approached by correlation post-hoc and we obtained r(whole
density, EMG total duration) = 0.90 and r(whole density, EMG total
number of chews) = 0.82, indicating that this is likely to be true at
least in this sample set. Dry matter and eating behaviour seem also
 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145
Fig. 11. Explanation of the reduction in working space by reduction of both horizontal (a) and vertical (b) amplitude during oral processing.
linked by correlation, but the drier products required less chewing rath-
er than more chewing and r(dry matter, EMG total duration) = −0.88.
If the water content of the product was driving saliva uptake, we would
expect the reverse pattern. Since moisture uptake and migration are
driven by water activity rather than by moisture content (or inversely
dry matter), we disclose this as a limitation of this work. In our sample
set, those differences seem highly driven by the manufacturing process,
Fig. 12. Vertical and Horizontal average amplitude of the jaw movement during oral
processing.
143
leading to different product structures, since in our case, low density
equaled high porosity from the extrusion process. This illustrates the in-
terplay between formulation and processing to reach a final food struc-
ture with controlled properties, beyond simple hardness.
This would not be seen in studies which aim to describe masticatory
performance using model systems such as the commonly used gels
since those are not sensitive to the action of salivary enzymes, neither
porous nor dry, but purely to “mechanical” oral processing. Similarly it
is interesting to note that jaw opening time is varying due the type of
products and throughout food oral processing. This has not been report-
ed before and is in contradiction with the recent findings of Grigoriadis
et al. [35] who reported that both the opening and closing phase were
decreasing during oral processing. We hypothesize that the transition
from a solid brittle food to a pasty sticky bolus in the cereal matrices
we investigated is the reason for this modulation of eating behaviour,
which would be different in the case of the gelatin gels used by
Grigoriadis et al. Such samples could potentially melt in mouth hence
reducing the need for mastication towards the end of oral processing
[35].
In addition, we observe on Fig. 7 that most of the products exhibit an
inverse U shape when it comes to the evolution of EMG activity. This
suggests that during the formation of the food bolus transition from a
solid brittle food to a pasty sticky bolus leads to an initial increase in re-
sistance of the bolus to external stresses. Then as the bolus gets
144 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145
Fig. 13. Horizontal and vertical average velocity of the jaw movement during oral processing.
hydrated with saliva and enzymatically broken down, EMG activity de-
creases throughout the chewing process. This interpretation of the data
then seriously questions the relevance of texture analysis over a single
or a couple of cycles, as we can see that first bite is often not at all rep-
resentative of the full chewing cycles (see Figs. 10 and 11), at least for
cereal based products. If tests are to be conducted on foods aiming to
trigger different oral processing patterns certainly techniques such as
EMG and KJM [31] seem more relevant than texture analysis. Indeed
what we demonstrate here using KJM at the individual cycle level is
that humans adapt their masticatory path resulting in small, significant
and more importantly non-correlated differences in jaw motion vertical
and horizontal amplitudes (as well as velocity), meaning that different
masticatory muscle groups and elemental motor functions are recruited
differently during eating of those foods.
From Fig. 12, one can see that the vertical amplitude is more or less
the same for all products, except the B1 cereal based food and the C1
which are statistically different from the other products (range/
LSD = 1.70). However the horizontal amplitude is more differentiating.
The products are well spread on this axis (range/LSD = 2.41). This find-
ing is in contradiction with those of Hedjazi et al. [17], who found that
all products were following a similar horizontal/vertical displacement
during mastication, and that those two dimensions were always linked
with each other. This might be due to either the fact that the products
tested were not different enough in texture, or that the single subject
used in this study influenced the results in this manner. The E1 yields
mainly vertical motion whilst the C1 yields mainly horizontal motion.
It is interesting to relate these differences in chewing behaviour with
the targeted consumer for those products. We know that at early stages,
children exhibit mainly a vertical motion during chewing [25], so prod-
uct B1 could be more adapted to young children mastication abilities,
since it is also one that destructures very rapidly (see slopes on Fig.
10) and yields a low number of chews/chewing duration (see Figs. 5
and 6). At the opposite, adults exhibit mature mastication with a rotary
motion which is then more capable of breaking down foods needing
shearing action like the C1.
Such differences still require to be linked with food structure, but
such findings offer exciting perspectives to design foods that will devel-
op or maintain masticatory capabilities with more refinement than of-
fered by a sensory profile or a simple mechanical test.
Acknowledgments
Our esteemed colleagues Drs. Christoph Hartmann (Nestlé Re-
search) and Jordan Green (Massachusetts General Hospital) are
thanked for fruitful discussions around this work. We also thank the
reviewers for their useful comments which have greatly improved the
quality of this article.
References
[1] C.G. Forde, N. van Kuijk, T. Thaler, C. de Graaf, N. Martin, Texture and savoury taste
influences on food intake in a realistic hot lunch time meal, Appetite 60 (2013)
180–186.
[2] C.G. Forde, N. van Kuijk, T. Thaler, C. de Graaf, N. Martin, Oral processing character-
istics of solid savoury meal components, and relationship with food composition,
sensory attributes and expected satiation, Appetite 60 (2013) 208–219.
[3] R.A. de Wijk, N. Zijlstra, M. Mars, C. de Graaf, J.F. Prinz, The effects of food viscosity
on bite size, bite effort and food intake, Physiol. Behav. 95 (2008) 527–532.
[4] A. Maier, C. Chabanet, B. Schaal, P. Leathwood, S. Issanchou, Food-related sensory
experience from birth through weaning: contrasted patterns in two nearby
European regions, Appetite 49 (2007) 429–440.
[5] R.S. Corruccini, A.F. Choudhury, Dental occlusal variation among rural and urban
Bengali youths, Hum. Biol. 58 (1986) 61–66.
[6] A.A. Khan, J.F.V. Vincent, Anisotropy in the fracture properties of apple flesh as in-
vestigated by crack-opening tests, J. Mater. Sci. 28 (1993) 45–51.
[7] P. Lillford, The materials science of eating and food breakdown, MRS Bull. 25 (2000)
38–43.
[8] L. Mioche, M.A. Peyron, Bite force displayed during assessment of hardness in vari-
ous texture contexts, Arch. Oral Biol. 40 (1995) 415–423.
[9] L. Mioche, Mastication and food texture perception: variation with age, J. Texture
Stud. 35 (2004) 145–158.
[10] M.-L. Jalabert-Malbos, A. Mishellany-Dutour, A. Woda, M.-A. Peyron, Particle size
distribution in the food bolus after mastication of natural foods, Food Qual. Prefer.
18 (2007) 803–812.
[11] R. González, I. Montoya, J. Cárcel, Review: the use of electromyography on food tex-
ture assessment, Food Sci. Technol. Int. 7 (2001) 461–471.
[12] Y. Ioannides, J. Seers, M. Defernez, C. Raithatha, M.S. Howarth, A. Smith, et al., Elec-
tromyography of the masticatory muscles can detect variation in the mechanical
and sensory properties of apples, Food Qual. Prefer. 20 (2009) 203–215.
[13] M. Devezeaux de Lavergne, J.A.M. Derks, E.C. Ketel, R.A. de Wijk, M. Stieger, Eating
behaviour explains differences between individuals in dynamic texture perception
of sausages, Food Qual. Prefer. 41 (2015) 189–200.
[14] W.E. Brown, D. Braxton, Dynamics of food breakdown during eating in relation to
perceptions of texture and preference: a study on biscuits, Food Qual. Prefer. 11
(2000) 259–267.
[15] H. Koç, E. Çakir, C. VInyard, G. Essick, C.R. Daubert, M.A. Drake, J. Osborne, E.A.
Foegeding, Adaptation of oral processing to the fracture properties of soft solids, J.
Texture Stud. 45 (2014) 47–61.
[16] W.E. Brown, D. Eves, M. Ellison, D. Braxton, Use of combined electromyography and
kinesthesiology during mastication to chart the oral breakdown of foodstuffs: rele-
vance to measurement of food texture, J. Texture Stud. 29 (1998) 145–167.
[17] M.A. Peyron, C. Lassauzay, A. Woda, Effects of increased hardness on jaw movement
and muscle activity during chewing of visco-elastic model foods, Exp. Brain Res. 142
(2002) 41–51.
[18] L. Hedjazi, S. Guessasma, C. Yven, G. Della Valle, C. Salles, Preliminary analysis of
mastication dynamics and fragmentation during chewing of brittle cereal foods,
Food Res. Int. 54 (2013) 1455–1462.
[19] H. Iguchi, J. Magara, Y. Nakamura, T. Tsujimura, K. Ito, M. Inoue, Changes in jaw mus-
cle activity and the physical properties of foods with different textures during
chewing behaviors, Physiol. Behav. 152 (2015) 217–224.
[20] B.J.D. Le Révérend, L.R. Edelson, C. Loret, Anatomical, functional, physiological and
behavioural aspects of the development of mastication in early childhood, Br. J.
Nutr. 111 (2014) 403–414.
 B. Le Révérend et al. / Physiology & Behavior 165 (2016) 136–145
[21] J.R. Green, C.A. Moore, J.L. Ruark, P.R. Rodda, W.T. Morvée, M.J. VanWitzenburg, De-
velopment of chewing in children from 12 to 48 months: longitudinal study of EMG
patterns, J. Neurophysiol. 77 (1997) 2704–2716.
[22] R.W. Steeve, C.A. Moore, J.R. Green, K.J. Reilly, J. Ruark McMurtrey, Babbling,
chewing, and sucking: oromandibular coordination at 9 months, J. Speech Lang.
Hear. Res. 51 (2008) 1390–1404.
[23] R.W. Steeve, C.M. Price, Investigating the use of coherence analysis on mandibular
electromyograms to investigate neural control of early oromandibular behaviours:
a pilot study, Clin. Linguist. Phon. 24 (2010) 485–501.
[24] R.W. Steeve, C.A. Moore, Mandibular motor control during the early development of
speech and nonspeech behaviors, J. Speech Lang. Hear. Res. 52 (2009) 1530–1554.
[25] E.M. Wilson, J.R. Green, The development of jaw motion for mastication, Early Hum.
Dev. 85 (2009) 303–311.
[26] E.M. Wilson, J.R. Green, G. Weismer, A kinematic description of the temporal charac-
teristics of jaw motion for early chewing: preliminary findings, J. Speech Lang. Hear.
Res. 55 (2012) 626–638.
[27] E.G. Gisel, Chewing cycles in 2- to 8-year-old normal children: a developmental pro-
file, Am. J. Occup. Ther. 42 (1988) 40–46.
[28] E.G. Gisel, Effect of food texture on the development of chewing of children between
six months and two years of age, Dev. Med. Child Neurol. 33 (1991) 69–79.
145
[29] L. Remjin, B. Groen, R. Speyer, J. van Limbeek, M. Nijhuis-vander Sanden, Reproduc-
ibility of 3D kinematics and surface electromyography measurements of mastica-
tion, Physiol. Behav. 155 (2016) 112–121.
[30] E. Çakir, H. Koç, C. VInyard, G. Essick, C.R. Daubert, M.A. Drake, E.A. Foegeding, Eval-
uation of texture changes due to compositional differences using oral processing, J.
Texture Stud. 43 (2012) 257–267.
[31] C. Loret, M. Walter, N. Pineau, M.A. Peyron, C. Hartmann, N. Martin, Physical and re-
lated sensory properties of a swallowable bolus, Physiol. Behav. 104 (2011)
855–864.
[32] A. Mishellany-Dutour, M.-A. Peyron, J. Croze, O. François, C. Hartmann, M. Alric,
et al., Comparison of food boluses prepared in vivo and by the AM2 mastication sim-
ulator, Food Qual. Prefer. 22 (2011) 326–331.
[33] R Development Core Team, R: A Language and Environment for Statistical Comput-
ing, 2011.
[34] M.-A. Peyron, I. Gierczynski, C. Hartmann, C. Loret, D. Dardevet, N. Martin, et al., Role
of physical bolus properties as sensory inputs in the trigger of swallowing, PLoS One
6 (2011), e21167.
[35] A. Grigoriadis, R.S. Johansson, M. Trulsson, Temporal profile and amplitude of
human masseter muscle activity is adapted to food properties during individual
chewing cycles, J. Oral Rehabil. 41 (2014) 367–373.