I regularly advise writers to use grammatical parallelism and visual formatting to influence document quality. (Use the links if you don’t know what I mean.) But I saw some evidence presented by colleagues at a recent conference that led me to refine that advice. Here’s the bottom line for those who don’t want the details:
- Use both parallelism and format to improve
readabilityreading efficiency and accurate identification and recall of information. - Use format alone to improve
readabilityreading efficiency and accurate identification and recall of information. - Don’t use parallelism alone to improve
readabilityaccurate identification and recall of information.
It pains me to offer you #3. It runs counter to what I’ve long held to be true. I suspect that’s true for many of you as well. But I can’t ignore the evidence. Ugh. Details below.
July 22, 2014 Update: I’ve tweaked the wording of my guidance based on more investigation of related research. See this post for details.
Some Background on the Research
The researchers did two studies where they asked people to read documents and then tested their recall by asking multiple-choice content questions without allowing them to look back at the documents.
Study 1. To test the effect of grammatical parallelism and visual formatting on readability, 100 people read a version of a passage from Ordnance Instructions for the United States Navy from 1866. (This limited the possibility of influence based on previous knowledge.) There were 4 document versions: (a) parallel [P+] and formatted [V+], (b) parallel and unformatted, (c) nonparallel and formatted, and (d) nonparallel and unformatted.
They found that accuracy was most affected by visual formatting and not much at all by grammatical parallelism. These results are the basis for my first two pieces of advice at the beginning of this post.
Study 2. To test the effect of grammatical parallelism and visual formatting on emotional response, 87 people read one of the four versions of the same passage from Study 1 followed by a question measuring their emotional response and then received either (a) parallel [Q+] or (b) nonparallel multiple-choice content questions again followed by a question measuring their emotional response.
They found the greatest inconsistency of emotional response in document versions without visual formatting and with grammatical parallelism of both the reading text and the multiple-choice questions. These results are the basis for my third piece of advice at the beginning of this post.
So What’s Wrong with Inconsistent Emotional Response?
If you’ve read this far, I’m pretty sure this is the question you’re asking. Here goes.
Unity is a fundamental quality of human perception. Humans actively seek and prefer experiences that we can interpret as unified. This idea from Gestalt psychology is one I wrote about extensively in a 1995 book on document design. The lack of unity or consistency in color is the reason Smashing Magazine used the website shown at right in its Ugly Showcase.
My colleagues, Nicole Amare and Alan Manning, argue that unity or consistency is especially important for types of textual elements they call “decoratives.” That includes grammatical parallelism (and color). Decoratives are aesthetic. Parallelism does not convey information itself. Rather, it evokes a feeling about that information. Many people describe parallel text with terms related to organization: tidy, orderly, tight, neat, uncluttered, etc. As Manning said,
Parallel but visually unformatted text evidently disrupts that unified feeling and is therefore less desirable.
Visually formatted but nonparallel text was not a problem, confirming visual format is far more salient to readers than parallelism. I suspect inconsistent feelings slow down cognitive processing and distract us from our quest for comprehension. I’ve long believed parallelism affects efficiency more than effectiveness in documents with a primarily informative function. But that research hasn’t been done — yet.
References
Amare, N. & Manning, A. (October, 2013). Grammatical and visual parallelism in business communication pedagogy. Association for Business Communication Convention, New Orleans, LA.
Amare, N. & Manning, A. (2013). A Unified Theory of Information Design: Visuals, Text & Ethics. Amityville, NY: Baywood.
Campbell, K.S. (1995). Coherence, Continuity & Cohesion: Theoretical Foundations for Document Design. Hillsdale, NJ: Lawrence Erlbaum.
I’m not disagreeing, but surely part of the difference in response has less to do with the parallelism (or non-) than the fact that the non-parallel options use language that is more descriptive/more common usage/less bureaucratic/more plain language. So the recall may be more related to that than to the presence or non-presence of parallelism.
Also, I do wonder about using a recall test to measure “readability.” Isn’t it really measuring recall? Surely there is information that we need to be able to recall (the 3 digit number to call in an emergency), but more often we simply need to be able to find the information within a document. (Of course, we do have to recall which document the information was in.) Just thinking aloud.
Thanks for sharing your thoughts, Susan. Before I guess about this, I’m hoping Manning or Amare will reply . . .
Thanks, Susan, for your comment, and I do see how the non-parallel passage appears more accessible but, unfortunately, I don’t see, in objective terms, how to support the claim that the visually formatted passage is in plainer English than the parallel passage. True, the parallel passage has all parallel passives, but evidence does not support the traditional assumption that passives are inherently less plain-English than actives (see Kim’s post on that from Oct 15 here on Proswrite). The visually formatted passage alternates passive and active sentences; otherwise, the wording and phrasing of the two passages is about the same. So, perhaps the perception of more common/plainer language of the non-parallel passage, I would suggest, is caused by the visual formatting itself, which is our very point.
Admittedly, recall is only an indirect measure of readability; however, it is a reasonable indirect measure: surely readers will remember more from readable text than they will from unreadable text. In an objective empirical study, we can’t directly measure a reader’s private experience reading a text, so all we have are indirect measures: accuracy of recall, reading speed, reader’s reported emotional response, and so forth. We welcome and strongly encourage others to run their own experiments to test our findings, using whatever measures they think are equally accurate or more accurate. Getting to the truth has to be an ongoing group effort.
In terms of parallel vs. nonparallel passage style, I find the nonparallel passage to be wordy and rambling, which makes me feel like the passage is less clear. However, there must truly be a ruse of clarity (see Barnard’s 2010 CCC article on clarity) because the parallel, non-visually formatted text actually almost lulls me into hypnosis with its marching drumbeat of parallel passives, whereas the syntactically chaotic passage keeps me awake because I get those line breaks.
Sad but true: it looks more and more like Randall Walker, MD and colleagues were on to something about the area of eye vision and what the mind could take in during a single gestalt of text lines (see page 2, figure 1: http://faculty.washington.edu/farkas/TC510-Fall2011/Walker-VSTF-IPCC-07.pdf)
More studies are needed: longer passages, timed reading, and longer wait times before testing recall. Even eye tracking would be helpful to measure, I think, when comparing the two passages.
So we editors have been wasting our time all these years making the elements in a list parallel?
Well, I don’t think so. To me, one lesson here is that we need to use common language when we build our parallel items. Another lesson is that visual chunking of information is additional and complementary. I think that first list works because it groups the information for readers.
As for testing, we have successfully timed how long it takes people to find a particular piece of information in a document. So it’s not recall, but rather their ability to use the navigation cues to be able to find the information. Those cues are not merely headings, but would include whether key words are placed in key positions within the sentence, e.g. the subject position.
I guess the real point here is that these techniques work together to make a document readable and usable.
wondering about men vs. women in the results? The non-parallel version seems to reflect a more narrative structure, and research normally shows that women are good with narratives. Men might have been better at it because of the subject matter; so theoretically, they could have two unconnected conflating variables.
Deanya, I don’t know about the demographics of the sample. It’s certainly a possible influence on the results. I’m hoping Amare or Manning will reply . . .