J. Anim. Sci. 2003. 81:2741-2750
© 2003 American Society of Animal Science
Relationships of consumer sensory ratings, marbling score, and shear force value to consumer acceptance of beef strip loin steaks
W. J. Platter*,
J. D. Tatum*,1,
K. E. Belk*,
P. L. Chapman
,
J. A. Scanga* and
G. C. Smith*
* Department of Animal Sciences and
and
Department of Statistics, Colorado State University, Fort Collins 80523
 |
Abstract
|
|---|
Logistic regression was used to quantify and characterize the effects of changes in marbling score, Warner-Bratzler shear force (WBSF), and consumer panel sensory ratings for tenderness, juiciness, or flavor on the probability of overall consumer acceptance of strip loin steaks from beef carcasses (n = 550). Consumers (n = 489) evaluated steaks for tenderness, juiciness, and flavor using nine-point hedonic scales (1 = like extremely and 9 = dislike extremely) and for overall steak acceptance (satisfied or not satisfied). Predicted acceptance of steaks by consumers was high (>85%) when the mean consumer sensory rating for tenderness, juiciness, or flavor for a steak was 3 or lower on the hedonic scale. Conversely, predicted consumer acceptance of steaks was low (
10%) when the mean consumer rating for tenderness, juiciness, or flavor for a steak was 5 or higher on the hedonic scale. As mean consumer sensory ratings for tenderness, juiciness, or flavor decreased from 3 to 5, the probability of acceptance of steaks by consumers diminished rapidly in a linear fashion. These results suggest that small changes in consumer sensory ratings for these sensory traits have dramatic effects on the probability of acceptance of steaks by consumers. Marbling score displayed a weak (adjusted R2 = 0.053), yet significant (P < 0.01), relationship to acceptance of steaks by consumers, and the shape of the predicted probability curve for steak acceptance was approximately linear over the entire range of marbling scores (Traces67 to Slightly Abundant97), suggesting that the likelihood of consumer acceptance of steaks increases approximately 10% for each full marbling score increase between Slight to Slightly Abundant. The predicted probability curve for consumer acceptance of steaks was sigmoidal for the WBSF model, with a steep decline in predicted probability of acceptance as WBSF values increased from 3.0 to 5.5 kg. Changes in WBSF within the high (>5.5 kg) or low (<3.0 kg) portions of the range of WBSF values had little effect on the probability of consumer acceptance of steaks.
Key Words: Beef Consumer Satisfaction Marketing Palatability Prediction Tenderness
 |
Introduction
|
|---|
According to the 2001 to 2004 Beef Industry Long Range Plan, improving consumer satisfaction with the palatability and consistency of beef products is one of the U.S. beef industrys key strategies for attaining the goal of increasing consumer beef demand by 6% by 2004 (NCBA, 2001
). Beef palatability research studies often use traits such as marbling score, Warner-Bratzler shear force (WBSF), and consumer or trained taste panel evaluations of tenderness, juiciness, and flavor as indicators of beef palatability. However, only a few large consumer studies (Savell et al., 1987
; Miller et al., 2001
; Lorenzen et al., 2003
) have examined the impact of relative differences in these traits on overall consumer acceptance of steaks, because of the difficulty and costs associated with conducting large-scale consumer taste panels. Development of reliable statistical models representing relationships between the most commonly measured beef palatability traits and overall consumer acceptance of steaks would be useful to the beef industry, especially when interpreting results of beef palatability studies that have not directly measured overall consumer acceptability ratings in consumer taste panels. Consequently, the current analyses were undertaken to quantify and document the impacts of changes in several commonly measured beef traits (marbling score, WBSF, and consumer panel sensory ratings for tenderness, juiciness, or flavor) on the probability of overall consumer acceptance of beef strip loin steaks.
 |
Materials and Methods
|
|---|
Sample
Data presented in this report were from the same source as data reported by Platter et al. (2003)
. Detailed descriptions of the cattle management history, experimental procedures through harvest, and strip loin sample collection were provided in the preceding report (Platter et al., 2003
). Briefly, strip loins (Institutional Meat Purchase Specification 180; USDA, 1988
) were collected from the right sides of 550 carcasses originating from crossbred steers of various biological types. Strip loins were transported to the Colorado State University Meat Laboratory, where a single 2.54-cm-thick steak was removed from the anterior end of each strip loin. The remaining portion of the strip loin was placed in a vacuum-sealed bag and aged at 2°C for 14 d. After the aging period, the strip loin sections were frozen and stored at -20°C. Strip loin sections were fabricated (in the frozen state) into 2.54-cm-thick steaks using a band saw (model 5700, Hobart, Troy, OH). The first steak from the anterior end of each strip loin was identified and placed in an individual vacuum-sealed bag for subsequent WBSF determination. The next two steaks were identified, individually vacuum-packaged, and stored for subsequent untrained, consumer taste panel analysis. Upon completion of the fabrication process, steaks were sorted for intended use and returned to frozen storage (-20°C). Samples used for sensory analysis were stored frozen for approximately 180 d.
Shear Force Determination
The 14-d aged steak from each strip loin section was removed from freezer storage and allowed to thaw for 24 h at 2°C. Steaks were cooked on an electric conveyor grill (model TGB-60, Magikitchn, Quakertown, PA) for 6 min and 35 s at a setting of 176°C to a target internal temperature of 70°C. After cooking, each steak was allowed to equilibrate to room temperature (22°C), and 6 to 10 1.27-cm-diameter cores were removed from each steak parallel to the muscle fiber orientation. Each core was sheared once, perpendicular to the muscle fiber orientation, using a Warner-Bratzler shear machine (G-R Electric Manufacturing Co., Manhattan, KS), and peak shear force measurements were recorded and averaged to obtain a single WBSF value for each steak.
Consumer Sensory Panels
Branson Research Associates (Bryan, TX) performed extensive demographic analyses of the Denver, CO, metropolitan area to select a test population representative of the age, income, and ethnic background of the U.S. population. Consumers were contacted by telephone and prescreened to ensure that they were at least 18 yr of age and consumers of beef products. Four testing sites, located in the Denver metropolitan area (Denver, Lakewood, Arvada, and Littleton), were used to conduct a total of 25 consumer panel sessions that included a total of 489 consumers. Because this sample population represented consumers from diverse age levels, income levels, and ethnic backgrounds, and included consumers categorized as "beef-eaters," the consumer demographic profile was considered to be an acceptable population to model relationships of consumer sensory ratings, WBSF, and marbling scores to overall acceptance of steaks by consumers.
Within a session, paired steaks from 22 different carcasses were selected for sensory evaluation. Steak identification numbers, as well as order of service to consumer panelists, were assigned randomly to each of the 20 consumers per session. Frozen steaks prepared for consumer panel evaluation were thawed at 2°C for 24 h and cooked for approximately 15 min on electric grills (model GGR64, Salton, Inc., Mt. Prospect, IL) designed to heat steaks from both sides, simultaneously, to a final internal temperature of 70°C. A Type K thermocouple (Omega Engineering Inc., Stamford, CT) was placed in the geometric center of each steak, and internal temperature was monitored during cooking using a microprocessor thermometer (model HH21, Omega Engineering Inc.). Steaks were cut into 1.3 x 1.3 x 2.5-cm portions, covered, and placed in a warming oven (49°C) until consumers were served. One steak from each pair of steak samples from each carcass identification number was prepared for serving during the first half of the session, whereas the other matched steak was prepared for serving in the last half of the session to minimize any changes in sensory attributes associated with holding samples for longer periods of time in a warming oven.
Consumer panel evaluation procedures used for this study were approved by the Colorado State University, Use of Humans in Research Committee. At each location, consumers were randomly seated at tables arranged in a circular order in a room containing standard fluorescent lighting. Instructions regarding the structure of the ballot and sampling procedures for the steak samples were provided verbally to the consumers in each session. Panelists were provided double-distilled, deionized water and saltless saltine crackers and were instructed to take a bite of cracker and a drink of water before evaluating each sample to cleanse their palates and to minimize sensory fatigue between samples.
Consumers rated each steak sample for like/dislike of tenderness, flavor, and juiciness using nine-point, end-anchored hedonic scales (1 = like extremely and 9 = dislike extremely). Additionally, consumers were asked if they were satisfied ("yes" or "no") with the overall eating quality of each sample.
Statistical Approach and Rational
The statistical technique chosen to model the relationship between consumer sensory ratings, marbling, or WBSF and overall acceptability of steaks by consumers was logistic regression analysis. Logistic regression analysis models the relationship between a binary or ordinal response variable (e.g., "yes" or "no" response) and one or more explanatory variables (Ott and Longnecker, 2001
). Logistic regression transforms the dependent variable into a logit variable (the natural log odds of the response variable occurring or not). After transformation of the dependent variable, logistic regression applies maximum likelihood estimation, and, in this way, can estimate the probability that a certain event will occur. Maximum likelihood estimation seeks to maximize the log likelihood that the observed values of the dependent variable will be predicted from the observed values of the independent variables. Unlike ordinary least squares regression, logistic regression does not require normally distributed variables, does not assume linearity of relationship between the independent variables and the dependent variable, and does not assume homoscedasticity (Hosmer and Lemeshow, 2000
). The binomial response variable in this case was the overall consumer acceptance ("yes" or "no") of a steak. Explanatory variables included WBSF value, marbling score, or mean consumer sensory rating for tenderness, juiciness, or flavor.
Numerous studies have documented the existence of a tenderness gradient across longissimus muscle steaks that may impact WBSF and sensory tenderness ratings of longissimus muscle (Smith et al., 1969
; Wheeler et al., 1996
; Kerth et al., 2002
). Additionally, the ASTM (1968)
Manual on Sensory Testing Methods, Special Technical Bulletin 434, stated that, although the variability of consumer preference testing is high, the use of larger numbers of panelists would improve the discrimination of the test. These reports suggest that both location of the sample within a steak and panelist bias may influence individual consumer sensory ratings. Averaging consumer sensory ratings by steak reduces the variance of consumer panel data (Polkinghorne et al., 1999
). Nevertheless, outlier sensory rating responses made by individual consumers can greatly influence average steak sensory ratings. Recently, researchers employed by Meat and Livestock Australia developed standards for a new grading system (Meat Standards Australia) to describe palatability of beef based solely on the use of sensory results from consumer testing (Thompson, 2002
). Outliers among consumer sensory ratings were prevalent in the dataset used for the development of Meat Standards Australia, requiring researchers to devise a method to minimize the impact of these observations on the palatability prediction equation (Polkinghorne et al., 1999
). In that analysis, average consumer sensory ratings were calculated after four responses (the two highest and the two lowest sensory ratings) out of a total of 10 individual consumer responses for each steak were removed from the dataset (Polkinghorne et al., 1999
). In our study, 7 to 10 consumers were asked if they were satisfied (yes or no) with the overall eating quality of each steak. Effects of individual consumer bias were minimized in our analyses by using an average consumer satisfaction rating, by steak (without removal of outliers), to designate whether the overall palatability of a steak was acceptable to consumers. Steaks were designated as acceptable, in our analyses, if two-thirds or more (an average of 66% or more) of the consumers sampling each steak indicated that they were satisfied with its overall eating quality.
Data Analysis
Analyses were conducted to examine the correlation structure of the data using the PROC CORR procedure, and logistic regression equations were developed using the PROC LOGISTIC procedure of SAS (SAS Inst., Inc., Cary, NC). Generalized adjusted coefficients of determination were calculated for each model using the RSQ option of the LOGISTIC procedure. Predicted probability values were obtained from each logistic regression model, and the accuracy of these predictions was tested against actual observations in the original dataset via a classification table using a procedure that approximates an unbiased "jackknifing" method (SAS, 1999
). The predictive accuracy of each model also was tested on a separate population (Wheeler et al., 2002
) by computing the predicted probability values of overall acceptance of steaks by consumers and comparing these predictions with actual observations via a classification table.
 |
Results and Discussion
|
|---|
Consumer Panel Participants
Frequency distributions for consumer demographic attributes are reported by category as a percentage of the total sample population (Table 1
). Approximately 56% of the consumer panelists were female and 44% were male. A high percentage (80%) of panelists had some form of postsecondary education, and about 52% were married. The age of consumers in the sample population ranged from 18 to 84 yr, and the mean age of consumers was 46 yr. Of the incomes reported, the highest percentage of consumers had a total household income of between $40,000 and $49,000/yr, but incomes ranged from less than $10,000/yr to more than $90,000/yr among panelists. A high percentage of consumers reported their ethnicity as Caucasian (76%), followed by African-American (11%), and Hispanic (8%). A very low percentage of the sample population reported an ethnicity of American Indian and Asian or Pacific Islander.
Survey responses of panelists consumption and eating preferences are presented in Table 2
. Seventy-five percent of the consumers in the sample population were the primary shopper of the household. Fifty-eight percent of panelists reported consuming beef as a portion of an evening meal three or more meals per week. Fewer consumers reported consuming pork (5%) and poultry (31%) as a portion of an evening meal three or more times each week. A small percentage (9%) of consumers reported consuming a vegetarian evening meal three or more times each week. A majority (93%) of panelists reported eating an evening meal outside of the home at least once per week. Fifty-two percent of the consumers in the sample population listed tenderness as the most important sensory attribute when purchasing beef, whereas 38% of consumers considered flavor and 11% of consumers considered juiciness, as most important. These results are remarkably similar to consumer survey responses reported by Huffman et al. (1996)
, where 51% of consumers listed tenderness, 39% of consumers listed flavor, and 10% of consumers listed juiciness as the most important beef sensory attribute when determining their eating satisfaction in a home or restaurant environment.
Marbling and Shear Force Characteristics of Sample
Distribution of marbling scores of steaks in the experimental sample is provided in Figure 1
. According to data from the 2000 National Beef Quality Audit, only 0.43% of carcasses from fed steers and heifers had Moderately Abundant or Abundant marbling scores and only 0.02% of carcasses had marbling scores of Practically Devoid (McKenna et al., 2002
). The mean marbling score of steaks used in this analysis was 458 ± 105 (Small58) with a range in marbling score from Traces67 to Slightly Abundant97, resulting in a sample that closely resembled the range in marbling scores of the majority of U.S. fed steer and heifer carcasses.

View larger version (18K):
[in this window]
[in a new window]
|
Figure 1. Frequency distribution of marbling scores of steaks in the test population (n = 550). Marbling score: 200 = Traces, 300 = Slight, 400 = Small, 500 = Modest, 600 = Moderate, and 700 = Slightly Abundant.
|
|
The distribution of WBSF values is presented in Figure 2
. The range for WBSF values was 2.33 to 7.51 kg, with a mean WBSF value of 4.35 ± 0.93 kg. Other studies (Morgan et al., 1991
; George et al., 1999
; Brooks et al., 2000
) have reported slightly lower means, but similar ranges, for WBSF values for retail strip loin steaks.
Correlation Analysis
Results of the analysis of the correlation structure of the data revealed moderate to high correlations (P < 0.05) among mean marbling scores, WBSF values, and mean consumer palatability ratings (Table 3
). The correlation between consumer tenderness ratings and WBSF values was moderately high (r = 0.63). Marbling scores were moderately correlated with WBSF, consumer tenderness ratings, consumer juiciness ratings, and consumer flavor ratings (r = -0.31, -0.27, -0.34, -0.22, respectively). High, positive correlations (r = 0.80 to 0.84) were observed among all consumer sensory ratings.
View this table:
[in this window]
[in a new window]
|
Table 3. Simple correlation coefficients of mean marbling scores, mean shear force values, and mean consumer palatability ratings
|
|
Interpretation of Logistic Regression Diagnostic Statistics
Statistics describing the strength of the relationship between the independent variables and the dependent variable, the discriminatory power, and the success of a regression equation are different for logistic regression than for ordinary linear regression (Hosmer and Lemeshow, 2000
). There is no direct analog to the coefficient of determination (R2) in logistic regression; however, a generalized R2 (R2 adj) that attempts to measure the strength of association of a logistic regression model has been proposed (Nagelkerke, 1991
). The area under the receiver operating characteristic curve (c-statistic) is a measure of the discriminatory power of the logistic equation (Hosmer and Lemeshow, 2000
). The value of the c-statistic can range from 0.50 (the models predictions are no better than chance) to 1.0 (the model can perfectly discriminate the observed responses in the sample). Classification tables are often used to tally correct and incorrect estimates of the logistic regression model and measure the predictive accuracy of the model (SAS, 1995
). A 2 x 2 classification table for binary response variables can be created by comparing the predicted outcomes of the model with the actual observations in the dataset. When the value of the estimated probability of an observation is greater than, or equal to, 0.50, the observation is classified as a predicted event (i.e., an acceptable steak). When the estimated probability is less than 0.50, the observation is classified as a predicted nonevent (i.e., an unacceptable steak). From the classification table, the probability that the model correctly classifies the sample data (percentage correct) can be calculated. Additionally, the ratio of the number of correctly classified acceptable steaks to the total number of acceptable steaks (sensitivity) and the ratio of the number of correctly classified unacceptable steaks to the total number of unacceptable steaks (specificity) can be derived from a classification table.
Sensory Rating Models
Figure 3
displays plots of the predicted probability curves for overall acceptance of steaks by consumers as derived from the cumulative logit response functions of average steak consumer ratings for tenderness, juiciness, or flavor. The R2 adj values for predicting overall acceptance of steaks by consumers were 0.574, 0.516, and 0.520 for models using average steak consumer ratings for tenderness, juiciness, and flavor, respectively. The discriminatory power of the sensory rating models was high (c-statistic = 0.869 to 0.891), and the models correctly classified a large percentage of the observations (percentage correct = 77.7 to 79.3%). Sensitivity percentages were higher than specificity percentages for consumer tenderness (83.2 vs. 75.2%), juiciness (78.6 vs. 76.7%), and flavor (82.1 vs. 75.6) rating models predicting overall acceptance of steaks by consumers (data not shown). Ordinary least squares regression analysis performed by Huffman et al. (1996)
indicated that consumer tenderness ratings accounted for the most variation (R2 = 0.56) in overall consumer palatability ratings for steaks prepared in a restaurant setting, whereas consumer flavor ratings accounted for the most variation (R2 = 0.67) in overall consumer palatability ratings for steaks prepared in the home. In that study, a three-variable ordinary, least squares regression model that included tenderness, juiciness, and flavor ratings accounted for 79% of the variation in consumer overall palatability ratings (Huffman et al., 1996
). Similarly, in the current study, a three-variable logistic regression equation including average steak consumer ratings for tenderness, juiciness, and flavor had a stronger relationship (R2 adj = 0.62) to overall steak acceptance and more discriminatory power (c-statistic = 0.908) than single consumer sensory rating models (data not shown).

View larger version (23K):
[in this window]
[in a new window]
|
Figure 3. Predicted probability of overall consumer acceptance of steaks by mean consumer rating for tenderness, juiciness, and flavor. Consumer like to dislike rating for tenderness, flavor, and juiciness as follows: 1 = like extremely and 9 = dislike extremely. The R2 adj is a generalized coefficient of determination. The c-statistic is the area under the receiver operating characteristic curve. Percentage correct is the percentage of observations in the dataset correctly classified by the logistic regression equation. The symbol "P" represents the predicted probability for a steak being rated as acceptable by consumers. The constant "e" equals the base of the natural logarithm (2.718282).
|
|
Predicted probability curves for overall acceptance of steaks by consumers were similar for all sensory rating models. This was expected, given the high correlations among consumer sensory ratings for tenderness, juiciness, and flavor (Table 3
). Predicted overall acceptance of steaks by consumers was high (>85%) if the average consumer sensory rating for tenderness, juiciness, or flavor for a steak was three, or lower, on a nine-point hedonic scale. Conversely, predicted consumer acceptance of steaks was low (
10%) when the average consumer rating for tenderness, juiciness, or flavor for a steak was five or higher. As mean consumer sensory ratings for tenderness, juiciness, or flavor decreased from three to five, the probability of overall acceptance of steaks by consumers diminished rapidly in a linear fashion. These results suggest that small changes in consumer sensory ratings for these sensory traits have dramatic effects on the probability of overall consumer acceptance of steaks. Additionally, these results suggest a mean consumer sensory rating for tenderness, juiciness, or flavor which is at, or beyond, the mid-point of a nine-point scale would result in a low probability of overall steak acceptance by consumers.
Marbling Score Model
The probability curve for overall acceptance of steaks by consumers, as predicted by the cumulative logit response functions of marbling scores, is presented in Figure 4
. The R2 adj value for predicting overall steak acceptance for the marbling score model was 0.053, suggesting a relatively weak relationship between marbling score and overall consumer acceptability. The marbling score model had a c-statistic of 0.574, and correctly classified 54.5% of the observations, indicating low discriminatory power of the model. These results are consistent with research that has shown marbling scores to have a low to moderate relationship with overall beef palatability ratings (Crouse and Smith, 1978
; Tatum et al., 1982
; Smith et al., 1984
).

View larger version (15K):
[in this window]
[in a new window]
|
Figure 4. Predicted probability of overall consumer acceptance of steaks by mean marbling score. Marbling score: 200 = Traces, 300 = Slight, 400 = Small, 500 = Modest, 600 = Moderate, and 700 = Slightly Abundant. The R2 adj is a generalized coefficient of determination. The c-statistic is the area under the receiver operating characteristic curve. Percentage correct is the percentage of observations in the dataset correctly classified by the logistic regression equation. The symbol "P" represents the predicted probability for a steak being rated as acceptable by consumers. The constant "e" equals the base of the natural logarithm (2.718282).
|
|
Marbling score is used as the primary predictor of beef palatability among carcasses of similar maturity characteristics in the USDA beef grading system (USDA, 1997
). Based on results of a large consumer retail beef study, Savell et al. (1987)
concluded that increasing the amount of marbling in top loin steaks had a positive impact on the eating quality of beef. Trained sensory panel results from other studies have indicated that overall palatability ratings of beef steaks generally increased as marbling score increased, but the differences in palatability ratings for each successive increase in marbling score were not always statistically significant (Tatum et al., 1980
; Smith et al., 1984
). Interestingly, the shape of the predicted probability curve for overall acceptance of steaks by consumers in this analysis was nearly linear over the entire range of marbling scores (Traces67 to Slightly Abundant97), and suggests that the likelihood of consumer acceptance of steaks increases approximately 10% for each full marbling score increase from Slight to Slightly Abundant (Figure 4
).
Shear Force Model
A majority of consumers in this (Table 2
) and other studies (Huffman et al., 1996
; CCA, 2002
) indicated that tenderness is the single most important palatability trait for determining overall steak acceptance. Warner-Bratzler shear force is routinely used by scientists as an objective measurement of meat tenderness and, despite criticism, has remained the most popular and accurate instrumental measurement of meat tenderness (Wheeler et al., 1997
). Otremba et al. (1999)
and Wheeler et al. (1996)
reported WBSF value correlations (r) of -0.68 and -0.85, respectively, with trained panelist overall tenderness ratings for longissimus steaks. Shackelford et al. (1995)
reported that a single variable regression equation using WBSF values explained 73% of the variation in trained panelist overall tenderness ratings for longissimus muscle steaks. Because tenderness is an important driver of overall steak acceptability and WBSF is used as an objective measure of tenderness, it is reasonable to assume that WBSF values could be used in a logistic regression model to predict overall acceptability of steaks by consumers.
Figure 5
represents a plot of the predicted probability curve for overall consumer acceptance of steaks resulting from the cumulative logit response functions of WBSF value. The strength of the relationship between consumer acceptance of steaks and WBSF values was moderate (R2 adj = 0.225). The WBSF model had moderate discriminatory power (c-statistic = 0.738) and correctly classified 66.7% of the observations. Boleman et al. (1997)
demonstrated that consumer perceptions of tenderness and overall satisfaction of beef top loin steaks were affected when steaks were segmented into categories based on WBSF value. Lorenzen et al. (2003)
reported low correlations of WBSF values with "in-home" consumer panelist ratings of tenderness (r = -0.26) and consumer overall likeness ratings (r = -0.18) for top loin steaks, but attributed the lack of relationship of these traits, in part, to the variation in steak preparation encountered during "in-home" consumer studies.

View larger version (13K):
[in this window]
[in a new window]
|
Figure 5. Predicted probability of overall consumer acceptance of steaks by mean shear force value. The R2 adj is a generalized coefficient of determination. The c-statistic is the area under the receiver operating characteristic curve. Percentage correct is the percentage of observations in the dataset correctly classified by the logistic regression equation. The symbol "P" represents the predicted probability for a steak being rated as acceptable by consumers. The constant "e" equals the base of the natural logarithm (2.718282).
|
|
Shackelford et al. (1991)
used trained taste panel responses to determine WBSF "threshold" values for predicting acceptable consumer steak tenderness ratings ("slightly tender" or higher). Results of that analysis indicated that a WBSF value of 4.6 kg would have a 50% chance, and a WBSF value of 3.9 kg would have a 68% chance, of being rated as acceptable in tenderness by consumers. Results of our analysis produced results similar to those of Shackelford et al. (1991)
, with predicted probabilities of consumer steak acceptance of 50 and 68% at approximate WBSF values of 4.4 and 3.7 kg, respectively. Many researchers have employed various WBSF values to describe consumer tenderness acceptability "thresholds" in beef palatability studies (Shackelford et al., 1995
; Miller et al., 2001
; Vote et al., 2003
). In a review, Meilgaard et al. (1999)
stated that some experts question the validity of sensory "thresholds" because they are ill-defined in theory, may not reproduce results well, and may not even exist.
An advantage of the current form of analysis describing the relationship of WBSF values to overall acceptance of steaks by consumers is that the relationship can be compared over a wide range of WBSF values. The predicted probability curve for overall acceptance of steaks by consumers was sigmoid shaped for the WBSF model, with a steep decline in predicted probability of acceptance as WBSF values increased from 3.0 to 5.5 kg. Changes in WBSF within the high (>5.5 kg) or low (<3.0 kg) portions of the range of WBSF values had little effect on the probability of consumer acceptance of steaks.
Validation of Logistic Regression Models
Consumer sensory ratings, marbling score, and WBSF value models developed in our analysis were tested against the population described by Wheeler et al. (2002)
. Predicted probabilities for observations in the validation dataset were calculated from the logistic regression equations developed from the original population and compared with actual consumer acceptance ratings via a classification table. Of the three sensory ratings models, the tenderness-rating model was the most accurate (78.7%), whereas the juiciness and flavor rating models were less accurate (63.7 and 68.7%, respectively) for determining whether two-thirds of consumers would have rated steaks as acceptable. The equations from the marbling score and WBSF models were 57.3 and 70.7% accurate, respectively, for determining whether two-thirds of consumers would have rated steaks as acceptable. Applying the logistic regression equations to a separate dataset for validation purposes produced results similar to those observed in the original analyses.
 |
Implications
|
|---|
The relationship between common measurements of beef palatability traits and overall consumer steak acceptance levels is extremely important when interpreting the results of beef palatability research. Marketing implications of small, but significant, differences in consumer sensory ratings, marbling scores, or shear force values can be difficult to interpret in some beef palatability studies. Results of our study may assist researchers in describing the potential effects of results from research studies of beef palatability on overall steak acceptability as perceived by consumers.
1 Correspondencephone: 970-491-6530; fax: 970-491-0278; E-mail: dtatum{at}lamar.colostate.edu.
Received for publication February 18, 2003.
Accepted for publication June 9, 2003.
 |
Literature Cited
|
|---|
ASTM. 1968. Manual on Sensory Testing Methods. STP 434. Am. Soc. Testing Material, Philadelphia, PA.
Boleman S. J., S. L. Boleman, R. K. Miller, J. F. Taylor, H. R. Cross, T. L. Wheeler, M. Koohmaraie, S. D. Shackelford, M. F. Miller, R. L. West, D. D. Johnson, and J. W. Savell. 1997. Consumer evaluation of beef of known categories of tenderness. J. Anim. Sci. 75:15211524.[Abstract/Free Full Text]
Brooks, J. C., J. B. Belew, D. B. Griffin, B. L. Gwartney, D. S. Hale, W. R. Henning, D. D. Johnson, J. B. Morgan, F. C. Parrish Jr., J. O. Reagan, and J. W. Savell. 2000. National beef tenderness survey1998. J. Anim. Sci. 78:18521860.[Abstract/Free Full Text]
CCA. 2002. Beef Consumer Satisfaction Benchmark Study. Canadian Cattlemens Assoc., Calgary, AB.
Crouse, J. D., G. M. Smith, and R. W. Mandigo. 1978. Relationship of selected beef carcass traits with meat palatability. J. Food Sci. 43:152157.
George, M. H., J. D. Tatum, K. E. Belk, and G. C. Smith. 1999. An audit of retail beef loin steak tenderness conducted in eight U.S. cities. J. Anim. Sci. 77:17351741.[Abstract/Free Full Text]
Hosmer, D. W., and S. Lemeshow. 2000. Applied Logistic Regression. 2nd ed. John Wiley and Sons, Inc., New York, NY.
Huffman, K. L., M. F. Miller, L. C. Hoover, C. K. Wu, H. C. Brittin, and C. B. Ramsey. 1996. Effect of beef tenderness on consumer satisfaction with steaks consumed in the home and restaurant. J. Anim. Sci. 74:9197.[Abstract]
Kerth, C. R., J. L. Montgomery, J. L. Lansdell, C. B. Ramsey, and M. F. Miller. 2002. Shear gradient in longissimus steaks. J. Anim. Sci. 80:23902395.[Abstract/Free Full Text]
Lorenzen, C. L., R. K. Miller, J. F. Taylor, T. R. Neely, J. D. Tatum, J. W. Wise, M. J. Buyck, J. O. Reagan, and J. W. Savell. 2003. Beef customer satisfaction: Trained sensory panel ratings and Warner-Bratzler shear force values. J. Anim. Sci. 81:143149.[Abstract/Free Full Text]
McKenna, D. R., D. L. Roeber, P. K. Bates, T. B. Schmidt, D. S. Hale, D. B. Griffin, J. W. Savell, J. C. Brooks, J. B. Morgan, T. H. Montgomery, K. E. Belk, and G. C. Smith. 2002. National beef quality audit-2000: Survey of targeted cattle and carcass characteristics related to quality, quantity, and value of fed steers and heifers. J. Anim. Sci. 80:12121222.[Abstract/Free Full Text]
Meilgaard, M. C., G. V. Civille, and B. T. Carr. 1999. Determining thresholds. Pages 123132 in Sensory Evaluation Techniques. 3rd ed. CRC Press LLC, Boca Raton, FL.
Miller, M. F., M. A. Carr, C. B. Ramsey, K. L. Crockett, and L. C. Hoover. 2001. Consumer thresholds for establishing the value of beef tenderness. J. Anim. Sci. 79:30623068.[Abstract/Free Full Text]
Morgan, J. B., J. W. Savell, D. S. Hale, R. K. Miller, D. B. Griffin, H. R. Cross, and S. D. Shackelford. 1991. National beef tenderness survey. J. Anim. Sci. 69:32743283.[Abstract]
Nagelkerke, N. J. D. 1991. A note on a general definition of the coefficient of determination. Biometrika 78:691692.[Abstract/Free Full Text]
NCBA. 2001. Beef Industry Long Range Plan 20012004. Natl. Cattlemens Beef Assoc., Centennial, CO.
Ott, R. L., and M. Longnecker. 2001. Multiple regression and the general linear model. Page 675 in Statistical Methods and Data Analysis. 5th ed. Duxbury, Pacific Grove, CA.
Otremba, M. M., M. E. Dikeman, G. A. Milliken, S. L. Stroda, J. A. Unruh, and E. Chambers IV. 1999. Interrelations among evaluations of beef longissimus and semitendinosus muscle tenderness by Warner-Bratzler shear force, a descriptive-texture profile sensory panel and a descriptive attribute sensory panel. J. Anim. Sci. 77:865873.[Abstract/Free Full Text]
Platter, W. J., J. D. Tatum, K. E. Belk, J. A. Scanga, and G. C. Smith. 2003. Effects of repetitive use of hormonal implants on beef carcass quality, tenderness, and consumer ratings of beef palatability. J. Anim. Sci. 81:984996.[Abstract/Free Full Text]
Polkinghorne, R., R. Watson, M. Porter, A. Gee, J. Scott, and J. Thompson. 1999. Meat Standards Australia, a PACCP based beef grading scheme for consumers: The use of consumer scores to set grade standards. Pages 1415 in Proc. 45th ICoMST, Yokahama, Japan.
SAS. 1995. Logistic Regression Examples Using the SAS® System, Version 6. 1st ed. SAS Inst. Inc., Cary, NC.
SAS. 1999. The logistic procedure. Page 1957 in SAS/STAT Users Guide. Ver. 8. SAS Inst. Inc., Cary, NC.
Savell, J. W., R. E. Branson, H. R. Cross, D. M. Stiffler, J. W. Wise, D. B. Griffin, and G. C. Smith. 1987. National consumer beef retail study: Palatability evaluations of beef loin steaks that differed in marbling. J. Food Sci. 52:517519.
Shackelford, S. D., J. B. Morgan, H. R. Cross, and J. W. Savell. 1991. Identification of threshold levels for Warner-Bratzler shear force in beef top loin steaks. J. Muscle Foods 2:289296.
Shackelford, S. D., T. L. Wheeler, and M. Koohmaraie. 1995. Relationship between shear force and trained sensory panel tenderness ratings of 10 major muscles from Bos indicus and Bos taurus cattle. J. Anim. Sci. 73:33333340.[Abstract]
Smith, G. C., Z. L. Carpenter, H. R. Cross, C. E. Murphey, H. C. Abraham, J. W. Savell, G. W. Davis, B. W. Berry, and F. C. Parrish Jr. 1984. Relationship of USDA marbling groups to palatability of cooked beef. J. Food Qual. 7:289308.
Smith, G. C., Z. L. Carpenter, and G. T. King. 1969. Considerations for beef tenderness evaluations. J. Food Sci. 34:612618.
Tatum, J. D., G. C. Smith, B. W. Berry, C. E. Murphey, F. L. Williams, and Z. L. Carpenter. 1980. Carcass characteristics, time on feed and cooked beef palatability attributes. J. Anim. Sci. 50:833840.[Abstract/Free Full Text]
Tatum, J. D., G. C. Smith, and Z. L. Carpenter. 1982. Interrelationships between marbling, subcutaneous fat thickness and cooked beef palatability. J Anim. Sci. 54:777784.[Abstract/Free Full Text]
Thompson, J. 2002. Managing meat tenderness. Pages 1727 in Proc. 48th ICoMST., Rome, Italy.
USDA. 1988. Institutional Meat Purchase Specifications for Fresh Beef. Agric. Marketing Serv., USDA, Washington, DC.
USDA. 1997. United States standards for grades of carcass beef. Agric. Marketing Serv., USDA, Washington, DC.
Vote, D. J., K. E. Belk, J. D. Tatum, J. A. Scanga, and G. C. Smith. 2003. On-line prediction of beef tenderness using a computer vision system equipped with a Beefcam® module. J. Anim. Sci. 81:457465.[Abstract/Free Full Text]
Wheeler, T. L., S. D. Shackelford, L. P. Johnson, M. F. Miller, R. K. Miller, and M. Koohmaraie. 1997. A comparison of Warner-Bratzler shear force assessment within and among institutions. J. Anim. Sci. 75:24232432.[Abstract/Free Full Text]
Wheeler, T. L., S. D. Shackelford, and M. Koohmaraie. 1996. Sampling, cooking and coring effects on Warner-Bratzler shear force values in beef. J. Anim. Sci. 74:15531562.[Abstract]
Wheeler, T. L., D. Vote, J. M. Leheska, S. D. Shackelford, K. E. Belk, D. M. Wulf, B. L. Gwartney, and M. Koohmaraie. 2002. The efficacy of three objective systems for identifying beef cuts that can be guaranteed tender. J. Anim. Sci. 80:33153327.[Abstract/Free Full Text]
This article has been cited by other articles:

|
 |

|
 |
 
B. A. Schneider, J. D. Tatum, T. E. Engle, and T. C. Bryant
Effects of heifer finishing implants on beef carcass traits and longissimus tenderness
J Anim Sci,
August 1, 2007;
85(8):
2019 - 2030.
[Abstract]
[Full Text]
[PDF]
|
 |
|

|
 |

|
 |
 
T. Okumura, K. Saito, H. Sakuma, T. Nade, S. Nakayama, K. Fujita, and T. Kawamura
Intramuscular fat deposition in principal muscles from twenty-four to thirty months of age using identical twins of Japanese Black steers
J Anim Sci,
August 1, 2007;
85(8):
1902 - 1907.
[Abstract]
[Full Text]
[PDF]
|
 |
|

|
 |

|
 |
 
S. L. Gruber, J. D. Tatum, J. A. Scanga, P. L. Chapman, G. C. Smith, and K. E. Belk
Effects of postmortem aging and USDA quality grade on Warner-Bratzler shear force values of seventeen individual beef muscles
J Anim Sci,
December 1, 2006;
84(12):
3387 - 3396.
[Abstract]
[Full Text]
[PDF]
|
 |
|

|
 |

|
 |
 
W. T. Choat, J. A. Paterson, B. M. Rainey, M. C. King, G. C. Smith, K. E. Belk, and R. J. Lipsey
The effects of cattle sex on carcass characteristics and longissimus muscle palatability
J Anim Sci,
July 1, 2006;
84(7):
1820 - 1826.
[Abstract]
[Full Text]
[PDF]
|
 |
|

|
 |

|
 |
 
Y. M. Cho, H. Choi, I. H. Hwang, Y. K. Kim, and K. H. Myung
Effects of 25-hydroxyvitamin D3 and manipulated dietary cation-anion difference on the tenderness of beef from cull native Korean cows
J Anim Sci,
June 1, 2006;
84(6):
1481 - 1488.
[Abstract]
[Full Text]
[PDF]
|
 |
|

|
 |

|
 |
 
W. J. Platter, J. D. Tatum, K. E. Belk, S. R. Koontz, P. L. Chapman, and G. C. Smith
Effects of marbling and shear force on consumers' willingness to pay for beef strip loin steaks
J Anim Sci,
April 1, 2005;
83(4):
890 - 899.
[Abstract]
[Full Text]
[PDF]
|
 |
|