The Categorisation of Occupation in Identified Skeletal Collections: A Source of Bias?

ABSTRACT

Identified skeletal collections, that is, skeletons for which sex, age at death and occupation at death are known, have been used to test methods for recording entheseal changes (EC). By testing methods on identified collections, the sensitivity of EC for recording activity levels can be ascertained prior to applying the methods to test hypotheses in archaeological contexts. However, the definition of occupational categories used for this research may, in itself, be a source of bias. The aim in this study was to test how categorising occupation affected the interpretation of EC. Male skeletons (n = 211) from two Portuguese identified skeletal collections were used. Three methods for categorising occupations, all of which have been previously published, were used each dividing occupations into five, three and two categories, respectively. Fibrocartilaginous entheses were recorded and EC scored as present/absent. Results showed that the method for categorising occupation affected the frequencies of EC found in occupational categories for specific entheses. Frequencies that were significantly different between occupational categories for one method were not necessarily significant for others. This demonstrates that the sensitivity of the occupational categorisation does affect the results. However, using logistic regression, we found age to have a greater effect than occupation. These results demonstrate the need for standardised occupational categories, as well as the importance of considering age. Copyright © 2012 John Wiley & Sons, Ltd.

Introduction

The use of identified skeletal collections for osteological analyses has been considerable in the past few years. This increased interest is due to the quantity and quality of data available about the lives of the individuals, which include sex, age at death and occupation (Rocha, 1995; Cardoso, 2006; Cunha & Wasterlain, 2007). As a result, these collections have been used to develop and test methods for age at death and sex assessment, as well as to test correlations between those variables and osteological changes observed in the skeleton that are attributed to disease or activity (Matos & Santos, 2006; Santos & Roberts, 2006; Alves Cardoso, 2008). For research into the latter, entheseal changes (EC) have been the most widely recorded (Cunha, 1995; Mariotti et al., 2004, 2007; Alves Cardoso, 2008; Alves Cardoso & Henderson, 2010; Villotte et al., 2010; Niinimäki, 2011; Campanacho et al., 2012; Milella et al., 2012). However, the assumption that EC are directly and exclusively linked to muscle use during activity has been questioned for over a decade (Jurmain, 1999; Jurmain et al., 2012). Nevertheless, testing these recording methods on identified collections continues, as it is seen as the best method for controlling for factors affecting EC, such as sex, age at death and occupation. However, whereas sex and age at death are reliable in these collections, occupation at death is not. Occupation at probate does not provide the full record of activities an individual performed during his or her life course. Activities, such as hobbies, cooking, cleaning and changes of occupation, through life are not described in these records (Cardoso, 2005; Caffell et al., 2012). Furthermore, when using occupation at death, in studies that aim to assess EC and activity, occupations are normally classified and grouped on the basis of how heavy and repetitive workload was during life, but again, these descriptions are also a source of bias dependent on the research objective and research design (Perréard Lopreno et al., in this issue).

Aims and objectives

The aim of this study was to test the hypothesis that EC frequency is dependent on the method for categorising occupation. If this is supported, it has implications for studies that have previously tested EC recording methods using these categories, the development of others and those using them to differentiate between labour patterns. The study will also test the effect of occupational grouping on age profiles, because EC frequency has been shown to increase with age. The working hypothesis is that those individuals described as non-manual workers, and frequently associated with higher socio-economic status, will have a higher mean age at death due to better access to dietary and medical resources. If this is upheld, then non-manual workers should also have a higher frequency of EC (due to the increased age), when compared with the working groups associated with lower socio-economic status, for example manual workers.

Materials and methods

Two Portuguese identified skeletal collections were used: the Luis Lopes skeletal collection and the Coimbra collection. The Luis Lopes skeletal collection, curated in the Museum of Natural History in Lisbon, represents a predominantly urban population. In 2006, this collection was composed of 1692 documented skeletons and 75 unidentified individuals (Cardoso, 2006). Since that date, new individuals have been incorporated into the collection (Hugo Cardoso, pers. comm.). On the basis of the detailed information available for 699 individuals, it was concluded (Cardoso, 2006) that the collection consists mostly of individuals of Portuguese nationality who were born and had died between 1805 and 1975. On the basis of the occupational profile of the collection, it was described as composed of individuals of lower to middle socio-economic status. Male individuals were predominantly working in sales (e.g. shop assistants), in services (e.g. were civil servants), or as artisans and in other skilled trades (e.g. carpentry or tailoring). Female individuals were mostly listed as domésticas (housewives/housekeepers), and these represented 85% of the total number of female individuals, with other individuals listed as maids, teachers or students (Cardoso, 2006; Alves Cardoso, 2008).

The collection from Coimbra is currently curated in the Museum of Anthropology in the Department of Life Sciences in Coimbra University. This collection is composed of individuals who were born and died between 1826 and 1938 (Rocha, 1995; Alves Cardoso, 2008). The profile of this collection represents a similar distribution to that of Luis Lopes collection, that is, the lower to middle social classes of society (Cardoso, 2006; Alves Cardoso, 2008). The majority of the male individuals were employed at the time of their death as waiters, farmers and unskilled workers, or as skilled workers, for example barbers, carpenters, tailors and shoemakers. Occupations relating to commerce and transport, and professions such as teachers and lawyers were also represented (Alves Cardoso, 2008).

The sample used in this study was composed of a total of 211 individuals, 104 male skeletons from the Coimbra collection and 107 male skeletons from the Lisbon Luis Lopes collection. These individuals were selected on the basis of their specific activity with the aim to have a similar number of individuals in occupational groups. This was not entirely possible to achieve because of skeleton preservation. Only male skeletons (n = 211) were used for this study, as female skeletons were most commonly referred to in the records as ‘doméstica’, that is, housekeeper/housewives. This categorisation of occupation was considered extremely ambiguous in relation to the actual tasks women were performing that could range from child care to farming activities (Alves Cardoso, 2008). Therefore, categorisation of female occupations was not appropriate, and this is one of the serious limitations of using documentary evidence of occupations at death to categorise work.

Three methods were used to categorise occupations (Roque, 1988; Alves Cardoso & Henderson, 2010 and Villotte et al., 2010), all of which have been previously employed in skeletal analyses. These methods were used to divide the sample to test the hypothesis that EC frequency is dependent on categorisation method and test their impact on age distribution. The three categories divide the occupations into five (Roque, 1988), three (Villotte et al., 2010) and two categories (Alves Cardoso & Henderson, 2010), respectively.

Table 1 presents the full list of occupations, their categorisation by method and the sample distribution. The first method, referred to forthwith as the Roque method (Roque, 1988), divides the sample into five categories (government, administrative and service industry, commerce and transport, skilled workers and artisans, farmers and servants, and unskilled workers). This method is based on contemporary Portuguese socio-economic status and, unlike the other two methods, was not devised by anthropologists studying past populations and activity-related stress. Therefore, this method does not consider the physical activity involved in an occupation, only the social and economic arena in which it takes place. Roque (1988), with his description of the social and economic constitution of the Coimbra city, developed a characterisation of Coimbra's population, thought to be comparable with that of Lisbon (Alves Cardoso, 2008).

Table 1. Occupations and categorisation by method

Occupation	N	Five-category		Three-category (non-manual, light manual and heavy manual)		Two-category (manual and non-manual)
Chauffeur	2	Commerce/Transport		Light manual		Manual
Coach driver	3	Commerce/Transport		Light manual		Manual
Hospital employee	1	Commerce/Transport		Light manual		Manual
Pharmacy assistant	1	Commerce/Transport		Light manual		Manual
Shop assistant	25	Commerce/Transport		Light manual		Manual
Stallman	1	Commerce/Transport		Light manual		Manual
Commercial agent	2	Commerce/Transport		Non-manual		Non-manual
Industrial	8	Commerce/Transport		Non-manual		Non-manual
Insurance worker	2	Commerce/Transport		Non-manual		Non-manual
Merchant	9	Commerce/Transport		Non-manual		Non-manual
Newspaper man	1	Commerce/Transport		Non-manual		Non-manual
Salesman	2	Commerce/Transport		Non-manual		Non-manual
Salesperson	1	Commerce/Transport		Non-manual		Non-manual
Farmer	3	Farmers/Servants		Heavy manual		Manual
Road mender	3	Farmers/Servants		Heavy manual		Manual
Servant	3	Farmers/Servants		Light manual		Manual
Building constructor	1	Government and Services		Heavy manual		Manual
Bank clerk	3	Government and Services		Non-manual		Non-manual
City ouncil employee	2	Government and Services		Non-manual		Non-manual
Civil servant	13	Government and Services		Non-manual		Non-manual
Clerk	6	Government and Services		Non-manual		Non-manual
Corporation employee	1	Government and Services		Non-manual		Non-manual
Court official	1	Government and Services		Non-manual		Non-manual
Owner/proprietor	5	Government and Services		Non-manual		Non-manual
Pharmacist	1	Government and Services		Non-manual		Non-manual
Scribe	1	Government and Services		Non-manual		Non-manual
Solicitor	1	Government and Services		Non-manual		Non-manual
Student	1	Government and Services		Non-manual		Non-manual
Teacher	3	Government and Services		Non-manual		Non-manual
Baker	6	Skilled workers/Artisans		Heavy manual		Manual
Blacksmith	1	Skilled workers/Artisans		Heavy manual		Manual
Bricklayer	8	Skilled workers/Artisans		Heavy manual		Manual
Carpenter	15	Skilled workers/Artisans		Heavy manual		Manual
Foundry worker	2	Skilled workers/Artisans		Heavy manual		Manual
Glass blower	1	Skilled workers/Artisans		Heavy manual		Manual
Mechanic	1	Skilled workers/Artisans		Heavy manual		Manual
Stoker	1	Skilled workers/Artisans		Heavy manual		Manual
Stonemason	1	Skilled workers/Artisans		Heavy manual		Manual
Sawyer	1	Skilled workers/Artisans		Heavy manual		Manual
Barber	5	Skilled workers/Artisans		Light manual		Manual
Basket weaver	1	Skilled workers/Artisans		Light manual		Manual
Electrician	5	Skilled workers/Artisans		Light manual		Manual
Fishmonger	1	Skilled workers/Artisans		Light manual		Manual
Gilder	1	Skilled workers/Artisans		Light manual		Manual
Locksmith	3	Skilled workers/Artisans		Light manual		Manual
Plumber	2	Skilled workers/Artisans		Light manual		Manual
Shoemaker	12	Skilled workers/Artisans		Light manual		Manual
Tailor	3	Skilled workers/Artisans		Light manual		Manual
Tanner	1	Skilled workers/Artisans		Light manual		Manual
Toothpick artisan	1	Skilled workers/Artisans		Light manual		Manual
Upholsterer	1	Skilled workers/Artisans		Light manual		Manual
Weaver	1	Skilled workers/Artisans		Light manual		Manual
Wireman	1	Skilled workers/Artisans		Light manual		Manual
Photographer	2	Skilled workers/Artisans		Non-manual		Non-manual
Carrier / worker	1	Unskilled workers		Heavy manual		Manual
Worker	24	Unskilled workers		Heavy manual		Manual
Caretaker	2	Unskilled workers		Light manual		Manual
Pantry keeper	1	Unskilled workers		Light manual		Manual
Total	211
		Grouping categories	n	Grouping categories	n	Grouping categories	n
		Government and Services	39	Heavy manual	69	Manual	146
		Commerce/Transport	58	Light manual	77	Non-manual	65
		Farmers/Servants	9	Non-manual	65	Total	211
		Skilled workers/Artisans	77	Total	211
		Unskilled workers	28
		Total	211

The second method is based on that published by Villotte et al. (2010), which divided the occupations into the following four categories: non-manual workers, manual workers, manual workers carrying heavy loads and manual workers probably involved in forceful manual tasks (Villotte et al., 2010: 22). However, while presenting and discussing their results (Villotte et al., 2010: 22), they refer predominantly to three categories: non-manual, light manual and heavy manual. It is these three categories that will be used for this study. As described by Villotte et al., their grouping criteria considered the descriptions of the occupations in the collections alongside the interpretation of the performance of these occupations in the past. The authors also took into account a medical perspective of occupational injuries. EC where then analysed using presence and absence (which correspond to the definition of presence and absence used in this paper). Villotte et al. (2010) did not find differences between the light manual and non-manual workers, but there was a difference between both of these groups and heavy manual workers.

The final method divides occupations into manual and non-manual workers (Alves Cardoso & Henderson, 2010). This grouping took into consideration the historical evidence for the activities performed (Roque, 1988; Alves Cardoso, 2008): occupations historically deemed more strenuous or physically demanding were grouped as manual, whereas those listed as non-manual were associated with less physically demanding activities. Within this grouping parameter, it is hypothesised, given that Villotte et al. did not find any differences between light manual and non-manual workers, that pooling heavy manual and light manual workers will mean that no statistically significant differences will be found between manual and non-manual workers.

Fibrocartilaginous EC were recorded as present or absent as described in Alves Cardoso & Henderson (2010) on the basis of anatomical descriptions of normal and abnormal entheses (Benjamin et al., 2002). The absence of EC is defined as no deviation from the normal smooth, well-defined enthesis (Benjamin & Ralphs, 1998; Alves Cardoso & Henderson, 2010). Any deviation from this, for example the presence of lytic lesions or bone formation, was recorded as present (Figure 1). This method is only applicable to fibrocartilaginous entheses and not to fibrous ones (Alves Cardoso & Henderson, 2010; Jurmain et al., 2012). There is currently no agreed definition for the normal appearance of fibrous entheses (Jurmain et al., 2012); therefore, only fibrocartilaginous entheses were used in this study. The entheses recorded represented both the upper and lower limbs. Entheses of the upper limb were recorded, as it is the arm, forearm and hand that are most commonly utilised for occupation-related tasks. The lower limb was included to study mobility as well as to record changes that may be due to occupational use of the lower limb, for example bending to pick up heavy loads. The entheses (all insertions, unless otherwise specified) recorded were m. subscapularis, m. infra- and supraspinatus (recorded as one enthesis as the fibres merge close to the attachment; Minagawa et al., 1998), common flexor origin, common extensor origin, m. biceps brachii, m. triceps brachii, m. brachialis, hamstring group (recorded as one enthesis), gluteus group (recorded as one enthesis), and m. triceps surae.

Figure 1. Insertion of the m. subscapularis in two left humeri (from the Luis Lopes Collection). (A) Normal appearance of the fibrocartilaginous insertion site. (B) Abnormal appearance with disruption to the normally smooth enthesis, that is, entheseal changes that include new bone formation.

The data analysis considered the importance of both age and occupation on EC frequency. Therefore, age distribution for the total sample was tested for normal distribution using a Shapiro–Wilk test prior to further decisions on appropriate statistical tests. The age distribution was not normal (W = 0.972; p < 0.005); therefore, non-parametric tests were used to test if there were statistically significant age differences between the occupational categories: Mann–Whitney U was used for the binary division (manual versus non-manual) and Kruskal–Wallis was used for the other categorisations.

The relationships between EC and occupational categorisation methods were tested using Chi-square tests, with the Fisher exact significance, so that the statistical assumptions for Chi-square test were not violated. Finally, logistic regression was used to test the effect of both occupational categorisation and age on EC. Statistical significance for all these tests was set at 95%.

Results

Table 2 presents the age distribution and tests of normality for age by occupational category. The analysis of age differences between occupational groups revealed no statistically significant differences in age in the five-category occupational grouping (H = 1.748, p = 0.782). However, once the sample is compacted into three occupational groups, there are statistically significant differences in age between groups (H = 7.896, p = 0.020) with non-manual workers having a significantly higher median age at death (56 years) than the light manual workers (45 years). The heavy manual workers had a median age at death (52 years). The trend for statistically significant differences in age between the occupational groups continues, as they become less detailed, with the two-category grouping non-manual workers having a higher median age at death (56 vs 47 years) than manual workers (U = 3765.0, p = 0.017).

Table 2. Age distribution and tests of normality (Shapiro–Wilk test) for age by occupational category

	N	Minimum	Mean	Median	Maximum	Standard deviation	Shapiro–Wilk	p-value
Bold indicates statistically significant values (95%).
Five-category
Government administration/Services	39	23	52.77	56.00	82	17.164	0.960	0.18
Commerce/Transport	58	20	51.79	49.50	85	18.888	0.955	0.03
Skilled workers/Artisans	77	20	49.13	46.00	84	17.771	0.960	0.02
Farmers/Servants	9	34	49.33	50.00	66	10.805	0.966	0.86
Unskilled workers	28	26	48.79	52.00	70	13.318	0.937	0.09
Three-category
Heavy manual	69	23	50.61	52.00	84	15.906	0.974	0.16
Light manual	77	20	46.71	45.00	83	16.737	0.965	0.03
Non-manual	65	23	54.86	56.00	85	18.096	0.955	0.02
Two-category
Manual	146	20	48.55	47.00	84	16.410	0.975	0.01
Non-manual	65	23	54.86	56.00	85	18.096	0.955	0.02
Total	211	20	50.50	50.00	85	17.154	0.972	<0.005

Table 3 presents the frequencies of EC by enthesis for each occupational categorisation method. Overall, the levels of EC presence are not high, but where present, statistically significant differences were found (Table 4). This can, for example, be seen in the left infra- and supraspinatus and the right triceps brachii: all demonstrating statistically significant differences between the occupational groups for all categorisation methods. No other consistent patterns were found. The gluteus muscles had statistically significant differences for the left side in the five-category method and the right side in the three-category method, but no statistical difference for the two-category method. The brachialis, common extensor origin and subscapularis all demonstrated statistically significant differences between occupations in the five-category method but for no other method. No further statistically significant differences were found.

Table 3. Description of the fibrocartilaginous entheses used and entheseal change (EC) frequency for each occupation categorisation method

	Subscapularis				Infra- and supraspinatus				Common flexor origin				Common extensor origin				Biceps brachii
	Left		Right		Left		Right		Left		Right		Left		Right		Left		Right
	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%
The infra- and supraspinatus insertions were recorded as one enthesis. n/N, number of individuals with presence of EC/total number of individuals where EC observation was possible. % refers to positive cases of EC within that specific subgroup. Total refers to the total sample.
Five-category
Government administration/Services	9/32	28.1	10/36	27.8	5/27	18.5	2/26	7.7	2/30	6.7	3/31	9.7	3/30	10.0	5/31	16.1	9/35	25.7	9/37	24.3
Commerce/Transport	11/54	20.4	14/51	27.5	5/45	11.1	8/46	17.4	0/45	0.0	2/43	4.7	8/44	18.2	9/45	20.0	18/55	32.7	19/54	35.2
Skilled workers/Artisans	17/72	23.6	19/74	25.7	6/72	8.3	7/67	10.4	1/68	1.5	3/63	4.8	7/69	10.1	10/63	15.9	24/75	32.0	18/73	24.7
Farmers/Servants	0/8	0.0	0/8	0.0	1/8	12.5	1/8	12.5	0/9	0.0	0/9	0.0	0/9	0.0	0/9	0.0	1/9	11.1	1/9	11.1
Unskilled workers	5/27	18.5	8/26	30.8	1/24	4.2	4/26	15.4	1/24	4.2	2/25	8.0	2/25	8.0	3/25	12.0	13/28	46.4	8/27	29.6
Three-category
Heavy manual	14/66	21.2	16/66	24.2	4/64	6.3	9/61	14.8	2/65	3.1	4/60	6.7	7/64	10.9	8/60	13.3	27/68	39.7	18/67	26.9
Light manual	12/70	17.1	19/70	27.1	5/65	7.7	7/65	10.8	0/59	0.0	1/58	1.7	4/61	6.6	7/59	11.9	19/75	25.3	18/71	25.4
Non-manual	16/57	28.1	16/59	27.1	9/47	19.1	6/47	12.8	2/52	3.8	5/53	9.4	9/52	17.3	12/54	22.2	19/59	32.2	19/62	30.6
Two-category
Manual	26/136	19.1	35/136	25.7	9/129	7.0	16/126	12.7	2/124	1.6	5/118	4.2	11/125	8.8	15/119	12.6	46/143	32.2	36/138	26.1
Non-manual	16/57	28.1	16/59	27.1	9/47	19.1	6/47	12.8	2/52	3.8	5/53	9.4	9/52	17.3	12/54	22.2	19/59	32.2	19/62	30.6
Total	42/193	21.8	51/195	26.2	18/176	10.2	22/173	12.7	4/176	2.3	10/171	5.8	20/177	11.3	27/173	15.6	65/202	32.2	55/200	27.5
	Triceps brachii				Brachialis				Hamstrings				Gluteus muscles				Triceps surae
	Left		Right		Left		Right		Left		Right		Left		Right		Left		Right
	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%	n/N	%
Five-category
Government administration/Services	2/32	6.3	7/31	22.6	2/36	5.6	2/35	5.7	11/37	29.7	14/35	40.0	12/37	32.4	11/33	33.3	12/33	36.4	11/32	34.4
Commerce/Transport	2/47	4.3	6/49	12.2	5/54	9.3	8/53	15.1	17/54	31.5	19/55	34.5	22/53	41.5	18/56	32.1	18/50	36.0	19/53	35.8
Skilled workers/Artisans	8/73	11.0	9/67	13.4	4/76	5.3	7/73	9.6	27/71	38.0	30/74	40.5	20/76	26.3	28/75	37.3	23/68	3.3.8	29/69	42.0
Farmers/Servants	0/9	0.0	0/9	0.0	0/9	0.0	0/9	0.0	2/9	22.2	2/9	22.2	1/9	11.1	2/9	22.2	2/5	40.0	3/6	50.0
Unskilled workers	2/26	7.7	2/25	8.0	1/28	3.6	2/28	7.1	12/28	42.9	10/28	35.7	7/28	25.0	6/27	22.2	10/26	38.5	9/27	33.3
Three-category
Heavy manual	5/65	7.7	6/63	9.5	2/69	2.9	4/67	6.0	26/65	40.0	26/67	38.8	18/69	26.1	15/68	22.1	20/58	34.5	26/62	41.9
Light manual	7/68	10.3	6/64	9.4	5/74	6.8	9/71	12.7	20/72	27.8	22/73	30.1	21/73	28.8	31/73	42.5	22/67	32.8	26/69	37.7
Non-manual	2/54	3.7	12/54	22.2	5/60	8.3	6/60	10.0	23/62	37.1	27/61	44.3	23/61	37.7	19/59	32.2	23/57	40.4	19/56	33.9
Two-category
Manual	12/133	9.0	12/127	9.4	7/143	4.9	13/138	9.4	46/137	33.6	48/140	34.3	39/142	27.5	46/141	32.6	42/125	33.6	52/131	39.7
Non-manual	2/54	3.7	12/54	22.2	5/60	8.3	6/60	10.0	23/62	37.1	27/61	44.3	23/61	37.7	19/59	32.2	23/57	40.4	19/56	33.9
Total	14/187	7.5	24/181	13.3	12/203	5.9	19/198	9.6	69/199	34.7	75/201	37.3	62/203	30.5	65/200	32.5	65/182	35.7	71/187	38.0

Table 4. Results of the statistical tests for the association between entheseal changes (EC) and occupational categories

	Five-category						Three-category						Two-category
	Left			Right			Left			Right			Left			Right
	n/N	χ^2	p-value	n/N	χ^2	p-value	n/N	χ^2	p-value	n/N	χ^2	p-value	n/N	χ^2	p-value	n/N	χ^2	p-value
Statistical significance (at 95%) is highlighted in bold. n/N, number of individuals with presence of EC/total number of individuals where EC observation was possible.
Subscapularis	42/193	3.359	0.7	51/195	3.222	0.030	42/193	2.221	0.134	51/195	0.189	0.743	42/193	1.891	0.184	51/195	0.041	0.860
Infra- and supraspinatus	18/176	3.346	0.038	22/173	1.975	0.068	18/176	5.632	0.013	22/173	0.450	0.477	18/176	5.559	0.025	22/173	0.0	1.000
Common flexor origin	4/176	4.448	0.032	10/171	1.842	0.100	4/176	2.141	0.149	10/171	3.102	0.089	4/176	0.823	0.583	10/171	1.794	0.288
Common extensor origin	20/177	3.640	0.069	27/173	2.58	0.036	20/177	3.250	0.075	27/173	2.657	0.079	20/177	2.652	0.121	27/173	2.008	0.118
Biceps brachii	65/202	5.114	0.033	55/200	3.357	0.118	55/202	3.376	0.071	55/200	0.485	0.517	65/202	0.0	1.000	55/200	0.446	0.608
Triceps brachii	14/187	2.780	0.140	24/181	4.364	0.020	14/187	1.894	0.146	24/181	5.376	0.019	14/187	1.569	0.243	24/181	5.375	0.030
Brachialis	12/203	1.995	0.101	19/198	3.604	0.046	12/203	1.854	0.216	19/189	1.803	0.213	12/203	0.898	0.515	19/198	0.016	1.000
Hamstrings	69/199	2.439	0.067	75/201	1.525	0.158	69/199	2.486	0.113	75/201	2.930	0.085	69/199	0.234	0.633	75/201	1.808	0.206
Gluteus muscles	62/203	5.715	0.025	65/200	2.546	0.061	62/203	2.229	0.144	65/200	6.687	0.014	62/203	2.109	0.183	65/200	0.003	1.000
Triceps surae	65/182	0.239	0.658	71/187	1.375	0.203	65/182	0.814	0.331	71/187	0.805	0.425	65/182	0.777	0.407	71/187	0.554	0.513

For those entheses with statistically significant associations with occupation, the effect of age was tested. The results of the logistic regression, in which both age and occupational category were considered, are presented in Table 5. For two of the tests, the model did not fit the data: the right brachialis in the five-category method and the right triceps brachii in the two-category method; therefore, these will not be discussed further. For those entheses found to have consistently significant differences when only occupation was taken into account, that is, the left infra- and supraspinatus and the right triceps brachii, it was found that age, and not occupation, was statistically significant. This was also found to be the case for most of the other entheses found to have statistically significant associations with occupation in the previous test, excepting the left common flexor origin for the five-category group for which neither occupation nor age were statistically significant. However, two entheses stand out: the left biceps brachii enthesis in the five-category group and the right gluteus muscles in the three-category group. For the left biceps brachii, both age and occupation were found to effect EC frequency. In contrast, only occupation was found to be significant for the right gluteus insertion.

Table 5. Results of logistic regression for those entheses and categories with statistically significant associations between entheseal changes and occupation

	Predictors	Wald statistic/df	p	Wald statistic/df	p
		Left		Right
The Wald statistic tests the statistical significance of each coefficient (age and occupational category) in the model. The Hosmer–Lemeshow goodness-of-fit test was not significant in the cases presented (p > 0.05).
Subscapularis	Age			30.845/1	<0.001
	Five-category			0.965/4	0.915
Infra- and supraspinatus	Age	17.731/1	<0.001
	Five-category	2.187/4	0.701
Infra- and supraspinatus	Age	18.403/1	0.791
Infra- and supraspinatus	Age	7.207/1	0.007
	Two-category	0.004/1	0.947
Common flexor origin	Age	1.985/1	0.159
	Five-category	1.455/4	0.835
Common extensor origin	Age			23.771/1	<0.001
	Five-category			0.221/4	0.994
Biceps brachii	Age	39.507/1	<0.001
	Five-category	8.689/4	0.069
Triceps brachii	Age			12.122/1	<0.001
	Five-category			2.048/4	0.727
Triceps brachii	Age			11.333/1	0.001
	Three-category			2.404	0.301
Brachialis	Age			14.860/1	<0.001
	Five-category			1.242/4	0.871
Gluteus muscles	Age	35.382/1	<0.001
	Five-category	3.627	0.459
Gluteus muscles	Age			0.070/1	0.791
	Three-category			6.249/2	0.044

Discussion

Testing methods on identified skeletal collections has become increasingly popular. This is due to the idea that normally unknown variables (in archaeological samples), such as sex and age at death, can be controlled for; this makes them ideal for testing the effects of these parameters. The study of EC associated with activity and labour patterns is amongst the most repeated analyses conducted on identified skeletal collection (Cunha & Umbelino, 1995; Mariotti et al., 2004, 2007; Alves Cardoso, 2008; Alves Cardoso & Henderson, 2010; Villotte et al., 2010; Niinimäki, 2011; Milella et al., 2012). However, researchers are increasingly becoming aware of the fact that identified skeletal collections may not be as illustrative of real life as previously thought (Hunt & Albanese, 2005; Alves Cardoso, 2008; Komar & Grivas, 2008; Henderson et al., in this issue). The concern is not only whether they are a representative sample of the population but the fact that the information provided in the documentary evidence associated with those collections omits many details that are relevant in a person's life and cannot be encapsulated in a death certificate. For instance, the documentary evidence does not include information on any changes in occupation during life (Cardoso, 2006; Caffell et al., 2012; Henderson et al., in this issue), age at which they began to work, any hobbies and their clinical history, and from a social and cultural viewpoint, it also fails to provide information on the historical and cultural settings in which these individuals were living. This archival and historical information (Cunha, 1995; Santos, 1999; Herring & Swedlund, 2003) is not readily associated with the skeletons, and therefore, further historical contextualisation of these individuals' life courses is necessary. Also, there is a growing concern that the research designs and the manner in which data and variables are coded may cause bias and compromise comparisons between studies (Perréard Lopreno et al., 2012).

Taking all these factors into consideration, the aim of this paper was to test the hypothesis that the frequency of EC is dependent on the method for categorising occupation. The results support this hypothesis. The categorisation criteria of the occupations into five, three or two occupation categories have shown that the significance of EC varies accordingly and, consequently, so does the interpretation of the results. This is sufficient to question the accuracy of past population reconstruction of behaviour and behavioural patterns, as well as sexual division of labour. However, the latter was not explored in this paper because of the limited data on female occupations, as previously discussed. This limitation is a problem for most European identified skeletal collections (Rocha, 1995; Mariotti et al., 2004; Cardoso, 2006; Alves Cardoso, 2008; Caffell et al., 2012; Milella et al., 2012). With regard to male individuals, the grouping criteria were either based on perceived activity levels (two-category and three-category) or social and economic hierarchies (five-category): all of which are inherently subjective. As seen in Table 1, when categories are collapsed from a five to a two-category group, specific occupations move between groups. For instance, stonemason, weaver and photographer are in the same category in the five-category (skilled workers/artisans); in the three-category group, stonemasons change from skilled workers/artisans to heavy manual workers, weaver to light manual and photographer to non-manual; in the two-category group, stonemason and weaver are considered manual and photographer non-manual. This change reflects perspectives on the interpretation of what occupation means, and it is based on the criteria used to categorise the occupations. In the end, the question of how to code and interpret the concept of ‘physical effort’ involved in occupation is dependent upon the social, economic and cultural settings within which it is interpreted. This same problem has also been found in gender studies (Fernandes, 2001; Geller, 2005; Sofaer, 2006; Alves Cardoso, 2008; Marques, 2009).

Within the current work, when EC are analysed by occupational category, without controlling for age, only infra- and supraspinatus (recorded as one enthesis) and triceps brachii remain significant throughout the three methods for grouping activity (Table 4). In the remaining cases, the significance levels vary according to entheses and categories. However, when age is considered as a predictor for EC presence, alongside occupation, it is clear that this association between EC and occupation was a false positive. In almost all cases in which statistical significance was found, age became the sole significant factor (Table 5). This observation is valid regardless of the occupation categorisation method; that is, it is not dependant on whether activity levels or socio-economic status are the underlying concepts used to create the categorisations. The working hypothesis that individuals described as non-manual workers would have a higher mean age at death was also supported (Table 2). The importance of age in EC studies is not a new finding, but its significant impact on EC has only recently has been taken into serious consideration (Cunha, 1995; Mariotti et al., 2004; Alves Cardoso, 2008; Alves Cardoso & Henderson, 2010; Villotte et al., 2010; Milella et al., 2012). Unfortunately, age assessment is extremely problematic in archaeological samples (Bocquet-Appel & Masset, 1982; Buckberry & Chamberlain, 2002; Falys, et al. 2006; Milner & Boldsen, 2012). Therefore, and faced with results that increasingly show the importance of age in EC interpretation, it is necessary to question the use of EC to reconstruct activity in archaeological populations (Jurmain et al., 2012). On the other hand, a simple association between EC and age fails to express the multifactorial aspect of EC formation (Jurmain et al., 2012), and simplistic interpretations of the association between EC and age (e.g. to develop a new ageing method) should be avoided (Tichnell, 2012).

The presumed straightforward interpretation of the differences between occupational categories and EC frequency as being due to socio-economic status is not possible. The clustering of specific occupations into larger occupational categories masks occupations that may be on the borderline or straddle socio-economic groups. For example, in the Portuguese context, individuals described as farmers may be representative of different socio-economic backgrounds and have differing occupational duties. The word lavrador/agricultor (farmer) could be used to name a number of different categories: tenant farmers, sharecroppers, landless day labourers or dependent poor. On the other hand, there are also farmers, called agricultores who are wealthy landowners, whereas in some cases, these individuals might be referred to as proprietário, that is, owner. (Roque, 1988; Vaquinhas, 1993; Lopes, 2003; Alves Cardoso, 2008). Additionally, it is possible that some of the higher status individuals worked their way up from manual to non-manual work, changing socio-economic status as they went. This could explain the lack of significance in EC frequency between the groups (Caffell et al., 2012). Occupation at death does not reflect all activities undertaken during life (Cardoso, 2005; Caffell et al., 2012) nor even the variability of occupations throughout life. Currently, only occupation at death is known for most identified skeletal collections, and this study demonstrates that further historical research is required to flesh out the lives of these individuals to fully understand the relationship between occupation, age and EC.

Conclusion

In summary, no occupational grouping method used in this study demonstrated a significant link with EC, whereas the results do reinforce the finding that age is a major confounding variable when studying any degenerative skeletal process. Studies of EC need to consider not only the recording methods used to assess EC but also the categories used to group the skeletons. It is necessary to choose an occupational grouping category appropriate to the research question prior to recording, analysis or interpretation of the data. Individuals described as non-manual workers and frequently associated with higher socio-economic status have higher mean values for age at death, and this may positively skew the EC frequency in these groups. The major conclusion of this paper is that the categorisation of occupation is a source of bias. Therefore, research should not blindly rely on occupation at death to test the relationship between EC and occupation.

Acknowledgement

This research was supported by the FCT — Fundação para a Ciência e a Tecnologia, reference: SFRH/BPD/43330/2008 and SFRH/BPD82559/2011.