Jessica Boynton

Disclaimer: This is the work of a third year undergraduate student in linguistics so... interpret it as such, please.

Theta-role Assignment for Transitive Participial Adjectives

Table of Contents

1 Introduction

This research project has various aims. The first is explicit – to study the existence of a certain construction in a child’s grammar in relation to the same construction in an adult’s grammar. In order to examine this aim’s scope, the construction itself must first be understood (Section 2), as well as the hypothesis concerning it (Section 3).

The second aim is to educate the researcher as far as research methodology is concerned. In order to examine this aim’s scope, options in research methodology must be understood (Section 4) and the pilot study undertaken must be surveyed objectively (Sections 5 & 6).

2 Understanding the Construction

A term-by-term breakdown of the construction at hand goes a long way in explaining its structure.

Theta-roles reveal the semantic position of arguments of a verb (or certain kinds of adjectives). They tell how exactly how the participants of an action participate. Many theta-roles exist, but the two that are most central to this study are ‘agent’ and ‘patient’ roles. The doer of an action, whether the action is in verbal or adjectival form, is the agent. The doee of an action is the patient.

As stated earlier, other theta-roles exist to communicate slightly different meanings; the causer theta role is comparable to agent, only the agent volitionally does the action while the causer does not. An experiencer is, at the right level of abstraction, the patient of a verb/adjective with an experiencer. For the purposes of this paper, causers will be treated as agents and experiencers will be treated as patients.

Transitivity communicates the number of participants (AKA arguments) of an action word. In formations using intransitive verbs/adjectives, there is only one argument. This typically makes the theta-role assignment quite straightforward. For example, in the expression ‘the dog sleeps’ the dog is the only argument and, as the doer of the action ‘sleep’, receives the theta role ‘agent’. Formations using transitive verbs/adjectives have two arguments – the doer and the doee. Therefore, in the expression, ‘the man eats the apple’, the man (the doer) is the agent and the apple (the doee) is the patient.

There are also verbs that take three arguments, called ditransitive verbs, as in the expression ‘give the gift to Mary’ but these are outside the scope of the project. This project focuses on transitive formations.

Participial adjectives are those special adjectives alluded to earlier; they derive from verbs and subsequently carry the verb’s argument structure. There are two types of participial adjectives in the English language: those that derive from passive/past participles, such as ‘eaten’ in either ‘the apple was eaten’ or ‘I have eaten’, and those that derive from present participles, such as ‘eating’ in ‘the man is eating’. As participial adjectives, these participles serve in the constructions ‘the eaten apple’ and ‘the eating man’ respectively.

The object of direct study is theta-role assignment in transitive participial adjectives. Stated simply, in formations like ‘the eaten apple’, is the noun the adjective modifies (in this case ‘apple’) the agent or the patient of the action (in this case ‘eat’)?

The fascinating phenomenon revealed by an analysis of passive/past participial adjectives and present participial adjectives is the asymmetric theta-role assignment. In the former, found in sentences like ‘the eaten apple’, the modified noun is the patient; in the latter, found in sentences like ‘the eating man’, the modified noun is the agent. This point is driven home when one tries, somewhat unwisely, to picture an eaten man or an eating apple.

There seem to be two constraints at work in these constructions, although further research may conflate them in some way. They prohibit agentive role assignment in passive/past participial adjectives, so ‘the eaten apple’ can’t mean that the apple is doing the eating, and prohibit patient role assignment is present participial adjectives, so ‘the eating man’ can’t mean that the man is receiving the eating (being eaten). In the interest of experimental control, this study focused on the former construction – the theta role assignment of transitive passive/past participial adjectives.

3 The Research Hypothesis

In order to design an experiment concerning the construction, one must first outline a research hypothesis. The constraint explained earlier serves as a starting point; a passive/past participial adjective cannot assign the theta role agent to the noun it modifies (the eaten apple cannot mean that the apple is doing the eating). To adult speakers this seems painfully obvious, to the extent that serious study of the formation may appear ridiculous. Nothing could be farther from the truth. The mere fact that the meaning of the sentence is painfully obvious reveals just how tenacious this constraint is. Arguably, that adult speakers can’t conceive of the alternative reading of the sentence ‘the eating apple’ (which would, in fact, translate to ‘the eaten apple’) means that that interpretation is absolutely ungrammatical in the adult grammar.

This research project asks whether children have the same constraint. While it is entirely possible that children interpret the sentence ‘the eaten apple’ the same way adults do, it is also entirely possible that the formation either means nothing to them or is ambiguous as far as theta role assignment is concerned. The former alternative to the adult grammar would mean that children hear ‘the eaten apple’ as verbal noise. The latter would mean that children hear ‘the eaten apple’ and can’t tell whether it means that the apple is being eaten or that the apple is eating. This project focuses on whether the construction is ambiguous for children.

The experimental hypothesis proposes that children have the same constraint that adults do; the null hypothesis proposes that they do not. The role of the experimental and null hypothesis will be further discussed in Section 4.

Now that the formation in question and the asserted hypothesis are understood, the experimental design can be outlined. First, the design type used will be explained (Section 4); then the design of this particular experiment will be detailed (Section 5).

4 The Truth Value Judgment Task

The truth value judgment task is a research design particularly useful in determining the ambiguity of a construction. In the task, the subject is presented with a scenario (in the form of a narrative, a dialogue, a play…) and is prompted to determine the accuracy of a carefully designed sentence describing some element of the scenario. This sentence, called the target sentence, uses the construction in question. When experimental design is sound, the subject’s response to the sentence can reveal the whether the construction is ambiguous or not.

The (potential) ambiguity is revealed in the subject’s response to a statement that is factually false on the grammatical interpretation of a construction, but factually true on the ungrammatical interpretation. For instance, in testing theta role assignment in transitive verbs, a researcher might ask the subject to judge the validity of the statement “the cat chased the dog” when, in the story, the dog, in fact, chased the cat. If the subject judges the statement as true, the theta role assignment may be ambiguous, or even opposite to the adult assignment. If, however, the subject judges the statement as false, the subject likely follows the same assignment rules as an adult.

The experimental design, however, must account for more factors than simple targets and simple responses. Uncontrolled elements in an experiment result in confounding data and faulty statistics. In order for an experiment to be successful (success meaning that it results in usable data, not necessarily that the data supports any particular hypothesis), certain felicity conditions must be met.

There must be two hypotheses considered in the experiment – the experimental hypothesis and the null hypothesis. The experimental hypothesis is the hypothesis to be tested; the null hypothesis assumes the experimental hypothesis is false. (See Section 3 for an outline of these hypotheses in this particular experiment) In order to avoid unwarranted support for the experimental hypothesis, the experiment must be designed so that any potential confounding factors bias the subject towards the null hypothesis.

For example, previous research indicates that subjects prefer giving affirmative responses; that is, they would rather answer yes than answer no. Therefore, in order to bias subjects toward the null hypothesis, target sentences in a truth value judgment task must assign the grammatical interpretation to a factually false. Then, if a subject answers ‘true,’ as would be expected with the bias toward affirmative responses, the null hypothesis would be supported. If a subject answers ‘false,’ the ungrammaticality of the sentence has overridden the bias toward affirmative responses, and the experimental hypothesis is strongly supported. Thus, while the response that supports the null hypothesis may also be due to this paralinguistic factor, the response that supports the experimental hypothesis is not.

Another potentially confounding factor is readiness of information – the fact that, all else being equal, the most recent utterance is most likely to coincide with the interpretation of a sentence. If the last sentence uttered before the target sentence contains a construction which grammatical interpretation matches that of the target sentence, the subject may interpret the target sentence correctly not because of their own grammar, but because that interpretation was recent in their mind and therefore more readily available. In order to bias subjects toward the null hypothesis, then, the last sentence before the target sentence should contain a construction which grammatical interpretation does not match that of the target sentence, but instead coincides with the that in the null hypothesis. In this way, an answer that supports the null hypothesis may also be due to this confounding factor; an answer that supports the experimental hypothesis is not.

Other factors must also be taken into consideration. The subjects should be familiar with the experimenter (and vice versa) to ease the experimental situation. In addition, the study should be entertaining and should engage the subjects’ interest. Furthermore, the pretense of the experiment should make sense; if, for instance, the subject is expected to answer a correct someone else’s comprehension, there must be reason to believe that the entity’s comprehension is in question and that the subject is the only one able to provide corrections.

In sum, in order for the Truth Value Judgment Task to be effective, the grammatical interpretation of the target sentence must be factually false, the semantic situation underlying the ungrammatical interpretation must be communicated just before the target sentence, the pretenses behind the experiment must make sense, and the study must be comfortable, with interesting material and familiarity between the researcher and the subjects.

Now that the Truth Value Judgment Task is understood, the design of the pilot study will be surveyed.

5 Pilot Study Design

Since the project aims to discover whether the construction in question is ambiguous for children, the Truth Value Judgment Task was the natural choice of research design. Remember that, in this breed of research design, subjects are presented with a situation (in the case of this particular project in the form of a story) then with a target sentence that contains the construction under study. The proposition of the construction must be factually false under the grammatical interpretation and factually true under the ungrammatical interpretation, to accommodate the extra-linguistic preference for affirmative answers.

The pilot study incorporated six short stories, each with a target sentence that contained a transitive passive/past participial adjective. One story, for instance, was about two little boys who play in the mud. They get dirty, and walk into the house and get the floor dirty too. Their mother demands that they wash the floor, and they decide that whoever enjoys mopping the most should wash the floor. After some deliberation about which prefers mopping, they decide that a particular boy enjoys is more. So, one boy washes the floor while the mother washes dirt off the other boy. The target sentence is: The washed boy likes to mop the most.

The target sentence is carefully constructed to be factually false under the grammatical interpretation. In this way, the answer ‘yes’ means that the subject either interpreted the sentence ungrammatically, or resorted to the preferred affirmative response. The answer ‘no’ means that the subject interpreted the sentence grammatically. However, more factors must be accounted for for this study to be effective.

Remember that, because of the extra-linguistic factor of availability of information, the most recently heard information may influence the interpretation of a construction. Therefore, the situation underlying the ungrammatical interpretation of the target sentence should be communicated in the last sentence before the target sentence. This was achieved in the pilot study by, in the given story for example, mentioning the washing boy last, when the target sentence addressed the washed boy. In this way, any bias due to the availability of information would result in erroneous support of the null hypothesis, not the experimental.

Another potential effect of availability of information and short term memory is that subjects may honestly not remember what happened in the story; therefore, either response would be somewhat suspect. This effect can be accounted for at some level by repeating the main points of the story just before the target sentence – a tactic used in the pilot study. It is comforting to note, however, that if the subjects were to forget the happenings of the story, the biases caused by preference for affirmative answers and recent information mean that confusion will result in affirmative answers based on the most recent information which, because of the experimental design, result in support of the null hypothesis.

The experimental situation must also make sense to the subjects – because the pilot study required the subjects to judge comprehension, a premise had to be set that made judgment acceptable and that made the subject the logical judge. The study, therefore, incorporated an owl puppet that was trying to learn to talk with people and was trying to understand the stories. The owl wanted to make sure he was getting it right, but was afraid to let grown-ups know if he was wrong. Therefore, the subjects, all children, had a right to correct the owl and were the logical candidates for the job. In the course of the procedure, then, the researcher told the story then prompted the owl to interpret it. The owl summarized the main points story, then said, “I know one thing about that story – (target sentence).”

Other considerations to be made were relatively specific to this study. Because the target sentences all incorporated passive/past participial adjectives and juxtaposed them against their present participial counterparts, and because the experiment was spoken, care had to be taken that the participial adjectives weren’t too similar in sound. For instance, the eating/eaten examples given previously were not possible because, in natural speech, the two sound almost identical. It would have been impossible to tell which form the subjects had in mind when interpreting the target sentence. Therefore, the forms juxtaposed for the pilot study’s six stories were: washing/washed, catching/caught, fearing/feared, hearing/heard, throwing/thrown and biting/bitten.

Also, because both interpretations of the theta role assignment had to be logical to meet felicity conditions, verbs had to be chosen where both participants could be animate. The verbs listed in the previous paragraph met this criterion also.

Another particular consideration was due to the aspect of the participial forms. The passive/past participial adjectives are perfective in nature, so the action they describe must be completed by the end of the story. However, the present participial adjectives are progressive in nature, so the action they describe must be continuing on at the end of the story. Moreover, because of the nature of the study, the two forms had to be derived from the same verb. Therefore, only verbs that have the potential to be both perfective and progressive could be considered, and the stories had to be crafted in a way that accounted for the aspect of the forms.

In order to more clearly illustrate the pilot study, and the techniques used to meet felicity conditions, story one’s protocol will be detailed. The remainder of the pilot study’s protocol is attached as an appendix.

* The study began with a brief explanation of the owl’s position (that it was trying to learn to understand English and needed corrections from a child because it was afraid to be wrong to grown-ups) and the story began:

Researcher narrative:

Two little boys were playing outside on a bright summer’s day when they came across a mud puddle. They knew they weren’t supposed to get dirty, but they couldn’t resist so they splashed and stomped in it and made a big mess. The first one got really covered in mud, but the second one was wearing coveralls so he didn’t get himself muddy at all. When they walked into their house, they got mud all over the floor too. When their mom walked in and saw it, she was mad and told them that they had to wash the floor. The boys decided that whichever one liked to mop most should wash the floor.

The first little boy said that he liked mopping a little, cause you could slosh water all around and make a mess. The second little boy said he liked mopping even more, because he liked to skate around on the wet floor. They agreed that the second boy liked to mop more. So, while their mother helped wash the dirt off the first little boy, the second little boy washed the floor.

* At this point in the protocol, the owl was prompted to tell what it thought happened in the story:

Puppet response:

That was a neat story. There were two little boys who got all muddy, then their mom made them clean up. The mom washed one little boy and the other little boy washed the mud off the floor because he likes to mop most. I know one thing about that story –

The washed boy likes to mop most.

* The subject was then prompted to judge the accuracy of the puppet’s response, then was asked why that judgment was given. Then the subject was asked whether (s)he wanted to hear another story; if yes, the protocol was repeated for another story (except for the explanation of the owl’s position, which only had to be done once).

This story presents two little boys and creates a situation where one is the agent of the verb ‘wash’ and one is the patient. Then, to accommodate constraints on short-term memory, the puppet responds with a summary of the story, ending the summary with a reference to the agentive boy to ensure that the ungrammatical interpretation is freshest in the subject’s mind. Then the puppet produces the target sentence – “the washed boy likes to mop most,” which is factually false under the grammatical interpretation, to accommodate the bias toward affirmative responses. The design, therefore, adequately biases the subjects toward the null hypothesis.

6 Pilot Study Results

The pilot study was conducted in conjunction with another pilot study (on passive constructions) and incorporated three subjects. S1 was six years old, S2 was five years old and S3 was three years old.

The first run-through was somewhat lacking – S1 was alert and responsive; the researcher, however, failed to ask follow-up questions. Also, the researcher intended to use cut out dolls to act out the stories while they were being told, but the dolls were unable to stand up on their own on the carpet (which constituted the research area). Therefore, while the responses are of interest, they are not usable as data.

S1 played audience to stories One, Two and Three (see appendix) and judged the ‘washing/washed’ target sentence as true and the other two (‘biting/bitten’ and ‘catching/caught’) as false. However, as explained before, justifications were neither requested nor given.

Luckily for the researcher, S2 needed no prompting to supply justifications for judgments. S2 listened to and judged two target sentences: the ‘fearing/feared’ sentence from story Four and the ‘hearing/heard’ sentence from story Five. Both sentences were judged to be true. S2 justified each judgment with a factually accurate reason, explaining why the agentive participant fit the description in the target sentence. It would seem that S2 may find the construction ambiguous.

S3 also listened to and judged two target sentences. With this subject, however, the researcher encountered an unexpected complication. The subject requested stories with particular characters, so the researcher found herself scrambling to create a story that would incorporate the desired characters but still culminate in one of the preplanned target sentences. Furthermore, after hearing the first story, which involved the characters getting muddy, all stories thereafter had to incorporate mud as well. So, in actuality, far more than two stories were told; only two of them lent themselves to appropriate target sentences.

In effect, S3 played audience to stories One and Three (‘washing/washed’ and ‘catching/caught’). S3 judged both targets as true. For story One, S3 justified the judgment by saying “because that’s what happened.” The researcher probably could have pushed for a more illuminating answer, but she was still reeling from the unexpected ad-libbed nature of this interaction. For story Three, S3 wavered between answering true or false, but clearly identified the ‘patient’ participant as the one described in the target sentence (in accord with a grammatical interpretation of the target sentence) before conceding that the owl was right. Then S3 clarified that it was really the researcher who was right. It would seem that S3’s responses were heavily influenced by the bias toward affirmative responses.

7 Conclusion

The results of the study were somewhat inconclusive; responses were too varied to conduce a firm analysis. This is partly due to the low number of participants (3) and partly due to flawed research design.

The low number of participants, and of target sentences that were judged by the participants, resulted in a small data set from which to draw generalizations. Future incarnations of the study will involve more subjects and will take measures to ensure that more target sentences will be judged by them; these measures will be described later.

The research design and execution was flawed in various ways. The primary problems were: lack of familiarity between the researcher and the subjects, confusing and potentially uninteresting presentation of stories, lack of cohesion in presentation, and lack of consistency in follow-up questions.

As stated in Section 4, the researcher and subjects should be somewhat familiar with each other to engender a comfortable environment for the study. However, since the subjects were being addressed in their own environment (the house of two of the children, frequently visited by the third child), this criterion was overlooked. And, indeed, the children seemed at ease with the experimental process. The researcher, however, was not. Unfamiliar with the personalities of the children, the researcher was visibly uneasy with the interaction stipulated by the research design. It is likely that previous interaction with the children would have improved the researcher’s performance.

The stories were somewhat confusing and lacking in intrigue. This was somewhat due to the content of the stories themselves, which by necessity was information-rich, but more due to the lack of action. Not only were the stories in narrative form, but also they were not acted out in any manner. Ideally, stuffed animals would have been used to act out the stories and, during the puppet’s response, would have been positioned in a way that reflects the happenings of the story. The use of figures was attempted, in the form of cutout dolls, but they wouldn’t stand on their own on the carpet and were therefore more cumbersome than helpful. In future incarnations of the study, stuffed animals will absolutely be used.

The stories also lacked cohesion – they were obligatorily broken up with target sentences, but they were also separate stories incorporating different characters. That gave the study a more drawn-out, less interesting feel, which resulted in the judgment of less target sentences. Future incarnations of the study will utilize the same characters for all the stories in an attempt to achieve cohesion.

There was also a lack of consistency in follow-up questions. Ideally, the follow-up questions should have been written into the protocol and repeated verbatim. Instead, the researcher counted on having her wits about her during the experimental process. As a result, she failed to prompt S1 to explain why truth value judgments were made the way they were, and failed to prompt S3 in a consistent manner (S2 provided the information without being prompted). Future incarnations of the study will incorporate consistent follow-up protocol.

In sum, the pilot study served little to reveal the grammatical constraints in childrens’ grammar, but served quite well to teach experimental design. Incorporating the modifications outlined above, the researcher will make another attempt at studying theta-role assignment of transitive participial adjectives.

Appendix

Story one:

Researcher narrative:

Two little boys were playing outside on a bright summer’s day when they came across a mud puddle. They knew they weren’t supposed to get dirty, but they couldn’t resist so they splashed and stomped in it and made a big mess. The first one got really covered in mud, but the second one was wearing coveralls so he didn’t get himself muddy at all. When they walked into their house, they got mud all over the floor too. When their mom walked in and saw it she was mad and told them that they had to wash the floor. The boys decided that whichever one liked to mop most should wash the floor.

The first little boy said that he liked mopping a little, cause you could slosh water all around and make a mess. The second little boy said he liked mopping even more, because he liked to skate around on the wet floor. They agreed that the second boy liked to mop more. So, while their mother helped wash the dirt off the first little boy, the second little boy washed the floor.

Puppet response:

That was a neat story. There were two little boys who got all muddy, then their mom made them clean up. The mom washed one little boy and the other little boy washed the mud off the floor because he likes to mop most. I know one thing about that story –

The washed boy likes to mop most.

Story two:

Researcher narrative:

Two kittens love to play outside. The first kitten goes outside a lot, but she also likes to stay inside and play with her toys. The second kitten goes outside all the time and stays inside a lot less than the other kitten. One day, the kitten who goes outside more got fleas, and they kept biting her and biting her. She complained and complained but couldn’t do anything about it. The other kitten, who stays inside more, just shook her head and went back to playing with and biting her toy mouse.

Puppet response:

That was a neat story. There were two kittens and one goes outside more but she got fleas. The other one stays inside and plays with her toys. I know one thing about this story –

The bitten kitten stays inside more.

Story three:

Researcher narrative:

Two monkeys worked for the circus, and for their big trick they would walk on a really thin wire way way up high. One night, while they were on the wire, the crowd was cheering and cheering and they wanted to know who was the crowd’s favorite. So they decided to each take a bow in turn and see who got the biggest applause.

The first one bowed and people clapped really loud, but then he fell off the wire. He fell far far down until the safety net caught him. Then the other monkey bowed and the crowd clapped even louder, so he was the favorite because he got the biggest applause. The crowd loved him so much that they even threw flowers at him and he caught them and kept bowing.

Puppet response:

That was a neat story. There were two monkeys on a high wire and one fell off. The other one got flowers. I know one thing about that story –

The caught monkey was the crowd’s favorite.

Story four:

Researcher narrative:

There were two monsters who were trying to scare their friends. The first monster decided he would sneak up behind his friends and roar right behind them to see how high they would jump. But when he did, his friends turned around and growled right back. He was so scared of them that he ran away and tripped and fell. The other monster laughed at him and decided to show him how it was done. So he hid behind a tree and waited for his friends to pass. When they did, he jumped out and roared really loud and they all ran away in fear.

Puppet response:

That was a neat story. There were two monsters and they tried to scare their friends, but one was scary and the other got scared. I know one thing about that story –

The feared monster tripped and fell.

Story five:

Researcher narrative:

Two girls loved to dance. The first one liked to clog-dance and the second one liked to dance ballet. One day, they were dancing together and the ballet dancer complained to the other one about how noisy her clog-dancing was. The clog-dancer didn’t respond, so the ballet dancer walked up to her and tapped her on the shoulder, telling her how loud she was. The clog dancer shook her head and motioned that she couldn’t hear and kept dancing. The ballet dancer covered her ears and kept dancing, but could still hear the clog dancer.

Puppet response:

That was a neat story. There were two girls and one clog-danced really loud but couldn’t hear anything. The other one danced ballet quietly and couldn’t stop hearing the loud girl. I know one thing about that story –

The heard girl danced ballet.

Story six:

Researcher narrative:

Two men wanted to see who was better at riding horses. They decided that whoever stayed on the horse for the longest was the better rider. So the two men two turns riding a horse to see who could stay on longer. The first man got on the horse and started riding, and he rode for a really long time and everybody thought that he must be the best. But then the horse threw him off and he landed kerplop in the dirt.

The second man got on the horse and started riding, and he rode for even longer than the first man. In fact, the horse never threw him off. Everybody agreed that he was the better rider. To show off how good he was at riding horses, the better rider started throwing tennis balls into a basket at the same time that he was riding the horse!

Puppet response:

That was a neat story. There were two men and they had a horse-riding contest, and they were both good riders but one could even throw tennis balls into a basket while he was riding. I know one thing about that story –

The thrown man is the better rider.