Questions of validity in mammary physiology: methodology and ethics.

An analysis of the type presented is highly unusual, but most timely in view of the rapid changes occurring in this and other branches of applied biology. Mammary physiology is here interpreted broadly to include cognate sciences, fundamental and applied aspects, and both medical and dairying contexts, though references to the latter predominate. To aid discussion, ‘validity’ is considered under four headings, i.e., empirical, experimental, epistemological and ethical: the categories are not absolute and each is, in a sense, contained in its succeeding category, like a set of Russian dolls. Several examples quoted are decidedly ‘historical’; but the principles they illustrate have equal validity today. Empirical validity This refers to perception and recording of sense data by the experimenter; transmission in publications; and interpretation as facts by other scientists. Conceptually, the process involved is simply data transfer. Although apparently unproblematical, in practice, many developments in physiology can be traced to identification of errors perpetrated at this level. For example, Graham et al. (1938) made deductions about the blood precursors of milk protein in goats which cannot be sustained by examination of their recorded data, as was pointed out in an exemplary review by Barry (1961). A more amusing case is the recording of information on milk composition for guinea-pigs, through mistranslation of the German word ‘Meerschwein’ (porpoise) (see Jenness & Sloan, 1970). A particularly apposite example is that cited by Widdowson (1984), i.e., the incorrect description of the treatment of rats in feeding experiments reported in the classic paper by Gowland Hopkins (1912), which led to the discovery of vitamins. Generally speaking, opportunities for error at the empirical level range from fraud (see Broad & Wade, 1983) to selective quotation of findings (citing the ‘best’ as ‘typical’). Many scientists would defend the exclusion of obviously aberrant values caused by operational error, but since this is usually a post hoc decision it is susceptible to unintentional abuse. Experimental validity Experiments in mammary physiology may be classified as: (1) in vivo non-surgical, (2) in vivo surgical, (3) in vitro. Category (1) includes: variations in milking frequency, intraductal infusion, teat cannulation, gland volume measurement by water displace-