27.1 Introduction

The organizing of a dictionary project involves many of the standard project management principles which apply to other types of projects. There are also aspects of dictionary projects which are very different to other projects, such as the small pool of personnel available and qualified to undertake lexicographical work – as compared to, say, a software project. Rather than presenting an exhaustive list of general project management principles, this chapter highlights those points most relevant to dictionary projects.

27.2 Preparation

Before looking at the tasks involved in dictionary compilation, let us look at tasks which need to be completed and components which need to be ready before the dictionary compilation or revision begins. This phase of the project doesn’t involve as large a team as will be needed in the compilation phases, but it does have a number of very large tasks which are likely to require three to six people to complete.

27.3 Identifying Core Tasks

This section lists the main tasks which may be encountered in a dictionary project, and the subsequent discussion of how to implement those tasks is based on the list.

The tasks here are based on the compilation of a bilingual dictionary from scratch. Such an event is unusual – a dictionary project may entail revision of an existing dictionary or compilation of a new dictionary based on a previous one. Also, some steps won’t apply to a monolingual dictionary. Where a major revision of a dictionary is undertaken, however, many of the tasks will still be executed, albeit on a reduced scale. The tasks described are also based on compilation of a fairly large dictionary, that is, with 40,000–50,000 headwords. For a smaller dictionary, the tasks can be scaled down accordingly. The aim here is to present the full scale and type of tasks relevant to dictionary compilation.

The four main phases in the compilation of a bilingual dictionary are source language, target language, editing, and publication. It is assumed that the preparation phase is complete and its outputs available.

27.4 Translating Tasks Into a Work Plan

To estimate the workload involved in the tasks identified in Section 27.3 above, the necessary editorial decisions will ideally have been made in relation to those tasks before the project starts. To translate these tasks into a work plan, the steps described in this section are recommended.

27.4.2 Estimating the Tasks

The first step in estimating tasks is to identify what those tasks are. For each of the main phases of the project, that phase needs to be broken down into the steps needed to complete it. For the source language phase, the main tasks might be something along the following lines: compilation of first draft by lexicographer; review by senior lexicographer and feedback to the compiler; revision of entry by lexicographer; second review by senior lexicographer. For the target language phase, the typical tasks would be: target language material added by translator; review by senior translator; revision of entry by translator; second review by senior translator. For the editing phase, the typical tasks would be: editing by editor, to produce draft entry; review by senior editor; revision of entry by editor; second review by senior editor; internal proofreading. And for the publication phase, the typical tasks would be final editing by editor or senior editor; review by (different) senior editor; external review and proofreading; uploading final entries to website and app.

Those task lists may vary from dictionary to dictionary; the key thing is for the project manager and the chief editor to identify all of the main tasks in the project, their sequence, and who will execute them. Once that is done, the next step is to estimate the work effort and duration for the tasks. Note that the structure implied by the tasks above is for a large dictionary project, executed under ideal conditions. In the real world, finance and human resources may be much constrained and not all of these roles may be filled – or the senior lexicographer, senior translator and senior editor roles may all be filled by one person!

Estimating compilation tasks should be based on lexical units, or LUs. As explained above, the LU count comprises not just the basic senses, but also idioms, phrasal verbs and inflected forms, all structures which can entail significant amounts of work. Attempting to estimate the number of LUs for the source language phase can be particularly tricky, since the framework compilation hasn’t yet started. But it’s essential to make the best estimate possible – based on trial compilations, other dictionaries in the same language or even other dictionaries in similar languages.

Having made an estimate of the number of LUs in the project, the next step is to define the daily rate for processing them, in each stage of the project. It’s crucial to get the daily (or weekly, or monthly) rates as accurate as possible. A large dictionary project might well have 100,000+ LUs, so the underestimation of a task may have major impacts on the project. For example, if the daily rate of LUs for a particular step is planned to be ten per day, but the rate achieved is only eight per day, then that step of the project will take 25 percent longer.

Companies or organizations experienced in dictionary production will usually be able to estimate these rates accurately. For organizations new to such tasks to get a solid basis for planning, it is strongly recommended to conduct trial runs before the start of the project proper. No small task, trial runs involve assembling a number of lexicographers, translators and/or editors, selecting a group of trial headwords of varying complexity, and then working through all of the steps in the different phases of the project, while carefully measuring the time taken. But it’s an essential task for any group embarking on a dictionary project for the first time. It will also have the side benefit of clarifying the style guides for the different phases of the project, as questions arise during the trial runs.

Figure 27.1 shows an example of such a trial. In this instance an approach to the translation phase was under consideration, whereby the translators would be asked first to translate most of the LUs spontaneously, that is, without reference to any other sources, but based solely on their knowledge of the target language. In a second pass, they were asked to cross-check with other sources – existing dictionaries and also a terminology database – and to record the time they spent on both translation activities. The trial group consisted of twelve entries containing 11,210 words of text in the source language; some entries were distributed to all five translators, while the rest were distributed to a subset of the team.

Figure 27.1 Statistics from trial translations.

In columns L and M we see the total time spent by all translators on each entry, for both the spontaneous translations and the researched translations. In columns N and O the mean time spent per translator is calculated for each entry, and these figures are used to extrapolate, in columns Q and R, the mean translation time per 100 words of source language. In column S in line 16, we now have an estimated total translation time per 100 words of the source language. This is used in lines 18 and 19 to extrapolate the total translation time needed for Group 1 of the entries, and in line 21 to extrapolate the translation time needed for all twenty-five groups in the dictionary. Such extrapolations are still somewhat rough estimates, but they give at least a ballpark figure which can be compared with the hours originally budgeted for the translation phase. Trials such as this can also be useful for organizations which have dictionary-production experience, but where a new language or a new task is to be handled.

Whether based on previous experience or on trials, the daily quota of LUs will vary widely in each phase of the project and for different steps within each phase. Some examples are shown in Table 27.1.

Table 27.1 Daily quota of lexical units in project phases and steps

Phase	LUs per day
Translation phase	40–64
Publication phase – editing of entries already published online, for inclusion in a printed version Editing by editor Review by senior editor	100 70

The rate of LUs would be much lower than those shown in the source language phase if the entries were to be compiled from scratch. It would also vary widely depending on the type of entry. The referential headwords described earlier will often be of a technical or quasi-terminological nature – headwords such as ladder, limekiln, and magnetize. Entries for such headwords are usually much quicker to draft, translate, and edit than for denotational headwords such as big, run, quickly. And the function words and super-large entries referred to earlier will be at the top end of the scale in work effort. In summary, the daily rate of LUs will vary according to entry type and the phase of the project. There must be a solid basis for the estimates in all of the steps and phases.

27.4.3 Assembling the Team

The bigger the core team of full-time, and ideally, permanent employees, the better, but the days when dictionary publishers had large teams of permanent staff is probably long gone. Even in state-funded organizations, such as national language bodies, it is increasingly difficult to finance such teams of lexicographers. The implementation of a dictionary project thus often involves assembling a team of freelance lexicographers, translators, and editors.

27.4.3.1 Recruiting

How contract staff are recruited will depend on the organization’s experience in dictionary projects; it may have a contact list of people it uses regularly. But if it doesn’t, or if it is undertaking a project in a new language, then a number of tasks will need to be undertaken to recruit personnel, including devising tests to measure applicants’ ability; advertising the roles to be filled; conducting and correcting the tests; possibly interviewing applicants; arranging training courses and creating batches of trial entries.

27.4.3.2 Training

Training may be needed not just for staff who are new to the project but for personnel who have been working on the project but are now moving into a new role – for example, translators who have shown the potential to do editorial work.

The era of COVID-19 has shown just how many activities can be carried out online, sometimes surprisingly efficiently. But we strongly recommend organizing training courses where everyone is in the same physical location, if at all possible. Although this involves more effort and expense, it has many benefits – there is much more feedback when people are in the same room: it’s often easier for people to ask questions in such an environment than in a screen-to-screen set-up; reading a person’s body language can help to assess whether they have truly understood the points being discussed; and often the informal conversations, between training sessions, can be valuable in discussing points or issues. Last but not least, it helps to foster an esprit de corps. For similar reasons, we would recommend holding regular workshops, even when no new training may be needed, to make sure that everyone is on the same wavelength.

One important point, whenever a training session has been completed, is the size of the follow-up work batches which you distribute to the participants; these should be much smaller than the normal work batches. So, if a normal batch of entries contains three to five days’ work, the post-training mini-batches might contain enough LUs for just one day’s work. Very often it will take several batches of putting the training into practice until it has been fully comprehended by the participants, and there is thus likely to be a much higher rate of corrections in the post-training batches. It is far more efficient for an editor to mark errors in a batch of forty LUs and then return it to the author than doing this in batches with 100 LUs.

Another point regarding trial batches is that it can be useful to give trial batches with the same entries to all participants and to compare the results delivered by each participant for the same entries during the training course.

A key point regarding training is to recognize which tasks can be taught and which ones can’t be; this is something that often varies from individual to individual. A simple example is the use of technology and the software tools which have now become an intrinsic part of dictionary projects – some people handle this well, while others struggle with it. In the latter case a decision may need to be made whether extra support can be given to certain personnel if their expertise in other areas is particularly valuable to the project.

There will also be other areas where training is not successful, as with examples relating to methodical processes and example sentences in a recent project. In processing referential entries, translators were asked to check a number of extant sources, in a designated order of priority, and to mark the selected translation according to whether it was an exact match for the source language term, whether it was found in a related entry, whether it was based on a related term, and so on. Essentially it was a ten-step process, which was well-documented in the Style Guide with plenty of examples. Most translators became expert in this process, with practice, but a small number just didn’t and would return a high rate of errors even after completing multiple batches. In the step of the project where the editors were to add example sentences, when needed, it was constantly stressed that such sentences should be as short as possible, while adequately illustrating the context needed. This did not prevent wordy examples like the following, where an example is needed using the words hold and meeting, and in which case the first eight words of this example would have sufficed: “we held a meeting with Ciara and Liam, on a rainy Wednesday afternoon, in a stuffy conference room.”

On a related topic, Figure 27.2 shows statistics gathered early in the editing phase of a project, as regards how many example sentences editors were adding to the entries in that phase, rather than how long the examples were. The aim was not to ensure that everyone had near-identical statistics, since these can vary significantly depending on the type of headword being handled, but to check for outliers. What’s interesting here is that Editor 5 is inserting less than half the mean number of examples, while Editor 1 is altogether trigger-happy. If those editors were unable to show an understanding of why they needed to increase or decrease, respectively, their rate of examples, there would have been no option but to assign them to different duties.

Figure 27.2 Statistics analyzing the rate of examples in entries.

To summarize, then – in the same way that no number of lessons would ever turn this writer into a good singer – there will be tasks in a project which some personnel simply cannot be trained to do. It may be necessary to recognize that for certain tasks, a person just doesn’t get it or doesn’t have the instinct needed for it and that they may need to be allocated to different tasks in the project.Two other points regarding recruitment need to be taken into account. First, the percentage of a contract staff member’s working hours during which the applicant will be available for work on the dictionary, if selected, is crucial. If the applicant is very experienced, they may be able to make a useful contribution even if available for as little as one day per week. But for personnel who are new to the work, or less experienced, anything less than two days per week may mean too long a gap between dictionary tasks, where the person is pretty much relearning from scratch each time they return to the dictionary work – a bit like a runner planning a marathon but training only one week every month: their fitness/expertise just won’t build up.

The time a freelancer works on the project also cuts both ways – it’s important to establish as far in advance as possible what each person’s availability is over the course of the project and to give freelancers notice of how long they will be needed.

Second, it is, of course, important to set a rate for the job which will attract good people, without being unnecessarily high.

One final word regarding assembling the team – it can help a lot if the publisher has links with a university and provides work experience for interns. This is a useful way of assessing whether particular students have an aptitude for lexicographical work and if they would be worth contacting for future projects.

27.5 Monitoring the Project

Unlike some other kinds of projects, with dictionary projects progress can be measured based on the number of LUs completed compared to the number targeted. Setting up plans and spreadsheets for all activities may require a fair bit of work, but once in place these IT tools can minimize the labor required to monitor progress. Figures 27.3 and 27.4 show how such tools might be used: they illustrate a plan which was revised at the end of Quarter 1, for a project phase which had eighteen months remaining.

Figure 27.3 Planned output for editors.

Figure 27.4 Planned editorial output for half a year of the project.

Figure 27.3 shows the targets which were set down for editors and senior editors, in lines 22–24. The entries are divided into three broad categories according to complexity – Category 1 entries are to be reviewed by senior editors, while Categories 2 and 3 are to be reviewed by editors. It should be emphasized that this phase of the project was to turn an already-published electronic dictionary into printed form, so if the daily rates seem high, it’s because the editorial heavy lifting had been done prior to the online publication. In this phase the activity was to select and edit the material for the print version.

Lines 31–35 and 42–44 of Figure 27.3 show that the availability of the senior editors and editors varies from one to three days per week, or 20 to 60 percent.

Figure 27.4 then shows the planned targets for each week during the second half of year 1, following implementation of the revised plan:

• Weekly targets shown as zero, against a shaded background, reflect planned unavailability of that person during a particular week (e.g. vacation, maternity leave);

• Weekly targets shown as zero, against a background of vertical bars, however, show unplanned availability of that person (e.g. illness, jury duty, reallocation to other tasks);

• All other weekly targets are calculated automatically, based on the number of working days in that week (line 7), the target LUs per day from the Figure 27.3 spreadsheet, and the percentage of the working week that person is available, again based on Figure 27.3;

• The actual number of LUs completed each week is entered manually;

• Where the cumulative number of LUs completed is less than the cumulative target, the shortfall is highlighted automatically in a designated color type, as in line 14 and parts of line 20;

• The cumulative figures for all the senior editors, in lines 41–45, are calculated automatically; similar figures (not shown in this example) are calculated for the editors.

The spreadsheet gives a clear snapshot, on a weekly basis, of progress. Of the three senior editors, two of them are behind target in July; by September, two senior editors are ahead of target, so that by the current week (week 40, highlighted in Figure 27.4), the project overall is (just barely) ahead of target, as illustrated by the negative percentage in the shortfall row in line 45.

Monitoring progress on a weekly basis, especially for recently started phases and tasks, is highly recommended. If such activities bed in after some time and are progressing on track, the intervals between checking may be longer. There are also other areas of the project which need to be monitored regularly – spending versus budget, quality control, risks and issues. It is good practice for the project manager to discuss these weekly in a short meeting with the chief editor and the senior editors.

27.6 Maximizing Efficiency

Since dictionary-writing is so labor-intensive, it’s worth highlighting other techniques which can help to minimize labor and maximize efficiency. Weekly monitoring is one of the important activities since it will highlight early any areas where progress is behind schedule. It can’t be over-emphasized that any effort spent streamlining all of the steps in a dictionary project will pay dividends. Some other techniques which can contribute to this effort are discussed below.

27.6.1 Statistics

A key step in maximizing efficiency is to know how efficient (or inefficient) the project is. The benefit of statistics has already been cited in discussing how many example sentences each editor inserts and how long the sentences are. Editors who insert too many examples or sentences that are too long create unnecessary work for senior editors, who will review and have to amend those examples. So it’s worth gathering statistics on these, as described above, and focusing on any outliers. In the same way, spotting early on those editors inserting too few examples, and addressing it with them, will reduce the need for feedback from senior editors when reviewing individual work batches.

Other useful statistics include time taken per batch, number of translations added, and comments added – by lexicographers, translators, and editors.

Figure 27.5 illustrates how these statistics can be used. The statistics are taken from early stages of the translation phase of a large project with over thirty translators. Almost none of them had previous experience of translating in a dictionary context, and most were recruited via an open competition. These statistics proved essential in confirming which team members would continue on the project and in identifying potential issues at an early stage.

Figure 27.5 Sample statistics from the translation phase.

27.6.1.1 Time Taken per Batch

Column E shows that most of the distributed batches were completed within a week. But Translators 5, 6, and 9 took five to seven weeks to complete a single batch. For one translator, this was explained by illness, but the other two had full-time jobs and, lacking the minimum time available needed to make a regular contribution, were dropped from the team.

This period of elapsed days is useful for identifying such cases but is not granular enough for other purposes. At various stages in the translation phase, all translators were asked to record the working time spent on each batch and to return that information when uploading a completed batch. This information was useful in confirming whether the estimated daily rates for translating LUs were realistic or might need adjustment and in highlighting where some translators were spending well over the set time in completing their batches.

Excessive time spent on batches could be indicative of an unclear understanding of the translation process or of the translator being too meticulous. The latter occurs regularly in dictionary projects and needs to be guarded against. There’s no such thing as a perfect dictionary entry – and no limit to the number of times a lexicographer, translator, or editor may be tempted to revise their work. Freelance personnel may even revise work to their own financial detriment: although they might get paid per batch or based on LUs or on the number of translations, they continue to spend time on completed entries, essentially unpaid work. It is vital to instill in the team an understanding of the balance between reaching for perfection and the practical budgets and deadlines of the project.

27.6.1.2 Number of Translations Added

In the translation phase of this particular project, in some contexts the insertion of a translation was mandatory (e.g. basic senses, phrases, phrasal verbs), while in others a translation was optional (e.g. for collocations). Before batches were distributed, blank translation fields were inserted into the entries by a computer batch job. After the batches were translated, the number of translation fields always showed an increase, as expected, due to the provision of multiple translations for the source-language word or phrase. The average increase in translation fields during this phase was 73 percent.

In lines 10 and 11 of Figure 27.5, the two batches for Translator 8 (in column I) show an increase in translation fields of 100 percent and 132 percent respectively. As with the number of example sentences, there’s no correct figure for this step, but a translator way outside the norm should prompt further analysis as to the reason. Supplying many more translations than can be used in a final entry creates unnecessary work for editors. Furthermore, as the translators in this case were being paid based on the number of translations supplied, it didn’t make sense to pay for translations which weren’t going to be used.

Lines 2 and 3 show the opposite issue, where Translator 2 has added only 13 percent and 15 percent of additional translations to their two batches. It may be that these were two batches with many referential headwords, which tend to be a lot simpler, or the translator may be skimping on the additional translations needed. In the latter case, it’s highly preferable to correct this early in the translation phase, rather than coming across it in the editing phase, where it could generate significant extra work for editors.

27.6.1.3 Comments Logged

In each phase of this project, lexicographers, translators, and editors could add comments to an entry for the attention of senior lexicographers, senior translators, or senior editors. In line 14 (Figure 27.5) Translator 11 has added thirty-one comments (column J), far above the average and, again, generating work for senior translators or editors. There may be good reason for the high number of comments – a particularly tricky or larger than usual batch, for example – but a translator who shows such patterns regularly may warrant investigation. On the other hand, Translator 10 has added zero comments. Again, there may be acceptable reasons, but it would be necessary to check a translator’s understanding of when and why such comments should be added if this pattern occurs regularly.

27.6.2 Using Software Tools

Needless to say, it behooves dictionary makers to take advantage of the powerful tools that digitization makes possible. Here are a few examples:

27.6.2.1 User-Friendly Screens

Such screens are helpful to those working on the dictionary project, rather than the end-users who will be the dictionary customers. Entry structures can become complex, so auto-inserting fields and color-coding them can make the job easier and clearer in every phase of the project, as in Figure 27.6, where various categories are colored differently – HWD in a different color from POS (N) and both different from EX. We previously mentioned a computer batch job used to insert the mandatory translation fields, and Figure 27.6 shows that those fields would also have their own color-coding. So the translator who opens the entry amnesty sees that the Translator ID field TRID (six lines down) must be populated and at least one translation for the first sense of the headword inserted. Note that translation fields have not been auto-inserted for the sample sentences nor the collocations because it’s not mandatory to translate them – the translator can decide whether any warrant a translation, using the translation Style Guide.

Figure 27.6 Fields auto-inserted and color-coded.

Note also that in this first sense the domain field (eleven lines down) is highlighted in a dark color to draw the translator’s attention to the fact that this is a referential, or term-like, sense, and should be handled accordingly.

Some of the dictionary-writing software packages available also allow translators and editors to tailor the entry view so that users can suppress fields they don’t need to see, and which clutter the display, homing in on elements they wish to focus on.

27.6.2.2 Auto-Generated Reports on Batches

When a translator or editor has completed a batch, computerized reports can check for missing mandatory fields, typos, and violations of the Style Guide. The sample report shown in Figure 27.7 is from the editing phase of a project and includes these features:

• Statistics – for translations, comments and example sentences, giving the editor a picture as to whether the balance looks right for the batch;

• Errors – highlighted (and here shown in boxes) must be corrected. They include locked entries – the editor has forgotten to upload the entry (or batch); invalid entries – where mandatory fields are missing, for example the part of speech label, or other DTD violation errors; TR with 000 – used for referential headwords and senses where no translation was found in the sources to be checked, and it’s up to the editor to insert a recommended translation and remove the “000”; and SRC field missing – again for referential headwords and senses, the source information must be included;

• Possible errors – the remaining labels, not necessarily errors, may be violations of the Style Guide and need to be checked by the editor who completed the batch. Some of the potential issues shown are: EX with capital – in this case the second example sentence is correct but the first is incorrect: it should start with a lowercase letter; EX formal – a prompt to the editor to check if the style should be more informal, for example, should the first example be “he’s three times as big as her” rather than “he is … ”; and TRCOMMENTs – comments from the translators, which should be dealt with by the editor and then removed from the entry.

Figure 27.7 Auto-generated report on an edited batch.

This report was sent to editors whenever they completed a batch. It allowed errors to be identified and corrected without intervention by another editor or senior editor, shortening the review and proofreading steps of the project accordingly. In summary, it moved tasks from the senior editors to the administration staff who generated and sent the reports.

27.6.2.3 Spellchecking

Whenever a group of new or amended headwords is to be added to the dictionary, just prior to the start of the proofreading step, the text fields can be extracted and checked by a spellchecker. It’s an easy task to extract all the source language fields into one text file or spreadsheet and all the target language fields into another file and to run these through a spellchecker for the respective languages. The flagged “errors” can then be checked manually and corrected where necessary. Many false errors will get flagged – proper nouns, abbreviations, etc. – but this is a more effective way to find typos than proofreading the new/amended entries in their entirety, necessary as the latter step still is. Better again is to have a spellchecker within the dictionary-writing software package, where possible.

27.6.2.4 Spot Checks

The structured nature of electronic entries means that a vast number of spot checks can be run to check for human error – even, over time, to more than fifty checks, run just before each upload of new and revised headwords. Among examples would be missing sound files; missing elements – for example, verbs lacking a “transitive” or “intransitive” label; duplicates – phrases which appear under more than one headword; incorrectly nested structures (thus likely to be displayed incorrectly).

27.6.2.5 Reversing the Source Language and the Target Language

Where the dictionary is bilingual and one side of it is complete or partially complete, consideration can be given to using software to reverse the completed entries, as a starting point for the other half of the dictionary. This works particularly well for referential headwords, as many such entries have only a single sense and often a single translation in many language pairs – parsnip is always panais in French, and panais always translates back to English as parsnip. This reversal can be extended to more complex headwords. (For descriptions of this for English–Slovenian, see Reference Krek, Šorli, Polonca, Bernal and DeCesarisKrek et al. 2008; for Estonian–English, see Reference Veldi, Dykstra and SchoonheimVeldi 2010.) While it will require extensive checking and supplementing, it can still entail less labor than starting the entry frameworks from scratch.

27.7 Other Tasks

The core tasks in a dictionary project (discussed earlier) account for most of the work, but a large number of ancillary tasks must also be included in the project plan, as outlined below.

27.8 Summary

To summarize the issues discussed in this chapter, the key areas in successfully executing a dictionary project are the following:

• A fine-grained categorization of the entries, enabling them to be grouped together efficiently in coherent work batches;

• Distribution of tasks at the lowest level needed to complete them competently, whether by junior, mid-range, or senior staff; removal of tasks from a senior level to another level where feasible;

• Constant monitoring of output, especially when new steps and phases start;

• High usage of tools and statistics to achieve the above steps and to identify potential issues early.

Figures 27.8 and 27.9 summarize, respectively, the roles and team structure in a dictionary project and the main tasks in a high-level sample project plan. Once again, these are geared toward a large dictionary project, executed under ideal conditions. With a smaller project or conditions less than ideal, the team structure and the plan will need to be tailored accordingly.

Figure 27.8 Project personnel.

Figure 27.9 Sample high-level project plan.

28.1 Introduction

This chapter provides an overview of dictionary editing both for scholarly historical dictionaries and for commercial synchronic dictionaries. The observations reflect my nearly thirty years of experience in lexicography, beginning in the literal hands-on process employing pencils and slips of paper and extending into the contemporary era of computer databases and digital corpora. My first stint as a lexicographer (1982–1989) took place at the Middle English Dictionary (MED), a historical dictionary covering 400 years of the English language from around 1100 to 1500, that is, from just after the Norman conquest of England to just after the introduction of printing to England. I then spent twenty-two years (1989–2011) editing commercial dictionaries, most notably the American Heritage dictionaries (AHD), ranging from dictionaries for preschoolers and early readers to the 2,100-page American Heritage Dictionary of the English Language, at Houghton Mifflin Company, later Houghton Mifflin Harcourt (HMH).Footnote ¹ For thirteen years I served as Executive Editor, overseeing the fourth (Reference PickettAHD4, 2000) and fifth (Reference PickettAHD5, 2011) editions of the department’s flagship dictionary, as well as most of its other publications. While these experiences in historical and synchronic lexicography obviously shared some deep commonalities in the search for the meaning of words, the circumstances and pressures involved were very different. Most editors of historical dictionaries focus their attention on a single, very long-term project, while editors of synchronic dictionaries, particularly when working for a large publishing corporation, typically juggle multiple projects at the same time.

This chapter elaborates the considerations that lexicographers of any stripe have to confront when conceiving, shaping, and editing a dictionary, from the scope of the project (what kind of language is to be covered) and its audience (what are the assumptions about the end user that will determine the kind of information presented in the dictionary’s entries and supplemental materials) to the use of evidence and the prioritization and assignment of work to staff members (see Convery, Chapter 27, this volume). Many of the preoccupations of commercial lexicographers, such as gaining the enthusiasm of marketing and sales staff and assessing the current status of competing titles, do not concern the historical lexicographer, for whom marketing (except as a pitch to potential funders and as an exercise in reassurance to academic committees) has no bearing and for whom the competition does not exist, has a very different scope, or is so old as to be obsolete. Yet the editorial process of making judgments about the meaning of words and of organizing and presenting that information in clear and understandable form are just as important for each kind of endeavor.

28.2 Scope and Audience

Editing a dictionary can only begin after making a host of interrelated decisions that will both shape the project and (hopefully) ensure its success. The senior editors will inevitably have to summarize these decisions in a presentation describing the project and submitted to possible funding sources, such as publishing executives, university officers and trustees, foundation boards, and government agencies.

Fundamentally, the editors, working with a publishing executive, must describe the scope of the dictionary – what language varieties will be “covered,” what vocabulary subgroups will be treated, and what the rationale for the scope (versus another range of vocabulary) should be. Flipped on its head, the scope will be defined by its limitations and what will not be included. Will the dictionary be limited primarily to a single national variety of the language? To what extent will other nations’ use of the language be covered? Will regional or nonstandard dialects be treated? For commercial dictionaries, editors must decide what a user will need to look up and what might interest users who want to learn more about the language. This latter category is subjective, of course, as we will explore later in the chapter. For historical dictionaries, the scope largely entails the chronological span of language to be covered, and in what detail, or how frequently that language should be attested.

Equally important is the publication’s audience or market. Who is the likely user, and what incentive is there for that user to purchase or subscribe to the dictionary? Most historical dictionaries have an obvious market in research libraries and scholars who study the time period addressed in the lexicon. The audience will likely encompass generations of scholars, as the product is assuredly a unique resource that will not be re-edited for a long time, owing to cost and the protracted nature of the editing. The funders’ interest lies largely in backing such a project as a tool for advancing knowledge and enhancing the reputation of the funder organization. Makers of commercial, synchronic dictionaries must convince purchasers in different markets that the product is a good investment at the proposed price point. These markets exist at various levels beyond the envisioned individual consumer, who may be a direct user or a relative or friend of the user (the “gift” market). Publishers must also convince booksellers, wholesalers, office products companies, “big box” stores, websites, libraries, college professors, schoolteachers, and school boards that the dictionary is a valuable resource well worth the price. Marketing and salespeople assemble different deals for each market, with discounts for volume purchases (see Morse, Chapter 29, this volume).

The trade (i.e. bookstore) business is unusual in that most retail outlets have the option of returning unsold books to the publisher for a refund, absorbing only the cost of return shipping. At the same time, publishing executives have learned to cast a critical eye on inventory that they must pay to warehouse, so coming up with a realistic estimate of the likely number of books that will sell in a given period is very important, as challenging as the task is. Publishers must also take into consideration the costs of printing the book. The greater the number of copies printed, the lower the per-unit cost, so ideally the publisher will pay to print a large number of copies that sell in a reasonable amount of time and will not have to be warehoused. This situation gives outsized importance to large chain retailers likely to take substantial orders and to outlets willing to make final, nonrefundable sales, such as wholesalers like Costco. Representatives of all these businesses have an ongoing dialogue with a publisher’s sales and marketing staff, whose opinions are taken very seriously. Selling or licensing an electronic dictionary obviously eschews the inventory and warehousing concerns that bedevil book sales, and here the customer base can be very different, primarily focusing on websites and businesses and other institutions. My point is that all of these factors come into play in the preliminary discussions that a dictionary publisher will have with the house’s sales and marketing executives about a proposed dictionary or reference title. These people must be on board and believe that the project makes financial sense and that the end product can be sold in adequate numbers to bring in enough revenue to be worth their time and effort.

Also fundamental to marketing, and therefore to editing, is an analysis of the existing competitive products for a proposed dictionary, a task that inevitably falls to editors since marketers don’t routinely spend time immersed in the various dictionaries currently for sale. Thanks largely to the phenomenal success of the dictionary format that the public sees as the hallmark of authority (headword, part of speech, definition, etc.), users have a difficult time distinguishing among available dictionaries. Businesspeople tend to view dictionaries as commodities that sell primarily by their affordability or lowest price possible. Still, the new product must have distinguishing features that salespeople can tout over and in contrast to those of competing products. That is, editorial decisions about the headword list and other content, structure, defining style, and special features (such as usage or synonym notes) are guided greatly by concerns about the dictionary’s ability to compete in a crowded marketplace. Depending on an existing dictionary’s previous sales, editors of a new edition must decide to what degree they must continue the look and feel of the existing product, and they will have to explain and argue for any departure from that look and feel if the brand has established success in sales (see Morse, Chapter 29, this volume). And on the basis of the new dictionary’s place relative to its competitors, editors may find themselves writing prefaces, jacket and web copy, marketing materials, media pitches and scripts, and posters to delineate and promote the brand. Editors of historical dictionaries are usually spared these responsibilities because there usually isn’t any competition to be concerned with.

Most dictionaries have to assume a certain level of education among users, who are expected to have some familiarity with dictionary conventions and abbreviations, alphabetical arrangement, parts of speech, certain fundamentals of grammar (like transitivity of verbs or mass/count distinctions in nouns), and pronunciation symbols. The value of the classic standardized dictionary format is that it makes entries easier to read – the user knows what kind of information to expect when examining an entry, and even though a dictionary includes an enormous variety of entry types, the user can comfortably move across the vocabulary gaining knowledge. Imagine the difficulty of using a dictionary in which the entry format varies unpredictably! The conventions of using and understanding a dictionary seem obvious once they are learned and internalized, but for children and certain other users, these conventions can be bewildering and must be included in the dictionary only partially or gradually over a series of increasingly sophisticated offerings, as for elementary, middle school, and high school.

Standardized dictionary format also functions as a kind of data compression technique, allowing enormous amounts of material to be compressed into individual entries (as long and cumbersome as they might sometimes seem). After all, any dictionary entry could be rendered as a discursive essay, which might be interesting to read but would probably overwhelm or annoy users as burdensome and tedious. Over the course of a lexicon, such discursion would expand the text to immense size. This is not to say that discursive writing has no place in dictionary entries. On the contrary, discursive or “full sentence explanations” (as in “When you wake or when someone wakes you, you become conscious again after being asleep.”) have the virtue of being more conversational and hence less intimidating, which may be why they originated in dictionaries for children. Sentence definitions provide “a convenient way of representing context, function, and pattern” of word usage (Reference MoonMoon 2009).Footnote ² Their drawback lies chiefly in their verbiage, which can lengthen an entry and make it more unwieldy to read, which is why it has mainly been used in learners’ dictionaries with a headword list of a few thousand words.

In short, most editorial decisions, such as ones in defining style, entail tradeoffs, among which may be fewer dictionary entries overall or more cumbrous navigability.

28.3 Schedules

Another facet of pre-defining work that shapes editing is the development of schedules for the different aspects of the project. For print dictionaries, assuming publication of the entire work at one time (and not piecemeal in sections, as many historical dictionaries are released), a publication date must be chosen. From that date, the phases of the project are walked back to the point when editing proper may begin. Time allotments must be estimated for shipping, manufacturing, page composition, and proofreading, as well as for the editorial components of the project: general defining, technical defining (typically including specialized vocabulary review by outside consultants), pronunciation, etymology, art research and review, special features, and front and back matter. These components must have budgets created for them as well, and it is crucial for a managing editor to track progress of editing against the projected schedules and expenses (of which some are non-editorial, such as the cost of paper, printing, and warehousing). If there are discrepancies, the project scope may have to be reduced or additional sources of revenue found. As we shall see, schedules and budgets may be extrapolated from an editorial time trial, in which editors work on certain kinds of entries from start to finish, as a reality check against dreamy visions of expeditious success. Experience with previous projects can inform the whole process and give management reasonable confidence in the plan.

Missing a publication date can have disastrous consequences for the publisher (see Morse, Chapter 29, this volume). Salespeople have to convince bookstores, office products vendors, and other sales points to take a certain number of copies. Vendors expect delivery at a certain time, such as June for the “back to school” market, and if the dictionary comes out later than promised, the vendors may refuse to receive as many products as agreed. They may even refuse delivery altogether. Diminished distribution of dictionaries puts the publisher’s substantial investment at increased risk. Commercial dictionary editors thus face intense deadline pressure, and they must carefully design their project so that it does not fall too far behind schedule, given a certain amount of wiggle room that sensible managing editors usually build into the schedules. By the same token, dictionary editors must also plan the project so that it does not become unwieldy, and end up noticeably incomplete, with little for marketers to crow about.

For historical dictionaries, deadlines mostly exist as meetings seeking funding continuations and renewals. The senior editors must demonstrate some kind of quantitative progress and be able to plausibly explain why the original project deadlines are unmet.

28.4 DTD and Database

The electronic aspects of dictionary making have their own schedules and budgets. The structure of the dictionary database, spelled out in a document called a DTD (“document type definition”) must be designed with an eye both to enable searching by end users and by editors themselves and to facilitate typesetting and display on a screen (see Convery, Chapter 27, this volume). DTD authors must decide which and how much information editors should key and which elements of what the user sees will be generated by a program that converts data in database fields (and sometimes the fields themselves) to type. So, for example, a variant spelling might be entered in one field that generates an italic “or” before its input data, or in another field that generates an italic “also,” reflecting relative frequency in comparison to the headword. The value here is in consistency of entry structure and in the minimization of keying errors that (assuming proofreaders catch them) will have to be corrected in page proofs at considerable expense. The DTD must anticipate every possible permutation of data fields in the myriad entries in the dictionary. The DTD must include metadata fields, such as a control number to ensure correct alphabetical order of headwords, a signoff to indicate which editors have worked on an entry and when, a classification field for designating the subjects the headword is used in, and fields that flag entries for various reasons, not least of which is to allow managers to track editorial progress. Also important is a messaging field to allow editors to annotate, explain, and query what has been done to an entry.

Figure 28.1 shows a portion of the AHD5 DTD written in SGML. The metadata fields (review, mdate, message, ok, etc.) do not appear in the end product but are for tracking which entries need review, identifying which editors have worked on an entry, and communicating between editors and reviewers. “Seg” stands for “segment,” and “rtseg” for “root segment,” the parent field containing the entry word, variant spellings, and pronunciation (these fields do not appear in the figure). A plus sign indicates that the field is required in an entry but can be repeated. So the pseg for part of speech is required, but some entries (in this entry scheme) have multiple parts of speech. A question mark (?) indicates that the field is optional. Thus, the etyseg for the etymology is optional, as not all entries get etymologies. An asterisk (*) indicates that the field is optional but can be repeated. So the pvseg for phrasal verbs can be repeated, as not all entries have phrasal verbs but some have many. The definition segment (not shown) is subfielded under part of speech.

Figure 28.1 AHD5 DTD for metadata and major divisions of an entry.

Figure 28.2 shows the DTD of the definition segment (“ds”) and its possible subfields. The definition segment is a subfield under the part-of-speech segment (as no definition can appear in an entry unless listed under a part of speech). The txt field (“tx”) contains the text of the definition, “qu” contains a quotation, and “so” is the source (author). Examples of usage or “verbal illustrations,” as they were referred to, appear in the “ill” field. The “xref” fields are for various kinds of cross reference: for example, “vref” for a variant spelling referring to an entry with the preferred spelling, “seeref” for a “see” cross reference, and “synref” for a cross reference to an entry that has a synonym paragraph.

Figure 28.2 AHD5 DTD for the definition segment of an entry, showing all possible subfields.

The DTD structure of a dictionary entry may seem unduly complex at first glance, but it is designed to enable a variety of outputs, both print and electronic, enhancing the product’s value. Additional search capabilities can be added as an electronic product is developed. The online OED, for example, has greatly facilitated entry navigation by displaying in a column on an entry page a list of links to compounds and phrases within that entry so users can jump directly to a component of an entry of immediate interest (such as the phrasal verb pick out or pick over in the entry for pick v¹). These compounds and phrases are data entered into various DTD fields. This functionality relieves the user from having to scroll through the entry to find the point of interest (see Figure 28.3). Hide and reveal functionality can also facilitate navigation, as by hiding examples and quotations so the definitions can be scanned quickly.

Figure 28.3 Initial screen showing OED entry pick, verb 1, first in a series of homograph entries, used with permission. Column to the right lists links to components within entry. “pick, v. 1.” OED Online, Oxford University Press, March 2021. www.oed.com/view/Entry/143373.

Editors work in a database, adding segments (or copying existing entries) and writing text where needed. Figure 28.4 shows most of the AHD5 database entry for the verb and noun straggle, a relatively simple entry. (The data in the metadata fields for sign-offs and messaging have been deleted for privacy concerns.) The DTD provides the structure, and the database provides the working environment for editing.

Figure 28.4 AHD5 database entry for the entry straggle.

Straggle has three definition segments, the first of which contains two subsenses (“sds”). All definitions are followed by quotations whose authors’ names appear in print, but other bibliographic information does not. Figure 28.5 shows the straggle entry as it appears in the online version of the dictionary.

Figure 28.5 The entry straggle as it appears in the online version of AHD5. “Straggle” from The American Heritage Dictionary, fifth edition. Copyright 2016 by Houghton Mifflin Harcourt Publishing Company.

Reprinted by permission of Houghton Mifflin Harcourt. All rights reserved.

Once a database structure has been created, software must be created to enable proper editing. This software should allow editors to search the database by any number of fields (by character strings in the definition field or labels in a label field), enabling group review of entries. Subject classifications such as medicine, physics, and sports need to be keyed in a metadata field (“cl”) so batches of entries can be extracted for editing by specialist editors or sent to expert consultants for comment. A single entry may have many classifications, depending on its definitions. Classifications are also useful for tracking entries that might be prone to typesetting errors because of a special character or some other issue. This search dimension frees editors from being tied to alphabetical runs, facilitates the use of consultants for vocabulary in different subjects, and helps ensure greater consistency of treatment across entries in disparate parts of the alphabetical list. Stylistic and label changes from one edition to another, or even at different stages in a single project, can be done quickly and efficiently.

In the pre-electronic era, this kind of work was cumbersome and difficult and had to be handled with handwritten messages on index cards (or some similar method) to be placed with the materials assembled for other headwords in the alphabet. These cards or sticky notes could get lost or be subject to misinterpretation, and of course weren’t searchable. A single editor could edit certain groups of words, to be sure, but such work was much more difficult to coordinate, opening the possibility for errors of omission or diminished oversight.

28.5 Evidence

Also preliminary and essential to editing is the assembly and organization of the evidence on which definitions, attestations, and related information are to be based. Traditionally, this is called a reading program. Individuals are assigned to read selected texts and identify or copy out passages containing a word to be filed under a dictionary headword. But before this laborious work of data collection, the dictionary staff must identify and select preferred editions and manuscripts of the works that will be read. These works must fall within the dictionary’s proper scope. That is, they must fall within the date range covered by a historical dictionary, such as the 1100–1500 range for the MED, and must represent the varieties in the dictionary’s purview (e.g. English, not Scots; Standard English, not slang). In historical lexicography, the works to be read must be assigned dates of authorial composition and ideally dates of publication (as when a manuscript was written or copied), and a method must be devised for indicating these dates when a passage is cited in the dictionary’s text. Also crucial is the assignment of a standardized designation of a work. The designation is really a reference to an entry in the bibliography, where more expansive information about the work is presented. The MED staff devised an ingenious, space-saving method for doing this, in which each work is cited by means of a “stencil.” For instance, Chaucer’s Troilus and Criseyde is designated by a1425 (c1385) Chaucer TC (book and line number follow), where the manuscript date “ante 1425” is followed by the date of composition “circa 1385,” the author’s last name, and an abbreviated title. The MED bibliography informs us that Fred Norris Robinson’s edition is being referenced by this stencil. Still further, texts may be assigned to particular regional dialects, if that is possible to do with any confidence. This is another dimension of a historical dictionary that requires an enormous amount of research prior to editing.Footnote ³

Once the slips containing examples of words have been copied out, they must be assigned to and assembled under headwords. At the MED, for instance, possible headwords were indicated on large file cards in oversized shoe boxes, and the slips assigned to them (sometimes only a few, sometimes hundreds or even thousands) were placed after the file cards. Editors typically chose a box of words to edit, and editorial progress was tracked by the number of boxes that primary editors had finished editing and by the number that review editors had reviewed. Once approved by review, the citations in each entry had to be proofread against the original texts to ensure mistakes in copying had not been made. Cross references for variant spellings were then generated and included in the sequence. Then, the full text of the entries, all handwritten on paper slips, had to be typed up as camera-ready copy, proofread, corrected, and sent to the printer.Footnote ⁴

The modern method of assembling evidence for dictionary editing takes the form of a corpus of digitized texts, where the number of words can run into the billions, and editors can examine the equivalent of citation slips through a KWIC (key word in context) concordancer, where the word under investigation appears in the center of the screen with strings of words supplying context on either side. A well-constructed corpus equipped with well-designed software tools can also allow an editor to analyze grammatical information, relative frequencies of use, levels of formality, and much more in ways that traditional editors could only dream of. The construction of a corpus is a tremendous undertaking but provides an immensely powerful tool for lexicographers to efficiently examine a word’s appearance in a huge variety of contexts.

Corpora must be designed to represent different subject areas (biology, cooking, medicine, military, etc.), types of sources (newspapers, reports, spoken language transcripts, etc.), and linguistic registers (formal, casual, intimate, etc.). The corpus compilers must select which texts they want in the corpus, and they must acquire legal rights to use the texts, and the texts must have a reliably digitized form to avoid introducing errors or causing omissions. In addition, texts must be tagged for parts of speech (programs exist that automate this process, but some manual error correction is usually required) and lemmatized so that different inflected forms of the same word can be viewed without being lost or confused with other words. All this work requires a number of non-editorial people with considerable skill in software and considerable understanding of language. A system of citation and bibliography similar to the one described for historical dictionaries above must be developed if the materials are going to be quoted in the dictionary itself.Footnote ⁵

Today, American lexicographers can take advantage of a variety of corpora available on the web, such as the Corpus of Contemporary American English (COCA), a sample of whose KWIC concordance (for the word saunter) is shown in Figure 28.6, and the Corpus of Historical American English (COHA). These resources would boggle the mind of editors for most of the history of American lexicography, some of whom spent considerable effort trying to convince their publishing executives to invest the enormous sums required to construct and maintain a corpus. But the longstanding unavailability of corpora for American lexicography forced American editors for decades to extemporize as best they could and find other ways to assemble materials electronically. A publishing house like Houghton Mifflin Harcourt owns the rights to a considerable library of fiction and non-fiction titles and of other publications (such as magazines and children’s books), and for many years these materials have been created and stored in digitized format as part of the standard publishing process. For the AHD, we decided to make the most of the materials we could get access to, and thanks to our department’s tech person, we were able to construct a corpus that included full published texts and our digitized citation files. Although very modest in comparison to corpora assembled over decades and containing billions of words, this in-house resource nonetheless presented a valuable tool for lexical research, especially when supplemented by online databases like newspaper archives and Google Books.

Figure 28.6 Sample screen KWIC concordance in COCA; the verb saunter highlighted in immediate contexts.

28.6 The Headword List

Let’s assume we have a dictionary database, editing software, and materials assembled for use in creating or updating a dictionary. The next step is to decide what the headword list will look like. Historical dictionaries build their headword lists from scratch, out of a reading program or an electronic corpus. Relative frequency of use may or may not be an issue. The MED, for instance, included every lexical item discovered in its reading program, even if the word was attested by a single citation or its meaning could not be made sense of and had to be guessed at. Certain items were excluded, such as elements of surnames and place names that were attested by themselves without further use in prose or poetry. On the other hand, for a historical dictionary spanning many centuries like the OED, there are simply too many rare attestations and ephemeral coinages for editors to spend time analyzing them, when so many other, more important words must be reckoned with.

With the plethora of existing dictionaries available, commercial dictionary editors can either purchase a headword list or even copy one as the basis for making a “new” dictionary. Most likely, a headword list already exists in a previous edition of a dictionary owned by a publisher, so editors face the task of enhancing or abridging the list, depending on the product being created. A word list can be extracted from a corpus or a digitized citation file of works read in the reading program. This list can then be compared to the headword list of an existing dictionary. The words that don’t match constitute a list of potential new words for a new edition. The editorial staff can then decide which of these words are worth including. In my experience, assigning subject areas of vocabulary to specific dictionary editors is exceptionally important. Such editors familiarize themselves with the vocabulary and treatment of certain subjects (such as biological terms or sports terms) in existing editions (and in competing dictionaries as well); they can then identify which new words are important enough for entry. At Houghton Mifflin Harcourt (HMC), a database engineer created such a list, and it was circulated among the editorial staff, who voted on the words they felt should appear in a new edition, initialing their selections. Senior editors then approved certain candidates for inclusion, and entries for those words were created in the database.

We typically made decisions about which candidates to include based on criteria such as frequency of use, saliency or sudden prominence, recommendations of subject-area experts, an editorial decision to include vocabulary from subject areas not covered in a previous edition, the need to fill in gaps of coverage, and just sheer joy of showing the creativity inherent in the lexicon. Many of these words were not immediately new, and many had established records of attestation and were either missed or discounted in the making of a previous edition, but usually quite a few were recent coinages. And of course many candidates fulfilled more than one determinative criteria. For Reference PickettAHD5 (2011), we added entries like MRSA and K street because of their prominence and frequency; crowdsourcing, podcast, and vlog because of their growing and expansive use; metabolome and zeitgeber because of their importance in their fields. We decided to add vocabulary from areas not much treated in previous editions, such as Indian clothing and world music, along with the basic lingo of the sport of lacrosse, which was rapidly growing in popularity in the US. Many of these then-recent coinages, especially compounds like face-plant, gap year, must-have, and nutjob, seemed unremarkable not long after. The same process was used in adding new senses of existing entries, from the real-estate sense of flip to the Internet sense of cloud.

Imagine, if you will, that you are one of the editors responsible for making these decisions. Let’s assume that we have assembled a raw list of words to be investigated further for defining. This list will change as editing progresses. New words will be added as they are discovered and deemed worth including; others may be dropped. Deleting headwords (normally to save space in print dictionaries) is never an easy choice, as one never knows when a word considered obsolete will resurface in the public eye. A premier example is chad, referring to the tiny bits of paper punched out of data cards back when such cards were used in computer technology. What could be more insignificant? the editors of AHD4 thought in the late 1990s. If ever there were a word that could be cut from a dictionary, chad was it! Somehow, in deference to those studying the history of computers, we left the word in. Who could have predicted that chad would figure prominently in the US presidential election of 2000? There are many obsolete words (and senses) that are tempting to remove from a dictionary, and editors have to gauge the dictionary’s commitment to the past. Almost no one knows what a shot tower is today – a tower used to make shot by pouring molten lead through a sieve and letting the lead fall into a vat of water. Who would need to look up such a word? A reader of Huckleberry Finn might, as a shot tower appears in a scene of this classic American novel. Is this reason enough to keep the word in the headword list? For AHD5, a dictionary conceived to be read as well as momentarily consulted, I thought so.

For historical dictionaries, entry words sometimes must be relocated in the alphabetical sequence, owing to the vagaries of spelling and the influence of dialect. The Middle English verb stonden ‘to stand,’ was originally placed in the sta- sequence, but examining the citation slips showed that the sto- spelling was predominant, and so the headword was repositioned later in the alphabet. Needless to say, cross references abound in historical dictionaries.

The next step in our pursuit of a headword list requires that editors decide what the list should look like – that is, whether each part of speech should be a separate lexical item that deserves its own entry with a different homograph number (e.g. are there three separate entries for spring verb, noun, adjective?) or whether different parts of speech will be bundled under a single entry word. Both approaches have advantages and disadvantages. Certainly, a historical dictionary will need to treat each part of speech as a separate entry since the task at hand is to show the development of each part of speech through time, and trying to bundle these different words under a single entry would result in entries that are so long as to be intimidating and possibly unusable, to readers. The drawback is that the number of homograph entries under this scheme can become extensive. In the OED, for example, the character sequence pat has eight homograph entries: five nouns, two verbs, plus a combined entry for adverb and adjective. Word separation of homographs is largely determined by etymology (e.g. school ‘institution for instruction’ comes from Latin and Greek, where school ‘group of fish’ comes from Middle Dutch). Homograph order may be determined by earliest attestation or by relative frequency or saliency. So, in the OED, page¹ referring to a boy or servant comes before page² ‘a leaf or side of paper.’ In AHD5, the sequence is reversed. Sometimes, in a commercial dictionary, saliency of homographs is obvious. For instance, in AHD5, the verb meet¹ ‘to come into the presence of,’ a substantial entry with many senses, precedes the skimpy adjective meet² ‘fitting or proper.’

Homographs are typically given etymologies in dictionaries by way of explaining the fact of there being more than one entry with the same spelling. The decision to enter separate parts of speech as homographs increases the effort to research and explain where a homograph comes from. In this regard, bundling multiple parts-of-speech under the same headword might be thought to ease the burden on the etymologist, although for AHD flagship products, etymologies were carried all the way back to their Indo-European roots and (for AHD4 and AHD5) to Proto-Semitic roots, aspects of etymology that are not treated in most other dictionaries.

28.7 Structure of Entries

After making decisions about the arrangement of the word list, the presentation of elements within an entry presents another series of decisions. Should senses be ordered by earliest attestation in the recorded literature, or should the sequence of senses be arranged by frequency of attestation (most frequent first in the sequence), or by some “logical” presentation supposed to show how one meaning (or group of meanings) is related to another? Again, each manner of presentation has its virtues and shortcomings. The dating of texts, and especially manuscripts, is often uncertain, and the survival of manuscripts is often the product of chance. Many words are assumed to have been in oral use for some period before the date of first recorded occurrence, and that first recorded use of a word is often an obsolete sense or one that sees rare use today. Forcing the reader of a commercial dictionary to plow through a long entry to find a predominant contemporary meaning might seem like a disservice.

On the other hand, allowing readers to see how recorded usage for a word develops (as in the OED) is a valuable lesson. A logical sequence of senses, in which groups of semantically related senses are collected together, probably with the most common one presented first, might seem like a fine solution if it weren’t nearly impossible to implement consistently. MED entries are structured on this “logical” principle primarily because of the difficulty in determining a reliable earliest occurrence for a word and because of the uncertainty of identifying which meaning grew from which (or from which of several meanings). There are many other questions of presentation of definitions that must be addressed. For instance, should verbs be treated in separate sequences for their transitive and intransitive properties, as if these were different parts of speech, or should the editors group transitive and intransitive senses together according to semantic closeness? Will a user appreciate the difference? The important point is that a dictionary entry must have some principle for being ordered, even if the principle cannot be applied with perfect consistency across entries.

The editors of the American Heritage dictionaries, created for educated adults rather than for children, preferred the “logical” principle, with transitive and intransitive entered as separate “sub” parts of speech. Merriam-Webster (MW) employs a hybrid approach in which the main senses of a word (as 1a, 2a, 3a …) are presented in historical sequence, but subsenses (1b, 1c, 1d …) are listed logically by semantic closeness under the first in the semantic group. The marketing people at HMH viewed the AHD logical scheme as a key selling advantage over MW. The editors at Oxford University Press devised an ingenious alternative to the traditional numbered sequence of definitions in their New Oxford Dictionary of English, adapted for the American market as the New Oxford American Dictionary. They identified “core meanings” sequenced by frequency evidence from their in-house corpus and listed subsenses under these “core” umbrellas. Only the core meanings were given numbers, simplifying the presentation of senses. These Oxford dictionaries placed greater emphasis on phrases and combinations than their other American competitors, no doubt reflecting Oxford’s longstanding experience in making dictionaries for English as a Second Language users. To deploy this treatment in a print dictionary, however, Oxford must reduce the number of definitions and eliminate some less prominent vocabulary. This “core meaning” innovation was touted as a main virtue in the earlier years of the dictionaries’ release in the 1990s, but now it does not appear among the products’ key features listed on the Oxford University Press website. Perhaps the value of this style of entry organization proved too difficult to explain succinctly to users.

28.8 Style Manual, Writing Sample, Time Trial

Having made higher level decisions about the nature of the headword list and the arrangement of elements in an entry, senior editors must elaborate rules for consistent treatment of all the possible elements in a style manual, as well as consistent treatment of similar kinds of entries (such as chemical elements or color terms). Information about proper coding or fielding in the database must be delineated to prevent errors and frustration when an improperly coded entry can’t be saved into the database. (The entry will not “parse,” as the lingo goes.) At HMH, the editors used a style manual of this sort, plus a defining manual providing guidance for proper defining, a manual for editing pronunciations using a set of symbols, and an etymology manual, since etymologies were a separate realm governed by a single editor navigating complex formatting and coding rules.Footnote ⁶

One might believe that, equipped with a style manual codifying the rules for editorial decision making, the editors could be let loose on defining the words assigned to them, whether by alphabetical sequence or by subject area. But no! First, the decisions in the style manual must be tried out in a sample to see if they are not too cumbersome to implement and produce a presentation that lends itself to easy use. The writing sample must include different kinds of entries that show the various elements described in the style manual. In fact, once a sufficient number of entries have been edited according to the style manual, editors using a database can for the most part ignore the manual, as it is much more efficient to clone an existing and properly fielded entry and repopulate its fields than to create an entry in a database from scratch, having to remember or look up all the standardized treatments for each element of an entry. The style manual remains important as an arbiter for rare conundrums and as a comforter for higher executives when they need to be reassured there is a method to the dictionary department’s madness.

The writing sample may be extended somewhat to function as a time trial to be assessed in relation to the schedules created for the project. After all, as ingenious as dictionary editors might be, they are easily deluded into believing what they can accomplish, and writers of all kinds never know what they are up against until they dive in and deal with reality rather than an imagined fairyland. For the MED, the trial of “the plan” for the project devised by Reference Kurath, Ogden, Palmer and McKelveyHans Kurath (1954) extended for the entire letter E, one of the smallest letters by entry count. It was only after the plan had been proven to be a workable one in E that the editors went back to work on letter A.

28.9 Design and Presentation

Once a writing sample has been created, the publisher can use it to assay various combinations of fonts to present the dictionary’s information in type on a page, and to extrapolate what the projected length of the printed book would be under a given type design. Typically, dictionaries have a sans serif font for headwords and other forms to be set in boldface (such as inflected forms and run-ons) and a serif font for the text of definitions, example sentences, and notes. Other fonts may be selected for image captions, note headings, and other design elements. The typeset design of a dictionary entry determines its readability, and readability is essential to marketability. A dictionary with an intimidating design or one that presents a column of dense type as a cloud of gray can make for a harder sell to buyers, whether to individuals or institutions. By the same token, an attractive combination of fonts can make for a far more inviting presentation on the page or screen and encourage users to explore the dictionary in greater depth. How art is presented on the page or display screen is equally important to the dictionary’s success as a commercial product (see also Adams, Chapter 9, this volume).

As for overall length, page and type design determine the number of words that can be included in a line, the number of lines that can fit to make a column, the amount of art that can be included, and the amount of material that can be included overall in a printed book, which will have a number of pages determined in part by the cost of paper and manufacturing for a given price point. Different combinations of fonts can produce a huge difference in how many entries a print dictionary can encompass. Consequently, the page and type design must be decided early in a dictionary project, and length estimates (or demonstrations, assuming a WYSIWYG program can be deployed) should be performed periodically during the editing (see also Nichols, Chapter 11, this volume).

Different fonts and font sizes remain key to readability, but also of concern are links to other entries or components of the dictionary, text and image boxes of various kinds, color screens (backgrounds) to help set off certain features from other parts of an entry. Electronic dictionaries don’t have length considerations per se (which is one of their main blessings for a lexicographer), yet aesthetic presentation for screen viewing is paramount to the success of projects with enormous amounts of text, so editors must also ensure the aesthetic appeal of display elements, which may include audio buttons for pronunciations and video players. For commercial dictionaries, advertisements might also have to be designed, possibly with the approval of the sponsor.

28.10 Task Assignments

An arrangement for reviewing edited work must also be established. This may seem obvious from the titles and positions different people have – executive editor, senior editor, editor, associate editor, etc. – but as the work is too complex and writing too easily misunderstood, every editor’s primary work should be reviewed by someone else on the staff, regardless of seniority. At the MED, there were usually two review editors, and sometimes experienced staff members were asked to review the work of junior ones in order to provide timely feedback. At AHD, senior editors did a lot of primary editing as well as reviewing, so entries had to be routed among editors according to a pre-established scheme. Typically, this took the form of a review sequence. First, one senior editor would review the structure and defining of another senior editor or of a junior editor. Once this work had been approved (often after some back and forth), the entry was forwarded to the pronunciation editor and the etymologist. Senior editors always reviewed each other’s work, although pronunciation and etymology were largely left to themselves, except in cases open to dispute. Associate editors were assigned groups of entries to define (say, grammatical terms or food terms) but normally did not review the work of others. Assistant editors did proofreading, more mechanical database edits, and some limited defining.

A commercial dictionary is really many books bundled into one – a lexicon, various note programs (usage, synonym, word histories, and sometimes others), illustrations and photographs, encyclopedic entries (biographical and geographical entries), front matter, back matter (parts of which linguists may author). All these facets of the dictionary have to be assigned to staff members and tracked by their own schedules. Different staff members must be given responsibility for different groups of entries – biology, medicine, physics, economics, food, sports, social terms, and so on – so the editing is often not (or not only) a march in alphabetical sequence but jumps around.

To be sure, similar entries should be handled by individual editors even in a historical dictionary, where the editing in alphabetical sequence is standard procedure. This is because the editing of one word (such as the preposition in) might inform the editing of another (on) or the words might be confused by overlapping spellings (such as Middle English verbs setten and sitten). This practice of defining by group, however, is often not possible because pressure to complete an alphabetical sequence and publish a section of the dictionary (as in a fascicle) can forbid the allocation of resources to parts of the lexicon that fall outside the alpha section at hand.

28.11 Prioritizing Work

With staff assignments now arranged, and a page and type design decided upon, editing proper is on the cusp of becoming a reality! But one more preliminary must be determined, at least for commercial dictionaries. The hard deadlines that editors face behoove them to prioritize the different facets of their work. This prioritization is driven largely by the need of marketing and salespeople to distinguish the product from competitors, discussed above. At the top, new words and senses will figure heavily in marketing materials. Also high on the list are new features that distinguish the product; for example, a different form of entry presentation such as the larger boldface “core meanings” that group senses together and make long entries easier to navigate in the Oxford commercial dictionaries. Other priorities might be highly visible elements of the dictionary, such as new and updated usage notes in AHD or a new “Our Living Language” note program explaining well-known features of nonstandard varieties of English.

With tasks prioritized, editors can avoid the disaster of running out of time before completing important marketable aspects of the book. Coming short on a competitive number of new words or lacking features that a salesperson can tout could destroy the dictionary department’s reputation with sales and marketing, and an unmotivated salesforce does not bode well for anyone.

28.12 Relations with Sales and Marketing

A brief digression will underscore the importance of the relations between the editorial group and the sales and marketing group, as infrequent as contact between them might be. At HMH, the dictionary publisher felt strongly that the department should have books coming out on each publishing list (that is, every spring and every fall), to keep the salesforce aware of the editors’ importance to their success. “So they don’t forget about us,” was the sentence I recall. This meant that aside from the substantial work the editors were doing on their most prestigious dictionaries, they had to create, revise, or edit other works. The HMH dictionary department was responsible for multiple lines of dictionaries, from children’s dictionaries and paperbacks to college dictionaries and usage books, all selling at different price points, and all of these books had to be maintained and given copyright updates (as well as new editions) at regular intervals. The department (of roughly fifteen, including production staff, for the thirteen years when I was executive editor) also published authored books, and most of these were edited, sometimes extensively, in house. On top of this, we sometimes got suggestions for dictionaries or other books from the marketing or office products departments, whose clients had predicted handsome sales, and the publisher viewed these requests as gift horses that could not be turned down.

During the five years we worked on AHD5, for instance, our staff of fifteen people published about forty other titles. To be sure, many of these projects were copyright updates of existing titles, such as our children’s line of dictionaries, where a recent copyright date (along with a new book cover) was considered important to sustaining sales. We usually treated copyright updates, typically undertaken when inventory was dwindling, as an occasion to make as many fixes and additions as possible, given a compressed time frame. If a page needed a fix we would look for enhancements that might be made on the same page, as we were paying to reset and reprint it anyway. We also added some recently prominent words to help with promotion. Some copyright updates of our more expansive dictionaries included 200–500 new words and senses that we were developing for the next edition.

A number of titles published during this period were short paperbacks created to help boost the AHD brand, such as our 100 Words series (e.g. 100 Words Every High School Graduate Should Know, 100 Words Almost Everyone Mispronounces, 100 Words Almost Everyone Confuses and Misuses). Most of these works were researched and written by our own staff members, adapting existing materials when possible. Editors in our department also worked on new books acquired from outside authors, or updates of previously published authored titles (as for business communication in the era of electronic messaging), under the assumption that these authored works would involve less work than a new dictionary. This pace of work was the norm throughout my entire thirteen-year stint as Executive Editor, and for some years before that.

In between these roughly five-year periods when we worked intensely on the flagship AHD dictionaries, we edited “derivative” products (e.g. a paperback edition, a college edition, a dictionary for elementary or middle school students), and we expanded the department’s line of products into new areas, such as medical dictionaries, science dictionaries (for adults, for middle schoolers, etc.), picture dictionaries (often requiring that editors conceive effective images and write instructions for artists), a technology dictionary, thesauruses, Spanish dictionaries, and so on. During these periods, we also published more authored books than we would when working on our flagship dictionary. Naturally, the idea was to mine the vocabulary in the dictionaries created during this period for entry in the next edition of the flagship dictionary, rather than duplicate the effort down the road. And of course, we continued our ongoing research into new lexical developments, and our ongoing suggestions for improving existing dictionary entries and features.

Many of the titles published during this time required preliminary database work (adapted structure, new features, etc.), and all required the usual sequence of tasks for print book publishing: page design and font selection, art research, table design and creation, cover and jacket design, and the proofreading and review of multiple passes of page proofs. Not all of this work was done by our department (we hired freelance artists, engaged book designers, and so on), but the great bulk of the work was done by our group of fifteen people. The pace of work, always under publication deadlines, was intense and exhausting, very different from the steady plugging away that editors do when working on a single historical dictionary.

While some of the titles we published at Houghton Mifflin Harcourt were sudden additions to our lists – such as a book proposal that came tumbling over the transom and seemed worthy, or some titles urged upon us as eminently saleable by the marketing or office products departments – nearly all of the dictionary department’s work had to be planned and projected for release on one of the spring or fall publishing lists announced by the Trade (bookstore) Division to booksellers and wholesalers and to other Divisions of the corporation (most notably School and College). Accordingly, the publisher, managing editor, and certain other senior staff members devised a five-year publishing plan for the department, which we shared on occasion with the senior marketing staff and which we regularly revisited and revised. Usually, the first two years of this plan were fairly realistic, given the length of time it takes to publish dictionaries, but looking out three years or more, the plan became increasingly aspirational. Sometimes, as the editors saw that work on the most important projects fell behind our projections, we had to deliver this bad news and ask that some other projects be delayed or dropped (to our publisher’s immense chagrin), but for the most part we stuck to the original plan as long as we could.

28.13 Global Edits

Once shielded by a prioritization scheme of this type, commercial lexicographers are now ready to begin defining – almost. Before the editors can be let loose on their arduous task, at least for a new edition of an existing dictionary, it makes sense to make global edits. These may include systematic editorial changes like revising the hyphenation of headwords, inflected forms, and run-ons. They may include stylistic changes to formulaic definitions (such as adjectival formulas, “Of, relating to, or …” or interjections, “Used to …”). They may include changes to an existing configuration of labels that don’t require editorial judgment (such as substandard to nonstandard, or obsolete to archaic), and so on. Some of these edits may have been intended for the late stages of the previous edition of a dictionary but were never got to, as time ran out. The key lesson is the importance of having a dictionary that is as consistent as possible for a starting point, so that editors are not introducing errors by copying or modeling work based on discarded practices.

28.14 Writing Definitions

The plan having been shown to be workable, assignments made, and priorities determined, now at last defining can begin. Guidelines for good defining (by genus term and qualifiers) have been laid out by Ladislav Reference ZgustaZgusta (1971, 257–258) and elaborated by Sidney I. Reference LandauLandau (2001, 157–189) and Andrew Reference HawkeHawke (2016, 176–202),Footnote ⁷ and I do not have space to recite them in detail here. The Defining Manual for AHD4 and AHD5 runs to thirty-two pages. Here are some selected headings from the document:

• Defining from Citations

• Discovering Meaning

• Overgeneralization or Excessive Sense-Splitting

• Defining, Not Discussing, a Word

• Analytic Definitions and Precision in Selection of Genus Terms

• Synonym Definitions

• Binding Substitute [i.e. definitions that govern a number of subsenses]

• Circularity and Repetition of Headword within Its Own Definition

• Qualifiers

• Defining by Part of Speech

• Defining Figurative Uses

There are two methods for creating definitions: from citations and by technical expertise. Most dictionary editors lack the expertise to define technical or scientific terms on their own, so whenever possible the advice of an expert in a given field is sought. For a revised edition, subject areas may be prioritized, as certain fields (like biology) are more given to change than others (like mathematics). Ideally, an editor will have some knowledge of the subject area and be able to identify a consultant willing to take a thoughtful approach and make revisions and suggest additions. The editor will then rework the consultant’s submissions to fit the project defining style for the title in question. Many words have a technical sense and are used more loosely or metaphorically by the general public; for instance, critical mass, parameter, and tornado. For these words, either multiple editors must work on the entry, or the technical definer must wear two hats.

The main form of defining is by citations – that is, by making inferences about examples of a word’s usage. This process has been ably described by Sherman M. Reference KuhnKuhn (1975, 13–18), Sidney I. Reference LandauLandau (2001, 200–202), Andrew Reference HawkeHawke (2016), and Iztok Reference KosemKosem (2016). Basically, the editor examines as many examples “as time allows,” in Kuhn’s generously vague phrase (as one can always seek more evidence unless the corpus of the language in question is very small), and then groups these examples by common property. This property is often determined by other words associated or collocated with the word being defined. For instance, citations for a verb may be sorted by what kind of subjects and objects are used with the verb. Personal or animate subjects (The horse jumped across the stream) may be sorted from inanimate ones (The needle on the dial jumped). Objects of various types tend to generate different meanings. Picking beans is different from picking the strings on a guitar, even though both motions involve using the fingers to touch or grasp something. Context can provide another basis for discovering meaning. As Kuhn puts it, “In nautical language, line has a special sense which it does not have in other fields.”

The process of explaining these groupings by definitions is usually a gradual one in which editors draft wording that they feel best describes the meaning common to each group. To be sure, in a historical dictionary, many subsenses may lack definitions as they list transparent collocations and phrases and do not need to restate the main definition for each subsense.

Developing a definition can be a mysterious process, and whether a separate sense emerges from the data an editor examines depends on how the editor conceives the intended user’s needs or interests. In Middle English, the noun tail can refer to the hindmost part extending from the rear of a mammal, reptile, fish, bird, or devil, and in the MED each of these senses has its own definition with an array of quotations demonstrating it. In a synchronic commercial dictionary, these variations are normally subsumed under a single, general definition. But it might be important to a scholar of the Middle Ages to see how medieval writers made mention of tails on different animals and especially on the devil. Many words (such as pronouns and function words) in a historical dictionary have numerous groupings and abundant quotations that illustrate the word’s grammatical development through time (e.g. as direct object, indirect object, the object of a preposition, etc.). In the MED, the entry for the preposition of, for instance, has thirty-two main senses (with numerous subsenses) illustrated by 1,934 quotations. In AHD5, of has twenty-one main senses (with two additional subsenses) illustrated with twenty-four examples.

Clearly, the main goal of a historical dictionary is to show in organized and accessible fashion how words are used through the period in question. Indeed, for A. J. Aitken, editor of the Dictionary of the Older Scottish Tongue, the definitions “remain subservient to the citations themselves,” serving “as finding-aids or sign-posts to particular sections of a long entry.” Definitions

Aitken’s skepticism about the precision of meanings in a historical dictionary (suggested by the quotation marks he employs) might dismay an editor who has anguished over the wording of a definition and worried that a definition that has been elaborated to justify a grouping of citations somehow fails to capture the breadth and nuances of meanings placed in a single semantic bucket. Consider, for instance, definition 1(a) for the noun shame in the MED:

This definition is phrased in an effort to encompass and account for the quotations presented beneath it as well as delineate it from other definitions in the entry, such as 2(a)

and further to distinguish it from 3(a)

as well as from other subsequent senses.

Aitken is hardly alone in championing the evidence in a historical dictionary. Kurath “was an arch-empiricist, skeptical about knowledge of semantics and editorial judgment” (Reference AdamsAdams 2009a, 337), and he devised MED’s spare entry format, modeled on that of bilingual dictionaries, to maximize the amount of evidence that can be presented on a page. Nonetheless, both Kuhn and Robert E. Lewis, who followed Kurath in overseeing the MED, found it necessary to supply more information about dialect and meaning than Kurath had originally thought wise (Reference AdamsAdams 2009a, 343–344). Emphasis on the importance of citations does not necessarily minimize the importance of skilled defining. In fact, Reference KuhnKuhn (1975, 13) goes out of his way to take issue with any diminishment of the importance of definitions in a historical dictionary. He conceives the editorial process, once citations have been sorted into tentative groupings, as having five branches: “the building of definitions, the ordering of definitions, the selection of quotations, the listing of forms, and the etymologizing.” For him “the building of definitions is the most difficult, probably the most important, and possibly the least satisfying of an editor’s tasks.” While definitions might be characterized as signposts for citation paragraphs, without clear definitions, Reference KuhnKuhn (1975, 21) insists, the reader is abandoned and must blaze a trail in a dark wood. He writes as if in direct rebuttal to Aitken:

Perhaps the difference really lies in how each editor envisions the audience of the work: as scholars who are adept readers of the language and quite familiar with its culture or as students aspiring to learn about those very things.

In a synchronic commercial dictionary, definitions make up the bulk of the text, and the reader seeks understanding primarily through them, so the reader must in even greater measure rely on the judgment of the definers. While sense distinctions tend to be more finely delineated, examples and quotations are limited to one or at most two per definition.

Different editors may write quite different definitions within the constraints of the dictionary’s style conventions, and editors certainly may construct varying schemes of definitional groupings based on the same raw data. As we have seen, the selection of citations in a historical dictionary is guided not strictly by semantic considerations but also by the need to show phrases, variant spellings, and inflected forms, as well as the word’s use in texts written in different dialects and in texts of different kinds (religious, medical, poetic, and so on) so the reader can get a sense of the word’s broader use and glimpse its connotations.

It’s important that the primary editors realize that their work is in preparation for a reviewer to review draft definitions and revise as appropriate. It is helpful if primary editors include brief notes explaining or justifying how they have crafted an entry (or even a particular definition) to spare the reviewer from trying to intuit what was intended. Primary editors should be encouraged to admit when they are baffled about how to treat a certain citational grouping, make their best effort, and then move on to something else, rather than sit and fret and waste time pulling an entry apart and reassembling it under a different scheme. I have encountered editors who are so in awe of the authority invested in “the dictionary” that they find themselves paralyzed when they discover that there are multiple ways of setting up an entry, and they can’t determine which is the correct and unchallengeable one, as if there really were one.

28.15 Reviewing Definitions

The reviewer then has to accept the definitions presented by the primary editor or revise the definitions, sometimes querying the editor regarding the phrasing. This is basic to all editing, I suppose. Even when guided by the most thoughtful training and the most detailed Style Manual, a definer can go wrong in many and sundry ways, and the reviewer must be on the alert to spot them. The reviewer must be cognizant of the basic principles of defining, such as ensuring that every word in a definition is in fact defined and that the definition is substitutable for the word in context, and that polysemous words are disambiguated.Footnote ⁸ Reference WoolfHenry Bosley Woolf (1973) surveyed typical pitfalls that can bedevil a definer’s work: incompleteness of information, imprecision in details, ambiguity caused by syntactic sloppiness, faulty parallelism, references and analogies outside the intended user’s experience, insensitivity, editorializing. To be successful at this game, the reviewer must don the guise of the intended user, a person uninitiated to the knowledge of specialized fields. The reviewer must make assumptions about the sophistication of the user (be they an educated adult or a child or a learner of the language) but must not assume too much. Further, the reviewer must be on guard that the wording of the definition does not imply something the primary editor did not intend, and so lead to misinterpretation, or that the definition might imply that clarifying explanation is forthcoming in a following sentence that is absent from the entry.

28.16 Verbal Examples and Quotations

In commercial dictionaries, definitions outside scientific or technical fields ought to be illustrated wherever possible by examples of usage or by quotations. This is especially important for polysemous entries, where simple examples can make the semantic distinctions vivid: The ice melted in the sun … Sugar melts in water … The crowd melted away after the rally … The blue melted into the green in the painting. Ideally, these examples are based on quotations, and here is where a corpus can be invaluable, as corpora enable the identification of patterns of usage and grammatical structures in a variety of contexts and can determine relative frequencies of usage. Corpus enthusiasts champion “authentic examples,” which reflect idiomaticity and typicality, and distrust examples invented by lexicographers out of their own sense of the language. I would guess that most lexicographers who have worked under intense time pressure would admit to inventing a bumper crop of examples in their careers. Often, for reasons of excessive length or peculiarities of context, examples must be edited to illustrate meaning succinctly. Corpora may ease the burden on the minds and imaginations of a lexicographer, who “becomes a validator, as opposed to an analyst” (Reference KosemKosem 2016, 92).

The historical lexicographer, of course, is limited to quotations to illustrate meaning and usage, and here some editing is often required as well to avoid overly lengthy quotations and to give the reader the essential information quickly. The task of selection is often determined by the dating of the texts. If there is only a single text attesting a word within a certain date range, the editor must include that quotation. Dan Michel of Northgate, the scribe (and author/translator) of the Ayenbite of Inwit, dates his own manuscript at 1340, and since there are relatively few manuscripts written between 1300 and 1350, his work is cited many times in the MED. The more common task is in selecting quotations from an abundance having the same or closely adjacent dates. Here, the editor must decide which contexts might be of value to the user, and for certain criteria can be elaborated, as when the word is defined by the writer or glossed in another language (Reference Kurath, Ogden, Palmer and McKelveyKurath 1954, 3). Other criteria for selecting a quotation might include that it shows use in a specific region (often reflected in the word’s spelling) or includes an unusual spelling or inflected form. Editors have some leeway in selecting contexts that seem especially revealing of the culture or are considered to be humorous or otherwise entertaining to a reader. Reviewers must assess all the quotations and determine if they actually do illustrate the definition in question or belong elsewhere in the entry.

For every quotation illustrating a definition in a historical dictionary, often dozens or hundreds are rejected as unnecessary, by having a date attested by another quotation. At the MED, the stack of rejected slips for many definitions was often thick. Yet every one of the quotations, whether kept or rejected, had to be considered and analyzed by primary editor and by reviewer, frequently by reading surrounding context in the manuscript or published text and sometimes by reading the Latin or French from which the English was translated. This means that writing definitions actually takes up less time than analyzing quotations, sometimes much less time. The published entry for the ME verb taken “to take,” for which I did the primary editing, has eighty-five main senses illustrated by 3,085 quotations, and for most of these quotations there were many rejects. A reviewer analyzes the rejects to see if some should be included or substituted for the ones the primary editor selected. Given such requirements, it’s hardly surprising that a historical dictionary can take many decades to complete.

A longstanding tradition in monolingual lexicography has held that where possible, quotations from well-respected writers should be used. For the MED, a quotation from Chaucer was always to be included if one or more of Chaucer’s works appeared among the citations. Today, honoring great writing in dictionaries is sometimes criticized as overly relying on the aesthetic judgment of the editors, and may even seem quaint, when simple authenticity would better serve. There is some merit to this criticism, if one assumes that the function of a dictionary is solely utilitarian, and that users should not be encouraged to read or browse. Online dictionary interfaces tend to foreground the list aspect of dictionaries and further atomize the vocabulary, making it harder to explore a series of entries, whereas in print books, readers perforce encounter at a single glance an entire page spread of entries, often with art or notes on usage, word history, or another aspect of language. The question is a significant one: Is the dictionary a resource through which a user might learn more about language than just a single meaning of a word via the quick lookup? Should users be encouraged to discover and become more sophisticated appreciators of the language, or should this tradition of dictionary use be forever discarded in the interest of immediacy?

For AHD’s flagship dictionary, we chose to highlight fine writing with quotations when we could. To be sure, this was not easily done in the days before we had access to an in-house corpus and to Google Books. Even with these tools, finding choice quotations by respected or well-known writers can be time-consuming. The quotation has to unambiguously illustrate the meaning in question and be both easily understandable and concise. Longer quotations can add excessive length to a book, foul the navigation of an entry, and tax a reader’s patience. But a brief quotation that hits the nail on the head can be far more compelling than yet another functional example, whether made up or adapted from a corpus. The reader is freed from being dependent on the judgment of the lexicographer and can see the word in real use that ideally affords pleasure as well as insight. The straggle entry shown in Figure 28.5 offers a case in point. It has quotations from Carson McCullers and George Orwell (taken from the AHD corpus, as these authors were published by HMH), plus more contemporary writers Jean M. Auel and Stephen King, identified via Google Books. In my view, these quotations help justify the various meanings, which could be collapsed or even ignored in a dictionary for a different audience.

28.17 Managing a Dictionary Staff

Dictionaries are really (or should be) ongoing research projects. As such, they need and benefit from having long-term trained and motivated employees. To be sure, dictionaries can engage freelance editors, but using freelancers assumes that an editor’s role is solely to use existing materials to define words and edit entries. I have found it invaluable to work as part of a team that is familiar with an existing edition’s lexicon, knows its strengths and weaknesses, and over time contributes notes for additions and improvements to a shared file. While there is no disputing the value of having access to a robust corpus, there is also enormous benefit in having a group of people committed to making the best dictionary possible. For commercial dictionaries, one way to do that is to ensure that staff members pay attention to what is happening in the language and record their observations as potential edits for the next edition.

To keep a stable dictionary staff requires that staff members be motivated. Dictionary work requires a considerable amount of education, the ability to concentrate for prolonged periods, and the confidence to make thoughtful editorial decisions with some swiftness. The work is often tedious, and people who thrive on social contact usually struggle on the job. It is easy for editors to feel bogged down or overwhelmed or exploited by a higher management that underpays them and is more interested in maximizing profit than in funding research into words or investing in their employees (see Convery, Chapter 27, this volume).

In my view, the best way to motivate editors is to give them reason to feel ownership in the projects they work on. For this to happen, they must feel that their opinions and creative energy are valued. They must feel that the books are their books too, that their decisions and actions shape the books, that they are indeed co-authors. At the same time, they must learn to develop editorial detachment so that, if an entry they have edited gets revised, they don’t take it personally.

Historical dictionaries have similar challenges regarding motivation, but since a project typically goes on for decades, and primary editorial work is restricted to a single project, the daily sameness can feel like a grind, with an editor sometimes working on the same word for many months. Needless to say, it can be difficult for a manager to sustain staff energy at a high level, even when people understand that they are creating a scholarly resource of immense value. Editors may hold part-time appointments (sometimes splitting appointments with a university academic department), further jeopardizing staff engagement, and as academia has a chronically depressed job market, staff members may find their long-term employment prospects precarious, if not bleak. Review of primary work can be delayed for many months (often owing to reviewers being outnumbered by junior staff members). This is another factor that can lead lower-level editors to feel disengaged from ownership of the project, and a sense of disengagement can be compounded by a review editor’s lack of experience in management. Reviewers of historical dictionaries are usually appointed on the basis of their scholarly work. Although editors may stand in awe of the scholarly accomplishments of their colleagues, many reviewers of historical dictionaries have never worked in an office or had staff reporting to them, and they may feel that getting management training at an advanced stage in their careers is a waste of time.

For their part, editors of commercial dictionaries must face the vagaries of the publishing industry, where rumors of layoffs, reorganizations, and the closing down of entire departments routinely skitter down the halls. It may well be, given the attrition of the commercial dictionary business in the US, that a permanent twilight has descended on monolingual lexicography, as much as it ought to be honored and preserved.Footnote ⁹

29.1 Introduction

A letter from Madeline Kripke’s remarkable archive of Merriam-Webster correspondence and documents, now at the Lilly Library at Indiana University, tells the story.

It’s Thursday, March 28, 1844, a drizzly morning in New York City as George Merriam, co-owner of G. & C. Merriam, sits in his hotel room and writes a letter to his brother Charles back in Springfield, Massachusetts. He tells him of his meeting with a man named Adams of the J. S. & C. Adams company based in Amherst, Massachusetts, which had acquired the unbound sheets of Noah Webster’s 1841 edition of An American Dictionary of the English Language after Webster’s death the year before. Adams is facing difficulties selling the dictionary and is ready to sell his rights. George tells Charles, “we may become interested.” But first he wants Charles to meet with Adams to “talk over the p’s and q’s” of the deal, and he proposes a plan.

Adams will soon head home to Amherst, and he will pass through Springfield on the way, so perhaps they can persuade Adams to stop in Springfield for a meeting. George knows that Adams wants to be home soon, and he fears it may be hard to get him to stop in Springfield. “Unless,” he says, “this long drizzling rain is the means of retarding him.” Amherst lies another twenty-five miles north of Springfield, and George figures that if the weather is bad, Adams may stop in Springfield rather than pressing on the rest of the way home. “He goes for comfort,” George says. If Adams does tarry in Springfield, the instructions to Charles are clear.

We don’t know whether the drizzle persisted or if the meeting took place, but we do know that a deal was struck soon after, allowing the Merriams to acquire the rights to Noah Webster’s dictionary with the clear idea of becoming its publisher “permanently.”

The moment is crucial. It marks the beginning of corporate dictionary publishing in the United States. Dictionaries had been reprinted in the US for years, and new dictionaries had been published on both sides of the Atlantic, but always with living authors who would do the writing, most of the promoting, and sometimes part of the financing. Now a publishing company would take over responsibility for the revision and updating of the dictionary, for its distribution and marketing, and for the required investment.

This was new. Although Samuel Johnson’s Dictionary went through many editions even after his lifetime, it never found a permanent publisher to keep it current and in print after his death (Reference LynchLynch 2016, 243). Joseph Worcester’s dictionary was clearly the rival of Webster’s, but other than some short-lived interest from Lippincott in the 1870s and 1880s (Reference MartinMartin 2019, 271), no publishing company wanted to become its permanent home after his death in 1865. It isn’t until the 1850s that another publishing company, Longman in the UK, committed to being the long-term publisher of another reference book, Roget’s Thesaurus. Nevertheless, this is the model that we are used to, and, in the case of dictionaries in the US, it begins on that drizzly morning in March 1844.

Publishers of dictionaries take on many responsibilities, including the direction of editorial policies and practices, but it is the noneditorial aspects of dictionary publishing that are explored in this chapter – the sales, marketing, and distribution of dictionaries and the decisions about how to make money from them. The chapter focuses on monolingual dictionaries for native speakers of English from American commercial publishers and has little to say about dictionaries from university presses or about children’s, bilingual, or learner’s dictionaries although these constitute a large part of the market. However, the business issues involved with all kinds of dictionary publishing are seen in sharpest relief in the adult monolingual category, and many of the strategies and activities seen in this category apply to other dictionary categories as well.

What follows is a highly selective history of dictionary publishing in the US, exploring just enough history to establish this premise: That while the activities associated with the sales, marketing, distribution, and financing of dictionaries have changed radically over the years, many of the challenges remain the same. Specifically, the chapter illustrates the persistence of five constants that have shaped dictionary publishing in the US since its inception:

1. 1. Dictionary publishing has always been expensive. The factors driving that expense change over time, but even in the digital era there are no signs that it is becoming any less expensive.

2. 2. Dictionary publishing has nearly always been highly competitive. There has seldom been a time in the history of American dictionary publishing when there haven’t been two or three or more dictionaries competing for the same user.

3. 3. Dictionary publishing in the US has always been aimed at the broad general public more than at the professional or scholarly market; hence, the dominant publishing strategy has been to lower prices in order to broaden distribution.

4. 4. Technology has always been a driver of publishing strategies. From the introduction of stereotyping printing plates in the nineteenth century to the full embrace of digital distribution in the late twentieth century, new technologies have played a significant role in driving down prices to broaden distribution.

5. 5. Aggressive sales and marketing have always been a part of dictionary publishing, but throughout the history of dictionary publishing in the US, the one marketing strategy that overshadows all others has been effective brand management.

29.2 A Word About Branding

There are countless elements that go into successful dictionary publishing – strong editorial content, an effective sales force, compelling packaging, attention-getting publicity, capital to invest – but the most important element in the US has always been branding. Most consumers are ill-equipped to decide which dictionary will meet their needs best. As Frederick C. Mish, editor-in-chief of the ninth, tenth, and eleventh editions of Merriam-Webster’s Collegiate Dictionary, has pointed out, even those of us who know a great deal about dictionaries “would be hard put to give a precise technical analysis of why one dictionary is better than another” (Reference CarterCarter 1987, 28). Yet consumers must decide, as there have always been competing dictionaries to choose amongst. Moreover, the choice is an important one, especially in the US, where the ability to understand and produce standard English is seen as a civic virtue, and dictionaries are viewed as an essential tool in creating that competency.

“A brand is a set of promises,” Scott Davis tells us in Brand Asset Management (Reference Davis2000, 3–7). “It implies trust, consistency, and a defined set of expectations.” This is crucial because while there are those in the publishing industry who tell us that dictionaries are a fungible commodity, many people care very much about whether their dictionary will live up to their expectations and deliver the reliable information and guidance they seek.

In light of the highly competitive nature of dictionary publishing, both on the bookstore shelf and on the Google results page, some of Davis’s other observations seem particularly relevant, among them these:

1. 1. A brand differentiates products and services that appear similar in features, attributes, and possibly even benefits.

2. 2. A brand is about confidence and security. Brands help consumers cut through the proliferation of choices available in every product and service category.

3. 3. Strong brands embody a clear, valued, and sustainable point of differentiation relative to the competition.

4. 4. Brands are shorthand that customers use to guide their all-important purchase decisions. And given a choice, they will stay loyal to a brand for a long time.

Because of the importance of branding to the sales and marketing of dictionaries, this chapter talks mostly about the experience of Merriam-Webster, the publisher that has made the most consistent and successful use of the branding since entering the market in 1844. Nevertheless, Merriam-Webster has never been alone in this strategy. Efforts have been made over the years to create brands around the names of Worcester, Funk & Wagnalls, American Heritage, Random House, Oxford, Encarta, and most recently Dictionary.com.

29.3 Setting the Stage: Dictionary Publishing Before 1847

We get an excellent view of the challenges faced by printers and publishers of this era from the “Annals of the Merriam Family,” a family history created by Homer Merriam, the youngest of the Merriam brothers, which includes an account of the printing and publishing business formed by his uncle Ebenezer Merriam in 1798 in West Brookfield, Massachusetts (Merriam, 21–40).

The biggest challenge was the distribution of books, most of which was done on the basis of a barter system, with copies of books being exchanged for goods or services or to acquire other books to sell in the publisher’s bookstore. To do this, small country printers took on printing jobs for larger publishers in major cities, with the country printers being paid with an allowance to print extra books beyond what was ordered and then either selling them in their own bookstores or using them as barter with other country stores or with their suppliers. Country printers could also take on printing jobs on their own and then sell those books in their stores or use them as barter (Reference LarkinLarkin 1986, 42–44).

This system often led to complicated multi-step exchanges for the original printer to end up with the goods, services, or cash needed to pay for the print run (Merriam, 71). Thus, it was important to print books that had wide general appeal so that they could be easily bartered – Bibles, hymnals, schoolbooks, almanacs, and dictionaries, for example (Reference LarkinLarkin 1986, 45). Ebenezer himself did a printing of William Perry’s Royal Standard English Dictionary in 1801 (Merriam, 25).

Another challenge was the hefty expense of printing books. The biggest drivers of out-of-pocket expense were the fonts and the cost of paper. A font that Ebenezer needed to print an edition of the New Testament in the early 1800s cost $2,667.67 (Merriam, 27), which is the equivalent of nearly $60,000 in 2020 US dollars. Paper was the other big element in the cost of the print run. Through the 1820s, paper purchases could usually be paid for through the barter system, but by the 1830s papermakers began insisting on cash, thus forcing publishers to find new ways to raise cash (Reference LarkinLarkin 1986, 65).

In this environment publishers often went deep into debt to print books, and if the project was delayed for any reason, creditors took legal or sometimes not-so-legal action to get paid back. Homer’s “Annals” include Ebenezer’s own chilling narrative of an angry torch-bearing mob descending on and vandalizing his print shop in the middle of the night as a way of extracting money owed to them (Merriam, 33–39).

It’s not surprising then that publishers were reluctant to take on publishing Noah Webster’s American Dictionary in 1828. The project encompassed 2,000 quarto pages and required the purchase of special fonts to print the etymologies, and the dictionary was certain to be a slow seller priced at $20 a copy. It’s not known how much it cost Webster’s publisher Samuel Converse to print the 2,500 copies. One estimate from a publisher who declined the project put the cost at about 400 British pounds (Reference MonaghanMonaghan 1983, 108) or well over $500,000 in 2020 US dollars. Converse probably did it for less than that, but the project did not turn out well for him, as he ended up losing money on it and went bankrupt in 1833 (Reference MartinMartin 2019, 107).

One way to make publishing dictionaries more financially sound has been to include a plan for publishing lower-priced abridgments. As Jack Lynch points out, unabridged editions may get the fame, but abridgments make the money (Reference LynchLynch 2016, 352–353). Converse secured the right to publish an abridgment in octavo format, but Webster refused to do the work himself. Webster’s son-in-law Chauncey Goodrich then became involved in the project, and he and Converse enlisted Joseph Worcester, who had just completed a revision of Johnson’s Dictionary and was at work on his own dictionary (Reference MartinMartin 2019, 92–93). The abridged edition was published in 1829, and with just 940 pages and priced at $6, it was an immediate success (Reference MartinMartin 2019, 102).

By the 1830s, the dictionary market in America was highly competitive. Johnson’s Dictionary had been imported into the country since 1818 (Reference GreenGreen 1996, 285), and by 1827 Worcester’s revised version was available. The best-selling dictionary was Worcester’s abridgment of Webster (Reference MartinMartin 2019, 144 and 268); however, the most important competitor to Webster was Worcester himself. His Comprehensive and Pronouncing Dictionary, published in 1830, was not a direct competitor to Webster’s dictionary, but his name was becoming a rival brand standing for a more conservative approach to dictionary-making and finding favor among those who preferred retaining British spellings and pronunciation. Out of that rivalry emerged the War of the Dictionaries, the first phase of which was fought in the 1830s.

Much has been written about the War of the Dictionaries, so it’s enough to say here this phase of the War involved Webster’s claim that Worcester had plagiarized 121 entries from the American Dictionary in creating the Comprehensive and Pronouncing Dictionary, a charge that Worcester refuted. The controversy mostly took the form of letters to the editor of the Worcester (Mass.) Palladium in 1834 and 1835 (Reference MartinMartin 2019, 130–139). If the only issue was 121 entries, the controversy seems overblown and scarcely lives up to Webster’s charge of “a gross plagiarism.” Perhaps Webster feared that Worcester’s dictionary would hurt sales of his own dictionary; Worcester’s dictionary was very different from Webster’s, however. It was a 343-page octavo measuring about 7½ inches tall. Webster’s had 2,000 pages in two quarto volumes and was nearly twelve inches tall. There should have been a place in the market for both.

The bigger issue was that for many years Webster had been promoting himself and his name as a brand, even if he might not have thought in those terms. His eighteen-month multi-state bookselling and lecture tour in 1785 and 1786 was designed to encourage sales of his Blue-Back Speller, but it also promoted his name as an authority on language (Reference OstlerOstler 2017). The Speller became an astounding success, probably selling more than twenty-four million copies by 1847 (Reference MonaghanMonaghan 1983, 219) and establishing Webster in the public eye as the preeminent expert in matters having to do with American English. It was that brand that needed protecting against Worcester. So as trifling as this plagiarism would have been, even if it had occurred, it gave Webster the opening to attack Worcester’s reputation as a way of protecting his brand. From today’s perspective it seems petty and cruel, but the episode represents the beginnings of brand management as a crucial ingredient in dictionary publishing.

Despite its high price, Reference WebsterWebster’s 1828 edition had sold out by 1836, and Webster began making plans to bring out a new edition. Recognizing that price was a problem, he published the new edition in 1841 as an octavo and reduced the price to $15. Still the new edition did not sell well, and many unbound sheets remained unsold when he died in 1843.

29.4 Enter the Merriams: 1844–1864

The Merriams’ acquisition of Webster’s dictionary in 1844 ushered in one of the most tumultuous and eventful two decades in the history of American dictionary publishing. The most noticeable feature was the resurgence of the War of the Dictionaries, and the most significant outcome was a series of competing dictionaries from Worcester and the Merriams that altered the shape of dictionary publishing in America. The 1847 edition of Webster’s dictionary established the model of the American unabridged dictionary as a one-volume quarto, priced under $200 in 2020 US dollars. It broadened the market for large full-featured dictionaries and made them commercially viable. Worcester’s two dictionaries, the Universal and Critical Dictionary (1846) and the Dictionary of the English Language (1860), pressured the Merriams to make needed revisions to Webster’s dictionary. The 1864 edition incorporated those revisions and solidified Webster’s as the preeminent dictionary brand in America.

None of this would be easy. The Merriams were acquiring the rights to a high-priced, slow-selling dictionary, now sixteen years since its original publication, written by a person whose name was well known both from the Speller and from the abridgment of his dictionary which continued to sell well (Reference Ellis and RettigEllis 1992, 288), but who was not universally admired, especially by those who preferred more conservative linguistic standards. The dictionary had problems, especially in matters of spelling, pronunciation, and etymology. To turn this dictionary into a success, the Merriams would have to do three things: (1) adapt to the new cash-based publishing model which was taking hold in the 1840s; (2) restore the value of the Webster name and brand, which had lost some luster over the years; and (3) undertake a major revision program correcting the many flaws in Webster’s dictionary.

The shift in the business model fundamentally changed the way goods and services were exchanged. Large publishers stopped having their books printed by small country printers and began printing on their own and expected to be paid in cash. Printers and publishers like Ebenezer Merriam who did not have access to books from large publishers to use as barter and lacked good ways to raise cash faced extinction (Reference LarkinLarkin 1986, 65). George and Charles fared better. Their business, founded in 1831, focused on publishing books for which there was proven demand, thus ensuring profitable print runs, and they were located in Springfield, Massachusetts, which was becoming a transportation hub, making it easier for them to distribute books outside their region (Reference LeavittLeavitt 1947, 43–45).

However, to make Webster’s dictionary a success, they had to lower its price, which at $15 was much too high. To get to the lower price, they reduced the number of pages and compressed the two volumes into one. They did this by dramatically changing the design of the pages. The 1828 and the 1841 editions featured large legible type, wide margins, and generous use of white space on the page. In the Merriam’s 1847 edition, the type is smaller, the margins narrower, and most of the white space is gone from the page. With these changes, the new edition was printed in a single quarto volume and sold for six dollars.

The other way they reduced cost was to stereotype the printing plates. The technology had been available since the early nineteenth century, but Webster had not used it (Reference BryanBryan 1887, 9). The process involved setting the type for the pages, making a mold of the printing form, and creating a metal printing plate by pouring hot metal into the mold. This freed up the individual pieces of type for other jobs and eliminated the need for re-typesetting when reprinting. This technology had been part of the Merriams’ business from the start. It was particularly useful for the books they were publishing (Bibles, law books, schoolbooks), which do not require much revision between printings and could be printed many times from the same plates.

The second piece of the strategy was to rely on the power of Webster’s name to maintain sales while the revisions were being made. In doing so, they would build not only on Webster’s success in establishing his name as the preeminent authority on American English but also on his idea that the new nation needed a single uniform language and that his dictionary was the sole arbiter for setting the standards of that language. Hence they did everything they could to promote an image of Webster consistent with that role. A glowing biography was added to front matter of their printings of anything dictionary-related and they included praise for Webster in their advertising, press notices, pamphlets, posters, and other promotional materials (Reference LeachLeach 1996, 29; Reference MartinMartin 2019, 192–194).

More darkly, they looked for opportunities to cast aspersions on Worcester’s name, which led to the resurgence of the War of the Dictionaries. The two frequently cited causes for the resurgence of the War – an unflattering reference to Webster in the preface to Reference Worcester and JosephWorcester’s 1846 dictionary and the appearance of a British edition of Worcester’s dictionary that used the name Webster on it – were probably not serious enough to set off twenty years of hostilities. The more important reason was that the Merriams knew that Worcester’s name and dictionaries were becoming significant rivals to Webster’s, and the Merriams were intent on there being no such rival to Webster’s dictionaries or his name as the foremost brand in American dictionary publishing (Reference MartinMartin 2019, 177–178).

The public side of all this is well documented. Less well known are the behind-the-scenes efforts to promote the sale of Webster’s dictionaries. These included gaining favor with politicians and government officials who could make large purchases, using the company’s clout in the market to pressure distributors to favor Webster over Worcester, and making use of skilled influence peddlers who were free to spend whatever cash was needed to secure sales of Webster’s dictionary (Reference Ellis and RettigEllis 1992, 290; Reference GreenGreen 1996, 331–332; Reference LeachLeach 1996, 41). Much of this may not have been out-of-bounds according to that era’s business practices, but it was upsetting at least to Charles. When word came in 1877 that Lippincott planned to reissue Worcester’s dictionary, thus perhaps reigniting the War, Charles retired rather than go through the ordeal again (Merriam “Annals,” 85 and 103).

The challenge for the Merriams became even more serious when Worcester’s A Dictionary of the English Language came out in 1860. It was widely praised, and at nearly 1,700 pages and more than 100,000 entries, it was bigger than any other dictionary in America. The Merriams were outselling Worcester, but strong sales of Worcester’s dictionaries were still a threat to their sales and their brand (Reference MartinMartin 2019, 268). The War was not going to end until the Merriams put out a better dictionary.

Hence the third piece of their strategy was to commit to a program of revisions to correct the many problems with Webster’s dictionary. Some improvements had been made in the 1847 edition, but many had to be deferred to get that edition out quickly. The plan was to follow up with a more thorough revision that would rid future editions of any remaining idiosyncrasies in matters of spelling and pronunciation, replace all the etymologies, and revise the chaotic structure of many of Webster’s entries by removing redundant definitions, establishing historical ordering, and gathering similar meanings of words together under one numbered sense.

Engaging in such a sweeping revision was bold. First, from the marketing standpoint, the Merriams had put great effort into promoting the Webster name, so it was going to be awkward to acknowledge that there were flaws in his work. Second, the scope of the revision was so vast that the company would have to employ a different production model from anything done before. The new Merriam-Webster dictionary would be the first English-language dictionary written by a large editorial team, with all the managerial challenges attending that approach.

Third, the project would require a substantial investment. We don’t know how much it cost to create the new dictionary, but the next edition, the 1890 Webster’s International Dictionary, cost more than $300,000 (Merriam, 112) or around $9 million in 2020 US dollars, and there is no reason to believe that the 1864 edition would have been much less expensive. Most of that investment was made during the Civil War, when there were serious questions about how the war and its aftermath would affect the economy and the dictionary market in particular.

In fact, the company’s finances were not in good shape by the end of the project. In his “Annals,” Homer tells us that when he joined the company in 1856 the business appeared “promising, safe, and sure to be profitable” but that by 1864 the value of the company had declined, and he considered leaving the business. He lists the factors that led to this decline: (1) buying the rights to the Royal Octavo abridgement and the University Edition (one of the other abridgments of Webster’s dictionary) from J. B. Lippincott & Co.; (2) the money spent to counteract the publication of Worcester’s Dictionary in1860; (3) the drop in sales of the unabridged dictionary brought about by the Civil War; and (4) the large investment required for the creation of the 1864 edition.

Homer says of the Royal Octavo abridgment and the University Edition that they paid “many thousand dollars more than they were worth” (Merriam “Annals,” 79), but here Homer is wrong. One of the smartest long-term investments made by the Merriams in the 1850s was acquiring the rights to all the abridgments of Webster’s dictionary, giving them a product line of ten different Webster’s dictionaries, with one for every grade level and price point. This would allow them to compete more effectively against Worcester in the school market and lay the basis for a much more robust publishing program.

The Merriams’ financial picture began to improve with the publication of the 1864 edition. It included 114,000 entries, more than any preceding dictionary, with new entries drawn from every aspect of American life. There were 3,000 illustrations and thousands of illustrative quotations, and all of Webster’s etymologies were replaced by new ones provided by Dr. Carl Augustus Friedrich Mahn, a respected German philologist.

Perhaps the most significant aspect of the 1864 dictionary was its turning away from Webster and becoming more like Worcester, even as marketing for the dictionary continued to rely on and promote Webster’s revered status. The new edition included features intended to appeal to those with a more conservative approach to language, including usage paragraphs at many entries, quoting from British commentators, and entries that favored British terms over their American equivalents. While the 1828 edition was notable for its inclusion of quotations from American writers, the preface to the 1864 edition made a point of mentioning quotations from British writers such as Scott, Coleridge, Lamb, Byron, DeQuincy, and Tennyson.

All of this posed a brand-management challenge. If Webster was such an authority, why is a thorough revision required? And if such a revision was required, what is its relation to Webster’s original work? The job of explaining this fell to the principal editor, Noah Porter. In his preface, he says that his rule in revising the dictionary was to make any change that he believed “would be desired by Dr. Webster himself, were he now living, and fully possessed of the principles which have been universally accepted by modern philologists and lexicographers.” In other words, Webster was not fully possessed of those principles, but this is the dictionary that he would have written if he had been (Reference Goodrich and PorterAmerican Dictionary 1864, vi).

Porter then describes the approach of the editors as they reworked the structure of Webster’s entries.

With this, Porter kept the reputation of Noah Webster intact. The one aspect of Webster’s work that had not come in for extensive criticism was the quality of the definitions, so Porter makes clear that those definitions are still part of the new edition. In this important regard, it was still Webster’s dictionary, so advertisements for the new dictionary could simultaneously celebrate Webster’s name while also saying that it was thoroughly revised and enlarged. Thus, the brand was preserved.

The new dictionary was published October 1, 1864 and met with near-universal praise. It solidified Webster as the preeminent dictionary brand for years to come; marked the end of the Worcester threat; and set a triumphal tone for Merriam’s marketing voice for the next several editions. “Get the Best,” “The National Standard,” and “Superior to All Others” were recurring elements in the advertising for the new edition. The caption on one ad from the period catches the sentiment: “An Old Sun Rising with New Splendor.”

29.5 Copyrights, Trademarks, and Book Titles: 1890–1913

29.6 The Birth of Paperback Dictionaries and Mass-Market Distribution: 1947

Little has been said over the years about the role of mass-market paperback dictionaries, and yet they constitute the most dramatic step in the strategy of lowering prices to increase sales, at least within the print era. Many very small abridgments had been published over the years, but they were mostly for students and seemed to be side shows next to the larger editions.

That changed with the so-called Paperback Revolution of the late 1930s and 1940s. Although some paperback books had been published earlier, the beginning of the modern era of mass-market paperback publishing in the US is usually considered to be Robert de Graff’s founding of Pocket Books and its publication in 1938 of a pocket-sized paperback edition of Pearl Buck’s The Good Earth for 25 cents (Reference DahlinDahlin 1997, 51). The following year he published a list of ten titles, including Agatha Christie’s The Murder of Roger Ackroyd, Emily Bronte’s Wuthering Heights, and Thornton Wilder’s The Bridge of San Luis Rey, each at 25 cents a copy. Success was immediate. Pocket Books sold 100,000 copies in the first month and ten million copies in the first year (Reference EnnisEnnis 1981, B19; Reference ShafferShaffer 2014).

The key to offering these lower prices was not only the smaller size, glued binding, and less-expensive groundwood paper, but also much-larger print runs, and key to those larger print runs was much wider distribution. When Pocket began, there were only 1,500 bookstores in America, mostly in big cities, so relying on bookstore sales would not provide the sales volume needed (Reference DahlinDahlin 1997, 51). However, Pocket published its books as if they were magazines and expanded its distribution to newsstands, grocery stores, drug stores, train stations, and other outlets, adding up to approximately 70,000 outlets by the time it sold its 100 millionth copy in 1944 (Reference ShafferShaffer 2014). The success of Pocket Books and the US operations of Penguin did not go unnoticed, and they were soon joined in this business by Avon, Bantam Books, New American Library, and Fawcett Books (Reference ShafferShaffer 2014).

In 1947, Merriam-Webster became one of the first dictionary publishers to enter the paperback market (Pocket had been selling a dictionary titled The Pocket Dictionary since 1942), but the origins of Merriam’s involvement go back earlier, to the Armed Forces Editions program during World War II. Under this program a wide variety of books, including mysteries, short-story collections, and biographies, was published in paperback and given to members of the armed forces. According to one book historian cited on the Merriam-Webster website, it was the largest book giveaway program in the history of publishing. One of the most popular books in the series was Merriam’s Webster’s New Handy Dictionary, a paperback version of the smallest dictionary it published (Dictionaries at War).

Following the war, Merriam created The Merriam-Webster Pocket Dictionary based on the Armed Forces Edition and licensed it to Pocket Books, which published the new dictionary at 25 cents a copy with an initial print run of 500,000 copies. The announcement from Pocket Books to its sales force carried a sense of excitement, referring to the book as a “miracle” and as the “biggest title in 25¢ book history” (Reference HoweHowe 1947):

Once again, the enthusiasm was not misplaced. The new dictionary sold a million copies in its first year and another million in its second year and would be a perennial best seller in every year to come. It was followed quickly by The New American Webster Dictionary from New American Library and the Thorndike Barnhart Handy Pocket Dictionary from the paperback division of Doubleday, and in time by dictionaries coming from other publishers. Webster’s New World’s paperback edition would be published by Warner Books, the Random House paperback by Ballantine, and the American Heritage by Dell. Eventually, the prices of paperback books went up, and their page count and entry count expanded, but they have remained strong sellers, often the best-selling dictionary in a publisher’s line of products.

The success of the paperbacks is well deserved. In many ways, they deliver on the benefits promised for the Collegiate Dictionary: convenience, simplicity, ease of handling, and attractive pricing. The entry count is lower than in a college-level dictionary, and the definitions are reduced to their basics, but this is adequate to many situations. The editor of Merriam’s 1974 paperback edition recognized the value of such a book when in his preface he hoped that users would find in it, in the words of Christopher Marlowe, “infinite riches in a little room” (Merriam-Webster 1974).

29.7 Dictionaries after Webster’s Third: 1961–2000

The relative quiet in the early twentieth-century marketplace ended in 1961 with the publication of Webster’s Third New International Dictionary, Unabridged. So much has been written about the controversy that followed that little more needs to be said here. However, a few aspects of the controversy are worth pointing out.

First, the controversy was reminiscent of the War of the Dictionaries. Once again conservative scholarly and literary elites were set against a headstrong innovator, in this case Phillip B. Gove, the editor-in-chief of Webster’s Third. The conservative elite preferred that traditional notions of linguistic propriety be upheld; the innovator wanted to recognize and destigmatize the language he saw as representative of most Americans.

Second, the controversy resulted from a significant shift in brand positioning for Merriam. Part of the success of the 1864 revision came from its moving away from Webster’s text and toward an accommodation with the literary elites, a strategy that continued through Webster’s Second in 1934. Webster’s Third reversed that strategy, both in terms of the dictionary itself but also in its marketing, especially in highlighting usage from popular culture.

Perhaps most importantly, despite the strong sales of Webster’s Third and much good publicity about it (Reference MortonMorton 1994, 281; Reference SkinnerSkinner 2012, 298), the controversy created the perception that damage had been done to the Webster brand and emboldened competing dictionary publishers to seek greater market share (Reference LandauLandau 1989, 74; Reference MortonMorton 1994, 279), which led to one of the most competitive periods in the history of American dictionary publishing, from the late 1960s through the end of the century. However, the competition moved from the unabridged category to the college dictionary market.

29.8 The Age of Also and Beyond: Dictionaries in the Digital Era

Dictionaries are of this world and subject to its laws. “The dictionary,” writes Bo Reference SvensénSvensén (2009, 427), “is not only a cultural phenomenon […] but also, and to a very great extent, a commercial product subjected to market conditions.” Dictionary authors and publishers thus confront issues of copyright, both in the making of dictionaries and in owning them. “The survival instinct” drives lexicographers “to consult all dictionaries relevant to their own work in order to make sure that they have not overlooked anything important” (Reference SvensénSvensén 2009, 428), but ethics requires that consultation stop short of misappropriation of dictionary text, especially features unique to a dictionary’s style. As Sidney Reference LandauLandau (2001, 402) notes, “neither facts nor ideas can be copyrighted – only the particular form in which they are presented” (see also Reference SvensénSvensén 2009, 429), so lexicographers have a broad gray area in which to consult and borrow without breaking the law. Lexicographers also ward off plagiarism of their own work insofar as they can.

Lexicographers and dictionary publishers also wrestle with trademark holders who fear that dictionary treatments of their trademarks will interfere with business. Such claims should fall on deaf ears, in the interest of accurate reporting of language facts, but dictionary publishers are usually trademark holders themselves, so, as with copyright, they play both sides of the board. When should a dictionary include a term that originated as a trademark and with what acknowledgment or disclaimer? What responsibility do dictionaries have for the viability of trademarks, if any? Neither law nor practice nor theory has answered any of these questions definitively yet.

Beyond these two primary legal concerns of dictionaries, other ethical and cultural obligations vex the conscientious lexicographer or dictionary publisher. Lexicography is almost always collaborative work, so dictionary editors and publishers struggle with the ethics of giving credit where it’s due. We often identify dictionaries with chief editors’ names, yet most of the work has been accomplished by others under the chief editor’s indirect supervision – editors in charge of defining, etymology, pronunciation, and so on, often receive little more credit than those who do the keyboarding. In fact, even menial roles are essential to making a dictionary and should not be overlooked when credit is apportioned. One wonders whether dictionaries (especially digital ones) couldn’t roll credits like the movies, accounting even for the accountants, caterers, and personal assistants of the stars. Print dictionaries could not acknowledge their peripheral makers so effusively – although there are exceptions, as described further on in this chapter – given the obvious constraints.

People make dictionaries and people use them, so dictionaries reflect the cultural knowledge and attitudes of their makers and users, however much editors attempt a sort of scientific objectivity. Sometimes, those attitudes are contestable and even insidious. When dictionary entries present words of racial or ethnic significance, gender significance, disability significance, etc., they can do so from blind spots of routine and protocol, and the resulting definitions, labeling, and, in historical dictionaries, choice of quotations may prove insensitive or objectionable to some people. Increasingly, dictionary users hold dictionaries responsible for such lapses. (On attitudes generally, see Finegan, Chapter 19, this volume.)

30.1 Copyright

Early English dictionaries were schoolroom aids. The genre was so new and examples of it so few that one would think it emerged pristine, yet plagiarism sullied the dictionary tradition immediately: Robert Cawdrey’s A Table Alphabeticall (1604), arguably the first English dictionary, drew its word list from Edmund Coote’s The English Schoole Master (1596) and borrowed other material from Thomas Thomas’s Dictionarium Linguae Latinae et Anglicanae (c1588). Henry Cockeram borrowed some material from Cawdrey in The English Dictionarie (1623) but even more from John Bullokar’s An English Expositor (1616) (Reference Starnes, Witt and NoyesStarnes and Noyes 1946, 13–36). The history of lexicography is rife with such borrowing, though contemporary copyright protections have mitigated the worst of it. “Dictionaries have always copied from one another,” Sidney Reference LandauLandau (2001, 402) editor of many dictionaries, general and specialized, agrees, “but no reputable dictionary today would take over entire sections of another work and print them verbatim, a practice common in the seventeenth century.”

Concerns about plagiarism troubled the lexicographical community throughout the nineteenth century, however. James Murray, comparing parts of his New English Dictionary with some of William Dwight Whitney’s Century Dictionary worried that Whitney was borrowing material inappropriately and contemplated legal action; he was exasperated again when he saw proofs of the Stanford Dictionary of Anglicised Words and Phrases, which also bore signs, he thought, of illicit appropriation of his work (Reference GilliverGilliver 2016, 213–215; Reference Ogilvie and AdamsOgilvie 2010). Across the Atlantic, Noah Webster’s earlier claims that Joseph Worcester plagiarized Webster’s dictionaries ignited the so-called War of the Dictionaries (Reference MortonMorton 1994, 45–51).

The wordlist tradition of Western lexicography introduced plagiarism almost inevitably, because, as Reference SvensénBo Svensén (2009, 428) observes, “Naturally, it would be absurd to ‘reinvent the wheel’ at the beginning of every new dictionary project,” and the scope of vocabulary to be treated in such a project might best be determined after careful inspection of other dictionaries’ coverage. Dictionaries of similar scale likely have similar wordlists. Until the nineteenth century, when national copyright laws solidified and even international copyright was better settled, John Reference ConsidineConsidine (2017, 238) suggests, “It is possible to become distracted from the concept of ‘tradition’ by the concept of ‘plagiarism.’” Reference 721HamannByron Ellsworth Hamann (2015) has articulated just such a tradition for dictionaries affiliated with and developed from those by Antonio Nebrija (1441–1522), circulating the originals around the globe while adapting them to time and place. Sooner or later, however, the publishing and reputational stakes are high enough that plagiarism threatens corporate and personal investments.

Relying too much on someone else’s wordlist can interfere with a dictionary’s success, for as Richard A. Reference SpearsSpears (1987, 129) observed of argot dictionaries, because they often “just repeat older material, they distort the real state of argot by ignoring new material in favor of the easy pickings of the old.” The same principle applies to the scope of more general dictionaries and partly explains why words almost no one uses (concatenate, neonate, ululate) end up in all the dictionaries while some that many use (shipper, sim, sus) appear in none. Latinate vocabulary survives, desiccated, at the expense of words in the flush of (relative) youth. In the digital age, publishers face quick replication of their wordlists and guard against it, for instance, by inserting fake words in them, which would expose any thoughtless perpetrator. Thus, the New Oxford American Dictionary entered *esquivalience (see Reference AlfordAlford 2005), which the unlucky plagiarist would need to identify and remove as a fake before publishing the stolen wordlist.

Dictionary publishers attempt to protect their property, but editors face a practical copyright problem at the entry level. Even before Samuel Johnson’s A Dictionary of the English Language (Reference Johnson1755) and especially afterwards, lexicographers illustrated meaning and usage with quotations. Copyright was so insecure at the time that Johnson didn’t give it a second thought, but had he published his dictionary under today’s laws, he’d have thought twice about quoting Joseph Addison and Richard Steele extensively from the Tatler (founded in 1709) and the Spectator (founded in 1711), which would have been under their copyright in 1755. Still, Johnson would have had some room to maneuver, as short quotations in the service of lexicography could be considered “fair use,” especially since inclusion in the dictionary transformed the quoted material for the purposes of another discourse.

Historical dictionaries, which are built on quotations, include many and longer extracts from published works than do general-purpose dictionaries. Historical dictionaries have small audiences and are made without expectations of profit – they fall somewhat outside the category of “commercial product subjected to market conditions.” General-purpose dictionaries reach mass audiences, and their publishers expect profit from dictionary sales. Quoting other people’s work in pursuit of profit one doesn’t share with them might run afoul of copyright law, so, the experts suggest, quote as little as possible to the purpose: “In practice, no more than a sentence is required to render the sense” (Reference LandauLandau 2001, 405). On one hand, examples must represent current usage, so within copyright; on the other hand, short illustrative sentences in a dictionary constitutes fair use. Or, at least, we assume it does: “Because the use of a quotation in a dictionary is protected by fair use,” Reference LandauLandau (2001, 405) advises, “the dictionary editor should never request permission to use any quotation,” so as not to validate claims of copyright infringement.

30.2 Trademarks

Trademarks do not comprise a large proportion of vocabulary presented in dictionaries, but they pose a problem for lexicographers that gets to the heart of the lexicographical enterprise. Trademarks are names of brands of goods or services and, for the purposes of competition in a free market, they belong to the provider of goods and services on the terms of trademark law. Trademark holders have a lot at stake in preserving the viability of their marks, which distinguish their products among competitors in the market. Infringement of another firm’s trademark potentially confuses consumers, who purchase a thing or service on the reputation of its name. Infringement of the name interferes with the legitimate trademark holder’s success in the market. Thus, trademark holders vigilantly protect their marks, not only from infringement, but from so-called “genericide,” because once a term has become generic rather than the name of a specific brand, it cannot be maintained as a trademark. Cases in which a trademark becomes a generic term – famous examples include the verb xerox “make a photocopy” and Webster for “dictionary” test a dictionary’s sense of obligation (see Reference LandauLandau 2001, 407, for more examples and 410–412 for Webster especially; see Morse, Chapter 29, this volume; and see also Reference Butters, Gibbons and Teresa TurrellButters 2008a, 241n9).

Reference LandauLandau (2001, 407) proposes a dictionary’s social responsibility: “If the object of a dictionary is to represent the language so that people unfamiliar with the meanings of words can find those meanings, doesn’t omitting trademarks from the dictionary subordinate the interests of dictionary users to those of trademark owners?” The very purpose of the dictionary is at the center of the question: whom does it serve, how, and on what terms? If editors of the dictionary in question have collected enough accurate, relevant data, and if they construct facts from that data that assist public understanding of the meanings of words, including some trademarks, then they have done their job, and they “cannot allow any special-interest group to determine what goes in [the] dictionary or how it is represented” (Reference LandauLandau 2001, 407). Inevitably, trademarks will end up in large dictionaries, because some of them satisfy criteria for inclusion in them (see Reference ButtersButters 2007, 509–511). But lexicographers do not include them on a whim: people use the word band-aid, not only to refer to the brand of adhesive bandages trademarked as Band-aid, but to other brands of adhesive bandages, and also metaphorically to things and situations that need fixing. When usage so far exceeds use of the trademark, Landau proposes, dictionaries cannot deny the evidence to satisfy the trademark holder (Reference LandauLandau 2001, 407; but see Reference Butters, Westerhaus, Curzan and EmmonsButters and Westerhaus 2004, 115–119).

Nevertheless, many trademark holders resist both inclusion of “their” words and any treatment of them that might endanger their protected status (Reference SvensénSvensén 2009, 428). Sometimes, lexicographers have bent backwards in trademark holders’ directions, as when the Dictionary of American English modified its treatment of Crackerjack “popcorn and peanut confection” upon complaint from Crackerjack’s lawyers, tipping in newly revised and freshly printed pages to replace the offending ones in all the warehoused printed copies of Volume 1 (see Reference AdamsAdams 2005b). The remedy was extreme, but commercial dictionaries often include disclaimers in front matter to ward off legal objections, such as this one from the Random House Webster’s College Dictionary (Reference Nichols2000, ii):

Lexicographers and dictionary publishers simultaneously avoid lawsuits and indicate good faith treatment of the target vocabulary by conveying a word’s trademark status whenever it’s known, within the limits of responsible research. Indeed, trademark use should be recorded just as surely as registers like “slang” or frequency, as in “obsolete” or “poetic” – trademark status is an important language fact, and labeling makes the fact visible to dictionary users.

In fact, dictionaries can’t be sued for including trademarks or defining them as they think the evidence warrants. Trademarks prevent confusion among consumers about the source of a product, but if a dictionary defines a term once exclusively a trademark as a generic term, that’s not the dictionary’s fault, but the fault of any number of speakers who may speak as they like, without regard for a trademark holder’s fortunes. A generic meaning does not automatically or absolutely lead to consumer confusion; after all, we distinguish meanings of complex words from one another all the time. Human cognition can probably manage using the same word both for a specific brand and as a generic term without confusion, and American trademark law, at least, allows for the possibility (see Reference Butters, Gibbons and Teresa TurrellButters 2008b, 514; see also Reference Butters, Gibbons and Teresa TurrellButters 2008b, 516n31, quoting unpublished work on the matter by Roger W. Shuy).

Still, trademark holders must live with trademark law as it’s written, and it requires them to police use of their marks. Because dictionaries aren’t themselves competitive in commerce with the trademarks they may enter and define, lawyers have had to devise other ways to keep extended generic meanings of trademarks out of the lexicographical record. One approach claims that dictionaries cause or at least participate in the “dilution” of trademarks: when a dictionary includes any information different from what serves the interests of a trademark holder, it may participate in the blurring or tarnishing of distinctions important to consumer perceptions.

The dilution doctrine has never succeeded legally, but it has a fine pedigree, first proposed in Frank Schechter’s “The Rational Basis of Trademark Protection” (Reference Schechter1927), which was published in the Harvard Law Review (see Reference BoneBone 2007). Later, some legal scholars proposed that dictionaries could serve as agents of dilution and that the law should stop them (see, for instance, Reference RobbRobb 1981), and some forensic linguists proposed that dictionaries can and should be brought into dilution claims on behalf of plaintiffs (Reference Butters, Gibbons and Teresa TurrellButters 2008). A dictionary entry cannot make a term generic, but it might participate in diluting the strength of a mark by noting a related, generic meaning. Of course, it can do that only because speech has already diluted the mark, so dictionaries aren’t the critical agents of dilution – their culpability lies only in reporting the language facts, at which point trademark holders and dictionaries wrestle over competing interests, the private commercial value of certain words on the parts of the former versus public information about a language and the cultures and subcultures to which it belongs.

Dictionaries’ social commitment to represent trademarks fairly leads to some odd defining practices, however. The Random House Webster’s College Dictionary (Reference Nichols2000) labels Kleenex as Trademark and carefully defines it as “a brand of facial tissue.” The fifth edition of the American Heritage Dictionary of the English Language (Reference Pickett2011; AHD5) takes a notably different direction, no label and the brief “A trademark for a soft facial tissue,” in which “soft” may be editorial overreach. Who determines which are the soft and which the rough facial tissues? And is “soft” part of the word’s meaning or a brand claim? If the latter, the Random House definition is at once more accurate and more cautious. The third edition of the New Oxford American Dictionary (Reference Stevenson and Lindberg2010; NOAD3) ventures a bit further, but not unproblematically: “trademark an absorbent disposable paper tissue,” which points usefully to bona fide characteristics of the trademarked product (disposable, paper), though they aren’t distinctive to the brand. A dedicated user of facial tissues, I admit that absorbency is not a quality on which I depend.

All the dictionaries cited so far enter Kleenex with a capital letter to indicate its status as a proper noun or name, a name for a brand. (Trademark lawyers and judges prefer to interpret trademarks as adjectives, for legal reasons beyond the scope of this chapter, but lexicographers see trademarks as nouns – “I need a kleenex” – sometimes used attributively – “I prefer Kleenex brand facial tissues.”) The fifth edition of the Shorter Oxford English Dictionary (Reference Brown2002) reiterates all of the qualities of all the other entries: “(Proprietary name for) an absorbent disposable paper tissue, used esp. as a handkerchief.” But it goes further and balances its social and lexicographical responsibilities better. Kleenex is used by many speakers for any brand of facial tissue, so the parenthetical onset to the entry accurately recognizes the fact while also indicating trademark status. Then, because the Shorter Oxford is selectively a quotations dictionary, it provides two literary quotations under Kleenex, both of which also capitalize Kleenex, thus affirming knowledge of the word as proprietary while noting that not everyone uses it accordingly. Very likely, lawyers for the quoted books’ publishers insisted on capital K Kleenex, which means that the law intervened in the evidence on display in the dictionary – dictionaries almost always quote edited text.

Dictionaries naturally worry about their copyrights and defend them. They will also be subject to the protective maneuvering of trademark holders and their lawyers, too, and will grapple with the occasionally competitive responsibilities to report language use accurately, respect others’ intellectual property, and serve users’ needs and interests. In other words, as with any socially relevant type of text, dictionaries cannot rise above the social structures in which they are embedded and on which they trade.

30.3 Recognition

Some problems that beset dictionaries are ethical rather than legal. Many modern dictionaries, while they have authors in chief editors (as with James Murray in the case of the OED and Philip Gove in the case of Webster’s Third), depend on a large cast of lexicographers, experts, clerical staff, and others. While dictionary staff are as hierarchical as any other modern Western organization, as with the others, the “lowly” contribute essentially to the project and its outcome. Imagine a dictionary text that had never passed a proofreader’s scrutiny. Considering matters of type and scale, many dictionaries have one, two, or a few authors, easily acknowledged on a title page. The college and unabridged dictionaries, however, and the large-scale historical dictionaries involve so many people that giving credit where it’s due is difficult. Reference LandauLandau (2001, 412–420) suggested guidelines for acknowledging the work of lexicographers (not necessarily everybody else; see Reference LandauLandau 2001, 416) “[b]ecause the professional livelihoods of lexicographers depend on such recognition and because dictionary makers, publishers, and users are best served by fostering a skilled pool of lexicographers through recognizing their work” (Reference LandauLandau 2001, 419). The irony of the second justification punctuates several chapters of this handbook.

If we want big, general dictionaries, even if they live perfectly virtual lives, credit for those who make the product we use will prove a challenge:

Dictionaries have addressed the problem of credit for lexicographical work in various and illuminating ways, both over time and concurrently, which suggests the scope and persistence of the problem.

In the English tradition, dictionaries started out as an often-anonymous business. Robert Cawdrey, author of the first English dictionary, A Table Alphabeticall (1604), signs the opening epistle to readers but doesn’t appear on the title page. The first professional lexicographer may have been John Kersey the Younger, the J. K. of A New Dictionary of the English Language (Reference Kersey1702), but we cannot be absolutely sure: Reference Starnes, Witt and NoyesStarnes and Noyes (1946, 69–75) hedged their bets, while Allen Walker Reference ReadRead (2003, 222–223) insisted on the identification. It may have been a way of distancing himself from the lowness of his material, but the first serious slang lexicographer in English, author of A New English Dictionary of the Terms Ancient and Modern of the Canting Crew (1699) is known to us only as “B. E., Gent.” Perhaps we should not make too much of anonymity at that point in publishing history – authorship didn’t mean then what it means today, or what it would come to mean in lexicography after a famous author, Samuel Johnson, published his dictionary in 1755.

Johnson was the origin of the problem, not the solution Landau or anyone else has proposed. Complicated mastheads were unnecessary in single-authored works, that is, most of the dictionaries published before Johnson’s. And Johnson’s dictionary presented itself as the product of its author, as Johnson surely was. He did not, however, write the dictionary on his own; he was not, taking all the work of making a dictionary into account, its only lexicographer. We know that he had at least six assistants or amanuenses: Francis Stewart, Robert Shiels (who had already been denied credit for his Lives of the Poets [1753]), Alexander McBean (who had previous experience as a lexicographer), William McBean, V. J. Peyton, and William Maitland – a Mr. Stockton is also a possibility (Reference HitchingsHitchings 2005, 64–68; Reference ReddickReddick 1990, 96–120). We have all assumed that Johnson’s quippy definition of lexicographer as “harmless drudge” amounted to a rare instance of self-effacement, but perhaps he had some other, uncredited drudges in mind. The story of the lexicographer and his assistants was by no means a purely Anglophone phenomenon: all major dictionaries require work beyond the capacity of an auteur (see, for instance, Reference LittréLittré 1897, 20–22).

The rise of “scientific” lexicography in the nineteenth century led to an early form of crowdsourcing (Reference Ogilvie and SafranOgilvie 2020, 55–58), and the very large crowd of readers, assistants, sub-editors, and specialists responsible (see Reference GilliverGilliver 2016) deserved credit for bringing the material of the Oxford English Dictionary (OED) together and making preliminary sense of it deserved credit, which many but not all received in the extensive list of them at the beginning of the dictionary’s first volume (xxi–xxv; see Ogilvie 2023). Besides approaching the English language scientifically, the research product, the dictionary itself, recorded its co-authors in much the same way as reports of research by teams of scientists, from Principal Investigator to the dishwashers. Big dictionaries mean big staff, even if not all were as big as the OED’s. Even the Grimm brothers – dictionary auteurs, if there ever were any – and their lexicographical heirs required some assistance (Reference HarmHarm 2020, 87).

Lexicography also increasingly resembled journalism, from their shared Grub Street addresses of the late seventeenth and early eighteenth centuries to the fortnightlies and dailies of the nineteenth. Authors received credit in by-lines – in the literary journals, reviews were often still anonymous, right into the twentieth century – but editors found a place in the organ’s masthead. Signing dictionary entries is utterly impractical, so the sort of recognition Landau has in mind led major twentieth-century commercial and large-scale academic dictionaries to masthead acknowledgment of their staff, from editors-in-chief to typesetters and proofreaders. Proofreading is so fundamental to modern lexicography that it seems rather mean not to consider those doing it lexicographers in some elastic sense. And while it hardly ensures that women engaged in lexicography receive fair public recognition or move up the ladder into the positions of most authority within a dictionary organization, it certainly improves on earlier, blatantly patriarchal tendencies to marginalize women’s lexicographical work (see Reference RussellRussell 2018, chapter 4, for a thorough examination of the problem).

Thus, when one opens AHD5, one first encounters the “Editorial and Production Staff” masthead (Reference Pickett2011, vii), from “Senior Vice-President, Publisher, Adult Trade and Reference” to “Editorial and Production Assistance,” with “Proofreaders” at the bottom of the first column, just after the “Senior Lexicographer,” “Senior Editors,” “Editor,” “Associate Editors,” and “Consulting Editors.” The “Vice-President, Executive Editor,” “Supervising Editor,” and “Vice President, Managing Editor” occupy their own line below the Senior Vice President and above the columns into which the staff they supervised were sorted. Turn the page (2011, viii–ix) and one finds “Special Contributors,” those who wrote notes of various kinds – for example, usage, language variation and change – “Special Consultants,” and “Etymology Consultants.” Other major dictionaries list many more in the last two categories than AHD5’s twenty-eight subject consultants. For instance, Webster’s Third (Reference Gove1961, 8a–12a) lists an astonishing 199 “Outside Consultants,” but it also lists many more “Associate Editors” (13), “Assistant Editors” (58), “Editorial Assistants” (66), and “Secretarial and Clerical Assistants” (31). It was a bigger dictionary with a more than correspondingly bigger staff.

As Landau notes, dictionaries of substance take some time to make, and large staff experience turnover, especially in the junior ranks – not every assistant editor or editorial assistant expects to lead a life in lexicography. Webster’s Third indicates dates of service for those who didn’t see the project through from start to finish. Among women on the Webster’s Third clerical staff, one sees time run in the opposite direction, by parenthetically indicating maiden names – some who checked the masthead would know them by one name, others by another. AHD5 (Reference Pickett2011, ix, and previous editions) lists contributors to the previous edition, acknowledging that much material in the latest edition entered with its predecessor, but as Landau points out, dictionaries cannot acknowledge contributors in perpetuity. One previous edition is a reasonable compromise and ensures that AHD mastheads won’t trespass on pages better given to entries.

Large academic dictionary projects face similar problems of giving credit, but with some differences. Historical dictionaries, for instance, take a long time to complete, in the cases of the OED, the Middle English Dictionary (1952–2001; MED), and the like, some seventy years or so. The MED was published only in fascicles, and editors were featured on the cover of each. Letter by letter, however, staff were acknowledged at the outset, with additions to the staff during editing of fascicles in that letter listed at the end. The last editor-in-chief, Robert E. Lewis, neglected to compile those headnotes and endnotes to letters in the revised Plan and Bibliography (2007), but David Reference JostJost (2020), once an MED editor, reworked the lists, with additional research, and indicated the letters for which the staff were acknowledged. That’s a bit misleading, because staff acknowledged under S were also working on T. Misalignment of this kind suggests the intractability of the credit problem for dictionaries that take decades to write and publish, some of which will only be resolved by archival research well after the fact.

Acknowledgment gives credit where it’s due – it’s fair. But it also responds to the human need for acknowledgment, and morale suffers when accomplishment is apparently overlooked or dismissed. Some famous (or infamous) cases (see Reference AdamsAdams 1998 for an example) caused low morale and came with consequences for the projects and lexicographers alike. Academic lexicographers need jobs once the dictionary is done or they’re done with the dictionary, whichever comes first, but the academy approaches credentials differently from the commercial publishing world. How can one point to significant lexicographical achievement in an anonymizing work like the MED? Joseph P. Pickett, who wrote the book-length entry for the verb taken ‘take’ didn’t receive credit for a book. With no book and limited teaching experience, he could not easily translate his lexicographical career into professorship. He and others, including Jost before him, migrated into commercial lexicography. How could the MED better credential its junior editors? A 1977 review of the project proposed:

The dictionary never adopted the practice, largely because it was cumbersome and because it would not have persuaded English departments that lexicographical research was like literary research. The approach would never work in commercial dictionaries, which have much shorter entries and too many layers of editorial collaboration to give credit in this way.

Is it possible to give some credit to everyone involved in a big dictionary project over decades, in all capacities? In fact, it is, as the Dictionary of American Regional English (1985–2013; DARE) has proved. In acknowledging contributions, it exceeds every other dictionary in the English tradition and, to my knowledge, all the dictionaries in every tradition. Besides the expected masthead (2012, v) and two pages of immediate “Acknowledgments” (2012, vi–vii), the fifth volume of DARE includes twenty-seven four-column pages crediting the contributions of “DARE Staff, Students, and Volunteers, 1965–2011” (2012, xii–xiii), “Contributors of Words and Wisdom, 1948–2011” (2012, xiv–xxx), and “Financial Contributions to DARE, 1965–2011” (2012, xxxi–xxxix). Commercial dictionaries don’t require financial contributions from granting agencies or individual donors, so Landau had not considered such a category of credit, nor do most dictionaries record and then publish the names of all those they consult or who, motivated by interest in the project, send in words and quotations for inclusion. Dictionaries of the kind, however, depend on those contributions – no dictionary without donations, so while donors aren’t lexicographers and are thus outside of Landau’s proposals, some dictionaries really should acknowledge the necessary assistance.

Masthead acknowledgment gives lexicographers and staff and other contributors their due, but mastheads also convey different lexicographical orientations, differences in the conception, editing, and production of major dictionaries. Webster’s Third invested in outside subject experts, not only because the vocabulary it treated could benefit from their expertise, but because the expertise had marketing value. American Heritage and Houghton Mifflin (and then Houghton Mifflin Harcourt, and then Houghton Mifflin again) distinguished their dictionaries, in contrast with Merriam-Webster’s, by focusing more on usage advice, so its usage panel receives more masthead attention (2011, x–xii, double column) than the subject and etymological experts. Again, drawing on the usage panel affects the dictionary’s contents, but it provides a marketing hook, as well. DARE, different from commercial dictionaries in nearly every respect, captured American regional English of, by, and for the people, and its approach to acknowledgments was similarly democratic (Reference AdamsAdams 2013, 178–182), the dictionary crowdsourced beyond even the OED’s breathtaking standard.

Mastheads are important historical documents, then, partly because they embody their dictionaries’ motivations but also because they allow us to trace the people who contributed to them. For historians of American lexicography, a quick glance at the Webster’s Third masthead explains a lot. Short-term assistant editors included Robert B. Costello (1960–1962), Audrey R. Duckert (1953–1956), J. Edward Gates (1956–1962), and Sol Steinmetz (1958–1961). Since Webster’s Third saw print in 1961, we can infer that Costello, Gates, and Steinmetz worked, not only on that tome but also the collegiate dictionary that followed in its wake. Duckert collaborated with F. G. Cassidy on plans for DARE and continued as a very active consulting editor on that project until her death in 2007. Gates went on to organize and lead the Dictionary Society of North America, in one capacity or another, for several decades. Costello and Steinmetz moved to Random House, where they had stellar careers, taking what they’d learned on Merriam-Webster dictionaries far from Springfield, Massachusetts. These facts are important because dictionaries don’t make themselves – people make them. Knowing who the people were and, in broad strokes, what they did, is essential to the history of lexicography, and the mastheads, or any form of acknowledgment, are crucial evidence in that history.

30.4 Ethics and Entries

So much for legal and ethical issues that affect the internal workings of dictionary business. But what of the ethical dimensions of lexicography for dictionary users and readers? Dictionaries represent facts about language, but the very phrase is a paradox: facts may be unimpeachable, but their representations are not, and the human work of writing entries entails ethical decisions about what to include in a dictionary and how to characterize the meanings of words. Dictionary authority has long rested on the notion that the work and its results are “scientific,” that is, systematic and objective. Choosing the best words to convey meaning, however, is less a matter of science than judgment. There’s nothing wrong with that, but dictionary audiences will reasonably question lexicographers’ judgments in some cases, especially when the judgments convey harmful social assumptions.

Although the effect of including, defining, assessing usage, and quoting in support of an entry’s argument for meaning can be malign, lexicographers’ intentions rarely are. Writing entries without cultural assumptions is nearly impossible when the words entered, defined, and described are those for people or their qualities. Questionable attitudes appear in English lexicography from nearly the earliest date, when Randle Cotgrave, in A Dictionary of the French and English Tongues (1611) entered and defined Brunette as “brownish, somewhat browne […] a nut-browne girle.” Nowadays, we hesitate to distinguish people by hair color, eschewing cultural memes like “Blondes have more fun” and wall-flower brunettes. We are sensitive enough now to identify brunets/brunettes as brown-haired “people” or, following the feminine suffix, as “women” rather than “girls.” The terms on which we define and illustrate meanings matter to dictionary readers.

Later dictionaries assert their prejudices more confidently and in all awareness. The most famous instance is probably Samuel Reference JohnsonJohnson’s (1755) definition of Oats: “A grain, which in England is generally given to horses, but in Scotland supports the people.” One finds it difficult to swallow this without a grain of salt – it’s no mere clinical definition, and no wonder that Scots have found it offensive. Allen Walker Reference ReadRead (1934b) noted that “we feed our horses what they eat for dinner” is a trope found continuously from Classical literature to that of Johnson’s day. Johnson simply had some fun with the trope. Despite Johnson’s well-known opinions about Scotland and its people, Reference ReadRead (1934b, 89) observes that the Oats definition is the only one for a word with Scottish associations that one might consider a slur. Perhaps Johnson’s Scottish amanuenses protected him from himself. Although he did write other definitions, contrary to his stated intention in Plan of a Dictionary (1747; quoted in Reference ReadRead 1934b, 88), that would “disturb the quiet of others,” Reference ReadRead (1934b, 88) finds it “strange that Johnson allowed any conscious bias to enter his Dictionary.” Well, unconscious bias flows through many a careful dictionary.

How does one account for and correct bias in a dictionary? As lexicography became more scientific, for example in the guise of the OED, crowdsourcing helped. The OED’s original editors and Oxford University Press were sensitive to public senses of propriety, so some profanities and obscenities – as understood then – were omitted (Reference BurchfieldBurchfield 1992, 83). James Murray, the OED’s first and principal editor, consulted experts so often that the Post Office placed a pillar box at his front door (Reference GilliverGilliver 2016, 268). Had common sense not resisted a Johnsonian definition of oats, advice from a specialist in field crops might have done so, but no major American commercial dictionary, at least, lists an agronomist among its consultants. Bad oats definitions might well have endured and proliferated, under Johnson’s outsized influence, but AHD5 reassures us that careful defining is sufficient to solve such problems: “1. […] a. Any of various grasses of the genus Avena, especially A. sativa, widely cultivated for their edible grains. b. The grain of any of these plants, used as food and fodder.”

Careful definition alone, however, may not sufficiently signal or explain socially problematic words, of which oats isn’t really one. Lexicographers may distinguish between a word’s lexical semantics and socially significant metadata (to borrow an idea from Reference Fogal, Harris and MossGeoffrey Nunberg 2018). One might make the latter legible to dictionary users in restrictive labeling, for instance. An objectionable word – four-lettered or otherwise – can be marked as such: in NOAD3, dyke ‘lesbian’ is labeled “offensive” and fuck “vulgar slang,” but the latter requires more explanation, the editors thought, so they included a brief usage note at the end of the entry:

If a dictionary includes words like these, it must position them culturally, as well as define them, so that public use of the words is well-informed and offensive use minimized.

Explicit metadata serve various purposes but also introduce further problems into the lexicographical record. For instance, dyke is “offensive,” but to whom? One assumes to lesbians who do not identify with the term, though some do. In fact, such labels complicate a dictionary’s “scientific” status, its commitment to describe lexical facts rather than prescribe usage. The label may make too broad a claim about the term’s offensiveness to lesbians, but it makes an even broader claim implicitly, that everyone should find the word offensive and so not use it – unless, of course, one intends to use it offensively. Is fuck the most taboo English word and to be avoided, or is it valuable because it’s an effective expletive, perhaps the most effective expletive in the language? What does “vulgar” mean, and how does it characterize those who use fuck? Are they vulgar? Is that characterization more or less offensive than use of fuck, and who makes that judgment? When a word is problematic, can all dictionary users agree on the facts or the usage advice? Here we may encounter a limit to lexicography, to its success, anyway.

Even if we exclude problematic words from dictionaries, they won’t disappear. And even if we cannot satisfy all dictionary users with labels and explanations, we have to do something to acknowledge the problems words cause culturally and even personally, the pain a slur inflicts on the slurred, especially when the definition or some other element of an entry represents bias of a majority, privileged community, the community (in America, at least) from which by far most lexicographers are drawn, thus the possibility of unconscious bias in treatment of culturally problematic words is significant. Uniform treatments of problematic words – profanity defined and labeled this way, slurs that way – by someone especially thoughtful about those classes of words may mitigate the problems.

NOAD3 (and the earlier editions) had just such an arrangement with Ronald Butters, who was a member of its advisory board and consulted on vulgar slang and homosexual terms (2010, viii–ix; see J. Reference AdamsAdams 2021, 202–203). One sees the consistency in his approach across entries: fag “male homosexual” labeled “informal, derogatory,” while faggot “male homosexual” is labeled “informal, chiefly offensive,” with a cross reference to a usage note at queer. I’d have thought cunt, rather than fuck, might have been the most taboo word in American English, but Butters did not write a usage note for the former parallel to that he wrote for the latter. Such approaches are typical of modern lexicography. On the MED, the same editor drafted entries for all the prepositions, so that the prepositional field was covered without gaps or contradictions while capturing the contrasts among their meanings. General-purpose dictionaries usually have one person define all the color terms for the same reason (see Reference WilliamsWilliams 2014).

But even this method cannot ensure sound lexicography. The Encarta World English Dictionary (1999) defines and labels fuck as follows: “(taboo offensive) 1. […] OFFENSIVE TERM an offensive term meaning to have sexual intercourse,” which crosses the line between definition and Nunberg’s metadata and, as each sense of the word suffers the same treatment, becomes risible in its insistent consistency. Butters’ treatment of dyke would be more credible had he not argued, as a forensic linguist, that Dykes on Bikes, for a lesbian motorcycle gang, should be eligible for registration as a trademark (J. Reference AdamsAdams 2021, 206). How could he be so inconsistent? The answer is simple: such judgments depend on context, the use of words in discourse; general-purpose dictionaries, which abstract both meaning and metadata for commercial reasons, cannot capture context in its complexity, so cannot easily or successfully account for problematic words. Readers are bound to find treatment of one or another word unsatisfactory.

Reference PtaszynskiMarcin Overgaard Pstazinsky (2010) suggested that problems with the efficacy of labeling could be solved by a hybrid formal/functional approach that answered users’ communicative needs. This assumes a limited value to dictionaries, that people use them to know what they should say or how they can say it best, or to evaluate others’ speech and writing according to norms reflected in entry-level treatment (although not always within an entry’s structure, as in a usage note). The assumption is that people use dictionaries for practical purposes rather than for political or cultural knowledge. Perhaps dictionary “users” do, but dictionary “readers” have other motives in mind. Original subscribers to the OED received fascicles tied in brown paper just like other periodicals – they read the dictionary for historical and cultural knowledge, as do most who consult historical dictionaries. Readers of general-purpose dictionaries aren’t always looking for advice about usage but rather determining whether one or another dictionary represents critical cultural facts adequately or whether they need to be challenged so that users don’t inadvertently acquire the wrong ideas along with the wrong usage advice.

All of this is just to say that the best-intentioned lexicographers cannot patch up lexical meaning and the cultural problems people speak into their words. A wide range of issues have invited criticism of dictionary entries over the last few years. Lindsay Rose Reference RussellRussell (2018, 83–87) surveys sexism in Anglophone entries, from decisions to include certain items to bias in definitions, and she refers to so many confirming studies that one would think the question closed and a new direction charted, though there have been dissonant moments in that otherwise full-throated chorus (see Reference Whitcut and HartmannWhitcut 1984). The sexism is often caught up in other cultural conflict, for instance, over the meanings of sexual intercourse (Reference LandauLandau 1974), abortion (Landau 1988), and pregnant (Reference Mackenzie and Mel’čukMackenzie and Mel’čuk 1988), terms not necessarily problematic in all cultures, but certainly so in late-twentieth and early-twenty-first century America. Women are badly treated, too, in treatments of the lexis of mental, as well as bodily health (Reference Adamska-SałaciakAdamska-Sałaciak 2021; Reference Pierce and AdamsPierce 2010), though issues of disability and illness broadly have figured in the discussion of dictionary deficiencies throughout the last half-century or so (see, e.g., Reference Jost and CrockerJost and Crocker 1987, on down syndrome; and Reference NorriNorri 2020, for an especially illuminating survey across terms and dictionaries). Racism and ethnocentric insensitivity have received considerable attention, too, since the development of cultural criticism of dictionaries (see Reference MurphyMurphy 1991 and Reference Murphy1998, as early examples).

Currently, dictionary readers and lexicographers are focusing on words for historically excluded and currently underrepresented groups, their members, and characteristics of the groups or their members (see, for instance, Reference 751StoneStone 2021). The rise of anti-racism in the wake of recent police killings of Black people in the United States certainly contributes to the attention paid, but the concern that people outside of those groups define relevant terms without regard for the views of people inside them reaches beyond race and ethnicity to issues of gender and sexual identity, as well. With these cultural commitments in hand, people turn to dictionaries as authorities on meaning and usage, sometimes to discover that they know better, or that the terms on which words are defined and otherwise described are hurtful or harmful. Kory Reference StamperStamper (2021, 188) notes that “the internet demands that lexicography change,” and indeed it does, because reader dissatisfaction registers quickly and lexicographers cannot ignore it – they can no longer reply patiently to supposedly misguided letters from occasional critics.

Criticism of dictionaries on cultural grounds, though newly motivated by twenty-first century issues, extends back into the twentieth century, when all kinds of people, readers of dictionaries and lexicographers alike, began to resist the routines of what Stamper calls “Govian lexicography,” the style – which was part attitude – initiated by the editor of Webster’s Third. So Reference StamperStamper (2021, 187–188) “would posit that lexicography is undergoing a moment of reckoning” about the nature of its authority, “much like it did when Webster’s Third was released. Both then and now, the way that dictionaries represent language was under scrutiny. Both then and now, what was in the dictionary was seen as a reflection not of how language changed, but of how culture changed as evidenced – or not – by language.” Some of the change isn’t change in social conditions but in people’s attitudes toward those conditions and their insistence that dictionaries not aid and abet preservation of those conditions by means of lexico-cultural misrepresentation.

I won’t venture to identify the onset of such criticism, but I think history pivoted on the point of Delphine Abraham’s 1997 public campaign against the Merriam-Webster’s Collegiate Dictionary definition of the n-word, amplified by Randall Reference KennedyKennedy’s (2002, 133–137) account of that troublesome word’s history (see also Reference RussellRussell 2021, 249). Russell gives examples of dictionary boycotts from around the world dating from the 1960s, but primarily from 1976 forward, that challenged entries for Jew, bed/bang, marriage, wog, wop, dago, Jap, Jerusalem, Palestinian, Arab, Bangkok, homosexual, Black, or cross-references to offensive synonyms, depending on the instance. At the beginning of the twenty-first century, we see a new kind of crowdsourcing, not in dictionary production but in critical dictionary reception. And criticism abounds beyond boycotts. Reference RussellRussell (2021, 238) suggests consulting Reference MartinMartin (2021) and Reference Ogilvie, Treharne and WalkerOgilvie (2017) “[f]or additional examples of public dictionary criticism.”

Some words in Russell’s boycotts perpetuate gender stereotypes, and the cross-references arguably reinforce those stereotypes, engaging users/readers in loops of denigration. Others are racial or ethnic slurs, for which the list of problems is much longer than the list of boycotts – at least, so far. “Normal” defining just doesn’t work for slurs because as Reference Fogal, Harris and MossNunberg (2018, 284) explains, “users of a slur don’t always see it as the expression of negative feelings, yet it can evoke an angry reaction from its targets even so,” and probably should, and probably should also lead to some sort of response, perhaps a boycott. “That’s why one can’t say,” in Nunberg’s view, “that the attitude conveyed by an utterance of a slur is prescribed by the linguistic convention governing its use. And it means that here, too, the impact of the utterance must be determined in part by what the speaker intends and in part by external considerations.” That certainly expects a lot of the relevant definitions, though one can see, in careful comparisons across editions of dictionaries, that lexicographers have attempted to satisfy those demands, at least some of the time (see Reference LandauLandau 1985).

Lexicographers have tried to balance the demands of semantic content and externalities, but because general-purpose dictionaries can spare little space for such balancing, they have only muddied the waters, from a semantic point of view, without providing sufficient explanation of the externalities, using coded techniques like labeling and cross-references. Geoffrey Reference Pullum and SosaPullum (2018, 184) has noticed and attacked these tendencies, preferring strictly lexical semantic definitions and believing that various hedges, such as “considered as,” “regarded as,” and “used as” are “intended to deprive us of certain key entailments,” such as “that a bitch (in the abusive sense […]) is a woman.” This approach, while it has semantics behind it, will not satisfy the boycotters, however, nor lots of other dictionary readers.

Pullum’s explanations of words rather than meanings is perfect, however: “One crucial point to grasp, I think, is that the many years of philosophers talking about the relations between words and things,” a distinction of which lexicographers are also fond,

General-purpose dictionaries cannot convey all this information in an entry; editorial policies determine what types of information lexicographers emphasize over others; the variety of dictionaries extends from those policies; people should refer to more than one dictionary of different kinds whenever they can, if they really want to know about a word. Dictionary entries are abstracts of meaning and metadata and when defining words as complex as slurs will not – cannot – tell the whole story such that a user with underinformed intentions or with a justified sense of injury will see how that fits into an encyclopedic account of a thing called a word.

Hidden within Pullum’s “history” are histories of almost all the other word properties he lists. Also, words don’t mean the same things to all people in all contexts. M. Lynne Reference Murphy, Livia and HallMurphy (1997, 35) noted this regarding “lavender lexical issues,” especially “the inclusiveness and exclusiveness of sexual minority labels,” which not all within those minorities use with the same denotative or connotative meanings, and which assist in boundary maintenance between the groups. Ordinary definitions cannot itemize such distinctions: one needs a Venn diagram of interlocking groups and meanings. Murphy also argues that group/meaning configurations change over time, partly because of user awareness of the words and their problems. Print dictionaries necessarily lagged behind social change. Now, notionally, we can revise online dictionaries at will, though they, too, operate with limited resources. Slurs and other problematic words challenge the dictionary’s capacity to keep up with the times.

30.5 Responsive Lexicography

But keep up they must, if for no other reason than that social attitudes toward dictionaries have changed significantly in the digital age (see Finegan, Chapter 19, this volume). Dictionary content is no longer self-sufficient. As Katherine Connor Reference MartinMartin (2021, 220) puts it, it has become “decontextualized”: “We can see two general tendencies as we move further away from the original print context here: in general, with each step away from print the number of users affected by the dictionary content tends to increase, and the original publisher’s control over it, tends to decrease.” Audience control rises as publishers’ control falls, so dictionaries must respond to users’ political sensibilities, if they are to assert authority over problematic vocabulary, the vocabulary most likely to attract sharply committed public scrutiny. Many argue that such is the dictionary’s responsibility, but even if one doesn’t accept it as an ethical imperative, it’s certainly a commercial one.

Publishers and editors know how to deal with trademark and copyright issues. Both the impulse to give credit where it’s due and the techniques for giving it have improved during the twentieth century. Changes in the dictionary audience and its expectations of dictionaries, however, tied to digital modalities and overtly political, are relatively new. Reference MartinMartin (2021, 220) asserts that “In the early 2020s, the progressive decontextualization of dictionary content coincides with a period of hyperpolarized cultural and political discourse during which words and their meanings have become a battleground and dictionary publishers themselves have sometimes entered the fray.” One wonders whether dictionaries can protect facts about the words they enter from dissent and worries that soon factions in the culture wars will have “their” dictionaries, much as they have “their” news outlets and web spaces and facts: definitions must confirm their cultural and political perspectives. A happier future for lexicography would see dictionaries navigate the Scylla of audience predilections and the Charybdis of word-knowledge, achieved more easily in the digital age than print dictionaries allowed.