A study is made of the effect of experimental errors of microbiological tests upon similarity values used in numerical taxonomy. For the coefficient SSM approximations are given for the mean and standard error of similarity values when some test results are erroneous, S', if the true value of similarity (S), the average probability in I, o data of erroneous tests results ( p ) , and the number of tests (n) are known.Formulae are given for estimatingp from an analysis of variance, and for including or excluding chosen factors that affect test reproducibility.When p is over about 10 % the error of similarity values becomes unacceptably large. Test error also increases the existing sampling error. It is generally better to employ many tests even if they are not as reproducible as desired, rather than to use only a few extremely reproducible tests.Examples are given suggesting that within one laboratory p can usually be kept below 5 %, but the discrepancies between laboratories may often be much larger.It is advisable to replicate some of the strains in numerical taxonomic studies, so as to afford a check on test reproducibility.

No abstract

A numerical taxonomic study was made of strains belonging to the genera Bordetella, Alcaligenes, Brucella, Acinetobacter, Moraxella, Neisseria, Haemophilus, Actinobacillus, and Pasteurella, together with a few strains from allied genera. They were examined for 139 characters covering a wide range of tests and properties. The strains fell into two major groups. Group A contained Haemophilus, Actinobacillus, and Pasteurella. Group B contained eight main clusters, representing, respectively, Bordetella pertussis, Bordetella parapertussis, Bordetella bronchiseptica combined with Alcaligenes faecalis, Brucella, Acinetobacter anitratus, Acinetobacter lwoffii, and two clusters containing Moraxella and Neisseria strains. Brucella and the three clusters of Bordetella (including Alcaligenes) were more similar to one another than t o the other clusters; it is suggested that the family Brucellaceae should comprise these three genera, and that the other genera should be placed elsewhere.Several families of bacteria listed in Bergey ' s ManuaI (13) represent aggregations of genera which at best can only be described as loose associations. Instead of encompassing organisms which overall are closely similar, they have been set up t o accommodate genera with one or two prominent characters in common. Such a group is Brucellaceae, or its predecessor Parvobacteriaceae (12,63). This family is the niche for those gram-negative, rod-shaped bacteria which require rich media for growth and are responsible for many human and animal infections. Their potential pathogenicity and the alleged difficulties of working with them account largely for their neglect by taxonomists. They tend to grow poorly or t o die out on the usual media, whereas the addition of blood t o increase growth often interferes with tests.Within Brucellaceae the genus Bordetella Moreno-L6pez 1952 has been one of the least studied. Three species are generally recognized: the type species B. pertussis (the whoopingcough bacillus), B. parapertussis, and B. bronchiseptica. Although these organisms have been placed in various genera (e-g., Haemophilus), in recent years their segregation as a separate genus has become generally accepted ( 1 3, 84). They do not require X factor (haematin) or V factor (nicotinamide adenine dinucleotide) as do species of Haemophilus. The only reason that B. pertussis requires blood is to remove toxic substances from growth media; the blood can be replaced by starch or charcoal (65) or by anionic-exchange resins (45).Bordetella-species are obligate aerobes, whereas species of Haernophilus are facultative anaerobes. The guanine plus cytosine (G + C ) ratios also differ, being 67 t o 70% for Bordetella and 38 to 42% for Haemophilus (20, 36, and findings reported below).The separation of Bordetella from Alcaligenes, Acinetobacter, and Brucella is more difficult to justify, and the main motive for the present work was to investigate this question. Many workers have noted points of similarity between Bordetella bronchiseptica and A k aligenes faecali...