Abstract
Depending on the perceived balance of risk and benefit, and on the perceived unnaturalness, some applications of gene technology appear more acceptable to the public than others. This study asks whether a person’s knowledge of biology affects their assessment of these factors differently. A random sample of the Danish population (n = 2000) was presented with questionnaires. The respondent’s knowledge was measured by a number of textbook questions on biology. The results indicated that knowledge increases the likelihood that a person will have differentiated opinions on medical and agricultural applications, but decreases the likelihood that he or she will differentiate between cisgenic and transgenic cereals. We discuss the implication that knowledge makes people more likely to base their acceptance on judgements of risks and benefits, rather than on judgements of naturalness. The article concludes that the effect of knowledge on acceptance cannot be generalised wholesale from one application, or method, to others.
1. Background
The viability of genetically modified (GM) food products is dependent on public approval. Such approval cannot, however, be taken for granted. Instead, numerous European population surveys have documented widespread opposition to the use of gene technology in food production (Gaskell et al., 2003). To explain this opposition, considerable attention has been devoted to the study of the public understanding of biotechnology. Here evidence showed that only a small percentage of the public adequately understands the techniques involved in genetic modification (Curtis and Moeltner, 2007; Gaskell et al., 2000; Steinhart, 2006). Moreover, evidence was found that scientific knowledge is positively correlated with support for science in general, but not necessarily with support for specific technological applications (Allum et al., 2008; Moerbeek and Casimir, 2005).
This general correlation, along with the currently limited level of public understanding, has led some authors to assert that public acceptance can be earned by providing more information about gene technology (Bonny, 2003). Against this assertion others have argued that although such information drives out indecision, it does not always lead to acceptance (Biotechnology and the European Public Concerted Action Group, 1997). Armed with scientific knowledge, people seem to become more willing to take a stand, but sometimes increased opposition appears to be the net result (Madsen et al., 2003).
The primary focus of this debate has been on the direct correlation between scientific knowledge and public acceptance. In reality, however, acceptance is likely to be grounded in specific reasons – e.g. reasons relating to the risk–benefit ratio and to the perception of GM organisms as too unnatural (Lassen and Jamison, 2006; Shaw, 2002; Streiffer and Hedemann, 2005). An investigation of whether increased scientific understanding affects these reasons differently may provide new insights into the relationship between knowledge and acceptance. It may also explain why the effect of knowledge on acceptance seemingly varies across applications.
If scientific knowledge leads people to attach greater weight to some (but not other) reasons for accepting or rejecting gene technology, the acceptability of applications with different characteristics ought to vary with different levels of that knowledge. As far as we have been able to discover, however, previous studies have not investigated whether the comparative acceptability of different applications is constant across knowledge levels, or whether differences in the extent and depth of people’s scientific learning makes them differentiate between acceptable and unacceptable in a different manner.
There is ample evidence that, at an aggregated level, the acceptability of certain applications exceeds that of others. Instead of rejecting gene technology as such, it appears that most people base their assessment on judgements about the means and the end of each specific application (Pardo et al., 2002). What matters here are not just the risks and benefits viewed in isolation, but the risk–benefit balance. Benefits that are seen as crucial will often offset risk perceptions (Frewer et al., 1997a; Gaskell et al., 2000). Faced by serious illness few people, for example, are going to refuse a drug, or a treatment, on account of its origin, whereas a refusal to buy GM food products would not carry great costs, and so here a person might be reluctant to accept even minor risks (Frewer and Shepherd, 1995; Gaskell et al., 2000; Moses, 1999). As a result medical applications of gene technology, which are typically perceived to be useful and relatively risk-free, are widely accepted, whereas the deployment of gene technology in food production remains highly controversial. In contrast with medical usages, agricultural applications of gene technology are generally regarded as less useful and more risky (Gaskell et al., 2003).
To investigate whether knowledge affects the significance an individual attaches to the risk–benefit ratio, we will therefore compare, across knowledge levels, the acceptability of medical and agricultural applications.
In addition to its dependence on the risk–benefit ratio, acceptability is known to be restricted by the perceived unnaturalness of the modification in question. One aspect of this unnaturalness is the phylogenetic distance between the modified organism and the source of the inserted genes. Because hybrids of incompatible kinds may be perceived as though they have no essence of their own (Wagner et al., 2010), most people seem to prefer plant-to-plant transfer to hybridisations between sexually incompatible species (Burton et al., 2001; Lusk and Rozan, 2006; Lusk and Sullivan, 2002; Onyango and Nayga, 2004). In a study of consumers in Mississippi, Lusk and Sullivan (2002) found that 80% would eat a vegetable with an extra gene from the same vegetable, while fewer than 25% indicated a willingness to eat vegetables genetically modified with a gene from a virus, bacterium, fungus or animal. In a similar vein, Knight (2009) concludes that while the combination of genes from similar plants may be relatively innocuous, many people would oppose animal-to-plant transfer, mostly on account of their belief that such a process is unnatural.
Against the background of this public concern that hybrids of incompatible kinds are too unnatural, new lines of GM crops are being developed as a more natural alternative to traditional transgenic crops (Myskja, 2006; Nielsen, 2003). These crops are called cisgenic crops. By definition cisgenic crops are modified only with native genes or genes derived from close relatives; by contrast transgenic crops are modified with species-foreign material (Jacobsen and Schouten, 2008). Because cisgenic plants do not transgress the borders of species many people do seem to find them more natural (Gaskell et al., 2010) (although it is important to note that they would not necessarily satisfy all relevant criteria of naturalness (Mielby et al., submitted)). Assuming that cisgenic crops are more natural than transgenic crops, in the eyes of the public, they can be used to test the relative importance of naturalness-based considerations across knowledge levels.
In order to investigate how scientific knowledge affects the impact, on acceptability, of erceived naturalness, and risks and benefits, this paper examines whether applications that are different in these respects become more or less acceptable with different levels of knowledge. Instead of asking people in abstract terms how concerned they are about unnaturalness, we seek to discover their opinions about cisgenic, and then transgenic, crops. To compare the importance of the risk–benefit ratio we ask how people with different levels of scientific knowledge evaluate medical applications relative to agricultural applications.
The outcome of such an investigation is obviously going to depend on the type of knowledge investigated. So much is apparent from previous debates over the effect of knowledge on acceptance, where knowledge has been conceptualised in different ways, with different results. Some studies, for example, have distinguished between direct knowledge (acquired in personal experience) and indirect knowledge (acquired through others’ testimony). Others have distinguished between objective knowledge (being informed) and subjective knowledge (feeling informed) (House et al., 2004). The first of these distinctions prompted investigations into whether information is processed differently depending on the modality and source of information (Curtis and Moeltner, 2007); the second raises the question whether acceptance is more strongly associated with subjective rather than objective knowledge (House et al., 2004; Lusk and Sullivan, 2002).
To date most Europeans have had little or no personal experience of GM products, so most attitudinal opinion surveys on gene technology in this region (including the Eurobarometer series) have focused on indirect knowledge, which is ordinarily acquired through the education system or the mass media. Again, European research has been chiefly concerned with objective knowledge, which is typically interpreted in terms of the ability to correctly answer a number of textbook questions about basic biology or genetics (House et al., 2004).
Though the present study sits in this tradition, we recognise that this sets a rather narrow criterion of what it means to be informed. Other, equally topic-related types of knowledge, such as knowledge of the prevalence of GM products, or their regulation and societal impact, may be equally or more relevant in the formation of people’s opinions. A person need not be conversant with basic biology to understand the broader social, ethical and political implications of gene technology. As Fischhoff and Fischhoff (2001) have argued, objective measures tend to emphasise the perspective of curricular science at the expense of practical perspectives that are in fact central to policy-making, even though people may feel quite comfortable judging policies on new technologies without claiming to possess special knowledge of the technologies themselves.
For the remainder of this paper we will, however, focus on the effect of variation in objective knowledge alone. We shall ask whether this type of knowledge leads people to differentiate in another manner, when asked about the acceptability of medical and agricultural applications, and about the acceptability of cisgenic and transgenic methods of transformation.
2. Data
An unsolicited questionnaire was sent to a random sample of 2000 individuals drawn from the Danish Civil Registration system. The sample was restricted to people who were resident in Denmark and above the age of 16. The questionnaire was sent on 8 October, 2009, and three weeks later a reminder was sent to those who had not responded. A total of 753 (37.7%) of the forms were returned. These were optically scanned by an external agency, and the resulting data set was analysed using SPSS v17.0 and SCD/DIGRAM v1.92.0, which was used for scale validation.
A comparison with protocol data from Statistics Denmark showed that there were differences in response rates depending on age, income category and education. Thus, among the respondents, persons below the age of 35 and persons belonging to the second lowest income category (low income jobs) were under-represented, whereas those in the lowest income category (typically people receiving student grants, pensions or unemployment benefits) and those of higher levels of education were over-represented.
Problematic though they are, neither of these biases is uncommon in this kind of survey research. Younger people are often more reluctant to participate in postal surveys. Again, response rates of questionnaires on complex issues are often sensitive to, and biased by, occupational and educational background. The effort required to fill in a questionnaire is simply less onerous for people accustomed to office work. People who are unemployed or retired, on the other hand, may have more time available, which may explain the over-representation of people in the lowest income category. For present purposes, the educational bias – with those with higher levels of education being over-represented – was of most concern, because it might have resulted in population averages of scientific knowledge being overestimated. As our aim was to investigate the associations between knowledge and acceptance rather than the prevalence of acceptance, we did not apply weights in the analysis.
3. Method
Empirical investigation of the question of whether scientific knowledge makes people differentiate in another manner requires operationalised definitions of “knowledge” and of “differentiation.” To facilitate data collection these definitions need to be fairly simple in their construction. Following the literature on objective knowledge, we therefore measured knowledge via a number of “true or false” questions summarised in a single knowledge scale. Each question included a “do not know” category which, in the data analysis, was categorised together with incorrect answers, so that responses were recorded using a binary variable of right and wrong. The standard argument for collapsing the categories “incorrect” and “do not know” into one is that answering a question incorrectly is merely a way of demonstrating that one does not know the answer.
This rationale, however, ignores the possibility that those admitting that they do not know are generally more cautious than those who – not knowing the answer – make optimistic use of the 50:50 chance of guessing the right answer. Future research could, of course, investigate whether or not this difference has a bearing on attitudes to gene technology. However, on measurements of political knowledge, which rely on a similar item format, Sturgis et al. (2008) finds no real evidence to support the idea that the “do not know” responses conceal a degree of partial knowledge. By and large, those saying they do not know really do not know (Luskin and Bullock, 2011).
The questionnaire included ten questions on basic biology and genetics. Of these, eight satisfied the assumptions of a unidimensional Rasch model (Fischer and Molenaar, 1995) and were subsequently included into the final knowledge scale. We tested the model assumptions by partial gamma coefficients using repeated Monte Carlo estimates for assessment of significance. The Benjamini-Hochberg procedure was used to correct for false significance caused by multiple testing. Answers to the questions included were positively associated with the same latent variable (scientific knowledge). Given the latent variable, they were found to be conditionally independent both of each other and of the following exogenous variables: gender, age, income, education and political affiliation.
Four of the questions were replicated from Miller (2006). A fifth question from the same study was translated into two separate questions, both of which satisfied the model assumptions. Instead of asking whether “Antibiotics kill viruses as well as bacteria,” we asked whether “Antibiotics kill all forms of viruses” and whether “Resistant bacteria are able to withstand the effect of antibiotics.” This reformulation was motivated by a concern that the original question could in principle be answered correctly even if one wrongly believed that antibiotics kill viruses and not bacteria.
Although open-ended questions, as argued by Miller and Kimmel (2001), are in some respects preferable to closed-ended questions, they are also more difficult to manage and tend to increase dropout rates. Bearing in mind the postal format of our questionnaire, we therefore adapted from Miller (2006) only the closed-ended questions on biology, and only those questions with the highest factor loadings (i.e. loadings above 0.5). In addition, we posed four novel questions, two of which made it into the final scale, namely: “Is it true or false that men and women normally have the same number of chromosomes?,” and “Is it true or false that it is called cell division when new hereditary genes are inserted into a cell?” Both of these questions fell in the difficult end of the scale, thereby enabling us to better differentiate between people with more than average scientific knowledge.
Table 1 lists the eight items included in the final scale. The mean values can be interpreted as the percentage of the participants who answered correctly for each particular item.
The knowledge scale. Items are sorted from easiest to hardest on the basis of mean item scores; the latter are interpretable as the percentage of respondents who answered each item correctly.
Alpha = 0.738.
Knowledge scores ranged from 0 to 8, with people answering correctly 5.0 (SD 2.1) questions on average. The mean score for those with a college-preparatory education was 6.0 (SD 1.6), whereas those without had a mean score of 4.5 (SD 2.1). In Denmark the phrase “college-preparatory education” describes a class of education which qualifies students for admission to a university. These educations are intended for young adults between ages 15 and 18, and typically last three years.
Although we intended the scale as a measure of objective knowledge, it is inescapably affected by subjective knowledge, i.e. by what people think they know. If a person knows the right answer to a question but chooses the “do not know” option because he doubts the correctness of that knowledge, he or she will be grouped together with people not knowing. As alluded to above, then, caution may in some cases lead to a lower scale score than random guesses.
Differentiation was operationalised in terms of variation in acceptance. Where a respondent was more accepting of one application of gene technology than of another we inferred, or attributed to that respondent, a differentiated view. Conversely, we inferred, or attributed, uniformity if two applications were regarded as equally acceptable. Thus defined, our operationalisation of “a differentiated view” embodied an element of preference: it was not, as the ordinary meaning of the term “differentiation” might suggest, merely a matter of being able to tell things apart. To determine people’s preferences, we compared the acceptability of medical applications of gene technology with the acceptability of two agricultural applications, GM food and animal feed – the relative acceptability of which we compared as well. To determine relative preferences for cisgenic and transgenic crops, the acceptability of a pair of applications was compared: one involving the insertion of genes from a bacterium (transgenesis) and the other involving the insertion of genes from a closely related plant (cisgenesis) into a cereal crop.
We measured acceptability by two series of questions, one asking whether the respondents regarded themselves as proponents or opponents of gene technology when it is applied to the above-mentioned areas, and the other asking whether people agreed or disagreed that they would not mind buying bread made from cereals modified with genes derived from the different donors. In response to both lines of question, answers were given on a 5-point scale. This scale ranged, in the first case, from “opponent” to “proponent,” and, in the second case, from “Strongly agree,” through “agree,” “neither agree nor disagree” and “disagree,” to “strongly disagree.”
Although it is possible that these categories were not fine-grained enough for us to detect the presence of minor preferences, they did allow an intuitive grouping of respondents as either having or not having differentiated opinions. Through a pair-wise comparison of the acceptability of each application or method, four new variables were computed. These variables recorded whether the respondents were equally accepting of the paired applications/methods.
Among the respondents, 57.3% differentiated between medicine and foods, whereas 50.5% differentiated between medicine and animal feed. The vast majority did so out of preference for medical applications. Only slightly more ambiguous was the preference for animal feed over foods, with 29.4% having differentiated opinions on these two agricultural applications. The vast majority of those 60.7% who differentiated between cisgenesis and transgenesis, favoured cisgenesis.
4. Data analysis
When the acceptability of two applications is compared, knowledge can be said to have an effect on how people differentiate only if that knowledge affects the acceptability of at least one of the paired applications. As part of the initial analysis we deployed one-way ANOVA to see whether this condition was satisfied. Realising that a causal relationship, should it exist, would run from knowledge to acceptance, we used the acceptability measures as predictors of scientific knowledge. We did so in large part because we were keen to use simpler, yet sufficient, statistical procedures. Unequal variance arose between groups, so Welch’s ANOVA was performed. To correct for multiple testing we used Tamhane tests with a significance level of 0.05 for post hoc analysis. In this part of the analysis the favourable response categories, which were not characterised by significantly different knowledge means, were collapsed. As the unfavourable categories did not have different knowledge means either, these too were collapsed. This made it possible to depict acceptability as a three-group construct. The results of these analyses are shown in Table 2 and Table 3; they will be discussed more fully below.
Average knowledge scores for different levels of sympathy. The distribution of respondents across groups is listed as well.
Numbers within parentheses are the standard deviations.
η2 = .113 F(2,735) = 47.03, p = <.001; Welch’s p = <.001.
η2 = .053, F(2,739) = 20.65, p = <.001; Welch’s p = <.001.
η2 = .068, F(2,737) = 26.78, p = <.001; Welch’s p = <.001.
Average knowledge scores for different levels of agreement. The distribution of respondents across groups is listed as well.
Numbers within parentheses are the standard deviations.
η2 = .035, F(2,737) = 15.04, p = <.001; Welch’s p = <.001.
η2 = .039, F(2,738) = 13.50, p = <.001; Welch’s p = <.001.
Having confirmed that scientific knowledge was indeed associated with acceptance, we proceeded to address the question whether that knowledge encourages people to differentiate in any other manner. Thus we entered the differentiation variables into binary logistic regression models to examine their association with knowledge. In addition to the knowledge scale, we included a binary measure showing whether the respondents had completed a college-preparatory education. Tables 4 and 5 show the regression models.
5. Results
Table 2 shows that 55.2% of participants had a positive attitude to genetic modification for medical purposes, and that the mean knowledge score for those participants was 5.73. This score is significantly higher than the mean score of both the opponents (4.43) and the undecided group (4.21). The mean scores of the opponents and the undecided were not, however, significantly different from each other. Similar trends were seen for both GM foods and animal feed, although in the latter case the mean knowledge score was significantly different across all three categories. For each application, the association between scientific knowledge and acceptance was evident and highly significant, making it very unlikely that the observed differences in mean scores had occurred by chance. The effect size was relatively small for both food and feed, and only slightly bigger for medical applications.
Additionally, Table 2 confirms that acceptability varies with the area of application. Compared to the number of participants in favour of genetically modified foods (25.9%) and animal feed (31.1%), medical applications was favoured by a larger percentage (55.2%). By contrast, the difference between modified foods and feed was less pronounced, animal feed being only slightly more acceptable.
Like Table 2, Table 3 shows that there are significant differences in the mean knowledge scores of the three groups. This applies to both cisgenic and transgenic transformation, which indicates that the effect of knowledge on acceptance of modified foods (reported in Table 2) does not disappear, or cease, when the method of transformation is specified. Post hoc tests revealed that the mean score of those who agreed that they would not mind buying bread made from cisgenic cereals was significantly higher than the mean score of those who neither agree nor disagree; but not significantly different from the mean score of those who disagree. However, the mean score of respondents accepting bread made from transgenic cereals was significantly higher than the scores we found in the other groups, and the latter scores were not significantly different from each other.
Supporting the assumed relationship between acceptability and method of transformation, Table 3 also shows that cereals modified with related genes are acceptable to a larger proportion of respondents than are cereals modified with genes derived from a bacterium.
Turning to the regression models shown in Table 4, we see that the likelihood of differentiating between medicine and food was slightly higher than the likelihood of differentiating between medicine and animal feed (for similar levels of knowledge and education). The two models are similar, however, inasmuch as both show the positive effects of scientific knowledge and of education. By contrast, neither knowledge nor education affected the likelihood of differentiating between foods and animal feed significantly. We did not find evidence of interactions for any of the models portrayed in Table 4. Hosmer and Lemeshow tests were non-significant for all models, indicating that the models fit the data well enough. Still, the models accounted only for fractions of the variability in the dependent variables.
Effects of knowledge and college-preparatory education on the likelihood of differentiating between different applications.
The model correctly predicts 60.8% against 57.4% in the baseline. Hosmer and Lemeshow = .925, Nagelkerke R2 = 0.061.
The model correctly predicts 57.3% against 50.4% in the baseline. Hosmer and Lemeshow = .596, Nagelkerke R2 = 0.058.
The model correctly predicts 70.5% against 70.5% in the baseline. Hosmer and Lemeshow = .195, Nagelkerke R2 = 0.005.
Where differentiation between transgenic and cisgenic cereals was concerned, the effect of knowledge seemed to depend on the level of education (see Table 5). Thus knowledge emerged as a relevant predictor only in connection with respondents with a college-preparatory education. The specific effect of scientific knowledge here was to make it less likely that a respondent would prefer one transformation method to another. Among those without a preparatory education there was no specific effect of this kind: knowledge was significantly associated with neither a raised nor lowered probability that a respondent would differentiate. However, it must be emphasised that, even among those with an education, knowledge explained very little of the variation and did not improve the classification rate. These relatively unspectacular results may in part be explained, perhaps, by the fact that we operationalised differentiation in terms of variation in acceptance, rather than in terms of variation in the ability to tell the two methods apart.
Effects of knowledge on the likelihood of differentiating between transgenic and cisgenic cereals. Participants are grouped according to their educational status.
The model correctly predicts 54.5% against 54.5% in the baseline. Hosmer and Lemeshow = .964, Nagelkerke R2 = 0.021.
The model correctly predicts 64.6% against 64.6% in the baseline. Hosmer and Lemeshow = .825, Nagelkerke R2 = 0.001.
Unsurprisingly, the link between education and scientific knowledge was evident throughout our investigations, in that a preparatory education improved the likelihood of a higher knowledge score. This is not surprising given that we measured knowledge through answers to a number of questions on topics typically taught in high schools. This link notwithstanding, the inclusion of both parameters – i.e. a higher level of education and scientific knowledge – added to the explanatory power of regression models representing the odds of a respondent differentiating between medical and agricultural applications. The fact that scientific knowledge affected the likelihood of a respondent differentiating between transgenic and cisgenic cereals only when subjects had a preparatory education is interesting: it indicates that education, in this case, not only increases a person’s mean level of knowledge, but also makes knowledge differences more significant – possibly because, within the educational system, one is encouraged to integrate scholarly knowledge in the process of opinion formation.
6. Discussion
A positive correlation between scientific knowledge and attitudes of acceptance was found for all of the investigated applications and for both methods of transformation. However, the strength of this correlation was relatively weak, and it varied from one application/method to another. Invariably, the mean knowledge score of proponents exceeded that of the undecided group. The mean knowledge score of opponents, on the other hand, was significantly higher than that of the undecided group only when it came to modified animal feed and bread made from cisgenic cereals. The mean knowledge score of proponents was similar to that of the opponents in the case of cisgenic cereals alone. It was significantly higher in all other cases.
One should be careful, however, not to over-interpret the implications of these relationships. Because the observed differences in scientific knowledge were between subjects rather than successive measures of attitudinal change in one and the same subject, no conclusions can be drawn as to whether a person is likely to become more positive if he or she is provided with additional information. Accordingly, our findings do not contribute to the debate about whether levels of public acceptance can be raised by providing people with additional information, or whether additional information simply makes people form opinions that are more definite. Rather, this study adds a new, more fine-grained perspective on the impact of people’s current knowledge on acceptance by examining whether knowledgeable people differentiate in a distinctive manner, using distinctive criteria.
Like most previous researchers, we framed our enquiry in terms of indirect objective knowledge of biological facts. We do recognise, however, that other types of knowledge exist, and that these may be equally, or more, important modulators of public acceptance. It seems plausible, however, that objective knowledge of basic biology and genetics makes it easier for people to tell apart different applications and different methods of transformation. Similarly, we would expect this type of knowledge to ease comparisons of the specific risks, benefits and moral considerations that apply to each application/method individually.
Interestingly, none of this meant that knowledgeable people were generally more likely to have a differentiated view on the acceptability of gene technology. Rather our results demonstrate that scientific knowledge renders only certain types of differentiation more common. More specifically, knowledge made people more likely to differentiate between medical and agricultural applications. It did not increase the likelihood of a respondent differentiating the modification of foods and animal feed. Moreover, when the acceptability of transgenic and cisgenic cereals was compared, knowledge slightly decreased the odds of differentiation, though only among those with a college-preparatory education.
At least three explanations of the fact that scientific knowledge makes people more likely to prefer medical over agricultural applications of gene technology deserve consideration. The first relates to the contexts in which people learn about biology and basic genetics. Since one would typically acquire such knowledge through an educational system that itself celebrates the achievements of science, it is possible that those who have received extensive technical and scientific training will thereby be socialised to value the advances of modern medicine. If learning about biology and genetics goes hand in hand with learning that science has revolutionised the treatment of illness, it seems only reasonable to speculate that those who achieved a higher knowledge score would, on the whole, have been inclined to be positive about the prospect of further medical developments. Historically, of course, a similar argument would apply to scientific achievements in agriculture. However, more recently the association of modern cultivation methods and husbandry as sources of biodiversity depletion and problems with animal welfare has perhaps directed the educational focus towards environmental and animal care issues, making the positive association less pronounced.
Even if scientific education, through its values, encourages students to respond positively to medical developments, institutional socialisation is not the only factor at work here. Another candidate explanation of the fact that scientific knowledge makes people more likely to prefer medical over agricultural applications of gene technology points to a selection bias in the educational system, which favours those people already attuned to a scientific value set. That is, people who are already imbued with scientific values are easier to educate, and are thus more likely to receive, and complete, scientific training. They are also more likely to seek out scientific training themselves, because the educational system conforms to their beliefs. And, obviously, people with scientific training will tend to achieve more highly in tests of scientific literacy.
A third explanation is that scientific knowledge makes it easier for people to compare the risks and benefits of each type of application, thus enabling them to make more specific value judgements. Scientific knowledge may, for example, increase the preference for medical usage by making people more aware that the benefits provided by agricultural applications have so far mostly been economic. To the public, such benefits are less important than societal benefits (Bonny, 2003; Gaskell et al., 2003). In a similar way, people with concerns about the environmental impact of gene technology might embrace medical applications more willingly if a scientific education enables them to see that medical applications do not necessitate the introduction of genetically modified organisms into the environment. Such considerations may further explain why we did not find scientific knowledge affecting the acceptability of GM foods and animal feed differently, since many of the risks associated with these applications are shared or harder to separate. After all, people who are concerned about eating GM foods might not be any less concerned about meat and dairy products from animals fed with modified feed – regardless of their knowledge of basic biology and genetics.
Although divergent opinions on medical and agricultural applications are in all likelihood connected with the usefulness of each application, we would not expect people to distinguish between cisgenic and transgenic cereals primarily on the basis that the former provide greater benefits. On the basis of a study of consumer attitudes to cheese production, Frewer suggests that consumers view production methods and the benefits arising from them as independent of each other, and hence evaluate them completely separately (Frewer et al., 1997b). Rather than being a response to a perception of benefit, the preference for cisgenic crops seems to rest primarily on the belief that cisgenesis is a more natural method of transformation – in the specific sense that it incorporates only native genes or genes derived from closely related organisms (Mielby et al., submitted). Conversely, many people seem to deplore transgenic transformation because they see the joining of genes from different species as too unnatural, even when benefits accrue (Knight, 2009; Shaw, 2002).
The positive correlation between scientific knowledge and the likelihood of a respondent taking a uniform view of transgenic and cisgenic cereals may therefore indicate that concerns about unnaturalness become less important in the presence of scientific knowledge. One might well imagine that a balancing of benefits against risks would be more liable to grab the attention of those with some technical and scientific training than, say, the issue of unnaturalness, which they might regard as “irrational and emotional.” To the developers of cisgenic crops this creates something of a dilemma. On the one hand, scientific knowledge ought to make it easier for people to tell cisgenic and transgenic cereals apart; on the other, the difference in naturalness tends to appear less significant in the presence of scientific knowledge. The result may be that those most likely to understand the difference between transgenesis and cisgenesis are least likely to care.
Another possible reason why knowledge makes people less likely to differentiate between cisgenic and transgenic cereals relates specifically to the use of bacterial genes to create transgenic plants. For, clearly, scientific education could make one’s initial response to transgenesis somewhat more positive simply by leading one away from the popular misconception that all bacteria are pathogenic (Pfister et al., 2000). A similar effect would not be expected in the case of cisgenic cereals, because genes from closely related organisms are less likely to give rise to the same negative associations. Again, it has been argued that cisgenic cereals are more like their conventionally bred cousins on the grounds that the process of cisgenesis does not involve genes not already available to conventional breeders (Schouten et al., 2006). If – intuitively, at least – cisgenic cereals are more familiar to the public, knowledge as a counter-measure of neophobia may in this case be of limited importance.
Let us summarise these brief considerations. We found that people who achieved a higher knowledge score were more likely to condition their acceptance of a GM application on its purpose, but slightly less likely to make a distinction between methods. One interpretation of this is that with differences in the perceived balance of risk and benefit, scientific literacy plays a greater role, whereas perceived differences in naturalness are more important in the absence of a basic knowledge of biology and genetics. In a similar vein, Sturgis et al. (2005) has argued that a scientifically better informed public would be less likely to agree that “changing genes should be forbidden as tampering with nature” and more likely to approve of medical applications – provided they are aimed at the treatment of serious illnesses.
Finally, it is an important implication of the findings presented here that conclusions concerning the effect of knowledge on acceptance cannot be generalised wholesale from one application, or method, to others. Equally importantly, it can be seen that any efforts to develop new generations of GM crops which either provide greater societal benefits or are more “natural” may affect opinion differently, depending on individual levels of knowledge and educational background.
Footnotes
Acknowledgements
Financial support from the Danish Food Industry Agency is gratefully acknowledged.
Author biographies
Henrik Mielby has a background in political science with a strong interest in science and technology studies, bioethics and the sociology of food. In 2011 he obtained his PhD for a thesis on public attitudes to genetically modified crops. He currently serves as a part-time lecturer in Sociology, and Philosophy of Science at the University of Copenhagen.
Peter Sandøe is professor in bioethics at the University of Copenhagen. He is the director of the Danish Centre for Bioethics and Risk Assessment (CeBRA), an interdisciplinary and inter-institutional research centre founded in January 2000. Since 1990 the major part of his research has been within bioethics with particular emphasis on ethical issues related to animals, biotechnology and food production. He is committed to interdisciplinary work combining perspectives from natural science, social sciences and philosophy.
Jesper Lassen is associate professor in sociology at the Institute of Food and Resource Economics, University of Copenhagen. His main research interest is the study of the interface between science, technology and society with particular focus on issues in relation to food and agriculture. His research and publications focus on public perceptions of risks (e.g. genetic technologies, zoonoses, pesticides etc.), conflicts between lay and expert perceptions, political processes in relation to food and agriculture, and participatory methods.