Peter Pawlowski

76-451

1 May 2000

Quasi-Pronouns and Phrasal Verbs

In pedagogical grammars of English, the placement of the prepositional particle of a phrasal verb is often said to be dictated by the word category of the direct object: pronouns locate themselves between the verb and its preposition, while common nouns follow both the verb and the preposition. In actual usage, however, such a simple distribution is not evident; frequently, common nouns are placed before the prepositional particle. This phenomenon may be accounted for by positing a new category of nouns, that of the quasi-pronoun. This category would consist of words which were previously considered to be common nouns, but seem to share the behavioral properties of pronouns. Intuitively, the most likely candidates for this category seem to be those nouns which exhibit semantic similarities to pronouns. Such traits could include referentiality, concreteness or animateness. My research attempted to test the hypothesis that such a word category exists and is composed of words described by these traits.

Procedure

My research consisted of two different approaches to data selection and analysis. For my first attempt at testing my hypothesis, my goal was to distinguish possible quasi-pronouns from a set of collected data. To accomplish this, I selected from a large number of transitive phrasal verbs a small number that are well-represented in the corpus in both their pre-prepositional direct object and post-prepositional direct object forms. From this data, I attempted to draw conclusions about my hypothesis by comparing collocate lists and selecting for further examination what I found to be the only common lexical item. In a further analysis of these phrasal verbs, I compared the frequency of occurrence of a certain subset of nouns in the two possible direct object positions. I further posited this subset to be potentially similar to that of quasi-pronouns.

In a different approach to testing this hypothesis, I examined the t-scores of three nouns I considered to be strong potential candidates for the category of quasi-pronoun as they occurred in the object position of each of a large number of phrasal verbs. With this approach I attempted to verify the status of preselected nouns as quasi-pronouns, rather than identify members of this category from a relatively small set of data.

Initial Analysis

For my initial phase of research, I selected a list of 33 potential transitive phrasal verbs occurring sufficiently often in the corpus. I then reduced this list to nine by using the query lines <verb@+2NOUN+prep> and <verb@+prep+2NOUN> and examining the raw results. I then isolated only the transitive, phrasal verb examples that were found (see Sample Data for results of the verb write down). Table 1 lists these verbs and the number of instances in which they occur with the two possible locations of direct object: pre-prepositionally (abbreviated "pre-prep.") and post-prepositionally ("post-prep."). From this reduced list, I intended to identify those nouns occurring as a pre-prepositional direct object of at least two of the nine verbs. After comparing the collocation listings for all nine of the phrasal verbs, I found that only the noun things occurred more than once. I then compared the frequencywith which this noun occurred before the preposition to its frequency after the preposition, the results of which can be seen in Table 2.

pre-prep post-prep

give up 2 5.56% 1 0.33%

let down 0 0.00% 0 0.00%

make out 2 14.29% 0 0.00%

set down 0 0.00% 0 0.00%

stir up 12 80.00% 0 0.00%

think through 10 71.43% 0 0.00%

think up 0 0.00% 1 1.67%

throw out 0 0.00% 0 0.00%

write down 9 25.00% 8 5.67%

Table 2: Frequency of things

The general infrequency of the noun things prevented any strong conclusions from being drawn. Additionally, two of the six verbs (think through, stir up) may occur frequently with this noun pre-prepositionally because of established idioms ("think things through", "stir things up"), which could negatively influence their statistical relevance. Several of the verbs (think through, think up, make out) were also poorly represented in terms of quantity of data. With these factors taken into consideration, a pattern is evident in the data: the majority of instances of things is consistently pre-prepositional. On the basis of this data and analysis, it is difficult to firmly establish whether the noun things can accurately be classified as a quasi-pronoun, although it does provide us with at worst some preliminary evidence.

Further examining things, I looked at t-scores for the noun as it occurred with each of the verbs (Table 3).

pre-prep post-prep

give up 1.3 1.3

let down - -

make out 1.37 1.37

set down - -

stir up 3.44 <1

think through 3.14 <1

think up - <1

throw out - -

write down 2.94 2.6

Table 3: T-scores for things

In the case of think up, we must consider the fact that the t-score is less than one for the one occurrence of the noun in the post-prepositional position to be statistically insignificant, given the insufficient amount of data for that verb. If we consider the remaining verbs, there is no case in which the t-score of the noun in the post-prepositional position is higher than the t-score for the pre-prepositional position. T-scores for the noun occurring with the verbs give up and make out suggest that the strength of collocation with these verbs is about equal. The t-scores for the verbs stir up and think up, however, are less than one for the noun in the post-prepositional position, which suggests a less than expected collocation. This is compared to a very strong collocation (t-score greater than three) in the case of the pre-prepositional position. The verb write down once again exhibits more similar behavior between both positions, and what had seemed to be a significant difference when examining frequency becomes shrinks when examining t-score rather than frequency. Although the evidence provided by the verbs stir up and think up is strong, the similarities in t-scores of the remaining verbs seem to outweighed it in significance and once again complicate the possibility of pursuing judgment of our hypothesis.

In the face of this inconclusive evidence, I reformulated my hypothesis and broadened its scope. My modified hypothesis posits that the category of quasi-pronouns is composed largely (or perhaps entirely) of plural and mass nouns. To test this new hypothesis, I examined the frequency of occurrence of this type of noun in both possible positions relative to the preposition (Table 4).

pre-prep post-prep

give up 2 5.56% 61 20.13%

let down 67 35.83% 22 36.07%

make out 8 57.14% 10 8.40%

set down 0 0.00% 1 1.92%

stir up 14 93.33% 35 28.00%

think through 12 85.71% 8 29.63%

think up 0 0.00% 7 11.67%

throw out 11 22.00% 0 0.00%

write down 15 41.67% 22 15.60%

Table 4: Frequency of mass/plural nouns

By approaching the data in this manner, I hoped to verify that there exists a pattern in the characteristics of those non-pronominal nouns which occur before the verb’s preposition. The generality of the nominal attributes I chose detracts from the potential significance of the results of the analysis, but narrowing my analysis to a more specific set of noun attributes (for instance by requiring that the noun be referential or concrete) would have required a detailed, contextual examination of every instance of a potential noun.

After counting the occurrences of mass and plural nouns in both of the possible positions, I found that, for five of the verbs, an object occurring between the verb and preposition is significantly more likely to be either a mass or plural noun than an object occurring after the preposition. In two of the remaining four cases, the frequencies of the target noun types in both positions were comparable. Finally, two of the verbs (give up, think up) occurred with mass or plural direct object more frequently after the preposition. Given the previous assertion that these results cannot be considered conclusive, these results seem significant in that a large majority of the target noun types preceded the verbal preposition at least as frequently as they followed it. Here our conclusion is weak, but it does suggest the characterization of the quasi-pronoun word category to be similar to the subset of mass and plural nouns.

My initial approach to testing this hypothesis was one with a goal of discovery: positing a pattern of usage and identifying it by searching its environment. This method’s lack of success in either confirming or refuting my hypothesis could indicate that my hypothesis is only partially applicable to the data. I consider more likely, however, that this method was simply unsuited to this particular task. The unwieldy nature of transitive phrasal verbs, specifically their frequent occurrence in intransitive or phrasal prepositional form, high idiomaticity relative to other constructions, and confusion with similar prepositional verbs and simple verbs with prepositional complements, creates a great deal of noise and difficulty when the goal is to indirectly reveal a presupposed pattern.

Secondary Analysis

As an extension of my initial research, I approached the problem of testing my hypothesis from a different direction; namely, I intuitively identified quasi-pronoun candidates and tested their occurrence, rather than attempting to use the data to identify them. After preliminary testing with several phrasal verbs, I found that none of the three nouns occurred very frequently, least of all with the less common phrasal verbs. Because of this, I chose from the original list of transitive phrasal verbs the five which occurred most frequently in the corpus in an effort to improve the likelihood of finding relevant data. I chose three potential quasi-pronouns to test with these phrasal verbs: the noun things, which had already shown some potential from previous analysis, the noun people and the noun stuff. I chose these nouns because of their commonness, the fact that they are all plural or mass nouns, and their referential nature. To find occurrences of the phrasal verbs with these nouns, I used the queries <verb@+1noun+prep> and <verb@+prep+1noun>. Although the t-score for the results of these queries would have been the most useful, considering we do not have figures for the total number of instances in which the verb occurs in either form, the high frequency of the verbs themselves caused the noun to be chosen as the node word, making t-scores unavailable. We are left with only the raw frequency of occurrence to analyze, which in most cases proved to be quite small.

With only two exceptions (take off things, take over stuff), the results support my hypothesis (Table 5). For most of these five phrasal verbs, the frequency in the corpus is in the range of thousands, yet we surprisingly find only very few instances containing any of the three target nouns. The most frequent combination of verb and noun was turn people on, which occurred only four times out of the 801 total occurrences of turn on. If we do take these figures as significant, however, it seems the strongest quasi-pronoun candidate of the three is not things, but people. The noun people does not occur following the preposition of any of the five phrasal verbs, but does occur ten times overall preceding the preposition. This contrasts with things, which occurs half as many times following the preposition as preceding it. Despite the minimal number of examples, this is the strongest evidence encountered thus far for the classification of people as a quasi-pronoun. Classification of the noun stuff, however, was inconclusive. The noun occurred only four times total in both positions, a frequency too small to be useful, even when compared to the frequencies of the other two nouns.

things people stuff

pre-p post-p pre-p post-p pre-p post-p

take in 3 1 2 0 0 0

take off 1 3 0 0 1 0

take out 2 0 3 0 2 0

take over 2 0 1 0 0 1

turn on 0 0 4 0 0 0

Table 5: Frequency of quasi-pronoun candidates

Conclusions

Even after performing several different analyses of the two different data sets, it remains difficult to draw any general conclusions regarding my hypothesis. Taken together, the evidence presented seems to reflect the existence of quasi-pronouns, but not to prove it. We have evidence that isolated nouns, for example things and people, often behave like pronouns, and we also have evidence that a large subset of nouns, namely plural and mass nouns, may behave like pronouns more frequently than not. But as for our ultimate goal, demonstrating the existence of a hitherto uncategorized subset of nouns which generally behave like pronouns, we have at best confirmed vague intuitions and provided a basis upon which further, broader analysis may be conducted.

Table 1: Raw and actual frequencies of phrasal verbs

Raw frequency Frequency after reduction

post-prep. pre-prep. post-prep. % of raw pre-prep. % of raw

bring by 149 159 - - - -

clean up 704 25 - - - -

cut down 473 46 - - - -

do over 84 128 - - - -

do up 100 216 - - - -

fill out 163 5 - - - -

fill up 177 12 - - - -

find out 1835 76 - - - -

give up* 2191 146 303 13.8% 36 24.7%

knock over 81 23 - - - -

let down* 189 198 61 32.3% 187 94.4%

let off 135 50 - - - -

let through 2 27 - - - -

look up 658 62 - - - -

make out* 317 493 119 37.5% 14 2.8%

make up 685 22 - - - -

pass out 121 7 - - - -

pass up 65 10 - - - -

push over 21 18 - - - -

set down* 124 69 52 41.9% 56 81.2%

set off 628 42 - - - -

stir up* 183 16 125 68.3% 15 93.8%

take in 1241 5687 - - - -

take off 1182 1312 - - - -

take out 1345 1091 - - - -

take over 2793 394 - - - -

tell off 24 28 - - - -

think over 31 53 - - - -

think through* 55 39 27 49.1% 14 35.9%

think up* 92 38 60 65.2% 3 7.9%

throw out* 439 105 160 36.4% 50 47.6%

turn on 801 710 - - - -

write down* 243 36 141 58.0% 36 100.0%

* verbs selected for further analysis

Sample Data

<write@+down+2NOUN>

you and I can sit here in August write down 20 names, and one of those

for me. Before you read any further, write down 20 answers to the question `

was to write, so that's what I did. I wrote down a lot of deep and personal

to yourself and <tc text=pause> write down a couple of sentences <ZGY>

to a bunch of records and then writing down a thing it just reminds me of

power of good. <tc text=pause> I've written down a few things on here which I

my heart for my ministry minimum I wrote down a thousand dollars. And I want

envied what he saw as his ability to write down a finished piece first time

her headaches. She asked me to write down a practical plan for her <f>

the Publisher. She asked me to write down a practical plan for her to

would spend on Christmas. Try and write down a figure straight away. Once

real life," I repeated, as if I were writing down a criticism. `Would you care

because I was unemployable. I tried writing down a list of things to do each

to a nurse to read her lips and write down a note. `I have decided to end

you read the rest of this section, write down a description of a friend or

American astronomer Frank Drake. He wrote down a famous formula - the Drake

is that you can't agree. Sit down, write down a list of rules if you have

week or start the diary today and write down all the drinks you have this

Think about it for a few days. Write down all the things that make you

draw up a list of mutual friends. We wrote down all the names we could think

think it might do you good anyway to write down all your figures and <F02>

I don't see them again. And then I wrote down all their names because I got

was so long ago - but I remember writing down all the names, then, when I

told that `give an account" meant write down all the facts and you might

ones who were actually expected to write down all this guff and report it

bit more but in math it's just like writing down answers and you know <ZF1>

<write@+down+2NOUN> t-score

and i to NODE down the name

one if and NODE a thoughts

is useful i NODE your story

asked yourself could NODE all list

each word you NODE their things

here and then NODE her wishes

also have just NODE his suggestion

first own he NODE every memories

her right paper NODE some answers

if me actually NODE my options

so which never NODE 20 notes

this he ve NODE things practical

unemployab like NODE assertiven ideal

sorbet you can NODE equations names

pondered had she NODE eng letters

trembled is we NODE whichever date

passers by heartf NODE influences word

essences forgets abbreviati NODE jokes own

tennessee drake fanfare NODE recommende good

indoors writ sabrina NODE memories first

repeating notebook transmissi NODE answers deas

strengthen technicall youngster NODE careful ramblings

<write@+2NOUN+down>

up all the orders. Why didn't you write the orders down?" the chairman

doesn't deal in cash. You should write all those things down." The

pass a piece of paper around and write names down <ZGY> erm <ZGY> In fact

process of composing a sentence to write that answer down in will be

of it's on it. Erm you can sort of write comments down the side of it. And

in Russia <ZF1> a <ZF0> and let's write all these numbers down er add them

he'd see a sentence and be unable to write that sentence down. So he was

in history to talk about things and write things down you're it help you a

Let me <F0X> <ZGY> <F01> I'm going to write your names down as well 'cos erm

161 <ZZ0> Number eleven <ZGY> <ZGY> write a list down of all the animals that

these days <M01> Mm. Okay. I'll write that comment down actually <F01>

announcement to make, he had to <43> write all the names down on a sheet of

should you want to go anywhere, just write the address down and give it to

to the bed and sat down again. Write the number down," Dennison ordered,

Frank won't write anything. I have to write things down for him." When I asked

way we would like to hear about it. Write your story down and send it to:

fact I've now got over 900 songs. I write words down as a release from it

negligence". This normally includes writing your PIN down and leaving it where

Are you writing the <ZF0> are you writing the dingbats down? <F0X> <ZGY>

the person at the bus stop has been writing your answers down

<M01> I know. <F01> and they hate writing 'em down. <M01> I know. There's

difficulty because this stress on writing things down <M01> Mm. <M02> erm

<M01> Mm. <M02> But very often we're writing things down because <ZGY> round

MILLS: So what are the advantages of writing all this information down? And

novel begins with a warning against writing these stories down. Gerald

<write@+2NOUN+down> t-score

i they to NODE things down down

paper i d NODE all comment pillows

this advantages you NODE the address names

a sentence i NODE your number numbers

diaries larger shade NODE that dingbats words

900 anywhere hate NODE desks pin informatio

sofa songs includes NODE em answers things

eleven unable letters NODE comments sentence because

sat stress book NODE names orders could

hear warning actually NODE words stories as

stop normally again NODE these names you

sun prize against NODE two answer for

okay paper ll NODE a story is

often remember should NODE words the

didn let re NODE list

mean sort him NODE months

go around just NODE lot

m things been NODE those

your going she NODE these

about down by NODE this

can ve of NODE the

had only and NODE

be after he NODE

and about on NODE

that has it NODE