Cannabis Cultivator Phenotype and Attribute Preferences in Legal Recreational and Medical Cannabis Cultivation Operations in Oregon and Colorado


 BackgroundAs recreational and medicinal Cannabis sativa production increases across the United States, concerns have arisen regarding declines in cannabis biodiversity. Studies have suggested genetic bottlenecking has occurred primarily because of breeder and cultivator preferences and practices which over-emphasize the selection of strains with high contents of the psychoactive cannabinoid, Δ-9-tetrahydrocannabinol (THC). No study to-date, however, has sought to systematically assess grower and breeder knowledge, preferences, and practices within legal operations within the United States. MethodsTwenty-one licensed recreational and medical cannabis growers in the states of Colorado and Oregon were provided with free lists to ascertain cultural domains of knowledge regarding cannabis phenotypes and attributes within commercial cannabis production operations. Semi-structured interviews were also conducted with participants to provide more nuanced explanations of free-list responses, and to clarify responses where necessary. Free-list results were initially ungrouped excepting those variables whose meanings had been triangulated utilizing interview data. Free list data was assessed utilizing Smith’s S formula to ascertain cultural salience for listed attributes and phenotypes. A second analysis was then conducted for grouped terpene-related data. Results Results from the ungrouped free list identified Yield as the most culturally salient category (.29) followed by High THC Content (.26) and Smell (.25). Overall, horticultural characteristics such as yield, plant structure, and morphology garnered were the most frequently listed, followed by chemometrics, economics, and phenology, with some overlap between these categorizations. Though THC content was described as important within commercial operations by most growers both within free-list responses and interviews, growers also expressed that this preference was due to state testing regulations and a misinformed consumer base, rather than grower partialities. Respondents also noted that they believed consumer preferences were changing as consumer demographics were beginning to trend towards “older” consumers. When terpene-related attributes within free-listed results were combined utilizing triangulation from semi-structured interviews to verify the free list data, terpenes became the most cultural-salient attribute listed by growers (.42). Conclusions The results from this study suggest that cannabis biodiversity may indeed be declining due to breeder practices that emphasize THC content, but that these practices are informed by state policies and consumer purchasing metrics rather than grower preferences.


Introduction
Over the course of the last decade, an increasing number of states have rati ed legislation legalizing the production, sales, and consumption of Cannabis sativa, creating one of the most lucrative agricultural industries within the United States. Amidst this rise in cannabis production and consumption, scholars from a wide range of disciplines have commenced in conducting research which offers a more comprehensive understanding of the circumstances and implications of these emerging cannabis markets. While research from the social sciences has indeed made inroads into the socio-cultural and political contexts and consequences of state-level cannabis legalization, most of this research has continued to focus almost exclusively on the criminology of cannabis-related activities ( Studies outside of the social sciences have highlighted however, that cannabis biodiversity has been steadily declining over the course of the last few decades, with some researchers suggesting that these deteriorations are due to cannabis grower and breeder practices (Clarke & Merlin, 2013;Clarke & Merlin, 2016; Mudge et al., 2018). As an increasingly large body of literature has identi ed the potential medicinal bene ts of a wide range of cannabinoids and terpenes found in the cannabis plant, the implications of those losses in terms of both agrobiodiversity (ABD) and pharmacology (Bridgeman & Abazia, 2017;Costa, 2007) indicate an increasing need for research which positions cannabis production within an agricultural, rather than solely economic or legal, framework.

The Ethnobiology of Agricultural Biodiversity
While certain practices and management decisions amongst farmers are necessarily ubiquitous (i.e., irrigation, weed and pest management, and nutrient management) farmer knowledge and how this knowledge is applied within agricultural practice varies amongst agricultural practitioners and can have divergent outcomes in terms of agrobiodiversity (ABD) for the broader landscape within which they are implemented (FAO, 2006;Nautiyal et al., 2008;Nazarea, 1998). A subset of overall biodiversity, ABD is de ned by the Convention on Biological Diversity, (2011) as "the outcome of the interactions among genetic resources, the environment, and the management systems and practices used by farmers" and is the foundation on which the crops that sustain human life are produced. Because of the wide breadth of factors that in uence ABD, pursuit of its preservation requires interdisciplinary approaches which include ethnobiological studies on how groups of people "interpret, conceptualize, represent, cope with, utilize, and generally manage their knowledge of those domains of environmental experience which encompass living organisms" (Ellen, 2006). Merlin (2013, 2016) offer the most comprehensive ethnobotanical assessment of desirable traits for cannabis cultivators in both traditional societies and within more recent decades. According to their research, the authors suggest without human intervention most cannabis populations contain approximately equal amounts of the psychoactive compound Δ 9 -tetrahydrocannabinol (THC) [1] and Cannabidiol (CBD), but that selection and breeding practices throughout the history of the humancannabis relationship has resulted in the preponderance of strains that include high THC and low CBD (Type I), or low CBD and high THC (Type III). Without testing technology, pre-modern cannabis farmers relied upon the effects of cannabis when consumed as well as traits such as color, smell, trichome development, and resin content (stickiness) when selecting strains to breed and cultivate, as they believed these attributes indicated high contents of psychoactive cannabinoids (i.e., High THC) which would intensify or increase the plants potential effects when consumed.
Alternatively, the authors note that while modern farmers continue to use similar traits to their cannabis cultivating and breeding forebearers such as large trichomes and "desirable" aromatic terpenes, the advent of cannabinoid quanti cation methods such as gas chromatography and high-performance liquid chromatography have resulted in an uptick in growers cultivating strains with reproducible pro les of target cannabinoids and an overall high total cannabinoid content (Clarke & Merlin, 2016;Lazarjani et al., 2020). Though modern testing methods permit growers to assess non-THC and non-CBD cannabinoid contents [Cannabinol (CBN), Cannabigerol (CBG), Cannabichromene (CBC)] within cannabis which have also been found to have medicinal properties (Appendino et al., 2008;DeLong et al., 2010;Weydt et al., 2005), Clarke & Merlin (2016) propose that the pursuit of "high cannabinoid content" continues to predominantly refer to High THC and High CBD contents which are obtained by growers through a mixture of the "heritability of cannabinoid pro les, hybrid vigor between genetically distinct gene pools and the highly focused, human arti cial selection for potency (p. 302)." The proposed hypothesis that THC content has increased in cannabis over the years at least in part because of breeder practices was substantiated in a 2018 study by Mudge et al. (2018) which found that cannabis biodiversity within Canada has been decreasing over the past few decades due to breeder and grower practices that emphasize(d) increasing the average content of THC. Mudge and colleagues specify that biodiversity declines are most pronounced amongst strains rich in the non-psychoactive cannabinoid CBD, even amidst the increasing popularity of CBD products within mainstream markets (Lamers, 2019 Despite these illuminating ndings, the fact remains that most studies on cannabis biodiversity and breeding practices have relied upon chemometric and metabolomic analyses of the cannabis plant itself rather than on data procured from cannabis breeders and growers as to their actual preferences and practices in relation to cannabis production. For example, while Mudge et al. (2019) explore the desirability of varying scent pro les (i.e., terpene contents) by growers in cannabis breeding and cultivation, and suggest that growers utilize scent pro les as "benchmarks" for determining which strains of cannabis contains high concentrations of cannabinoids, they also note that the assumed association growers make between terpene and cannabinoid content is "anecdotal" in nature, rather than based on scienti c inquiry (p. 787).
As state-legal cannabis production continues to expand within the United States, shifting legal, economic, and technological frameworks dictate the need for contemporary ethnobiological studies on cannabis cultivator preferences, practices, and knowledge in relation cannabis production. Utilizing a qualitative approach, this study seeks to begin the process of ameliorating the dearth of research available on cannabis producers by investigating the knowledge and preferences of cannabis cultivators in commercial recreational and medical cannabis operations in the states of Colorado and Oregon. In identifying the culturally informed values, beliefs, and knowledge which in uence cannabis production, this study contributes to the emerging areas of scholarship concerned with the impacts of the integration of cannabis production into the broader environmental and social landscapes within which it is occurring.
[1] Δ 9 -tetrahydrocannabinolic acid (THCA) is the naturally occurring cannabinoid within the cannabis plants which acts as the precursor for the psychoactive Δ 9 -tetrahydrocannabinolic (THC). While most state regulations dictate that total THC content reported on packaging must be calculated using the following formula: THC total = (%THCA) x 0.877 + (%THC) [OAR 845-026-0100 (15)] THC, rather than THCA or THC total, is the colloquial term utilized by cannabis growers and will be utilized here.

Methods
To ascertain cannabis cultivator knowledge and preferences in commercial operations, this study utilized a qualitative approach consisting of semi-structured interviews (N=21), and a phenotype free-listing exercise (N=21) with state-legal cannabis growers in the states of Colorado and Oregon. Research participants were recruited utilizing cannabis-focused pages on social media platforms including Instagram, CannaBuzz, and Facebook, as well as snowball sampling.

Sample
Of the interview/free list participants included in this study (N=21), fourteen (67%) worked for licensed cannabis facilities in Oregon, and 7 (33%) were licensed to cultivate cannabis in Colorado. Of the respondents 6 (29%) reported working for facilities licensed for recreational cannabis production, 5 (24%) reported working for exclusively medical facilities, 5 (24%) reported working for mixed recreational and medical facilities, and 1 (5%) reported a mixed hemp and recreational cannabis facility. Participants predominantly identi ed as white, between the ages of twenty six and forty four, and were approximately equally split in terms of gender-identity ( Figure 1).  Though respondents were requested to be currently working for a legally licensed recreational and/or medical cannabis production facility in either Colorado or Oregon, upon starting their interviews, 4 participants located in Oregon reported that they were not currently employed by a licensed company. All 4 respondents, however, were personally licensed to grow cannabis in their state and had worked for commercial, non-hemp, cannabis facilities in the past. These respondents were asked to answer interview questions using their most recent legally licensed non-hemp employer or operation as a basis for their responses and are listed under "not currently licensed" in Figure 2.  . Within the free listing process, participants are asked to record or list all entities or attributes that they can think of within the domain that is of interest to the researcher. The data produced by participants is then analyzed to assess the frequency with which listed variables are included, as well as the positioning of the variable within lists with the expectation that those items that are listed rst and more frequently are more culturally salient than those that are listed less frequently or later in the list (Borgatti, 1998;Miranda et al., 2007).
Free lists differ from open-ended surveys as they require participants to recall information about a cultural domain, rather than about their subjective feelings or experiences. Within this study, participants were asked about their preferences in cannabis cultivation which according to Quinlan (2019), could potentially fall within the domain of open-ended surveying rather than true free listing. The parameter of selecting preferences within commercial operations speci cally, however, engaged participants to think outside of their personal preferences to what is desirable for market-bound cannabis products.
Interview participants (N=21) were provided with Phenotype Free-Listing Forms at the time of their interviews and were asked to record preferred phenotypic attributes for cannabis plants included in their commercial grow operations. Free-listing data was collected prior to interviews to avoid the possibility of the contamination of data by the researcher during the interviews, which occurs when researchers inadvertently lead interviewees in their responses by priming participants with suggestions or mentions of possible responses (Quinlan 2019). Interviewees whose interviews were conducted either over the phone or via Zoom were asked to verbally list desired cannabis phenotypes and/or attributes which were then recorded by the interviewer. Interviewees whose interviews were conducted in-person were provided with the forms directly, which they then lled out by hand. Participants were provided with no other instructions than to "list the cannabis attributes or phenotypes that are most important or desirable to you when selecting which strains to include in your commercial operations". Though there was the potential for data contamination by providing examples of phenotypes or attributes, after conducting the rst few interviews, the examples of Color, Smell, THC Content, and CBD content were provided in the instructions as examples based on the responses provided by the rst few participants. The inclusion of these examples was required as participants were often unsure what was being asked of them when they were requested to list "phenotypes or attributes" of cannabis plants.
The distinction between the nature of the free-listing exercise and an open-ended survey was substantiated by participants involved in the exercise as many interviewees following the distribution of the free-list instructions asked for clari cation as to whether the variables being listed should fall under their own personal preferences for cannabis or those that they looked for in the course of their job. Though personal preferences were not included here, respondents would often specify that if they owned the grow or the cannabis being produced was for personal use, they would select different phenotypes or attributes.

Free List Analysis
Thompson and Juan (2006) working from Borgatti (1998), note that the frequency of mention of an item or variable is the simplest way to determine the salience of a variable within a knowledge domain as the most cited terms "tend to denote more locally prominent or salient items" (Silva et al. 2014: p. 23). To create a more comprehensive salience analysis however, the position of the variable within a list may also be included utilizing the Smith's S formula ( Figure 3) which combines frequency of mention with the rank of the variable (Thompson and Juan 2006). The positioning of a variable within a list is of importance in addition to frequency of mention, as participants are most likely to list rst the terms or variables that are most familiar to them thus indicating salience (Quinlan 2005).
Free list responses were analyzed using ANTHROPAC 1.0-Freelists which applies Smith's S formula to assess salience. In cases where respondents listed two variables on a single line (e.g., Taste and Smell), responses were separated with the rst listed variable placed at a higher rank (i.e., 1-Taste, 2-Smell). According to Quinlan (2005) and Borgatti (1998) selecting which items are salient within a free list is often complicated by the fact that most lists result in "a huge number of items that are each mentioned by just one person" (Borgatti, 1998), and suggest that the most effective method for selecting salient items is to utilize natural "breaks" in the data in terms of frequency of mention using a scree plot. Based on the outcomes of the free list-analysis (see Figure 6) which resulted in a natural break at the 10% mark, this cut-off was utilized to determine which items were culturally salient and should be included in the nal analysis.
The analysis of free list data provided by respondents produced some unique challenges considering the interrelationship between cannabis phenotypes and chemometrics, particularly in relation to terpenerelated attributes. Terpenes or "all natural compounds built up from isoprene subunits that predominantly originate from plants" (Breitmaier, 2006 directly referenced, and tangentially alluded to, terpene contents. For example, while some growers listed Unique Terpene Pro les and Wide-Ranging Terpenes, others referred to terpene-related attributes such as Smell Spectrum or more speci c avor/scent attributes such as Fruity or Gassy plant. Borgatti (1998) argues that within free lists, subdomain items and like-concepts may be treated as synonymous when clarifying information is provided by respondents. Semi-structured interviews were thus conducted following the free listing portion of the data collection which offered the opportunity to collect a "more exhaustive and contextualized" set of responses from participants regarding phenotype and strain preferences within commercial operations(Zambrana et al., 2018: p. 201). Free list response meanings that were veri ed using the interview data before being combined with "like" responses have been marked as such within the results (see Appendix A).

Interview Analysis
Interviews were conducted following the free-listing activity, in person, over the phone, or over Zoom. Interviews lasted between one to two and a half hours during which cannabis growers were asked questions regarding their preferences and practices in relation to cannabis strain selection and cultivation. Interviews were recorded, transcribed, and then coded for data related to cultivator preferences and practices utilizing the NVIVO qualitative data software program. [1] Both respondents in this category identi ed as mixed race Native American, American Indian, or Paci c Islander & white.

Results
The free list analysis resulted in fty-two unique attributes or phenotypes of interest to growers in their commercial operations with nineteen items producing frequency scores greater than 10%. When items were ranked by salience score rather than frequency, 4 additional items (Temperature Resilience, High Terpene Content, and Unique Terpene Pro le) with frequency scores of 9.5% emerged as being more culturally salient than other attributes with a >10% frequency. As these items were approximately within the 10% cut-off and exhibited higher salience scores, they have also been included within this analysis.
The results of the free list portion of this study are pictured in Figure 4, with listed attributes organized by highest (left) to lowest (right) salience scores.
The top twenty-three attributes by frequency of mention and salience were sorted into the broader categories of Horticultural, Phenology, Chemometrics, and Economic ( Figure 5). Most listed attributes by participants fell within the horticultural domain, followed by Chemometric, Economic, and Phenology. The following analysis presents the free listing data by category and highest to lowest salience scores. Some items have been combined within the analysis however, based on similarity of the items or the relationship between the items described by participants (e.g., Bud Size and Bud Density, Smell and Taste). The results presented here are supported by ndings from the interviews which expand grower's rationale for selecting attributes their importance to commercial cannabis production. Interviews were also employed to clarify broad or unclear terminology utilized by growers and assisted with the grouping

12
Rec and Med I CO When you get your license it's not how many pounds you can grow. You can grow a million pounds. There's no limit on the amount you can produce but the limit is on the number of plants, so if you're a businessperson, you're not going to pick all tiny plants that are the best because you're not going to make any money of items, where appropriate. For example, while one grower noted that "THC Content" was of importance in their free list, it was unclear what type of THC content was preferable (high or low). Interview data with this grower revealed that High THC content was of interest to the consumer ("since the consumer is chasing this rst high, growers have been like well then I'm going to fucking give it to you" [#5]) and caused the participant to adjust their strain selection processes accordingly indicating that the item THC Content could reasonably grouped within the category of High THC. Items whose meanings were clari ed using the interview data and were subsequently grouped with other like terms within the free list analysis are denoted as such within Appendix A.
Horticultural Yield: Frequency (43%), Salience (.29) Yield resulted in both the highest frequency and salience scores within the free listing analysis. Yield was also emphasized within interviews with twelve (57%) of the twenty-one respondents explicitly mentioning it as an important attribute when considering what to grow in a commercial operation ( Figure 6). Growers who discussed yield expressed consensus that maximizing production was desirable within commercial contexts as this resulted in greater nancial returns, indicating a single versus multi-directional approach in grower preferences and practices. Smell garnered the third highest salience score out of free-listed attributes with seventeen (81%) respondents also directly discussing the importance of smell, fragrance, or aromas within the course of their interviews. Likewise, while only 24% of respondents included avor/taste within their free lists thirteen (62%) growers speci cally noted the importance of avor and taste within the interview data. As the sample data in Figure 7 illustrates, the smaller frequency of responses that speci ed smell and/or taste/ avor in the free lists is likely due to the interconnectedness of these variables and the tendency of some growers to use these terms interchangeably [#1, #9, #12]. Growers were also aware that taste and smell were a result of terpene content which was potentially inferred through the inclusion of items such as Fruity Terpenes, Gassy Terpenes, or Exotic Fruit Terpenes. The lack of speci cation as to the preferred outcome (smell, taste, or both) associated with these terpenes however, required that these attributes remain as standalone items within the free list analysis.  one grower summarizing the association by stating that "the nose knows". Participant [#18] who is a well-known breeder within the Paci c Northwest echoed this sentiment by explaining that they tended to breed for " avors, smells and experiences" rather than cannabinoids, as everyone's cannabinoid system is different and so breeding for THC and CBD would be "useless".
Disease Resistance: Frequency (29%), Salience (.20) Within free listed responses, 3 participants speci cally listed resistance to the fungal plant pathogen Powdery Mildew (Golovinomyces cichoracearum), and 2 respondents speci ed that they sought strains that exhibited mold resistance. Powdery Mildew Resistance was also the most frequently mentioned disease-related trait within interviews with 8 (38%) of growers suggesting that they seek out plants with higher levels of resistance to this common fungal infection. 2 growers discussed di culties in contending with the fungus botyrtis (boytrytis cinerea); however only 1 grower speci ed that they selected for strains with botrytis resistance, while the other grower simply noted that the fungal pathogen was a problem. Similarly, while 11 (52%) of growers expressed concerns about contending with mold in their grows, only 2 respondents within the free list exercise speci ed that they selected for mold-resistant strains though 4 (19%) growers noted that mold-resistance was a desired attributes within the interview process.
Height: Frequency (33%), Salience (. 19) Preferences regarding height were also itemized both within the free listing exercise and interview data, though levels of speci city in responses varied with 1 grower noting that "Height" was of interest, and 1 grower suggesting that "Bigger" plants were desirable, speci cally in outdoor production. Because interview data was not able to substantiate the more nuanced meaning associated with these terms, these responses were analyzed individually within the free list analysis. 33% of respondent's however, explicitly listed Short(ness) as a desirable trait. The import of the short stature of cannabis for growers was elaborated upon by interview participants who explained that shorter plants were preferable as indoor growers are limited by the height of the ceilings of the buildings in which they are producing [#7, #14] while outdoor growers in Oregon in particular, expressed a need for shorter plants (under 8 feet) as these were more likely to be able to withstand inclement weather and were easier to manage [#4, #9] (Figure 8). While large and dense owers could arguably be grouped within the larger umbrella of Yield considering the ower is the commercially valuable part of the plant, the separation of these attributes within some grower's lists indicated a preference for Large or Dense Buds for reasons separate from overall production. Interview data con rmed the validity of this approach with growers explaining that, while large buds were commercially desirable ("No one wants to get a bag of little popcorn stuff, it's like here's an 8 th of your BIG bud"[#14]), strains with larger buds were also preferable because "big buds are obviously cheaper to harvest [#13]," as they required "less trimming and maintenance overall…and drying space" [#14]. Dense Buds were described as of importance in terms of assisting with moisture content though some growers reported that less dense buds were more appropriate in damper climates such as Oregon, where the development of mold was a concern.
Vigor: Frequency (29%), Salience (.14) Vigor was indicated as being desirable by almost a third of producers, however what was meant by the term Vigor varied amongst interviewees, with some growers describing vigor as speed of growth [#13, #18], and others describing vigor as pest and disease resistance [#13] (Figure 9). One respondent associated vigor with branching patterns [#3] explaining that the spacing of branches and how many leaves were being produced could indicate "how vigorous" the plant would ultimately be. , and so were less desirable than plants that grew more sparsely (Figure 10). Similar to the reporting style exhibited in relation to disease resistance, growers who discussed pest resistance were more likely to generally discuss common pest pressures rather than indicate that they selected strains that exhibited pest resistance. For example, while 6 (29%) growers noted that contending with pests was a central responsibility within their grows, only 2 of these respondents revealed that they selected strains that were genetically adapted to contend with pest pressures. Common pests identi ed by growers included russet mites, aphids, spider mites, and thrips with spider mites being mentioned by the largest percentage of growers (29%).
Trichome Development: Frequency (19%), Salience (.1) Though most traits were addressed at a higher frequency within interviews than in free list responses, trichome development was included by only 4 (19%) of respondents within their interviews. Of these references, 3 addressed the quantity of trichomes with growers indicating that higher quantities of trichomes were desirable as trichomes are associated with cannabinoid and terpene content, and 1 grower proposing that the color (milky) of the trichomes was the most important component of this attribute.
Easy/Low Inputs: Frequency (19%), Salience (.086) Strains that performed well in low input environments or that were not heavy "feeders" were described as desirable to growers as a lower need for inputs also meant less labor in feeding the plants and reduced cost when purchasing nutrients. 1 grower also expressed that they felt that the utilization of lower inputs resulted in a better product that "processed better and smoked better [#19]" in addition to being cheaper to produce.
Temperature Resilience: Frequency 9.5%, Salience (.082) Temperature (heat) resilience was listed by only 2 respondents within the free listing portion of the exercise, both of whom were outdoor growers. Though many respondents discussed temperature and climate within their interviews, only 1 outdoor grower suggested that they selected strains that were speci cally acclimated to high heat. The lack of attention to temperature resilience may at least be in part due to the preponderance of indoor growers involved within the study as indoor growers have greater control over the humidity, air ow, and temperature of their grow environments. the respondents, only one grower indicated that he speci cally selected for more "pure" or "extravagant" colors such as purples, violets, or blues. Other growers however, generally speci ed that color was of importance, but without elaboration.

Phenology
Flowering Time: Time to ower/harvest was also noted within the free listing portion of the data collection process with 29% of respondents including this variable. In interviews, Short Flowering Time was speci cally expressed as being desirable for outdoor growers in Oregon because it assisted in ensuring that product was harvested before the end of the growing season ( Figure 11).   Figure 12: Chemometrics-High THC A common narrative expressed by growers regarding High THC was that the drive to produce High THC strains was a result of consumer, rather than grower, preferences and was the result of the miseducation or a lack of education on the part of the consumer and cannabis sales representatives (i.e., "budtenders") ( Figure 13). Interestingly, while some growers acknowledged catering to consumer demands and selecting and propagating High THC strains [#2, #14, #16] others were more likely to speak about the push for Higher THC strains more generally or as a practice applied by other growers [#5, #6, #13, #17, #18, #20] rather than something that they themselves engaged in.  We're still in a world where testing labs are getting death threats because somebody sent samples there that they were sure was 30% THC and it came back as like 22%.
6 Rec I OR Now I feel like more people are just caring about how much they can sell rather than the quality of it, and how much percentage of THC they can get into it so they're pumping all this random stuff into it and crossing these random crosses that don't go together that give people that experience. Participant: I don't think so, no. People seem to think that but no, I don't really think so. All our stuff tested at the last fall harvest, at 20% or above. Between 20 and 25%.
As a key informant, respondent #18's opinion regarding the lack of import of High THC content in producing a high-quality end cannabis product, but the import of THC to consumers, was of particular interest, particularly as this sentiment was echoed by all other growers who directly addressed the value of High THC excepting one participant [#15] indicating a cultural consensus.
Narratives regarding trends in increasing the production of lower THC strains also emerged that addressed the widening of the cannabis consumer base to include "older" clientele, who were less likely to be interested in strains with high THC content ( Figure 14). Rec I OR I think there needs to be some marketing to older people because I know so many old people that don't like to smoke because they've always been given the wrong product and those low-THC strains would probably be really good for people While THC content was noted as being important by most growers, 17 (81%) of respondents also directly discussed CBD within the course of their interviews though all producers were engaged in the non-hemp recreational or medical markets. Of these respondents, 4 noted that they were speci cally engaged in producing cannabis strains that were higher in CBD content. As one grower explained [#16] (Figure 15), low CBD strains have increased in popularity due to the inability of hemp growers to produce strains for recreational markets.  Nevertheless, only 13 (62%) of growers mentioned the importance of terpenes directly within their interviews as opposed to 76% who directly discussed or referenced THC. Also, while the cannabinoids THC and CBD were frequently discussed by name within interviews, only 6 (29%) interview participants named speci c terpenes within the interview data, and of these responses 3 participants [#5, #6, #10] were responsible for 70% of the total frequencies (See Figure 16). Of these mentions, no respondents speci cally associated aromas or tastes with a particular terpene but rather spoke of the terpenes more generally as components of the plant (Figure 17). As in the case of terpene-related data, the relatively low salience score produced for Market Demand was complicated by the integration of discussions on consumer preferences situated within the data on cannabinoid and terpene content. THC content speci cally for example, was associated by growers with consumer preferences. As THC content was the second most culturally salient category, Market Demand may be assumed to have a more signi cant weight than what is indicated by the free list responses.
Uniqueness in Grows: Frequency (14.3%), Salience (.053) The desire for strains or genetics that varied from what growers currently had in cultivation in their grows, spoke to attempts to meet the demands of a wider consumer base. Growers were aware that variations in terpene and cannabinoid pro les are of interest to different consumers within the market with one grower noting that: "Breeders are essentially just chefs and we have different palettes and stuff that we like so you know if you go to a big city there's a bunch of steak restaurants but they're de nitely cooked differently at different restaurants. And you can nd a chef that cooks steaks just the way you like it. That's what our customer bases are doing, they're just looking for the chef where they agree with the palette, and they like what they see and taste." Interestingly, while some growers had suggested that the advent of testing had limited biodiversity by placing THC content at the front and center of the consumer's consciousness, a different participant suggested that testing had increased biodiversity as growers had to be conscientious that the cannabinoid and terpene content of the plant were going to be made publicly available: "Some people really enjoy just growing a couple like the black market grows that I've helped with or worked at, we've grown just two or three just because the ower isn't being tested at all through a lab. So really the black market grower is looking for a really pungent smell and the weight of the nug and the size of it versus….for example Jager is a high-yielding plant and a lot of black market growers love to grow this plant. It has a little bit of the possibility for mold, and the numbers are a little average when you get it tested, but because the yields are so much, most growers opt for that strain in the black-market industry. I think generally the farms that I've worked on in the rec market, the farms have really wanted to get variety in their strains because they want to keep it exciting for the market so if someone doesn't like a strain they have many others to choose from."

Discussion
The results of the research presented here support the hypothesis proposed by Merlin (2013, 2016) that cannabis growers are most familiar with the cannabinoids CBD and THC, and that strain selection and breeding practices within state-legal commercial cannabis operations tend to favor the maximization of these compounds, and in particular THC. The results of this study also suggest however, that the desirability of these attributes, rather than stemming from grower preferences, has predominantly originated from the consumer and market trends. Participants suggested that consumer preferences for THC were the result of a lack of knowledge within the consumer base regarding non-THC and non-CBD cannabinoids, combined with state-mandated testing regulations which require the reporting of these major cannabinoids on cannabis packaging. Regulations which prevent consumers from smelling cannabis products prior to purchase were also indicated as being responsible for an emphasis on THC content as opposed to terpene content within cannabis markets, as scent was associated with a consumer's ability to "intuit" which strains contained appropriate chemical compounds for their unique endocannabinoid systems.
Data which addressed terpene content within the cannabis plant produced some unique challenges considering the interrelationship between cannabis phenotypes and chemometrics, and speci cally how growers chose to express preferences for various terpene-related attributes. Terpenes or "all natural compounds built up from isoprene subunits that predominantly originate from plants" (Breitmaier, 2006: p lists, grower's responses both directly referenced, and tangentially alluded to, terpene contents. For example, while some growers listed Unique Terpene Pro les and Wide-Ranging Terpenes, others referred to terpene-related attributes such as Smell Spectrum or more speci c avor/scent attributes such as Fruity or Gassy. When results were combined to re ect all terpene-related attributes however (Appendix B), Terpene Content became the most culturally salient category within this analysis. Growers, furthermore, did not indicate that they consciously associated terpene-related attributes such as aroma with cannabinoid content as suggested by Mudge, Murch, and Brown (2019), rather, the terpenes were valued for their aromatic and taste outputs.

Conclusions
The cannabis plant has over one hundred and fty terpenes and one hundred cannabinoids whose form and function are still little understood (Booth & Bohlmann, 2019). Though participants in this study indicated that terpene content was of considerable importance when selecting for desirable cannabis phenotypes and attributes, most were also unlikely to describe or identify associations between speci c terpenes (e.g., linalool, myrcene, pinene, etc.) and aromas, avors, or effects. The omission of this association does not necessarily mean that cannabis growers are ignorant of the value of terpenes or that their understanding of how the chemometric components of Cannabis sativa interact. Cannabis producers hold knowledge that is most often based on experience and experimentation. Reliance on the domain of knowledge embedded in practice by cannabis growers then, make them key actors and Free List Item Frequency and Salience Scores