Jack Brown, Jim B. Davis, Angela P. Brown, Donna

A. Erickson & Lindy Seip


Dept. of Plant, Soil and Entomological Science, University of Idaho,

Moscow, ID 83844-2339, USA.  Email:




US canola production (Brassica napus L. or B. rapa L.) is affected by severe insect damage, poor competitiveness with weeds, heat and drought stress, and pod shatter.  Yellow mustard (Sinapis alba L.) offers the potential to overcome these agronomic deficiencies.  However, available yellow mustard cultivars are for condiment use, and do not have oil quality characteristics necessary for an oilseed crop. Also, yellow mustard cultivars have high glucosinolate levels in the seed meal, which limits the value of the meal as a feed. Breeding lines of yellow mustard have been developed at the University of Idaho that have either canola-quality oil (less than 2% erucic acid) or industrial-quality oil (greater than 50% erucic acid), and low glucosinolate content (less than 30 mmol/g of defatted seed meal).  Eight yellow mustard parents with low, intermediate, or high erucic acid content, and either low or high glucosinolate content were hybridized in a full diallel design. F1, F2 and F3 seed were tested for total glucosinolate content and fatty acid composition to investigate inheritance. A simple additive/dominance model was adequate for explaining genetic variation in oleic and erucic acid content.  Additive was more important than dominant gene action for oleic and erucic acid, and heritability was very high for oleic and erucic acid content.  High erucic/low oleic acid content was dominant to low erucic/high oleic.  Low glucosinolate content in seed was found to be maternally controlled by a single recessive gene.  Glucosinolate content was highly heritable at F2 (h2n = 0.88) and F3 (h2n = 0.84) generations, and it was not correlated to any fatty acid type at F3 but was negatively correlated with erucic acid content in F2 families. This study shows that combining desirable oil characteristics with low glucosinolate content is possible in yellow mustard.


KEY WORDS: Genetics, oil quality, glucosinolates, inheritance, plant breeding.




Canola oil continues to increase in popularity among U.S. consumers because it fills a health market niche, having the lowest level of saturated fat of all edible oils and the second-highest level of mono-unsaturated fat (PGAL, 1987).  In 1995, 409,500 metric tonnes of canola seed and 425,466 metric tonnes of canola oil were imported to the U.S.A., mainly from Canada (USDA, 1995).  By the end of this decade, the estimated demand for canola oil in the U.S.A. will exceed one million metric tonnes of oil (Powell, 1994).  Demand for industrial rapeseed oil products (oil high in erucic acid content) also has risen over the past 5 yr (Glaser, 1996).  Although not as dramatic as canola, importation of rapeseed oil has increased in recent years, with approximately 14,000 metric tonnes of oil being imported into the U.S.A. annually (Glaser, 1996).


In contrast to demand and use, canola and industrial rapeseed production in the U.S.A. has been very low.  In 1993 the total US canola/rapeseed acreage was less than 185,000 ha (Powell, 1994).  The gap between production and demand provides an enormous incentive to increase canola and industrial rapeseed production in the U.S.A.


Agronomic deficiencies of canola/rapeseed species (Brassica napus L. and B. rapa L.) under U.S. agricultural conditions include: susceptibility to heat and drought stress; damage by insects such as flea beetles (Phyllotreta cruciferae Goeze), diamondback moth (Plutella xylostella), cabbage seed pod weevils (Ceutorhynchus assimilis Payk.) and aphids (Myzus persicae, Brevicoryne brassicae, etc.); and susceptibility to fungal diseases.  If vegetable oil production from Brassica spp. is to increase in the U.S.A., better-adapted cultivars need to be developed.


Several years ago, work began at the University of Idaho to identify potential crops that can be included in rotation with small grain cereals that predominate in the dry-land regions of the Pacific Northwest.  Yellow mustard (Sinapis alba L.) showed excellent potential based on small plot trials, with seed yields exceeding 3,500 kg ha-1 (Gareau et al., 1990), heat and drought tolerance (Brown et al., 1995a), and rotational benefits as good or better than other Brassica oilseed crops (Guy, 1994).  Yellow mustard crops produce over 6,000 kg ha-1 of crop residue; enough to maintain adequate ground cover residue for the subsequent crop and helps prevent soil erosion (Gareau and Guy, 1995).  In addition, mustard residue easily breaks apart and is spread well by combine harvesters (Brown, et al., 1995a).


Research in Canada suggested that yellow mustard is resistant to the flea beetle (Lamb, 1980; Bodnaryk and Lamb, 1991), a major early season pest of Brassica spp. crops.  Studies at the University of Idaho confirmed this resistance (Brown et al., 1999a).  Furthermore, yellow mustard is highly tolerant to late season pest damage caused by diamondback moth larvae and aphids and is completely resistant to the cabbage seedpod weevil (Brown et al., 1999b), an important pest of Brassica oilseed crops.  In commercial production, yellow mustard produces high seed yields without insecticides (Esser, 1998), and in small plot trials it out-yielded either B. napus or B. rapa which received several insecticide applications (Brown et al., 1995b).


Effective weed control is a major problem in canola/rapeseed production (Chandler et al., 1984, Davis et al., 1998).  Preliminary studies show that yellow mustard is highly competitive with both broad-leaf and grassy weeds and can been grown without herbicide application (Esser et al., 1996).


It has been suggested that oilseed cultivars can be developed from yellow mustard (Rafiullah, 1994).  However, intermediate erucic acid content (approximately 23-35%) makes the oil of yellow mustard unsuitable for either edible (less than 2% erucic acid content) or industrial uses (greater than 45% erucic acid content).  Also, until recently, little variation existed in the world yellow mustard germplasm for oil fatty acid profile (Liu, 1987). 


Although several studies have reported inheritance of fatty acid composition in other oilseed Brassica species, none have examined inheritance of oil quality in yellow mustard.  Erucic and eicosenoic acid contents in rapeseed (B. napus and B. rapa) are controlled by the genotype of the developing embryo and not by the sporophyte (Dorrel and Downey, 1964; Harvey and Downey, 1964; Stefansson and Hougen, 1964; Kondra and Stefansson, 1965).  In addition, erucic acid content is controlled in these species by a few major genes, and can be treated as a qualitative characteristic in genetic research and breeding (Liu, 1984). 


Seed meal is a major by-product of Brassica oilseed production and can provide excellent livestock feed (Josefsson, 1972).  Protein content of whole seed is about 22-24% (Larsen and Sorensen, 1985) and it has a favorable amino acid composition (Ohlson and Anjon, 1979; Sarwar et al., 1984; Larsen et al., 1985).  Available yellow mustard germplasm has been selected for high glucosinolate content in the seed, which gives mustard its spicy flavor and value.  However, when glucosinolate compounds are enzymatically de­graded, toxic by-products (isothiocyanates, thiocyanates and sulfur compounds) cause metabolic disturbances when fed to non-ruminant livestock (Kondra and Stefansson, 1970).  High glucosinolate levels therefore need to be reduced in oilseed yellow mustard cultivars to avoid detri­men­tal effects on livestock consuming the meal, thus adding value to the oilseed yellow mustard crop. 


To date, very few studies have examined inheritance of glucosinolate content in yellow mustard.  Glucosinolate content in Brassica juncea and B. napus is maternally controlled (Kondra and Stefansson, 1970).  Raney et al. (1995) reported that from 1000 F2 plants from a cross between `Saber' (a cultivar with high glucosinolate content in seed meal) and a low glucosinolate line `92-6669', 13 plants with a total glucosinolate content of less than 10mm/gram were selected.


Since 1992 researchers at the University of Idaho, using a combination of mutagenesis and backcrossing, have developed yellow mustard breeding lines with less than 3% erucic acid content, and others with greater than 54% erucic acid content.  Using genotypes obtained from Poland (Krzymanski et al., 1991), the University of Idaho breeding group also developed breeding lines with very low total glucosinolate content (Tang et al., (1995); Brown et al., 1999c).  This provides the basic germplasm materials to examine the inheritance of fatty acids in yellow mustard.


The objectives of this study were to investigate the inheritance of fatty acid composition and glucosinolate content in yellow mustard, and determine the feasibility of developing oilseed cultivars of yellow mustard.




Eight yellow mustard parents from existing condiment cultivars and oilseed breeding lines developed at the University of Idaho were selected for use in this study based on their fatty acid profile and glucosinolate content (Table 1). 


Table 1.  Fatty acid profile and glucosinolate composition of eight parents used in diallel crossing design.                                                                                                                                                                   


                                    Fatty acid composition                         Glucosinolate composition§

Parent          16:0    18:0    18:1     18:2     18:3       20:1       22:1       Prog       Sina     T.Gluc    Ttape

                ------------------------- % ----------------------------   ------ mmol/gram ------  - 0 to 5 -

UI.3553        2.7      1.9    66.2a‡    10.6    11.6a       4.1         2.9e          7.5cd   323.7ab   331.1ab        5.0a

UI.3568        3.9      2.2    63.4a      12.0    11.2ab     4.1         3.2e          6.9d    333.4a     340.2a          5.0a

Mustang        2.4      0.8    18.4e     11.1      9.3c        4.6       53.4a          9.6c     286.2bc   295.7bc        5.0a

UI.Israel       2.1      0.8    18.2e      11.9      9.3c        3.0       54.7a          2.4e     309.8ab   312.2ab        5.0a

UI.LG.3        2.2      0.9    32.3cd      9.6      8.7d       9.1       39.9bc      27.7a          0.7d       28.4d          0.5b

UI.LG.2        2.6      1.3    39.7b     10.4    10.7b     10.8       25.9d       12.4b          0.2d       12.6d          0.0b

Kirby            2.6      1.0    27.2d       9.6    11.0ab      9.1       39.5b          2.4e     228.6c     231.0c          4.7a

Gisilba          2.6      1.2    31.6c        9.5      8.2d     11.8       35.1c          7.7cd   249.3bc      257.0bc          4.5a


Significance   ns       ns        *           ns        *           ns             *          *            *            *            *



* = P<0.05; ns = not significant.  16:0 = palmitic acid, 18:0 = stearic acid, 18:1 = oleic acid, 18:2 = linoleic acid, 18:3 = linolenic acid, 20:1 = eicosenoic acid, 22:1 = erucic acid. § Prog = progoitrin = 2-hydroxy-3-butenyl glucosinolate, Sina = sinalbin = p-hydroxybenzyl glucosinolate; T. Gluc = total glucosinolate; Ttape = Tes-tapeÒ reading. 

Means within columns with different superscript letters are significantly different (P<0.05) according to Tukey's multiple range test.



Erucic acid (22:1) content of the eight parents ranged from 2.9 to 54.7%, while oleic acid (18:1) content varied from 18.2% to 66.2%.  Other fatty acids, such as eicosenoic acid (20:1), linolenic acid (18:3) and linoleic acid (18:2) showed less variation, and were not significantly different.  The eight parents fall into three types based on fatty acid profile: (1) parents ('UI.3553' and `UI.3568', developed at the University of Idaho from germplasm originating in Canada, with low (2.9 to 3.2%) erucic acid content; (2) two parents (`Mustang' and `UI.Israel') with high (more than 53%) erucic acid content.  Mustang was developed in Europe as an oilseed cultivar and has relatively high oil content while UI.Israel was developed from a wild accession from Israel; and (3) four parents with intermediate (25.9 to 39.9%) levels of erucic acid content, `UI.LG.3', `UI.LG.2', `Kirby' and `Gisilba'.  Genotypes UI.LG.2 and UI.LG.3 were developed at the University of Idaho to have low total glucosinolate content, while condiment cultivar Kirby was developed in England and condiment cultivar Gisilba was developed in Germany. 


The eight parents were crossed in all possible hybrid combinations (including reciprocals) in an 8 x 8 full diallel design, Method I, as described by Griffing (1956).  Cross pollinations were made by hand in the greenhouse at the University of Idaho over the winter of 1994.


F1 seeds of the 56 crosses and 8 parents were grown in the greenhouse in the spring of 1995, in a randomized complete block design with 4 replicates and plot size of 2 plants per replicate.  In each replicate, both plants of each cross or parent were tested for fatty acid content separately.  F2 plants were grown in the same design as the F1 trial in the fall of 1995.   


Growing conditions of F1 and F2 plants were similar.  Individual plants were grown in 20-cm plastic pots.  Potting soil was Sunshine Mix (Fisons West Corp., Vancouver, BC, Canada), containing 75% peat moss, 15% perlite and 10% vermiculite.  Plants were grown under a 16 h light/8 h dark cycle.  High intensity discharge multivapor halogen lamps (1000 watt) were used for supplemental illumination.  Temperature was maintained at 22+3o C during the day and 16+3o C at night.


Bud-selfed seeds were produced by hand pollination of all parents, F1, and F2 plants (i.e., to produce F2 and F3 seeds, respectively).  Yellow mustard is primarily wind pollinated and has a sporphytic self-incompatibility system that is only functional after flowers open (Hemingway, 1976), and bud-selfing was necessary to avoid unwanted cross pollination.  During the flowering stage 10 buds which were close to flower opening on secondary racemes of each plant were selected for bud-selfing.  After opening with alcohol-cleaned tweezers, buds were hand-pollinated using pollen from opened flowers of the same plant.  Bud-selfed seeds were used for fatty acid analysis and to plant subsequent generations. 


Fatty acid profile of the bud-selfed eight parents was determined using a bulk seed procedure, while seeds from bud-selfed hybrid progeny were tested using a single seed procedure, where four single seeds were analyzed from each plant, and the data averaged before analyses.   Bulk seed fatty acid composition was determined on a 250-mg sub-sample from each parent.  Fatty acid composition was determined from four separate F1 seeds from each progeny/parent according to the method described by Hammond (1009), Raney, et al. (1987) and Christie (1992) with a Hewlett Packard 5890 Series II Gas Chromato­graph.


The primary seed storage fatty acids in oilseed Brassica species are palmitic acid (16:0), stearic acid (18:0), oleic acid (18:1), linoleic acid (18:2), linolenic acid (18:3), eicosenoic acid (20:1), and erucic acid (22:1).  Only oleic acid, linolenic acid and erucic acid showed significant differences among progeny.  Since these fats are the most important to determine the end use of vegetable oil, only these fatty acids are discussed in more detail.


Qualitative determination of glucosinolate content was assessed by a glucose Tes-tape® procedure (Smith and Donald, 1988).  Five seeds from each plant (i.e., a total of eight plants from each cross combination) were placed on a piece of tape that had been secured, adhesive side up, on a table top.  The seed was then crushed and a drop of distilled water (± 50 ml) was added to each sample.  After three minutes, a 2-cm section of glucose sensitive test paper (M-73, Eli Lilly and Co., India­napo­lis) was placed over each crushed seed sample.  A two minute period was allowed for chemical reaction and associ­ated color change.  Visual assessment scores on a 0 to 5 scale (0 = no or very low glucosin­ola­tes and 5 = very high glucosinola­tes) were used to quantify glucosinolates levels.  Quantitative evaluation of total glucosinolate content and glucosinolate profile were conducted using gas chromatography (Varian 3700 Gas Chromato­graph) and the trimethy­lsilyl (TMS) method (Daun and MacGregor, 1991).


Data were analyzed using the computer packages CHIP (Brown, 1984) and POTSTAT (Brown and Caligari, 1988).  General combining ability (GCA) and specific combining ability (SCA) analyses were carried out based on the method outlined by Mather and Jinks (1977).  In these analyses, parents were treated as random effects because the intent was to make inference on situations arising when any high erucic acid yellow mustard was crossed to one with low erucic acid content.  Error estimates in analyses were calculated from 2 (or 4) seeds within each plant, 2 plants within each replicate, and 4 replicates, assuming a nested design, where seeds and plants were nested within replicate blocks.  Analyses of relationships between array variances (Vi) and array/non-recurrent parent covariances (Wi) were carried out according to the methods of Jinks and Hayman (1953), using an algorithm described by Morley-Jones (1965).  Additive genetic variance (A) and dominant genetic variance (D) were estimated using the procedures of Mather and Jinks (1977) where: A = 4/7(Vp+Wi+V); D = 4Vi-A;  Vp = the variance between parental performance;  Wi = the average array non-recurrent parent covariance;  Vi = the average array variance;  and V = the variance of array means;  From these parameters, narrow-sense heritability (h2n) was estimated as:


                                                 h2n = 1/2A/(1/2A+1/4D+s2e),


where s2e is the error variance obtained from the analysis of variance.  Estimates of D in F3 progeny were adjusted to account for greater homozygosity.




Oil quality - fatty acid profile


Relationships among fatty acid profiles were examined by simple correlation.  Erucic acid content was negatively correlated with all other fatty acids examined (Table 2).  Highest negative correlation was between erucic acid and oleic acid.  Oleic acid content was correlated to high palmitic, stearic, linolenic and eicosenoic acid.


Reciprocal effects were significant for linolenic acid only in F1 seed (Table 3).  No reciprocal effects were detected in the inheritance of either oleic or erucic acid.  Therefore, the order that crosses were made did not affect the resulting fatty acid profile for oleic and erucic acid.  However, the variance component due to maternal general combining ability (GCA) tended to be greater than paternal GCA for oleic and erucic acid in each generation.  Highly significant general combining ability (GCA) of female and male parents were observed for oleic acid and erucic acid content in F1, F2 and F3 generations.  Specific combining ability (SCA) was also highly significant for erucic and oleic acid content, although considerably greater variation among progenies was explainable by GCA compared to SCA.  In F1 progeny, maternal parent GCA was significant (P<0.05) and paternal parent GCA highly significant (P<0.001) for linolenic acid content.  In the two later generations, GCA was not significant for linolenic acid content.  Overall greater additive effects (GCA) were found for the inheritance of erucic and oleic acid than for linolenic acid.


Table 2.  Correlation coefficients between seven fatty acids from progeny (averaged for F1, F2 and F3 families) from a 8 x 8 diallel crossing design.


Stearic                  0.56 *

Oleic                     0.45 *           0.67 *

Linoleic                 0.58 *           0.19              0.14

Linolenic               0.28 *           0.28 *           0.27 *           0.06

Eicosenoic             0.00              0.34 *           0.21 *           0.05            -0.05

Erucic                  -0.60 *          -0.72 *          -0.84 *          -0.37 *          -0.36 *            -0.32 *


                         Plamitic         Stearic           Oleic           Linoleic        Linolenic        Eicosenoic


* = P<0.05; all other correlation coefficients are not significant.


Table 3.  Mean squares from the analyses of variance of oleic acid, linolenic acid and erucic acid content in F1, F2 and F3 progeny from an 8 x 8 diallel crossing design.


          Source                          d.f.           Oleic acid           Linolenic acid         Erucic acid                    


F1       Replicate blocks              3                 39.8                     2.6                    92.6

           GCA Female                   7             3056.4*                  24.9*               4523.6*

           GCA Males                    7             1852.8*                  57.1*               4160.8*

           SCA (F x M)                49               169.6*                  11.5*                 180.3*

           Reciprocal                     28                 58.8 ns                 4.6*                   53.1 ns

           Error                          445                 38.52                   2.10                  55.22


F2       Reps                               3               240.5*                  10.8 ns              265.5 ns

           GCA Female                   7             1875.5*                  22.6 ns            3114.1*

           GCA Male                      7             1307.4*                  28.6 ns            2517.2*

           SCA (F x M)                49               268.0*                  15.9*                 346.1*

           Reciprocal                     28               115.1 ns                 8.3 ns              150.0 ns

           Error                          445                 86.21                   7.24                115.06


F3       Replicate blocks              3               137.4 ns               13.2 ns              193.5 ns

           GCA Female                   7             2920.6*                  19.4 ns            3976.2*

           GCA Male                      7             2076.4*                  19.1 ns            2580.4*

           SCA (F x M)                49               310.0*                  14.8*                 284.0*

           Reciprocal                     28               147.7 ns                 6.8 ns              195.4 ns

           Error                          445               132.76                   5.86                153.86


* = P<0.05; ns = not significant.



Significant differences were observed among array means (i.e. average performance of all progeny with a common parent) for oleic, linolenic and erucic acid content (Table 4).  Hybrid progeny with low erucic acid content were derived from crosses where one, or both, parents had low erucic acid content.  Parental performance and average progeny performance were highly correlated for oleic and erucic acid (Table 5).  Correlation coefficients between parents and progeny for linolenic acid were positive, but not significant at the 5% level.


Table 4.  Array means of oleic acid, linolenic acid and erucic acid.


Parent                    Oleic acid                 Linolenic acid                 Erucic acid

                             -------------------------- % ----------------------------


UI.3553                 40.43 a†                        11.33 ab                        18.58 c

UI.3568                 41.68 a                         12.33 a                         18.52 c

Mustang                 22.92 c                         10.82 bc                        36.22 ab

UI.Israel                 20.57 c                         10.93 b                         44.82 a

UI.LG.3                 33.22 ab                        11.27 ab                        27.03 bc

UI.LG.2                 34.45 ab                        12.13 a                         24.67 bc

Kirby                      28.33 bc                        10.45 c                         33.50 ab

Gisilba                    31.33 ab                        10.98 b                         27.55 bc


Means within columns with different superscript letters are significantly different (P<0.05) according to Tukey's multiple range test.


Table 5.  Correlation coefficients obtained by correlation of average progeny performance (array means) and parental values of oleic acid, linolenic acid and erucic acid.


Generation        Oleic acid         Linolenic acid               Erucic acid


F1                        0.95 *                    0.55 ns                    0.96 *

F2                        0.93 *                    0.32 ns                    0.94 *

F3                        0.95 *                    0.29 ns                    0.95 *


* = P<0.05; ns = not significant.



The relationship between Vi’s and Wi’s from each parent was used to determine the mode of inheritance (Mather and Jinks, 1977).  Regression of Vi on Wi produced regression slopes that were not significantly different from one for oleic acid and erucic acid (Table 6).  However, the regression slope for linolenic acid was significantly lower than one at F2 and F3.  Therefore, inheritance of oleic and erucic acid can be explained by a simple additive/dominance model while the inheritance of linolenic acid is influenced by other factors (i.e., linkage or epistasis).


The intercept on the Wi axis of the regression line was significantly greater than zero for oleic and erucic acid indicating the additive genetic variance is greater than dominance variation (Table 6).  The scatter diagram of Vi onto Wi showed that parents with highest erucic acid content were located nearest to the origin in the diagram (i.e. parents which produce lower Vi and Wi values) while parents with lowest erucic acid content were located furthest from the origin (i.e. with largest Vi and Wi values).  Therefore, high erucic acid content is dominant over low erucic acid content.  Similar results were obtained for oleic acid content where high oleic acid content was recessive to low oleic acid content.


Additive (A) and dominant (D) genetic variances were estimated for oleic acid, linolenic acid and erucic acid content (Table 7).  As expected from the analyses in Table 3, and the relationship between Vi and Wi in Table 6, additive genetic variance for oleic acid and erucic acid content was larger than dominant genetic variance.  Dominant genetic variance was, however, a greater component in the inheritance of linolenic acid.  Averaged for the three generations examined, narrow-sense heritability of oleic acid, linolenic acid and erucic acid content was 0.69, 0.37 and 0.69, respectively.  Therefore, greatest selection response will be likely for increased erucic acid or oleic acid content, and least selection response will be obtained by selecting for increased or decreased linolenic acid content.


Table 6.  Regression slope and intercept obtained by regression of array variance (Vi) onto array/non-recurrent parent covariance (Wi).


Generation                     Oleic acid                   Linolenic acid               Erucic acid


F1         intercept            201.54 *†                         4.96 ns                 376.31 *

            slope                     0.82 ns                       0.33 *                       0.77 ns

F2         intercept              43.49 *                         -4.36 *                     33.75 *

            slope                     0.82 ns                         0.77 ns                     1.00 ns

F3         intercept              88.89 *                         -3.89 *                     35.72 *

            slope                   0.68 ns                          0.69 ns                     0.97 ns


intercept comparison to zero.  slope comparison to one.

* = P<0.05; ns = not significant.



Table 7.  Additive genetic variance (A), dominant genetic variance (D), and narrow-sense heritability (h2n) from 8 x 8 diallel crossing design.


Generation         Statistic                Oleic acid                  Linolenic acid            Erucic acid


       F1                   A                    958.63                          13.43                       1514.21

                             D                    356.20                          27.16                         556.65

                             h2n                       0.79                            0.43                             0.80

       F2                   A                    841.37                           6.48                       1325.62

                             D                    732.52                          19.52                       1286.57

                             h2n                       0.51                            0.39                             0.57

       F3                   A                    515.31                            5.30                        659.28

                             D                    434.95                          18.69                         536.27

                            h2n                       0.67                            0.29                             0.71



The 56 hybrid combinations examined in the diallel crossing design were grouped into six different classes according to whether each parent had low erucic acid content (L), intermediate erucic acid content (I) or high erucic acid content (H) and fatty acid composition of 400 progeny were examined from each cross type.  Variation within the L x I, L x H, I x I and I x H was continuous for erucic and oleic acid content suggesting polygenic inheritance rather than inheritance by one or two genes.  Average erucic acid content ranged from 8.6% in L x L crosses to a high of 52.4% in H x H crosses (Table 8).  Erucic acid content of L x I crosses were lower than L x H crosses while I x I crosses were lower in erucic acid content than I x H crosses.  Greatest variation within progeny from different cross types was observed in hybridization between low erucic acid x high erucic acid.  Crosses between low erucic acid parents and intermediate erucic acid parents (i.e., L x L or L x I) were capable of producing progeny with a fatty acid profile identical to canola.  It was also noted that only L x L crosses produced progeny with extremely high (greater than 70%) oleic acid content.  Similarly, extremely high erucic acid content (as would be required for industrial oil) was only obtained in I x H and H x H crosses.


Table 8.  Population statistics of oleic acid and erucic acid of F2 seed from six types of crosses in an 8 x 8 diallel crossing design.

                                                                            Cross type

                                     L x L         L x I         L x H          I x I          I x H         H x H            

Erucic acid

Mean                                8.6          21.6          33.2            27.1         41.8          52.4

s.e. mean                         1.45          1.15          2.58            2.19         1.11          1.68

Median                              5.2          19.1          36.2            24.9         42.4          56.8

Minimum                           0.2            0.1            0.7            12.9           8.8            4.0

Maximum                        43.8          56.5          55.3            52.6         62.2          62.5

Frequency     <1%§         0.18          0.01          0.00            0.00        0.00           0.00

                     <10%         0.65          0.21          0.13            0.00        0.01           0.04

                     >50%         0.00          0.05          0.10            0.10        0.40           0.60

                     >60%         0.00          0.00          0.00            0.00        0.05           0.27

Oleic acid

Mean                              52.3          38.7          31.7            32.9         24.0          18.4

s.e. mean                           1.6          1.05          2.34            1.89         0.81          1.38

Median                            52.4          40.0          28.8            32.3         23.8          15.4

Minimum                         20.0          11.6          13.3            16.4          8.4            8.2

Maximum                        71.3          66.3          60.6            47.5         54.9          59.1

Frequency    >50%         0.60          0.16          0.10            0.04         0.02          0.00

                     >60%         0.24          0.05          0.01            0.00         0.00          0.00

                     >70%         0.06          0.00          0.00            0.00         0.00          0.00


High erucic acid parent (H); Low erucic acid parent (L); Intermediate erucic acid parent (I). 

§ Proportion of 400 progeny less than 1% erucic acid content.  proportion of 400 progeny greater than 50% oleic acid content. 



Glucosinolate content


Glucosinolate content was under maternal control, where glucosinolate content of F1 seed was determined by the genotype of the female parent (Table 9).  The first segregation for glucosinolate content was observed in the F2 seed generation (equivalent to the F1 hybrid genotype) (Table 10).  In both generations, low glucosinolate content (L) x L  always resulted in low glucosinolate progeny.  Similarly with high (H) x H progeny always resulted in high content in both generations.  L x H, or H x L crosses showed little segregation for glucosinolate content at F2 where the majority of progeny produced higher glucosinolate content.  Greater segregation was noted in the F3 generation although progeny means tended towards higher glucosinolate content.  This suggests that high glucosinolate content is dominant to low content.


Mean squares from general combining ability (GCA) of maternal and paternal parents were not significantly different at F2 (Table 11).  In addition, reciprocal effects were not significant, indicating that L x H progeny were equal to H x L progeny.  GCA of female and male parents was highly significant (P<0.001) at F2, collectively accounting for over 50% of the total variation between progeny means.  However, specific combining ability was also highly significant (P<0.001) in this generation.  At F3 GCA of female and male parents was again highly significant (P<0.001) while SCA was not significantly different from the replicate error.  Overall, the analyses of variance indicated that the inheritance of glucosinolate content is highly additive.



Table 9.  Glucosinolate Tes-tape scores of F1 seed from reciprocal crosses.


L x H crosses                    Score                H x L crosses                      Score


UI.LG.2 x UI.3553                1                    UI.3553 x UI LG.2                   5

UI.LG.2 x UI.3568                1                    UI.3568 x UI.LG.2                   5

UI.LG.2 x Mustang                0                    Mustang x UI.LG.2                   5

UI.LG.2 x UI.Israel                0                    UI.Israel x UI.LG.2                   5

UI.LG.2 x Kirby                     0                    Kirby x UI.LG.2                       5

UI.LG.2 x Gisilba                   0                    Gisilba x UI.LG.2                      5

Mean                                  0.33                                                               5.00


L x H crosses = low glucosinolate parent as female and high glucosinolate parent as male; H x L crosses = high glucosinolate parent as female and low glucosinolate parent as male.

0 = very low glucosinolate; 5 = very high glucosinolate score, according to Tes-tape rating.


Table 10Glucosinolate Tes-tape scores of F2 and F3 seed from 64 hybrid combinations in a diallel design.


F2 seeds       UI.LG.3     UI.LG.2       UI.3553     UI.3568     Mustang  UI.Isr.         Kirby      Gisilba       Mean

UI.LG.3          0.3        0.1           5.0         4.8         2.0         5.0          4.4       4.5       3.3c‡

UI.LG.2          0.6         0.0           3.6         3.1         4.4         4.9          4.1       2.8       2.9c

UI.3553          3.1         4.0           5.0         4.6         4.8         4.5          4.9       4.6       4.4ab

UI.3568          4.3         4.1           5.0         4.9         4.8         4.8          3.8       4.4       4.5ab

Mustang          4.8         4.5           4.0         3.3         4.1         5.0          4.9       4.3       4.3ab

UI.Israel          4.9         5.0           5.0         4.3         4.5         4.9          5.0       4.8       4.8a

Kirby              3.8         3.0           4.9         4.0         3.6         4.5          4.9       4.3       4.1b

Gisilba             3.8         3.3           5.0         5.0         4.8         5.0          5.0       3.8       4.4ab

Mean             3.2c        3.0c          4.7ab      4.2ab      4.1b        4.8a         4.6ab    4.2ab     4.1


F3 seeds       UI.LG.3     UI.LG.2       UI.3553     UI.3568     Mustang  UI.Isr.         Kirby      Gisilba       Mean

UI.LG.3          0.3         0.1           5.0         1.3         2.5         2.0          2.5       3.8       2.2b

UI.LG.2          0.3         0.5           3.5         2.8         2.3         3.8          3.8       4.5       2.7b

UI.3553          1.3         4.5           5.0         4.5         4.0         4.8          4.5       4.3       4.1a

UI.3568          2.8         5.0           4.5         4.8         4.8         5.0          4.8       4.8       4.5a

Mustang          3.8         3.8           4.8         5.0         5.0         4.8          4.8       4.5       4.5a

UI.Israel          2.8         3.0           4.8         5.0         4.0         5.0          4.5       4.8       4.2a

Kirby              5.0         3.8           5.0         4.8         4.8         5.0          5.0       5.0       4.8a

Gisilba             3.8         2.8           5.0         4.5         5.0         5.0          4.3       4.0       4.3a

Mean             2.5b        2.9b         4.7a        4.1a        4.0a        4.4a         4.3a      4.4a      3.9


mean of four replicates each of 2 plants (i.e. a total of eight plants).  Array means with different superscript letters are significant (P<0.05) according to Tukey's multiple range test.



The relationship between Vi and Wi was examined through simple regression.  In both F2 and F3 generations the regression slope was not significantly different from one, indicating that a simple additive/dominance inheritance model would be adequate to explain the observed variation.  The regression line in the F2 progeny is highly biased due to the Vi and Wi values of the two very low glucosinolate parents (UI.LG.2 and UI.LG.3), while the high glucosinolate parents were clustered around the origin.  The location of the very low glucosinolate parents (i.e. farthest from the origin) confirms that high glucosinolate content is dominant to very low content.  The regression intercepts on the Wi axis were both positive indicating that additive genetic variance was larger than dominant variation.  Indeed estimates of A and D (Table 12) show that additive variance effects were more than 11 times greater than dominant variance effects at F2 and over 19 times larger at F3.  Narrow-sense heritability estimates were very high (Table 12) indicating that response to selection for low glucosinolate content should be good.


Table 11.  Mean squares from analyses of variance of glucosinolate scores.


                                                                          F2                                           F3


Source                                    d.f.         M.Sq.          % Total              M.Sq.         % Total

Replicate blocks                       3              0.87 ns           1.6                  3.31 ns             1.2

GCA     Female parents          7            14.65***          20.6                19.68***            16.6

               Male parents             7            13.59***          19.2                28.97***            24.4

SCA§      Females x Males      49              3.35***          33.0                  2.31 ns           13.6

Reciprocal effect                    28              0.82 ns            -                     2.26 ns              -

Replicate error                     189              0.70             26.6                  1.19                44.2


percentage of total sum of squares.  General combining ability;  § Specific combining ability. 

*** = P<0.001; ns = not significant.


Table 12.  Additive genetic variance (A), dominant genetic variance (D), and narrow-sense heritability of glucosinolate scores.


Statistic                                        F2                                   F3


     A                                        14.60                              14.56

     D                                          1.30                                0.75

     h2n                                      0.88                                0.84

     h2n                                      -                                     0.77


estimated from A, D and error variance.   estimated by regression of F2 onto F3 performance.



The 56 crosses examined in the 8 x 8 diallel design were divided into four types (L x L, L x H, H x L and H x H) and the performance of each type compared (Table 13).  In both generations, the variation within progeny was greatest in L x H or H x L crosses compared to L x L or H x H.  Variation between L x L and H x H was similar between the two generations although greater within progeny variation existed in F3 compared to F2.  At F2 almost all progeny from L x L crosses had very low glucosinolate content (a score of less than 1).  Almost all progeny from the other three cross types showed consistently high glucosinolate content.  At F3 the L x L and H x H cross progeny performances were similar although more than one quarter of the progeny between L x H or H x L crosses showed very low glucosinolate content.  From the 200 L x H progeny and 200 H x L progeny examined (and as no significant reciprocal effects have been noted) a total of 118 phenotypes showed very low glucosinolate content and the remaining 282 had high glucosinolate content.  Comparing the observed ratio of low:high glucosinolate (1:2.4) to the expected ratio (1:3) if a single recessive gene controls low glucosinolate inheritance, resulted in a non-significant c2 value of 3.12.  When the observed ratio of low:high progeny was compared to the ratio expected in a similar two-gene situation (i.e. 1:15), a significant c2 value was found.


Table 13.  Population statistics of glucosinolate score of F2 seed and F3 seed from four types of crosses in an 8 x 8 diallel crossing design.


                                                                            Cross type

                                      L x L               L x H                 H x L               H x H              


Mean                              0.38b                4.02a                  4.03a                4.56a

s.e. mean                        0.006               0.177                 0.168               0.072

Median                             0.0                   4.5                     4.5                   5.0

Minimum                           0.0                   1.2                     1.1                   2.4

Maximum                          1.5                   5.0                     5.0                   5.0

Frequency <1                 94.3                  0.1                     0.0                   0.0


Mean                              0.25c                3.09b                  3.50b                4.68a

s.e. mean                        0.063               0.328                 0.290               0.074

Median                             0.0                   4.0                     4.1                   5.0

Minimum                           0.0                   0.0                     0.0                   2.4

Maximum                          1.1                   5.0                     5.0                   5.0

Frequency <1                 97.6                 27.0                   31.9                 0.00                      


Low glucosinolate content parent (L); High glucosinolate content parent (H).   percentage of 400 progeny less than glucosinolate score of 1.





Oleic acid and erucic acid content in yellow mustard are highly correlated, and a reduction of one type is almost directly related to an increase in the other.  The inheritance of oleic and erucic acid is similar, where GCA is greater than SCA, heritabilities are high, a simple additive/dominant model is adequate to explain the variation, and high erucic acid content is dominant to low erucic acid content.  Low, or high, erucic acid content was not adversely correlated to other agronomic characters, although the longer fatty acid chains of eicosenoic and erucic acid were related to higher oil content.  Plant breeders should be able to develop yellow mustard cultivars with high oleic acid content (greater than 70%) and with low (less than 1%) erucic acid content (i.e., canola quality).   Similarly, high erucic acid cultivars with greater than 60% erucic acid content should be attainable.


A single recessive gene controls the inheritance of very low glucosinolate content in yellow mustard.  Inheritance of higher glucosinolate content appears to be more quantitative and continuous in nature.  No canola-quality seed meal glucosinolate progeny (i.e. less than 30mmol/g of defatted seed meal) were found from crosses that did not include UI.LG.2 or UI.LG.3 as parents.  Inspection of the glucosinolate content of these two parents and their low glucosinolate progeny suggests that a single gene is indeed completely blocking the production of sinalbin glucosinolate although blocking this single syntheses pathway results in increased progoitrin (Table 14).  It is also interesting that these very low glucosinolate yellow mustard lines produced small amounts of benzyl and 2-phenylethyl glucosinolates, which have never been documented to exist in yellow mustard cultivars. Glucosinolate profiles of the very low glucosinolate germplasm will require further examination before a firm conclusion can be made on this observation.


In order to develop oilseed yellow mustard cultivars, breeders must combine low glucosinolate seed meal characters with desirable oil quality (fatty acid profile) and high oil content.  Little relationship exists between fatty acid profile and glucosinolate content, while the relationship between low glucosinolate content and high oil content would be advantageous in breeding for oilseed yellow mustard crops.


Breeding at the University of Idaho has been aimed at oilseed yellow mustard cultivar development.  Yellow mustard breeding lines have been developed with high oil content (35-38%), and with either canola oil quality or industrial oil quality combined with glucosinolate content as low as exists within canola quality cultivars (Table 14).  Initial research has shown that these breeding lines have retained insect and disease resistance, heat and drought tolerance features of traditional condiment yellow mustard, and hence they offer tremendous potential as low input rotation crops.  In addition, release of such cultivars will expand the acreage of canola and industrial rapeseed into regions of higher temperature and lower rainfall or into areas where insecticides of fungicides are not available.


Table 14.  Glucosinolate composition (mmol/gram of defatted seed meal) and fatty acid profile (%) of selected genotypes from the University of Idaho breeding program.


                                 Glucosinolate                                                  Fatty acid profile                         

                  Prog   Sinal   Benz   Phenyl   Total          16:0    18:0    18:1     18:2    18:3    20:1    22:1

              -------------- mmol/gram -------------   --------------------------- % ---------------------------

UI.9.1.8        7.8       0.0    4.9       0.3     13.0            3.0      1.2     18.5       6.7      7.6       5.8     57.2

UI.4.1.1      10.2       0.0    0.9       0.1     11.2            2.9      0.8     10.1     10.6    10.3       3.9     61.5

UI.4.1.7      13.6       0.0    2.8       0.1     16.5            2.6      0.9     10.0       7.5      9.9       5.5     63.6

UI.6.1.8       9.7       0.0    3.3       0.3     13.3            2.8      0.6     10.7     11.1    10.9       2.5     61.4

UI.3.1.3        6.5       0.0    4.1       0.2     10.8            4.8      1.2     63.0     12.7    14.1       2.7       1.4

UI.6.3.1        9.6       0.0    7.8       0.9     18.3            3.9      1.1     71.9       8.2    13.0       1.6       0.3

UI.6.3.2        7.0       0.0    2.2       0.3       9.5            4.6      1.1     67.9     10.7    13.0       2.1       0.6

UI.1.2.8      15.8       0.0    4.4       0.3     20.5            3.2      1.1     69.8     14.6    10.2       1.5       0.6


Gisilba§         7.7   249.3    0.0       0.0   257.0            2.6      1.2     29.6       9.5      8.2     11.8     33.3


Prog = Progoitrin = 2-hydroxy-3-butenyl glucosinolate; Sinal = sinalbin = p-hydroxybenzyl glucosinolate, Benz = benzyl glucosinolate, phenyl = 2-phenylethyl glucosinolate.  16:0 = palmitic acid, 18:0 = stearic acid, 18:1 = oleic acid, 18:2 = linoleic acid, 18:3 = linolenic acid, 20:1 = eicosenoic acid, 22:1 = erucic acid.   § Condiment yellow mustard.





The authors would like to acknowledge thanks to Wendy Lawrence for invaluable assistance in fatty acid determination and data entry.




Bodnaryk, R.P. and R.J. Lamb.  1991.  Mechanisms of resistance to the flea beetle Phyllotreta cruciferae Goeze in mustard seedlings Sinapis alba L.  Can. J. Plant Sci. 71:13-20.

Brown, J.  1984. A new data base computer package for plant breeders.  Euphytica 33: 935-942.

Brown, J. and P.D.S. Caligari.  1988.  POTSTAT. User’s manual.  S.C.R.I. Pentlandfield,  Scotland, UK.

Brown, J., S.O. Guy, and J.P. McCaffrey.  1995a.  Developing production systems for yellow condiment mustard (Sinapis alba L.) as an alternative crop in rotation with small grain cereals in the dryland regions of the Pacific Northwest.  In: Proc. of the 1994 STEEP Conference, Richland, Washington, January, 1994.

Brown, J., J.P. McCaffrey, J.B. Davis, B.L. Harmon, D.E. Erickson and A.P. Brown. 1995b.  The effect of insect damage on productivity of Brassica napus, B. rapa, B. juncea and Sinapis alba.  pp.  In: Proceeding of Oilseed Rape Conference in Cambridge, July, 1995.

Brown, J., J.P. McCaffrey, J.B. Davis, B.L. Harmon, D.E. Erickson and A.P. Brown. 1999a.  Yield reduction in Brassica napus, B. rapa, B. juncea and Sinapis alba caused by late season insect infestation.  J. Agric. Sci. (in press)

Brown, J., J.P. McCaffrey, D.E. Erickson, A.P. Brown, J.B. Davis and B.L. Harmon. 1999b.  Yield reduction in Brassica napus, B. rapa, B. juncea and Sinapis alba caused by infestation by flea beetle (Phyllotreta cruciferae Goeze).  (Submitted to press)

Brown, J., J.H. Tang, J.B. Davis, L. Seip, A.P. Brown & D.E. Erickson. 1999c.  Inheritance of seed meal glucosinolate contant in yellow mustard (Sinapis alba L.).  Crop Sci. (in press).

Chandler, J.M., A.S. Hamill and A.G. Thomas.  1984.  Crop losses due to weeds in Canada and the United States.  Weed Sci. Soc. Am. Champaign, IL 61820.

Christie, W.W.  1992.   Preparation of fatty acid methyl ester.  Inform.  3: 1031-1034.

Daun, J.K., and D.I. McGregor.   1983.  Glucosinolate analysis of rapeseed (canola) - method of the Canadian Grain Commission Grain Research Lab (2nd ed.).  Agriculture Canada, Winnipeg.

Davis, J.B., J.S. Brennan, J. Brown and D.C. Thill.  1999.  The effect of weed seed contamination on the quality of oil and seed meal of canola (Brassica napus L.).  Weed Sci. (in press).

Dorrel, D.C. and R.K. Downey.  1964.  The inheritance of erucic acid content in rapeseed (B. campestris).  Can J. Plant Sci. 44: 499-504.

Esser, A.D. 1998.  Agronomic and economic feasibility of yellow mustard (Sinapis alba L.) as an alternative crop in the dryland region of the Pacific Northwest.  MS thesis,  University of Idaho.

Esser, A.D., J. Brown and J.B Davis.  1996.  The competitiveness of canola (Brassica napus L. ) and yellow mustard (Sinapis alba L.) to contamination by broad-leaf and grassy weeds.  In: Proc. of the Pacific Northwest Canola Meeting, Great Falls, Montana, November, 1996.

Gareau, R.M., D.L. Auld and M.K. Heikkinen.  1990. Evaluation of seven spe­cies of oilseeds as spring planted crops for the Pacific Northwest. p.1-13  In: Prog. Rep., Univ. Idaho, Moscow (277).

Gareau, R.M. and S.O. Guy.  1995. Condiment mustard, dry pea and lentil residue production and overwinter decomposition.  In: Proc. Pacific Northwest Canola Conf., Coeur d'Alene, Idaho, November 5-7, 1995.

Glaser, L.K.  1996. Crambe: an economic assessment feasibility of providing multiple-peril crop insurance.  Federal Crop Insurance Corporation, November 1996. 

Griffing, B.  1956.  Concept of general and specific combining ability in relation to diallel crossing systems.  Australian J. Bio. Sci.  9: 462-493.

Guy, S. 1994.  Canola benefits to rotational wheat crop performance. p.41-44   In: Proceedings of the Pacific Northwest Canola Conference, November 9-11, Spokane, Washington, 1994.

Hammond, E.G.  1991.  Organization of Rapid Analysis of Lipids in Many Individual Plants.  12: 321-330.   In H.F.  Linskens et al. (ed.) Modern Methods of Plant Analysis,  Essential Oils and Waxes.  Springer-Verlag Berlin, Germany.

Harvey, B.L. and R.K. Downey.  1964.  The inheritance of erucic acid content in rapeseed (B. napus).  Can. J. Plant. Sci. 44: 104-111.

Hemingway, J.S.  1976.  Mustards, Brassica ssp. and Sinapis alba (cruciferae). p. 56-59.  In: N.W.Simmonds (ed.) Evolution of Crop Plants.  Longman Inc., NY.

Jinks, J.L. and B.I. Hayman.  1953.  The analysis of diallel crosses.   Maize Genetics Newsletter 27: 48-54.

Josefsson, E.  1972.  Nutritional value and use of rapeseed meal.  p. 354-378.  In L.A. Appelqvist et al. (ed.) Rapeseed: Cultivation, Composition, Processing and Utilization.  Elsevier Publ. Comp. NY.

Kondra, Z.P. and B.R. Stefansson.  1965.  Inheritance of erucic acid and eicosenoic acid content in rapeseed oil (B. napus).  Can. J. Genet. Cyto. 7: 505-510.

Kondra, Z.P. and B.R. Stefansson.  1970.  Inheritance of the major glucosinolates of rapeseed  (B. napus) meal.  Can. J. Plant Sci. 50: 643-647.

Lamb, R.J. 1980.  Hairs protect pods of mustard (B. hirta, `Gi­sil­ba') from flea be­e­tle fe­ed­ing damage.  Can­. J. Pl. Sci. 60:1439-1440.

Larsen, L.M., and H. Sørensen.  1985.  The value of oilseed rape production in Denmark and the EEC.  In:  H. Sørensen (ed.) Advances in the Production and Utilization of Cruciferous Crops.  Martinus Nijhoff / Dr W. Junk Publishers, Denmark.

Liu, H.L.  1984.   Rapeseed Genetics and Breeding.  Shanghai Scientific Press.  Shanghai, P.R. China.

Liu, H.L.  1987.  Rapeseed Cultivation.  Shanghai Scientific Press.  Shanghai, P.R. Chian.

Mather K. and J.L. Jinks.  1977.  p.68-98  Introduction to Biometrical Genetics.  Diallels, Chapman and Hall, London.

Morley-Jones R.M.  1965. Analysis of variance of the half diallel table.  Heredity.  20(1): 117-121.

Ohlson, R. and K. Anjou.  1979.  Rapeseed protein products.  J. Am. Oil Chem. Soc. 56: 431-437.

Procter and Gamble Analytical Laboratories (PGAL).  1987.  Dietary fat in nutrition and cardiovascular health:  A research summary.  Company publication.

Powell, G.R., 1994.  The future of Canola.  In: Proceedings of the Pacific Northwest Canola Meeting, November 9-11, Spokane, Washington, 1994.

Rafiullah, S.  1994.  Use of diallel analysis to determine the inheritance of agronomic and quality characteristics in yellow mustard (B. hirta (L) Moench).  Ph.D. dissertation,  University of Idaho.

Raney, J.P., H.K. Love, G.F.W. Rakow and R.K. Downey.  1987.  An apparatus for rapid preparation of oil-free meal from Brassicaceae seeds.  Fett Wissenschaft Technologie / Fat Science Technology.  6: 235-237.

Raney, P., G. Rakow, and T. Olson.  1995.  Development of high erucic, low glucosinolate Sinapis alba.  Vol. 2 p.452-454. In:  Proceedings of 9th International Rapeseed Congress, Cambridge, UK.

Sarwar, G., R. Blair, M. Friedman, M.R. Gumbmann, L.R. Hackler, P.L. Pellett, and T.K. Smith.  1984.  Inter- and intra-laboratory variability in rat growth assays for estimating protein quality of foods.  J. Assoc. Off. Anal. Chem. 67: 966-981.

Stefansson, B.R. and F.W. Hougen.  1964.  Selection of rape plants (B. napus) with seed oil practically free from erucic acid.  Can. J. Plant Sci. 44: 359-364.

Smith, D.B. and C.N. Donald.  1988.  The measurement of glucosinolates in oilseed rape by glucose release.  Plant Varieties and Seeds 1:121-130.

Tang, J.H., J. Brown, and J.B. Davis.  1995.  Developing canola quality cultivars in yellow mustard (Sinapis alba).  In: Proc. Pacific Northwest Canola Conf., Coeur d'Alene, Idaho, November 5-7, 1995.

USDA, 1995.  Foreign Agricultural Trade of the United States, July, 1995.