Predictability of Meaning in Written Speech: "Cloze task" and "Shannon’s Guessing Game" as an Illustration

TABLE OF CONTENTS

1. Introduction

2. Methodology and Execution
2.1. Cloze Task
2.2. Shannon’s Guessing Game

3. Results of the studies
3.1. Cloze Task
3.2. Shannon’s Guessing Game

4. Conclusion

5. Bibliography

1. Introduction

The term paper is to examine predictability in written speech based on two central experiments that underline the possibility of predicting answers and blank spaces. As examples arise Shannon’s Guessing Game implemented by Claude Shannon in 1951 and Cloze Task by Taylor implemented in 1953. The tests are to show in how far grammatical and phonological structures of one’s mother tongue as well as contextual information influence the choice of words in connected reading concerning phrases on one side and small texts as bigger units on the other side. As the subjects filled in the Cloze Task first, the thesis will stick to the same order as the questionnaire used for the survey. Hence, the term paper will give an overview on main features of the two methods, their origin and the way the experiments are carried out. In the further course, the approach this paper is based on, a questionnaire completed by native speakers of English, is explained and the data the analysis focuses on is outlined, examined and related to concepts of language perception and construction in the field of psycholinguistics. The term paper thus verifies how the choice of words can be traced back to conventionally learned rules and outlines the impact this has combined with knowledge about the context on the predictability of the lexical structure of a text and its meaning. Finally, the research paper is to give an overview of the phenomena present in the experiments it refers to and mentions which factors influence perception and the way phrases are built.

2. Methodology and execution

2.1 Cloze Task

The first test the thesis is based on is the Cloze Task . It was first introduced by Taylor in 1953 (cf. McLeod 1970, 117) and represents one of the very first approaches to explain predictability of lexical structures (cf. Lowder 2018, 1167). It consists in eliminating target words in a systematical way to find out in how far experimental groups are able to fill in the empty spaces and to make correct guesses (cf. Bickley 1970, 232). Normally, the Cloze Task is known for very limited vocabulary and describes a coherent situation in order to achieve a high level of predictability (cf. Lowder 2018, 1167). Hence, the study is to point out how many test subjects chose a particular word in a certain context (cf. Lowder 2018, 1167). Researchers have found out that the contextual information facilitates guessing the correct word as the participants activate words related to a certain topic before even perceiving it (cf. Lowder 2018, 1167). For the experiments executed for this term paper five native English speakers with American origin were chosen. The participants were given a digital questionnaire via email including a small coherent text with 91 words, the participants’ responses included, for the Cloze Task.

2.2 Shannon‘s Guessing Game

Shannon’s Guessing Game represents the second experiment. Implemented by Shannon in 1951, the test consists in giving a masked text to the experimental group, who has to guess the meaning of a sentence word after word (cf. Warren 2013, 40). The number of attempts regarding the guesses is established at the beginning of the test. If the participants do not get the right word before this threshold is surpassed, the solution of the word is revealed (cf. Warren 2013, 40). Thus, the guesses become easier as more contextual information is available (cf. Leech 1990, 57).

The three sentences provided for the Guessing Game were included in the Word document as well. As the participants filled in the document alone, the task was accompanied by instructions which asked the test persons not to scroll down to the next page before having typed in the guesses for the first space. Every guess was made on a singular page and the subjects were asked to type in three guesses for the respective space. The correct answer was provided on the next page of the file where the next three guesses had to be done. As examples arose the clauses: You are a very charming girl. , This is the only way to solve the conflict. and I am not a member of the group. For the first two phrases the first word you and this were revealed before guessing while group was the given word in phrase 3.

3. Results of the studies

3.1 Cloze Task

Yesterday, my _____ found ______ _______ full of sweets. As he has a ____ tooth, he ____ all without sharing ____ prey with us. But soon he _____ _____ a lesson: due ____ the sugar, his _____ began _____ _____ and he couldn’t ______ anymore.

______ _______ has happened before when we went to ________ last ________. He did not want _____ _____ us anything, although he had ________ it. We wanted to take ______ and hid his prey ______ ________ ________ and laughed when he ____ _____ find it.

The experimental group proposed nouns as the possessive pronoun my preceding the blank space introduces a noun phrase. As the space consists only in one word, the option of guessing an adjective before a noun for a more complex structure of the noun phrase could not be applied. The following verb found can be understood as an attribute implying that the word searched for must be a someone, a person or a creature being able to find something. Due to the personal pronoun he in the second and third phrase, which reveals the gender of the correct word, the participants suggested terms that name male beings. When filling in the second and third gap, the test persons expected a noun phrase of a more complex type as they could perceive more than one blank space. As a determining element is necessary before using an adjective for further descriptions, the subjects decided to propose mainly articles. The research results have shown that the undetermined article a seemed to be more plausible for a description as the participants expected the narrator to speak about something unknown for the protagonist. Nonetheless, in one case the proposal did not consist in an article but a possessive pronoun: my. It is worth noticing that this possessive was used in the part of the text read before. A reason to explain the choice of this particular element is the pre-activation of elements that are expected based on the context the blank space is found in (cf. Lowder 2018, 1167).

Concerning the following gap, adjectives were inserted as the combination of a + _____ + noun, in this case tooth , implied a complex construction of a noun phrase. Apart from the fact that to have a sweet tooth is a determined expression in English to describe a person preferring rather sweet, sugared than salty meals (cf. Merriam Webster) the choice of sweet by some of the test persons is influenced by the noun sweets in the first sentence. When reading the word sweets, the brain activated words from familiar fields in order to have easier and faster access to them and thus used the contextual information about the sentence structure and vocabulary the lemma is surrounded by (cf. Lowder 2018, 1179). Another aspect worth to be mentioned is the verb phrase in the third clause. The majority of the participants proposed a construction out of auxiliary + main verb. The verbs chosen can be considered synonyms and there was not a high difference between the proposed answers. Nonetheless, most of the subjects decided for would learn as it could be considered a rather frequent verb in comparison with figured out . While the conventionally learned expression due to something did not show any differences between the answers given by the subjects, the nouns proposed to follow his in the same phrase were limited to body parts but showed different suggestions. Two participants proposed teeth as the hurting body part of the protagonist in the text. The plural of tooth mentioned in the second phrase of the text might have influenced the participants’ perception. Apart from that, teeth could have been activated when tooth was read, and such a pre-activation of stimuli in the context of a word facilitates the access of the brain to certain topics and isotopes (cf. Lowder 2018, 1167). In the following gap, the majority of the participants guessed that the protagonist was not able to eat anymore, only one case suggested chew, which seems to be a rather uncommon in comparison with eat . The fact that the subject was also the one who guessed teeth instead of stomach explains the choice.

The new paragraph was mainly introduced by a noun phrase, which means that most of the test persons proposed rather a determined article or undetermined one followed by a noun. The choice of the article depended on the noun proposed and its relation to the situation and context. While one proposed the hunting, the suggestion of another test subject an ache could be related to the verb hurt used in the preceding paragraph. This indicates the pre-activation of terms relating to the isotope of pain. Another case chose a connective and a demonstrative pronoun but this. The subject had proposed an element substituting a noun and tried to make the text more complex by using a connective. This might have been implied by the beginning of a new paragraph. The remaining gaps of the phrase show another phenomenon. The experimental group guessed either an infinitive as a possible element following the preposition to. Two participants instead proposed a locative complement by putting places such as Malta or California . As all of the participants chose a temporal adverbial for the last gap, a syntactical construction of subject + verb + object + place + time is created. Nevertheless, the fact that the majority of the group chose an infinitive shows the frequent combination with the preposition to. For the next gap the participants showed a high level of predictability because the guesses consisted of verb forms only as this is the expected concept. High frequent verbs like to have and to do dominated rare forms like to get (cf. Ringbom 1998, 193). After the reveal of solve, the experimental group expected a noun phrase and therefore proposed a form of determiner which could be an article like the and a but as well determining elements like that or this. For the noun at the end of the phrase, words from the isotope of ‘ situation that can be solved’ were proposed. Apart from choices like dilemma, mystery and puzzle, which differed from participant to participant, issue and problem appeared in every survey, which implies a high frequency of these terms, as well in relation to the term conflict proposed by the researcher as it did not appear in any of the responses.

[...]