Wild or Domestic? Biometric Variation in the Cat Felis silvestris Schreber

International Journal of OsteoarchaeologyInt. J. Osteoarchaeol. 17: 581–595 (2007)Published online 12 April 2007 in Wiley InterScience

13
(www.interscience.wiley.com) DOI: 10.1002/oa.9
* Correspondence to: DeparYork, Kings Manor, York YOe-mail: [email protected]

Copyright # 2007 Joh

Wild or Domestic? BiometricVariation in the Cat Felis silvestrisSchreber

T. P. O’CONNOR*

Department of Archaeology, University of York, Kings Manor, York YO1 7EP, UK

ABSTRACT Investigation of modern biometric data indicates that it may be possible to distinguish wildcatsfrom house cats in many instances. Applying the log-ratio (log-difference) technique toarchaeological samples from medieval northern Europe, and to mixed samples of wildcatsand house cats, shows that the differentiation may not always be clear, and the possibility thatsome samples include hybrids is discussed. The technique is applied to samples from theOrkney Islands to demonstrate that single wildcat specimens can be identified in smallsamples. Copyright � 2007 John Wiley & Sons, Ltd.

Key words: w ildcat; domestic cat; Felis silvestris; size-index scaling; medieval Europe

Introduction

Cats are one of the most widespread and familiaranimals of the domestic realm, and theirassociation with people extends back at leastinto the Neolithic period (Davis, 1989; Vigneet al., 2004). Opinions regarding the originaldomestication of cats range from the view thatthey were deliberately domesticated by people asuseful predators of household vermin (e.g.Zeuner, 1963; Kratochvil & Kratochvil, 1976;Clutton-Brock, 1999) to the behavioural modelproposed by Todd (1978), which has catsadopting people as a useful source of food, andbeing first tolerated then encouraged and fed bypeople. Serpell (1988, 1996) has made the pointthat people seem to find benefit in thecompanionship of other species, quite apart fromany resource value those animals may have, andthat the furry, purry amiability of cats fits them

tment of Archaeology, University of1 7EP, UK.

n Wiley & Sons, Ltd.

particularly well to the role of companion animal.That being the case, it is remarkable how littleresearch has tackled the archaeology of therelationship between people and cats, comparedwith, for example, the copious literature oncaprines or dogs (e.g. Vigne et al., 2005;Crockford, 2000)One factor in this relative neglect is that any

systematic study would have to differentiatebetween the bones of wild and domestic cats. Thelatter term itself may be problematic. A ‘specia-tion’ model of animal domestication, such asunderlies the attribution of Linnaean binomials todomestic forms (Gentry et al., 2004), requiresseparation of the breeding populations of thedomestic animals and their wild precursors;indeed, this is one of the defining aspects ofanimal domestication. However, domestic andferal cats are conspecific and sympatric with aparticularly widespread and polytypic species,Felis silvestris, the modern range of which extendsthroughout Europe into western Asia and throughthe entire length of Africa (Sunquist & Sunquist,2002). Breeding between feral cats and wildcats isa major conservation problem today in parts of

Received 18 August 2006Revised 11 December 2006Accepted 11 December 2006

582 T. P. O’Connor

Europe (Daniels et al., 2001; Pierpaoli et al.,2003; Yamaguchi et al., 2004; Kitchener et al.,2005; Lecis et al., 2006). It is a reasonablepresumption that the genetic boundary betweenthe two ecomorphs has always been as blurred asthe behavioural boundary. Given that, and inorder to avoid the quite specific connotations of‘domesticated’, this paper uses the term ‘housecats’ to refer to the generally smaller ecomorphthat associates itself with human settlements,without any presumptions about the phyloge-netic relationship between this form and free-living wildcats. The first point to make aboutthe zooarchaeological data is that it may beunrealistic to expect a clear distinction betweenwildcat and house cat bones because ofthe probability that hybrid forms existedwherever and whenever the two ecomorphs weresympatric.The investigation reported here forms part of a

larger study of the archaeology of cats. Theattribution of cat bones to house cat or wildcat isat the heart of investigating their zooarchaeol-ogy, and this paper limits itself to investigatingthe feasibility of making such attributions withconfidence by the use of conventional biometrictechniques, and to investigating inter- andintra-sample variability in ancient and modernsamples of cat post-cranial bones. WithinEuropean zooarchaeology, it has been theconvention to attempt differentiation of wild-and house cats by demonstrating the smaller sizeof the latter ecomorph (e.g. Harcourt, 1979: 154;Johansson & Huster, 1987: 44–7). Whether thisapparent size difference derives from a markeddifference in selection pressures in synanthropicpopulations, or reflects the origin of house cats inthe more lightly-built African form of Felis silvestris(formerly F. lybica; see Clutton-Brock, 1999), or acombination of both founder effect and severalmillennia of drift, is a debatable point. For thepresent purposes, it serves as a starting point forinvestigation.

Materials and methods

Cat bones typically occur at rather low frequen-cies in zooarchaeological assemblages: it isunusual for them to constitute more than 1%

Copyright # 2007 John Wiley & Sons, Ltd.

of an assemblage. Accordingly, it is exceptionalfor any one element of the cat skeleton to yield auseful biometric sample, making it necessary tocombine data from as many elements as ispracticable. A technique that allows just such acombination of data is the use of log-ratio orlog-difference values (Meadow, 1999). Bond &O’Connor (1999) used this approach to inves-tigate size variation in cat bones from medievalYork, using a single cat skeleton as the ‘standard’measure against which to rescale the archae-ological data. It is generally better practice to usemean values from a number of skeletons as thestandard than to use a single individual whichmay be atypically proportioned. The largestreadily-accessible samples of biometric datapertinent to this paper are those published byKratochvil (1973, 1976a). The former studyinvestigated size and morphological variationwithin and between samples of 51 wildcat and60 house cat skulls, whilst the latter studied thepostcranial skeletons of 31 wildcats and 76 housecats. In both studies, the wildcats were frompopulations in the western Carpathians and thehouse cats from the Brno region, then inCzechoslovakia. Kratochvil’s aim was to describeand quantify consistent differences between thewildcat and house cat samples, testing andvalidating criteria proposed by previous authorssuch as Rohrs (1955) and Schauenberg (1969),and to measure the degree of intra-populationvariability of different characteristics. This latterpoint is important in a zooarchaeological contextas Kratochvil’s data on variability may help toresolve whether a given archaeological samplerepresents a rather variable sample of one orother ecomorph or a mixture of the two. The dataalso show that sexual dimorphism is negligible inthe postcranial skeleton and in the cranialskeleton of house cats, but measurable in thecranial skeleton of wildcats (Kratochvil, 1976b).Again, this is a valuable observation when tryingto explain variation in zooarchaeologicalsamples. Given the lack of sexual dimorphismin postcranial elements, this present study usesKratochvil’s male and female house cat datacombined to constitute a ‘standard’ datasetagainst which to compare archaeological datausing the log-ratio (log-difference) technique(Meadow, 1999).

Int. J. Osteoarchaeol. 17: 581–595 (2007)DOI: 10.1002/oa

Biometric Variation in the Cat 583

Log-ratio values were calculated using thealgorithm:

Log-ratio ¼ log10ðobserved=standardÞ (1)

The house cat data were used as a standard, inpreference to the wildcat data, because thesample size is appreciably larger for Kratochvil’shouse cat sample, and because a preliminary scanof archaeological data indicated that the greatmajority of specimens were closer in size to thehouse cat size range than to the wildcat range.The modern data are summarised in Table 1,which is derived from Kratochvil (1976a).

Table 1. Sample means, standard deviation, and number oKratochvil (1976a), with codes following von den Driesch (1

House cat mean SD n

Atlas GL 17.1 1.13 61Atlas BFcr 22.00 1.29 61Atlas BFcd 15.55 1.00 61Atlas GB 34.16 2.77 61Axis LCDe 23.78 1.81 57Axis SBV 11.45 0.89 61Scapula DHA 63.25 4.39 61Scapula HS 68.19 4.61 61Scapula SLC 12.02 0.80 61Scapula GLP 13.7 0.95 61Scapula LG 11.48 0.77 61Scapula BG 9.03 0.65 61Humerus GL 96.46 4.89 62Humerus Dp 20.32 1.31 62Humerus SD 6.64 0.68 62Humerus Bd 17.91 1.16 62Radius GL 92.17 4.86 61Radius Bp 8.05 0.62 63Radius SD 5.24 0.44 62Radius Bd 12.51 0.79 61Ulna GL 108.9 5.65 63Ulna DPA 11.08 1.03 63Ulna BPC 8.74 0.7 63Sacrum GL 25.18 1.66 58Sacrum GB 28.3 1.91 57Pelvis GL 43.59 2.59 63Pelvis LAR 10.96 0.81 63Pelvis SH 10.9 0.9 63Femur GLC 105.58 5.98 63Femur Bp 20.1 1.23 63Femur DC 9.74 0.6 63Femur SD 8.28 0.8 63Femur Bd 18.42 1.15 63Tibia GL 111.32 5.65 63Tibia Bp 19.35 1.22 63Tibia SD 7.21 0.68 63Tibia Dd 9.4 0.58 63Fibula GL 103.72 5.63 62Astragalus GH 15.81 0.79 62Calcaneus GL 29.22 1.78 62


Some of Kratochvil’s measurements are notused here, either because they are only rarelyavailable or recorded in zooarchaeologicalmaterial, or because of ambiguity in equatinghis measurements with those defined andillustrated by von den Driesch (1976) and usedby most zooarchaeologists.Further modern data were acquired from house

cats and wildcats in the collections of the authorand the University of York (Table 2). The housecats are fully documented as to sex and age atdeath. The wildcats are not fully documented, butwere stated on acquisition to have originated as‘road-kill’ specimens in highland Scotland, and to

f cases for house cat and wildcat samples derived from976). All measurements in mm

Wild cat mean SD n Log(wild/house)

19.46 1.19 19 0.05625.44 1.09 19 0.06317.79 1.07 19 0.05838.91 2.78 18 0.05727.83 1.43 19 0.06812.57 0.72 19 0.04180.88 4.6 19 0.10785.96 5.04 19 0.10114.13 0.89 19 0.07016.74 1.08 19 0.08714.5 1.06 19 0.10111.08 0.71 19 0.089

119.08 6.17 19 0.09124.66 1.85 19 0.0848.04 0.52 19 0.083

22.18 1.63 19 0.093115.96 5.14 19 0.10010.16 0.6 19 0.1016.3 0.5 19 0.080

15.39 1.08 19 0.090134.47 6.25 19 0.09213.16 0.95 19 0.07511.16 0.83 19 0.10630.09 2.15 15 0.07732.28 1.88 18 0.05752.7 2.49 20 0.08212.92 0.79 20 0.07113.18 0.94 20 0.082

132.7 6.77 19 0.09924.03 1.58 19 0.07811.51 0.78 19 0.0739.55 0.57 19 0.062

21.87 1.43 19 0.075140.19 5.82 19 0.10023.18 1.46 19 0.0788.4 0.7 19 0.066

10.97 0.8 19 0.067131.04 5.17 18 0.10217.82 1.57 19 0.05234.44 1.6 19 0.071


Table 2. Length measurements and corresponding log-ratio values for wild- and house cat specimens fromUniversity of York (EAU numbers) and author’s referencecollections

GL(C) GLlog

EAU776 Wild Fibula 122.2 0.071EAU776 Wild Tibia 130.6 0.069EAU776 Wild Radius 107.0 0.065EAU776 Wild Femur 122.0 0.063EAU776 Wild Ulna 124.0 0.056EAU776 Wild Humerus 109.0 0.053EAU777 Wild Tibia 122.7 0.042EAU777 Wild Radius 101.0 0.040EAU777 Wild Ulna 118.7 0.037EAU777 Wild Femur 113.3 0.031EAU783 Wild Radius 98.3 0.028EAU783 Wild Fibula 109.6 0.024EAU783 Wild Tibia 117.2 0.022EAU783 Wild Femur 111.0 0.022EAU783 Wild Humerus 101.4 0.022Hobbit Humerus 100.0 0.016Hobbit Radius 95.0 0.013Hobbit Ulna 112.2 0.013Hobbit Femur 108.5 0.012Hobbit Tibia 114.0 0.010Hobbit Fibula 105.7 0.008Juno Humerus 95.9 �0.003Juno Radius 91.4 �0.004Juno Tibia 109.0 �0.009Juno Fibula 101.5 �0.009Juno Ulna 106.0 �0.012Vicki Tibia 108.1 �0.013Vicki Radius 89.4 �0.013Vicki Fibula 100.4 �0.014Vicki Ulna 105.3 �0.015Juno Femur 102.0 �0.015Vicki Femur 101.3 �0.018Vicki Humerus 92.5 �0.018EAU129 Humerus 89.4 �0.033EAU129 Tibia 102.8 �0.035EAU129 Ulna 100.2 �0.036EAU129 Fibula 95.3 �0.037EAU129 Radius 83.9 �0.041Omega Ulna 97.7 �0.047EAU129 Femur 94.4 �0.050Omega Radius 82.0 �0.051Omega Fibula 91.7 �0.053Omega Tibia 98.0 �0.055Omega Humerus 84.3 �0.059Omega Femur 91.7 �0.061

Specimens designated ‘wild’ were attributed on pelagecharacters. The specimens have been sorted by size. Ofthe house cats, Hobbit was an intact male; the others wereall female.


have pelage characteristics of ‘wild’ F. silvestris.Given the uncertainty that surrounds the diag-nosis of wild- and house cats in regions of closesympatry (Kitchener et al., 2005), it is possible


that one or more of these cats is a hybrid (seediscussion of the data below). None the less,these data comprise a sample of modern cats frommainland Britain that includes behaviourally wildand domestic individuals. All measurements weretaken using sliding callipers to a precision of0.1mm, following conventional zooarchaeologi-cal procedures.

Archaeological datasets were acquired fromthe author’s own records and from published data.Data from sites in York and Lincoln, recorded bythe author between 1979 and 1999, have beenused to assemble three substantial post-Romansamples:

� Y
orkVik – York Viking Age (or Anglo-Scandinavian), 10th to early 11th century AD.Specimens from 16–22 Coppergate (O’Connor,1989).
� Y
orkMed – York High Medieval, 12–13th
century. Specimens from 16–22 Coppergateand The Bedern (Bond & O’Connor, 1999).

� L
incVik – Lincoln Anglo-Scandinavian, 10th toearly 11th century. Specimens from Flaxengatetimber building phases (O’Connor, 1982).
Further data were acquired from the broadly‘Viking Age’ (9–11th century) assemblage fromHaithabu, Germany (modern Hedeby; Johansson& Huster, 1987) and from the medieval (12–14th

century) assemblage from Schleswig, Germany(Spahn, 1986). These two examples were chosenbecause the exceptionally large size of theassemblages allows investigation of intra-samplevariation.

Further data were acquired from a post-medieval assemblage from Kilton Castle, in thenorth of England (author’s unpublished data; seealso O’Connor, 1987). These cats were part of amixed assemblage of cats and rodents recoveredfrom the lowest deposits in a well, and apparentlyderived from 16th century ‘clearing up’ of thecastle. The cats are included here as examinationof the morphology of cat crania in thisassemblage indicates the presence of wildcatsand house cats, making the size range andvariability of the post-cranial bones particularlyrelevant to this paper.


Table 3. Log-ratio values for minimum, maximum and95% Confidence Interval values from Kratochvil (1976a)for (a) house cats; (b) wildcats

Min. Max. 95% CI

(a)Tibia GL �0.058 0.064 �0.006Radius GL �0.060 0.056 �0.006Humerus Bd �0.054 0.095 �0.007Femur Bd �0.050 0.091 �0.007(b)Tibia GL �0.028 0.032 �0.008Radius GL �0.035 0.032 �0.009Humerus Bd �0.049 0.029 �0.015Femur Bd �0.054 0.035 �0.013


Results

To start with an internal comparison, Kratochvil’swildcat sample means were re-expressed aslog-ratio values against the house cat standard(Table 1). Unremarkably, all values were positive,indicating the wildcats to be larger on allmeasured parameters. The log-ratio values weregenerally lowest for measurements taken on theatlas, axis, pelvis and tarsals; that is, the differenceis most marked in bones of the limbs, rangingfrom 0.062 (femur SD) to 0.107 (scapula HS).From the limb bones, length measurementsgenerally gave larger log-ratio values thanbreadth measurements. Lengths range from0.091 (humerus GL) to 0.102 (fibula GL).Disregarding the length measurements, theforelimb generally gave higher values than thehindlimb: forelimb breadth values range from0.080 (radius SD) to 0.101 (radius Bp), andhindlimb values from 0.062 (femur SD) to 0.078(femur Bp, tibia Bp). To summarise, comparisonsof length and breadth measurements should bemade separately. This conclusion has beenreached in log-ratio based biometric studies ofother species, notably sheep (Davis, 1996). Wemay expect limb bones, especially lengthmeasurements, to show the clearest differen-tiation of wild and house cats, and hindlimbbreadth measurements to show the least differ-entiation, although even these measurementsshow the wildcat sample to be significantly largerthan the house cats (t-test analyses on untrans-formed data are given in full in Kratochvil,1976a). Although the comparison data, and allarchaeological data, have been derived from‘adult’ (i.e. fully fused) limb elements, thepossibility of some post-fusion growth cannotbe wholly discounted. This should be mostevident in the diaphysis, and hence in SDmeasurements, which should not be taken asdiagnostic without supporting evidence fromepiphysial or length measurements. Kratochvil(1976a) noted in passing that diaphysial breadthmeasurements showed the highest coefficients ofvariation in his datasets.In order to compare the scale of those log-ratio

values with the degree of variability of the housecat sample, four measured variables were inves-tigated in more detail: tibia GL, radius GL,


humerus Bd and femur Bd. These four werechosen to give a mix of fore- and hindlimbelements, and of lengths and breadths. Theminimum and maximum values given byKratochvil for each of the four variables wereconverted to log-ratio values based on the housecat mean values, thus showing the highest andlowest log-ratio values that the house cat samplewould have given had the specimens been plottedindividually. The results are shown in Table 3.The values for range show that the largestindividual house cat specimen would typicallyhave fallen into the wildcat range, althoughsomewhat below the mean. The particularly highmaximum values for humerus and femur Bd reflectexceptionally large values for one male house cat.The much narrower confidence limits of thehouse cat sample show that this sample typicallyhas quite tightly-clustered data with a few outliersreflected by the values for the range. Apart fromthe humerus and femur Bd values, the data wouldlead us to expect a zooarchaeological sample ofhouse cats to include the odd one or two givingvalues up to �0.06, with the great majority wellinside this limit. Differences in sample meansbetween wildcat and house cat samples in Table 1range from 0.062 (femur SD) to 0.107 (scapulaHS; see above), appreciably greater in scale thanthe confidence limits calculated here. Thisindicates that it should be possible to differentiatemost wildcats from most house cats on log-ratioscaled size alone, but with some overlap betweenexceptionally large house cats and likewise small


Table 4. Length measurements and corresponding log-ratio values for cat bones from post-medieval deposits atKilton Castle, Yorkshire, sorted by size

Context GL(C) GL log

Kilton <28 Radius 114.9 0.096Kilton 28–30 Radius 114.9 0.096Kilton 28–30 Tibia 136.2 0.088Kilton <28 Ulna 133.0 0.087Kilton <28 Femur 128.6 0.086Kilton 28–30 Ulna 132.5 0.085Kilton 28–30 Tibia 135.0 0.084Kilton 28–30 Femur 127.6 0.082Kilton 28–30 Radius 105.3 0.058Kilton 28–30 Tibia 126.6 0.056Kilton 28–30 Tibia 126.1 0.054Kilton No depth Femur 118.4 0.050Kilton 28–30 Femur 118.2 0.049Kilton No depth Ulna 121.8 0.049Kilton <28 Humerus 104.7 0.036Kilton 28–30 Femur 112.9 0.029Kilton <28 Tibia 118.4 0.027Kilton 28–30 Ulna 113.5 0.018Kilton <28 Ulna 113.4 0.018Kilton <28 Fibula 103.6 �0.001Kilton <28 Tibia 111.0 �0.001Kilton <28 Humerus 96.0 �0.002Kilton 28–30 Tibia 104.2 �0.029


wildcats. The same calculations were performedon the numerically smaller wildcat sample,standardising the minimum, maximum and 95%confidence limits against the wildcat sample meanso as to obtain modified values that are directlynumerically comparable with those obtained forthe house cat sample by the same procedure(Table 3). The lower values for the minimum andmaximum show that there were fewer outliers inthe wildcat sample, whereas the slightly highervalues for the 95% confidence limits reflect thesmaller sample size (hence higher values for thestandard error of the mean, hence largerconfidence limits). None the less, the wildcatresults confirm a useful but by no means completeseparation of the two samples, if anythingindicating that it is more likely that an outlyinglarge house cat would be mistaken for a wildcatthan that an exceptionally small wildcat would begrouped with the house cats.Table 2 lists length data from the additional

house cats and wildcats. The three wildcats haveseparated from the house cats, and specimensfrom each of the wildcats have grouped together.EAU776 is the largest of the three by some way.EAU777 and EAU783 are appreciably smaller,particularly the latter. There is no substantial‘break’ in the data between the smallest of thewildcats and the largest of the house cats. Giventhe difficulty of differentiating ‘pure’ wildcats inmodern Scottish populations from pelage char-acters alone, it is at least possible that EAU783 isactually a hybrid. None the less, this sampleusefully shows the range and variance that may beexpected of a sample that contains both wild andhouse cats.The archaeological sample from Kilton Castle

was thought to contain both ecomorphs. Table 4lists the length data for comparison with Tables 1and 2. The first point to note is that the largest ofthe Kilton cats are very large, nearing themaximum values for Kratochvil’s wildcat samplegiven in Table 1. At the other end of the range,the smallest are below the house cat standardvalues. As with the mixed modern sample inTable 2, there is no discernible ‘break’ in the datathat could be taken to mark the boundarybetween wild and domestic forms. Figure 1 showsquite clearly the continuity of the Kilton databetween the largest and smallest specimens. The


Kilton data seem to confirm that the samplecontains both wildcats and house cats, but thecontinuity indicates a considerable size overlapbetween the two ecomorphs.

The medieval archaeological samples werechosen in part as substantial datasets, and in partwith the expectation that samples from medievaltowns would represent populations predomi-nantly of house cats, with maybe an occasionalspecimen of wildcat deriving from animals huntedfor their fur. Figures 2 and 3 show the log-ratiosize distribution of the two York samples, andshow a distinct difference in size distributionbetween two medieval phases in the same city.The Viking age sample, although small innumber, shows a size variation either side ofthe standard, with perhaps three particularly largecats among the breadth measurements (Figure 2).The two specimens exceeding 0.060 in Figure 2are both ulnae. This value is close to the valuediscussed above as likely to differentiate mosthouse cats from wildcats. The third largestspecimen is a radius at 0.048, within the rangeseen in the reference collection wildcats(Table 2). Pending further discussion below,


Figure 1. Kilton Castle length and breadth measurements against house cat standard. This figure is available in colouronline at www.interscience.wiley.com/journal/oa.


the two largest specimens can be described as‘probably wildcat’, with the large radius possiblyrepresenting a hybrid or an exceptionally largehouse cat.Comparing the York Viking age sample with

the medieval sample, the difference is quite


striking. On the medieval length measurements,just three specimens are within �0.01 of thestandard, and the remaining 45 are all below. Onbreadth measurements, just eight are within�0.01 of the standard and the remaining 58are all below. The medieval sample is clearly of


Figure 2. Log-ratio length and breadth distributions for York 10th to early 11th century cats. This figure is available incolour online at www.interscience.wiley.com/journal/oa.


cats that were substantially smaller than thestandard sample: indeed, around 20 of thebreadth sample are as far below the house catstandard as we might expect wildcat breadthmeasurements to be above the standard. Thesefigures indicate a distinct decrease in the size of


cats in York between the Anglo-Scandinavian andHigh Medieval periods (either side of theNorman Conquest). Figure 4 shows length andbreadth results for the Lincoln sample, and a sizedistribution more akin to the York medievalsample than to the contemporary 10th–early 11th


Figure 3. Log-ratio length and breadth distributions for York 12th to 13th century cats. This figure is available in colouronline at www.interscience.wiley.com/journal/oa.


century sample. One cat stands out in the Lincolnbreadth data, with a radius at 0.044. This is not aparticularly high value, although the specimen isdistinctly larger than the remainder of thedistribution. Given that, this specimen could beanother ‘probable wildcat’. More to the point,


the Lincoln data suggest that the size distributionof the York 10th century cats is distinctive to thecity, and not to the period of time.Figures 5 and 6 show the distribution of length

data from the very large assemblages at Haithabuand Schleswig. The range and distribution of the


Figure 4. Log-ratio length and breadth distributions for Lincoln 10th to early 11th century cats. This figure is available incolour online at www.interscience.wiley.com/journal/oa.


Haithabu sample is unexceptional, being quitesimilar to the contemporaneous samples fromYork and Lincoln. The Schleswig sample, incontrast, shows a remarkable range, from 0.105 to�0.138. The five largest specimens, with log-ratio values of 0.096 and above, overlap with


Kratochvil’s wildcat data, with the referencecollection wildcats, and with the largest speci-mens in the Kilton Castle sample. The sixthlargest specimen has a value of 0.077, so there issomething of a break in the data between the fifthand sixth largest specimens, equivalent to, for


Figure 5. Haithabu cat lengths against house cat standard. This figure is available in colour online at www.interscience.wiley.com/journal/oa.

Figure 6. Schleswig cat lengths against house cat standard. This figure is available in colour online at www.interscience.wiley.com/journal/oa.


example, about 5mm difference in the length ofthe femur. It seems feasible, therefore, that thelargest five specimens, at least, in the Schleswigdistribution derive from wildcats. Indeed, a graph


of humerus GL in the original report showsa distinct ‘shoulder’ of larger cats, althoughthis interpretation is not offered (Spahn, 1986,Fig. 24). However, that does not explain the


Table 5. Sample mean values of log-ratio lengthmeasurements and standard deviations of those meanvalues

Sample Mean GL log SD n

Reference collection 0.000 0.038 45Kilton Castle þ0.048 0.036 23York Viking Age �0.012 0.028 8Haithabu �0.015 0.023 248Schleswig �0.026 0.054 208Lincoln Viking Age �0.028 0.024 21York medieval �0.053 0.027 48


remarkably small specimens in the sample, and thequestion of intra-sample variability merits furtherdiscussion.

Discussion

What becomes clear from a consideration of themodern data and the archaeological samples isthat it would be unrealistic to expect a cleardifferentiation in biometric data between wildcatsand their house cat contemporaries. Furthermore,the size distribution of house cat populationsvaried distinctly from time to time and from placeto place. Table 5 summarises the variationobserved in length measurements using the meanlog-ratio values for each sample, and the standarddeviation of that mean. Although this manipula-tion of log-ratio values is unconventional, it servesa useful purpose here in summarising both theoverall size difference between samples in thisinvestigation, and giving us a measure of thedegree of intra-sample variation. The mean valuesshow that the largest average size occurs, notsurprisingly, in the reference collection sampleknown to contain both wildcats and house cats,and in the Kilton Castle sample inferred tocontain both. That result, at least, tends tosupport the interpretation of the Kilton sample.Apart from Schleswig, the other archaeologicalsamples show a consistency in their standarddeviation values, those values being lower than inthe two ‘mixed’ samples. With more conventionaldata, the standard deviation values could beconverted to a coefficient of variation (V¼ 100�SD/mean). However, as the mean values from


these data range either side of zero, such acalculation would be meaningless (i.e. for thereference collection sample, V¼ 100� 0.038/0.000¼1). Indeed, the absolute numerical valueof the calculated standard deviations are of littlesignificance, but the results at least confirm that‘mixed’ samples tend to show a higher intra-sample size variation than do archaeologicalsamples arguably predominantly composed ofhouse cats. Schleswig is therefore highly anom-alous, with a mean value in the same range as theother Viking and medieval archaeologicalsamples, and a size distribution consistent withthe other archaeological samples (Figure 6), but afar higher standard deviation than even the‘mixed’ reference sample.

It has been argued above that the largest fewcats in the Schleswig sample may have beenwildcats, but that does not explain the highoverall variance. The Schleswig sample spans anappreciable period of time, from the later 11th

century into the 14th century. Although somedistinction is made in the published reportbetween ‘young’ and ‘old’ phases within thatperiod, the biometric data are not subdivided. It isclear that cat bones were about equally dis-tributed between the two phases, and that therewas no substantial change in the spatial distri-bution of cat bones between the phases (Spahn,1986: 15–17). What we have for Schleswig,therefore, is a pooled sample spanning severalcenturies. It is quite possible that some sizechange occurred in house cats during this periodof time, greatly increasing the intra-samplevariability. The very small size of some of thecats is notable, suggesting quite strong selectionpressures acting on adult body size (e.g. Spahn,1986, Fig. 25, which illustrates two skulls at eitherend of the size range). Whether those pressuresfavoured smaller body size, or whether we areseeing a relaxation of pressures that wouldnormally favour a larger body size, is notapparent. With a medium-sized animal such asthe cat, it is difficult to predict the direction inwhich body size would tend to move given arelaxation of the selection pressures associatedwith a ‘wild’ lifestyle.

It remains to ask whether the results of thisinvestigation have any practical application in thediagnosis of cat bones from archaeological


Table 6. Howe, Orkney: measurements of cat bones from Late Iron Age Phase 8, as given in Smith (1994), with log-ratiovalues relative to the house cat standard

GL(C) Bp SD Bd GL log Bp log Bd log

Howe Radius 102 9 6 14 0.044 0.048 0.049Howe Radius 95 8 6 13 0.013 0.003 0.017Howe Radius 93 8 6 12 0.004 0.003 �0.018Howe Femur 104 20 0 17 �0.007 — �0.035Howe Femur 104 19 0 17 �0.007 — �0.035Howe Tibia 109 19 0 14 �0.009 �0.008 —

Table 7. Earls Bu, Orkney: measurements of cat bonesfrom late Viking Age deposits with log-ratio values relativeto the house cat standard. Data courtesy of I. Mainlandand J.F. Harland

Log-ratio value

EB humerus Bd¼ 17.7 �0.005EB humerus Bd¼ 21.5 0.079EB radius Bp¼ 7.6 �0.025EB tibia Bp¼ 17.9 �0.034EB astragalus GH¼ 14.2 �0.046EB astragalus GH¼ 15.5 �0.008EB calcaneum GL¼ 27.8 �0.022EB calcaneum GL¼ 27.7 �0.023


samples, given that these generally occur in smallnumbers, precluding the analysis of intra-samplevariation that has been possible here. Table 6 listsmeasurements of cat limb bones from Late IronAge (5th to 8th century AD) structures overlyingthe broch at Howe, Orkney (Smith, 1994). Oneradius stands out from the remainder, sufficientlyso to suggest that this could perhaps be a smallwildcat, the remaining specimens deriving fromhouse cats. The log-ratio values for length andbreadth measurements on the large radius are notclearly in the modern wildcat range, but doapproach the range of values seen in largersamples in which the presence of wildcat is eitherknown (e.g. Table 2) or suspected (Table 4). Thisspecimen lies just outside the 2s range of theYork Viking sample. Perhaps of more significanceis the apparent confirmation that the majority ofspecimens are from house cats. Although theHowe specimens are later in date than the Romanperiod in more southerly parts of Britain (i.e.late 1st to early 5th century AD), Orkney was wellbeyond the sphere of Roman influence. Even ifhouse cats had been introduced to Britain inRoman times, this is unlikely to have resulted intheir introduction to Orkney. Therefore, theidentification of house cats at Howe contradictsthe oft-repeated assertion that house cats ‘came toBritain with the Romans’, at the northern end of adistribution that began with domestication inMiddle or New Kingdom Egypt (e.g. seeKratochvil & Kratochvil, 1976; Teichert, 1977;Serpell, 1988; Malek, 1993). A second example isgiven in Table 7, which lists measurements of catbones from the high-status late Viking Age site atEarl’s Bu, Orkney (Batey et al., 1993). Again, onespecimen, a humerus, stands out as being wellwithin the range to be expected of wildcat, with


the remaining specimens all being typical ofrather small house cats in the range seen at Yorkand Haithabu. As these Orkney specimens comefrom insular sites, it should be noted that cats, incommon with most carnivores, do not exhibitdwarfing in island populations (Meiri et al., 2004).Finally, limb bone measurements from a cat

recovered from the Roman (1st–2nd century AD)site at Quseir el-Qadim, Egypt, have been testedagainst the house cat standard. This cat wasreported by von den Driesch & Boessneck (1983),who drew particular attention to its size:‘extremely large for a house cat’ (p. 208). Thesame authors drew attention to the often largesize and ‘wild’ proportions of house cats in ancientEgypt (Boessneck & von den Driesch, 1982). Thecat was wrapped in linen and lain on woollentextile (appropriately of tabby weave) in a nichewithin a major building. By adding this egregiouscat to the analysis, the aim was to test thelimitations of the procedure. Conversion of limbbone lengths and breadths to log-ratio values



against the house cat standard gave a range ofstrongly positive values, from 0.043 (radius Bd,although note 0.026 for calcaneum GL) to0.079 (femur GLC). On the face of it, thisis a wildcat-sized animal, although calculationsof log-ratio values against Kratochvil’s wildcatsample means (Table 1) gave consistentlynegative values (�0.013 tibia Bp to �0.047radius Bd). Von den Driesch & Boessneck werequite confident that this was a house cat, citingmorphological traits of the cranium and mand-ible, and the photographs that accompany theirpaper clearly show a shortened angular processon the mandible, typical of house cats. Evidently,any investigation of house cats from Roman orearlier Egyptian sites would need to pay closeattention to context as well as to biometry.

Conclusions

This investigation has demonstrated that differ-entiation of wildcats and house cats may befeasible using a conventional and relativelyunsophisticated biometric technique. However,examination of intra-sample range and variationwithin large zooarchaeological samples showsthat it may be unrealistic to expect a clear sizedistribution break between the two ecomorphs, aresult that is consistent with modern observationsthat sympatric wildcats and house cats interbreedfreely to produce a continuum of hybrid forms.None the less, application of the results to twotypically small samples has shown that specimensof wildcat may be identified in samples consistinglargely of house cats, albeit with differing degreesof confidence. The Quseir el-Qadim cat was anexceptional beast, and serves as a warning that thecontext of the finds should be taken intoconsideration. The next step is to attempt theconverse, namely to identify scarce house catsamong samples that largely consist of wildcats(O’Connor, in prep.). Acquiring the data for suchan analysis is more difficult than accessing thesubstantial medieval samples used in this paper,but further investigation is essential if ourknowledge of the domestication and adoptionof this most familiar of companion animals is toproceed on a more sound evidential basis.


Acknowledgements

I am grateful to Ingrid Mainland and JenniferHarland for access to data from Earl’s Bu, andto referees for their comments on earlier drafts.

References

Batey CE, Harry RC, Morris CD. 1993. Excavations atthe Earl’s Bu, Orphir, Orkney 1993. Glasgow UniversityArchaeological Research Division: Glasgow.

Boessneck J, von den Driesch A. 1982. Studien ansubfossilen Tierknochen aus Agypten. MunchnerAgyptologische Studien 40: 1–172.

Bond JM, O’Connor TP. 1999. Bones from MedievalDeposits at 16–22 Coppergate and Other Sites in York.Council for British Archaeology, Archaeology ofYork 15/5: York.

Clutton-Brock J. 1999. A Natural History of DomesticatedMammals, (2nd edn). Cambridge University Press:Cambridge.

Crockford S (ed.). 2000. Dogs Through Time: AnArchaeological Perspective, BAR International Series889. Archaeopress: Oxford.

Daniels MJ, Beaumont MA, Johnson PJ, Balharry D,Macdonald DW, Barratt E. 2001. Ecology andgenetics of wild-living cats in the north-east ofScotland and the implications for the conservationof the wildcat. Journal of Applied Ecology 38: 146–161.

Davis SJM. 1989. Some more animal remains from theAceramic Neolithic of Cyprus. In Fouilles recentes aKhirokitia (Chypre) 1983–1986, Le Brun A (ed.). ADPF,Edition Recherches sur les Civilisations: Paris;189–221.

Davis SJM. 1996. Measurements of a group of adultfemale Shetland sheep skeletons from a singleflock: a baseline for zooarchaeologists. Journal ofArchaeological Science 23: 593–612.

Gentry A, Clutton-Brock J, Groves CP. 2004. Thenaming of wild animal species and their domesticderivatives. Journal of Archaeological Science 31:645–651.

Harcourt RA. 1979. The animal bones. In Gussage AllSaints. An Iron Age settlement in Dorset, Wainwright GJ.Dept of Environment Archaeological ReportNo. 10. Her Majesty’s Stationery Office: London;150–160.

Johansson F, Huster H. 1987. Untersuchungen an Skelet-tresten von Katzen aus Haithabu (Ausgrabung 1966–1969).Karl Wachholtz, Berichte uber die Ausgrabungen inHaithabu 24: Neumunster.



Kitchener AC, Yamaguchi N, Ward JM, MacdonaldDW. 2005. A diagnosis for the Scottish wildcat(Felis silvestris): a tool for conservation action for acritically-endangered felid. Animal Conservation 8:223–237.

Kratochvil Z. 1973. Schadelkritieren der Wild- undHauskatze (Felis silvestris silvestris Schreb. 1777 undF. s. f. catus L. 1758). Acta Scientiarium Naturalium Brno7(10): 1–50.

Kratochvil Z. 1976a. Das postkranialskelett derWild- und Hauskatze (Felis silvestris und F. lybica f.catus). Acta Scientiarium Naturalium Brno 10(6): 1–43.

Kratochvil Z. 1976b. Sex dimorphism of the domesticcat (Felis lybica f. catus L.) on the skull and on themandible. Acta Veterinaria Brno 45: 159–167.

Kratochvil J, Kratochvil Z. 1976. The origin of thedomesticated forms of the genus Felis (Mammalia).Zoologicke Listy 25(3): 193–208.

Lecis R, Perpaoli M, Biro ZS, Szemethy L, Ragni B,Vercillo F, Randi E. 2006. Bayesian analysis ofadmixture in wild and domestic cats (Felis silvestris)using linked microsatellite loci.Molecular Ecology 15:119–131.

Malek J. 1993. The Cat in Ancient Egypt. British MuseumPress: London.

Meadow RH. 1999. The use of size-index scalingtechniques for research on archaeozoologicalcollections from the Middle East. In HistoriaeAnimalium ex Ossibus, Becker C, Manhart H, PetersJ, Schibler J (eds). Marie Leidorf Verlag: Rahden;285–300.

Meiri S, Dayan T, Simberloff D. 2004. Body size ofinsular carnivores: little support for the island rule.American Naturalist 163: 469–479.

O’Connor TP. 1982. Animal Bones from Flaxengate c.870–1500. Council for British Archaeology, Archae-ology of Lincoln: 18/1 London.

O’Connor TP. 1987. Why bother looking at archae-ological wild mammal assemblages? Circaea 4(2):107–114.

O’Connor TP. 1989. Bones from Anglo-Scandinavian Levelsat 16–22 Coppergate. Council for British Archaeology,Archaeology of York 15/3: London.

Pierpaoli M, Biro ZS, Herrmann M, Hupe K,Fernandes M, Ragni B, Szemethy L, Randi E.2003. Genetic distinction of wildcat (Felis silvestris)populations in Europe, and hybridization withdomestic cats in Hungary. Molecular Ecology 12:2585–2598.


Rohrs M. 1955. Vergleichedne Untersuchungen anWild- und Hauskatze. Zoologische Anzeige 155:53–69.

Schauenberg P. 1969. L’identification du Chat forest-ier d’Europe Felis s. silvestris Schreber 1777 par unemethode osteometrique. Revue Suisse de Zoologie76(2): 433–441.

Serpell JA. 1988. The domestication and history of thecat. In The Domestic Cat. The Biology of its Behaviour,Turner DC, Bateson P (eds). Cambridge UniversityPress: Cambridge; 151–158.

Serpell JA. 1996. In the Company of Cnimals. A Study ofHuman-Animal Relationships. CUP Canto edition:Cambridge.

Smith BB (ed.). 1994. Howe: Four Millennia of OrkneyPrehistory. Excavations 1978-1982. Society of Antiqua-ries of Scotland: Edinburgh.

Spahn N. 1986. Untersuchungen an skelettresten von Hundenund Katzen aus dem mittelalterlichen Schleswig. AusgrabungSchild 1971–1975. Karl Wachholz; Ausgrabungen inSchleswig, Berichte und Studien 5: Neumunster.

Sunquist M, Sunquist F. 2002. Wild Cats of the World.University of Chicago Press: Chicago.

Teichert M. 1977. Fundnachweise von Wild- undHauskatzenknochen aus ur- und fruhgeschichtli-cher Zeit. Hercynia, Leipzig 14: 212–216.

Todd NB. 1978. An ecological, behavioural, geneticmodel for the domestication of the cat. Carnivore 1:52–60.

Vigne J-D, Guilane J, Debue K, Haye L, Gerard P.2004. Early taming of the cat in Cyprus. Science 304:259.

Vigne JD, Peters J, Helmer D (ed.). 2005. The FirstStages of Animal Domestication. Oxbow Books: Oxford.

von den Driesch A. 1976. A Guide to the Measurement ofAnimal Bones from Archaeological Sites, PeabodyMuseum Bulletin No. 1: Harvard.

von den Driesch A, Boessneck J. 1983. A Roman catskeleton fromQuseir on the Red Sea coast. Journal ofArchaeological Science 10: 205–211.

Yamaguchi N, Driscoll C, Kitchener AC, Ward JM,Macdonald DW. 2004. Craniological differen-tiation between European wildcats (Felis silvestrissilvestris), African wildcats (F. s. lybica) and Asianwildcats (F. s. ornate): implications for their evolutionand conservation. Biological Journal of the LinnaeanSociety 83: 47–63.

Zeuner FE. 1963. A History of Domesticated Animals.Hutchinson: London.


Wild or Domestic? Biometric Variation in the Cat Felis silvestris Schreber

Documents

Transcript of Wild or Domestic? Biometric Variation in the Cat Felis silvestris Schreber