Relationships Between Descriptive Beef Flavor Attributes and Consumer Liking

NOVEL BEEF FLAVOR RESEARCH Relationships Between Descriptive Beef Flavor Attributes and Consumer Liking Rhonda K. Miller*, Chris R. Kerth, and Koushik Adhikari Rhonda Miller, Ph.D. 2471 TAMU Department of Animal Science Texas A&M University College Station, TX 77843-2471 rmiller@tamu.edu INTRODUCTION Beef flavor has been defined as an important component of beef demand. Beef flavor, however, is not a single attribute, but is composed of multiple attributes that can be dynamic. The beef industry took the first big step in addressing beef flavor by funding the development of the beef flavor lexicon (Adhikari et al., 2011) that identified major and minor beef flavor components. Systems for identifying beef flavor cannot be developed if we do not know what beef flavor is or how the human perceives beef flavor. Now that the beef lexicon has been developed, understanding how components of beef flavor are related to consumer acceptance, volatile aromatic compounds, and meat chemical components can assist different segments of the beef industry in understanding how to enhance, control or maximize beef flavor to increase beef consumption. Recent research has established relationships between consumer acceptance, descriptive sensory attributes of flavor from the Beef Lexicon, juiciness and tenderness, volatile aromatic compounds, and meat chemical components. Results from three major projects that have examined these relationships in light and moderate to heavy beef eaters will be discussed. Additional information on beef flavor drivers for millenials that are either light or moderate to heavy beef eaters compared to non-millenials will be presented. RESULTS AND DISCUSSION The Beef Flavor Lexicon for whole muscle beef cuts was developed and can be found at Adhikari et al. (2011) and in AMSA (2014). This lexicon identifies beef flavor attributes and provides references for anchoring trained descriptive panelists on each attribute. Additionally, AMSA (2014) provides guidance on how to train a panel to use the lexicon. The first question asked after development of the lexicon was can it be used to determine flavor differences in beef that would expectantly differ in beef flavor based on past research. We designed a project that examined the effect of USDA Beef Quality Grades (Choice and Select), cooking method (gas grill, clam shell grill, and crockpot), internal cook temperature endpoint (58, 70 and 80 C), and cut (top loin steaks, top sirloin steaks, clod steaks, bottom round steaks and bottom round roasts) (Miller and Kerth, 2012). A portion of these data are presented in Table 1 for top loin, top sirloin and clod steaks. Using major beef flavor attributes, grade classification did not differ (P>0.05) for the three steaks, except high ph top loin steaks had higher fat-like flavor than Select or Choice top loin steaks. However, cooking method and internal cook temperature endpoint impacted beef flavor attributes to a great extent than USDA Quality grade. When steaks were cooked on a gas grill, top loin steaks had higher (p<0.05) levels of beef identity, brown/roasted, and fat-like flavor attributes; top sirloin steaks had higher (p<0.05) brown/roasted and fat-like flavor attributes; and clod steaks had higher (P<0.05) brown/roasted and lower (p<0.05) bloody/serumy and metallic flavor attributes then steaks cooked on a clamshell grill. As internal cook temperature increased, top loin and clod steaks had higher (p<0.05) levels of brown/roasted and lower levels of bloody/serumy and top loin steaks had lower (p<0.05) levels of metallic flavor attributes. Previous research had shown that cooking method and internal cook temperature affect beef flavor and we concluded that the Beef Flavor Lexicon was adequate for identifying flavor differences in whole muscle beef cuts. While the previous research documented that the Beef Flavor Lexicon was an effective tool for understanding beef flavor differences, research to understand factors that drive beef flavor and how these factors where related to consumer acceptance had not been conducted. We designed two research projects. The first project selected consumers that ate beef 3 or more times per week and these consumers were classified as heavy beef eaters. Sixteen treatments were used and are defined in Table 2 American Meat Science Association 17

Table 1. Least squares means for selected beef descriptive flavor attributes for beef top loin steaks, top sirloin and clod steaks segmented by grade classification, cooking method and internal cook temperature endpoint from Miller and Kerth (2012). Top Loin Steaks Top Sirloin Steaks Clod Steaks Beef Brown/ Bloody/ Fat- Beef Brown/ Bloody/ Fat- Beef Brown/ Bloody/ Fat- Effect identity Roasted Serumy like Metallic identity Roasted Serumy like Metallic Identity Roasted Serumy like Metallic Grade Classification 0.59 0.49 0.86 <0.0001 0.38 0.73 0.49 0.51 0.69 0.79 0.35 0.10 0.58 0.92 0.99 Select 4.8 1.8 1.6 1.4 a 1.8 5.1 2.0 1.7 1.6 2.0 4.7 1.5 2.2 1.9 1.9 Choice 5.0 1.9 1.7 1.5 a 1.9 5.0 1.7 1.9 1.6 2.1 4.9 1.8 2.0 1.9 1.9 High ph 5.0 1.6 1.7 2.0 b 1.7 - - - - - - - - - - Cooking Method 0.004 <0.0001 0.15 <0.0001 0.48 0.09 0.01 0.25 0.01 0.63 0.08 <0.0001 0.02 0.79 0.01 Gas grill 5.2 b 2.2 b 1.9 1.8 b 1.8 5.2 2.4 b 1.6 1.8 b 2.0 4.9 2.1 b 1.8 a 1.9 1.8 a Clamshell grill 4.6 a 1.3 a 1.6 1.4 a 1.8 4.8 1.3 a 2.0 1.4 a 1.2 4.7 1.1 a 2.3 b 1.9 2.1 b Internal Cook Temperature Endpoint, C 0.11 0.001 0.0002 0.30 0.005 0.36 0.33 0.09 0.53 0.10 0.31 <0.0001 <0.001 0.35 0.40 58 4.8 1.4 a 2.2 b 1.7 2.1 b 4.9 1.7 2.3 1.7 2.3 4.7 1.2 a 3.1 c 2.0 2.0 70 4.7 1.6 a 1.9 b 1.6 1.7 a 4.9 1.6 1.9 1.6 2.1 5.0 1.6 b 1.9 b 1.8 1.8 80 5.2 2.6 b 1.0 a 1.6 1.6 a 5.3 2.2 1.3 1.5 1.7 4.8 2.1 c 1.3 a 2.0 1.9 Root Mean Square Error 0.63 0.58 0.66 0.29 0.39 0.11 0.57 0.93 0.84 0.32 0.54 0.37 0.38 0.58 0.32 0.27 abcd Mean values within a column and effect followed by the same letter are not significantly different (P > 0.05). to vary in USDA Beef Quality Grade, cut, cooking method, internal cook temperature endpoint and meat ph. Subprimals were obtained from beef carcasses (n=10 per cut) and steaks and roasts were cut within a subprimal and cut for descriptive flavor evaluation using the Beef Flavor Lexicon and consumer sensory evaluation in four cities (Houston TX, Philadelphia PA, Portland OR, and Olathe KS). Using the major beef flavor attributes (all beef flavor attributes were determined and are reported in Miller and Kerth, 2013 and Glascock, 2014), beef flavor differed (p<0.05) across treatments for the beef flavor attributes presented except live-like flavor (Table 2). Similarly, consumer liking ratings for overall, flavor, beef flavor and grill flavor differed (P<0.05) by treatment (Table 3). Consumers liked steaks cooked on the grill compared to steaks cooked on the clamshell grill, and roasts cooked in crockpots were lowest in liking ratings. High ph top loin steaks cooked on the clamshell grill and steaks cooked on the grill to 58 C were not liked as well as comparable steaks with normal ph. These results showed that consumers and trained descriptive flavor attributes panelists reported differences in beef flavor due to treatments, but these results did not show how consumer and trained descriptive flavor attributes were related. To understand relationships between beef descriptive flavor attributes and consumer sensory attributes for heavy beef eaters, partial least squares regression analysis was conducted (Figure 1). Consumer liking attributes were closely clustered with overall liking indicating that there was a strong relationship between consumer liking ratings or consumers liked beef samples where they rated high for overall flavor, beef flavor and grilled flavor. Descriptive flavor attributes most closely clustered with overall consumer liking was fat-like flavor with brown/roasted, butter, and salty somewhat related to overall consumer liking. The treatments most closely related to consumer liking were Choice top loin steaks cooked on a grill to 58 and 80 C, Choice top sirloin steaks cooked on the grill to 80 C, and high ph top loin steaks cooked on the grill to 80 C. Bloody/serumy, metallic, bitter, sour, sour milk and medicinal were related and most closely associated with Choice top sirloin steaks cooked on the clamshell grill to 58 and 80 C and cooked on the grill to 58 C. Beef identity, sweet, overall sweet and umami were closely related beef flavor 18 68 th Annual Reciprocal Meat Conference

Table 2. Beef flavor descriptive attributes (0=none; 15=extremely intense) least squares means for 16 beef cuts adapted from Miller and Kerth (2013), Glascock (2014) and Kerth and Miller (2015). Beef Brown/ Bloody/ Fat- Liver- Treatment identity roasted serumy like Metallic like P value g <0.001 <0.001 <0.001 <0.001 <0.001 0.17 Choice Top Loin steaks Clamshell grill, 58 C 10.1 bcd 0.8 abc 2.5 de 1.3 cd 2.6 cd 0.0 a Clamshell grill, 80 C 10.8 def 0.9 abcd 1.6 b 1.2 cd 2.0 ab 0.4 b Grill, 58 C 10.4 cde 1.4 d 2.7 e 1.5 de 2.5 cd 0.0 a Grill, 80 C 11.4 ef 2.2 e 1.8 b 1.3 d 2.2 b 0.1 a High ph Top Loin steaks Clamshell grill, 58 C 8.8 a 0.6 ab 2.7 e 1.6 e 2.2 bc 0.0 a Clamshell grill, 80 C 9.7 abc 1.0 abcd 1.5 ab 1.3 d 1.6 a 0.2 ab Grill, 58 C 8.8 a 0.6 ab 2.7 e 1.6 de 2.2 bc 0.0 a Grill, 80 C 11.1 def 2.1 e 1.6 b 1.9 e 1.7 a 0.0 a Choice Top Sirloin steaks Clamshell grill, 58 C 9.4 ab 0.9 abcd 3.3 e 1.4 de 2.8 d 0.1 a Clamshell grill, 80 C 10.2 bcd 1.1 bcd 1.7 b 0.9 ab 2.1 bc 0.1 ab Grill, 58 C 9.7 abc 1.1 bcd 3.0 e 1.5 de 2.9 d 0.1 a Grill, 80 C 11.6 f 2.5 e 1.7 b 1.0 abcd 2.1 b 0.0 a Choice Bottom Round roasts Crockpot, 58 C 9.9 bc 0.5 a 2.7 de 1.1 bcd 2.6 d 0.3 ab Crockpot, 80 C 11.0 def 1.2 cd 1.1 ab 1.0 abc 1.7 a 0.2 ab Select Bottom Round roasts Crockpot, 58 C 9.2 a 0.4 a 2.3 ce 1.0 abc 2.5 c 0.0 a Crockpot, 80 C 11.3 ef 0.9 abcd 0.9 a 1.0 abc 1.7 a 0.4 b Root Mean Square Error 1.04 0.61 0.60 0.31 0.38 0.28 abcdef Mean values within a column followed by the same letter are not significantly different (P > 0.05). g P - value from analysis of variance tables. attributes. Negative descriptive beef flavor attributes, attributes in the opposite quadrants to overall consumer liking, were cardboard, warmed over flavor, liver-like, sour aromatic and musty. Interestingly, Choice and Select bottom round roasts cooked to either 58 or 70 C, Choice top loin steaks cooked on the clamshell grill to 80 C and high ph top loin steaks cooked on the clamshell grill were most closely associated with negative descriptive beef flavor attributes. These were the first data to segment beef descriptive flavor attributes from the Beef Flavor Lexicon into positive and negative consumer flavor attributes and these data also reported groupings of beef descriptive flavor attributes. The aforementioned study used heavy beef eaters to understand relationships between consumer and descriptive flavor attributes and these data will help the beef industry to understand how flavor impacts heavy users of their product. The next question was if light beef eaters, consumers who eat beef 1 to 2 times per week, had the same relationships between consumer liking and beef descriptive flavor attributes. A study was conducted using steaks and roasts from 20 treatments (Table 4) with consumers in College Park PA, Olathe KS and Portland OR. An expert, descriptive flavor attribute sensory panel also evaluated the same samples. The effect of treatment for beef descriptive flavor attributes and consumer attributes are reported in Tables 4 and 5, respectively. Treatments were as defined in the study using heavy beef eaters, but four additional treatments were added that included Choice tenderloin steaks cooked on a grill or a clamshell grill to either 58 or 80 C. Trained descriptive and consumer sensory panels also rated juiciness and tenderness in this study. Steaks and roasts differed (P<0.05) in the major beef flavor descriptive attributes and results were similar to those reported in Table 2. Consumer sensory attributes differed (p<0.05) by treatments and consumers tended to rate Choice tenderloin steaks grill to either 58 or 80 C highest (p<0.05) for liking across attributes. These results indicated that light beef eaters responded similarly to heavy beef eaters by treatments for beef flavor. Principal component analysis was conducted to understand the relationship between the 20 treatments and light beef consumer sensory liking traits (Figure 2). Overall consumer liking was most closely related to overall flavor, beef flavor and grilled flavor liking. Juiciness and tenderness liking were closely related, and while closely clustered with American Meat Science Association 19

Table 3. Least squares means for consumer attributes for 16 beef cut, grade, ph and internal temperature endpoints treatments Miller and Kerth (2013) and Glascock (2014). Overall Flavor Beef flavor Grill flavor Effect like/dislike like/dislike like/dislike like/dislike P value f <0.0001 <0.0001 <0.0001 <0.0001 Choice Top Loin steak Clamshell grill, 58 C 6.2 d 6.1 d 6.2 cd 5.4 c Clamshell grill, 80 C 5.7 c 5.7 cd 6.0 c 5.1 c Grill, 58 C 7.3 f 7.2 e 7.2 e 7.2 e Grill, 80 C 6.8 e 6.8 e 6.7 de 6.8 e High ph Top Loin steaks Clamshell grill, 58 C 5.6 bc 5.3 bc 5.5 b 5.1 c Clamshell grill, 80 C 5.5 bc 5.4 bc 5.7 bc 4.8 b Grill, 58 C 5.7 c 5.6 c 5.8 bc 5.5 c Grill, 80 C 6.5 de 6.3 de 6.5 d 6.3 d Choice Top Sirloin steaks Clamshell grill, 58 C 5.8 c 5.8 cd 6.0 c 4.9 bc Clamshell grill, 80 C 5.2 b 5.2 bc 5.5 bc 4.7 bc Grill, 58 C 6.8 e 6.9 e 7.0 e 6.4 de Grill, 80 C 6.0 cd 6.1 d 6.3 cd 6.3 de Choice Bottom Round roasts Crockpot, 58 C 4.8 ab 5.0 b 5.3 ab 4.2 ab Crockpot, 80 C 4.8 ab 5.0 ab 5.3 ab 4.6 b Select Bottom Round roasts Crockpot, 58 C 4.8 ab 4.9 ab 5.0 ab 3.9 a Crockpot, 80 C 4.5 a 4.5 a 4.8 a 3.9 a RMSE e 6.5 2.02 1.99 2.00 abcdef Mean values within a column and effect followed by the same letter are not significantly different (P > 0.05). f P - value from analysis of variance tables. Figure 1. Principal component analysis of trained descriptive flavor attributes, cooking treatments and consumer liking for heavy beef eaters where Se = Select, Ch=Choice, HpH=high ph, GF=Clamshell grill, Gr=Grill, CP=Crockpot, LM=Top loin steak, GM=Top sirloin steak, BF=bottom round roast, and WOF=Warmed over flavor Miller and Kerth (2013) and Glascock (2014). 20 68 th Annual Reciprocal Meat Conference

Table 4. Beef flavor descriptive attributes (0=none; 15=extremely intense) least squares means for 20 beef treatments across cooking methods, USDA Quality Grade, ph and internal temperature endpoint treatments adapted from Miller et al. (2014). Beef Brown/ Bloody/ Fat- Liver- Treatment identity roasted serumy like Metallic like P-value j <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0003 Choice Tenderloin steaks Clamshell grill, 58 C 5.5 bcd 0.9 abc 2.0 ef 1.3 efg 2.0 fgh 0.4 bcdef Clamshell grill, 80 C 5.7 de 1.0 bc 0.9 b 1.4 fgh 1.6 abc 0.4 bcdef Grill, 58 C 6.4 f 2.0 f 1.7 cde 1.5 fghi 2.0 defgh 0.2 abcd Grill, 80 C 6.9 gh 2.2 f 1.1 bc 1.6 ghi 1.7 bcd 0.1 ab High ph Top Loin steaks Clamshell grill, 58 C 4.9 a 0.7 ab 2.8 g 1.6 ghi 2.0 efgh 0.6 f Clamshell grill, 80 C 5.3 abc 0.8 abc 1.3 bcd 1.2 def 1.5 ab 0.4 bcde Grill, 58 C 5.8 de 1.5 e 2.4 fg 1.6 ghi 1.8 bcdef 0.6 f Grill, 80 C 6.3 f 2.3 fg 1.2 bc 1.7 hi 1.4 ab 0.4 bcdef Choice Top Loin steak Clamshell grill, 58 C 5.3 abc 0.7 abc 2.5 fg 1.5 fghi 2.1 gh 0.3 abcde Clamshell grill, 80 C 5.5 cd 1.1 cd 0.9 ab 1.2 def 1.5 ab 0.3 abcde Grill, 58 C 6.5 fg 1.9 f 2.1 ef 1.7 i 1.9 cdefg 0.2 abc Grill, 80 C 7.2 h 2.6 g 1.1 bc 1.5 fghi 1.7 bcd 0.0 a Choice Top Sirloin steaks Clamshell grill, 58 C 5.8 de 1.4 de 2.5 fg 1.0 cde 2.2 h 0.3 abcde Clamshell grill, 80 C 6.1 ef 2.0 f 1.0 b 0.9 bcd 1.7 abc 0.2 ab Grill, 58 C 5.2 ab 0.6 ab 1.8 de 0.8 abc 1.8 cdefg 0.3 bcdef Grill, 80 C 5.1 ab 0.7 abc 1.1 b 0.8 abc 1.7 abc 0.5 def Choice Bottom Round roasts Crockpot, 58 C 5.2 abc 0.6 a 1.2 bc 0.9 abcd 1.7 bcde 0.5 ef Crockpot, 80 C 5.0 a 0.6 a 0.3 a 0.7 abc 1.4 a 0.5 ef Select Bottom Round roasts Crockpot, 58 C 5.0 a 0.5 a 1.3 bcd 0.7 ab 1.9 cdefg 0.6 f Crockpot, 80 C 5.0 ab 0.6 a 0.3 a 0.6 a 1.5 ab 0.5 cdef RMSE k 0.47 0.40 0.59 0.36 0.32 0.30 abcdef Mean values within a column and effect followed by the same letter are not significantly different (P > 0.05). j P-value from analysis of variance tables. k Root Mean Square Error GF=Clamshell grill clamshell grill American Meat Science Association 21

Table 5. Least squares means for consumer attributes (1=dislike extremely; 9=like extremely) for 20 beef treatments across cooking methods, USDA Quality Grade, ph and internal temperature endpoint treatments adapted from Miller et al. (2014). Overall Flavor Beef flavor Grill flavor Juiciness Tenderness Treatment like/dislike like/dislike like/dislike like/dislike like/dislike like/dislike P-value m <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 Choice Tenderloin steaks Clamshell grill, 58 C 6.7 jkl 6.6 hi 6.5 hi 5.7 f 7.2 k 7.7 jk Clamshell grill, 80 C 6.2 hij 6.0 efg 6.1 fgh 5.5 ef 6.1 efg 7.2 i Grill, 58 C 7.1 l 6.9 i 6.8 ij 6.5 gh 7.0 jk 7.9 k Grill, 80 C 7.0 l 7.0 i 7.0 ij 7.0 hi 6.5 ghi 7.3 ij High ph Top Loin steaks Clamshell grill, 58 C 5.2 bcde 5.2 bc 5.3 abc 4.7 bc 6.1 efg 5.9 ef Clamshell grill, 80 C 5.4 cdef 5.3 bc 5.6 cdef 5.0 cd 5.7 de 5.5 de Grill, 58 C 6.3 efgh 5.8 defg 5.9 defg 5.6 f 6.8 ijk 6.4 fg Grill, 80 C 6.5 ijk 6.3 gh 6.3 gh 6.5 gh 6.5 ghij 6.5 gh Choice Top Loin steak Clamshell grill, 58 C 6.1 ghi 6.2 fgh 6.2 gh 5.3 def 6.7 hijk 6.6 h Clamshell grill, 80 C 5.7 efgh 5.9 defg 6.0 efg 5.3 def 5.3 d 5.6 de Grill, 58 C 6.8 kl 6.9 i 7.0ij 6.8 ghi 7.0 jk 6.4 fg Grill, 80 C 6.8 kl 7.1 i 7.1 j 7.2 i 6.1 efg 6.2 fg Choice Top Sirloin steaks Clamshell grill, 58 C 6.7 kl 6.9 i 7.0 ij 6.5 gh 7.1 k 6.5 fgh Clamshell grill, 80 C 5.6 defg 6.2 fgh 6.3 gh 6.4 g 4.5 c 4.0 bc Grill, 58 C 5.8 fgh 5.7 cdef 5.9 efg 5.0 cde 6.3 fgh 6.4 fgh Grill, 80 C 5.1 bcd 5.4 bcd 5.6 bcde 5.0 cd 4.1 bc 4.7 b Choice Bottom Round roasts Crockpot, 58 C 5.0 bc 5.1 ab 5.4 abcd 4.2 ab 6.1 eg 5.4 cd Crockpot, 80 C 4.5 a 5.0 ab 5.2 abc 4.5 ab 3.7 ab 4.0 a Select Bottom Round roasts Crockpot, 58 C 4.8 ab 4.9 ab 5.1 ab 4.1 a 5.9 ef 5.2 cd Crockpot, 80 C 4.4 a 4.7 a 5.1 a 4.2 a 3.3 a 3.9 a Root Mean Square Error 2.00 2.00 1.97 1.91 2.04 2.13 abcdefghijkl Mean values within a column followed by the same letter are not significantly different (P > 0.05). m P-value from analysis of variance tables. overall liking, juiciness and tenderness liking were not as closely related to overall liking as the flavor liking attributes. This indicated that flavor attributes were slightly more highly related to overall liking than juiciness and tenderness liking. Consumers liked grilled Choice tenderloin steaks cooked to 58 and 80 C, grilled Choice top loin steaks cooked to 58 C, and grilled Choice top sirloin steaks cooked to 58 C. Consumers did not like Choice and Select bottom round roasts cooked to 58 or 80 C, Choice top sirloin steaks cooked on a clamshell grill to 80 C, and high ph top loin steaks cooked on a clamshell grill to 58 or 80 C. To understand how beef descriptive flavor attributes were related to consumer sensory attributes and treatments a partial least squares regression analysis was conducted and reported in Figure 3. Beef descriptive flavor attributes most closely clustered with consumer overall liking were fat-like, sweet, overall sweet, and salty attributes. Beef identity, umami, brown/roasted and burnt flavor attributes were related to overall consumer liking, but they were not as highly related. Juiciness and muscle fiber tenderness attributes were closely clustered to consumer juiciness and tenderness liking. These results show that trained and consumer panelists rate these attributes similarly and that they are positive attributes to overall consumer liking. Metallic and bloody/serumy flavor attributes were more closely clustered with juiciness liking and descriptive juiciness ratings indicating that treatments with higher levels of metallic and bloody/serumy were juicier. Negative descriptive flavor attributes were musty, cardboard, sour aromatic, green hay-like, and liver-like. These attributes were more closely related to the negative 22 68 th Annual Reciprocal Meat Conference

Figure 2. Principal component biplot of consumer liking sensory attributes (in blue) and 20 treatments (in green) where 111 = tenderloin steaks grilled to 58 C; 112 = tenderloin steaks grilled to 80 C; 121 = tenderloin steaks clamshell grill to 58 C; 122 = tenderloin steaks clamshell grill to 80 C; 211 = high ph top loin steaks grilled to 58 C; 212 = high ph top loin steaks grilled to 80 C; 221 = high ph top loin steaks clamshell grill to 58 C; 222 = high ph top loin steaks clamshell grill to 80 C; 331 = Choice bottom round roasts cooked in a crockpot to 58 C; 332 = Choice bottom round roasts cooked in a crockpot to 80 C; 431 = Select bottom round roasts cooked in a crockpot to 58 C; 432 = Select bottom round roasts cooked in a crockpot to 80 C; 511 = Choice top loin steaks grilled to 58 C; 512 = Choice top loin steaks grilled to 80 C; 521 = Choice top loin steaks clamshell grill to 58 C; 522 = Choice top loin steaks clamshell grill to 80 C; and 611 = Choice top sirloin steaks grilled to 58 C; 612 = Choice top sirloin steaks grilled to 80 C; 621 = Choice top sirloin steaks clamshell grill to 58 C; 622 = Choice top sirloin steaks clamshell grill to 80 C for light beef eaters adapted from Miller et al. (2014). treatments previously discussed. These results indicated that light beef eaters rated beef flavor similarly to heavy beef eaters and that differences in juiciness and tenderness, whether evaluated by a trained or consumer panel, were similarly rated. To more fully understand if light and heavy beef eaters respond to beef flavor differences or have similar liking for beef flavor attributes, the data from the light and heavy beef eaters studies were combined and analyzed using partial least squares regression (Figure 4). In addition to examining the effect of beef consumption frequency, the effect of age (millennial versus non-millennial) could be addressed with this analysis as consumer demographics were collected for the two studies. Consumer information from the National Cattlemens Beef Association (data not published) has indicated that millennials (ages 18 to 35) do not eat beef as frequently as non-millennials. As nonmillennials are the beef consumers of the future, it is important to understand if relationships between consumer liking and beef flavor attributes are the same. In other words, does usage (light versus heavy) or age (millennial versus non-millennial) impact how consumers perceive beef flavor and do differences in beef flavor drive consumer acceptance similarly. Therefore, data were segmented into four subclasses, heavy beef eaters that were millennials and non-millennials, and light beef eaters that were millennials and non-millennials. The data are presented as two biplots (Figure 4a and 4b). As juiciness and tender- American Meat Science Association 23

Figure 3. Partial least squares regression biplot (R 2 = 0.72) of trained descriptive flavor attributes from the Beef Lexicon (in red), consumer sensory attributes (in blue), and 20 treatments (in green) where 111 = tenderloin steaks grilled to 58 C; 112 = tenderloin steaks grilled to 80 C; 121 = tenderloin steaks Clamshell grill to 58 C; 122 = tenderloin steaks Clamshell grill to 180 C; 211 = high ph top loin steaks grilled to 58 C; 212 = high ph top loin steaks grilled to 80 C; 221 = high ph top loin steaks Clamshell grill to 58 C; 222 = high ph top loin steaks Clamshell grill to 80 C; 331 = Choice bottom round roasts cooked in a crockpot to 58 C; 332 = Choice bottom round roasts cooked in a crockpot to 80 C; 431 = Select bottom round roasts cooked in a crockpot to 58 C; 432 = Select bottom round roasts cooked in a crockpot to 80 C; 511 = Choice top loin steaks grilled to 58 C; 512 = Choice top loin steaks grilled to 80 C; 521 = Choice top loin steaks Clamshell grill to 58 C; 522 = Choice top loin steaks Clamshell grill to 80 C; and 611 = Choice top sirloin steaks grilled to 58 C; 612 = Choice top sirloin steaks grilled to 80 C; 621 = Choice top sirloin steaks Clamshell grill to 58 C; 622 = Choice top sirloin steaks Clamshell grill to 80 C for light beef eaters adapted from Miller et al. (2014) and Kerth and Miller (2015). 24 68 th Annual Reciprocal Meat Conference

Figure 4. Partial least squares regression biplot (R 2 =0.65) to (a) predict consumer sensory attributes (in blue) using trained descriptive beef flavor attributes (in red) for (b) heavy and light beef eaters, and millennial and non-millennial consumers (in green) from Miller and Kerth (2013) and Miller et al. (2014), respectively,. American Meat Science Association 25

ness attributes were evaluated in the light beef eater study only, these variables were removed from this analysis and for simplicity, some beef flavor descriptive attributes were not included. In Figure 4a, the relationship between consumer liking and beef descriptive flavor attributes across the two studies is presented. As reported for each study, fat-like is the beef flavor attribute most closely related to overall consumer liking. Brown/roasted and salty beef flavor attributes were closely related to consumer grill and flavor liking with umami somewhat closely related to the aforementioned flavor attributes. Cardboardy and liverlike were negatively associated with consumer liking ratings. Overall sweet, bloody/serumy, metallic, beef identity, sweet and bitter while positively related to consumer overall liking, these attributes clustered together and were negatively influenced by factor 2 or the vertical axis of the biplot. These relationships were similar as previously discussed even though some attributes changed clusters or relationships, mainly the relationship between bloody/ serumy and metallic with beef flavor identity. The biplot to understand if consumer segments affected how consumers rated beef in the two studies is presented in Figure 4b. Two major clusters are apparent. The upper cluster is composed of light millennial and non-millennial consumers and the lower cluster is comprised of heavy millennial and non-millennial consumers. Plots in Figure 4a and Figure 4b are on different scales. The scale used in Figure 4b is larger (-5 to 4 for the horizontal axis and -3 to 3 for the vertical axis) than the scale for Figure 4a (-1 to 1 for both axes). Data are presented in Figure 4b with wider axis to reduce clustering. When these data were placed on the -1 to 1 scale, the data cluster closely around the center of the biplot and are unreadable. Interpretation of data in both biplots should be based on a -1 to 1 scale for both axis. If data in Figure 4b were presented on a -1 to 1 scale, consumer segments would not be apparent. While light beef eaters would tend to be in the upper part of the cluster and heavy beef eater would be in the lower part of the cluster, differences would be very minor and these data would indicate that light and heavy beef eaters evaluated beef flavor similarly. The data was presented on an expanded scale to illustrate that within light and heavy beef eater groups, there was not an effect of millennial versus non-millennial. Within the light and heavy beef eater clusters, millennial and non-millennials were equally distributed indicating that age did not impact consumer ratings of beef flavor and overall liking. These data were combined from two studies that used different consumers, and the data therefore have study and light versus heavy beef eaters confounded. The analysis utilized partial least squares regression to remove auto correlations between independent and dependent variables; however, due to confounding the slight difference in light versus heavy beef eaters may be partially explained by study. To validate these findings, a study where light and heavy beef eaters are recruited and test beef in the same session is needed. CONCLUSIONS Beef flavor is a strong component of overall consumer liking and fat-like is the beef flavor descriptor that consistently is most highly related to overall consumer liking for light and heavy beef eaters. Beef flavor has a stronger relationship to overall consumer liking than juiciness and tenderness; however, juiciness and tenderness are major contributors to consumer overall liking. Beef flavor descriptive sensory attributes from the Beef Flavor Lexicon are related to consumer overall liking, with some attributes positively related to consumer liking (fat-like, followed by brown/roasted, beef identity, bloody/serumy, metallic, umami, sweet, and salty) and other attributes negatively related to consumer liking (cardboardy and liver-like). Light and heavy, millennial and non-millennials beef eaters tend to rate beef liking and beef flavor similarly; however, a study that examines light and heavy, millennial and non-millennials beef eaters simultaneously using the same beef samples is need to validate these findings. ACKNOWLEDGEMENT These studies were supported by funds from the Beef Checkoff. REFERENCES Adhikari K, Chambers E IV, Miller RK, Vázquez-Araújo L, Bhumiratana N and Philip C. 2011. Development of a lexicon for beef flavor in intact muscle. J Sensory Stud 26: 413-420. AMSA. 2014. Research Guidelines for Cookery, Sensory Evaluation, and Instrumental Tenderness Measurements of Meat. American Meat Science Association. www.meatscience.org/sensory. Glascock, RA. 2014. Beef Flavor Attributes and Consumer Perception. M.S. Thesis, Texas A&M University. Miller RK and Kerth CR. 2012. Identification of compounds responsible for positive beef flavor. Final Report. National Cattlemens Beef Association, Centennial, CO. Miller RK and Kerth CR. 2013. Beef flavor attributes and consumer perception. Final Report. National Cattlemens Beef Association, Centennial, CO. Miller RK, Luckemeyer T and Kerth CR. 2014. Beef flavor attributes and consumer perception II. Final Report. National Cattlemens Beef Association, Centennial, CO. 26 68 th Annual Reciprocal Meat Conference