Sensory and instrumental analysis of longitudinal and transverse textural variation in pork longissimus dorsi

MEAT SCIENCE Meat Science 68 (2004) 611–629 www.elsevier.com/locate/meatsci

Sensory and instrumental analysis of longitudinal and transverse textural variation in pork longissimus dorsi Stine Hansen a

a,*

, Thomas Hansen a, Margit Dall Aaslyng b, Derek V. Byrne

a

Department of Food Science, The Royal Veterinary and Agricultural University, Rolighedsvej 30, DK-1958 Frederiksberg C, Denmark b Danish Meat Research Institute, Magleg
ardsvej 2, DK-4000 Roskilde, Denmark Received 19 December 2003; received in revised form 19 May 2004; accepted 20 May 2004

Abstract In the present study sensory and instrumental analysis of the textural properties of pork longissimus dorsi were performed and analysed using multivariate data analytical methodologies. The aims were to determine how textural properties of musculus longissimus dorsi varied transversely, longitudinally, between left and right muscles, and as a function of ageing. By training the panel in the descriptive texture profile method the panellists were able to clearly detect, discriminate and describe the textural variation in the samples, except for the transverse variation. Sensory evaluation revealed a non-significant trend (P > 0:05) of decreasing tenderness from dorsal (nearest the spinal column) to medial sampling position. When analysed instrumentally, using a modified Warner–Bratzler shear force apparatus, the transverse muscle variation was found to increase in force (N) required for deformation from dorsal to lateral sampling position. Overall, these two methods agreed in that the dorsal position was the more tender of the three positions investigated. Lengthwise muscle variation was highly defined when assessed using sensory analysis, whereas instrumental analysis was unable to detect this variation. The sensory analysis revealed that tenderness decreased and hardness increased as the sample position approached the caudal end of the muscle. Additionally, sensory analysis revealed a major turnover point in textural properties at the end of the ribcage area in that tenderness showed a marked decrease in this area and hardness and juiciness increased correspondingly in the same area. Both sensory and instrumental analysis showed that muscles from left and right side of the carcass differed significantly (P < 0:05) in their textural properties. The right side muscle was clearly defined into stages more so than the left side muscle. In addition, the right loin was found to be harder per se than the left loin, which was postulated to be caused by a greater amount of work performed by right muscles compared to left muscles. Significant differences (P < 0:05) in sensory textural attributes were observed overall with increased ageing. The variation within muscles, which contributed to the overall change in texture with ageing, was found to be due to changes in the longitudinal variation, in that the individual chop variation observed at the cranial end became less pronounced when the meat was aged. Differences observed between the cranial and caudal end remained, regardless of ageing for 4 or 7 days. No changes could be seen in the transverse texture during ageing. In general, sensory and instrumental analyses were found largely to be predictive indexes of each other. However, these two methods could not be said to be causally predictive in that they did not measure the same physical properties of the meat. For instance sensory determined tenderness is a result of the type and rate of deformation and the heterogeneity of the sample assessed, whereas instrumental measurement is a result of resistance to shearing. The present study showed significant variation between longitudinal locations and this variation is critical when designing sensory texture profiling experiments of meat from loins. Moreover, the textural differences between left and right loin muscles must be considered when texture prediction is the objective if an applicable conclusion is to be drawn. Ó 2004 Elsevier Ltd. All rights reserved.

1. Introduction *

Corresponding author. Tel.: +45-3528-3274; fax: +45-3528-3190. E-mail address: [email protected] (S. Hansen).

0309-1740/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.meatsci.2004.05.013

Textural properties are of great importance in pork meat when it comes to consumer acceptance, and as

612

S. Hansen et al. / Meat Science 68 (2004) 611–629

such texture measurements in this context need to be investigated in a manner that clearly specifies what is meant by sensory texture as opposed to instrumentally measured texture. Only then it is possible to define what is meant by the textural variation in a meat versus the aspects of this textural variation that can be utilised for quality prediction through instruments. When measuring meat texture two methods are usually employed, namely sensory evaluation and instrumental determination in terms of compression. Instead of using these two methods as a reflection of each other it should be emphasised that these methods measure different properties of meat and therefore may result in low correlations. Compression is considered qualified for prediction of sensory perception of consistency during mastication, in that, food is destroyed when chewed upon. This method, however, is not a good predictor of a foods’ breaking characteristics. Often a local maximum will be seen in the curve, which is normally defined as force at breaking. However, in practice this is not the case, in that, the material starts to break long before this exact point (Qvist, 1998). Rosenvold, van den Berg, Andersen, Johansson, and Lundstr€ om (2002) emphasise that instrumentally determined tenderness only resembles bite resistance to shearing, whereas human textural perception is more complex. Poor correlations between sensory and instrumental determinations of tenderness are due to different deformations, deformation rates, and the intrinsic heterogeneity of the biological material (Spadaro, Allen, Keeton, Moreira, & Boleman, 2002). Texture testing instruments are calibrated to respond linearly to the intensity of the tested mechanical property, which does not apply in the human perception of texture (Szczesniak, 1987). According to Spadaro et al. (2002) linearity will only occur if the biological material is homogeneous. Tornberg (1996) suggests that the nonlinearity between sensory and instrumental assessments is due to non-linearities in sensory analysis and the fact that muscle fibre orientation is easier to control in instrumental than in sensory evaluations. Previous texture investigations have shown that m. longissimus dorsi varies in tenderness lengthwise as well as transversely. Weir (1953) investigated the variation in m. longissimus dorsi tenderness lengthwise using both Warner–Bratzler shear force (WBSF) and an experienced sensory panel. Overall, Weir (1953) found that the anterior and posterior parts of m. longissimus dorsi were more tender than the central parts. Møller and Vestergaard (1986) investigated effects of carcass suspension on pork tenderness in three locations of the loin. Warner–Bratzler shear force was influenced by sarcomer length and they found that location 1 (1st to 4th lumbra vertebra) was least tender and location 3 (9th to 11th rib) was most tender for both normal meat and meat exposed to pelvic suspension. This is in agreement with

the findings of Weir (1953). Rust, Olson, Shuler, and Thomson (1972) studied the positional differences existing between chops in the medial loin region (4 chops, 2.5 cm in thickness, were sliced anterior and posterior to the last rib) WBSF and a trained sensory panel for the evaluations. They found no significant differences lengthwise of the medial part of the loin. This is in conflict with the findings of both Weir (1953) and Møller and Vestergaard (1986). Alsmeyer, Thornton, and Hiner (1965a) investigated transverse differences of tenderness in pork m. longissimus dorsi. Slice tenderness evaluator (STE) puncture readings showed that the lateral position in pork was more tender than the medial position. Alsmeyer, Thornton, and Hiner (1965b) investigated six locations from pork m. longissimus dorsi (two each from the dorsal, medial and lateral areas) to see if there were any transverse tenderness variations. Again they found that the lateral locations were most tender, and the medial locations the least tender. Contrary to these findings Rust et al. (1972) found, that dorsal positions were significantly more tender than the lateral positions. This is in agreement with the findings of Onate and Carlin (1963) who found that m. longissimus dorsi were significantly more tender near the spinal column than at the outer edge when measured instrumentally. Sensory and instrumental analysis carried out under similar conditions, e.g., similar cooking methods and temperature regimes give high comparability and thus, higher correlations (Szczesniak, 1987). The degree of cooking has a large effect on the toughness of meat, which makes the degree of doneness essential in this type of study (Rosenvold et al., 2002; Wheeler, Shackelford, & Koohmaraie, 1997). Furthermore, the importance of muscle fibre orientation for instrumental as well as for sensory evaluation of tenderness has been clearly demonstrated by Otremba et al. (1999), who found that blade type and coring methods influenced correlations between sensory scores and WBSF values. Thus, samples should be presented and assessed with the same muscle fibre orientation to obtain the highest relationships between WBSF measurements and sensory scores. With respect to sensory descriptive texture profiling the serving order is essential to obtain results useful for valid interpretation. According to Meilgaard, Civille, and Carr (1999) using a balanced block design minimizes the interactions between panellists, samples, and time. Every panellist should evaluate all samples on an equal number of occasions and the samples should be served in a balanced manner before and after every other sample. The objective of the present study was to investigate the texture variation in pork m. longissimus dorsi (LD) through sensory and instrumental analysis. Specifically, the aims were to determine how the transverse and longitudinal texture varied within and between left and

S. Hansen et al. / Meat Science 68 (2004) 611–629

right muscles. A further objective was to determine the dynamic nature of texture variation occurring in and between left and right muscles of pigs and as a function of ageing. Overall, the predictive and causal association of these methodologies was considered paramount, such that confusion and misrepresentations as to what textural variation in pork meat means from a sensory as opposed to an instrumental perspective.

2. Materials and methods 2.1. Animals Pigs were selected at random from an abattoir production line (Danish Crown, Ringsted, Denmark). From a group of 58 pigs 31 were selected with regard to weight at slaughter, meat percentage, and ultimate pH. All animals weighed between 73 and 79 kg at slaughter (warm carcass weight), contained 58–62% lean meat (Commission Regulation (EC) No. 3127/94, 1994), and had an ultimate pH (pHu ) between 5.5 and 5.8. The ultimate pH was measured between the 4th and 5th or between the 5th and 6th lumbar vertebra in both left and right side of the cold carcasses approximately 20 h after slaughter using a Knick Portamess 751 pH-meter (Knick GmbH & Co., Berlin, Germany) and a direct insertion probe electrode (Ingold lot 406-M3, Mettler Toledo,

613

Urdorf, Switzerland). Both left and right LD from the pigs, a total of 62 muscles, were excised. During excision subcutaneous fat and bones were removed. For sensory and instrumental analysis a total of 54 LD muscles were used. 2.2. Sample variation The LD muscles (average weight 2.8 kg) were removed from both sides of the carcass 22–24 h after slaughter, divided into two halves at the 13th rib (cranial and caudal ends) and vacuum packed (R€ oshermatic CE94, Fomaco, Køge, Denmark) in oxygen impermeable bags. The muscles were aged on a polystyrene tray for 0, 4, and 7 days, respectively, at 4 °C before they were frozen at )20 °C. The start of ageing was from the time of excision. Muscles were held for a maximum of 4 weeks in frozen storage. An overview of sample variation is presented in Fig. 1(a) and (b). It should be noted that when the left LD from one animal was used for sensory analysis the right LD from the same animal was used for instrumental analysis and vice versa. 2.3. Sample preparation The LD muscles were taken from the freezer approximately 24 h before cooking. To avoid a long period of thawing the cranial LDs were placed at room

Fig. 1. (a) A total of 54 loins were evenly divided into three ageing periods (0, 4, and 7 days) and over two analytical methods (sensory and instrumental). Both left and right loins from the same animal were aged equally. One side was used for sensory profiling while the other was used for instrumental analysis. Numbers refer to chops cut from the cranial and caudal ends at 20 mm intervals, the 9th and the 10th chop cut from each side of the 13th rib. (b) On each chop there are three sample positions (excised in different ways depending on method of analysis) denoted a (dorsal), b (medial), and c (lateral). All panelists were exposed to all sources of variation in a balanced block design.

614

S. Hansen et al. / Meat Science 68 (2004) 611–629

temperature for 90 min and the caudal LD muscles were placed at room temperature for 120 min as they are larger (average weight of cranial and caudal ends were 1.33 and 1.45 kg, respectively). After this initial thawing period the LD muscles were placed at 4 °C for 22 h (Gram F 425 refrigerator, Gram A/S, Vojens, Denmark). The LDs’ were unwrapped, dried on the surface with cloth towels and weighed (Satorius 43600 balance, Goettingen, Germany). When a core temperature above 0 °C was reached, as determined with a Testo 926 digital thermometer (Testoterm, Buhl & Bundsoe, Virum, Denmark), both the cranial and caudal ends of the loins were sliced (Bizerba VE8 slicer, Bizerba, Germany) into 20 mm thick chops (each weighing approximately 128 g) from the 13th rib upwards and downwards (producing 9 chops in each direction). The chops were placed on a plastic tray, covered with a plastic bag to prevent surface drying, for up to 90 min at 4 °C prior to cooking. 2.4. Sample heat treatment The pork chops were pan cooked (Friberg FBLB 40/6, Friberg Verkst€ ader AB, Sweden) in neutral oil (grape seed oil) at 155 °C for approximately 8 min until a core temperature of 65 °C was reached as determined by a Testo 926 digital thermometer. The chops were turned every 2 min to ensure even cooking. 2.5. Sample excision The transverse sample cores for the sensory texture analysis (see Fig. 1) were excised by using a template (4
5 cm) and then divided into three equally sized pieces (approximately 17 mm wide, 20 mm thick, and 40 mm long). The cores from each chop were identified according to their position on the chop from inner to outer edge, a (nearest spinal column, dorsal), b (medial), and c (lateral). Plates were marked with random numbers and preheated in an Electrolux ar 10 esp/menu oven (Electrolux Stork€ ok AB, Alings
as, Sweden). For instrumental analysis transverse sample cores were excised parallel to the muscle fibre orientation to

the chop surface as recommended in AMSA (1995). The three sample positions from each chop were excised using a scalpel. Each piece was cut parallel to the muscle fibre orientation to the chop surface with a thickness of 10 mm. The cut surface was then placed on the table and a 20 mm wide piece was excised from the middle still parallel to the fibres (Fig. 2). The cores were identified as for sensory texture analysis. 2.6. Sensory descriptive texture profiling 2.6.1. Training The assessors (8 females and 1 male, 45–62 years of age) were trained with reference to the texture profile method (Meilgaard et al., 1999; ISO 11036, 1994). Training consisted of three sessions of approximately 2 h each. All assessors had previously been part of a meat-profiling panel and had a large amount of texture profiling experience with cooked meat products (Danish Meat Research Institute (DMRI), Roskilde, Denmark). Because of the panels previous experience with meat texture profiling, training was limited to ensuring panel agreement on descriptor definitions and ensuring that the assessors were able to detect the inherent variation present in the samples. In the training phase the assessors were instructed in biting with molar teeth at the cut chop surface. The panel was provided with a list of eight terms for objective measurement. The eight attributes and their definitions are presented in Table 1. The list of attributes was developed from the literature (ISO 11036, 1994; Meilgaard et al., 1999). The assessors were all familiar with the use of 150 mm, unstructured line scales, thus these scales were introduced from the beginning of the training. The line scales were anchored to the left with ‘none’ and to the right with ‘very much’. A data collection system for automatic acquisition of the assessor scores was used (FIZZ, Bio systems, France). This scoring method was used throughout all training sessions. The design variation of the samples appears in Fig. 3. For all training sessions subsets of the total variation were used (see Fig. 4(a)–(c)). For session 1 and 2 sam-

Fig. 2. Schematic representation of the method of core excision for instrumental analysis. The samples were cut parallel to the muscle fibre orientation to the chop surface with a thickness of 10 mm. The cut surface was then placed on the table and a 20 mm wide piece was excised from the middle still parallel to the fibres. The samples were measured perpendicular to the fibre orientation. The bold lines represents the (fried) chop surface.

S. Hansen et al. / Meat Science 68 (2004) 611–629

615

Table 1 List of descriptive sensory terms presented in order of appearance Terma

Sensory definition

Hardness Juiciness Fibrous Tenderness Cohesiveness Crumbliness Chewiness Mastication remnant

Force required to bite completely through the sample Amount of wetness/juiciness released from sample Amount of fibres appearing during mastication Easiness with which the meat is divided into fine particles during mastication Degree to which sample returns to original shape after a certain time period Impression of grittiness and dryness Length of time required to masticate a solid product into a state ready for swallowing Amount of particles left in mouth at a state ready for swallowing

a

Described in ISO 11036 (1994) and in Sensory evaluation techniques (Meilgaard et al., 1999).

Fig. 3. The overall sample set for each left and right loin. The numbers designate ageing periods (0, 4, and 7 days), big letters designate cranial (N, neck) and caudal (H, hip) regions, and small letters designate transverse sample position (a, b, and c).

ples, which spanned the main sources of variation present, were used, as the main aspect was to familiarise the panellists with the descriptors and to ensure that they understood the use of the line scale. 2.6.2. Profiling Sensory descriptive texture profiling (ISO 11036, 1994; Meilgaard et al., 1999) was carried out over six 2-h sessions by the trained panel (see Section 2.6.1). All

sessions took place on weekday mornings in the sensory laboratory at The DMRI, Denmark, which is arranged according to the international standards ISO 8589 (1988) and ASTM (1986). The design was balanced between the different ageing periods (0, 4, and 7 days), between left and right side muscles, and between cranial and caudal ends. A total of 54 chops from 27 different animals were assessed during profiling by each panellist. From each ageing period nine muscles, five left and four right side muscles were evaluated. Each assessor evaluated 2 chops from each muscle, one from the cranial and one from the caudal region. Each chop contained all three transverse sample positions dorsal, medial, and lateral. The sample cores were placed on heated plates to maintain temperature and evaluated immediately. Overall the balanced block design ensured an even distribution of samples, in that each of the panellists assessed five different chops from the cranial and the caudal region, respectively, for left muscles and four different chops from the cranial and the caudal region, respectively, for right muscles. Thus, all panellists were exposed to all experimental sources of variation. 2.7. Instrumental texture using modified Warner–Bratzler shear force Prior to transverse sample core excision for instrumental analysis the chops were wrapped in commercially

Fig. 4. Subsets used for (a) Training session 1, (b) training session 2, and (c) training session 3. The solid circles (Þ refer to the samples used for training. For codes see Fig. 3.

616

S. Hansen et al. / Meat Science 68 (2004) 611–629

available aluminium foil and rested for 4 h at room temperature. After 2 h the chops were turned in order to absorb as much exudate as possible. The samples were treated in this manner to ensure an equalisation in temperature and amount of exudate remaining, prior to measuring. After resting excess cooking exudate was removed and the chops were wrapped in new foil until excision and instrumental measuring. The sample cores were excised as previously described (see Section 2.5). The samples were measured in order from a (dorsal), b (medial), and c (lateral), using a modified WBSF apparatus. The modified WBSF consisted of a Texture Analyser (Stabel Micro Systems, UK) modified with Volodkewich shear blade (flat). It was run at a constant chart speed of 24 mm min1 during measuring. Samples were sheared perpendicular to the fibre direction. The maximum force (N), the distance (mm) at the maximum force, and the force (N) required for 80% compression (8 mm) were used as measures of the meat hardness. 2.8. Data analytical methods Generalised Procrustes Analysis (GPA, Gower, 1975) was performed on the raw data from the final training session using Matlab 6.5 (MathWorks Inc., USA). From GPA the assessor mean term correlation vector lengths were extracted. The derivation of the aforementioned vectors provided co-ordinates for each individual assessors terms in the group average configuration (Dijksterhuis & Gower, 1991). This enables separate biplots to be made for each sensory term. The assessor mean term correlation vector lengths shown in the biplots gives information about the assessors’ use of the actual term. A tight cluster of vectors, all pointing in the same direction, indicates high agreement on the meaning of the specific term. A scattering of the vectors pointing in different directions indicates a low agreement for this particular term. Generally speaking the lengths of the vectors are proportional to the ease of scoring, where a long vector indicates that the assessor was confident in scoring the actual term and a short vector alternately may indicate less certainty (Dijksterhuis & Gower, 1991; Byrne, O’Sullivan, Dijksterhuis, Bredie, & Martens, 2001). Furthermore, GPA was also performed on the raw data from the profiling sessions in order to level correct for the assessors use of line scales. Partial least squares regression (PLSR) is a bilinear, data analytical method, which allows investigation of multivariate data. It deals with two matrices designated X and Y, respectively. The principle of PLSR is to relate the variation in the response variables (Y-variables) to the variations of several predictors (X-variables), the purposes being explanatory or predictive. When performing PLSR the initial score vectors from X and Y are used to predict the Y-variables via linear regression. The information present in the first principal component

(PC) is subtracted and subsequent PCs extracted from the data. This is done such that the variation in X relevant for predicting Y is determined (Esbensen, 2001; Martens & Martens, 2001). This results in the first components being those most relevant for predicting Y-variables. Interpretation of the relationship between X- and Y-data is then simplified in that this relationship is concentrated on the smallest possible number of components (Esbensen, 2001). The explained variance is utilised to determine the optimum number of PCs to be used in the model. Additionally, the validated variance can be used as a way to determine how well a single variable is taken into account in an analysis. Validation techniques, e.g., crossvalidation, allow for estimation of the prediction error in future predictions expressed by a validation variance (Esbensen, 2001). In the present study, a special use of ‘two block’ PLSR known as APLSR (the ‘ANOVA like use of PLSR’) was utilised. In APLSR the X-variables are 0/1 experimental design variables (for products, assessors, replicates etc.), whereas the response data matrix, be it sensory and/or instrumental data, is set as the Y-matrix. This form of PLSR projects the response variables onto the design variables in order to determine to which degree each of the design variables in X contribute to the variation in the response variables Y (Martens & Martens, 2001). For contextual validation in the regression analyses, the conventional loading plots were replaced by plots of correlation loadings. This allowed easier interpretation since it revealed both the structures in the data and their degree of fit at the same time (Martens & Martens, 2001). The contribution from certain variables can be pacified in order for them to be non-influential on the remaining variables in the correlation loading plot. This is a purely graphical aid, which eases the interpretation of the plots (Martens & Martens, 2001). Projection methods (as PLSR) take advantage of variances and co-variances to build models where the influence of a variable is determined by its variance, and the relationship between two variables may be summarised by their correlation. While variance is sensitive to weighting, correlation is not. This provides us with a possibility of still studying the relationship between one variable and the others, while limiting this variables influence on the model. This is achieved by giving this variable a very low weight in the analysis (Martens & Martens, 2001). In relation to significance testing at the 5% level, i.e., P < 0:05, a re-sampling technique termed ‘jack-knifing’, which is part of cross-validation, can be utilised. Jackknifing allows for significance testing of the relationship between the X- and Y-matrices. Uncertainty limits of 2 standard uncertainties for each of the estimated variables are subsequently determined (Martens & Martens, 2001). All multivariate analyses were performed using the Unscrambler Software, Version 8.0 (CAMO ASA,

S. Hansen et al. / Meat Science 68 (2004) 611–629

Trondheim, Norway). In all regression analysis level corrected data were analysed centred with full crossvalidation and with both the X- and Y-matrices standardised (unless otherwise stated) in all cases. Level correction was performed by using GPA such that the relative difference between samples were retained (Dijksterhuis & Gower, 1991). 3. Results 3.1. Data analysis strategy Initially, multivariate APLSR and GPA analysis were performed on the sensory data from the training to gain insight as to the validity of the sensory terms used for the descriptive texture profiling. Subsequently, APLSR was carried out to investigate the associations between sensory data and relevant design variables in order to investigate the transverse, longitudinal, and ageing variation individually. Additionally, univariate line-plotting was performed to illustrate the textural variation of LD lengthwise. Furthermore, APLSR of the total variation was performed to determine if overall, the variation between left and right side muscles was significant. In relation to instrumental analysis multivariate APLSR was carried out to investigate the relationships between the design and the instrumental data. Moreover, the predictive relationships between sensory and instrumental data were determined directly by PLSR. 3.2. Training vocabulary validity over sessions The analyses of the training data were aimed at determining if the sensory terms chosen fulfilled the various criteria for descriptor reliability as set down by Byrne, Bak, Bredie, Bertelsen, and Martens (1999a) and subsequently their applicability was investigated by Byrne, Bredie, and Martens (1999b, 2001). These criteria and how they can be assessed and fulfilled are presented in Table 2. The main aim was to determine the reliability and validity of the terms that were present in training and subsequently used in profiling (for descriptors see Table 1). The design variation is shown in Fig. 3. For all training sessions subsets of the design variation were used (Fig. 4(a)–(c)). Table 2 Criteria for descriptor reliability Criteriaa

Fulfillment assessed by

1. Be relevant to the product

Prior knowledge/panel discussion/APLSR APLSR/GPA Prior knowledge/panel discussion/GPA/APLSR

2. Discriminate between samples 3. Have cognitive clarity a

Byrne et al. (1999a, 1999b) and Byrne et al. (2001).

617

APLSR multivariate analysis was performed to determine if the sensory panel improved in their understanding of the sensory vocabulary utilised in training over the three sessions. A validated explained variance was found for the models based on session 1 (15%) and 3 (39%), respectively (not shown). These were used as indications of cognitive clarity. The validity of the models, as indicated by the validated explained variance, increased with training, meaning that the sample set was assessed with greater reliability. The validity of the models increased as the training progressed, indicating the panel behaved uniformly and were in greater agreement on the use and meaning of the descriptors. To gain information on the cognitive clarity of the descriptor list for texture profiling, in a data analytical sense, GPA was performed. From GPA the assessor mean term correlation vector lengths were extracted and plots of PC1 versus PC2 were derived from each term (not shown). The panel showed high agreement for the terms tenderness, hardness, fibrous, mastication remnants, and chewiness. In general the vectors were oriented in the same direction and were of similar lengths, indicating a thorough understanding of the terms and a confidence in the assessment. All but one assessor had scored cohesiveness in the same way, indicating a general understanding of this term by the panel, though another assessor seemed uncertain in scoring the term (short vector pointing in the same direction as the others). For the term crumbliness two assessors seemed to have misunderstood the term. The biplot for juiciness showed a scattering of the vectors, which indicate a low understanding of the term. Of the nine assessors six scored similarly, however only one showed confidence doing so. The remaining three assessors clearly had a different interpretation of the terms meaning. Overall training session 1 did not meet the various criteria for descriptor reliability. However, in the final training session the majority of the sensory terms (tenderness, hardness, fibrous, mastication remnant, and chewiness) appeared relevant for the product variation and were able to discriminate between samples and showed cognitive clarity. Of the remaining terms both cohesiveness and crumbliness generally fulfilled the various criteria. Discrimination and cognitive clarity and thus, understanding could not be ascribed for the term juiciness from either APLSR or GPA. This indicated that the training was successful in that the panellists were able to describe the samples in a validated manner. 3.3. Sensory descriptive texture profile analysis 3.3.1. Transverse muscle variation APLSR analyses were carried out for left and right loins separately within each ageing period (0, 4, and 7 days) to relate the design (averaged over individual chop numbers) and the sensory responses (level corrected for

618

S. Hansen et al. / Meat Science 68 (2004) 611–629

assessors). All variables were standardised to 1=Sdev . The models explained 39% and 24% validated variation in the first two PCs, respectively. The correlation loading plots (PC1 versus PC2) for both left and right loins can be seen in Fig. 5(a) and (b), respectively. These plots showed that the sensory terms and 0 days of ageing spanned PC1. The ageing periods 4 and 7 days and transverse positions (a, b, and c) spanned PC2. These overall trends were revealed in both left and

right loins. By jack-knife uncertainty testing of the estimated regression coefficients in the APLSR model ageing was found to be significant (P < 0:05), whereas transverse positioning did not show any significant difference. Transverse variation seemed more distinct in right loins than in left loins, however both displayed the same trends in variation such that, juiciness and tenderness were highest in the dorsal position (a) and lowest in the medial position (b) (results not shown).

Principal Component 2 (Y-explained variance 1%)

1.0

7 days

0.5 a

b 0 days

Juiciness

0.0

Mastication remnant Chewiness Cohesiveness

Crumbliness

Tenderness

Hardness Fibrous

c

-0.5 4 days

-1.0 -1.0

-0.5

0.0

0.5

1.0

Principal Component 1 (Y-explained variance 39%)

(a)

Principal Component 2 (Y-explained variance 1%)

1.0 7 days

0.5

b c

0.0

Hardness Chewiness Mastication remnant

Tenderness

0 days

-0.5

a 4 days

-1.0 -1.0

(b)

Crumbliness Juiciness

Cohesiveness Fibrous

-0.5

0.0

0.5

1.0

Principal Component 1 (Y-explained variance 24%)

Fig. 5. Transverse sample variation in (a) left and (b) right LD. ANOVA partial least squares regression (APLSR) correlation loading plot of the first two principal components (PCs). Main design variables (cranial/caudal and ageing) and the sample variables (position a, b, and c) in the X-matrix and sensory terms in the Y-matrix. Ellipses represent r2 ¼ 50 and 100%.

S. Hansen et al. / Meat Science 68 (2004) 611–629

3.3.2. Longitudinal muscle variation APLSR analyses were carried out for left and right loins separately within each ageing period (0, 4, and 7 days) to relate the design (averaged over transverse positions within individual chops) and the sensory responses (level corrected for assessors). All variables, with the exception of cranial and caudal, were standardised. The contribution from the variables cranial and caudal was pacified. It was deemed reasonable to average over transverse sample positions, as no significant differences were found within the chops cross-sectional area (Fig. 5). Furthermore, xy-line-plots for single variables displaying ratings of tenderness, hardness and juiciness as a function of chop number were derived in order to visualise the trends in lengthwise variation. Of these figures only tenderness was displayed. Again data was averaged over sample position (a, b, and c). In addition, the data from left and right loins were averaged within each ageing period, respectively (five left and four right loins per ageing period). This was performed in order to see if there were any marked differences between left and right loins for each ageing period. The APLSR models explained 14% and 15% of the variation in total in the first two PCs, respectively. Fig. 6 displays the correlation loading plots (PC1/PC2) for left and right loins, respectively. For both sides in general it can be seen that increased ageing, described by the majority of the sensory terms (i.e., tenderness to hardness, fibrous, chewiness and mastication remnant) spanned PC1. PC2 was found to be explained by the longitudinal variation from cranial to caudal, as described by crumbliness to cohesiveness, respectively. Moreover, across PC2 the individual chop variation from cranial to caudal, chop 1–18, is apparent. It was clear that the cranial end was significantly (P < 0:05) more tender than the caudal end. Through investigation of the regression coefficients (results not shown) for crumbliness and cohesiveness, it seemed that these were the aspects of tenderness and hardness that separated the cranial and caudal ends of the LD. The caudal end was in general more cohesive and less crumbly than the cranial end. Moreover, the caudal end was more chewy, fibrous, hard, and left more mastication remnants when ready for swallowing, than the cranial end. In addition, through interpretation of the orthogonallity of the correlation loading plot it was apparent that ageing and longitudinal variation were interacting. Such that ageing had a larger effect on the cranial end than on the caudal end in terms of promoting tenderness. The individual chops in the left and right sides generally showed a gradual change in tenderness from the cranial to caudal end of the loins (Fig. 6). Moreover, this stepwise fall in tenderness from the cranial to the caudal end appeared to be more clearly defined in the right side into a number of stages com-

619

pared with the general decrease in tenderness in the left side from the cranial to the caudal end. Upon specific examination of this individual chop variation in relation to tenderness, through line-plotting, the overall decrease in tenderness which occurred when the sample location approached the caudal end in both left and right side muscles was again clearly seen (Fig. 7). Also as with the APLSR plots (Fig. 6) left and right loins had a general decrease in tenderness, but the decrease was much higher and more clearly defined into stages in the right loins (Fig. 7(a) versus (b)), respectively. A greater decrease in tenderness between individual chops was seen in the cranial compared to the caudal end at 0 days of ageing particularly. When the meat was aged, larger individual chop variation occurred in the caudal end. It seemed that tenderisation occurred equally in the caudal region, but became less pronounced in the cranial region when the meat was aged. Overall for all three ageing periods the first 7 chops were rated similarly in tenderness within each ageing period. At chop 8, a marked decrease in tenderness was apparent and from chop 11/12 onwards the tenderness was rated as being at a constant or slightly increasing level (see Fig. 7(a) and (b)). As tenderness decreased hardness was found to concurrently increase (results not shown). For all three ageing periods an overall increase in juiciness occurred when approaching the caudal end, the only exception being non-aged, right side muscles, which overall decreased in juiciness (results not shown). In general the cranial end was most juicy at 0 days but decreased with ageing. The caudal end was least juicy at 0 days of ageing but increased with ageing. A marked change in juiciness occurred between chop 8 and 10, which was also the case for tenderness and hardness. However, this showed an opposing trend compared to tenderness, in that juiciness increased. Bigger variation between individual chops was seen for juiciness compared to tenderness. The variation in tenderness cannot be explained totally by the variation in juiciness. Therefore, these descriptors reflect differing aspects of similar variation. 3.3.3. Ageing APLSR analyses were carried out for left and right loins separately to relate the design factors 0, 4, and 7 days to the sensory terms. The models explained 84% and 52% variation in the first two PCs of the left and right sides, respectively. In both left and right side non-aged meat samples (0 days) were significantly (P < 0:05) harder, more chewy, fibrous, and had a higher level of mastication remnants left when ready to swallow, than aged meat. Meat aged for 4 and 7 days, respectively, were largely equal in tenderness, thus no significance could be ascribed.

620

S. Hansen et al. / Meat Science 68 (2004) 611–629

However, the APLSR correlation loadings plot (Fig. 8) indicated that in general the variable crumbly may differentiate between 4 and 7 days of ageing, 7 days being most crumbly. Non-aged meat was found to be least juicy. Similarity in ratings of juiciness were observed at 4 and 7 days of ageing. As mentioned earlier tenderness increased in both the cranial and caudal ends of the LD, but it seemed that the ageing effect masked some of the individual chop variation in the cranial end, whereas the

individual chop variation slightly increased in the caudal end. Overall a similar increase in tenderness occurred at both the cranial and caudal ends when aged, meaning that these remained different regardless of ageing within this interval. In relation to ageing, left side muscles had a higher explained variance than right muscles. This is because the variation in the cranial and caudal ends was explained to a greater degree in the right side, as seen in

Principal component 2 (Y-explained variance 1%)

1.0

C11

0.5 C8

C18 C16 C1 C3 Caudal C13 Mastication remnant 7 days Cohesiveness Chewiness Tenderness 4 days C6 C17 Fibrous Crumbliness C4 Hardness C12 C2 Cranial C7 C10 C15 C9 C14 C5 Juiciness

0.0

0 days

-0.5

-1.0 -1.0

-0.5

0.0

0.5

1.0

Principal Component 1 (Y-explained variance 12%)

(a)

Principal Component 2 (Y-explained variance 1%)

1.0

Caudal

0.5

Crumbliness

C5 C6 Juiciness

0.0

Tenderness 4d ays

C4

C13 C15 C12 C18 Fibrous C17 C10 C11

C7 C9

C1

Chewiness Hardness Mastication remnant

Cohesiveness C16

C2 C8

0 days

C3

-0.5

Cranial

-1.0 -1.0

(b)

C14

7 days

-0.5

0.0

0.5

1.0

Principal Component 1 (Y-explained variance 12%)

Fig. 6. Longitudinal sample variation in (a) left and (b) right LD. ANOVA partial least squares regression (APLSR) correlation loading plot of the first two principal components (PCs). Main design variables (cranial/caudal and ageing) and the sample variables (various chops) in the X-matrix and sensory terms in the Y-matrix. Ellipses represent r2 ¼ 50 and 100%.

S. Hansen et al. / Meat Science 68 (2004) 611–629 14

621

14

13

13

4 days

12

4 days

12

11

11

10

10

Tenderness

Tenderness

7 days

9 8

0 days

7

9 8 7

6

6

5

5

4

0 days

4

1

(a)

7 days

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18

Chop number

1

(b)

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18

Chop number

Fig. 7. Longitudinal variation of tenderness in (a) left and (b) right loins within each ageing period. The dotted line (  ) lines represent 0 days of ageing. The solid ( – ) and the dashed (- - -) lines represents 4 and 7 days of ageing, respectively. Data is averaged over transverse sample positions (a, b, and c) within each chop. Arrows indicate the change point at chop 8.

the longitudinal analysis. Thus, as the variation from the cranial and caudal end was excluded from modelling, the right side had less variation associated with ageing (Fig. 8). Qualitatively both sides were similar, in that the same descriptors are associated with aged and non-aged samples. Quantitatively a little more distinction in the variables in the right loins seemed to occur. 3.3.4. Sensory determined textural variation in longissimus dorsi based on the total design APLSR analysis was carried out using the total design variation (transverse sample positions, longitudinal chop variation, cranial and caudal end, ageing and left and right side LD) to visualise the variation between left and right side muscles in relation to the sensory responses. All variables, with exception of cranial and caudal, were standardised. The contribution from the variables cranial and caudal was pacified. The main aim here was to gain an overview of all sources of variation in relation to the sensory variation seen in the present study. The model explained a total of 11% variation in PC1 and PC2. Jack-knife uncertainty testing revealed that most of the design factors differed significantly (P < 0:05), the only exceptions being the transverse sample positions and a few individual chops. In the previous sections, each of the main design factors (transverse variation, longitudinal variation, variation as per ageing) were analysed individually to determine how the texture of m. longissimus dorsi varies in relation to the sensory responses. The majority of the variation as described in the previous sections are in general clearly present in the APLSR correlation loadings plot based on all design variables and the sensory descriptive terms (Fig. 9).

In Fig. 9, displaying the correlation loading plot (PC1/PC2) for the total design variation, it can be seen that increased ageing, described by the majority of the sensory terms (i.e., higher levels of tenderness for aged meat to hardness, fibrous, chewiness and mastication remnant describing non-aged meat) spanned PC1. PC2 was found to describe the longitudinal variation from cranial (N) to caudal (H), i.e., a decrease in tenderness. Moreover, across PC2 the individual chop variation, from chop 1 to 18, was apparent. Transverse variation showed a trend (non-significant) for the dorsal position (a) to be the most and the medial position (b) the least tender. It appeared from both Figs. 6 and 9 that chops from the anterior part of the LD were more tender than chops from the posterior part. Additionally, both plots revealed that individual chop variation occurred. Fig. 9 showed two clusters of chops, representing cranial and caudal end, respectively. The cranial and caudal end and most of the individual chops (chop 1, 3–6, 8, 10–12, 14–18, a total of 14 out of 18 chops) turned out to differ significantly in the model based on the full design. What was not apparent in this model was how the variation occurred from chop 1–18 in left and right LD, respectively, where the major turn-over point in texture took place, and how ageing affected tenderness ratings in the cranial and caudal end of left and right LD, respectively. It was apparent that ageing and longitudinal variation were interacting as described earlier. Description of the sensory aspects of ageing appeared from Fig. 9 (as it did in Figs. 6 and 8) in that 0 days of ageing resulted in the highest ratings of hardness, chewiness, fibrous, and mastication remnant. Meat aged for 4 days seemed a bit more tender than meat aged for

S. Hansen et al. / Meat Science 68 (2004) 611–629

Principal Componenet 2 (Y-explained variance 1%)

622

1.0 4 days

0.5

Tenderness

0.0

Crumbliness

Cohesiveness Chewiness Mastication remnant

Juiciness

0 days Fibrous

Hardness

-0.5

7 days

-1.0 -1.0

-0.5

0.0

0.5

1.0

Principal Componenet 1 (Y-explained variance 83%)

(a)

Principal Componenet 2 (Y-explained variance 1%)

1.0 4 days

0.5

0.0

Cohesiveness

0 days

Crumbliness Juiciness

-0.5 7 days

-1.0 -1.0

(b)

Tenderness

Fibrous Mastication remnant Chewiness Hardness

-0.5

0.0

0.5

1.0

Principal Componenet 1 (Y-explained variance 51%)

Fig. 8. Texture variation in (a) left and (b) right loins as per ageing. ANOVA partial least squares regression (APLSR) correlation loading plot of the first two principal components (PCs). Design variables (ageing, left/right) in the X-matrix and sensory terms in the Y-matrix. Ellipses represent r2 ¼ 50 and 100%.

7 days. Crumbliness was the aspect of tenderness that differed between 4 and 7 days of ageing. Each ageing period was found to be significant in the model based on the full design. Of major interest in this analysis was that the left and right side muscles were found to vary considerably in that right side muscles were significantly (P < 0:05) more hard and cohesive than left side muscles (Fig. 9). However, this figure did not show how left and right loins differed. Qualitatively left and right were very

similar for all variations. However, the intensity of the descriptors was scored higher and was more prominent in the right side relative to the left. Thus, a quantitative significant difference (P < 0:05) was found in this overall analysis, between the left and right side muscles (Fig. 9). 3.4. Instrumental analysis as predicted by the design Initially, descriptive statistics displaying means and ranges of instrumental hardness were derived for left

Principal Componenet 2 (Y-explained variance 1%)

S. Hansen et al. / Meat Science 68 (2004) 611–629

623

1.0

Right Cranial

0.5

4 days Tenderness

0.0

7 days Crumbliness

C8 C3 C1 a C7 C9 C11 C4 C2 Juiciness C16 C18 Cohesiveness C6 c Mastication remnant C5 C12 Fibrous Chewiness 0 days b C10 Hardness C15

C13 C14

-0.5

C17

Caudal Left

-1.0 -1.0

-0.5

0.0

0.5

1.0

Principal Componenet 1 (Y-explained variance 11%) Fig. 9. Overall textural variation. ANOVA partial least squares regression (APLSR) correlation loading plot of the first two principal components (PCs). Total design (transverse positions and various chops, cranial/caudal, ageing, left/right) in the X-matrix and sensory terms in the Y-matrix. Ellipses represent r2 ¼ 50 and 100%.

Table 3 Descriptive statistics of Warner Bratzler instrumental hardness data Animal side

Ageinga

Positionb

Mean

SDc

Mind

Maxe

Left

0 0 0

a b c

94.94 92.40 96.39

21.18 14.83 17.24

44.71 59.44 62.64

151.79 137.11 145.17

4 4 4

a b c

68.80 68.97 71.07

13.39 13.57 16.16

33.22 41.75 39.36

93.85 109.25 105.66

7 7 7

a b c

65.38 66.01 69.22

14.21 15.88 16.01

34.02 39.76 33.42

101.90 116.24 122.58

0 0 0

a b c

93.78 94.07 93.43

27.48 27.14 25.89

49.54 47.81 48.27

155.10 160.18 178.26

4 4 4

a b c

77.36 80.40 82.02

18.91 24.49 23.85

38.45 46.61 36.91

162.78 147.23 156.02

7 7 7

a b c

88.64 91.06 97.54

23.41 24.92 31.87

51.82 54.35 55.82

169.82 186.08 177.39

Right

a

0, 4 and 7 days. Chop transverse positions: a, dorsal; b, medial; c, lateral. c Standard deviation. d Minimum. e Maximum. b

and right muscles in order to display numerically the trends in the transverse variation and the influence of ageing (Table 3). The data from left and right loins were

averaged over chops within each ageing period, respectively (four left and five right loins per ageing period). It was apparent from the table that left and right side

624

S. Hansen et al. / Meat Science 68 (2004) 611–629

muscles differed, the left side muscles being the least hard both on average and when looking at the maximum force used for compression. Furthermore, APLSR analysis was carried out using the total design variation (transverse sample positions, longitudinal chop variation, cranial and caudal end, left and right LD, and ageing) in order to differentiate the instrumental responses (maximum force, distance at the maximum force, and the force used for 80% compression) with respect to the design variation. Again the variables cranial and caudal were pacified, whereas the rest of the variables were standardised. The model explained a total of 9% validated variation in PC1 and PC2 (Fig. 10). Jack-knife uncertainty testing revealed that the design factors transverse position a (dorsal) and c (lateral), and a number of individual chops (1, 2, 9, 10, and 16) differed significantly (P < 0:05). Additionally the design variables left/right and ageing also showed significant (P < 0:05) differentiation. The longitudinal variation (cranial/caudal and most of the individual chops) were found to be non-significant (P > 0:05). Overall, in Fig. 10 the instrumental variables spanned the first PC whereas PC2 spanned the transverse and longitudinal sample variation (individual chops and cranial/caudal ends). A third PC spanned ageing and yet again a forth PC spanned left and right muscles (results not shown). Cross-sectional sample variation differed significantly from dorsal to lateral positions, dorsal being the least hard (less force required for compression). The medial position was found to be non-significantly different (P > 0:05), but a tendency for this sample po-

sition to be harder than the dorsal position and less hard than the lateral position were seen. Overall, this was also apparent in the descriptive statistics presented in Table 3. Moreover, it appeared from Table 3 that the standard deviations (SD) were lower for left side muscles than for right side muscles indicating less variation in the left side muscles compared to the right side muscles. Instrumental analysis showed a trend (non-significant) in decreasing hardness when approaching the caudal end of the muscle. From the APLSR correlation loadings plot (Fig. 10) it clearly appeared that the longitudinal variation could not be described by the instrumental measurement, even though a slight tendency (non-significant) for the cranial end to be harder than the caudal end was seen. Left and right muscles differed significantly (Fig. 10). Right side muscles were found to be harder than left side muscles when measured instrumentally (see Fig. 10; Table 3). Overall, the instrumental analysis showed the same variation as was found in the sensory results (Fig. 9), the only exception being that the longitudinal variation could not be described by the instrumental measurement. 3.5. Sensory and instrumental predictive and causal analysis APLSR analysis was utilised to investigate the relationship between the total design variation (transverse sample positions, longitudinal chop variation, cranial and caudal end, left and right LD, and ageing) and the

Principal Component 2 (Y-explained variance 1%)

1.0

7 days

0.5 b Right C17

C10

Max distance C13 C1 Caudal C9 C8 C12 C11 C14 C5 C6 C2 C18 4 days C4 Cranial C3 C16 C7 c a C15

0.0

80% compression Max force

Left

-0.5 0 days

-1.0 -1.0

-0.5

0.0

0.5

1.0

Principal Component 1 (Y-explained variance 10%) Fig. 10. Instrumental analysis. ANOVA partial least squares regression (APLSR) correlation loading plot of the first two principal components (PCs). Total design (transverse positions and various chops, cranial/caudal, ageing, left/right) in the X-matrix and instrumental terms in the Y-matrix. Ellipses represent r2 ¼ 50 and 100%.

S. Hansen et al. / Meat Science 68 (2004) 611–629

625

Principal Component 2 (Y-explained variance 1%)

1.0

Cranial

Right

0.5 C8 C9 C4 C5 C7 Max force C6 c C3 80% compression C1 Distance max Juiciness C10 C2 C11 Cohesiveness b Crumbliness a C15 Mastication remnant Chewiness Fibrous Hardness C12 C16 C17 0 days C18 C13 C14

7d ays Tenderness

0.0

4 days

-0.5 Left

Caudal

-1.0 -1.0

-0.5

0.0

0.5

1.0

Principal Component 1 (Y-explained variance 10%) Fig. 11. Instrumental and sensory predictive and causal analysis. ANOVA partial least squares regression (APLSR) correlation loading plot of the first two principal components (PCs). Total design (transverse positions and various chops, cranial/caudal, ageing, left/right) in the X-matrix and sensory and instrumental terms in the Y-matrix. Ellipses represent r2 ¼ 50 and 100%.

sensory and instrumental responses. The variables cranial and caudal were pacified, whereas the rest of the variables were standardised. The model explained a total of 11% variation in PC1 and PC2. The sensory properties hardness, fibrous, mastication remnant and chewiness were positively correlated with the maximum force and the force used at 80% compression. Sensory and instrumental measurements therefore were found to be predictive of each other in that they spanned the same dimensions in the correlation loadings plot shown in Fig. 11. However these methods did not measure the same qualitative properties of texture. Deformation in terms of force was measured instrumentally, whereas human perception of hardness was evaluated using sensory analysis.

4. Discussion The main aim of the training was to determine the reliability and validity of the terms present. The terms should be relevant for the product, discriminate between samples, and have cognitive clarity. These three criteria for descriptor reliability were not met in training session 1. However, training session 3 fulfilled all three. However, it should be emphasised that discrimination and cognitive clarity could not be ascribed for the term juiciness from either APLSR or GPA. According to Szczesniak and Ilker (1988) perception of juiciness is influenced by the force with which the juice is extracted from the meat during mastication, the amount of juice released on the

first bite and subsequent bites, and the contrast between the liquid and the solid phase. Perception is furthermore influenced by the hydrating effect of saliva production during mastication. The reason for using this term despite the lack of discriminative effect and cognitive clarity was based on the fact that juiciness is conventionally considered to be a very desirable attribute in meat and therefore a valuable term when considering acceptance of the product. In the present study, this did not appear to be the case and it may be wiser to concentrate on tenderness in this respect as it is clearly understood by panellists. Overall the training was successful in that panellists attained the ability to discriminate the various textural properties between samples. Three days of training was sufficient for the panel to be able to discriminate all sources of variation, except cross-sectional position, which was later found to be non-significant in profiling. Training session 3 resembled the subsequent profile, in its separation of variation, which was the ultimate validation of the training. In the sensory determined transverse muscle variation, a trend (non-significant) in quantitatively greater juiciness and tenderness ratings from the dorsal position (nearest the spinal column) to the medial position, was found. This was similarly found, and to be significant, in a study conducted by Rust et al. (1972). Onate and Carlin (1963) also found that LD was significantly more tender near the spinal column than at the outer edge when measured instrumentally. Instrumental analysis showed that hardness increased from the dorsal to the lateral position, both being

626

S. Hansen et al. / Meat Science 68 (2004) 611–629

significantly different. It should however be mentioned that the medial position could not be ascribed any significance level. The sensory and instrumental analysis agreed that the dorsal position was the most tender, but neither analysis was able to determine if the medial or the lateral position was the least tender of the three positions investigated. This difference between locations may be caused by morphological or metabolic differences within the muscle. The metabolic differences may be explained by different functionality of the different sections of the muscle, one being more supportive (for the spine) and another being more active in physical movements resulting in different patterns of postmortem energy metabolism. Regarding the subtle differences found for the medial and lateral positions, with respect to the sensory and instrumental methods used, it may be that the differences in texture are less pronounced and therefore the actual differences may be overshadowed by the differences in muscle fibre orientation in the samples used in these analytical methods. According to Otremba et al. (1999) the highest correlations between sensory and instrumental measurements would be obtained if samples were assessed the same way in sensory and instrumental analysis, meaning that muscle fibre orientation in presentation for measurement should be highly controlled. However, this would not reflect how a human would assess the tenderness of the sample normally, it would be more reflected in the sensory assessments of the present study where the orientation was more randomised. Transverse variation was better described instrumentally than from a sensory perspective, in that significance levels could be ascribed. This was most likely because fibre orientation when assessed in sensory analysis was more inconsistent in its orientation when placed in the mouth by panellists. The trend found in the present study disagrees with the findings of Alsmeyer et al. (1965a, 1965b), who found that the lateral position was significantly more tender that the dorsal and medial positions. Due to the inconclusive nature of existing research in relation to transverse muscle texture variation further investigations are required to determine if the decreasing trend in tenderness from dorsal to lateral position can be clarified and the significance determined. This may be achieved via a design fully orientated towards the textural differences between transverse positions of m. longissimus dorsi pork chops. In such an experiment factors that may be important to this variation, e.g., degree of cooking, panel training, experience and number of panellists used, should be given due consideration. However, as it stands the lack of cross-sectional significant difference in variation is notable for descriptive texture sensory analysis. The result implies that a sample can be taken at any point across a chop, (a, b, or c) and

considered representative of the transverse sensory variation in the chop as a whole. Furthermore, industrial on-line measurements need not be orientated to a specific cross-sectional position in the muscle in relation to tenderness prediction as all appear equally representative of the sensory variation. In the sensory determined longitudinal muscle variation a significant decrease in tenderness was found from the anterior to the posterior part of the muscle. This was also shown by Weir (1953). Specifically Weir (1953) found that position 1, 2, and 8 (2nd through 8th rib and sacral region) were the most tender, whereas position 4 and 5 (last of the rib region and anterior of the lumbar vertebrae region) were considerably less tender. A decrease in tenderness was observed from position 2 to 4. The latter correlates well with the findings of the present study in that position 2–4 in Weirs study corresponds to the cranial region including chop 10 in the present study. Møller and Vestergaard (1986) found that location 1 (1st to 4th lumbar vertebrae) was least tender and location 3 (9th to 11th rib) was most tender. This correlates well with the findings of this study in that location 3 and 1 roughly correspond to chop 4–7 and 12–15, respectively. However, it should be noted that Rust et al. (1972) found no significant variation lengthwise when investigating 4 chops, 2.5 cm thick, sliced anterior and posterior to the last rib. It may be that other factors (i.e., gender, slaughter weight, heat treatment, panel training, experience and number of panellists used, etc.) are important to the variation in the conflicting studies. It may also potentially be ascribed to changes in muscle fibre orientation throughout the muscle. This would not be easily recognised in instrumental measurements as they are conducted perpendicular to fibre orientation. However, the tenderness, hardness etc. in sensory analysis would be affected by changes in fibre orientation. As to the differences between cranial and caudal regions a study conducted by Rees, Trout, and Warner (2002) showed that the pH decline rate was fastest in the cranial region (measured at the 5th thoracic vertebra) compared to the medial (12th thoracic vertebra) and caudal (5th lumbar vertebra) regions. Furthermore, they showed that the temperature decline was slowest in the cranial region, suggested to be due to the greater amount of subcutaneous fat in this region which acted as an insulater of the muscle. These results are indicative of differences in energy metabolism in the different regions, which may contribute to the variations in meat tenderness as it may influence the amount of bound water which influences the perception of hardness (e.g., Henckel, Karlsson, Oksbjerg, & Petersen, 2000; Briskey, 1964; Bendall & Swatland, 1988). However, the reason for the cranial end to be the most tender may also be related to the anatomical location of the muscles in that different amounts of work may lead to differences in

S. Hansen et al. / Meat Science 68 (2004) 611–629

energy metabolism along the muscle. The cranial end is strongly supported by the ribcage and thus less work is performed in this muscle part. When performing more work muscle glycogen stores are lowered and thus the pH decline postmortem will occur to a lower degree. This results in more bound water and subsequently meat texture is perceived as increasing in hardness (e.g., Henckel et al., 2000; Briskey, 1964; Bendall & Swatland, 1988). Variation between individual chops was found when assessed by the panel but not when measured instrumentally. Instrumental analysis showed a trend (nonsignificant) in decreasing hardness when approaching the caudal end of the muscle. From the correlation loadings plot (Fig. 10) it appeared that the longitudinal variation could not be described by instrumental measurement, even though a slight tendency (non-significant) for the cranial end to be harder than the caudal end was seen. This is opposite to what was found and was significant and clearly interpretable in the sensory evaluation. This may be due to the heterogeneity of the meat. Texture testing instruments are calibrated to respond linearly to the intensity of the tested mechanical property (Szczesniak, 1987), but linearity will only occur if the biological material is homogeneous (Spadaro et al., 2002). When using sensory profiling no linearity is provided which gives better predictions of the actual texture variation perceived. Because the instrumental measurements lacked information about the longitudinal texture variation, specific examination of the texture was only carried out using data from the sensory analysis. When tenderness ratings were determined as a function of individual chops for each of the three ageing periods it appeared that the first 7 chops showed similarity in tenderness level for 0, 4, and 7 days respectively. A marked decrease in tenderness took place at the region around the end of the ribcage. Further on the tenderness was rated as being at a constant or slightly increasing level. As tenderness decreased hardness was found to concurrently increase. The turn-over point in tenderness could be applied to a general textural turnover point in that hardness and juiciness also showed marked changes in this specific region of the muscle. A markedly higher difference between individual chops was seen in the cranial than in the caudal end at 0 days of ageing. When the meat was aged bigger individual chop variation occurred in the caudal end. It seemed that the individual chop variation in the cranial end was less pronounced, whereas the variation in the caudal end was altered when the meat was aged. Sensory and instrumental data revealed significance differences between left and right muscles. The left muscles were found to be significantly more tender than the right muscles in both the sensory and instrumental analysis. The textural variation lengthwise was better

627

described when assessed by the panel compared to when measured instrumentally. Of considerably importance regarding this discrepancy, the shear force apparatus only measured deformation in terms of the force (N) used for compression of the sample, whereas the sensory profiling described what caused the differences in the longitudinal variation. Right muscles showed bigger internal variation throughout the loin, than did left muscles. Moreover, it appeared that the longitudinal variation was more distinct and interacting in the right loins compared to the left loins. Both left and right loins were significantly described by sensory descriptors when jack-knife testing was performed. Qualitatively left and right loins displayed the same trends but differed quantitatively in the levels of variation. Weir (1953) also found that tenderness varied significantly between left and right sides of the same animal. It was not outlined what the differences were nor was any attempt to explain the differences between left and right side made. Plausible reasons for the differences between left and right side muscles is either differences in physiological prerequisites or differences in response to the physical or mental stress the animal was exposed to in connection with the slaughtering procedure. The cranial ends showed similarity in tenderness ratings in both sides of the animals, which may be due to the fact that this region is held in place by the ribcage and therefore less affected by physical stress compared to the relatively less supported caudal end. Differences in the energy metabolism, resulting in variation in texture, drip loss, and colour in the different parts of LD are known to occur (Bendall & Swatland, 1988). The higher energy metabolism in the right LD may result from a greater level of work this muscle performs compared to the left LD. The characteristics of postmortem pH decline are determined by the physiological conditions that reside in the muscle at the time of slaughter and can be related to lactate production, or more specifically, to the capacity of the muscle to produce adenosine triphosphate (ATP) (Bendall, 1973). It is obvious that different parts of the muscle do different types of work, which may be the explanation for the differences in the basic metabolic processes in the muscle parts and thus the resultant pH related textural variation. The caudal end is significantly (P < 0:05) less tender in the right side muscles, indicating that these have performed more work than the left side muscles. In relation to this right versus left difference it may be that the majority of the pigs in the present study were ‘right-footed’. If this was the case it would explain why the right side muscles varies considerably more internally than the left side muscles. This differentiation between left and right muscles from the same animal may also in part have been caused by the hanging of the carcass. However, experiments with hanging have shown that there still is a

628

S. Hansen et al. / Meat Science 68 (2004) 611–629

difference between left and right side muscles regardless of hanging (Henckel, 2003: unpublished data). Thus, hanging cannot be assigned as the main contributing factor to the left and right side texture differences. Cranial and caudal muscle ends differ significantly (P < 0:05) in texture thus, the overall tenderness cannot be predicted from a single sample position in the left and/or right side. This study suggests that as the right loins are more defined in their longitudinal variation they could be more accurately utilised for lengthwise variation for sensory predictability purposes. For right side sensory predictability purposes three sample positions may be necessary. The present study indicates that reliable and representative predictability should arise if sampling at the last rib (chop 10), 12 cm anterior, and 12 cm posterior to the last rib (chop 4 and 16, respectively) are chosen. A marked decrease was observed at the end of the rib cage, whereas chop 4 and 16 were a more or less constant representative of the cranial and caudal regions, respectively. Since variability was more clearly defined and described in right loins, these should be used for predictability purposes in this muscle. In the left side two positions, one at the cranial and one at the caudal end would be sufficient. The right muscle may be considered representative of the sensory variation in the left muscle, whereas the left muscle lacks certain information to predict the variation in the right muscle. It may be deemed reasonable to determine the cause of difference between the left and right side muscles via a design fully orientated towards this. Following from this the meat industry should consider the fact that differences between left and right side muscles are present when performing on-line measurements to predict the overall texture of meat. From an industrial perspective further research should be carried out to determine if these differences are pronounced early postmortem and can be utilised in more representative texture prediction on-line. Both sensory and instrumental analyses were able to differentiate significantly between 0, 4, and 7 days of ageing. As expected 0 days of ageing was positively correlated to hardness. Ageing for 4 and 7 days, respectively, resulted in equally tender meat. The sensory analysis revealed that 4 and 7 days of ageing were differentiated by the levels of crumbliness and cohesiveness, 7 days of ageing leading to more crumbly and less cohesive meat.

Acknowledgements The authors thank Poul Henckel, Danish Institute of Agricultural Sciences, Research Centre Foulum, for shareing his knowledge about energy metabolism within muscles. The authors also acknowledge the ‘ConSense’

and ‘Sense-Index’ projects and the financial support of the The Directorate for Food, Fisheries and Agri Business (DFFE), under the Danish Ministry of Food, Agriculture and Fisheries.

References Alsmeyer, R. H., Thornton, J. W., & Hiner, R. L. (1965a). Some dorsal–lateral location tenderness differences in the longissimus dorsi muscle of beef and pork. Journal of Animal Science, 24, 526– 530. Alsmeyer, R. H., Thornton, J. W., & Hiner, R. L. (1965b). Crosssectional tenderness variations among six locations of pork longissimus dorsi. Journal of Food Science, 30, 181–183. AMSA (1995). Research guidelines: Cookery, sensory evaluation, and instrumental tenderness measurement of fresh meat. American Meat Science Association. ASTM (1986). Physical requirements. Guidelines for sensory evaluation laboratories, STP 913. Pennsylvania: American Society for Testing and Materials. Bendall, J. R. (1973). Cold-contracture and ATP turnover in the red and white musculature of the pig post mortem. Journal of Science of Food and Agriculture, 23, 55–71. Bendall, J. R., & Swatland, H. J. (1988). A review of the relationship of pH with physiological aspects of pork quality. Meat Science, 24, 85–126. Briskey, E. J. (1964). Etiological status and associated studies of pale, soft, exudative porcine musculature. Advances in Food Research, 13, 89–178. Byrne, D. V., Bak, L. S., Bredie, W. L. P., Bertelsen, G., & Martens, M. (1999a). Development of a sensory vocabulary for warmed-over flavor: Part I. In porcine meat. Journal of Sensory Studies, 14, 47– 65. Byrne, D. V., Bredie, W. L. P., & Martens, M. (1999b). Development of a sensory vocabulary for warmed-over flavor: Part II. In chicken meat. Journal of Sensory Studies, 14, 67–78. Byrne, D. V., O’Sullivan, M. G., Dijksterhuis, G. B., Bredie, W. L. P., & Martens, M. (2001). Sensory panel consistency during development of a vocabulary for warmed over flavour. Food Quality and Preferences, 12, 171–187. Commission Regulation (EC) No. 3127/94 of 20 December 1994 amending Regulation (EC) No. 2967/85 laying down detailed rules for the application of the community scale for grading pig carcasses (pp. 0043–0044). Official Journal L 330, 21/12/1994. Dijksterhuis, G. B., & Gower, J. C. (1991). The interpretation of generalised procrustes analysis and allied methods. Food Quality and Preference, 3, 67–87. Esbensen, K. H. (2001). Multivariate data analysis – in practice. An introduction to multivariate data analysis and experimental design (5th ed.). Norway: CAMO ASA. Gower, J. (1975). Generalized procrustes analysis. Psychometrica, 40, 31–55. Henckel, P., Karlsson, A., Oksbjerg, N., & Petersen, J. S. (2000). Control of post mortem pH-decrease in pig muscles: Experimental design and testing of animal models. Meat Science, 55, 131–138. ISO (1988). Sensory analysis: General guidance for the design of test rooms. International Organization of Standardization, ISO 8589. Geneva, Switzerland. ISO (1994). Sensory analysis – methodology – texture profile. International Organization of Standardization, ISO 11036. Geneva, Switzerland. Martens, H., & Martens, M. (2001). Interpretation of many types of data X () Y: Exploring relationships in interdisciplinary data

S. Hansen et al. / Meat Science 68 (2004) 611–629 sets. In Multivariate analysis of quality. An introduction (pp. 139– 145). London, England: Wiley. Meilgaard, M., Civille, G. V., & Carr, B. T. (1999). Descriptive analysis techniques. In Sensory evaluation techniques (3rd ed., pp. 187–200). Boca Raton, Florida: CRC Press. Møller, A. J., & Vestergaard, T. (1986). Effects of altered carcass suspension during rigor mortis on tenderness of pork loin. Meat Science, 18, 77–87. Onate, L. U., & Carlin, A. F. (1963). Relation of physical and sensory evaluations of pork loin quality to backfat thickness. Food Technology, 17, 123–126. Otremba, M. M., Dikeman, M. E., Milliken, G. A., Stroda, S. L., Unruh, J. A., & Chambers IV, E. (1999). Interrelationships among evaluations of beef longissimus dorsi and semitendinosus muscle tenderness by Warner–Bratzler shear force, a descriptive-texture profile sensory panel, and a descriptive attribute sensory panel. Journal of Animal Science, 77, 865–873. Qvist, K. B. (1998). Uniaksial kompression, TPA, tension, penetration og punktur. In Reologi -med henblik p
a mejeriprodukter. Denmark: DSR Forlag. Rees, M. P., Trout, G. R., & Warner, R. D. (2002). Tenderness, ageing rate and meat quality of pork M. longissimus thoracis et lumborum after accelerated boning. Meat Science, 60, 113–124. Rosenvold, K., van den Berg, F., Andersen, H. J., Johansson, L., & Lundstr€ om, K. (2002). Beef Warner–Bratzler shear force measure-

629

ments in relation to sensory-determined tenderness. Does measurement temperature influence the interpretation? In Proceedings of the 48th international congress of meat science and technology (48th ICoMST) (pp. 234–235, Vol. 1). Rome, Italy. Rust, R. E., Olson, D. G., Shuler, R. O., & Thomson, G. M. (1972). Pork loins: Effect of location of chops in the medial loin region and position of cores in cross section of the longissimus muscle on tenderness and juiciness. Journal of Animal Science, 35, 548–550. Spadaro, V., Allen, D. H., Keeton, J. T., Moreira, R., & Boleman, R. M. (2002). Biomechanical properties of meat and their correlation to tenderness. Journal of Texture Studies, 33, 59–87. Szczeniak, A. S. (1987). Correlating sensory with instrumental texture measurements – an overview of recent developments. Journal of Texture Studies, 18, 1–15. Szczeniak, A. S., & Ilker, R. (1988). The meaning of textural characteristics – juiciness in plant food stuff. Journal of Texture Studies, 19, 61–78. Tornberg, E. (1996). Biophysical aspects of meat tenderness. Meat Science, 43, 175–191. Weir, C. E. (1953). The variation in tenderness in the longissimus dorsi muscle of pork. Food Technology, 7, 500–501. Wheeler, T. L., Shackelford, S. D., & Koohmaraie, M. (1997). Standardizing collection and interpretation of Warner–Bratzler shear force and sensory tenderness data. Proceedings of Reciprocal Meat Conference, 50, 68–77.