TL;DR: Both were and was are used when battery of tests is their subject, including in scholarly publications as shown below. Sometimes the choice of number depends on the intended meaning. There may be a relatively recent trend of were becoming a more common choice, but both are frequent.
Your intuition is correct — or at least, it accords with how I would myself say it:
A battery of tests were conducted.
That’s because in English, premodifying phrases such as these act like partitive determiners. That means that they do not alter the grammatical number of the noun that they’re being applied to. So if the noun is plural to start with, it remains plural for purposes of grammatical agreement with a finite verb that it governs as a subject.
All of these take were for me in this situation:
- A lot of my friends were surprised.
- A few of my friends were surprised.
- A number of my friends were surprised.
- A bunch of my friends were surprised.
- A couple of my friends were surprised.
- A pair of my friends were surprised.
- A dozen of my friends were surprised.
- A handful of my friends were surprised.
- A subset of my friends were surprised.
- A great many of my friends were surprised.
- The rest of my friends were surprised.
- Both of my friends were surprised.
- Few of my friends were surprised.
- Some of my friends were surprised.
- Half of my friends were surprised.
- Two thirds of my friends were surprised.
- Ninety percent of my friends were surprised.
- Many of my friends were surprised.
- Most of my friends were surprised.
- All of my friends were surprised.
- None of my friends were surprised.
It doesn’t matter that words like lot, dozen, handful are themselves singular and being used with an indefinite article in front of them. That’s just a red herring and you should not be distracted by it. They are not the logical head of the noun phrase that constitutes the subject of the sentence; the thing they are premodifying is, and it is that thing which governs subject–verb agreement. That’s why both these situations can grammatically occur and are correct:
- A lot of this book is unreadable.
- A lot of these books are unreadable.
As you see, the grammatical number of the subject is not changed by the premodifying a lot of portion: the verb agrees with either book or books as appropriate.
I therefore submit to you that your a battery of is no different from saying a bunch of, and so does not change the number of your subject.
He was rushed to the hospital immediately and a bunch of tests were conducted.
He was rushed to the hospital immediately and a battery of tests were conducted.
Now, it is also possible in English to use collectives as singulars if they are thought of as one unit instead of as their separate members taken individually. That’s why both the couple is and the couple are wind up being grammatical. The two versions merely reflect a different nuance as expressed by the speaker.
It is therefore also possible to use battery in a singular way when the entire group is considered as a singular entity, such as:
The first battery of tests done on Monday was inconclusive, but the second battery of tests from this morning leaves no room for doubt.
Just because you can use battery of tests sometimes in the singular and sometimes in the plural is no contradiction in English. It is how collectives work. This is the same principle that makes both of these grammatical:
- The council is united in their support for the measure.
- The council are divided in their support for the measure.
The perception of whether battery of tests should be thought of as singular or plural may depend on whether it is thought of as a single “definite” thing or just a bunch of them. In the example with the first battery, it feels more like one thing as a whole, but with a battery of, it feels more like a bunch of. However, even a bunch of can go both ways:
- A bunch of my friends were surprised.
- A bunch of grapes was all I ate that day.
Notice how both uses make perfect sense. It is even possible to swap them if you mean something else. The speaker’s intended meaning governs the number in these cases, not some mathematical law.
Postscript: Lies, Damned Lies, and N-Grams
Responding to a comment by JR, who presented a Google N-Gram that seems to show a growing use of the plural, I looked into several scenarios. Most of all these show that you cannot trust the raw n-gram results further than you can throw them.
As JR rightly observed, one must be extremely cautious in leaping to any conclusions based on Google N-grams, but it does appear that the premodifier use leading to plural verbs has been on the rise in the last fifty years, putting it nearly even with the other use:
Click on the graph to go directly to the n-gram.
That’s a ratio of around 7 : 6, but that does not mean one is “right” and one “wrong”. They can apply to different situations, ones where each is correct in that circumstance depending on the intent of the writer.
I again stress that you should not put too much trust in raw Google n-grams, let alone in that particular n-gram, for if you restrict the corpus to American English, the were case seems to disappear. Then again, it also disappears if you restrict it to British English! That calls into question the entire classification of published works that Google is using here.
Well, either that or else the Antipodeans talk way different than the rest of us. Me, I’m blaming Google, not our friends from the Southern Hemisphere. :)
Batteries of tests are not an especially common topic in general publication. However, they are a common topic in scholarly publications as indexed by Google Scholar. Here we find no shortage of citations where a plural verb follows battery of tests:
Not all results are as clear as this one from the British Journal of Medical Psychology, which you may be surprised to learn was published in Great Britain:
- A case can be made for the view that this factor is not one of overinclusion
and that the battery of tests do not provide a measure of overinclusive thinking.
Nor either are they all as clear as this example from the British Journal of Clinical Psychology, again by utter chance published in Great Britain:
- The same battery of tests were carried out with fifty-five people of 70 years and over living in their own dwellings, of comparable intelligence as measured by Raven's Matrices and of comparable social competence as measured by the Orpington Schedule after Doll (1953).
There are also plenty of American examples to be found, so one should not think this merely some peculiarity of British speech. Here is one from the Journal of the American Physical Therapy Association, which I hope nobody is surprised was published in America:
- A small battery of tests were sensitive enough to discriminate between people with PD who fall and those with no history of falls.
And here is another, this time from the American Heart Association, published in — need I really say it? — yes, in America:
- The proliferation of techniques has led to the development of non-invasive vascular laboratories (NIVL) in which a battery of tests are employed.
There are plenty more where those come from, but so what? Much more important is that when you look at those plural citations or those for the corresponding singular, you quickly notice something quite interesting: the way Google indexes things you cannot tell that battery of tests is actually the subject governing the verb that would appear to follow it!
Here are some of the situations of that which I encountered in examining the actual data:
- Battery of tests could be the last part of a compound subject, which of course demands a plural verb no matter how battery of tests is treated, and so tells you nothing about its own number.
- Battery of tests could be the object of a preceding verb and therefore completely unrelated to the verb that follows it.
- Battery of tests could be the object of a preceding preposition and also therefore completely unrelated to the verb that follows it.
- Indeed, in some of the Google results, battery of tests even concludes one sentence and the verb that follows is the first word of the next sentence that comes along. They aren’t even in the same sentence! That’s because punctuation doesn’t count in these searches.
For all those reasons and more, any blind interpretation of Google N-Grams results is futile, because those results do not show that the battery of tests mentioned is actually the subject of the verb that comes after it. Often it is not.
So the very chasing of a quick and easy popularity metric is misleading in the extreme. It won’t work. More serious analysis would be required.
However, none of that matters: you can play Google games till the cows come home and it changes nothing.
The bottom line is that premodifying phrases like a set/number/lot/bunch/battery of can sometimes be singular and sometimes be plural, depending not only on the noun that follows them but also more importantly on the perception and intent of the speaker. That means that as so often occurs in actual language, more than a single possibility is perfectly admissible.
In my own examples I have used the forms here that I would use in my own idiolect and in this particular time and place. I make no guarantees that this applies to others’ idiolects (although the curated scholarly results suggests it may), nor do I swear not to change things around myself under alternate circumstances, just as I did with the first battery of tests.
As has been so often demonstrated, collective nouns in English can go either way; it just depends.