When passives are easier than actives.

I met WR when I helped set up a recording session in a clinic in Sheffield. I was a PhD student at that time. WR had been diagnosed with primary progressive aphasia, which is a type of dementia that first manifests as a language impairment. WR was a friendly, gentle person who looked young for being 62. His language production was poor. When he spoke, he had the tendency to connect words with "is a" in an ungrammatical manner ("Mary is a holiday is a Turkey"). He preferred using pen and paper, and while his written language was also poor, communication was better through it. There was no sign that WR had problems beyond language. His non-verbal IQ was above average and as far as I can tell he was leading a very active life. At the time we carried out our research, WR's brain showed relatively small signs of degeneration. When Rosemary Varley and I discussed MRI scans with a radiologist at Royal Hallamshire Hospital, he said that if WR had come in with different problems, such as chronic migraine, it may have gone undetected at first glance. But it was there. Scans showed grey matter reduction in frontotemporal areas both left and right.

During a supervision meeting, Rosemary mentioned WR's sentence comprehension results. She tested WR on sentences which are "semantically reversible", such as The man pushes the elephant and The elephant pushes the man. With these sentences you cannot guess who is doing the action solely based on single words and your world knowledge. Elephants can push men and men can push elephants. To understand who is doing what to whom, it is important to interpret the sentence structure. WR performed well when sentences were in the passive voice (e.g., "The man is pushed by the elephant"), but not when they were in the active voice (e.g., "The elephant pushes the man").

It is safe to say that almost no researcher or clinician would expect this profile. WR's comprehension pattern was the opposite of what one finds in textbooks and in most research reports on people with grammatical impairment. Passives are considered more difficult than actives for a number of reasons, and therapies are designed based on these assumptions. Let's have a closer look at the active (1) and the passive (2):

1. The man pushes the elephant.
2. The elephant is pushed by the man.

Passives are longer than actives and contain more grammatical morphology (the auxilliary verb, the verb suffix, the preposition). They also have what is called a "non-canonical" word order. In English it is common for the "agent", meaning the entity carrying out the action (in the above case "the man"), to appear first. In passives, the agent is usually named last, if it is mentioned at all. Passives can omit the agent, which is why the phrase "mistakes were made" has become such a cliché in politics. Apart from structural differences, passives are also rarer than actives and encountered more often in written than in spoken language. Some native speakers with lower educational status have difficulties understanding passives. At the same time, researchers have a reputation for being terrible writers because they use too many passives, often because they try to avoid writing "I" or "we" in order to sound more objective, even though it is obvious that *someone* carried out their work.

We can find more differences when we look at child development. Children learn actives before they learn passives, and the process of acquiring passives seems to go through more development stages. Some linguistic theories assume that to produce or understand the passive, one first needs to access a structure which looks more like the active and then "transform" it to get the passive voice.

The classic agrammatic profile in aphasia describes poor comprehension of passives, and better (or good) comprehension of actives. Researchers working on grammar in aphasia often rely on theories that predict that someone unable to comprehend actives should not be able to comprehend passives at all.

WR, and a very small number of other cases in the literature, challenge these theories.

In our experiments we presented sentences to WR (e.g., The man pushes the elephant) alongside with pictures (e.g., one picture in which the elephant pushes the man, and one in which the man pushes the elephant). One explanation for WR's behaviour could be chance: if some patients make random guesses in these tasks, some will eventually perform like WR did. However, WR's strong performance on passives was consistent across experiments, while his performance on reversible actives remained at chance levels. This means that he likely made a guess when the sentence was in the active voice, but knew what he was doing when encountering a passive.

One therefore has to wonder in what way passives may be easier than actives, and from talking to other people in the field I dare say that most never asked themselves this question. The following are thoughts I developed with Rosemary Varley and Ewa Dąbrowska. For this, let's look at more abstracted forms of actives and passives.

1. X VERBs Y.
2. Y is VERBed by X.
3. By X Y is VERBed.

My favorite explanation is that actives rely more on word order, while passives rely more on processing grammatical morphology. The active sentences The lion kills the man and The man kills the lion mean very different things despite having the same words. It is the order which determines who is the "doer". In passives, the effect of order are more restricted as elements of the sentence are nested within structures which provide very clear cues. The man is killed by the lion means the same as (the more poetic sounding) By the lion the man was killed.

This order reversal works because the passive is based on two grammatical frames. The first is "X is VERBed", and in this frame X is never the agent - it never carries out the action. By contrast, the X in the active frame "X VERBs" can have several functions. X can be the agent (The man pushes the lion) but can also have other functions (e.g., in This dress sells like crazy it's not the dress which conducts the sale, and in The gloves allow him to climb the gloves are not doing the climbing).

The other part of the passive is the preposition "by" followed by a noun; "by Y" most of the time means that Y is doing something regardless of the rest of the sentence. While the preposition "by" has other functions, such as a spatial indicator (the key by the vase), more than two thirds of the time we use "by" to mark the agent (other examples are The news report a strike by lorry drivers and Hamlet, by Shakespeare). Interestingly, there is some evidence that suggests that for those who know how to interpret passives, processing can be faster than processing of actives.

In subsequent experiments we found that WR indeed made use of both the "X is VERBed" and the "by Y" grammatical frames, even when presented in isolation. In one experiment we showed him active sentences with a "spatial by" (e.g., The man by the woman kills the lion), which he interpreted as a cue for agency (e.g., regarding the woman as the killer).

At the same time, WR's ability to process linear order, which is more important for actives, seemed severely disrupted. We saw this not only when testing sentence comprehension, but also when presenting him structured sequences of meaningless shapes. WR was unable to detect the order regularities within these sequences. We suggest that WR's brain degeneration resulted in some kind of "linear order sensitivity". While rarely reported, and certainly not in the focus of clinical research, there is other work that suggest that such an impairment may exist not only in aphasia, but also in some children with language difficulties.

WR's language profile is very rare in the scientific literature. The few other cases were largely ignored. I do believe that these profiles are rare in everyday clinical experience as well. At the same time it is possible that these cases are underreported because they go against what's considered "typical" in aphasia. Selection and publication bias may result in some interesting data ending up in a drawer, simply because researchers don't know what to do with them. When talking to others about WR, I heard of one such unpublished case: after therapy, one patient became better on passives, but not actives. Data that surprise and challenge us belong disseminated, not dismissed. I hope that our work on WR will help fight this bias, and lead to a greater focus on the unexpected.