Filter words dig into prose like ticks on a sheep’s back. When I draft, I don't even realise I'm using them — and I suspect many authors are the same. Come revision time, and they are one of the first things I look for with the aim of re-writing them out of existence.

What are filter words?

But firstly, what are these horrible things of which I speak?

Filtering shapes the reader's experience through the lens of a character. While we as writers want readers to engage with our characters, excessive filtering falls into the trap of showing not telling. By filtering a reader's experience, you're inhibiting their ability to directly engage with the story — at least that's the theory. It also slows things down because filtering is generally more verbose.

So, here's a list of the most common filter words:

  • to see
  • to hear
  • to think
  • to touch
  • to wonder
  • to know
  • to realise (realize, to Americans)
  • to watch
  • to look
  • to seem
  • to feel (or feel like)
  • to decide
  • to sound (or sound like)
  • to notice
  • to be able to
  • to note
  • to experience

Surprised? These verbs are bland, boring and can add unnecessary filler to your prose. They, and the constructions they inspire, are also ubiquitous. Who hasn't read a paragraph like:

Peter felt a sickening feeling as he realised the foreclosure was immanent. He saw a black sedan ambling up the driveway towards the house, and he knew it was the bank manager. He heard his wife sobbing in the corner. He looked at the rifle mounted above the mantle and decided to shoot the thieving bastard dead.

Crikey, that was more grim than I intended! The words in bold type filter the story through Peter's point of view.

I've written thousands of paragraphs like this while drafting a story. I do so because it's a means of keeping the POV character in my head. But come revision and its time to craft that steaming pile of tell into something more appropriately show.

If we remove them, we allow the reader to experience the scene with more immediacy and less abstraction, like so:

Peter's guts twisted. Foreclosure — the day had come. A black sedan ambled up the driveway towards the house, a fleet car owned by the bank. His wife sobbed in the corner. The rifle beckoned from its mount above the mantle. I’m gonna shoot that thieving bastard!

I hope you agree that the second paragraph reads and feels better (more visceral given the tone). With the filtering gone, you're able to experience the scene as Peter does, as the events unfold. It also afforded me the chance to vary the sentence structure and add Peter's inner monologue.

Still, before continuing to the practical stuff with Scrivener, let's temper our enthusiasm.

Is filtering really that bad?

No.

Personally, I don't subscribe to the belief that any particular construction or style is necessarily bad. Filtering has its place, as does the other evils of prose, even the much maligned passive voice, or worse…the dreaded adverb. English is flexible, so don't let some tosser tell you one way or another is better. Good style is in the eye of the beholder — universities aren't wasting grant money researching the effect of different writing styles on the reading public.

As a reader (and this distinction is important), I value variety. I don't like prose that's boiled down and denuded of all colour, rhythm, and variation. Writing it also sucks as much as reading it.

Sometimes filter words are necessary and even desirable. There are situations where you will want the reader to experience a character's thoughts and perception through their eyes/ears/thoughts. This might be for cause and effect: a character has an experience and must react. Maybe hearing his wife cry make Peter angrier or ashamed or come to his senses — in that case, having Peter hear then process his wife's cries is justified, otherwise there's an awkward disconnect.

Removing filter words can also create ambiguity, particularly if there are several characters at play. Even if you write deep POV (current flavour of the month), you still need to establish who has POV and then maintain clarity as the scene progresses. Everyone's had the jarring experience of reading a passage or a piece of dialogue and not knowing who it's about or by. Clarity is just as important — if not more so — than conciseness.

So, scrutinise every word in your prose, but as always, your judgement and experience as an author (and reader) is what counts.

Right, so how do I nuke them?

Revising a hundred thousand-word manuscript is daunting and time-consuming. If every paragraph in your manuscript is like my example above, you have a monumental task before you. You'll have to trudge through, line by line, marking and editing as you go. Maybe you'll use the Find option to locate each word, one after another, in each document of your project. Tedious, right?

What if there was a way to highlight all the problem words with a single operation?

Well, in Scrivener there is! Scrivener has a powerful search engine, and it supports regular expression syntax (regex). Normally, the domain of developers, regex allows you to search for patterns of text rather than their literal value. If you've never used them before, they can seem extremely daunting, but fortunately the pattern we need for our purposes is probably the easiest one you can write.

So to begin, we're going to perform a project Search, but with a couple of options selected. Note, that I’m using Scrivener 3 on macOS, but I assume this works on Scrivener 3 for Windows.

The options we want to enable are Text and Regex, which means we want to search in the text of a document, and enable Regex syntax. You can enable these option in the Search widget’s drop-down menu as shown below.

Scrivener search options
Scrivener search options

With our options selected, we can now enter our search pattern. As I noted, this kind of search (looking for multiple words) is about the easiest regex you can do. Here’s one based on the list of filter verbs I mentioned above.

(felt|realised|knew|saw|heard|looked|thought|touched|wondered|watched|seemed|decided|sounded|noticed|noted|experienced)

Paste it into the search field and hit Return. As I’m searching across the entire project, the Binder now shows a list of all documents where one or more of these words show up. What I love about this feature is Scrivener’s ability to hightlight every matching word matching the search, for the entire manuscript.

Scrivener search results and word highlights
Scrivener search results and word highlights

Obviously, you aren’t limited to these filler words, and can use this construction to highlight anything, such as character names, locations, weasel words and jargon, and my personal favourite— profanity.

Optionally, we can save the search as a Collection. Not only does this allow us to return to our edit later, but Scrivener automatically updates the collection as we write and edit. I'm hoping as you read this, you'll have a light bulb moment — using this technique you can create a collection for anything, the only limit is your imagination and what’s possible with regex.

More patterns

The pattern above barely scratches the surface of what regular expressions can do. Here’s a couple more that I find useful in my editorial process.

Duplicate words

Duplicating words is a terrible habit of mine. I’ll often repeat words when I draft, thanks to an occasional disconnect between my brain and fingers, or when I’ve had to leave the keyboard temporarily. I mostly repeat pronouns and common words like it and said, which my brain seems to ignore.

This regex pattern locates most instances of word duplicates in your manuscript.

\b(\w+)\s+\1\b

Gerunds

Gerunds are sentences beginning with ing, and while I don’t mind them as much as other authors, they can often result in a dangling modifier. So, I find it useful to highlight them when I’m editing.

This regex patter locates any sentence that begins with a word ending in ing.

^[a-zA-Z]*ing\s+

Note that it will also pick up words like something, not just verbs, but I still find it a useful pattern.

Concluding thoughts

Powerful doesn’t begin to describe this feature. The combination of regex and collections makes Scrivener an editorial powerhouse, far exceeding the abilities of general word processors like Microsoft Word. Build up a collection of searches to help catch your common mistakes, or highlight constructions you want to improve, and you’ll save a lot of time in editing.

Hopefully, I’ve given you enough to get started. Yes, regex can be hard to learn, but if you wrangle text for a living, or for pleasure, it’s worth the effort.

If are struggling though, and have a particular pattern in mind, reach out, and I’ll help you write what you need.