Relationship between instrumental texture analysis and objective sensory evaluation
In this study, sensory properties of formulated sausages were evaluated in three ways (a) by instrumental means (absolute values), (b) a trained sensory panel (objective values), and (c) by an untrained sensory panel (hedonic values) to imitate consumer's perception. However, one must be careful when using instrumental data to imitate oral processes (Szczesniak and Hall 1975; Rosenthal 1999). Oral processing is a complicated time-dependent process (Hitchings and Lillford 1988), and therefore, sensory testing has been performed in all formulations. Both objective (i.e., absolute hardness) and hedonic tests (i.e., acceptability of hardness) were performed for all mechanical attributes, in order to obtain data on the objective magnitude of each parameter in relation to its consumer acceptance. Therefore, it was necessary first to establish the relationship between the absolute values obtained by instrumental analysis and the objective values obtained by the trained panelists. This was done by performing a redundancy analysis of variables which has been successfully used in food products by our team in an attempt to quantify relationships between the sensory and mechanical profiles (Raphaelides et al. 1995, 1998). By changing the proportions of some basic ingredients of the product, the texture profile is expected to alter, ultimately affecting the sensory attributes of food. In that sense, the redundancy analysis aims to locate and describe the predominant mechanical variables which control the response of the sensory attributes.
Overall dependent sensory and instrumental variables explained 89.7% of the total variation, regarding the first two major axes. Three criteria of variable importance for the redundancy analysis were chosen: forward selection of mechanical variables, t-values of the regression coefficients, and intersect correlation coefficients of mechanical variables with axes 1 and 2 (Ter Braak and Smilauer 2002). Hardness was assessed as the most important mechanical variable (R2 = 0.48, t = 3.19, r = 0.544 with axis 1), followed by chewiness (R2 = 0.28, t = 1.86, r = 0.237 with axis 2).
Potential relationships between the two sets of variables are shown in Figure 3: Mechanical hardness correlates strongly in a positive manner with objective sensory hardness and consistency and negatively with objective chewiness and fattiness. Mechanical chewiness correlates fairly well in a positive direction with objective sensory chewiness, while both color measurement methods (sensory and colorimetric) correlate very strongly and positively to each other.
Figure 3. Biplot based on redundancy analysis of sensory profile (solid lines) with respect to instrumental variables (dashed lines). The lines display the approximate correlation coefficients between the two sets of variables. Longer arrows are more important in producing effects.
Download figure to PowerPoint
Principal component analysis
PCA was performed on objective and hedonic data in order to distinguish the most important variables (Sharma 1996). Results (Figs. 4, 5) indicate that the first two principal component axes extracted 78.4% and 61.8% of the total variation for the objective and hedonic variables, respectively. The right half of the horizontal axis (axis 1) is explicitly described by a first group of variables, namely objective fattiness, elasticity, and chewiness, whereas a second group, objective consistency and hardness, describe the left half. In Figure 4, arrows forming small oblique angles indicate highly positive correlation coefficients, while obtuse angles indicate highly negative relationships. Loading values (Table 2) suggest that both groups of variables correlate very strongly between variables and with axis 1.
Table 2. Loading factors (correlations) between sensory variables and principal component axes and correspondent values between principal component analysis (PCA) scores and unstructured sensory scale
|AXIS 1||AXIS 2||Low (3–6 cm)||Moderate (6–9 cm)||Adequate (9–12 cm)||Fair (12–15 cm)|
|Chewiness||0.85|| ||−2.2||−0.5||−0.5||0.8||0.8||2.2|| || |
|Elasticity||0.67|| ||−2.2||−0.9||−0.9||0.9||0.9||2.2|| || |
|Color|| ||0.92||−2.2||−1.1||−1.1||1.0||1.0||1.8|| || |
|Consistency||0.70|| ||−2.4||−1.4||−1.4||0.9||0.9||2.2|| || |
Figure 4. Loading plot of objective sensory variables by the principal component analysis. Arrows indicate the strength of each variable importance and the number in brackets the percentage contribution of each axis to the total variation.
Download figure to PowerPoint
Figure 5. Loading plot of hedonic sensory variables by the principal component analysis. Arrows indicate the strength of each variable importance and the number in brackets the percentage contribution of each axis to the total variation.
Download figure to PowerPoint
Taking into account the results from PCA, it is possible to treat all five variables of axis 1, as one. The new variable is now axis 1, which reflects responses of elasticity, chewiness, fattiness, consistency, and hardness. High values of axis 1 correspond to high values of fattiness, elasticity, and chewiness, and low values of consistency and hardness and vice versa. The above are in very good agreement with previous findings where increases in biting force and hardness were associated with decreases in elasticity and chewiness (Petridis et al. 2010). This was attributed to the disruption of the protein gel due to local phase separations between starch and caseinate and/or meat protein. This is further reinforced by observations in caseinate–myofibrillar protein in meat products (Su et al. 2000; Barbut 2007), which suggest local phase separations between the two components. Objective red color appears to be the unique and strong positive contributor for the formation of axis 2 (r = 0.92).
PCA on hedonic data showed that consistency, elasticity, and chewiness correlate strongly and positively with each other (Fig. 5) and also with axis 1 (Table 2). High scores of axis 1 correspond to high values of those variables. The above correlation is readily explicable, as elasticity and chewiness are expected to be closely correlated by definition (Szczesniak and Hall 1975; Bourne 1978). The second axis is formed by the hedonic variables red color and fattiness rather loosely. It should be pointed out that, according to the above, objective consistency correlates negatively to elasticity and chewiness, whereas hedonic consistency correlates positively to the same attributes. This is a valuable indication of the necessity for the differentiation between the objective magnitude of an attribute (i.e., directly comparable to instrumental texture analysis) and its hedonic counterpart.
Analysis of the effects of mixture components on objective and hedonic sensory properties
Mixture experiments are performed in many product-development designs (Piepel and Cornell 1994). Two or more ingredients (components) are mixed in various proportions, and many attributes, sensory and/or mechanical, of the resulting products are recorded. The measured attributes (responses) can depend either on the proportion of components present in the mixture or on the total amount of the mixture.
The effect of the independent variables, caseinate (X1), starch (X2), and lard content (X3) was analyzed by regression using equation (1). Instead of regressing the independent variables for each separate dependent variable (objective and/or hedonic red color, hardness, consistency, fattiness, chewiness, and elasticity), regression was performed for the four axes (AXIS1_O, AXIS1_H, AXIS2_O, and AXIS2_H). According to PCA results, Axis1_O includes the objective attributes hardness, consistency, fattiness, chewiness, and elasticity, and Axis2_O the objective red color attribute, while Axis1_H and Axis2_H the corresponding hedonic variables.
The response of objective variables (AXIS1_O) is best described by the three components effects and the interaction term starch × lard:
The different mixture amounts did not affect the panelists' sensory stimuli, but starch and lard show negative synergy (negative sign in the term).
The mixture amounts are very effective in describing the effect of the response objective “red color intensity” (AXIS2_O), alone or combined with interaction terms:
In higher lard proportions (amount 27%), sodium caseinate and starch decrease the red color intensity (negative signs in the terms), but sodium caseinate, when combined with starch and lard in higher lard amounts, increase the red color intensity. These results should be interpreted with caution, however, as the gap between predicted and determined R2 values is large (29.9% and 57.6%, respectively).
Mixture amounts are also important for the hedonic variables (AXIS1_H):
Starch or sodium caseinate reduces the acceptability of the hedonic variables. Sodium caseinate, when combined with the other components in higher lard proportions, increases the acceptability of the variables under study.
Acceptability toward red color (AXIS2_H) was rejected, despite providing good fit with the mixture components, due to the discrepancy between the predicted and determined R2 values (10.0% and 43.1%, respectively, results are not shown).
In order to optimize the caseinate, starch, and lard content for consumer acceptance, it was necessary to arrange the PCA scores of the objective and hedonic sensory properties to the unstructured scale (0–15 cm) used for the respective sensory properties (Table 2). Results can be better visualized with the contour plots in Figures 6-8. These plots show how a response variable relates to the three ingredients, based on a model equation. Lard is very important for increasing the sensory intensity of fattiness, chewiness, and elasticity, and for decreasing the intensity of hardness and consistency (AXIS1_O), the latter two reaching maximum intensities at very high caseinate and starch proportions (5%) (Fig. 6 and Table 2). Increasing caseinate and starch content was expected to increase consistency. According to Su et al. (2000), the fat globules of a frankfurter batter are confined locally within the denser nonmeat (caseinate) protein matrix. This means that the chances for fat coalescence during cooking may be reduced so that emulsions with high fat and water-binding properties are formed. Thus, the products with firmer texture are expected. As the term “amount” was not found statistically significant for AXIS1_O (eq. 2), the two mixture designs appear identical in Figure 6 after eliminating the mixture-amount effect.
It should be noted that the contours are fairly symmetrical perpendicular to the lard vertex, indicating a fairly linear effect of the lard component. Deviations from linearity are due to the incorporation of the second-order term starch × lard in the equation (2). This suggests that, practically for all fat contents, the two macromolecular groups act in the same way toward fattiness, chewiness, elasticity, hardness, and consistency. This is in complete agreement with our findings derived from a different experimental (semiqualitative) design (Petridis et al. 2010), implying that both starch and caseinate interact similarly with the myofibrillar matrix of the sausage. Reduction in hardness should be related to phase separation between starch/caseinate and the myofibrillar protein.
Starch and pork meat protein appear not to mutually interact with each other upon heating in temperatures similar to the ones applied in the present experiments (Li and Yeh 2002). It is also reported that increased starch content reduces the elastic modulus of mixed starch–whey protein isolate gels due to a weak starch matrix formation between the two components (Aguilera and Rojas 1996). Such phase separations are normally concentration dependent. Modified starch enhances water binding (Ruusunen et al. 2003), thus increasing the polymers effective concentration. The rheology of whey protein–starch systems shows concentration-dependent transitions from solid like to liquid like (Vu Dang et al. 2009). In these cases, protein and starch appear to phase separate, with one phase dispersing into the other, reducing its elastic modulus. One can argue that starch phase separates and interferes with the continuous meat protein gel, reducing elasticity and its related parameters such as chewiness.
According to the rescaled values in Table 2, fattiness could be characterized in the mixture samples as “moderately intense,” chewiness and elasticity as “moderately to fairly intense,” and finally consistency and hardness as “moderately to fairly intense,” in opposite direction of the former variables (due to the negative signs of the scores).
Red color is important only in the higher proportion amounts (Fig. 7), showing an adequate intensity level at a composition of 2.5% caseinate, 0% starch, and 24.5% lard.
Adequate acceptability toward chewiness, consistency, and elasticity rises, as it reaches high lard proportions in the higher mixture-amounts design (Fig. 8). Moderate levels of acceptability are encountered mostly in the lower mixture-amounts design.
The ensuing step in our analysis has been the construction of response trace plots as to allow conditions for optimization. Trace or component effects plots show how each ingredient affects the response relative to a reference blend. The center point has been selected as the reference blend (Table 1; Figs. 1 and 9-12). At this point, for both mixture amounts (17% and 27%), the objective variables fattiness, chewiness, and elasticity have reached a moderate range of intensity (−0.20; Table 2 and Fig. 9), whereas the consistency and hardness have already switched to “adequate.” Lard is the most important component, that is, due to its large blend proportion range and distance of response change (extending its influence along the whole scale of both axes in the graph). Thus, increases in the lard proportion lead to an increase in the sensory intensity of score values for fattiness, chewiness, and elasticity, whereas a lard proportion of less than 22.0% or 13.89% (for each of the two mixture amounts) increases the hardness and consistency. This is in good agreement with previous data which suggest that objective sensory attributes such as elasticity and cohesiveness of frankfurters rises monotonically with fat content (Ritzoulis et al. 2010). The importance of the other two components is minor, because the range of reference blend proportion and Y-axis distance changes is fairly small.
Figure 9. Response trace plots for objective variables including both the mixture-amount designs. As the proportion of lard in the mixture increases (and the other mixture component decrease), the intensity rating of AXIS1_O increases.
Download figure to PowerPoint
Figure 10. Response trace plots for the red color intensity at the higher mixture-amount designs. As the proportion of caseinate in the mixture increases (and the other mixture component decrease), the rating of red color intensity decreases.
Download figure to PowerPoint
Figure 11. Response trace plots for hedonic variables at the lower mixture-amount design. As the proportion of lard in the mixture increases (and the other mixture component decrease), the acceptance rating of AXIS1_H increases.
Download figure to PowerPoint
Figure 12. Response trace plots for hedonic variables at the higher mixture-amount design. As the proportion of caseinate or starch in the mixture increases (and the other mixture component decrease), the acceptance rating of hedonic variables decreases.
Download figure to PowerPoint
The intensity of red color reaches the upper moderate level at a response value of 0.60 and only for the higher mixture-amounts design (Fig. 10), as the lower one (17%) was found not to have significant effects (Fig. 7). The red color intensity is reduced linearly with starch addition. In order to achieve optimal red color intensity, caseinate has to reach its maximum level (2.5%) and lard should be close to the maximum level (22%). Dependence of the red color intensity with caseinate has been previously reported by this group (Petridis et al. 2010), where caseinate was found to render the product more opaque, in a manner roughly comparable to that reported by Liu et al. (2007), attributed to the lowering of L-values in breakfast pork sausages due to water removal and subsequent reduction of the diffused color. Caseinate is a known water-binding material for meat products (Tsai et al. 1998; Pietrasik and Jarmoluk 2003). It can bind water from the gel matrix, reducing light diffraction, hence lowering the L-value of the sausages. As far as the effect of lard toward red color intensity is concerned, Pietrasik (1999) reports that redness values a* were inversely proportional to fat content, due to the increase of yellow-hue components of lard. The lower fat levels in the work in question coincide with the intermediate-to-high fat levels in the present experiment. In the overlap region between the two works, the objective red perception does indeed reduces with the increase in fat.
The center point for the acceptability of variables at the lower mixture amount (lard + caseinate + starch = 17%) corresponds to the moderate range of chewiness, consistency, and elasticity (response value −0.44, Fig. 11). Lard is again the most important component. At lower amounts of fat (negative deviation from the central point), the acceptability levels are low and remain so up to a content of levels of 13.85%. From that point onward, the acceptability increase is monotonic with lard content, approaching eventually maximal acceptable scores at the maximum amount of lard 17% corresponding to the range of “adequate.” Close to the center point caseinate proportion, 1.57% has reached its maximum acceptable score while starch acceptable scores decrease up to fairly high concentrations.
A different behavior is observed for the higher mixture amount (lard + caseinate + starch = 27%) (Fig. 12). At the reference blend point, all the component proportions (2.5 – 2.5 – 22%) approach their maximal acceptable scores which correspond to the range of “adequate.” Further increases in caseinate and starch proportion reduce the acceptability. It is noteworthy that caseinate and starch act in practically the same way toward acceptability.