Wednesday, March 08, 2006

Do People Actually Use the "Strict Father" and "Nurturant Parent" Models?

I've talked about Lakoff on this blog a lot. Way, way, way too much. But in talking about him incessantly, I've focused on conceptual metaphor theory in general, and ignored one of his more widely known, and controversial claims in his political writings: the existence, and use, of the "Strict Father" and "Nurturant Parent" metaphorical models. These models get a lot of talk elsewhere in the blogosphere and popular press, though, and while the status of these models, and conceptual metaphor theory, is independent of the status of framing analysis in general, people often seem to form their opinion of framing analysis based on their opinion (often negative) of the two metaphors. So, I've been looking for an excuse to talk about the metaphorical models themselves, and the recent publication of a paper on them has given me just such an excuse. So, here goes.

When evaluating the claims of cognitive linguists in the conceptual metaphor tradition, it's not always clear how you should go about it. There are two main problems. First, there's not a whole lot of empirical evidence for most, if not all of the bigger claims, which makes it difficult to know exactly what you should be evaluating. Second, from the start of the conceptual metaphor paradigm, there have been serious reservations about its methodologies1, among psychologists and linguists, which makes it difficult to know just how to take the evidence that does exist. When evaluating Lakoff's political writings, it's easy to be completely overwhelmed by both problems. For his larger claims, such as that the "Strict Father" and "Nurturant Parent" metaphorical models serve as prototypes, or idealized cognitive models, Lakoff provides very little evidence, and what evidence he does provide is in the form of a few linguistic examples that have been analyzed by one person, Lakoff himself. This raises two problems. The first is the common problem of circularity in linguistic analysis. The cognitive linguistic argument says, in essence, that metaphors in language are evidence of metaphors in thought, and this is evidenced by the existence of metaphors in language (see Murphy, 1996, citation in footnote 1). The second problem, raised by Sandra and Rice (see footnote 1), is that it's not clear whether the metaphors that cognitive linguists discover through their linguistic analysis are in the minds of the producers of the utterances in which the metaphors were found, or in the minds of the linguists who found them.

The simple answer to all of these problems would be to run carefully designed experiments to test specific predictions of conceptual metaphor hypotheses. And a very, very few such experiments have been run, mostly in the lab of one psychologist, and in one domain: the domain of time-space metaphors2. And even these are open to very plausible alternative explanations. But as of yet, there is absolutely no experimental evidence for the existence, much less the use, of metaphorical models like the "Strict Father" and Nurturant Parent" models, and aside from the tiny amount of linguistic evidence that Lakoff presents in Moral Politics, there's no non-experimental evidence that they exist, either. Lakoff, of course, is not worried by this. In that book, he readily notes that the existence of metaphorical models "does not have the degree of confirmation that one would expect of more mature theories" (p. 158, or, I would say, preliminary working hypotheses!), and goes on to blame this not on the theory itself, but on the available methodologies, writing:
No experimental paradigms of the complexity needed to test this hypothesis now exist.
and
Survey research has not yet developed an adequate methodology to test for the presence of complex metaphorical cognitive models such as these.
And of course, he's right: there are no experimental paradigms or survey research methods of sufficient complexity to test his hypotheses. The reason no such paradigms or methods exist is that we haven't yet learned how to test hypotheses that make no predictions! Fortunately for us, though, one researcher didn't see the lack of straightforward predictions from the hypothesizing of metaphorical models as a problem, and actually went out to look for the models in actual speech. The study is primarily in the form of linguistic analysis, which means that it could potentially suffer from the problems mentioned above, but the researcher, Alan Cienki, used techniques designed to avoid them. But I'll get to all of that in a moment. First, a refresher course on the Strict Father and Nurturant Parent models (SF and NP respectively, from here on out).

The Strict Father and Nurturant Parent Metaphors

Good summaries of the two metaphors can be found at the Rockridge Institute, Lakoff's think tank. The two metaphors are both instances of the larger "Nation as a Family" metaphor. Here is how the SF model is described at Rockridge:

The father's job is to protect and support the family. Children are to respect and obey him. The father's moral duty is to teach his children right from wrong, with punishment that is typically physical and can be painful when they do wrong. It is assumed that parental discipline in childhood is required to develop the internal discipline that adults will need in order to be moral and to succeed. Morality and success are linked through discipline. This focus on discipline is seen as a form of love—"tough love."

The mother is in the background, not strong enough to protect and support the family or fully discipline the children on her own. Her job is to uphold the authority of the father and to care for and comfort the children. As a "mommy," she tends to be overly soft-hearted and might well coddle or spoil the child. The father must make sure this does not happen, lest the children become weak and dependent.

Competition is necessary for discipline. Children are to become self-reliant through discipline and the pursuit of self-interest. Those who succeed as adults are the good (moral) people and parents are not to "meddle" in their lives. Those children who remain dependent—who were spoiled, overly willful, or recalcitrant—undergo further discipline or are turned out to face the discipline of the outside world.

When everyone is acting morally and responsibly, seeking their own self-interest in a self-disciplined fashion, everyone benefits. Thus, instilling morality and discipline in your children is also acting for the good of society as a whole.

The political implications of this metaphor, from the same Rockridge article, are as follows:
  • Protect the country and its interests in a dangerous world by maximizing military and political strength;
  • Promote unimpeded competitive economic activity so that both the disciplined moral people and the undisciplined immoral ones are able to receive what they each deserve, based on their own choices;
  • Maintain order and discipline, through severe enforcement of the rules if necessary
Examples of metaphors within the SF model, from Moral Politics, include (with examples given in the book) MORALITY IS STRENGTH ("having the moral fiber to resist evil"), BEING GOOD IS BEING UPRIGHT ("an upstanding citizen"), and MORALITY IS PURITY (undesired behaviors are seen as "unclean").

The NP model is described at Rockridge as:
In the Nurturant Parent family, it is assumed that the world is basically good. And, however dangerous and difficult the world may be at present, it can be made better, and it is your responsibility to help make it better. Correspondingly, children are born good, and parents can make them better, and it is their responsibility to do so. Both parents (if there are two) are responsible for running the household and raising the children, although they may divide their activities. The parents' job is to be responsive to their children, nurture them, and raise their children to nurture others. Nurturance requires empathy and responsibility.
The political extensions for this model listed at Rockridge are too numerous to include in this post, but you can check out the link to learn about them. Central metaphors in this model (again from Moral Politics) include MORALITY IS EMPATHY, MORAL ACTION IS NURTURANCE, and MORGROWTHOTH IS PHYSICAL GROWTH.

It is very important to understand, both when trying to evaluate the models themselves, and for understanding the study I'm going to describe in a minute, that neither the SF or NP models are supposed to exist, in their entirety, in any single person's head (except Lakoff's). Instead, they are meant as prototypes, around which liberal (in the case of the NP model) and conservative (in the case of the SF model) world-views cluster in a family resemblance fashion. Liberals may represent some features of the NP model, and some of the SF, but overall they will tend to represent more features of the NP model than the SF model. The reverse is true of conservatives. This is why the two models may seem a bit like caricatures: they are, in fact, meant to be caricatures of a sort.

The Study

Cienki's study, published last year in the journal Cognitive Linguistics, went looking for the use of these two models in the speech of the two candidates in the 2000 presidential election, using their three televised debates as a corpus3. He gives four reasons for using the debates as data for his analysis (pp. 283-284):
  1. "They not only consist of spoken data from Bush and Gore on the same subjects, they provide approximately the same amount of linguistic output from each."
  2. "The lexical density [the number of different words] per debate did not differ statistically between Bush and Gore."
  3. "The debates took place late in the campaign. By this time the positions of the candidates had become more fixed, and so the debates can be seen as fair representations of the candidates' views."
  4. "They consist of a variety of speech genres, from the more formal and serious to the occasional joking, and from statements which were more prepared and often re-used to responses which were more spontaneous."
The first stage of the study consisted in looking for instances of specific SF and NP metaphors, which were taken from Lakoff's own list in Moral Politics (I'll put a list of the metaphors, from the appendix of the paper, in the next post). Here are examples of metaphors from the debates, as listed in the paper (p. 287):
IMMORALITY IS IMPURITY. Gore: "By giving parents the tools to protect their children against culturpollutionion."

ACTION IS THE NURTURANCE OF SOCIAL TIES. Bush: "And that's a case where we need to use our influence to have countries in Africa come together and help deal with the situation.
In order to overcome the problem in linguistic analysis, described above, of finding only metaphors that exist in the linguist's head, rather than the speaker's, two researchers coded the debates, and their agreement was high (near 100%). Here is Cienki's description of the results (p. 287):
What is noteworthy is that the central conceptual metaphors of the SF and NP models were expressed so seldom through metaphorical language in the debates. In the approximately 41,000-word corpus, there were only 48 expressions in total which directly expressed any of the 43 metaphors in the Appendix [see the next post]. The paucity of these metaphoric expressions contributed to their salience when they did occur, which is likely a factor in the very high agreement reached between coders.
In all, Bush used 29 expressions of the 43 metaphors, seven of which were from the NP model, and 22 from the SF model. Gore used a total of 19 expressions of the metaphors, 14 of which were from the NP model, and 5 of which were from the SF model. So, it appears that while the models' metaphors are rarely used, they are used in a pattern consistent with liberals (in this case, Gore) being more inclined to the NP model, and conservatives (Bush) being more inclined to the SF model.

Now, a skeptic like me would take the extreme paucity of metaphorical expressions from the two models as evidence that they are, in fact, not the models that shape the world views of liberals and conservatives. I certainly can't imagine how the data above could be taken as evidence for their existence. But Cienki doesn't want to let them go, writing that "the explanatory power provided by the models and their comprehensivenedeterter one from jumping to such conclusions." He argues that metaphors can exist at different levels of "schematicity," by which he means "the range of source or target domain concepthataht are consistent with a given metaphorical mapping" (p. 288). Put differently, metaphors exist at different levels of abstraction. Metaphors also differ in their productivity, by which Cienki means "the number of roughly synonymous expressions by which the metaphorical mapping is manifested, that is, how successful a schema is in manifesting its generality in linguistic expressions" (p. 289). In other words, how many linguistic expressions you can get from a particular metaphor. Cienki argues that the reason behind the dearth of metaphorical expressions from the SF and NP models may be that the models are high in schematicity, but low in productivity. That means they subsume a lot of other metaphors, but they don't allow for the production of many particular metaphorical expressions directly from the models themselves. In short, if the first prediction you come up with from a metaphorical models hypothesis doesn't work out, come up with a new one.

In order to test this new prediction, the second part of the study consisted of a search for "expressions in the corpus which do not directly reflect the central SF or NP metaphors, but which logically follow from the SF or NP model, that is, they may be entailed by them" (p. 289). The new prediction is thus that we'll find a lot of entailments, both metaphorical and non-metaphorical, and that this will indicate the presence of the SF and NP metaphors themselves in the minds of the speakers. If you're like me, you're probably thinking that we're getting further away from any possibility of non-circularity, by undertaking this type of analysis, but the results are worth looking at. Here are examples of the entailments (both metaphorical and non-metaphorical) found in the debates (pp. 290-291):
  1. SF metaphorical entailment. Bush: "I think it's important for NATO to be strong and confident."
  2. NP metaphorical entailment. Gore: "So I want proceedced carefully to cover more people."
  3. SF non-metaphorical entailment. Bush: "We need to send five percent [of tax revenues] back to you that pay the bills.
  4. NP non-metaphorical entailment. Bush: "There needs to be more community health clinics where the poor can go get health care."
The explanations of theexamplesles are as follows (from p. 291). For example 1., "The phrase supports the SF ideal of individual self-reliance, and in this case, the organization is metaphorically an individual person." For example 2, "This is consistent with the NP priority of nurturance, and also of fair distribution." Example 3 "reflects the emphasis on the importance of individual control of one's well-being; managing one's money is thus a matter of individual responsibility, and taxation is viewed as government interference." And example 4 "reflects the concern for others typical of the NP priorities of empathy and fair distribution." In all, very few metaphorical entailments were found. Bush used a total of 48 metaphorical entailments, 27 from the SF model, and 21 from the NP model, while Gore used a total of 45, 12 from the SF model, and 33 from the NP model. As with the direct metaphorical expressions of the model, both candidates used more metaphorical entailments from the model consistent with their political ideology than with their opponent's, though both used some of each, and Bush's totals from both models were pretty similar. Non-metaphorical entailments were a little more frequent. Bush used a total of 469 non-metaphorical entailment expressions, 269 from the SF model and 200 from the NP model, while Gore used 276 total, 68 from the SF model and 208 from the NP model. Once again, both used entailments from both models, but used more from the model consistent with their political ideology. However, Bush's ratio of SF to NP entailments was small in this case as well.

Again, if you're a skeptic like me, you'll note two things about this data. First, in a 41,000 word corpus, there were, again, very few metaphorical entailments, and many more non-metaphorical ones than metaphorical ones. This implies two things to me. One, it looks like people aren't using metaphors with the frequency that conceptual metaphor theory would have us believe, and two, it doesn't look like conceptual metaphors are structuring the way people talk, and given the linguistic arguments, think about political issues. Second, we again run into the inherent problem of circularity in using the non-metaphorical expressions, hese are simply expressions of political ideology, and Lakoff's analysis came up with the metaphors from similar expressions, meaning, in the end, that the argument from non-metaphorical expressions is that we see the positions in people's speech because the metaphorical models exist, and we know the metaphorical models exist because we see the positions in people's speech. So I'm not sure that it's actually possible to interpret this data, and I'm pretty damn sure that it's not possible to interpret it as supporting the existence of the SF and NP models, or conceptual metaphor theory in general.

In discussing these results, Cienki notes that neither candidate seems to be "very prototypical adherent[s] to either the SF or NP model" (p. 293). In fact, one of the interesting features of these results, in my mind, is the fact that Bush used both models frequently (his ratio of SF to NP was around 1.3:1 for both metaphorical and non-metaphorical entailments), whereas Gore's was at least 3:1 for both). This seems to be inconsistent with one of Lakoff's more practical claims, namely that conservatives have done a better job of framing issues in their language, and that, in fact, Bush is a master of this. Here he appears to be framing the issues in both conservative and political language (according to the SF and NP models), while Gore is much more consistent.

The final part of the debate is meant to get past this circularity. Instead of looking at speech, it looks at gestures. The argument is, if we find the use of the metaphors in gestures, then we have evidence that lies outside of the linguistic circle. Nevermind the fact that in doing so we create a new circle, and thus have cognitive linguists twirling hoops on both arms. At least the evidence isn't linguistic! The methods for the gestural analysis are kind of complex, and I don't want to try to explain them in detail here (this damn post is long enough as it is). Instead, I'll just give you a few examples of gestures described in the paper, and then the overall results of the gestural analysis. Here are two examples (from pp. 298-301; the gestures are described in bold below the statements during which they occurred, and Cienki's explanations are listed in italics beneath those descriptions; {} are placed around the words that occurred "during the main stroke of the gestures"):
Bush: "If we're a humble nation, but {strong}, they'll welcome us in.
2 hands barrier, fingers spread, bounce forward from self
.
Consistent with source domains of the SF metaphors... (such as STRENGTH, WHOLENESS, and BEING UPRIGHT).
Gore: "The power of example is America's {greatest power} in the world."
2 hands palms up, cupped.
No metaphorical description given.
The results of the analysis showed that Bush used many more "metaphoric" gestures than Gore (43 and 13, respectively), and that each candidate used more gestures when using expressions entailed by the metaphorical model consistent with their political ideology. I'll be damned if I know what that means, and I doubt anyone does, but Cienki takes it as evidence of the existence of the two models.

Overall, I think you have to give Cienki a pat on the back for even attempting something like this. Elsewhere (e.g., in this manuscript), he continues to search for methods to test for the existence and use of political metaphorical models. Unfortunately, the paper described in this post contains the only data published thusfar. But you have to respect a guy who attempts to test a theory that doesn't provide any testable predictions, even if that means that when one prediction doesn't pan out, he has to come up with new ones (which may even be inconsistent with the first prediction) to continue testing the theory. Of course, if you're a fan of the SF and NP models, the data in this paper cannot be encouraging. The models were rarely used, metaphorical entailments were rarely used, and the non-metaphorical entailments that were found in the speech of the two candidates can only be used to argue for the existence and use of the metaphorical models through circular reasoning. The gestural analysis, while interesting, doesn't seem to say much of anything. Plus, as I said above, Bush's use of entailments from both models with near equal frequency argues against one of Lakoff's central claims about "what conservatives know that liberals don't."

My sincere hope is that, given the complete lack of evidence for the SF and NP models, people stop trying to use them for now, and instead focus on integrating the basics of framing analysis, with insights from research on knowledge representation, into their political speech. I know Lakoff won't stop peddling them, but a lack of evidence has never stopped him (as his quotes above indicate). We're simply going to have to ignore him for a while, until he, Cienki, or someone else, comes up with some data thunequivocallybly supports the existence and importance of these two metaphorical models. I know I'm not holding my breath. I don't think you should be either.

1E.g., Murphy, G. 1996. "On metaphorical representation", Cognition 60: 173-204; Sandra, D., & Rice, S. (1995). Network analyses of prepositional meaning: Mirroring whose mind—the linguist’s or the language user's? Cognitive Linguistics, 6, 89-130.
2E.g., Boroditsky, L. (2000). Metaphoric structuring: Understanding time through spatial metaphors. Cognition, 75, 1-28.
3Cienki, A. (2005). Metaphor in the "Strict Father" and "Nurturant Parent" cognitive models: Theoretical issues raised in an empirical study. Cognitive Linguistics, 16(2), 279-312.

2 comments:

Bora Zivkovic said...

I was busy as hell when you posted this so I just skimmed it. I still intend to give it a more thorough read and to comment on it on my blog in a couple of days.

Anonymous said...

This is a very nice critique, and a wonderful real-world example of basic and subtle problems in the business of doing science on consciousness and cognition. I think there quite possibly might be good science to be found in studying metaphor, but I don't think we're there yet.

More fundamentally, I find I am unclear about what kind of knowledge comes from linguistic analysis in general. Neuroscience, beginning with the incredibly low-resolution technology of FMRI (compared to the number of neurons) hopes to one day posit an explanation (model) of consciousness that is entailed by basic physical science. By contrast, linguistic analysis is essentially backwards engineering from observed variables to construct a model that explains what we see in how people actually speak. But a model is just a model, a complex mental construct, even if it actually is testable. If we assume that there is a way that language and cognition actually *do* work, then we have to verify this model against what seems to be known about the rest of reality, including everything that is known about the relationships between the brain and cognition (so-called "neural correlates of consciousness").

In short, I often find myself wondering what precisely the relationship of this sort of work to reality is. If the point of building such models is to structure our thought to help us see patterns, that's perfectly useful, but then we need to be careful about reificiation.