rsif.royalsocietypublishing.org

ResearchCite this article: Karme A, Rannikko J,

Kallonen A, Clauss M, Fortelius M. 2016

Mechanical modelling of tooth wear. J. R. Soc.

Interface 13: 20160399.

http://dx.doi.org/10.1098/rsif.2016.0399

Received: 20 May 2016

Accepted: 20 June 2016

Subject Category:Life Sciences – Earth Science interface

Subject Areas:biomechanics, bioengineering

Keywords:chewing machine, microwear, plant material,

phytoliths, grit, tooth wear

Authors for correspondence:Aleksis Karme

e-mail: aleksis.karme@helsinki.fi

Janina Rannikko

e-mail: janina.rannikko@helsinki.fi

†These co-first authors contributed equally to

this study.

Electronic supplementary material is available

at http://dx.doi.org/10.1098/rsif.2016.0399 or

via http://rsif.royalsocietypublishing.org.

& 2016 The Author(s) Published by the Royal Society. All rights reserved.

Mechanical modelling of tooth wear

Aleksis Karme1,†, Janina Rannikko1,†, Aki Kallonen2, Marcus Clauss3

and Mikael Fortelius1

1Department of Geosciences and Geography, Division of Biogeosciences, and 2Department of Physics, Division ofMaterials Physics, University of Helsinki, Helsinki, Finland3Clinic for Zoo Animals, Exotic Pets and Wildlife, Vetsuisse Faculty, University of Zurich, Zurich, Switzerland

AK, 0000-0003-4852-2636; JR, 0000-0002-5542-4201; MC, 0000-0003-3841-6207;MF, 0000-0002-4851-783X

Different diets wear teeth in different ways and generate distinguishable

wear and microwear patterns that have long been the basis of palaeodiet

reconstructions. Little experimental research has been performed to study

them together. Here, we show that an artificial mechanical masticator, a

chewing machine, occluding real horse teeth in continuous simulated chew-

ing (of 100 000 chewing cycles) is capable of replicating microscopic wear

features and gross wear on teeth that resemble wear in specimens collected

from nature. Simulating pure attrition (chewing without food) and four

plant material diets of different abrasives content (at n ¼ 5 tooth pairs per

group), we detected differences in microscopic wear features by stereomicro-

scopy of the chewing surface in the number and quality of pits and scratches

that were not always as expected. Using computed tomography scanning in

one tooth per diet, absolute wear was quantified as the mean height change

after the simulated chewing. Absolute wear increased with diet abrasive-

ness, originating from phytoliths and grit. In combination, our findings

highlight that differences in actual dental tissue loss can occur at similar

microwear patterns, cautioning against a direct transformation of microwear

results into predictions about diet or tooth wear rate.

1. IntroductionThere has been considerable debate regarding the relative role of extrinsic ‘grit’

and internal minerogenic plant parts (phytoliths) as drivers of tooth wear.

Hardness of enamel, opal phytoliths and external grit have been in the centre

of the debate. Although opal phytoliths in pasture and fodder plants were

measured harder than enamel [1], the role and hardness of phytoliths have

been repeatedly questioned, partly on the basis of modern experimental work

[2–5]. Despite the conclusion that phytoliths cause plastic deformation rather

than removal of enamel, a recent ground-breaking study shows that phytoliths

do in fact impact enamel [5]. Very recently, particles softer than enamel were

reported to generate grooves, resulting from tissue loss and not plastic defor-

mation, on enamel in macro- (aluminium and brass spheres) and nanoscale

(amorphous silicon dioxide spheres) [6].

There is now growing recognition that an experimental study of dental wear

is needed to resolve the long-running debate, and more experimental work is

now being published. In pioneering work, Maas [7,8] constructed two exper-

imental studies for microwear focusing on species differences in enamel

microstructure, direction of shearing forces relative to enamel prisms, size of

the abrasive particles and magnitude of the force. The effect of individual

particles on dental enamel has also been studied experimentally, but not in a

manner that would have generated cumulative patterns similar to those caused

by long-term chewing [4,9]. The effect of different diets was explored in rabbits

(Oryctolagus cuniculus) with respect to three-dimensional texture analysis and

microwear, where grass diets resulted in scratch-dominated microwear patterns

and lucerne diets resulted in more pitted patterns [10]. Low silica content in the

diet also resulted in more variable surface textures and microwear patterns.

Exposing rabbits and guinea pigs (Cavia porcellus) to diets designed to reflect

moving part

25 mm arm with a lower tootharm with an upper tooth

(b)

(a)

Figure 1. Mechanical masticator. Panel (a) shows the machine and teethattached. Panel (b) shows a scheme of the machine. (Online version in colour.)

rsif.royalsocietypublishing.orgJ.R.Soc.Interface

13:20160399

2

different proportions of internal and external abrasives (sand),

and measuring the resulting effects manually and by computed

tomography, differences in wear and in the responding tooth

growth could be demonstrated [11,12]. Most recently, the

effects of external grit on the microwear signature were demon-

strated in live sheep [13] and angles of approach in chewing

were shown to affect microwear [14].

Microscopic study and analysis of worn tooth surfaces

[15,16] (microwear) have become a standard tool for identify-

ing the diet preferences of fossil and living vertebrates.

Microwear analysis has been widely used to study diet

among ungulates [17–19] and other animals [20–23].

Among ungulates, individuals consuming browse (dicotely-

donous plants) can be distinguished by their microwear

patterns from those consuming graze (monocotelydonous

plants). The main pattern, reconfirmed in a large number of

studies, is that browsers have pit-dominated microwear

while grazers have scratch-dominated microwear [15,17,24].

Among another widely studied group, the primates, hard

food consumers have a mainly pit-dominated microwear

pattern and tough food consumers a more scratch-dominated

microwear pattern [25,26].

Here, we present the microwear analysis results and actual

dental tissue loss (i.e. tooth wear rates) of a chewing exper-

iment with the standardized pelleted foods used in two

previous live animal studies [11,12]. The experiment was con-

ducted under controlled laboratory conditions, using

mechanical modelling (chewing machine) [27] to occlude real

horse teeth submerged in food in a simple chewing cycle. In

designing the mechanical masticator, it was not intended to

develop a system that mimics the conditions in the actual

animal mastication with detailed precision; for example, we

did not intend to emulate the consistency and properties of

saliva, or to exactly replicate the direction and detailed move-

ment of antagonistic dental contact. Instead, the aim was a

simplified system that would allow changing various factors

in mastication, for example different standardized feeds, to

test whether expectations based on a simplistic interpretation

of microscopic and macroscopic wear could be confirmed.

The aim of this study was to produce microwear features

seen in nature, i.e. pits and scratches, and to quantify macro-

scopic wear with real teeth and diets, in order to achieve a

detailed picture of the wear process and its components.

2. Material and methodsThe experiment was conducted with four different diets and

pure attrition with 25 pairs of horse molar teeth. Horse teeth

were used because they are relatively easy to obtain, large and

the three dental tissues, enamel, dentine and cement, are all vis-

ible on the surface at the same time. The third and fourth

premolars, as well as the first and second molars, were used,

as their shape is more symmetrical than teeth on either end of

the cheek tooth row.

Five pairs of teeth were made to chew in each diet and a set

of five teeth were also made to chew in water only, simulating

pure attrition (ATTR) without a buffering layer of food. Chewing

was conducted with a mechanical masticator (figure 1) built at

the University of Helsinki, Department of Geosciences and

Geography and Department of Physics [27], and the methods

were developed in pilot studies [27,28].

The pelleted diets were based on lucerne, L (or alfalfa,

Medicago sativa), which by nature contains very low levels of

internal abrasives [29], grass, G, which contains higher levels of

internal abrasives, and grass with the addition of rice hulls,

GR, which contains again higher levels [29]. The fourth pelleted

diet included grass and rice hulls (internal abrasives) and added

sand (external abrasives), GRS. These diets were chosen because

of the aforementioned attributes, and because the exact same

diets had been used in previous studies with live animals,

where they had been demonstrated to cause differences in

tooth wear [11,12]. Abrasives, measured as acid detergent insolu-

ble ash (ADIA) in dry matter, increased in diets without sand

from L (5 g kg21), to G (16 g kg21) and to GR (24 g kg21). A

basic assumption was that most of the abrasives are silica phyto-

liths [30]. Lucerne, which is a flowering plant in the pea family, is

essentially phytolith-free and is similar to browse in this respect.

GRS contained sand grains (mean diameter of 233 mm) as an

external abrasive, which increased the ADIA value and abrasive-

ness (ADIA 77 g kg21). A complete ingredient list and nutrient

composition of diets can be found elsewhere [11]. Teeth were

submerged in a relatively thick food bolus to achieve a cushion

effect during chewing, avoiding pure non-buffered attrition. In

total, 1 l of dry pellets was mixed with 3 l of water, which had

3 g salt per 1 l to avoid the expansion of the plant cells.

Before the chewing procedure, teeth were prepared to fit the

machine (figure 2a). After the roots had been cut off, the teeth

were glued into three-dimensionally printed polylactic acid

rings with epoxy, in order to fit them in the mechanical mastica-

tor. Preparations were finished by cutting the occlusal surfaces at

an angle of 338. Before chewing, the surfaces of the teeth were

polished with an abrasive disc grinding machine (grain size

68 mm, CAMI 220, ISO 6344 P220) (figure 2b). Teeth were posi-

tioned in the mechanical masticator so that scratches due to

chewing (parallel to chewing direction) could be differentiated

from the initial polishing striation (perpendicular to chewing

direction) (figure 2b,c). Fading of initial polishing striations was

used as a rough indicator of wear rate.

All tooth pairs were made to chew submerged in the diet–

water mixture for 6 h and 30 min, which equals roughly 100 000

chews (260 chews per minute) resulting in a speed of

108 mm s21, comparable to real chewing speeds of horses [31].

The mechanical masticator simulates a repeated, full-occlusion

single stroke movement between a pair of upper and lower

teeth with surfaces flattened to a single plane [27,28]. A motor

moves the other arm of the machine, which has a lower tooth

lingual sideroots

(a)

(b) (c)

0.4 mm 0.4 mm

buccal side

PLAring 33º

Figure 2. (a) Schematic of the preparation of a tooth (modified from Bertin et al. [28]). First, the roots are cut off, then the tooth is placed inside a polylactic acid(PLA) ring with epoxy and finally the occlusal surface is cut. (b) Polished occlusal surface where striation from the polishing procedure goes from the upper leftcorner to lower right corner. (c) Occlusal surface after 6 h and 30 min of chewing. Microwear counting area (0.4 � 0.4 mm2) is marked with a black square. Scalebars, 0.4 mm.

rsif.royalsocietypublishing.orgJ.R.Soc.Interface

13:20160399

3

attached, in a horizontal back and forth movement. An upper

tooth is attached to another arm, which is flexible. As the lower

tooth contacts the upper tooth, the upper tooth gives way slightly,

and the lower tooth slides along it, simulating occlusal contact

during the power stroke of natural chewing. To avoid the food

particles and sand particles from settling to the bottom of the

food container, we used a mechanical mixer and compressed air

conducted to the bottom of the container to keep the food

circulating.

Silicon moulds (President Plusw regular body, Coltene/

Whaledent AG, Altstatten, Switzerland) were made from a tooth

surface before and after chewing. Transparent epoxy casts (Epo-

Fixw, Struers Inc., Cleveland, Ohio, USA, base liquid and

hardener) were made out of the moulds and examined with a

stereomicroscope (SZX 10, Olympus Industrial, Essex, UK). The

casts were orientated so that the chewing scratches were 458from the centre line and illuminated from below in such a way

that a totally reflecting surface could be viewed through the micro-

scope. We decided to use a light microscope instead of scanning

electron microscope (SEM), because of easier access, affordability

and reduced time. Light microscopes have been used successfully

in other microwear studies [18,19,24,32]. Note that light micro-

scope results cannot be compared with SEM results directly due

to different magnification.

Images from each tooth were taken from two standard

locations in the enamel band in the middle of the teeth with a

Colorview III camera (Soft Imaging System GmbH, Munster,

Germany) attached to the microscope. All images were taken

with 32� magnification. Images were processed with Corel

Paint Shop Prow X5 (Corel Corporation, Ottawa, Canada) to

improve the visibility of wear features. Images were made

greyscale and negative and local tone mapping was used to

adaptively enhance the contrast.

The number of pits and scratches, which were generated in

the chewing procedure, and initial polishing striations, which

were generated in the polishing of the occlusal surfaces with

the grinding machine, were recorded from a 0.4 � 0.4 mm2 area

inside central enamel bands (figure 2c) with Microware 4.02

(P. Ungar, Fayatteville, Arkansas, USA). Pixel coordinates of

the major and minor axis of each feature were extracted, which

allowed us to calculate the absolute lengths and widths with

known pixel size. We refer to the major axis as ‘length’ and the

minor axis as ‘width’. A feature was categorized as a pit if the

ratio of length and width was less than four, and a scratch if

the ratio was more than four. Features were categorized to

small (length less than 20 mm), large (length 20–50 mm) or very

large (length more than 50 mm) pits [33] and thin (width less

than 15 mm) or wide (width more than 15 mm) scratches [32,34].

JMPw Pro 10 (SAS Institute Inc., Cary, North Carolina, USA)

was used to analyse the results statistically and for visualization.

Number of pits, number of small pits, number of very large

pits, number of wide scratches, pit length and pit width were

log-transformed to obtain normal distribution and equal var-

iances. Normal distribution was tested with Shapiro–Wilk test

and variance equality between groups with Levene’s test. Results

were analysed with analysis of variance (ANOVA) when groups

were normally distributed and their variances were equal, and

with Kruskal–Wallis test when normal distribution within or

equal variances between groups were not obtained. Differences

between groups (with normal distribution and equal variances)

were analysed with Tukey–Kramer HSD test. Wilcoxon method

was used to analyse the initial polishing striation count as it did

Table 1. Means and standard deviations of various microscopic measures of tooth wear in horse cheek teeth used in an artificial mastication system using fourdifferent diets (L, lucerne; G, grass; GR, grass and rice hulls; GRS grass, rice hulls and sand) as well as without food (ATTR attrition). All features were measuredfrom an area of 0.16 mm2. Superscripts a – c within rows indicate significant differences between diets (Wilcoxon method for initial polishing striationsremaining, and Tukey – Kramer HSD for all others).

measure (unit) L G GR GRS ATTR

scratches (n) 24.40+ 4.36a 19.60+ 6.01a 23.50+ 5.60a 17.25+ 6.68a 7.00+ 2.81b

pits (n) 24.85+ 11.18a 33.45+ 15.96a,b 16.20+ 8.12a 41.00+ 22.50a,b 83.30+ 41.93b

small pits (n) 18.30+ 7.99 24.80+ 12.26 12.65+ 7.91 18.55+ 12.77 44.35+ 36.47

large pits (n) 6.05+ 3.81a,b 8.20+ 4.19a,b 3.20+ 1.04a 17.95+ 9.33b 34.65+ 8.81c

very large pits (n) 0.50+ 0.47a 0.45+ 0.21a 0.35+ 0.22a 4.50+ 2.50b 4.30+ 3.08b

thin scratches (n) 20.85+ 3.39a 17.15+ 5.10a,b 22.50+ 5.73a 12.50+ 4.23b,c 5.25+ 2.70c

wide scratches (n) 3.55+ 2.97 2.45+ 1.24 1.00+ 0.59 4.75+ 2.93 1.75+ 0.73

pit length (mm) 18.20+ 2.30 14.17+ 6.83 35.59+ 27.33 23.82+ 11.20 26.86+ 8.42

pit width (mm) 12.32+ 2.17 9.82+ 4.86 20.10+ 12.88 14.47+ 6.91 15.55+ 4.07

scratch length (mm) 441.53+ 115.73a 304.97+ 149.73a,b 448.60+ 82.14a 137.72+ 83.65b 180.62+ 67.80b

scratch width (mm) 9.55+ 3.38 7.57+ 3.67 7.72+ 0.73 10.00+ 5.23 14.09+ 3.92

polishing striations left after chewing (n) 5.20+ 1.55a 7.75+ 2.50a 3.80+ 3.14a,b 0.95+ 1.02b,c 0.25+ 0.56c

Table 2. Statistical evaluation. p-Values from ANOVA and Kruskal – Wallis test show significant differences (less than 0.05*).

measure (unit) log-transformationLevene’s testsp-value

normal distributionof groups

p-values from ANOVA andKruskal – Wallis testkw

scratches (n) no 0.6143 yes 0.0003*

pits (n) yes 0.9847 yes 0.003*

small pits (n) yes 0.5109 yes 0.4913

large pits (n) no 0.1758 yes ,0.0001*

very large pits (n) yes 0.1191 yes ,0.0001*

thin scratches (n) no 0.2094 yes ,0.0001*

wide scratches (n) yes 0.3507 yes 0.1018

pit length (mm) yes 0.0533 no 0.222kw

pit width (mm) yes 0.0997 yes 0.4188

scratch length (mm) no 0.372 yes 0.0002*

scratch width (mm) yes 0.183 no 0.058kw

polishing striations left after chewing (n) no 0.1787 no 0.0009kw*

rsif.royalsocietypublishing.orgJ.R.Soc.Interface

13:20160399

4

not meet normal distribution for each group. We used a p-value of

0.05 as a level of significance.

X-ray microtomography (computed tomography or CT scan-

ning) was used to evaluate the actual amount of wear in one

specimen from each diet group by scanning before and after

the chewing experiments. The equipment used was a custom-

made mCT device nanotomw (PhoenixjXray systems þ Services

GmbH, Wunstorf, Germany). The parameters used in the ima-

ging were as follows: voxel size 37.5 mm, tube voltage of

150 kV and current 80 mA filtered with 0.5 mm of Cu, with 0.58angular step over 3608. Three-dimensional reconstructions were

made using datosjx rec software (PhoenixjXray systems þ Ser-

vices GmbH), thresholding reconstructed volumes and surface

extraction using the software VGStudio MAX 1.2.1 (Volume

Graphics GmbH, Heidelberg, Germany). Alignment, visualiza-

tion and occlusal surface extraction in three dimensions were

made with Rapidform XOS3 (3D Systems Inc., Rock Hill, South

Carolina, USA) [35] and the amount of wear was quantified

in ArcMap 10.0 (ESRI Inc., Redlands, California, USA) by

calculating the mean height for the whole surface in three

dimensions both before and after chewing, and subtracting the

worn surface height from the unworn one.

3. ResultsEvery tooth had both macroscopic and microscopic wear

marks after chewing; means, standard deviations and differ-

ences of pits and scratches can be seen in table 1 and the

results of statistical tests in tables 2 and 3.

Many characters distinguish ATTR from the other diet

groups. It had fewer scratches and more large pits than any

of the other diet groups, and it also had fewer thin scratches

and higher number of very large pits than all non-sandy

diets. ATTR had a significantly higher number of pits and

shorter scratch length than L or GR (tables 1, 2 and 3).

Tabl

e3.

Stat

istica

leva

luat

ionof

diffe

renc

esbe

twee

nall

diet

s.p-

Valu

esfro

mTu

key–

Kram

erHS

Dan

dW

ilcox

onm

etho

dsh

owsig

nific

ant

diffe

renc

es(le

ssth

an0.

05*)

.

mea

sure

(uni

t)L-

GL-

GRL-

GRS

L-AT

TRG-

GRG-

GRS

G-AT

TRGR

-GRS

GR-A

TTR

GRS-

ATTR

scrat

ches

(n)

0.61

070.

9987

0.24

080.

0004

*0.

7679

0.95

30.

0092

*0.

3619

0.00

07*

0.04

22*

pits

(n)

0.92

950.

730.

7011

0.02

37*

0.28

760.

9869

0.11

810.

1242

0.00

15*

0.27

56

small

pits

(n)

——

——

——

——

——

large

pits

(n)

0.98

190.

9503

0.05

01,

0.00

01*

0.71

930.

1426

,0.

0001

*0.

0108

*,

0.00

01*

0.00

36*

very

large

pits

(n)

0.98

130.

9892

0.00

28*

0.01

15*

10.

0005

*0.

0021

*0.

001*

0.00

4*0.

9519

thin

scrat

ches

(n)

0.67

180.

974

0.04

73*

0.00

01*

0.33

220.

4667

0.00

28*

0.01

32*

,0.

0001

*0.

104

wid

esc

ratch

es(n

)—

——

——

——

——

—

pitl

engt

h(m

m)

——

——

——

——

——

pitw

idth

(mm

)—

——

——

——

——

—

scrat

chlen

gth

(mm

)0.

2689

10.

0014

*0.

0061

*0.

2265

0.12

090.

3545

0.00

11*

0.00

48*

0.96

43

scrat

chw

idth

(mm

)—

——

——

——

——

—

polis

hing

striat

ions

lefta

fterc

hew

ing

(n)

0.14

250.

2087

0.01

17*

0.00

95*

0.09

470.

0119

*0.

0097

*0.

0578

0.00

97*

0.23

930

0

25

50

75

100

125

150

5 10 15 20scratches (n)

LGGRGRSATTR

pits

(n)

25 30 35

Figure 3. A bivariate plot presenting average number of pits plotted againstaverage number of scratches in horse cheek teeth used in an artificial mas-tication system using four different diets (L, lucerne; G, grass; GR, grass andrice hulls; GRS, grass, rice hulls and sand) as well as without food (ATTR attri-tion only). Ellipses show 90% CIs. (Online version in colour.)

rsif.royalsocietypublishing.orgJ.R.Soc.Interface

13:20160399

5

GRS could be distinguished from other plant material diet

groups by having a higher number of very large pits and from

L and GR by having shorter scratches and fewer thin scratches.

GRS also had significantly higher number of large pits than

GR. Polishing striations left after chewing were much fewer

in GRS and ATTR than in non-sandy diets (tables 1–3).

There were no statistical differences between the groups

in the number of small pits, number of wide scratches, pit

length, pit width and scratch width.

Non-sandy diets, excluding ATTR, did not show statisti-

cally significant differences between tested characters. In

general, GR had numerically fewer pits in every size category

than L and G, and longer features than G. L and G were very

similar. G tended to have more initial polishing striations left

after chewing than GR. Average numbers of pits and

scratches in every tooth pair are plotted in a bivariate plot

in figure 3.

The results from our CT scanning experiments describing

tooth crown height loss from the whole occlusal surface

(difference of surface mean heights between before and

after chewing) of the single teeth were 8.6 mm for ATTR,

2.7 mm for L, 60.3 mm for G, 66.5 mm for GR and 133.5 mm

for GRS (figure 4).

4. Discussion and conclusionOur approach was to combine analyses on microscopic wear

features and actual dental tissue loss to form a detailed view

of tooth wear through mechanical modelling. The mechanical

masticator is not meant to mimic natural mastication per-

fectly. Rather, it provides a simple and well-defined

mechanical modelling of repetitious chewing events and

allows easy execution of precisely controlled experiments

using different foods and machine settings. Our results

show that the relationships between food properties and

lucerne, Lheight change0.003 mmwear rate per year0.18 mmlifespan (12 cm)658 years

grass, Gheight change0.060 mmwear rate per year4.1 mmlifespan (12 cm)29 years

grass + rice, GRheight change0.067 mmwear rate per year4.5 mmlifespan (12 cm)27 years

GR + sand, GRSheight change0.134 mmwear rate per year9.0 mmlifespan (12 cm)13 years

attrition, ATTRheight change0.008 mmwear rate per year0.58 mmlifespan (12 cm)206 years

wear rateGRS>>GR>G>>ATTR>L

Figure 4. Tooth crown height changes during the chewing experiments. Surface reconstructions from CT scanned teeth, where colour change from yellow to red(light grey to dark grey in greyscale version) describes the mean height loss deviation. All the dietary groups have information about the mean height change, yearlywear rate and lifespan expectation calculated with 12 cm tooth height and 5.4 days spent per 100 000 chews. (Online version in colour.)

rsif.royalsocietypublishing.orgJ.R.Soc.Interface

13:20160399

6

wear patterns partially corroborate expectations derived

from teeth of live animals. Some results, however, are unex-

pected, giving rise to questions about our understanding of

diet–teeth interactions.

In this study, attrition in pure water led to a microwear

signal similar to that seen in browsers: a high number of

pits, but only few scratches [24,36] (figure 3). Also almost all

initial polishing striations were removed. This supports the

hypothesis that occlusal wear facets are caused by attrition

[37] and that non-abrasive diets can be recognized from the

facets caused by attrition-dominated wear [38]. The extent to

which such pure attrition resembles the wear that occurs in

natural chewing cannot be further explored here, but differ-

ences are likely to occur because natural attrition will always

be tempered to some extent by the presence of food.

External grit, especially quartz in soils, has been known to

wear animals’ teeth [1,39,40]. As expected, our results show

that external grit, in this case sand particles with mean grain

size of 233 mm, heavily damages tooth surface, detaches

enamel and thus primarily causes pitting. Material loss was

equally drastic, causing a reduction of tooth surface mean

height of over 130 mm during the 100 000 chewing events.

The fact that sand caused pitting rather than scratching

shows that scratch-dominated wear cannot be explained by

‘grit’ of large grain sizes. Similar results have been obtained

with living sheep, which have been fed with fine-grained

(diameter of 180–250 mm) and medium-grained (diameter of

250–425 mm) sand [25]. It is argued that external dust causes

wear, as open grass areas collect relevant amounts of dust par-

ticles [2]. Our experiment did not deliberately include ‘grit’ of

fine particle size (i.e. dust), although we cannot positively

exclude that fine dust was included in the amount of abrasives

measurable as ADIA. We therefore cannot assess the degree to

which scratches could also be generated by dust.

Phytoliths have long been considered as a major agent of

wear [1,13] and grazing has been considered the main driver

of the evolution of hypsodonty [3]. Our grass and grass–rice

diets, rich in ADIA levels and known to contain phytoliths,

did cause parallel scratches on enamel. That does not necess-

arily mean that phytoliths are the only wear agent, as other

components of the food matrix like plant fibres, or properties

of the food matrix that catch detached enamel particles or

dust and keep them in the ingested bolus as additional abra-

sives, could also be responsible. Amorphous silica, which

was used as an analogue of phytoliths in a recent study [6],

has a role in the formation of microscale scratches. The diet

containing most phytoliths also overwrote the striation from

the initial polishing more quickly than the other non-sandy

diets, suggesting tissue removal. The removal of tissue hypoth-

esis is also supported by other research [41]. To assess whether

the effects observed in microwear represent actual tissue loss,

we additionally quantified the amount of tissue removed

during the chewing experiment in one tooth per diet by CT

scanning. In the groups containing phytoliths, but not grit

(G and GR), the mean tooth height loss was more than

60 mm, which represents real tissue removal and thus strongly

suggests a wear effect of phytoliths.

An unexpected finding of our experiment was that the diet

made from lucerne, which lacks phytoliths, generated a micro-

scopic pattern similar to that generated by grass (figure 3). We

expected the lucerne diet to behave as a browse diet and there-

fore to cause pitted wear, but this did not occur. There are

possible explanations for the lack of correspondence between

the predicted and the observed results: the experiment

set-up changes the signal, or pelleted lucerne is a bad represen-

tative of ‘browse’. Lucerne as used here did not have the

heterogeneous structure of leaf and stem components typical

for browse, as it was ground and mixed with water to form

a homogeneous, pasty material. In our set-up, lucerne does

not behave as it would in real mastication, because of the cush-

ion effect caused by evenly distributed food material; this

creates a barrier between the teeth preventing most of the

1 mm 1 mm 1 mm

1 mm1 mm

(b) (c)(a)

(d) (e)

Figure 5. Images of an enamel band from every diet group and attrition after chewing. Scale bars, 1 mm. (a) Lucerne, (b) grass, (c) grass – rice, (d ) grass – rice –sand and (e) attrition.

rsif.royalsocietypublishing.orgJ.R.Soc.Interface

13:20160399

7

attrition. Natural browsing would probably be something

between our pit-dominated theoretical attrition and the

lucerne diet. Tooth height loss during the chewing exper-

iments with lucerne and theoretical attrition was minimal

compared with the other categories, only less than 10 mm,

being more of a polishing effect than gross wear.

The scratched microwear pattern in the teeth chewed in

the lucerne diet indicates that also something else than phy-

toliths might scratch enamel. The pellet ingredient table [11]

shows that the lucerne diet includes pure lignocellulose

next to other components that contain acid detergent fibre

and acid detergent lignin. Fibres have been suggested pre-

viously to be responsible for polishing dental surfaces [13];

our results also raise the possibility that some plant fibres

might cause scratches. More comparative and experimental

work is needed to explore this possibility.

One major factor often mentioned in microwear analyses

is the question of overwriting and surface turnover. The

microwear pattern is often thought to represent foods eaten

during a few weeks, though in grazers turnover of the surface

can occur in days or hours [42] and in primates even in min-

utes [43]. Unlike other artificial chewing studies, our

experiment had the teeth do about 100 000 chews. Using

data on chewing per gram dry matter, as depending on

diet neutral detergent fibre (NDF) [44] and the daily dry

matter intake linked to a specific diet NDF [45], a 500 kg

horse would reach 100 000 chews within a range of 3.4–5.4

days. Owing to the specific arrangement of initial polishing

striations to the chewing direction in the machine, we could

also visualize and calculate how many of these initial polish-

ing scratches were still observable on the 0.4 � 0.4 mm2 area

after 100 000 chews.

The mean number of polishing striations left after chewing

shows that the grass and lucerne diets caused less, and

the grass–rice–sand diet more overwriting than the most

phytolith-containing grass–rice diet. Rabbit and guinea pig

studies with the same diets [11,12] also showed that absolute

wear rate increased from lucerne to grass to grass–rice and

to grass–rice–sand. This was also evident in our CT scan

data. Similar results can be seen in ruminants; browsers have

lower wear rates than grazers, and mixed-feeders are in

between [42,46]. Teeth in attrition chewing had the lowest

number of initial polishing scratches left after chewing; this is

caused most probably by the attritional polishing effect that

notably leads to little actual tissue loss. The more wear food

particles cause, the quicker they overwrite existing microwear

patterns. This will lead to problems if the diet is reconstructed

from microwear analysis only, as already a short-term shift can

change the whole microwear pattern very quickly. This is very

clear when one compares microwear results and the amount of

actual wear quantified in our CT scans.

For the sake of future discussion, we present a possible

analysis methodology for microscopic wear features in the

electronic supplementary material, S1, with non-parametric

density plots of all, not classified, individual features described

by pure feature lengths and widths analysed by modal cluster-

ing based on density contours. This methodology would

resemble texture analysis [10,21] but is made using light

stereo microscopy with image and statistical analyses.

Internal structures of teeth, i.e. the arrangement of enamel,

dentine and cement, affect how teeth are worn down. Differen-

tial wear of dental tissues carries a dietary signal that has so far

been little explored. Most studies focus on enamel wear,

though in many herbivores wear of all three dental tissues

together is important. For example, as dentine and cement

wear away, they leave the brittle enamel exposed to fracturing.

Although microwear was studied quantitatively on enamel

only, we observed wear of dentine and cement. Dentine and

cement were worn more than enamel by diets containing abra-

sives (figures 4 and 5b–d). This is evident because enamel

bands begin to emerge above their surrounding area. By con-

trast, lucerne and attrition diets (figures 4 and 5a,e) caused

very light wear on all tissues and with three tissues wearing

at very similar rates, a feature typically observed on natural

wear facets (figure 5e).

Our microwear results demonstrate that all diets studied,

including pure attrition, caused microwear features (micro-

scopic wear features) to the tooth surface. Diet-specific

microwear patterns between lucerne, grass and grass–rice

could not be distinguished statistically with the characters

rsif.royalsocietypublishing.orgJ.R.Soc.Interface

13:20160399

8

measured, though in general, grass–rice, which contained

higher amount of phytoliths, left longer features and fewer

pits than grass. As seen in nature [24,36], our grass-based

diets generated lots of fine and long scratches on enamel.

Attrition and sandy diet could readily be distinguished

from the other diets by their microwear pattern. A surprising

result was that the lucerne diet, lacking both phytoliths and

grit, caused a striated wear surface similar to that caused

by grass.

In combination with our measurements on actual tooth

wear, these microwear findings highlight that differences in

actual dental tissue loss can occur at similar microwear pat-

terns, as in lucerne versus grass diet, cautioning against a

direct transformation of microwear results into predictions

about diet or tooth wear rate. If similar microscopic wear pat-

terns occur at different absolute dental tissue loss levels, this

indicates differences in microwear signal accumulation rates.

The wear rate on lucerne was 50 times lower than that on

grass–rice–sand, leading to more microwear accumulation

and partly polishing.

The diet containing most abrasives including grit, grass–

rice–sand, had twice the wear rate of grass and grass–rice

diets, and over 10 times the wear rate of lucerne and attrition.

For a horse with a theoretical molar crown height of 12 cm,

with the food intake and chewing frequency of 100 000 per

5.4 days as mentioned before, this would lead to maximum

expected lifespans (fully worn down teeth) of 13 years for

grass–rice–sand compared with 27 years for grass–rice, and

29 years for grass (figure 4). The maximum lifespan for a dom-

estic horse is approximately 25–30 years, similar to the grass

and grass–rice diets, and the maximum lifespan for a zebra

in the wild is approximately 16–18 years [47–49], similar to

the diet with heavy grit load, grass–rice–sand. Zebras in cap-

tivity, with a presumably distinctively lower grit load in their

diets [50], have about the same lifespan as domestic horses

[47–49]. In attrition, a horse molar would wear down

in about 200 years, and even more extremely on lucerne, in

more than 600 years. These ‘browsing’ diets with lucerne

and pure attrition have wear rates that explain why certain

animals can live with very brachydont teeth.

To conclude, simplified mechanical modelling simulates

mastication and tooth wear with plausible rates. Clear differ-

ences in dietary components and their corresponding wear

rates were not always distinguishable by microscopic wear

features. Plant material, phytoliths, grit and teeth are all

agents of tooth wear.

Data accessibility. The datasets supporting this article have beenuploaded as part of the electronic supplementary material. Morespecific data can be obtained from Aleksis Karme and JaninaRannikko.

Authors’ contributions. Al.K. designed the methods and Al.K. and Ak.K.the equipment. Al.K., M.F. and M.C. designed the experiment. J.R.and Al.K. ran the experiment and performed the analyses. Ak.K.and Al.K. performed the CT scans. J.R. and Al.K. analysed theresults. J.R., Al.K., M.F., M.C. and Ak.K. prepared the manuscript.M.F. and Al.K. first had the idea of building a simple chewingmachine. All authors gave final approval for publication.

Competing interests. We declare we have no competing interests.

Funding. Al.K. had funding from the Finnish Cultural Foundation.

Acknowledgements. We would like to thank the Finnish Cultural Foun-dation, Doctoral Program in Geosciences and the Academy ofFinland for funding the authors. We appreciate the constructive com-ments from the referees that helped us to elevate the level of ourarticle. Authors would also like to acknowledge Thomas Bertinwho did some preliminary experiments with the chewing machine,Szabolcs Galambosi for assistance in the design, Vainion TeurastamoOy for providing the horse skulls, Helena Korkka for facilitatingtooth preparation and Jukka Ukkonen for manufacturing repairparts for the chewing machine. A special acknowledgement belongsto Pauli Engstrom, who was working with the design and assemblyof the chewing machine, contributing in a very significant way.

References

1. Baker G, Jones LHP, Wardrop ID. 1959 Cause ofwear in sheeps’ teeth. Nature 184, 1583 – 1584.(doi:10.1038/1841583b0)

2. Sanson GD, Kerr SA, Gross KA. 2007 Do silicaphytoliths really wear mammalian teeth?J. Archaeol. Sci. 34, 526 – 531. (doi:10.1016/j.jas.2006.06.009)

3. Damuth J, Janis CM. 2011 On the relationship betweenhypsodonty and feeding ecology in ungulatemammals, and its utility in palaeoecology. Biol. Rev.86, 733 – 758. (doi:10.1111/j.1469-185X.2011.00176.x)

4. Lucas PW et al. 2013 Mechanisms and causes ofwear in tooth enamel: implications for hominindiets. J. R. Soc. Interface 10, 20120923. (doi:10.1098/rsif.2012.0923)

5. Lucas PW et al. 2014 The role of dust, grit andphytoliths in tooth wear. Ann. Zool. Fenn. 51, 143 –152. (doi:10.5735/086.051.0215)

6. Xia J, Zheng J, Huang D, Tian ZR, Chen L, Zhou Z,Ungar PS, Qian L. 2015 New model to explain toothwear with implications for microwear formation anddiet reconstruction. Proc. Natl Acad. Sci. USA 112,10 669 – 10 672. (doi:10.1073/pnas.1509491112)

7. Maas MC. 1991 Enamel structure and microwear: anexperimental study of the response of enamel toshearing force. Am. J. Phys. Anthropol. 85, 31 – 49.(doi:10.1002/ajpa.1330850106)

8. Maas MC. 1994 A scanning electron-microscopicstudy of in vitro abrasion of mammalian toothenamel under compressive loads. Arch. Oral Biol.39, 1 – 11. (doi:10.1016/0003-9969(94)90028-0)

9. Danielson DR, Reinhard KJ. 1998 Human dentalmicrowear caused by calcium oxalate phytoliths inprehistoric diet of the lower Pecos region, Texas.Am. J. Phys. Anthropol. 107, 297 – 304. (doi:10.1002/(SICI)1096-8644(199811)107:3,297::AID-AJPA6.3.0.CO;2-M)

10. Schulz E, Piotrowsk V, Clauss M, Mau M, MerceronG, Kaiser TM. 2013 Dietary abrasiveness isassociated with variability of microwear and dentalsurface texture in rabbits. PLoS ONE 8, e56167.(doi:10.1371/journal.pone.0056167)

11. Muller J, Clauss M, Codron D, Schulz E, Hummel J,Fortelius M, Kircher P, Hatt J-M. 2014 Growth andwear of incisor and cheek teeth in domestic rabbits(Oryctolagus cuniculus) fed diets of different

abrasiveness. J. Exp. Zool. 321, 283 – 298. (doi:10.1002/jez.1864)

12. Muller J, Clauss M, Codron D, Schulz E, Hummel J,Kircher P, Hatt J-M. 2015 Tooth length and incisalwear and growth in guinea pigs (Cavia porcellus)fed diets of different abrasiveness. J. Anim.Physiol. Anim. Nutr. 99, 591 – 604. (doi:10.1111/jpn.12226)

13. Hoffman JM, Fraser D, Clementz MT. 2015Controlled feeding trials with ungulates: a newapplication of in vivo dental molding to assess theabrasive factors of microwear. J. Exp. Biol. 218,1538 – 1547. (doi:10.1242/jeb.118406)

14. Hua L-C, Brandt ET, Meullenet J-F, Zhou1 Z-R, UngarPS. 2015 Technical note: an in vitro study of dentalmicrowear formation using the BITE Master IIchewing machine. Am. J. Phys. Anthropol. 158,769 – 775. (doi:10.1002/ajpa.22823)

15. Walker A, Hoeck HN, Perez L. 1978 Microwear ofmammalian teeth as an indicator of diet. Science201, 908 – 910. (doi:10.1126/science.684415)

16. Rensberger JM. 1978 Scanning electron microscopyof wear and occlusal events in some small

rsif.royalsocietypublishing.orgJ.R.Soc.Interface

13:20160399

9

herbivores. In Development, function and evolutionof teeth (eds PM Butler, KA Joysey), pp. 415 – 438.New York, NY: Academic Press.

17. Solounias N, Teaford M, Walker A. 1988 Interpretingthe diet of extinct ruminants: the case of a non-browsing giraffid. Paleobiology 14, 287 – 300.(doi:10.1017/S009483730001201X)

18. Merceron G, Viriot L, Blondel C. 2004 Toothmicrowear pattern in roe deer (Capreolus capreolusL.) from Chize (Western France) and relation to foodcomposition. Small Ruminant Res. 53, 125 – 132.(doi:10.1016/j.smallrumres.2003.10.002)

19. Rivals F, Takatsuki R, Albert RM, Macia L. 2014Bamboo feeding and tooth wear of three sika deer(Cervus nippon) populations from northern Japan.J. Mamm. 95, 1043 – 1053. (doi:10.1644/14-MAMM-A-097)

20. Ryan AS. 1981 Anterior dental microwear and itsrelationship to diet and feeding behavior in threeAfrican primates (Pan troglodytes troglodytes, Gorillagorilla gorilla, and Papio hamadryas). Primates 22,533 – 550. (doi:10.1007/BF02381245)

21. Ungar PS, Grine FE, Teaford MF. 2008 Dentalmicrowear and diet of the Plio-Pleistocene homininParanthropus boisei. PLoS ONE 3, e2044. (doi:10.1371/annotation/195120f0-18ee-4730-9bd6-0d6effd68fcf )

22. Gomes Rodrigues H, Merceron G, Viriot L. 2009Dental microwear patterns of extant and extinctMuridae (Rodentia, Mammalia): ecologicalimplications. Naturwissenschaften 96, 537 – 542.(doi:10.1007/s00114-008-0501-x)

23. Whitlock JA. 2011 Inferences of Diplodocoid(Sauropoda: Dinosauria) feeding behavior fromsnout shape and microwear analyses. PLoS ONE 6,e18304. (doi:10.1371/journal.pone.0018304)

24. Solounias N, Sembrebon GM. 2002 Advances in thereconstruction of ungulate ecomorphology withapplication to early fossil equids. Am. MuseumNovitates 3366, 1 – 49. (doi:10.1206/0003-0082(2002)366,0001:aitrou.2.0.co;231)

25. Teaford MF, Walker AC. 1984 Quantitative differencesin dental microwear between primate species withdifferent diets and a comment on the presumeddiet of Sivapithecus. Am. J. Phys. Anthropol. 64,191 – 200. (doi:10.1002/ajpa.1330640213)

26. Ungar PS. 1996 Dental microwear of EuropeanMiocene catarrhines: evidence for diets and toothuse. J. Hum. Evol. 31, 335 – 366. (doi:10.1006/jhev.1996.0065)

27. Karme A, Kallonen A, Galambosi S, Engstrom P,Fortelius M. 2012 Artificial chewing with real teeth.72nd Annual Meeting of the Society of VertebratePaleontology, Raleigh, NC, USA, 17 – 20 October 2012.

28. Bertin T, Karme A, Fortelius M. 2013 Experimentalstudy of dental microwear using a mechanicalmasticator. (M1 internship report), Ecole NormaleSuperieure de Lyon/University of Helsinki.

29. Wohlbier W. 1983 Handelsfuttermittel. Band 2. TeilA Futtermittel pflanzlicher Herkunft. Stuttgart,Germany: Eugen Ulmer.

30. Epstein E. 1994 The anomaly of silicon in plantbiology. Proc. Natl Acad. Sci. USA 91, 11 – 17.(doi:10.1073/pnas.91.1.11)

31. Bonin SJ, Clayton HM, Lanovaz JL, Johnston T. 2007Comparison of mandibular motion in horseschewing hay and pellets. Equine Vet. J. 39, 258 –262. (doi:10.2746/042516407X157792)

32. Merceron G, Blondel C, De Bonis L, Koufos DG, ViriotL. 2005 A new method of dental microwearanalysis: application to extant primates andOuranopithecus macedoniensis (Late Miocene ofGreece). Palaios 20, 551 – 561. (doi:10.2110/palo.2004.p04-17)

33. Mihlbachler MC, Beatty BL, Caldera-Siu A, Chan D,Lee R. 2012 Error rates and observer bias in dentalmicrowear analysis using light microscopy.Palaeontol. Electron. 15, 12A. ( palaeo-electronica.org/content/2012-issue-1-articles/195-microwear-observer-error)

34. Merceron G, Costeur L, Maridet O, Ramdarshan A,Gohlich UB. 2012 Multi-proxy approach detectsheterogeneous habitats for primates during theMiocene climatic optimum in Central Europe.J. Hum. Evol. 63, 150 – 161. (doi:10.1016/j.jhevol.2012.04.006)

35. Karme A, Kallonen A, Matilainen VP, Piili H,Salminen A. 2015 Possibilities of CT scanning asanalysis method in laser additive manufacturing.Phys. Proc. 78, 347 – 356. (doi:10.1016/j.phpro.2015.11.049)

36. Fraser D, Mallon JC, Furr R, Theodor JM. 2009Improving the repeatability of low magnificationmicrowear methods using high dynamic rangeimaging. Palaios 24, 818 – 825. (doi:10.2110/palo.2009.p09-064r)

37. Butler P. 1952 The milk molars of thePerissodactyla, with remarks on molar occlusion.Proc. Zool. Soc. Lond. 121, 777 – 817. (doi:10.1111/j.1096-3642.1952.tb00784.x)

38. Fortelius M, Solounias N. 2000 Functionalcharacterization of ungulate molars using theabrasion-attrition wear gradient: a new method forreconstructing palaeodiets. Am. Museum Noviates3301, 1 – 36. (doi:10.1206/0003-0082(2000)301,0001:FCOUMU.2.0.CO;2)

39. Healy WB, Ludwig TG. 1965 Wear of sheep’s teeth I:the role of ingested soil. N. Z. J. Agric. Res. 8, 737 –752. (doi:10.1080/00288233.1965.10423710)

40. Ludwig TG, Healy WB, Cutress TW. 1966 Wear ofsheep’s teeth III: seasonal variation in wear andingested soil. N. Z. J. Agric. Res. 9, 157 – 164.(doi:10.1080/00288233.1966.10420770)

41. Rabenold D, Pearson OM. 2014 Scratching thesurface: a critique of Lucas et al. (2013)’s conclusionthat phytoliths do not abrade enamel. J. Hum. Evol.74, 130 – 133. (doi:10.1016/j.jhevol.2014.02.001)

42. Solounias N, Fortelius M, Freeman P. 1994 Molarwear rates in ruminants: a new approach. Ann. Zool.Fenn. 31, 219 – 227.

43. Teaford MF, Oyen OJ. 1989 In vivo and in vitroturnover in dental microwear. Am. J. Phys.Anthropol. 80, 447 – 460. (doi:10.1002/ajpa.1330800405)

44. Janis CM, Constable EC, Houpt KA, Streich WJ,Clauss M. 2010 Comparative ingestive mastication indomestic horses and cattle: a pilot investigation.J. Anim. Physiol. Anim. Nutr. 94, e402 – e409.(doi:10.1111/j.1439-0396.2010.01030.x)

45. Meyer K, Hummel J, Clauss M. 2010 Therelationship between forage cell wall content andvoluntary food intake in mammalian herbivores.Mamm. Rev. 40, 221 – 245. (doi:10.1111/j.1365-2907.2010.00161.x)

46. Damuth J, Janis CM. 2014 A comparison of observedmolar wear rates in extant herbivorous mammals. Ann.Zool. Fenn. 51, 188 – 200. (doi:10.5735/086.051.0219)

47. Jones ML. 1977 Longevity of mammals in captivity.Int. Zoo News 159, 16 – 19.

48. Spinage CA. 1972 African ungulate life tables.Ecology 53, 645 – 652. (doi:10.2307/1934778)

49. Smuts GL. 1974 Age determination in Burchell’szebra (Equus burchelli antiquorum) from the KrugerNational Park. J. S. Afr. Wildl. Manage. Assoc. 4,103 – 115.

50. Taylor LA, Muller DWH, Schwitzer C, Kaiser TM,Castell JC, Clauss M, Schulz-Kornas E. 2016Comparative analyses of tooth wear in free-rangingand captive wild equids. Equine Vet. J. 48,240 – 245. (doi:10.1111/evj.12408)