Sensory evaluation of hardness, chewiness, springiness and adhesiveness of cheeses were correlated with instrumentally determined mechanical properties by Imoto et al. ( 1 979). They showed that the correlation changed with the compression ratio used. When trying to correlate sensory hardness with compression force, high correlation was found over all the compression ratios studied (20 to 80%) , whereas other parameters could only be satisfactorily correlated within a certain range of compression ratios. The authors suggested that the optimum compression ratio to be used in instrumental evaluation of cheeses should be selected for each property, u sing the fitness test for correlation between instrumental and sensory data. Results for processed cheese showed distinctly different mechanical responses to changing compression ratios than natural ripened cheeses.
The same textural characteristics studied by lmoto et al. ( 1 979) , plus cohesiveness and gumminess, were measured by Chen et al. ( 1 979a) for 1 1 types of cheese. Obj ective measurements were reported to have correlated closely with panel scores, with correlation coefficients ranging from 0 . 79 to 0.85. The obj ective measurement of adhesiveness was correlated negatively with the sensory values. This was attributed to the use of the plunger technique instead of the flat plate attachment for determining adhesiveness.
Sensory-instrumental texture correlation was found to be poor in a study with Cheddar and Cheshire cheeses (Green et al. , 1 985) . Sensory data obtained from trained panels proved to be better at discriminating b etween cheeses than the data from the compression tests. The lack of good correlation (coefficients under 0 . 70) was attributed to the fact that compression does not really mimic the kind of food failure that occurs in the m outh, which is basically cutting/ biting. Biting as opposed to simple compression b ecomes increasingly important when hard foods are b eing evaluated (Meullenet et al. , 1 997). The results from the instrumental tests
Literature Reviev) Ll 3 reported b y Green et al. ( 1 985) should b e considered with caution. As the authors pointed out, reproducibility among the replicates was fairly low. Besides, no transfonnation of the force-defonnation curves to true stress H encky strain was perfonned to account for the change in shape of the testing samples during compression.
Results from a study on food compression by soft machines (Campanella & Peleg, 1 988) indicated that the testing of foods with soft devices can serve as a means to incorporate the effects of the tissue defonnability into textural evaluations. As pointed out by the authors, the test conditions, particularly with respect to defonnation level and rate, should be as close as p ossible to those existing in sensory evaluations. Meaningful relationships between sensory and instrumental evaluation require , at least in principle, that the rheological properties of the sensory system be taken into account. It is reported , however, that the mathematical modelling of the sensory processes is very problematic.
A review on the relation b etween instrumental and sensory evaluation of the rheological and fracture properties of cheese showed that attributes like finnness, cohesiveness, graininess and springiness were correlated with instrumentally accessible p arameters (Zoon, 1 99 1 ) . Adhesiveness and chewiness, however, showed no clear reported correlation with instrumental p arameters to that date.
Jack et al. ( 1 993a) investigated the relationship between rheology and composition of Cheddar cheeses and sensory texture perceived by consumers. Free choice profiling was chosen as the technique for the sensory evaluation, since it requires almost no training time and is cheaper to run than other tests. The research showed that analysis of compositional and instrumental data did not discriminate between 1 9 samples of cheese in the same way as consumers perceived texture. Therefore , correlation between sensory and instrumental-chemical analyses was found to be limited and of restricted predictive capability. Such lack of agreement could have been caused by the fact that instrumental and sensory tests measured different properties, as mentioned in their discussion. Besides, the kind of association sought, between obj ective and consumer response , is likely to produce different
Literature R evieu) 1 14 outcomes to what would b e achieved had a trained , expert panel been used. An example of that can be found in Moskowitz et al. ( 1 979).
In another attempt to correlate sensory and instrumental data for Cheddar cheeses, Jack et al. ( 1 993b) applied double compression (TPA) , electromyography (EMG) and quantitative descriptive profiling to a range of commercial cheeses. Although it is reported that panellists were familiar with the product, little is mentioned about the extent of training they had. EMG traces of masticatory muscle activity were reported to b e unique for each subj ect for different samples, which led to inconsistent correlation with compression results and with sensory responses. The measured Instron variables discriminated between the cheese samples and , according to the authors, were good "predictors" of sensory perception.
The same authors, in a subsequent study, analysed the textural changes perceived in Cheddar cheese during mastication (Jack et al. , 1 994) . Their objective was to develop appropriate sensory methodology to account for mastication, while considering the relationship of such methodology with force-deformation properties. Texture profiles for the different samples changed during chewing in progressive profiling testing, but it was observed that swallowing varied from after 6 chews to 1 9 chews, depending on the panellists. Instrumental measurements were obtained usmg double compression to 60% in cycling mode, and , as expected, proved to correlate poorly to sensory data. This study emphasises the fact that mastication is of great importance in relation to texture and its perception, as previously reported by Brown et al. ( 1 996) , Cardello & Segars ( 1 989) and Christensen ( 1 984) .
Instrumental texture profile analysis results and sensory scores for reduced fat Cheddar cheeses correlated well with hardness (r :;;: 0 . 95) and springiness (r "" 0.95), as reported by Bryant et al. ( 1 995) . The Instron determination of adhesiveness correlated p ositively with sensory ratings (r =
0.73), b ut instrumental and sensory de terminations for cohesiveness did not correlate (r = -0.4 1 ) . The authors pointed out that uniform lubrication of compression plates and surface friction are critical factors for high correlation. Lubrication with mineral oil was used for the tests. Brennan &
Literature Review 1 1 5
BOUIne ( 1 994) reported that compression of foods in the mouth followed the nonlubricated pattern, even though lubrication was provided by saliva. Correlations between sensory and nonlubricated instrumental tests are not reported by Bryant et al. ( 1 995) .
Reggianito grating cheese was assessed for texture usmg sensory and instrumental measurements, with good correlations found between sensory results and parameters from a compression test (Hough et al., 1 996) . Strain at the breaking point was found to be the instrumental parameter that best correlated to visual , manual and oral sensory textural attributes. The correlations were analysed using partial least squares regression. Moisture level in the cheese was considered in the correlation procedure and proved to be a useful parameter for prediction of sensory texture.
Wium et al. ( 1 997) characterised Feta cheeses with different textures using uniaxial compression and observed that, among the four rheological parameters derived from the test, stress at fracture, modulus of deformability and work to fracture described the same type of information in the data set. Hencky strain at fracture described a different type of information by itself, as shown by the Principal Component Analysis results. Correlation of the rheological parameters with six sensory attributes, three oral (firmness, stickiness and crumbliness) and three nonoral (firmness, brittleness , spre ad ability) was reported. Firmness (oral and nonoral) was the sensory attribute best predicted by instrumental measurements, with stress at fracture being the best predictor at all deformation rates. Strain at fracture did not correlate well with data from sensory texture analysis. Combination of all four rheological parameters improved correlation coefficients slightly, but not enough to justify compromising the simplicity of the model with stress at fracture alone. Use of shear testing (frequency sweep, strain sweep and relaxation) to try and improve the predictive precision of the models showed that their contribution was only marginal and that, by themselves, these measurements were not very useful for sensory texture prediction (Wium &
Qvist, 1 998) .
Two different varieties of cheese, one semi hard (Appenzeller) and one very hard (Parmiggiano Reggiano) , were used for investigation of sensory
Literature Review 1 16 texture and instrumental (rheological and chemical) data in a European interlaboratory experiment (Noel et al. , 1 998) . Even though the results obtained can not be extrapolated to other cheese varieties, they showed that sensory textural attributes can be related to instrumental values when evaluated by p artial least square regression. Firmness and friability correlated p ositively to strain at fracture and dry matter and negatively to the modulus of deformability. Elasticity and deformability correlated to the same p arameters in the opposite direction, while dry matter appeared to correlate with adhesiveness. Dry matter, together with the force at peak height of 1 0% compression, was also found to be an influential objective variable for high correlation with sensory hardness, brittleness, cohesiveness and adhesiveness in several French cheeses (Antoniou et al. , 2000).
Pesenti & Luginbuhl ( 1999) used , among other methods, uniaxial compreSSIOn and stress relaxation/ creep tests to quantitatively assess cohesion in Gruyere-type cheeses of known sensory cohesiveness. Transient tests (relaxation and creep) seemed not to be as powerful as uniaxial compression and tension in distinguishing the two levels of cheese cohesiveness. Within the parameters obtained in uniaxial compression, stress at fracture, strain at fracture and work to fracture depended significantly (a. <
0 .0 1 ) on the level of cohesion. Significance was not found , however, for the cohesion-dependency of the modulus of deform ability (Young's modulus).
The results of a comprehensive work involving several natural and processed cheeses, evaluated instrumentally (TPA, frequency sweep and creep) and sensorially (in the hand and the mouth) were recently reported by Drake et al. ( 1 999b) . These authors observed that fundamental and empirical rheological techniques worked equally well in predicting sensory textural response. The u se of frequency sweep or creep parameters alone, however, led to very poor prediction of sensory textural attributes in that study. Overall, mouth and hand firmness were the terms best predicted. In a previous study, hand and mouth evaluated sensory attributes correlated highly and worked equally well for differentiation of cheese texture (Drake et al. , 1999a) . Drake et al. ( 19 99b) also reported that instrumental measurements predicted sensory attributes of processed cheeses better than natural cheeses.
Literature Review ll 7