It is true that common indexing wisdom is to avoid repeating locators within an array. Repetition, with locators duplicated across subheadings, can bloat an index. An array may be difficult to read, or confusing to readers, if there are too many similar subheadings side-by-side. If there is a strict length limit, there may also not be enough space for duplication.

And yet, as with many conventions in indexing, “it depends.” I recently wrote an index in which I did consider repeating locators within an array. I think it is worth teasing out why this may sometimes be appropriate.

Framing Discussions

To start, it is important to recognize that discussions can be framed multiple ways. This affects how subheadings are phrased.

One option is to state the literal concept, object, organization, or person that is being discussed. For the book I recently indexed, which was about Anglican Church leaders and fertility-manipulation schemes during the interwar period, this could be something like,

birth control: Anglican leaders on

This tells the reader who is commenting (Anglican leaders) and what they are commenting on (birth control), but it does not tell us anything about the substance of the discussion.

Another approach is thematic, to focus on that substance. For example,

birth control: class and; empire and

These two approaches can also be combined, though such subheadings can become unwieldy and each element of the subheading may not appear on every page. For example,

birth control: and Anglican leaders’ concerns about class and empire 

My point here is not to advocate for one framing over another. I use different approaches, often within the same index. It depends on what I think it most relevant for the reader and the array. To cut down on length and to keep arrays succinct, it often helps to pick a single frame. As is oft said, the goal is to point readers to the information, not to rewrite the text.

Other times, however, it may make sense to include multiple frames—multiple subheadings—because the different frames carry equal weight and because the discussions—and locators—while overlapping, are also distinct. This can help tease out the different elements, especially when discussions are complicated and intertwined. So, something like,

birth control: Anglican leaders on; class and; empire and

Alternate Entry Points

An alternative to having parallel subheadings and duplicate locators within an array is to create entry points elsewhere. Perhaps there is not enough space in the birth control array to include subheadings for class and empire, but those can both be (and should be) main entries in their own right. When considering the different elements of a discussion, and the different ways a discussion can be framed, think about what other headings and subheadings can be created elsewhere. An index is the perfect place to deconstruct and scatter discussions.

Multiple Discussions Per Page

It may also happen that a single page contains multiple distinct discussions. This is especially true the more words there are per page, depending on how the book has been typeset. If this is the case, then index to the discussions. Don’t worry if locators are duplicated. 

The trick here is distinguishing what is truly a distinct discussion or fact, rather than a granular subset of the same discussion. Does the discussion still hold together when you telescope out?

Focus on the Story

Lastly, consider the story that the array needs to tell. What is it, about this subject, that is important for the reader to know? This is not about rewriting the book in the index, but rather about effectively pointing readers in the right direction. Having a focus is also relevant for space, both because space for the index may be limited and in the sense that a succinct array is easier to scan. 

This also comes back to framing. What is the most effective frame for readers to recognize what they are seeking? As mentioned, often a single frame is sufficient, focusing on whatever is most relevant. In other instances, it may be better to include multiple frames if they hold equal weight. Either way, test the subheading(s) against the larger story or point that the book is making about that topic. 

Repeating locators within an array, across two or more subheadings, is usually a question of how that information is being framed. It is often best to avoid duplication, to avoid similar, parallel subheadings or an array that is too long. In other instances, in which discussions are complicated with intertwined subjects and themes, parallel subheadings may be justified. It is a matter of seeing the whole story that needs to be conveyed, and then deciding the best way to point readers towards that story. As with so much when indexing, it is less a question of right or wrong, and more a matter of weighing the relative merits and applicability of different approaches.

It is a big book, and it is taking a while to get through. It is also not the only large volume I’ve indexed this year. I feel like I’ve done more than usual, including an Oxford handbook on Baptist studies, an edited collection on Black Canadian art and literature, and a biography of Canadian poet Margaret Avison. These are books that are the equivalent of two or three regular-length volumes. 

Large books can be challenging to index. I have a theory that the length of the index grows exponentially with the length of the book (while, maybe not quite exponentially, but the length does grow), because the longer the book, the more details there are to index, the more locators there are, and the more likely subheadings are needed. Subheadings take up more space and take more time to write and edit, and so the work goes on and on. Also, the larger the book, the more there is to keep in mind. I need to remind myself to reset my expectations for how long the index will take. 

I still want to complete the index efficiently and as quickly as possible, even if the overall number of days and weeks is longer. I’ve been reflecting on what it is I do that is different compared to indexing a shorter book. This is what I’ve gathered so far:

• I write more notes for myself, such as brief summaries for each chapter. Because there is so much more to remember, I don’t want to forget what the beginning of the book was about by the time I reach the end. These notes can also help me see the overarching storyline and themes running through the book.
• I slow down and consciously remind myself to connect back to the big picture. With so many pages to get through, the temptation is to rush and to focus on picking up the easy entries—the low-hanging fruit—of people, places, reactors, chemical elements, etc…Those details are important and need to be picked up, but why are they important? What is the larger discussion or story? I need to make sure that I am creating headings and subheadings at those broader levels too, which means slowing down and understanding what all this is actually about.
• I write subheadings at a broader level. I still pick up plenty of smaller details, but I think broader subheadings are both easier to understand and navigate, and can help keep the array at a manageable length. For main characters and topics, which are going to have a lot of subheadings anyway, indexing less granular may be the difference between forty subheadings versus sixty or eighty, for example. Forty subheadings is still a lot, while still being easier to scan. 
• As I mentioned, broader subheadings can also be easier to navigate. This is because they encompass more and clearly express the larger context. Say a person is involved in constructing the National Research Experimental (NRX) reactor, is quoted reflecting on the reactor, and is also present for the reactor’s meltdown. A more granular approach would be to have separate subheadings for each of those three elements, which are arguably distinct. But in a large array, those three subheadings may sort apart from each other and a reader may not see all of them. Three distinct subheadings also take up more space (see point above). Better, I think, for a single subheading at a broader level, “National Research Experimental (NRX) reactor and,” to capture all of those discussions. The details of construction and meltdown can still be delineated as subheadings under the reactor itself. 
• While indexing at a broader level, I still want main headings and subheadings to be clear and specific. Indexing at a less granular level is not synonymous with being vague. If I go back, say when editing the index, and I don’t know what the subheading is referring to, then I can’t expect the reader to know. The meaning needs to be clear. Which reactor? Which relationship? Which concept or discovery? I try to make my first pass as clear as possible, so I have less work on my second pass and so I don’t leave myself confused when I come back around again.
• Lastly, I want to acknowledge that I use subcontractors on this large projects. I usually ask subcontractors to pick up the names of people, and perhaps also other common elements, such as, for the Avison biography, titles of poems and other works. Having a subcontractor handle these aspects frees up my headspace to focus on the subject entries and the overall structure of the index. It also helps me to get through my own draft more quickly, because I have less to pick up. So many thanks to the indexers I work with. With so much material to get through, I find it can make a noticeable difference.

In one sense, indexing a large book is no different from indexing any other book. I still pick up the same types of entries and my process doesn’t change too much. On the other hand, everything that makes indexing a regular monograph difficult is magnified simply because there is more of it. There are more details, more entries, more subheadings, more cross-references. I find writing the index requires a bit more planning, more notes, and more patience. Because the big picture is so big, it can be difficult to focus on, and so I need to make an even more conscious effort to keep referring back to what the book is about.

The big picture—the story that is being told—is the underlying aboutness, which needs to be visible in the index. I don’t want to get lost while indexing, and I don’t want the reader to be lost either. I need to slow down, think through my decisions, make sure I am seeing as clearer as I can in that moment, so that the reader can hopefully see through me.

Today’s question is about granularity:

How should I judge what level of granular information the reader will be looking for? Too many subheadings with the same locator is a red flag. Is there any other good marker? How do I know if it is too granular, or not granular enough? 

That is an excellent question. Thank you for asking.

I suspect this is a common issue that many indexers face. Part of the challenge is that the answer is so often, “It depends!” It depends on the audience, on the subject matter, on the type of book, and how much space is reserved for the index. Because there are so many factors involved, I’m not sure if I can give a satisfactory answer, but I’m going to give it a try.

What is Granularity?

To begin with, what is granularity, anyway?

Granularity refers to the level of detail. In indexes, this is how detailed entries and arrays are. The level of granularity can vary, from large or wide-angle details to small, specific details. When discussing how granular an entry or index should be—whether it is too granular or not granular enough—this is what we are talking about, the level of granularity. 

As I discuss in my book, I like to conceptualize granularity in terms of the hierarchy of information. Most books contain multiple levels of information, which can be mirrored in the index. At the top—the broadest level of information—is the metatopic, which is what the book as a whole is about. Below that are the supermain discussions, which are the main arguments or points which compose the metatopic. Below those are the regular discussions, which continue to break down the main discussions. At the bottom are the small, specific details which further support and illustrate. Each level is a different level of granularity. 

My view is that all levels of the hierarchy of information should be in the index. A well-written index should be balanced. It is not a question of finding the single correct level of granularity, but of creating entry points for all levels. 

Indexing at multiple levels of granularity provides a number of benefits. It serves different readers, both those searching for specific details and those searching at a broader level. Multiple levels of granularity can also facilitate navigation through the index, whether through nesting more granular subheadings under broader main headings or cross-references from broader to specific arrays. Including both broad and specific headings and subheadings also helps readers—and indexers—avoid getting lost in the weeds. Where I’ve seen granularity go wrong is when the index is primarily a collection of specific details, such that it is difficult to see how the details relate to each other or why they are relevant. Small, specific details can definitely be relevant and important to index, but they also need context, and readers also need context. Indexing at all levels of granularity provides that. 

Telling Stories

I like to think about indexing as telling a story. Each array encapsulates what the book has to say about that particular subject. Especially if the array contains subheadings, the subheadings need to work together to present what the book has to say.

Granularity can factor in a couple of different ways.

One is the appropriate length of ranges. A common rule of thumb is to add subheadings if a discussion extends beyond five pages. This is generally good advice, as searching more than five pages can be a big ask for a reader. But what if the range is seven or eight pages? What if there are no clear points within that range to differentiate into meaningful subheadings? What is space is tight? What if one or more pages are mostly filled with figures? Instead of treating those five pages has a hard limit, I treat it as a guideline. My first goal is to see the whole discussion in context. Sometimes, indexing to the full discussion, even if that means a slightly longer range, is enough. A broader level of granularity points readers to all they need to know.

Another way to consider granularity and storytelling is to consider how broad or specific subheadings can be. If the array is for a supermain discussion, some of the subheadings should probably be for regular discussions. Subheadings can be used to nest levels of granularity. 

To be clear, I am not suggesting that all subheadings need to be of the same granularity or that all subheadings need to be one level below the main heading. Subheadings can serve different purposes, and different types of subheadings can coexist within the same array. It is okay to mix in a few specific details or relationships to other supermain discussions, for example. What I am saying is to pay attention to the story that the array is telling, which is most clearly told in terms of the main points made about the subject and in terms of the relationships between that subject and other subjects. These main points and relationships can often be mapped to levels within the hierarchy of information, which is to say, levels of granularity. 

Spreading Out Granularity

So an index can contain different levels of granularity, and levels of granularity can be used to tell stories and indicate relationships.

Entries and arrays for different levels of granularity can also be spread out throughout the index. It is not always necessary to fit all possible subheadings under a main heading.

Say the array is for a supermain discussion. The book  makes seven main points—regular discussions—about the supermain discussion. Each of those points can become a subheading, which combined provides a solid overview of the subject. Each of those seven points can be further broken down into specific people quoted and illustrations made, but do those people and illustrations need to be in this array? I would suggest not. Keeping this array at a broader level of granularity allows it to be clear and compact. Readers will still find those smaller details if they read through to those pages. Those details can also be standalone entries elsewhere in the index, still findable if readers are looking for them specifically.

I hesitate to be too prescriptive here because every book, index, and array is different. Sometimes arrays do function as a place to gather certain types of detail and it is important to get every mention. But my point remains: levels of detail can be spread out. All information should be findable in the index; it does not all need to be findable in the same place. Play around with how granularity is distributed throughout the index. What can be double-posted? What doesn’t need to be double-posted? How can cross-references be used? What is the function, or story, of each array? 

Audience and Type of Book

The level of granularity can also be affected by the audience and type of book.

I know the common advice is that trade books can be more lightly indexed than scholarly books. I think that is true-ish. My tendency is to write comprehensive, detailed indexes for both trade and scholarly books. Perhaps better to say that scholarly books tend to contain more detail, which lends itself to a more detailed index, while trade books can—but not always—be written in broader strokes. When indexing, match the level of granularity to the book.

How the book is written, and the subject matter, can also affect granularity. Some books are written as broad overviews, whether covering a wide time period, or multiple people, countries, or organizations. For these, keeping subheadings at a broader level can be helpful to prevent the reader—and the indexer—from getting lost in the details. Similarly, if an array has a lot of subheadings—say for a main character or key organization—keeping subheadings at a broader level can be a way to cut down on the number of subheadings needed. Subheadings should still be clear and specific—indexing at a broader level should not be mistaken for being vague. Rather, indexing at a broader level means pointing towards the larger point or context. Specific detail may be better suited for standalone entries. 

How books will be used can also impact granularity. Reference books, for which the index will be a primary entry point, should contain detailed indexes, with lots of subheadings and entry points. Scholarly books often fall into this category as well. Books that are written more as a light or casual read are probably fine with a less granular approach. 

Space Constraints

Space can be a real issue for some indexes. There are not enough lines for the index you want to write. This affects granularity. I think it is still important to try to include entries at all levels of granularity, but if space is a constraint, some entries will need to be cut and compromises made.

If space is tight, I tend to prioritize the big picture. I want the supermain and regular discussions to be clearly present. These are the main points that the book makes, and in my view, the most relevant for readers to find.

Indexing at a broader level can also help save space because the broader entry can incorporate or cover multiple smaller entries. Say there is a discussion about railways, which includes mentions of multiple railway companies. In an ideal index, both “railways” and the specific companies are picked up, but if space is a constraint, I choose “railways” and leave out the specific companies. Readers will still find those companies if they search under “railways,” whereas if I only include entries for the specific companies, readers will have a harder time seeing the whole picture. 

Another consideration is how close or far an entry is from the metatopic. The closer to the metatopic, the more willing I am to keep entries for specific details. The further apart, the more likely I am to cut and instead rely on entries at a broad level. 

When space is tight, I also tend to use fewer subheadings and I allow longer strings of locators. I still try to include subheadings if more than ten or twelve locators, as I don’t want undifferentiated strings to be too long, but subheadings take up space and I’d rather squeeze in more main entries. This is another way to be less granular, by removing the clarity that subheadings can provide. Not ideal, but it can be a worthwhile compromise.

Finding the right level of granularity can be tough. The first challenge is learning to see the different levels. Once you can identify them, it becomes easier to zoom in and out, and to decide how different levels should fit together and be distributed.

To close, I want to reiterate that there is no single, perfect level of granularity that an index should be at. Indexes should contain multiple levels of granularity, and the right mix of levels will depend on the audience, the subject matter, how the book is written, and how much space the index has to play with. But once you can see the different levels, and to see the levels as tools or as building blocks, granularity can be a valuable lens through which to see and construct an index.

Someone replied to that reflection and brought up the important and related issue of synonymous and similar terms. Thank you for your email! Synonyms seem like a natural follow-up, so I am going to write about that today. 

Synonyms vs. Similar Terms

When I think about synonyms, I mostly think about concepts and things. These are different from names. Name variants either belong to the same person or organization or they don’t. With concepts and things, there can be much more nuance.

I like the phrase “synonymous and similar terms” because it reflects that nuance. Some terms are interchangeable. Other terms are similar yet distinct. The context of the discussion can also be important; terms may be synonymous in one book and distinct in another. So when dealing with concepts and things, it is not simply a matter of identifying synonyms but also of determining how they are used.

Terms can vary for a number of reasons. Synonyms exist in the English language and an author may use more than one term in their book. Authors in an edited collection may also choose different terms.

Preferred terms can change as society becomes more attentive to the impact and nuance of what terms mean and how terms impact people. For example, the recent move from “slaves” to “enslaved people.” In Canada, in regards to Indigenous Peoples, terminology has shifted over the past several decades from “Indian,” to “Aboriginal,” to “Indigenous.” Indigenous Peoples are also differentiated, in Canada, between the First Nations, Inuit, and Métis. 

Terms can also vary across countries and cultures. In regards to Indigenous Peoples, the term Native Americans is not typically used in Canada, while common in the US.

Terms can also reflect important distinctions. In an edited collection I recently indexed, the terms and phrases “Two-Spirit,” “Indigiqueer,” and “Two-Spirit, trans, and queer Indigenous” where all used in reference to queer Indigenous people. Though the authors typically used these terms together to discuss queer Indigenous people more broadly, the authors also wanted to acknowledge that queer Indigenous people can identify in different ways. 

Three Strategies

There are three main strategies for handling synonymous and similar terms. These can be used alone or combined.

1. Choose a preferred term to gather together discussions and create cross-references from the other terms. This works best if the book predominantly uses a single term, or if conventions and understandings have shifted to prefer one term over another. This strategy is also best when terms are truly synonymous. All discussions are gathered in a single array. Use cross-references to redirect readers from the other terms. 
2. Create separate arrays for each term, with cross-references between them. This works best if terms are similar, not quite synonymous. Keeping the arrays separate emphasizes the distinctions, while the cross-references alert readers that more can be found elsewhere. 
3. Combine terms into a single heading. This works best if terms are used interchangeably in the text. For example, if terms are synonymous and there is no clear preference. Or, if terms are distinct but mostly discussed together. In the queer Indigenous example above, the volume editors and I settled on the main heading “Two-spirit and Indigiqueer” to acknowledge that these are distinct identities that are nonetheless discussed together. When combining terms, headings can either be double-posted or, if subheadings are needed, a cross-reference included from the second term. 

Include All Alternative Terms?

How proactive should indexers be finding and including alternative terms? Is it acceptable to stick to terms in the text or should the indexer look beyond?

This is a judgement call that each indexer will need to make. If synonymous and similar terms are in the text, then yes, those should definitely be in the index. Preferred terms in the index should ideally reflect the text.

If terms are not in the text, it is not necessarily the indexer’s responsibility to find and include. Part of this is due to the limits of the indexer’s knowledge. I am not well-versed on terminology in Australia or the UK, for example, so if a British or Australian term is not in the book, I probably won’t think of it. I may also miss synonyms if the subject matter is outside of my range of expertise.

The inclusion of synonyms and similar terms should also be reasonable for the intended audience. If the book is being published in Canada for a Canadian audience, cross-references from American or British terms are probably unnecessary. I usually include the term “Native Americans” in indexes for American publishers and audiences but not for a Canadian audience.

If several synonyms exist, I think it is good enough to include what most readers are likely to search for, rather than trying to exhaustively find and include every possible term. In Canada, “Indigenous” is now well established, so I no longer feel the need to include a cross-reference from “Aboriginal,” unless the book uses the term. Similarly, when considering the terms “gay,” “LGBT,” “queer” and “2S+LGBTTQ*,” among others, does the audience really need a cross-reference from each? Maybe, maybe not. 

What I am trying to find is a balanced approach. While trying to be thorough, I am also mindful of my own limitations. I am also aware of diminishing returns, for myself if trying to think of and include every last synonym, and for readers, if the chances of readers searching for a particular term are miniscule. I am trying to provide the most reasonable access for the most readers; there may be a few outliers that I can’t control. 

Synonymous and similar terms can be difficult to spot and parse. Pay attention to how terms are used in the text, pick a strategy, and make sure that similar terms are somehow connected so that distinctions and similarities are clear to readers. 

I’ve been learning indexing in my free time (and admittedly have plenty of it these days!) and would love some book recommendations to practice on — ideally titles that already have professional indexes so I can compare my work. I’ve heard that even between two skilled indexers, their results can differ by about 20–30%, which makes sense since indexing is as much an art as it is a technical craft. Do you have any suggestions for how I can tell whether my index is on the right track? When comparing mine to the published version, what kinds of differences are worth noticing or learning from?

Thank you, Clara. That is a great question. I think there is a lot here to unpack.

You are right that each indexer is unique and will not create the exact same index as someone else. This can make comparisons between indexes tricky because differences between indexes are not necessarily wrong.

I am not going to recommend specific titles to practice on. I don’t know what books are available to you, or what your background and interests are. But I do have a few ideas for how to practice effectively.

Even though each indexer will write a different index, all good indexes should still be grounded in indexing best practices. So to start, I recommend becoming familiar with best practices, both as a lens for analyzing other indexes and as a guide when writing your own.

The American Society for Indexing (ASI) has a handy PDF of best practices available here (I’ve had some trouble with the link, so if it doesn’t work, look for the best practices PDF in the sidebar). For a more concise listing, I recommend looking at the criteria for both the ASI Indexing Awards and the Indexing Society of Canada/Société canadienne d’indexation (ISC/SCI)’s Ewart-Daveluy Award. These are what the judges are looking for when assessing nominations, and are also good guides for what to incorporate in your own indexes.

Once familiar with the qualities of an excellent index, start dissecting published indexes. You can start with the award winners. How do these indexes reflect best practices? What is similar across all of these indexes? What is different? Get a feel for how these indexes work and what you may want to borrow. ISC/SCI makes available all winning indexes, here. For some reason, ASI only makes available some (and primarily older) award-winning indexes, which you can find here.

In addition to prize-winning indexes, examine any index that you can find on your bookshelf, including terrible indexes. How do they exhibit best practices, or not? If an index seems terribly written, can you pinpoint why? How could those indexes be improved? I think understanding what doesn’t work can be just as valuable as understanding what does work. Something can be learned from any index, whether well-written or not. 

Now that you have a better feel for indexing best practices and how they are used—or not used—try incorporating best practices into your own indexes. 

One way is to deliberately practice. Identify one or two best practices to work on and be intentional about improving those aspects as you draft and edit your index. As you become more comfortable with those aspects, move on to incorporating another best practice.

You can also borrow techniques that you find in other indexes. I did this when I indexed Canada’s Army: Waging War and Keeping the Peace, Third Edition, by J.L. Granatstein. The book contains a lot of detail about dozens of military units across multiple wars. I struggled to figure out an effective way to organize all of that information into a structure that was comprehensive yet still easy to search. I browsed at least a half-dozen military histories before I found an index that I liked and which I thought I could adapt. If you find a particular approach that you like, by all means try using it yourself. Remember, though, that not all techniques work in all contexts. You may need to adapt techniques to fit the contents of the book or try a different technique altogether. Part of being a good indexer is having a few strategies on hand to pick and choose from.

You can also practice different approaches within the same index. You don’t need to rewrite the index multiple times. Rather, conduct a thought experiment. What is another way to structure the index? How would you trim the index by a quarter, if space was a limiting factor? By half? How could you handle cross-references differently? How would these decisions affect the index, for better or worse?

One final word of advice: When practicing, practice on easier books, whatever easier means for you. Indexing techniques are the same whether the book is difficult or easy, so I think better to practice on something easier so that you can be more focused on indexing decisions rather than getting bogged down trying to understand the book. As you become more familiar with indexing, it will become easier to apply those same skills to more difficult books. This was my experience, as I wrote my first several indexes while working in-house for a trade publisher. By the time I indexed my first scholarly book, it was less intimidating and felt like a natural progression because I had experience to build upon. In contract, I remember one new indexer I mentored who floundered after choosing a book to practice on that even experienced indexers would have struggled with. That indexer did have a background in that subject, but even so, it proved too challenging when combined with practicing indexing skills. So do yourself a favor and choose an easier book. Harder books can wait.

It is important to learn how to analyze your own indexes and to know how your indexes compare. To do that, it is important to understand and recognize best practices, both in your own indexes and in indexes written by others. So spend some time reviewing best practices and using them as a lens to analyze. Learn to identify specific strategies and techniques, and then apply those lessons to your own indexes. If you can do that, you are well on your way to writing excellent indexes yourself.

Counterintuitively, I enjoy indexing because it is difficult. Sure, there are plenty of days I complain about a chapter being too obtuse or I’m anxious because an index is taking longer to edit than I anticipated. But if writing an index were easy, I’d lose interest and want to do something else.

I’ve been thinking of this in the context of a recent article by Rebecca Solnit, published in The Guardian. Solnit writes about the lies that Big Tech tells us, including about AI, and what we can do to resist.  One of the lies that really resonated with me is that our work is too difficult. Why should we have to do hard things ourselves? Better to give it to AI to do on our behalf. 

What Solnit points out, as does Oliver Burkeman in his book Four Thousand Weeks (which I also highly recommend), is that it is often the friction in our work and daily interactions that makes life meaningful. By friction, they don’t mean conflict, but rather things like needing to interact with a human barista to order a coffee as opposed to ordering through an app or screen. For me, tasks with good friction would also include mowing the lawn or shoveling snow, which, while involving time and effort on my part, are also opportunities to be physically active outdoors in the fresh air away from my computer screen. Or, the friction involved in writing an index. For me, at least, it is the intellectual labor of problem-solving and learning interesting things, along with feeling like I am part of a larger team working together to produce the finished book, that makes indexing a meaningful and worthwhile endeavor. 

Of course, no one wants their life to be all friction. What is a pleasurable challenge for one person may be someone else’s mind-numbing drudgery. For me, while I probably could design a new website for myself, it is not my top interest or skillset and I’m happy to hire someone else. I also see this with authors and publishers who hire me to write an index. Could they write the index themselves? Probably, given enough time and maybe some coaching. But they already have enough difficult things to do and hiring an indexer is easier. 

So the value of work, for our satisfaction and well-being, lies partly in its difficulty and our ability to overcome that difficulty, yet we don’t want everything in our lives to be difficult. Doesn’t this swing back around to the value of AI or other Big Tech tools? Why hire a web designer or an indexer when you can instead hire an AI tool to do the work for significantly cheaper and faster?

I think this speaks to another lie that Solnit identifies, which is the scarcity of people. It is too hard to find someone qualified and available. They probably don’t exist anyway. Better to use AI, which is always available.

I think this lie is quite insidious and perhaps a self-fulfilling prophecy. If we, as a society and industry, don’t nurture human talent, is it any wonder that it may disappear? But that human talent does exist. And the friction of interacting with other people, to delegate and work together, makes the work more meaningful. It cuts through the isolation of being siloed alone with my computer and whatever apps Big Tech sells me.

I am concerned about the new AI indexing tools which are being developed. There is the concern of whether those tools will ever develop to the point of being legitimate competition to the quality of human-written indexes. There is the concern of whether some publishers, in their push to industrialize publishing and squeeze out all the profit for themselves, will use AI anyway, despite quality issues, simply because it is cheaper and faster. Those are excellent concerns and worth paying attention to.

But what feels even more existential to me is the question, do I even want to work with AI indexing tools? Even if I could leverage such tools to work five or ten times faster, indexing ten or fifteen books per month instead of three to five, would I still enjoy the work? Would writing an index still be a satisfying intellectual puzzle? Or would I be delegating most of that intellectual labor to the machine, and churning out indexes too quickly to meaningfully absorb and process what I am reading and piecing together?

I think the answer is no, I would not enjoy that kind of work.

I have to admit I am starting to think about how I may exit indexing if the industry changes too much and human-written indexes are no longer valued. What else could I possibly do where I can continue to problem-solve, be creative, and still feel like I am part of a community of people?

Thankfully, I don’t think I need to make a decision yet. The silver-lining, perhaps, is that the publishing industry is vast and segmented. Even if some segments turn to AI tools—and I’m pretty sure some will, the same segments that are already heavily invested in contracting to offshore companies and industrializing the production process as much as possible—I’m hopeful that not all segments will. Some publishers are still committed to creating quality books and to working with human freelancers. 

I don’t want to be completely against AI. It sounds like there are some areas in which it is genuinely useful. And I don’t want to alarm you too much, especially if you are new to indexing. I want to believe there is still a future for human indexers, and that the work will continue to be challenging and meaningful in all the right ways.

I think it is also important to try to be aware of new developments and to think through all implications, beyond the hype and lies that Big Tech wants us to focus on. 

If AI tools are adopted by indexers and in the publishing industry, they need to be accurate and reliable, obviously. But more than that, they need to support human work and human workers. Replacing indexers (and editors, designers, etc…) does not make for a healthy work environment, nor, I would suggest, for books that are written and produced with humans in mind. For an industry and world in which humans can thrive, people need to be kept at the forefront. 

I’ve been thinking about the saying that two indexers can index the same book and write two very different indexes. Which is true. This was demonstrated during the panel moderated last year by Jolanta Komornicka, at both the Indexing Society of Canada/Société canadienne d’indexation (ISC/SCI) and Society of Indexers (SI) conferences. For the panel, a number of experienced indexers indexed the book The Claremont Run: Subverting Gender in the X-Men, by J. Andrew Deman (University of Texas Press, 2023), and compared and discussed their decision-making, including how their indexes differed. I was not part of the panel, though I very much enjoyed the discussion at the ISC/SCI conference in Vancouver.

For all the differences—and each indexer did have their own approach—the overlap was also striking. It reminded me of something Enid Zafran said, that even though two indexers will write different indexes, there should still be significant overlap because the text itself remains the same. Everything that is in the book still needs to somehow be reflected in the index.

I am going to call what’s in the book aboutness, to use index-speak. In my book, Book Indexing: A Step-by-Step Guide, I define aboutness as “understanding what the text, at all levels, is about.” Meaning, what is the book as a whole about? What are the main concepts? Who are the relevant people? What organizations are discussed, or court cases cited, etc…? All of these discussions and mentions need to be in the index. How they appear in the index may differ, but one way or another, the contents of the book should be recognizable and findable. 

So there is this tension, in writing an index, between the objective and subjective elements. The text is the text. The rules and conventions are the rules and conventions. But the application and interpretation will vary.

I’m conscious of this tension when I subcontract. I want certain elements handled a certain way, while also aware that the subcontractor may have a different view or may be dealing with a different preference if they also subcontract for another indexer.

I think this tension also exists for anyone who has served on a jury for an indexing award. When I was on the Award Committee for the Ewart-Daveluy Award, some submissions were written differently than how I would have written them. Yet the award is not for, “who can best mimic how the judges would have written the index.” The award is for excellence, firstly for how well the index reflects the book, and secondly for how well the elements and strategies used in the index work. As a judge, I need to put aside my personal approach and judge submissions on their own merits, accepting that best practices can be interpreted and implemented in a variety of ways.

So what does all this mean for you or I as indexers?

First of all, the objective side of indexing is important. It is important to understand the book and accurately point back to it. The formatting and other conventions exist for a reason. Rules can be broken, but any deviation should be well thought out and justified.

It is also okay, however, to develop your own style. I know it sounds weird to discuss indexing in terms of style, but I think our personal preferences and methods do add up to make our indexes unique. If we are going to push back against a bland average that AI seems to excel at, maybe we need to talk more about what makes our human-written indexes stand out.

So as you practice and write indexes, consider. Of various approaches available, which do you like the best and why? What works better in certain scenarios and not others? What looks more elegant? What saves space? What do you want to borrow from others, and what do you disagree with? While keeping an eye on the goal of creating excellent indexes for readers, put your own stamp on your indexes too.

Repeating this out loud to myself, it sounds like a trick question. The value of an index is tied, in part, to its accuracy, right? Of course an index needs to be accurate.

Considered another way, I don’t expect a human written index to be 100% accurate. Perhaps instead of “accurate,” better to say “correct.” Humans are fallible, including myself. I know, or at least suspect, that errors make their way into the indexes that I write. Clients will occasionally point out the odd misspelling or omission that they spot during their review. For which I apologize and gladly correct.

But I still strive for the indexes I submit to be as accurate as possible. I believe that my editing process is thorough and I am confident that I catch most errors before indexes ship. So is the accuracy of my indexes good enough?

I have a theory that if a reader finds more than 3-4 errors or barriers in an index, then the reader will lose trust in the index and will stop using it. I don’t actually know if this is true or not; I haven’t scientifically tested it. And, a user may simply be unlucky enough to hit upon the handful of errors that do exist, even if the index overall is quite good. But I think there is something true about people not using tools if they lack trust in the tool’s usefulness. An index needs to prove itself to be trustworthy, and part of that is being accurate.

Before I go further, what do I mean by accuracy? A subject index, after all, has an inherently subjective aspect in the decisions about what to pick up, how to phrase headings and subheadings, how to structure the index, and how best to fit the space available. How can accuracy even be measured?

To start, as a baseline, locators need to match the text. For a subject index, the the broad sweep of the book needs to be recognizable in the index. How exactly the contents of the book are translated into the index will vary from indexer to indexer, and also depend on constraints such as space, but generally speaking, the reader should have an accurate sense of the book from the index. For other types of indexes, such as a name or scripture index, the general rule is that every mention is picked up, and so in that context, accuracy also encompasses thoroughness.

If a reader can trust that an index is accurate in these senses, then I believe they are more likely to use the index. If a reader loses trust because they are finding errors or omissions, my theory is that they will stop using the index and the index is now wasted words on the page. 

I’ve been thinking about all of this because I was contacted last week by a new company, IndexerLabs, which is developing AI indexing tools. IndexerLabs has already released a tool specifically for scripture indexes and appear to also be developing a tool for subject indexes.

For scripture indexes, IndexerLabs claims an accuracy rate of 99.1%, compared to the original, human-written indexes. It sounds impressive, doesn’t it? One of the criticisms of AI indexing tools, from the American Society for Indexing (ASI)’s AI committee, and others, is that these tools are neither accurate nor thorough. Has IndexerLabs managed to crack that problem?

My concern, though, is that an accuracy rate also implies an error rate. For a short index, with about 300 scripture references, a 0.9% error rate would lead to about 2-3 incorrect entries. For 600 references, about 5-6 incorrect entries. For 1500 references, about 13-14 incorrect entries. And so on, as the number of scripture references increases.

Personally, I am not comfortable with there being 13-14 incorrect entries, or more, even for a large book and index. I would expect myself to do better. If I subcontracted, I would expect the subcontractor to do better. Not a 100%, given that we are human, but better. 

Humans have the potential to recognize their mistakes and to build in ways to double check and correct. When I write a scripture index, or any index, I frequently stop—a few seconds at a time—to confirm that locators and headings are accurate. So far as I know, AI finds and decides based on an algorithm and probability. It does not review its own work.  

If an AI tool has a discernible error rate, then how much error is acceptable? Am I supposed to still approve the work? Is the author and publisher supposed to accept that a certain percentage of the index is likely inaccurate? Or, do I try to edit the output to find and correct errors? Editing someone else’s index can be very tedious and time-consuming, especially if I don’t trust the work and feel compelled to check every entry. So editing the work could negate the time-savings. 

My purpose in writing this isn’t to beat on IndexerLabs. I have not yet tried their tool. I also acknowledge that I have, in the past, expressed interest in an AI tool to pick up specific, easily identifiable entries, such as scripture references, that can be time-consuming to pick up. While I remain skeptical about AI’s ability to index, and I believe that any AI tool should be thoroughly tested, I also think it is important to keep a somewhat open mind and to be aware of new developments.

I am concerned, though, when the completeness or readiness for publication of an index is framed in terms of its accuracy as measured by a percentage. While agreeing that indexes absolutely need to be accurate, if accuracy becomes something quantifiable, does that normalize a certain amount of error?

Being human, I know that no index is perfect, including the indexes I write. I’ve made my peace with that. I also still think that humans are best positioned to write accurate, truth-worthy indexes. Humans best understand the end user and the end goal of the index, and humans can learn how to find their own errors and to course correct.

I don’t want to live in a world or work in an industry in which error and substandard quality is accepted simply because the tool is cheap, convenient, and fast. We can do better, as a society and industry. Humans can do better. 

I’ve been thinking about this question over the last couple of weeks. There have been parallel discussions on the Indexer’s Discussion Group (IDG) and CindexUsers email lists about AI and indexing, in response to Ben Vagle’s announcement of his latest AI indexing tool, Indexia. If you are not part of those email lists, you are welcome to join and go back to read.

You may recall that Ben released a different AI indexing tool earlier this summer, AI-indexing, which I mentioned here. This new tool seems to be an upgrade or a rebrand; I’m not sure how it is different, to be honest. It is described as a “professional indexing assistant,” capable of “generat[ing] a fully structured initial draft with terms, contextual subentries, and page references—giving you a professional starting point in minutes, not weeks.” I believe Indexia is still in the beta stage. I have not tried it.

In the ensuring discussions and Ben’s responses to questions, I was struck by the emphasis on term extraction. In Ben’s program, AI identifies and extract terms, along with their context and locators, which the human indexer can then edit. A “fully structured initial draft” is promised, but what does structure mean in this instance? Looking at the examples on the Indexia website, I wonder if structure actually means format, as in alphabetical sort and punctuation and tabs in all the right places. 

I thought again about term selection and structure last week while corresponding with a client. When the author contacted me, they also sent a list of terms that they had already written. I thanked them for the list, and explained that I would take it into consideration while also reading their book and selecting terms myself. The author then asked about my process for creating “the list.” Which is a great question and I was happy to answer. But I was struck by the use of the word “list.” When I index, I don’t think of it as compiling a “list” but rather as writing an index.

In both of these instances, I wonder if there is a misconception that an index is essentially a list. It is as if structure is a byproduct of arranging headings and subheadings in a certain way. No more complicated than that.

Flip through a handful of nonfiction books on your bookshelf and you are likely to find at least one index that does look like a list. There is no, or very little, sense of movement and relationship between terms. There are no, or very few, cross-references. Subheadings may be sparse. All of the headings—names and concepts—exist on one level. 

I like to think of those indexes as flat. They give the impression that not much is happening.

For some books, a list-like approach is indeed appropriate. I am not trying to say that every index needs to be complicated. But if the author or indexer assumes that an index is simply a glorified list, then the resulting index is probably not going to be very good. What is missing is a clear structure.

I like to think about structure in terms of the hierarchy of information. Information within any given book is typically layered or nested. At the top is the metatopic, which is what the book as a whole is about. Below the metatopic, often given their own chapters, are the supermain discussions, which are the key facets or arguments of the metatopic. Below the supermain discussions are the regular discussions, which flesh out the supermains. At the bottom are all the little details and examples which further give life to the book.

Mirroring the book, the index should somehow reflect each of these layers. The metatopic, supermain, and regular discussions should each have their own arrays. The relationships between topics and levels should be clear through subheadings or cross-references. Instead of a list, the index becomes an interconnected web, allowing the reader to enter the text through multiple entry points. This is what I mean when I say that the index needs a coherent structure.

What the structure actually looks like for any given index will vary. While I often have an approach in mind when I begin an index, I don’t recommend trying to shoehorn an index into a specific template. It is important to pay attention to how the book is actually written and how the information fits together.

That said, by way of example, here are three different structures for three different types of books:

• A book with a strong, central metatopic, like a biography or a corporate history. This will likely have an extensive metatopic array for the central subject. I don’t like breaking up a person’s life and scattering those chunks throughout an index. Other main topics and details, which can stand alone, would be their own arrays, revolving around the central subject.

• Other books, say a business book on leadership, can more easily be broken down into their constituent supermain discussions. Each leadership quality, in this example, would receive its own array with subheadings. The metatopic array, for leadership, would likely be short, mostly consisting of cross-references to the various qualities. This is a more decentralized structure, anchored by the various qualities. 
• Other books may be even more decentralized. A hiking guide, for instance, is essentially a list of hiking trails, along with details of scenic mountains, lakes, or rivers, which readers may also search for. A metatopic array for hiking is probably not useful, unless there is generic hiking advice or safety tips which need to go somewhere. Supermain arrays may also not be relevant, though if the hikes are organized by region or theme, then perhaps those regions or themes should also be reflected in the index. Instead, the emphasis in the index should be on the individual hikes, along with other relevant details. This may end up looking like a list, and that’s okay because it suits the content. 

In the three examples above, the index structure varies considerably, from an extensive metatopic array anchoring the index to the emphasis being on all the details. The appropriate structure depends on what the book and audience needs. What remains true in each case is that the choice of structure is deliberate. 

When structure is ignored, or assumed to be a byproduct, or thought to be synonymous with format, I fear that the index is not going to be nearly as useful or easy to navigate. A list is fine. It’s also inert, doesn’t indicate relationship or the depth of complexity that exists within a book. I know that the index is supposed to point to information, not reproduce the book, and I think that giving the index a clear and deliberate structure is the best way to point towards information, rather than compiling a list which is easy to gloss over. 

A good structure is also part of the value that a human indexer can bring. As human indexers, we do more than extract terms. We can organize and structure those terms to provide context and to facilitate navigation and retrieval. We can go beyond a list.

So be human. Be creative. Be intentional. When writing an index, let structure guide your decision-making, alongside term selection. 

If I may dare say, let’s help make the world a better place, one excellent index at a time.

Today’s question comes from Kimberly:

 I’ve been giving each heading a subheading right from the start, with the idea that it might be easier to delete unneeded ones than to go back and figure out what they should all be. Do you think this is more likely to speed me up or slow me down?

Great question, Kimberly, and thank you for asking. I appreciate that you are thinking ahead and anticipating that subheadings may be needed. Adding subheadings early can certainly be more efficient, while that information is fresh in your mind, than waiting until later. But is it more efficient to add subheadings to all headings? I’m not convinced about all. 

If you look at most indexes, most arrays lack subheadings. The need for subheadings varies from book to book, but generally speaking, while all arrays should be relevant for the reader and point towards substantive information, only a few arrays are discussed extensively enough, or can be readily divided into subcategories, to warrant subheadings. So spending time creating subheadings for all headings runs the risk of creating a lot of subheadings which will later need to be deleted. Both writing and deleting subheadings takes time (though you are right that deleting can be easy).

A better approach is to try to anticipate which arrays will need subheadings and which will not. This way you can still be proactive about creating subheadings where needed while avoiding unnecessary ones.

Here are two ideas for how to determine whether subheadings are needed:

• Before beginning, get a sense for what the book is about. Maybe scan the table of contents, read the introduction, flip through a few chapters to see how the book is written. What is the book about? What are the main arguments, themes, or topics? For example, books are often organized so that each chapter has its own topic, which will likely need subheadings. Are there types of entries that may require subheadings? In a corporate history, for example, key leaders will probably receive a lot of mentions, enough to need subheadings. In a work of literary criticism, certain key authors and books will probably also need subheadings. Try jotting down a list of likely key topics. When indexing, be on the lookout for them and add those subheadings.
• Run a quick keyword search when you first encounter a name or topic. If you only get three hits, then no subheadings. If you get more than six or ten hits, or whatever your threshold for adding subheadings is, then yes, subheadings are probably necessary. Running quick searches does take a moment, but still quicker than adding and deleting.

If you do realize, partway through, that an array should have subheadings after all, don’t panic. Begin adding subheadings from the point you make that decision. If a few previous locators are still fresh in your memory, go ahead and backtrack to add those subheadings in too. Otherwise, it’s okay to fill in a few missing subheadings at the end, while editing the index. I find that adding two, four, or even six subheadings doesn’t actually take that much time. While it is good to be proactive, missing a few or changing your mind later isn’t that big a deal either. 

Another consideration is that I don’t always know what the subheading should be until I’ve read the whole book and better understand the discussion. So while creating a subheading from the start is a good idea, it is not always feasible. If you find yourself struggling to phrase the subheading or identify what the discussion is about, it’s okay, and may even be faster, to move on and come back to it later.  

Anticipating whether or not subheadings are needed is not an exact science. I still sometimes realize that a discussion is smaller or larger than I thought. But if I can correctly anticipate whether subheadings are needed for most arrays, adding or deleting a few subheadings is still quicker than adding or deleting for all arrays. 

The rough draft is not supposed to be perfect. Yes, trying to get entries right from the start can pay off later with less editing, and I think that maintaining a sense of momentum is also important. I don’t know what is in the book until I finish reading it, and I can’t finalize the index until I know what is in the book, and so while I want to put down my best attempt on the page while drafting, I also want to keep moving until I have a full draft and can assess what I have done. Subheadings are part of this process. Have some strategies to do your best while drafting, while also knowing that it will not be perfect and that it is okay to circle back and revise as needed.