Theory and Object in Law: the Case for Legal Scholarship as Indirect Speech


On the final count, Westerman encourages legal scholars to continue doing what they have always done, to wit doctrinal law; only they could do with an update and apply some more rigorous techniques of empirical research to get their facts (or effects) right. She thinks that what legal scholars do, and can do, is largely determined by the categories and the reasonings inherent to the object they are investigating: law, or rather some specific legal order, as it struggles with the ever changing facts of socio-political life. This order emerges and re-emerges without a premeditated plan or method, let alone a methodology, much like, after some tinkering, order re-emerged in her mother’s cupboard, according to parameters of integrity and continuity.2

I want to take issue with this – rather Savignyian3 – thesis, that scholarship on law (‘legal science’ if you wish) is predetermined by the categories and the reasonings in law. My alternative thesis is that the two are certainly related (as legal scholarship is a hermeneutic enterprise), but in a much more complicated way than Westerman suggests. Legal scholarship is indeed supposed to offer a theory of a legal order, but not necessarily in the terms of that legal order.

Let us start at the common place, where Westerman and I agree. Legal scholars are well advised to improve on their scientific knowledge of how facts are construed. Here is an example, derived from Van Fraassen.4 Let me slightly change the context and the story for our purposes. Suppose a city considers reorganising a department of civil servants due to its sub-standard performance. A group of civil servants in that department brings charges against the city, claiming that lighting and ventilation condition in their workplaces harm their wellbeing, hence their ability to perform according to standard (let’s call that their productivity), hence their individual chances of tenure, or promotion, or getting fired in a reorganisation. Here, I continue with Van Fraassen:

The city hired a statistician who showed conclusively by means of sampling that the productivity among workers in ill-lit and ill-ventilated spaces was no less than among workers in general (or in better lit, better ventilated spaces) – the productivity level was the same in both groups. So the complaint was concluded to be baseless.

Then the civil servants asked for an opinion by another scientist:

(. . .) (T)he second statistician broke the data down by looking separately at women and at men. She showed clearly that among women, the productivity was less for workers in the ill-lit and ill-ventilated spaces than elsewhere. She also showed that among men, the productivity was less for workers in ill-lit and ill-ventilated spaces than elsewhere! So relevance of working conditions did not show up until there was a subdivision by this third factor (gender). How is this possible? That is precisely Simpson’s paradox; correlations can be washed out, or on the other hand brought to light, by averaging in different ways. Here is the solution to the puzzle: under all conditions the women were more productive than men working under the same conditions, but the women were predominantly working in poor conditions.

For those who want to check, I also quote Van Fraassen:

To make this concrete, imagine a very small situation, involving only 4 men and 7 women. Under good conditions the women produce 8 items per hour and the men 4. Under bad conditions the women produce 4 items per hour and the men 2. But two men and two women work in good conditions, with two men and five women assigned to bad working conditions. In the bad workplace, the production is 2(2) + 5(4) = 24 items per hour. In the good workplace, the production is 2(4) + 2(8) = 24 as well, precisely the same.5





5 × 4 items = 20

2 × 8 items = 16


2 × 2 items = 4

2 × 4 items = 8




I propose to extend Van Fraassen’s example by one legal turn. The city dismisses this reasoning by saying that the whole statistical exercise shows only one thing: if men would just work harder, there would be no issue of good and bad working conditions in the first place, as the department would perform according to expectations. These working conditions are not just good or bad, they are ‘good enough’ for all practical purposes, if only men would be less lazy in general. The city admits that the present situation is possibly harmful for female workers in particular, it promises to take measures to improve male performance and postpone reorganisation, but it denies that it is liable. So according to the municipality, there is no (relevant) correlation between production and working conditions, whereas the civil servants say there is.

Here I pause and ask Westerman what it means here that ‘the legal system functions as a theoretical framework that selects the facts and highlights them as legally relevant ones’?6 It is only trivial to say that there are no legal rules on averaging in, for instance, the Dutch, the German, or the French legal system, as far as I know. Indeed the whole concept of averaging is unknown in law, and no judge would hold a defendant liable by dint of not having his statistics right. What is more important, however, is this: the legal order does not give us any explicit ruling on which statement of fact is the more appropriate one considering the rules it upholds. And yet, although statistics will not be part of the law’s language, it may be part of the language that says what the law says. Scholars with sufficient knowledge of the Dutch legal order, for instance, will be able to predict that officials will have much more sympathy for the second way of averaging than for the first. It is also clear why: this has to do with equal opportunities, non-discrimination, etc, ie with norms that are explicitly promoted by law. The connection is there, however, not by virtue of the authority of the legal order, but by virtue of what legal scholars find pertinent in that legal order. So Westerman is right in claiming that legal scholarship feeds on the categories of the legal order, but not in the sense that these pre-determine the categories of legal scholarship.

Now let us suppose – for a second example – that this case goes to court, where there are three judges who will decide by majority vote in chambers. The rule they have to apply in this case is that if the working conditions are of unequal quality, and if unequal conditions bring harm to individual civil servants by affecting their productivity, the city is liable for having these bad working spaces renovated before a certain date. We can easily see that these judges may get into what contemporary literature on judgment aggregation calls a discursive dilemma.7


Judge A

Judge B

Judge C

Bad working conditions?




Harm done?



City liable?


How should these votes be counted? What is majority vote in this case? If we go by the conclusions, the majority vote will be that the city is not liable. However, if we go by the premises, we see that the majority of judges think that there are bad working conditions, and that there is