Editorial Errors and Standard Deviation
A social science proof of the universal validity of statistics

Dr. Georg Strudelfest

In a recent editorial exchange on Speculative Grammarian’s loosely-protected internal discussion site WikiSpeki, one of our esteemed editors wrote the following:

(1)   Five of our authors are deviant.

Naturally, this audacity provoked anger, irritation, and other pleasurable emotions, and a bit of a row ensued. Somewhere around response #458, the original poster finally clarified that s/he had meant to write

(2)   Five of our articles are deviant.

Although falling somewhere short of self-evident, this claim nonetheless passed without further comment, and is thus deemed to be the final word on the matter. No further edits to the page in question have been made in over two weeks.

While desperately searching for an article to fill out an otherwise meager issue, I read back through the entire discussion and discovered the minimal pair here represented as (1) and (2). And of course, the relevance of this data to the age-old problem of how to prove the validity of the hypothetical construct “standard deviation” sprang immediately to mind. As all statisticians know, and assiduously refrain from discussing, this issue has called into question the very core of the discipline since it was first raised by Bestertester and Carbuncle in 1976: if standard deviation is an indefensible artificial construct, then indeed all of statistics is at least as indefensible, and perhaps more so.

It will be immediately apparent to those who have read this far that (1), produced when in fact (2) was intended, is an example of an Editorial Error (EE). Furthermore, I take it as self-evident that

(3)   (1) > (2)

is a prototypical example of an EE, and therefore that an adequate analysis of (3) will in fact shed light on all instances of EE. Furthermore, as EE are clearly a subtype of performance error (PE), any EE instantiates all PE and by extension all human behavior (HB), which of course consists primarily of PE on most normal days.

To turn now to the proposed analysis, I note that the Editor in Error referred to five authors (example 1), when s/he meant to refer to five articles (example 2). The margin of this error falls within one standard deviation, based on the semantic categories of Lakoff’s Women, Fire and Dangerous Things, which I take to be the most advanced typology of human categorization available at the moment. Thus, the standard deviation is demonstrated to hold for all HB.

In many cases, a slightly more abstract formulation allows for a more insightful analysis. Such is true here, with the formulation of (3) inviting us to make an alternate analysis from a historical point of view. For this analysis, of course, we must take Greenberg’s mass comparison as the standard. I need not bother showing the correspondences, but I will anyway:

(4a)   A   TH   R   S
(4b)   A   TC   L   S

So even under this much more rigorous standard, this EE is easily within a standard deviation. Such is the power of applying linguistic theory properly to an otherwise difficult-to-explain phenomenon.

It is apparent, then, that the distribution of EE shows that all HB phenomena have “normal” distributions. This is critical for showing that common statistical tests can be used in any situation.

	My Name, the Performative—Virginia Bouchard
	The Speculative Grammarian Survey of Grammar Writers—The Writing Process—Morris Swadesh III
	SpecGram Vol CLXVII, No 3 Contents