Monday, April 23, 2007

S 59 Debate: More on the Direness of the Empirical Anti-smacking Literature

This post follows up a post from last week. Read that first.

1. Some researchers try a little harder than Gershoff to be open-minded about "child effects" but the deep unwillingness to take that sort of causal structure seriously is evident in many of those cases too. Reflecting on criticism of Gershoff from roughly our direction, George Holden (Perspectives on the Effects of Corporal Punishment: Comment on Gershoff (2002)) writes:
"For children, initial child misbehavior is sometimes used as an alternative hypothesis—parents spank because they have a difficult child (Larzelere, 2000). But where did the difficultness originate? Is it a reflection of inappropriate parenting practices at an early age, the child’s temperament, or perhaps a “third” variable? Do parents who spank engage in other behaviors (e.g., yelling) that contribute to the negative effects? Over time, what other child-rearing practices do parents use that may exacerbate, maintain, or even ameliorate possible prior effects of corporal punishment?" (Our italics)
Indeed, these are very interesting questions that are worthy of study. But surely we can agree that, while we're conducting those studies or waiting for appropriate models and methodologies to emerge, we shouldn't bury the important hypothesis (or hypothesis schema) that holds that, while there are many later factors that affect child behaviors, there's also a relatively robust behavioral dispositional vector coming from the child to which parents respond, and with which they may see themselves as wrestling for years, using a wide range of different techniques and strategies.
(Who knows how many dimensions that behavioral vector has? "Temperament" like "intelligence" ultimately just names a research problem.)

Moreover, we can ask the "where it came from?" question about anything – intelligence, sexual identity, etc. – but for many purposes it just does make sense to treat these sorts of macro-features as inputs to adult-decision-making, e.g.,

  • "If you have a very intelligent child then do x"
  • "If you have a gender-variant (-incongruent) child then do x"
and so on. Indeed one suspects that if there ever is a good empirical anti-smacking argument (albeit one that would seem to be more advice-giving than criminalizing) it'll probably have the form:
  • "Even if you have a highly-naughty child then don't bother smacking them. Save yourself and them the bother. You can't change them or discipline them 'straight', you'll just make them hate you (more) if you do. You just have to ride it out."
But that argument requires taking child naughtiness seriously, controlling for levels of it (whatever it is), and so on. Until the child development field does this – a symptom of which would be formulating recommendations like that described – then it's still back in the Dark Ages, and in fact has almost nothing to contribute to discussions of better and worse ways to raise children.

2. Murray Straus's 2000 paper "Corporal Punishment of Children and Adult Depression and Suicidal Ideation" [Chap 4 in McCord (ed.) Coercion and Punishment in Long Term Perspective (Cambridge Univ. Press)] positively correlates amounts of adult depression and adult suicidal thoughts with amount of remembered/reported corporal punishment as a teenager, after controlling for a number of obvious factors such as adult socio-economic status, adult marital violence, heavy drinking and so on.
Naturally, however, one wonders, among other things, whether there aren't underlying character traits – e.g., self-destructiveness, volatility, bipolarity, etc. – that cause both the behaviors that elicit parental slaps (say) in teenage years, and higher rates of adult depression, etc..
Straus tries to face up to this question by first conceding a version of our general complaint as it applies to his earlier studies correlating corporal punishment with increased aggression and crime (we'll issue a caveat in a moment):
"[T]he causal direction of the [earlier] findings is ambiguous. Rather than corporal punishment causing the child to become aggressive or delinquent, it could be the other way around. Parents typically spank a child because he or she hit another child or hits them. Thus, it is just as plausible to say that the child's aggression or other misbehavior caused the spanking as it is to argue that spanking teaches aggression." (first italics mine)
And how is Straus's new study any different?
"The problem of which is cause and which is effect may be less serious in the present case because the hypothesized outcome (depression and thinking of suicide) is probably something for which parents do not usually hit a child. Nevertheless a longitudinal or experimental study is needed to deal with the question of causal direction." (my italics)
I agree that it's a good bet that a parent who slapped her adolescent's face in 1990, say, wasn't even partially caused to do so by the depression that the adolescent will experience in 2007! And while it might be true that the 1990 parent takes herself to be acting in the long-term best interests of her child, it's surely unlikely that she made any detailed projections of her child's possible mental states 17 years down the road and calculated her actions on that basis ("Hell, I thought she'd most likely be dead well before then!"). [This point about Straus closely parallels some observations made about another study here, but I haven't read the study in that case myself to know whether the underlying issues are the same in the two cases.]

More seriously, Straus's remarks do not so much as entertain, let alone give any reason to disbelieve, the plausible (no backward causation required!) common-cause hypothesis, e.g., a slap in 1990 is a response to manifestations of character trait C which manifests itself in depression in 2007 etc.. That's a serious problem (i.e., because if we could control for C then Straus's correlations may – we'd predict, would – disappear).

One very simple indicator of the literal shallowness of Straus's discussion here is that he takes his basic challenge to be a matter of determining causal
direction. But it's the more general problem of inferring causal structure including common causes, common effects, and so on that is the real challenge he faces. In this way, even though Straus does, I guess, basically grasp the problem with his earlier studies, he misses important points even there. Consider again his concession:
"Parents typically spank a child because he or she hit another child or hits them. Thus, it is just as plausible to say that the child's aggression or other misbehavior caused the spanking"
Despite the talk of the "child's aggression" it's clear, on closely reading, that Straus really only means actual, prior aggressive behaviors. No serious common causes are entertained, rather everything's flattened down to behavioral surfaces. If you stay on that behaviorist surface – thereby defining questions of deeper causal structure and influence out of the picture – then you probably can kid yourself that only questions of causal direction between your favorite behavioral variables remain to be answered. And that's what Straus does.

Now the bizarro feature of Straus's own appreciation of his later studies – that there so much as could be a direct causal influence back from future behavior (broadly construed) into the past (if only then to exclude such a possibility as silly) – snaps into focus. If all causal influences have to run between surface behaviors of one kind or another then if some of that surface is in the future then suddenly your problems of causal direction dissolve, i.e., since traffic between past and future is all one way. "That's progress!" you say to yourself.

But it isn't progress. Only relative to your dopey and implausible behavioristic coding up of the situation does shifting some variables off into the future significantly narrow the space of legitimate causal hypotheses, let alone knock out any serious contenders for the causal structure that's involved.

On one level Straus covers himself. He says "The problem of which is cause and which is effect may be less serious in the present case" and he does call for further study ("Nevertheless a longitudinal or experimental study is needed"). On another level, however, this is a clear case in which a researcher is doctrinally incapable of fairly accommodating or acknowledging let alone assessing obvious alternatives to his own framework. Until such a researcher gets his act together and faces up more clearly to the limitations of his research we should ignore and, where that doesn't work, ridicule him. His research is, after all, premised on a hidden assumption which we might coarsely characterize as follows:
"Y'know all that stuff that you think is probably in outline correct about raising your children? Well let's just suppose all of that's false. In that case I can show that smacking is bad..... Hey!....Come back! Why are you walking away from me.....?"


Dave said...

you may want to read - Ive done a rebuttal to Jim Anderton's "smacking causes animal abuse" release

plague said...

Thanks. The "Animal Abuse and Youth Violence" report you link to and especially the exact passage you quote seems to have a slightly different focus: the link they're saying they don't know is causal is the link between being an animal abuser and later becoming a smacker and wife-beater, etc., i.e., rather than being a smackee causing one to be an animal abuser or the converse.

That report also cites uncritically the early 1990's work by Straus that Straus in 2000 says has no causal interpretation. Of course, my note argued that Strauss's later work does no better.

Anonymous said...

I should digg your article so more people can see it, very useful, I had a hard time finding the results searching on the web, thanks.

- Thomas