An Experiment in Comprehender`s Use of Information in SentencePrediction
An Experiment in Comprehender`s Use of Information in SentencePrediction
Sentence processing occurs in readers and listeners whenever theyencounter a word phrase or an utterance. Sentence processing is theact of taking in a sentence and understanding its meaning. The mindlike any other strategic device requires having a vast number ofpredictions to work effectively. Sentence understanding is a complexprocess faced with some challenges such as semantic, Lexical andstructural ambiguities. Ambiguity can be of two kinds localambiguity and global ambiguity. The former meaning an ambiguity thatpersists for a short period as it is quickly solved this kind ofambiguity occurs in sentences with only one meaning. Globalambiguity, on the other hand, lasts even after the full sentence iscompleted. A sentence becomes globally ambiguous if it has more thanone meaning an example of such a sentence is `The soldier shot thecriminal with a rifle.` The sentence could mean that the soldier hada rifle which he used to shoot the criminal or the criminal who wasshot by the soldier had a rifle.
This experiment`s main objective is to observe and analyze the use ofcontextual information by comprehenders in making predictions andanticipations of a sentence`s meaning in real time. The study isimportant as it will provide readers with new information on howsentence comprehension works and the role of prediction andanticipation in grasping the meaning of a sentence.
Role of Prediction in Comprehension
An individual is more likely to catch a stone flying in his or herdirection if the person can anticipate the trajectory of the stone.Similarly, the brain can better comprehend a sentence if it canpredict or anticipate the next word in a sentence whether in spokenor written language. It has been argued by multiple credibleresearchers that the brain utilizes prediction in many otherfunctions such as body movements and visual perception. It has beenobserved that sentences that have a predictable word sequence areeasier to understand than those with unpredictable meanings. Anexample of a sentence with a predictable word sequence is, thepoliceman caught the thief. A policeman is likely to catch a criminalor thief. An unpredictable sentence is such as the thief caught apoliceman. In experience, a thief is not likely to catch a policemanhence making the sentence less predictable. It was also observed thatlisteners and readers alike made more eye movements in sentences thatwere predictable and more so in sentences that contain food in them.The observations show that an individual is capable of making a vastnumber of predictions on the fly to comprehend a sentence faster andwith more ease, this is especially true for a native speaker of alanguage. The context of a sentence hence becomes a significantfactor in sentence processing and prediction. Different contextualinformation is however predicted differently, and some contextualinformation is predicted faster than others. For the sake of thisexperiment different sources of context will be examined and analyzedso as to make conclusions on how contextual information is processed.This will also assist in understanding the impact of context inunderstanding and prediction of a sentence (Hörmann, 2013).
Aim of the Experiment
In this experiment, we will be analyzing the comprehender`s use of aword`s meaning and its organizational position of the previous wordin association with his or her experience information to make lexicalsemantic predictions. We will also be investigating if the meaning ofthe previous word being a verb has any effect on the speed andrelevance of a prediction compared to the previous word being a noun.Common verbs such as `evict` already portray the image of a landlordand tenant, the relationship between a landlord and tenant makes theprediction further. It is more likely than not that an individualwill predict the sentence as something close to `the landlord evictedthe tenant.` The context of the verb evicts make it next toimpossible for a reader to predict the sentence as the tenantevicting the landlord.
The study will next analyze previous works by other authors toprovide context for the experiment.
The earlier study about the different perceptions about language fromvarious individuals was based on a tool known as Event RelatedPotentials (Nieuwland, Martin, and Carreiras, 2013). However, theprimary area of study in this case, is the N400 which simplifies thenegative ERP and it ranges between 250ms and 400ms. The N400 responseis caused by any word in the sentence which may include either noun,verbs or adjectives. This reaction has been closely associated withthe memory of a human being during the conversation. Even thoughresearch in the past has shown an inverse relation between the N400response and the interpretation of a word, the probability that aspeaker uses any given word in any context is operationalized(Nieuwland, Martin, and Carreiras, 2013). This entails the severaltrials before the speaker becomes fluent in making complete fragmentsof sentences that make meaning. Additionally, many factors can affectthe probability that a speaker will prefer an individual clause ascompared to others that already exist. Some of these factors arenegation, the knowledge of the world, sentence structure, andmessage-level representation (Small, Cottrell &Tanenhaus, 2013).
Conversely, recent studies have suggested that not all informationwithin the structure of a sentence may affect the comprehender`sforecast instantly. For instance, in the study conducted by, Kukonaand colleagues, eye-tracking that reported the eye movement data(Riordan, Dye, & Jones, 2015). This finding claimed thatlisteners cannot use information from previous disagreements topredict upcoming arguments (Nieuwland, Martin, and Carreiras, 2013).This situation was right under circumstances where there exist twosentences one active and another passive though they convey the samemessage. If the listeners use the information from the activesentence to predict the meaning of the next argument which ispresented in the passive voice, then, the participants in the actualsentences are likely to become the agents in the passive sentence(Riordan, Dye, & Jones, 2015). Besides this, the passivesentences provide differences in the way words are spelled and thetime between the pronunciation of a verb and the targeted noun ascompared to active sentences (Nieuwland, Martin, and Carreiras,2013). These effects have been known to delay the impact of theprediction offered by the interpreter of the information.
Although the information displayed in the argument seems to delay theforecast by the comprehender, this information is essential as itaids in interpreting the themes present about the verb utilized.Additionally, more practice by the speaker produces a largerpositivity as the verb which is realized late and is referred to asthe P600 effect (Nieuwland, Martin, and Carreiras, 2013). Even thoughthe P600 effect is still the subject of discussions, the rehearsalsdone by the speakers before the actual argument produces the P600effect thus indicating that comprehenders process the necessaryinformation about the evidence in interpreting the sentences. Thisobservation has been found to coincide with the premise that itrequires some moments for the prior knowledge about an argument toaffect the listener`s or reader`s prediction about an upcoming verb.The hypothesis was put on tests when the researchers utilized theObject-Verb structure in the Mandarin Chinese language (Nieuwland,Martin, and Carreiras, 2013). Structural roles of the parts of speechwere manipulated by reversing the original order that existed before.This manipulation in either way resulted in tripling the timeinterval linking the beginning of the following meaning and the verb.
The Present Study
In this study, two Event Related Potentials (ERP) experiments werecarried out to study the lexical implication and effect of theprevious terms on the structure of the sentence (Pozzan, andTrueswell, 2014). This was done concerning the comprehendersprediction of a verb in the final clause. Embedded object questionswith an OSV word order were utilized in executing these experimentssince it presented the lexical implication together with thefundamental roles of opinions before the verb (Pozzan, and Trueswell,2014). During the investigation, the order of the preverbal argumentswas reversed alongside changing one of the previous verbs argumentswith a suitable noun. Through the utilization of the N400 response,it was realized that there is the simplicity of accessing a similarword in the long-standing recollection of a speaker (Boland et al.2016)
The extent to which the initial prediction by the comprehender aboutan upcoming verb can be attributed to the lexical meaning of thesentence. The original prediction can also be directed towards thefundamental function of the argument. This experiment was carried outby contrasting the consequences of rehearsing an argument and thereplacement on N 400 amplitudes on a verb (Riordan, Dye, & Jones,2015). In the case that the knowledge on the lexical meaning hasadditional impacts on a comprehenders` prediction of the next verbthan their essential functions, a time interval should then beintroduced before presenting the opposing views (Nieuwland, Martin,and Carreiras, 2013). During such moments, the predictions by thecomprehender are more sensitive towards the actual meaning ascompared to the fundamental roles of the sentence.
While going on with our experiments, we separated the influence ofthe two kinds of contextual information on the prediction of the verbby manipulating the content of the sentence in two different ways(Riordan, Dye, & Jones, 2015). With the incorporation of thesechanges, the probability of the verb close was also seen to undergosome transitions thus resulting in two arguments which include debaterole reversal and argument substitution. The operation of a reversalof the argument caused the differentiation of the role played by theinformation delivered by the case through interchanging thearrangement of the targeted verb and all other things were heldunvarying (Nieuwland, Martin, and Carreiras, 2013). Some of thesentences presented to the participants include: The soldier shot thevillain with a shotgun. After putting the sentence trough argumentreversal, the participants will be presented with the sentence: thesoldier shot the villain using a rifle. The same sentence will bepresented to the participants after a debate role reversal as: Thevillain shot the soldier with a rifle.
On the other hand, the argument substitution manipulation causeddifferences in the lexical meaning of the arguments by substitutingone of the actual discussions with a noun while the other case andthe targeted verb are held constant. Collectively, the two cases arematched in different ways to provide a situation where comparisonscan be made on the fundamental functions and lexical sense to thecomprehenders` prediction of the next verb (Pearl &Braunwald,2015).
Hi:Contextual information plays a key role in determining thecomprehenders` prediction and anticipation of sentence meaning
Ho:Contextual information is irrelevant in comprehenders` prediction andanticipation of sentence meaning
Theresearch assumes that contextual information is key in comprehensionof a sentence. The study expects that the experiment will prove thehypothesis and show the relationship between context and sentencecomprehension as well as the prediction of the next verb in thesentence.
Hi:Different contextual information is processed differently and affectspredictions in different ways.
H0:All contextual information is processed in the same way.
The experiment will incorporate fifty participants (male participants=20, female participants = 30, the mean age of participants = 22 witha range of 18 to 30) from the general public including members of theuniversity. All participants will be native speakers of the Englishlanguage. The participants will all have a standard vision and nohistory of neurological disarray. The participants will also have abias to the right hand according to the Edinburgh HandednessInventory.
The identity and results of the participants will be withheld inagreement with the ethical standards of the research.
The experiment will make use of embedded object questions as opposedto the subject-verb objective(SVO). This means that the syntax andidentity of the roles will be exposed before the verb in thesentence. The sentences used in the experiment were developed inthree stages. The first two hundred word pairs, half of this pairshad role reversed arguments and the other half had normal arguments.The third sentence pairs had argument substitution this means theobject of the sentence was changed. Argument substitution means thatthe sentences will remain the same with only one variation in a nounor the point of the argument. The only word was changed for thissentences for example, instead of the use of the word thief, theword criminal was used in its place.
The participants will be comfortably seated in a hall with eachparticipant an average of one meter away from the next participant.The sentences are provided to the participants a word at a go on ablack font displayed on a white screen. The words will appear on thescreen for an average of 2.5 seconds then a blank space forapproximately 1.5 seconds. Once the sentence ended which wasindicated by a full stop, a blank space appeared for 8 seconds, andthe participant was asked if the sentence was possible and provided aspace to indicate yes or no. The sentences are to be presented to theparticipants in blocks of 50 sentences after which they are to begiven a thirty-minute break.
The test is designed to find out if a comprehenders` first semanticanticipation of the next verb is determined by the structural roleand the lexical meaning of the sentence. From previous studies, wehave already concluded that the kind of information presented to thecomprehender can determine the prediction of the next verb. Theexperiment goes further to determine whether the context of asentence affects predictions made by comprehenders and to what extentthey affect it if at all they do have an effect.
Boland, J. E., Kaan, E., Valdés Kroff, J., &Wulff, S. (2016).Psycholinguistics and variation in language processing.LinguisticsVanguard, 2(s1), 3-12.
Hörmann, H. (2013). Psycholinguistics: an introduction toresearch and theory. Springer Science & Business Media.
Nieuwland, S., Martin, &Carreiras, M. (2013). Event-related brainpotential evidence for human processing during sentencecomprehension. Brain and Language, 126(2), 151-158.
Pearl, L., &Braunwald, S. (2015). Language in mind: Anintroduction to psycholinguistics by Julie Sedivy (review). Language,91(4), e181-e183.
Pozzan, L., &Trueswell, J. (2014). Limitations of real-timeacquisition sentence grammar processing: Comparing verb-initial andverb-final languages. CAPEL 2014, 19.
Riordan, B., Dye, M., & Jones, M. N. (2015). Grammatical numberprocessing and forecast eye movements are less coordinated in Englishlanguage interpretation. Frontiers in Psychology, 6.
Small, S. L., Cottrell, G. W., &Tanenhaus, M. K. (Eds.). (2013).Lexical Ambiguity Resolution: Perspective from Psycholinguistics,Neuropsychology and Artificial Intelligence. Morgan Kaufmann.