Poverty of Stimulus in Ethics: Some Criticisms

Kyle Swan has an interesting post on PEA Soup entitled "Poverty of the Moral Stimulus." In it, he presents the case for a "poverty of stimulus" argument in ethics that is similar to the one used in linguistics. Kyle sees this as a natural extension of the linguistic argument, and I'm inclined to agree. Nativism is, after all, a Platonic position, and Plato was a nativist about ethics. Here is the primary evidence he presents for the use of a poverty of stimulus argument in ethics:

Just as empiricist accounts of learning are inadequate to explain children’s linguistic competence, so also are they inadequate to explain children’s moral competence (Dwyer 1999; Harman 2000). For example, children exhibit the capacity to make genuinely moral judgments since they are sensitive to the moral/conventional distinction in their use of terms like “wrong”. It has been shown that children as young as 3 years old will regard moral transgressions as more serious than conventional transgressions and regard the former as more universal, independent of any authority figure, and closely correlated with considerations of harm (Smetana and Braeges 1990; Smetana 1993; Nichols 2002a). So if children are told that some households have no rule against, say, running in the house or eating dessert before dinner, then even if their parents do have rules against these practices, most children will judge that it is not wrong for children who live in the households where the rules do not exist to do these things. However, children will judge that hurting another is wrong, even if they are told that some parents have no rule against hurting others. They make similar claims across different cultures. And this recognition emerges early in the developmental psychology of children – earlier than we should expect if children learn it only from explicit moral instruction from caregivers.

And here is the argument:

Absent a detailed account of how children extrapolate distinctly moral rules from the barrage of parental imperatives and evaluations, the appeal to explicit moral instruction will not provide anything like a satisfactory explanation of the emergence of mature moral competence. What we have here is a set of complex, articulated abilities that (i) emerge over time in an environment that is impoverished with respect to the content and scope of their mature manifestations, and (ii) appear to develop across the species (Dwyer 1999, 173).

Anyone familiar with the nativism in linguistics will notice some glaring problems with both the evidence and the argument presented here. First of all, the analogy to the poverty of stimulus argument in linguistics can't be meant to go very far. The Chomskyan version not only argues that children don't receive enough input to account for the level of sophistication in their early knowledge of grammatical principles, but also argues that no amount of input could account for this. I don't think this is the claim that ethicists want to make. Furthermore, the types of linguistic structures proposed to account for early grammatical knowledge, which generally consist of rules for the combination and movement of linguistic objects, are probably not the sorts of innate moral knowledge that ethicists would want to propose. Still, there are other, less Chomskyan versions of the poverty of stimulus argument, and not everyone buys the Chomskyan architecture either¹. So, there may still be parallels between the two domains.

Still, even if we assume that ethicists don't mean that ethical knowledge cannot be learned, no matter what the input, the evidence and argument still suffer from several major flaws. Here are some of them:

The first, and most glaring, is that it seems to be a simple argument from incredulity, as the first sentence in Dwyer's argument indicates. It may be the case that there currently are no accounts of how children learn certain ethical facts like the moral/convention distinction, but this does not mean that they aren't learned.

One of the biggest problems with the argument is that it seems to display an ignorance of human learning, particularly learning during development. This is evident in the focus on "explicit instruction" in both Dwyer's formulation of the argument and Swan's post. A quick perusal of the literature on the poverty of stimulus argument in linguistics² will make it clear that, if children develop a linguistic competence through learning from positive input, it is implicitly, rather than through implicit instruction. In fact, relative to the overall amount of learning that children do in their first years of life, the amount of learning they do through explicit instruction is pretty small. It stands to reason that most of their moral knowledge would be learned implicitly, trhough observation and imitation (mimesis), as is the case for most complex behavioral knowledge. I see no reason to believe that something like the moral/conventional distinction couldn't be learned this way, which leads me to the next problem.

The primary evidence for the poverty of stimulus in ethics seems to rest on Dwyer's and Swan's belief that the positive input is insufficient to account for the data, but neither presents a systematic study of this input. In linguistics, systematic studies of "motherese," or the speech caregiver's use when talking to young children, have usually been taken as one of the best pieces of evidence for the poverty of stimulus (though see the references in footnote 2 for counterarguments). As long as we don't adopt the Chomskyan position that no amount of input will be sufficient to account for a certain type of knowledge, we will need to have a thorough description of the input to know whether it is insufficient. I, personally, can come up with all sorts of possible types of input that might account for the moral/conventional distinction data that Swan presents. For instance, caregivers may talk about hurting others differently than they talk about conventional rules. Even subtle differences in attitudes toward moral vs. conventional rules may produce subtle distinctions in the knowledge of a child over the first 3 years of life. Furthermore, it's likely that caregivers' behaviors reflect these distinctions, and that children, over time, are able to pick up on these differences, even if they are not able to articulate them. None of this requires explicit instruction, and none of it can be ruled out without a comprehensive study of the input.

Another problem, which also comes up in the nativism debate lin linguistics, relates to whether what might be innate, if anything, about moral knowledge is specific to moral knowledge (modular) or instead consists of more domain-general cognitive capacities. It may be that moral knowledge bootstraps onto larger cognitive mechanisms, and that moral rules, and parental methods for conveying these rules, have evolved culturally (or biologically) to take advantage of these domain-general abilities ³.

It may be that certain moral rules, e.g. reciprocity⁴, are in fact innate. These innate rules would likely be treated as universal (though behavior would not be consistent with this, of course). However, the existence of these innate moral rules would not imply that moral knowledge, such as a capacity to distinguish between non-innate moral rules and conventional rules, is innate. The data presented by Dwyer and Swan is not inconsistent with this interpretation.

All in all, I think the use of the poverty of stimulus argument in ethics is interesting, but premature. A great deal of research, and a better formulation, are needed before empiricists take such an argument seriously. Right now, I see no reason to think that an empiricist account of ethical development can't account for the data. Each of the problems listed above (with the possible exception of the last one) has been mirrored in the nativism debate in linguistics. For that reason, I think ethicists who want to make poverty of stimulus arguments in their field would do well to familiarize themselves with the linguistics argument and the trials it has gone through over the past few decades. The analogy to the linguistics version of poverty of stimulus may not be meant to go very far, but it can still be instructive in this way.

¹ Ray Jackendoff's version of innate linguistic structure, for instance, differs markedly from Chomsky's. You can read about it in his excellent book, Foundations of Language: Brain, Meaning, Grammar, Evolution.
² For instance, the work of Michael Tomasello on language acquisition, references to which can be found here. His book The Culturual Origins of Human Cognition presents the arguments in detail, though he has produced even more convincing evidence since its publication.
³ Zuidema, for instance, makes this argument in linguistics, in this paper.
⁴ See e.g., The Evolution of the Golden Rule, Gretchen Vogel, Science Feb 20 2004: 1128-1131.