Sensory Evaluation of Alternative Turkey Genotypes

Two genotypes were raised for this study: conventional, fast-growing Broad Breasted White (Fast) and a slow-growing, “heritage” Standard Bronze (Slow). These birds were raised on a small commercial operation in Kansas. The fast-growing turkeys were obtained from a commercial source as day-old poults. The slow-growing turkeys were placed in May 2007 and the fast-growing turkeys in mid-July 2007. Forty turkeys of each genotype were raised for 14 and 24-28 wk, respectively, with the goal of raising the birds to a similar slaughter weight. The genotypes were raised in separate flocks and only females were used. The same feed was used for both genotypes. The turkeys were harvested in late October 2007. They were transported to a small processing plant one hour away. After processing, carcasses were transported under refrigeration to the University of Arkansas Pilot Poultry Processing plant where they were placed in storage at 33 F. All carcasses were stored in refrigeration until further analysis (described below).

At 2 d postmortem, 16 carcasses of each genotype were placed on cones and cut up for parts yield analysis and deboning of breast fillets. Breast and skin samples were taken for nutrient analysis and frozen until analysis. The remaining meat was repackaged and vacuum sealed for the consumer focus group and placed back in storage at 33 F. The pH was also measured at 2 d postmortem. The right breast fillet was weighed after deboning and the color of the meat was measured. Breast fillets were cooked on racks in pans to an internal temperature of 165 F. The cooked fillets were reweighed to determine cook loss, an indicator of water-holding capacity. The tenderness was measured by an instrumental method, Meullenet-Owens razor shear (MORS; Cavitt et al., 2004).

Descriptive analysis was conducted on breast and thigh meat at the University of Arkansas Department of Food Science sensory facility by a trained meat descriptive panel (15 members; Sensory Spectrum Inc., Chatham, NJ). Initial orientation was held to refine flavor and texture definitions. At 8 d postmortem, 4 turkeys of each type were cooked whole for flavor descriptive analysis. Carcasses were placed on a rack in a shallow pan and roasted at 325 F. After the breast was browned, a loose tent of aluminum foil was placed over the breast to prevent overcooking. The carcass was roasted until the thigh reached an internal temperature of 180 F. Carcasses were cooled for 30 m before cutting. Breast samples were taken of cooked meat for nutrient analysis. Meat was cut into 0.75-inch, bite-size cubes and was not seasoned or salted. The descriptive panel used 15 descriptive flavor terms to describe the white meat: sweet, salty, sour, bitter, cooked white meat, white meat fat, brothy, blood serum/metallic, organy, sulfury, sweet aromatic, other, astringent, and associated aftertastes. Dark meat flavor was described with similar flavor terms.

At 9 d postmortem, 4 Fast and 4 Slow turkeys were cooked for texture analysis. The trained panel used descriptive textural attributes to evaluate tenderness characteristics of breast meat. Initial hardness, cohesiveness, moisture release were evaluated in the first bite stage, while hardness of mass, cohesiveness of mass, fibrousness, and number of chews to swallow were evaluated in the chewdown stage. All intensities were expressed to one significant digit on 15-point numerical scales, with higher scores indicating higher attribute intensities.

Panelists for the descriptive tests were randomly presented samples from all treatment groups in duplicate utilizing a randomized complete block design. Panelists were randomly presented samples from all treatment groups in duplicate. Between each sample, panelists were instructed to cleanse their palate with distilled water and unsalted crackers. A 15-minute break period was allocated to the panelists halfway through the session.

At 10 d postmortem, a consumer test (40 panelists) was conducted on the breast and thigh meat from the Slow and Fast turkeys, with the addition of a third treatment, frozen retail turkey (Retail). The Retail turkey was purchased at a local grocery store and thawed at 33 F. It was cut up at the University of Arkansas processing plant like the other treatments. The Retail turkey contained additional ingredients (water, salt, phosphates, and spices) from injection as is common in retail turkeys in the US. These turkeys are fast-growing birds from indoor production systems and the inclusion of this treatment helped elaborate differences between specialty and conventional turkey products. Four turkeys of each type, Retail, Fast, and Slow (12 total), were cooked for consumer analysis. At 11 d postmortem, the procedure was repeated with an additional 40 panelists (80 panelists total). Whole turkeys were roasted as described above, and breast and thigh meat cut into bite-size cubes.

The consumer panelists were recruited via a panelist database housed at the University of Arkansas Food Science Department, and selection of panelists was based on consumption patterns of white and dark poultry meat and on the consumption of natural and/or organic poultry products. Institutional Review Board protocol was followed. Consumer panelists were each served samples from the different treatments one at a time and were instructed to cleanse their palates between samples. They were asked to evaluate their overall liking of the product, liking of appearance, texture, and flavor, and appropriateness of color, tenderness, juiciness, and flavor. They were also asked to indicate the likelihood that they would buy the product and if they would pay more for the product than for their usual poultry product. Nine-point hedonic scales were used to assess overall liking and liking of appearance, texture, and of flavor (1 = dislike extremely, 5 = neither like nor dislike, 9 = like extremely). Just-About-Right (JAR) scales were used to assess the appropriateness of color (1 = much too light, 2 = too light, 3 = just about right, 4 = too dark, 5 = much too dark), the appropriateness of tenderness (1 = much too tough, 5 = much too tender), the appropriateness of juiciness (1 = much too dry; 5 = much too juicy), and the appropriateness of flavor (1 = much too weak, 5 = much too strong). Samples were randomized by product type and then by meat type (breast meat or thigh meat). The use of Just-About-Right (JAR) scales are useful for diagnostics, because hedonic scales do not allow determination of the appropriateness of intensity of the attribute (Meilgaard et al., 1999). These questions were based on the differences determined earlier by the descriptive panel.

For the consumer focus group and community outreach, a nonprofit organization that focuses on local food, Ozark Slow Food, organized a “turkey tasting” event at a local restaurant in which these same 3 turkey treatments were sampled. Consumers with an interest in natural and local food were targeted and were contacted in various ways. Specifically, Ozark Slow Food developed a series of three informational pieces for the public on specialty poultry production with outdoor access, endangered farm breeds and their cultural contest as “heritage” breeds, and the “turkey tasting” consumer event. These informational pieces were posted over a period of 3 weeks to the OzarkSlowFood email listserver with a distribution of 120 recipients. The turkey-tasting event was announced on public calendars and the local public radio station. The public radio station also interviewed UA researchers about the project, as well as a local heritage turkey producer, and the interview was broadcasted in a show called Ozarks at Large on November 2, 2007 the day before the event. The turkey tasting event was also announced throughout Northwest Arkansas with posters at natural food stores and announcements by Ozark Slow Food at local events and at farmers markets. The community outreach and announcements of the event was done in the weeks before Thanksgiving.

The turkey tasting event was held at a local upscale restaurant, Basil’s Restaurant in Rogers, AR, on November 3, 2007, which was 5 d postmortem to the slaughter of the turkeys. Meat was cut up and prepared according to recipes selected by the restaurant chefs but all meat was cooked to 165 F.

An introductory session was held first to provide basic information and definitions on specialty turkeys. A presentation on Slow Food was also provided by visiting professor Dr. Paulo Sambo from the University of Padova. Following the presentation, participants completed a survey, sampled and ranked the three turkey products, conventional (marinated), fast-growing free-range turkey, and slow-growing heritage free-range turkey, in a blind test. The roasted breast and thigh meat samples were offered as appetizers with bread. After completing the survey and ranking, a full holiday meal made from locally-raised, seasonal ingredients was served. The participants had the opportunity to further sample the turkey products as part of their meal.

The meat quality data were analyzed as a t-test. The descriptive panel sensory data were analyzed as a randomized complete block design with 2 experimental units per treatment in each block. Genotype was treated as a fixed effect and panelist as a random effect. The consumer panel sensory data were subjected to analysis of variance. The JAR scores were compared using a chi-square test for equality of distributions except in those cases where small expected counts may have substantially affected the approximate p-value from the chi-square. In those cases, Fisher’s exact test was used (Fleiss, 2003).

In regard to the consumer survey and ranking, for the binomial distributions from the yes/no questions, a 95% Confidence Interval of the proportion of successes or positive answers was calculated. For multiple category questions, respondents who checked more than one answer were removed. We used a chi-square test for equality of distributions except in those cases where small expected counts may have substantially affected the approximate p-value from the chi-square where Fisher's exact test was used. If there was a significant difference among categories, we calculated confidence intervals for the difference between each possible pair of proportions based on the properties of a multinomial distribution. The variance estimate of the difference included a correlation because all questions were answered by same people. If 0 was not in the Confidence Interval, the proportions were different.

For ranking data regarding the reason for buying and of product preference, we used the Friedman test which is a nonparametric two-way analysis of variance of ranks and used the Person as the blocking variable. To separate the ranks, we used the average of the ranks (Least Squares Means) rather than the means because not all individuals answered all questions. P-values smaller than 5% were considered to be statistically significant. All statistical analyses were carried out using SAS version 9 (SAS Institute, Inc., Cary, NC).