For the past 41 months, Speculative Grammarian’s Office of Linguistic Documentation has conducted an extensive survey of linguists who have published descriptive grammars. Over 600 grammar writers responded to our extensive questionnaire, covering all areas of data-
We share these results as a service to our beloved field of linguistics, which we know relies heavily on published grammars. After all, if you can’t trust your data, who can you trust?
Previously published in this series: Report 1: Phonology
Grammar writers report that they experienced a certain degree of liberty when choosing specific data items to include. The most commonly-
This pressure also impinged upon the grammar writers’ analytical choices, and so we treat data and analysis together in this report.
All of the grammar writers who returned our survey reported that their published work included at least one example that they had completely made up, and had never checked with a native speaker of the language. Typically, this occurred after the writer had left the fieldwork location and lost contact with the relevant informants. During the writing process, every writer encountered at least one instance in which they could not find a suitable example for a critical morphosyntactic argument, and found that they had no option other than to construct the perfect bit of data themselves. Most writers confessed to having done this only a handful of times, typically two or three at most, with a single bit of constructed data being the most common. (Aside from phonology, of course. As we have already reported, nearly all grammar writers made up all the data in their phonology chapters.) A representative explanation was provided by one respondent, who wrote, “a couple of times I was sure I had the perfect example but couldn’t find it in my notebooks. I made sure that I only constructed examples that followed all the syntactic rules I had internalized during my fieldwork.”
Furthermore, all writers told us that they had included at least one example in their published descriptions which their native speaker consultants had judged marginally grammatical. These examples were of two types: those necessary to defend theoretical points which the grammar writers felt they were obliged to make (see below); and those which they felt were simply due to mistakes or misunderstandings on the part of the informants. The following comment was illuminating: “I’m sure that if my consultant had enough linguistic training, he would have agreed that this sentence was grammatical. Since I only had a few weeks to work with him, I didn’t have time to teach him the syntactic theory.”
Conversely, though, all writers reported that they had, during the revision process, removed examples which turned out to not only fail to support the point they were included to demonstrate, but which actually contradicted those points. Every writer confessed to removing multiple examples from their text, typically in at least five instances. A number of writers noted that they could not remember how many times they had done this, as they kept no records when deleting during the editing process. These two comments were typical of the ones we received: “I’m sure the speaker must have made a mistake when he said that one; it just doesn’t follow the rule;” and “I must have written that one down wrong.”
Data and analysis are of course closely linked, the one depending utterly on the other. The grammar writers we surveyed seemed to be split just about evenly as to whether they felt that analysis depends on data or the other way around.
Most writers admitted to using at least one term for a major phenomenon that they did not actually understand. Terms which were mentioned by multiple writers included converb, perfect, antipassive, particle, and focus. In many cases, the writers used these terms because they had previously been used by a prominent scholar in the relevant language family, and our respondents felt that not using these terms would lower the acceptance of their work. In a large minority of cases, however, grammar writers reported that under pressure from various deadlines they had simply chosen the first term they could think of to label a phenomenon which they did not in fact understand to begin with; as one writer put it, “it doesn’t much matter what I call that construction, since I haven’t the faintest idea what it actually does.”
In fact, over 80% of grammar writers reported that they had followed important terminological choices of a prominent living scholar (never a deceased one), even though they were absolutely convinced that the labels were incorrect. This was most common among graduate students, though junior faculty members also reported similar behavior, normally giving explanations such as “Professor X could prevent me from ever getting anything published in the field if I publicly rejected the term he/she invented.”
A much higher percentage (around 93%, to be specific) reported that they had used terms and analytical categories from a prominent syntactic theory that they actually did not know anything about. Most respondents suggested that this was for the benefit of dissertation committees or, in the case of junior faculty, tenure and promotion committees. None of our respondents, however, agreed to be quoted on this point.
Finally, special mention needs to be made of one specific terminological problem that was noted by fully 58% of our respondents: these individuals reported that they were not actually sure of the difference between perfect and perfective, and thus were not sure whether the term they had used was the correct one. All of these writers, however, reported that not a single reviewer of their published grammars had ever commented on this problem.