Murray Straus' comments on Baumrind's data>
August 27, 2001

I heard Diana Baumrind's paper and responded to it to several reporters after. See the Saturday NY Times for Saturday 25 August. I said that her study is excellent, but despite that there are clear reasons for not accepting her conclusions:

1. The most crucial data (the longitudinal part) is based on only 79 cases. This would be enough for many purposes. But in this case, she has split those cases into many small cells, some with as few as 6, 7, and 8 cases. With n's that small, it takes a huge differences in a dependent variable (such as internalizing or externalizing behavior problems) to be statistically dependable ("significant"). That plays into her hand because she wants to show that spanked children are not worse off.

Her handout table also fails to show the mean scores after adjustment for the controls. It is quite possible that if the means were given, they would they show that, although the differences are not significant because of the small n's, they are there, i.e. that the more spanking, the worse the outcome for the child.

The most extreme example of the small number of cases problem is the children whose parents did not spank at all. Diana used this group to counter my claim (based on many studies by me and others) that unspanked children are, on average, the best behaved and the least hassle for parents. She said that her study refutes this because the unspanked children in here study did not have fewer behavior problems, higher IQ, etc. than the spanked kids. But it turns out that there are only 3 such children in her study, so naturally there is no "significant" difference between the unspanked and the spanked children.

2. Baumrind's study is the seventh longitudinal study to appear since the first one in 1997. Although none of the previous studies controls as much as Baumrind's, all controlled for the most crucial sources of error, and together (even though not in any one study), they controlled for all of them. All of the previous six studies found harmful effects for CP. That is an almost unprecedented degree of consensus in research, and especially research on the effects of parenting practices. When 6 out of 7 of the best studies get parallel results, one should not rush to judging that the six are wrong because the 7th gets different results.

3. Finally, for the sake of this discussion, let us assume that her seventh study does call into question the previous results, making them less conclusive. Is it then improper for professionals to advise parents not to spank, as Baumrind argues? On the contrary, the professional OBLIGATION to advise no spanking remains. This is because there is such clear evidence that other methods of teaching and correcting children work at least as well, even in the short run. The situation is analogous to the FDA advising (about 2 weeks ago) against the use of a certain anti-cholesterol medication because of strong but not conclusive evidence of harmful side effects. They withdrew the drug, even though the evidence of harmful side effects was NOT conclusive, because there are two or three other drugs which are just as effective, but do not have a risk of those side effects. In short, there are alternative medicines for misbehavior that are at least as effective, so parents should avoid a medicine for which there is strong, even though not conclusive, evidence of harmful side effects. (See below for some of the evidence on the equal effectiveness of other medicines for misbehavior.)

4. For those of you who read the NY Times article, I want to point out that the review of previous research by Larzeler is highly misleading. (If you want to read the review, it is in Pediatrics for 1996, not in a 1999 journal). First, Larzelere's review does not include any of the longitudinal studies because they did not start appearing until 1997. More important, it presents those studies in misleading ways.

Larzelere concluded that the “eight strongest studies” all show “beneficial effects.” However, when one reads the actual studies, what they show is that, without exception, non-corporal methods were equally effective (see section A3 of this chapter).

Seven of the “the eight strongest studies” refer to short-term compliance with a parental request. Of these seven, five compared CP with alternatives. All five found the alternatives to be just as “beneficial” as CP in obtaining compliance by the child. For example, Day & Roberts (1983) compared spanking as a back-up for leaving a time-out chair with placing the child in a room with a waist-high barrier held across the door for ONE minute. They found that "both spank and barrier procedures were equally effective at increasing compliance" (p. 141), and that "There was no support for the necessity of the physical punishment….” (p. 150). A replication (Roberts and Powers 1990) obtained the same results. Of course, taken literally, Larzelere is correct in saying that these studies found “beneficial outcomes” in the sense of the spanked children complied with the parental requests. But I think it is misleading to not have more clearly stated that non-corporal discipline worked just as well.

Only one of the eight “strongest” studies was about long-term effects (Bernal, Duryee, Pruett, and Burns 1968). It does show a beneficial effect of spanking. However, it takes a leap of faith in CP to regard it as a “strong” study. It is about a single case. Moreover, that case was a child with severe conduct disorder, and possibly schizophrenia. So even that one case does not provide information that applies to children in general. Most important of all, only a minor part of the intervention was the use of CP. The major part was training the mother in how to respond appropriately, such as when to not take the bait provided by this child’s misbehaviour. The mother was also trained to reinforce positive behaviour and to issue commands confidently and consistently. Thus spanking was confounded with other interventions and the study provides no evidence that the spanking part of the intervention was what improved the child’s behaviour. In fact, the intervention might have been even more effective if the spanking part of the intervention had been omitted.

Larzelere also reviewed ten prospective studies and 17 retrospective studies. He summarizes the prospective studies as follows: “Three (30%) found predominantly detrimental effects, whereas the other seven (70%) found predominantly neutral outcomes. In short, the results were either no benefit for spanking or determent. Of the retrospective studies, nine (53% of the 17) found predominantly detrimental outcomes, 7 (41%) found predominantly neutral outcomes, and 1 (6%) found predominantly beneficial outcomes.” So, contrary to Larzeler’s implication of beneficial effects, examination of the actual studies, reveals that none of the 10 prospective studies and only one of the 17 retrospective studies found a beneficial outcome. Moreover, there were substantial percentages (30% and 53% with detrimental outcomes).

Since the publication of Larzelere’s review, almost a revolution has occurred in research on the effects of CP. There are now six prospective studies. All show that the long-term effect of CP is counterproductive in the sense of higher rates of misbehavior two and 4 years later for children who were spanked versus lower rates for children whose parents avoided CP (Gunnoe and Mariner 1997; Straus, Sugarman and Giles-Sims 1997), or that CP had harmful side effects, including slowing the rate of children’s cognitive development (Straus and Paschall 1999) and violence by adolescent boys towards their parents (Brezina 1999) or towards a dating partner (Simons, Lin and Gordon 1998). Five of these studies are summarized in the concluding chapter to the 2nd edition of Beating The Devil Out Of Them (Straus 2001b).

Return to Murray Straus Index
Return to The Baumrind Fallacy
Return to Table of Contents